From 9fe6e9e7b58944037714442384075c17cfde1c56 Mon Sep 17 00:00:00 2001 From: Jan Kara Date: Mon, 18 Mar 2024 17:32:09 +0100 Subject: nfsd: Fix error cleanup path in nfsd_rename() Commit a8b0026847b8 ("rename(): avoid a deadlock in the case of parents having no common ancestor") added an error bail out path. However this path does not drop the remount protection that has been acquired. Fix the cleanup path to properly drop the remount protection. Fixes: a8b0026847b8 ("rename(): avoid a deadlock in the case of parents having no common ancestor") Signed-off-by: Jan Kara Reviewed-by: Jeff Layton Acked-by: Al Viro Signed-off-by: Chuck Lever --- fs/nfsd/vfs.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c index 6a4c506038e0..daf5221ef922 100644 --- a/fs/nfsd/vfs.c +++ b/fs/nfsd/vfs.c @@ -1853,7 +1853,7 @@ retry: trap = lock_rename(tdentry, fdentry); if (IS_ERR(trap)) { err = (rqstp->rq_vers == 2) ? nfserr_acces : nfserr_xdev; - goto out; + goto out_want_write; } err = fh_fill_pre_attrs(ffhp); if (err != nfs_ok) @@ -1923,6 +1923,7 @@ retry: } out_unlock: unlock_rename(tdentry, fdentry); +out_want_write: fh_drop_write(ffhp); /* -- cgit From 15922f5dbf51dad334cde888ce6835d377678dc9 Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Tue, 19 Mar 2024 11:29:03 +1100 Subject: xfs: allow sunit mount option to repair bad primary sb stripe values If a filesystem has a busted stripe alignment configuration on disk (e.g. because broken RAID firmware told mkfs that swidth was smaller than sunit), then the filesystem will refuse to mount due to the stripe validation failing. This failure is triggering during distro upgrades from old kernels lacking this check to newer kernels with this check, and currently the only way to fix it is with offline xfs_db surgery. This runtime validity checking occurs when we read the superblock for the first time and causes the mount to fail immediately. This prevents the rewrite of stripe unit/width via mount options that occurs later in the mount process. Hence there is no way to recover this situation without resorting to offline xfs_db rewrite of the values. However, we parse the mount options long before we read the superblock, and we know if the mount has been asked to re-write the stripe alignment configuration when we are reading the superblock and verifying it for the first time. Hence we can conditionally ignore stripe verification failures if the mount options specified will correct the issue. We validate that the new stripe unit/width are valid before we overwrite the superblock values, so we can ignore the invalid config at verification and fail the mount later if the new values are not valid. This, at least, gives users the chance of correcting the issue after a kernel upgrade without having to resort to xfs-db hacks. Signed-off-by: Dave Chinner Reviewed-by: Christoph Hellwig Reviewed-by: "Darrick J. Wong" Signed-off-by: Chandan Babu R --- fs/xfs/libxfs/xfs_sb.c | 40 +++++++++++++++++++++++++++++++--------- fs/xfs/libxfs/xfs_sb.h | 5 +++-- 2 files changed, 34 insertions(+), 11 deletions(-) (limited to 'fs') diff --git a/fs/xfs/libxfs/xfs_sb.c b/fs/xfs/libxfs/xfs_sb.c index d991eec05436..73a4b895de67 100644 --- a/fs/xfs/libxfs/xfs_sb.c +++ b/fs/xfs/libxfs/xfs_sb.c @@ -530,7 +530,8 @@ xfs_validate_sb_common( } if (!xfs_validate_stripe_geometry(mp, XFS_FSB_TO_B(mp, sbp->sb_unit), - XFS_FSB_TO_B(mp, sbp->sb_width), 0, false)) + XFS_FSB_TO_B(mp, sbp->sb_width), 0, + xfs_buf_daddr(bp) == XFS_SB_DADDR, false)) return -EFSCORRUPTED; /* @@ -1323,8 +1324,10 @@ xfs_sb_get_secondary( } /* - * sunit, swidth, sectorsize(optional with 0) should be all in bytes, - * so users won't be confused by values in error messages. + * sunit, swidth, sectorsize(optional with 0) should be all in bytes, so users + * won't be confused by values in error messages. This function returns false + * if the stripe geometry is invalid and the caller is unable to repair the + * stripe configuration later in the mount process. */ bool xfs_validate_stripe_geometry( @@ -1332,20 +1335,21 @@ xfs_validate_stripe_geometry( __s64 sunit, __s64 swidth, int sectorsize, + bool may_repair, bool silent) { if (swidth > INT_MAX) { if (!silent) xfs_notice(mp, "stripe width (%lld) is too large", swidth); - return false; + goto check_override; } if (sunit > swidth) { if (!silent) xfs_notice(mp, "stripe unit (%lld) is larger than the stripe width (%lld)", sunit, swidth); - return false; + goto check_override; } if (sectorsize && (int)sunit % sectorsize) { @@ -1353,21 +1357,21 @@ xfs_validate_stripe_geometry( xfs_notice(mp, "stripe unit (%lld) must be a multiple of the sector size (%d)", sunit, sectorsize); - return false; + goto check_override; } if (sunit && !swidth) { if (!silent) xfs_notice(mp, "invalid stripe unit (%lld) and stripe width of 0", sunit); - return false; + goto check_override; } if (!sunit && swidth) { if (!silent) xfs_notice(mp, "invalid stripe width (%lld) and stripe unit of 0", swidth); - return false; + goto check_override; } if (sunit && (int)swidth % (int)sunit) { @@ -1375,9 +1379,27 @@ xfs_validate_stripe_geometry( xfs_notice(mp, "stripe width (%lld) must be a multiple of the stripe unit (%lld)", swidth, sunit); - return false; + goto check_override; } return true; + +check_override: + if (!may_repair) + return false; + /* + * During mount, mp->m_dalign will not be set unless the sunit mount + * option was set. If it was set, ignore the bad stripe alignment values + * and allow the validation and overwrite later in the mount process to + * attempt to overwrite the bad stripe alignment values with the values + * supplied by mount options. + */ + if (!mp->m_dalign) + return false; + if (!silent) + xfs_notice(mp, +"Will try to correct with specified mount options sunit (%d) and swidth (%d)", + BBTOB(mp->m_dalign), BBTOB(mp->m_swidth)); + return true; } /* diff --git a/fs/xfs/libxfs/xfs_sb.h b/fs/xfs/libxfs/xfs_sb.h index 2e8e8d63d4eb..37b1ed1bc209 100644 --- a/fs/xfs/libxfs/xfs_sb.h +++ b/fs/xfs/libxfs/xfs_sb.h @@ -35,8 +35,9 @@ extern int xfs_sb_get_secondary(struct xfs_mount *mp, struct xfs_trans *tp, xfs_agnumber_t agno, struct xfs_buf **bpp); -extern bool xfs_validate_stripe_geometry(struct xfs_mount *mp, - __s64 sunit, __s64 swidth, int sectorsize, bool silent); +bool xfs_validate_stripe_geometry(struct xfs_mount *mp, + __s64 sunit, __s64 swidth, int sectorsize, bool may_repair, + bool silent); uint8_t xfs_compute_rextslog(xfs_rtbxlen_t rtextents); -- cgit From f2e812c1522dab847912309b00abcc762dd696da Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Tue, 19 Mar 2024 09:36:28 +1100 Subject: xfs: don't use current->journal_info syzbot reported an ext4 panic during a page fault where found a journal handle when it didn't expect to find one. The structure it tripped over had a value of 'TRAN' in the first entry in the structure, and that indicates it tripped over a struct xfs_trans instead of a jbd2 handle. The reason for this is that the page fault was taken during a copy-out to a user buffer from an xfs bulkstat operation. XFS uses an "empty" transaction context for bulkstat to do automated metadata buffer cleanup, and so the transaction context is valid across the copyout of the bulkstat info into the user buffer. We are using empty transaction contexts like this in XFS to reduce the risk of failing to release objects we reference during the operation, especially during error handling. Hence we really need to ensure that we can take page faults from these contexts without leaving landmines for the code processing the page fault to trip over. However, this same behaviour could happen from any other filesystem that triggers a page fault or any other exception that is handled on-stack from within a task context that has current->journal_info set. Having a page fault from some other filesystem bounce into XFS where we have to run a transaction isn't a bug at all, but the usage of current->journal_info means that this could result corruption of the outer task's journal_info structure. The problem is purely that we now have two different contexts that now think they own current->journal_info. IOWs, no filesystem can allow page faults or on-stack exceptions while current->journal_info is set by the filesystem because the exception processing might use current->journal_info itself. If we end up with nested XFS transactions whilst holding an empty transaction, then it isn't an issue as the outer transaction does not hold a log reservation. If we ignore the current->journal_info usage, then the only problem that might occur is a deadlock if the exception tries to take the same locks the upper context holds. That, however, is not a problem that setting current->journal_info would solve, so it's largely an irrelevant concern here. IOWs, we really only use current->journal_info for a warning check in xfs_vm_writepages() to ensure we aren't doing writeback from a transaction context. Writeback might need to do allocation, so it can need to run transactions itself. Hence it's a debug check to warn us that we've done something silly, and largely it is not all that useful. So let's just remove all the use of current->journal_info in XFS and get rid of all the potential issues from nested contexts where current->journal_info might get misused by another filesystem context. Reported-by: syzbot+cdee56dbcdf0096ef605@syzkaller.appspotmail.com Signed-off-by: Dave Chinner Reviewed-by: "Darrick J. Wong" Reviewed-by: Mark Tinguely Reviewed-by: Christoph Hellwig Signed-off-by: Chandan Babu R --- fs/xfs/scrub/common.c | 4 +--- fs/xfs/xfs_aops.c | 7 ------- fs/xfs/xfs_icache.c | 8 +++++--- fs/xfs/xfs_trans.h | 9 +-------- 4 files changed, 7 insertions(+), 21 deletions(-) (limited to 'fs') diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c index abff79a77c72..47a20cf5205f 100644 --- a/fs/xfs/scrub/common.c +++ b/fs/xfs/scrub/common.c @@ -1044,9 +1044,7 @@ xchk_irele( struct xfs_scrub *sc, struct xfs_inode *ip) { - if (current->journal_info != NULL) { - ASSERT(current->journal_info == sc->tp); - + if (sc->tp) { /* * If we are in a transaction, we /cannot/ drop the inode * ourselves, because the VFS will trigger writeback, which diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c index 1698507d1ac7..3f428620ebf2 100644 --- a/fs/xfs/xfs_aops.c +++ b/fs/xfs/xfs_aops.c @@ -503,13 +503,6 @@ xfs_vm_writepages( { struct xfs_writepage_ctx wpc = { }; - /* - * Writing back data in a transaction context can result in recursive - * transactions. This is bad, so issue a warning and get out of here. - */ - if (WARN_ON_ONCE(current->journal_info)) - return 0; - xfs_iflags_clear(XFS_I(mapping->host), XFS_ITRUNCATED); return iomap_writepages(mapping, wbc, &wpc.ctx, &xfs_writeback_ops); } diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c index e64265bc0b33..74f1812b03cb 100644 --- a/fs/xfs/xfs_icache.c +++ b/fs/xfs/xfs_icache.c @@ -2039,8 +2039,10 @@ xfs_inodegc_want_queue_work( * - Memory shrinkers queued the inactivation worker and it hasn't finished. * - The queue depth exceeds the maximum allowable percpu backlog. * - * Note: If the current thread is running a transaction, we don't ever want to - * wait for other transactions because that could introduce a deadlock. + * Note: If we are in a NOFS context here (e.g. current thread is running a + * transaction) the we don't want to block here as inodegc progress may require + * filesystem resources we hold to make progress and that could result in a + * deadlock. Hence we skip out of here if we are in a scoped NOFS context. */ static inline bool xfs_inodegc_want_flush_work( @@ -2048,7 +2050,7 @@ xfs_inodegc_want_flush_work( unsigned int items, unsigned int shrinker_hits) { - if (current->journal_info) + if (current->flags & PF_MEMALLOC_NOFS) return false; if (shrinker_hits > 0) diff --git a/fs/xfs/xfs_trans.h b/fs/xfs/xfs_trans.h index 3f7e3a09a49f..1636663707dc 100644 --- a/fs/xfs/xfs_trans.h +++ b/fs/xfs/xfs_trans.h @@ -268,19 +268,14 @@ static inline void xfs_trans_set_context( struct xfs_trans *tp) { - ASSERT(current->journal_info == NULL); tp->t_pflags = memalloc_nofs_save(); - current->journal_info = tp; } static inline void xfs_trans_clear_context( struct xfs_trans *tp) { - if (current->journal_info == tp) { - memalloc_nofs_restore(tp->t_pflags); - current->journal_info = NULL; - } + memalloc_nofs_restore(tp->t_pflags); } static inline void @@ -288,10 +283,8 @@ xfs_trans_switch_context( struct xfs_trans *old_tp, struct xfs_trans *new_tp) { - ASSERT(current->journal_info == old_tp); new_tp->t_pflags = old_tp->t_pflags; old_tp->t_pflags = 0; - current->journal_info = new_tp; } #endif /* __XFS_TRANS_H__ */ -- cgit From e9e9fbeb83f65d3d487e0a0838c0867292c99fb2 Mon Sep 17 00:00:00 2001 From: Steve French Date: Sun, 24 Mar 2024 00:01:02 -0500 Subject: smb3: add trace event for mknod Add trace points to help debug mknod and mkfifo: smb3_mknod_done smb3_mknod_enter smb3_mknod_err Example output: TASK-PID CPU# ||||| TIMESTAMP FUNCTION | | | ||||| | | mkfifo-6163 [003] ..... 960.425558: smb3_mknod_enter: xid=12 sid=0xb55130f6 tid=0x46e6241c path=\fifo1 mkfifo-6163 [003] ..... 960.432719: smb3_mknod_done: xid=12 sid=0xb55130f6 tid=0x46e6241c Reviewed-by: Bharath SM Reviewed-by: Meetakshi Setiya Signed-off-by: Steve French --- fs/smb/client/dir.c | 7 +++++++ fs/smb/client/trace.h | 4 +++- 2 files changed, 10 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/smb/client/dir.c b/fs/smb/client/dir.c index 89333d9bce36..d11dc3aa458b 100644 --- a/fs/smb/client/dir.c +++ b/fs/smb/client/dir.c @@ -612,11 +612,18 @@ int cifs_mknod(struct mnt_idmap *idmap, struct inode *inode, goto mknod_out; } + trace_smb3_mknod_enter(xid, tcon->ses->Suid, tcon->tid, full_path); + rc = tcon->ses->server->ops->make_node(xid, inode, direntry, tcon, full_path, mode, device_number); mknod_out: + if (rc) + trace_smb3_mknod_err(xid, tcon->ses->Suid, tcon->tid, rc); + else + trace_smb3_mknod_done(xid, tcon->ses->Suid, tcon->tid); + free_dentry_path(page); free_xid(xid); cifs_put_tlink(tlink); diff --git a/fs/smb/client/trace.h b/fs/smb/client/trace.h index f9c1fd32d0b8..5e83cb9da902 100644 --- a/fs/smb/client/trace.h +++ b/fs/smb/client/trace.h @@ -375,6 +375,7 @@ DEFINE_SMB3_INF_COMPOUND_ENTER_EVENT(get_reparse_compound_enter); DEFINE_SMB3_INF_COMPOUND_ENTER_EVENT(delete_enter); DEFINE_SMB3_INF_COMPOUND_ENTER_EVENT(mkdir_enter); DEFINE_SMB3_INF_COMPOUND_ENTER_EVENT(tdis_enter); +DEFINE_SMB3_INF_COMPOUND_ENTER_EVENT(mknod_enter); DECLARE_EVENT_CLASS(smb3_inf_compound_done_class, TP_PROTO(unsigned int xid, @@ -415,7 +416,7 @@ DEFINE_SMB3_INF_COMPOUND_DONE_EVENT(query_wsl_ea_compound_done); DEFINE_SMB3_INF_COMPOUND_DONE_EVENT(delete_done); DEFINE_SMB3_INF_COMPOUND_DONE_EVENT(mkdir_done); DEFINE_SMB3_INF_COMPOUND_DONE_EVENT(tdis_done); - +DEFINE_SMB3_INF_COMPOUND_DONE_EVENT(mknod_done); DECLARE_EVENT_CLASS(smb3_inf_compound_err_class, TP_PROTO(unsigned int xid, @@ -461,6 +462,7 @@ DEFINE_SMB3_INF_COMPOUND_ERR_EVENT(query_wsl_ea_compound_err); DEFINE_SMB3_INF_COMPOUND_ERR_EVENT(mkdir_err); DEFINE_SMB3_INF_COMPOUND_ERR_EVENT(delete_err); DEFINE_SMB3_INF_COMPOUND_ERR_EVENT(tdis_err); +DEFINE_SMB3_INF_COMPOUND_ERR_EVENT(mknod_err); /* * For logging SMB3 Status code and Command for responses which return errors -- cgit From 8876a37277cb832e1861c35f8c661825179f73f5 Mon Sep 17 00:00:00 2001 From: David Howells Date: Wed, 27 Mar 2024 14:13:24 +0000 Subject: cifs: Fix duplicate fscache cookie warnings fscache emits a lot of duplicate cookie warnings with cifs because the index key for the fscache cookies does not include everything that the cifs_find_inode() function does. The latter is used with iget5_locked() to distinguish between inodes in the local inode cache. Fix this by adding the creation time and file type to the fscache cookie key. Additionally, add a couple of comments to note that if one is changed the other must be also. Signed-off-by: David Howells Fixes: 70431bfd825d ("cifs: Support fscache indexing rewrite") cc: Shyam Prasad N cc: Rohith Surabattula cc: Jeff Layton cc: linux-cifs@vger.kernel.org cc: netfs@lists.linux.dev cc: linux-fsdevel@vger.kernel.org Signed-off-by: Steve French --- fs/smb/client/fscache.c | 16 +++++++++++++++- fs/smb/client/inode.c | 2 ++ 2 files changed, 17 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/smb/client/fscache.c b/fs/smb/client/fscache.c index c4a3cb736881..340efce8f052 100644 --- a/fs/smb/client/fscache.c +++ b/fs/smb/client/fscache.c @@ -12,6 +12,16 @@ #include "cifs_fs_sb.h" #include "cifsproto.h" +/* + * Key for fscache inode. [!] Contents must match comparisons in cifs_find_inode(). + */ +struct cifs_fscache_inode_key { + + __le64 uniqueid; /* server inode number */ + __le64 createtime; /* creation time on server */ + u8 type; /* S_IFMT file type */ +} __packed; + static void cifs_fscache_fill_volume_coherency( struct cifs_tcon *tcon, struct cifs_fscache_volume_coherency_data *cd) @@ -97,15 +107,19 @@ void cifs_fscache_release_super_cookie(struct cifs_tcon *tcon) void cifs_fscache_get_inode_cookie(struct inode *inode) { struct cifs_fscache_inode_coherency_data cd; + struct cifs_fscache_inode_key key; struct cifsInodeInfo *cifsi = CIFS_I(inode); struct cifs_sb_info *cifs_sb = CIFS_SB(inode->i_sb); struct cifs_tcon *tcon = cifs_sb_master_tcon(cifs_sb); + key.uniqueid = cpu_to_le64(cifsi->uniqueid); + key.createtime = cpu_to_le64(cifsi->createtime); + key.type = (inode->i_mode & S_IFMT) >> 12; cifs_fscache_fill_coherency(&cifsi->netfs.inode, &cd); cifsi->netfs.cache = fscache_acquire_cookie(tcon->fscache, 0, - &cifsi->uniqueid, sizeof(cifsi->uniqueid), + &key, sizeof(key), &cd, sizeof(cd), i_size_read(&cifsi->netfs.inode)); if (cifsi->netfs.cache) diff --git a/fs/smb/client/inode.c b/fs/smb/client/inode.c index d28ab0af6049..91b07ef9e25c 100644 --- a/fs/smb/client/inode.c +++ b/fs/smb/client/inode.c @@ -1351,6 +1351,8 @@ cifs_find_inode(struct inode *inode, void *opaque) { struct cifs_fattr *fattr = opaque; + /* [!] The compared values must be the same in struct cifs_fscache_inode_key. */ + /* don't match inode with different uniqueid */ if (CIFS_I(inode)->uniqueid != fattr->cf_uniqueid) return 0; -- cgit From 99dc2ef0397d082b63404c01cf841cf80f1418dc Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Tue, 26 Mar 2024 17:00:20 -0400 Subject: NFSD: CREATE_SESSION must never cache NFS4ERR_DELAY replies There are one or two cases where CREATE_SESSION returns NFS4ERR_DELAY in order to force the client to wait a bit and try CREATE_SESSION again. However, after commit e4469c6cc69b ("NFSD: Fix the NFSv4.1 CREATE_SESSION operation"), NFSD caches that response in the CREATE_SESSION slot. Thus, when the client resends the CREATE_SESSION, the server always returns the cached NFS4ERR_DELAY response rather than actually executing the request and properly recording its outcome. This blocks the client from making further progress. RFC 8881 Section 15.1.1.3 says: > If NFS4ERR_DELAY is returned on an operation other than SEQUENCE > that validly appears as the first operation of a request ... [t]he > request can be retried in full without modification. In this case > as well, the replier MUST avoid returning a response containing > NFS4ERR_DELAY as the response to an initial operation of a request > solely on the basis of its presence in the reply cache. Neither the original NFSD code nor the discussion in section 18.36.4 refer explicitly to this important requirement, so I missed it. Note also that not only must the server not cache NFS4ERR_DELAY, but it has to not advance the CREATE_SESSION slot sequence number so that it can properly recognize and accept the client's retry. Reported-by: Dai Ngo Fixes: e4469c6cc69b ("NFSD: Fix the NFSv4.1 CREATE_SESSION operation") Tested-by: Dai Ngo Signed-off-by: Chuck Lever --- fs/nfsd/nfs4state.c | 36 +++++++++++++++++++++++++----------- 1 file changed, 25 insertions(+), 11 deletions(-) (limited to 'fs') diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c index ee9aa4843443..5fcd93f7cb8c 100644 --- a/fs/nfsd/nfs4state.c +++ b/fs/nfsd/nfs4state.c @@ -3831,15 +3831,20 @@ nfsd4_create_session(struct svc_rqst *rqstp, else cs_slot = &unconf->cl_cs_slot; status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0); - if (status) { - if (status == nfserr_replay_cache) { - status = nfsd4_replay_create_session(cr_ses, cs_slot); - goto out_free_conn; - } + switch (status) { + case nfs_ok: + cs_slot->sl_seqid++; + cr_ses->seqid = cs_slot->sl_seqid; + break; + case nfserr_replay_cache: + status = nfsd4_replay_create_session(cr_ses, cs_slot); + fallthrough; + case nfserr_jukebox: + /* The server MUST NOT cache NFS4ERR_DELAY */ + goto out_free_conn; + default: goto out_cache_error; } - cs_slot->sl_seqid++; - cr_ses->seqid = cs_slot->sl_seqid; /* RFC 8881 Section 18.36.4 Phase 3: Client ID confirmation. */ if (conf) { @@ -3859,10 +3864,8 @@ nfsd4_create_session(struct svc_rqst *rqstp, old = find_confirmed_client_by_name(&unconf->cl_name, nn); if (old) { status = mark_client_expired_locked(old); - if (status) { - old = NULL; - goto out_cache_error; - } + if (status) + goto out_expired_error; trace_nfsd_clid_replaced(&old->cl_clientid); } move_to_confirmed(unconf); @@ -3894,6 +3897,17 @@ nfsd4_create_session(struct svc_rqst *rqstp, expire_client(old); return status; +out_expired_error: + old = NULL; + /* + * Revert the slot seq_nr change so the server will process + * the client's resend instead of returning a cached response. + */ + if (status == nfserr_jukebox) { + cs_slot->sl_seqid--; + cr_ses->seqid = cs_slot->sl_seqid; + goto out_free_conn; + } out_cache_error: nfsd4_cache_create_session(cr_ses, cs_slot, status); out_free_conn: -- cgit From c40845e3195d074b34f8f8e400e28c9403a06588 Mon Sep 17 00:00:00 2001 From: Arnd Bergmann Date: Tue, 26 Mar 2024 15:47:16 +0100 Subject: kbuild: make -Woverride-init warnings more consistent The -Woverride-init warn about code that may be intentional or not, but the inintentional ones tend to be real bugs, so there is a bit of disagreement on whether this warning option should be enabled by default and we have multiple settings in scripts/Makefile.extrawarn as well as individual subsystems. Older versions of clang only supported -Wno-initializer-overrides with the same meaning as gcc's -Woverride-init, though all supported versions now work with both. Because of this difference, an earlier cleanup of mine accidentally turned the clang warning off for W=1 builds and only left it on for W=2, while it's still enabled for gcc with W=1. There is also one driver that only turns the warning off for newer versions of gcc but not other compilers, and some but not all the Makefiles still use a cc-disable-warning conditional that is no longer needed with supported compilers here. Address all of the above by removing the special cases for clang and always turning the warning off unconditionally where it got in the way, using the syntax that is supported by both compilers. Fixes: 2cd3271b7a31 ("kbuild: avoid duplicate warning options") Signed-off-by: Arnd Bergmann Acked-by: Hamza Mahfooz Acked-by: Jani Nikula Acked-by: Andrew Jeffery Signed-off-by: Jani Nikula Reviewed-by: Linus Walleij Signed-off-by: Masahiro Yamada --- drivers/gpu/drm/amd/display/dc/dce110/Makefile | 2 +- drivers/gpu/drm/amd/display/dc/dce112/Makefile | 2 +- drivers/gpu/drm/amd/display/dc/dce120/Makefile | 2 +- drivers/gpu/drm/amd/display/dc/dce60/Makefile | 2 +- drivers/gpu/drm/amd/display/dc/dce80/Makefile | 2 +- drivers/gpu/drm/i915/Makefile | 6 +++--- drivers/gpu/drm/xe/Makefile | 4 ++-- drivers/net/ethernet/renesas/sh_eth.c | 2 +- drivers/pinctrl/aspeed/Makefile | 2 +- fs/proc/Makefile | 2 +- kernel/bpf/Makefile | 2 +- mm/Makefile | 3 +-- scripts/Makefile.extrawarn | 10 +++------- 13 files changed, 18 insertions(+), 23 deletions(-) (limited to 'fs') diff --git a/drivers/gpu/drm/amd/display/dc/dce110/Makefile b/drivers/gpu/drm/amd/display/dc/dce110/Makefile index f0777d61c2cb..c307f040e48f 100644 --- a/drivers/gpu/drm/amd/display/dc/dce110/Makefile +++ b/drivers/gpu/drm/amd/display/dc/dce110/Makefile @@ -23,7 +23,7 @@ # Makefile for the 'controller' sub-component of DAL. # It provides the control and status of HW CRTC block. -CFLAGS_$(AMDDALPATH)/dc/dce110/dce110_resource.o = $(call cc-disable-warning, override-init) +CFLAGS_$(AMDDALPATH)/dc/dce110/dce110_resource.o = -Wno-override-init DCE110 = dce110_timing_generator.o \ dce110_compressor.o dce110_opp_regamma_v.o \ diff --git a/drivers/gpu/drm/amd/display/dc/dce112/Makefile b/drivers/gpu/drm/amd/display/dc/dce112/Makefile index 7e92effec894..683866797709 100644 --- a/drivers/gpu/drm/amd/display/dc/dce112/Makefile +++ b/drivers/gpu/drm/amd/display/dc/dce112/Makefile @@ -23,7 +23,7 @@ # Makefile for the 'controller' sub-component of DAL. # It provides the control and status of HW CRTC block. -CFLAGS_$(AMDDALPATH)/dc/dce112/dce112_resource.o = $(call cc-disable-warning, override-init) +CFLAGS_$(AMDDALPATH)/dc/dce112/dce112_resource.o = -Wno-override-init DCE112 = dce112_compressor.o diff --git a/drivers/gpu/drm/amd/display/dc/dce120/Makefile b/drivers/gpu/drm/amd/display/dc/dce120/Makefile index 1e3ef68a452a..8f508e662748 100644 --- a/drivers/gpu/drm/amd/display/dc/dce120/Makefile +++ b/drivers/gpu/drm/amd/display/dc/dce120/Makefile @@ -24,7 +24,7 @@ # It provides the control and status of HW CRTC block. -CFLAGS_$(AMDDALPATH)/dc/dce120/dce120_resource.o = $(call cc-disable-warning, override-init) +CFLAGS_$(AMDDALPATH)/dc/dce120/dce120_resource.o = -Wno-override-init DCE120 = dce120_timing_generator.o diff --git a/drivers/gpu/drm/amd/display/dc/dce60/Makefile b/drivers/gpu/drm/amd/display/dc/dce60/Makefile index fee331accc0e..eede83ad91fa 100644 --- a/drivers/gpu/drm/amd/display/dc/dce60/Makefile +++ b/drivers/gpu/drm/amd/display/dc/dce60/Makefile @@ -23,7 +23,7 @@ # Makefile for the 'controller' sub-component of DAL. # It provides the control and status of HW CRTC block. -CFLAGS_$(AMDDALPATH)/dc/dce60/dce60_resource.o = $(call cc-disable-warning, override-init) +CFLAGS_$(AMDDALPATH)/dc/dce60/dce60_resource.o = -Wno-override-init DCE60 = dce60_timing_generator.o dce60_hw_sequencer.o \ dce60_resource.o diff --git a/drivers/gpu/drm/amd/display/dc/dce80/Makefile b/drivers/gpu/drm/amd/display/dc/dce80/Makefile index 7eefffbdc925..fba189d26652 100644 --- a/drivers/gpu/drm/amd/display/dc/dce80/Makefile +++ b/drivers/gpu/drm/amd/display/dc/dce80/Makefile @@ -23,7 +23,7 @@ # Makefile for the 'controller' sub-component of DAL. # It provides the control and status of HW CRTC block. -CFLAGS_$(AMDDALPATH)/dc/dce80/dce80_resource.o = $(call cc-disable-warning, override-init) +CFLAGS_$(AMDDALPATH)/dc/dce80/dce80_resource.o = -Wno-override-init DCE80 = dce80_timing_generator.o diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile index 3ef6ed41e62b..4c2f85632391 100644 --- a/drivers/gpu/drm/i915/Makefile +++ b/drivers/gpu/drm/i915/Makefile @@ -33,9 +33,9 @@ endif subdir-ccflags-$(CONFIG_DRM_I915_WERROR) += -Werror # Fine grained warnings disable -CFLAGS_i915_pci.o = $(call cc-disable-warning, override-init) -CFLAGS_display/intel_display_device.o = $(call cc-disable-warning, override-init) -CFLAGS_display/intel_fbdev.o = $(call cc-disable-warning, override-init) +CFLAGS_i915_pci.o = -Wno-override-init +CFLAGS_display/intel_display_device.o = -Wno-override-init +CFLAGS_display/intel_fbdev.o = -Wno-override-init # Support compiling the display code separately for both i915 and xe # drivers. Define I915 when building i915. diff --git a/drivers/gpu/drm/xe/Makefile b/drivers/gpu/drm/xe/Makefile index 5a428ca00f10..c29a850859ad 100644 --- a/drivers/gpu/drm/xe/Makefile +++ b/drivers/gpu/drm/xe/Makefile @@ -172,8 +172,8 @@ subdir-ccflags-$(CONFIG_DRM_XE_DISPLAY) += \ -Ddrm_i915_gem_object=xe_bo \ -Ddrm_i915_private=xe_device -CFLAGS_i915-display/intel_fbdev.o = $(call cc-disable-warning, override-init) -CFLAGS_i915-display/intel_display_device.o = $(call cc-disable-warning, override-init) +CFLAGS_i915-display/intel_fbdev.o = -Wno-override-init +CFLAGS_i915-display/intel_display_device.o = -Wno-override-init # Rule to build SOC code shared with i915 $(obj)/i915-soc/%.o: $(srctree)/drivers/gpu/drm/i915/soc/%.c FORCE diff --git a/drivers/net/ethernet/renesas/sh_eth.c b/drivers/net/ethernet/renesas/sh_eth.c index 475e1e8c1d35..0786eb0da391 100644 --- a/drivers/net/ethernet/renesas/sh_eth.c +++ b/drivers/net/ethernet/renesas/sh_eth.c @@ -50,7 +50,7 @@ * the macros available to do this only define GCC 8. */ __diag_push(); -__diag_ignore(GCC, 8, "-Woverride-init", +__diag_ignore_all("-Woverride-init", "logic to initialize all and then override some is OK"); static const u16 sh_eth_offset_gigabit[SH_ETH_MAX_REGISTER_OFFSET] = { SH_ETH_OFFSET_DEFAULTS, diff --git a/drivers/pinctrl/aspeed/Makefile b/drivers/pinctrl/aspeed/Makefile index 489ea1778353..db2a7600ae2b 100644 --- a/drivers/pinctrl/aspeed/Makefile +++ b/drivers/pinctrl/aspeed/Makefile @@ -1,7 +1,7 @@ # SPDX-License-Identifier: GPL-2.0-only # Aspeed pinctrl support -ccflags-y += $(call cc-option,-Woverride-init) +ccflags-y += -Woverride-init obj-$(CONFIG_PINCTRL_ASPEED) += pinctrl-aspeed.o pinmux-aspeed.o obj-$(CONFIG_PINCTRL_ASPEED_G4) += pinctrl-aspeed-g4.o obj-$(CONFIG_PINCTRL_ASPEED_G5) += pinctrl-aspeed-g5.o diff --git a/fs/proc/Makefile b/fs/proc/Makefile index bd08616ed8ba..7b4db9c56e6a 100644 --- a/fs/proc/Makefile +++ b/fs/proc/Makefile @@ -5,7 +5,7 @@ obj-y += proc.o -CFLAGS_task_mmu.o += $(call cc-option,-Wno-override-init,) +CFLAGS_task_mmu.o += -Wno-override-init proc-y := nommu.o task_nommu.o proc-$(CONFIG_MMU) := task_mmu.o diff --git a/kernel/bpf/Makefile b/kernel/bpf/Makefile index 368c5d86b5b7..e497011261b8 100644 --- a/kernel/bpf/Makefile +++ b/kernel/bpf/Makefile @@ -4,7 +4,7 @@ ifneq ($(CONFIG_BPF_JIT_ALWAYS_ON),y) # ___bpf_prog_run() needs GCSE disabled on x86; see 3193c0836f203 for details cflags-nogcse-$(CONFIG_X86)$(CONFIG_CC_IS_GCC) := -fno-gcse endif -CFLAGS_core.o += $(call cc-disable-warning, override-init) $(cflags-nogcse-yy) +CFLAGS_core.o += -Wno-override-init $(cflags-nogcse-yy) obj-$(CONFIG_BPF_SYSCALL) += syscall.o verifier.o inode.o helpers.o tnum.o log.o token.o obj-$(CONFIG_BPF_SYSCALL) += bpf_iter.o map_iter.o task_iter.o prog_iter.o link_iter.o diff --git a/mm/Makefile b/mm/Makefile index e4b5b75aaec9..4abb40b911ec 100644 --- a/mm/Makefile +++ b/mm/Makefile @@ -29,8 +29,7 @@ KCOV_INSTRUMENT_mmzone.o := n KCOV_INSTRUMENT_vmstat.o := n KCOV_INSTRUMENT_failslab.o := n -CFLAGS_init-mm.o += $(call cc-disable-warning, override-init) -CFLAGS_init-mm.o += $(call cc-disable-warning, initializer-overrides) +CFLAGS_init-mm.o += -Wno-override-init mmu-y := nommu.o mmu-$(CONFIG_MMU) := highmem.o memory.o mincore.o \ diff --git a/scripts/Makefile.extrawarn b/scripts/Makefile.extrawarn index 3ce5d503a6da..c5af566e911a 100644 --- a/scripts/Makefile.extrawarn +++ b/scripts/Makefile.extrawarn @@ -114,6 +114,8 @@ KBUILD_CFLAGS += $(call cc-disable-warning, format-overflow) KBUILD_CFLAGS += $(call cc-disable-warning, format-truncation) KBUILD_CFLAGS += $(call cc-disable-warning, stringop-truncation) +KBUILD_CFLAGS += -Wno-override-init # alias for -Wno-initializer-overrides in clang + ifdef CONFIG_CC_IS_CLANG # Clang before clang-16 would warn on default argument promotions. ifneq ($(call clang-min-version, 160000),y) @@ -151,10 +153,6 @@ KBUILD_CFLAGS += -Wtype-limits KBUILD_CFLAGS += $(call cc-option, -Wmaybe-uninitialized) KBUILD_CFLAGS += $(call cc-option, -Wunused-macros) -ifdef CONFIG_CC_IS_CLANG -KBUILD_CFLAGS += -Winitializer-overrides -endif - KBUILD_CPPFLAGS += -DKBUILD_EXTRA_WARN2 else @@ -164,9 +162,7 @@ KBUILD_CFLAGS += -Wno-missing-field-initializers KBUILD_CFLAGS += -Wno-type-limits KBUILD_CFLAGS += -Wno-shift-negative-value -ifdef CONFIG_CC_IS_CLANG -KBUILD_CFLAGS += -Wno-initializer-overrides -else +ifdef CONFIG_CC_IS_GCC KBUILD_CFLAGS += -Wno-maybe-uninitialized endif -- cgit From a5e3dce493d4b12b74000b6a99b6712afa5d1a4d Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Fri, 22 Mar 2024 00:15:41 -0400 Subject: bcachefs: Fix assert in bch2_backpointer_invalid() Backpointers that point to invalid devices are caught by fsck, not .key_invalid; so .key_invalid needs to check for them instead of hitting asserts. Signed-off-by: Kent Overstreet --- fs/bcachefs/backpointers.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'fs') diff --git a/fs/bcachefs/backpointers.c b/fs/bcachefs/backpointers.c index 8cb35ea572cb..21b3ae7df824 100644 --- a/fs/bcachefs/backpointers.c +++ b/fs/bcachefs/backpointers.c @@ -44,6 +44,11 @@ int bch2_backpointer_invalid(struct bch_fs *c, struct bkey_s_c k, struct printbuf *err) { struct bkey_s_c_backpointer bp = bkey_s_c_to_backpointer(k); + + /* these will be caught by fsck */ + if (!bch2_dev_exists2(c, bp.k->p.inode)) + return 0; + struct bpos bucket = bp_pos_to_bucket(c, bp.k->p); int ret = 0; -- cgit From 8aad8e1f659fcea1b24072e816e434e4cd12382d Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Fri, 22 Mar 2024 04:01:27 -0400 Subject: bcachefs: Fix journal pins in btree write buffer btree write buffer flush has two phases - in natural key order, which is more efficient but may fail - then in journal order The journal order flush was assuming that keys were still correctly ordered by journal sequence number - but due to coalescing by the previous phase, we need an additional sort. Signed-off-by: Kent Overstreet --- fs/bcachefs/btree_write_buffer.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+) (limited to 'fs') diff --git a/fs/bcachefs/btree_write_buffer.c b/fs/bcachefs/btree_write_buffer.c index 5cbad8445782..baf63e2fddb6 100644 --- a/fs/bcachefs/btree_write_buffer.c +++ b/fs/bcachefs/btree_write_buffer.c @@ -11,6 +11,7 @@ #include "journal_reclaim.h" #include +#include static int bch2_btree_write_buffer_journal_flush(struct journal *, struct journal_entry_pin *, u64); @@ -46,6 +47,14 @@ static inline bool wb_key_ref_cmp(const struct wb_key_ref *l, const struct wb_ke #endif } +static int wb_key_seq_cmp(const void *_l, const void *_r) +{ + const struct btree_write_buffered_key *l = _l; + const struct btree_write_buffered_key *r = _r; + + return cmp_int(l->journal_seq, r->journal_seq); +} + /* Compare excluding idx, the low 24 bits: */ static inline bool wb_key_eq(const void *_l, const void *_r) { @@ -357,6 +366,11 @@ static int bch2_btree_write_buffer_flush_locked(struct btree_trans *trans) */ trace_and_count(c, write_buffer_flush_slowpath, trans, slowpath, wb->flushing.keys.nr); + sort(wb->flushing.keys.data, + wb->flushing.keys.nr, + sizeof(wb->flushing.keys.data[0]), + wb_key_seq_cmp, NULL); + darray_for_each(wb->flushing.keys, i) { if (!i->journal_seq) continue; -- cgit From 688d750d10aa9c4fb71c5154521c775f94c887e0 Mon Sep 17 00:00:00 2001 From: Thomas Bertschinger Date: Wed, 20 Mar 2024 21:42:42 -0600 Subject: bcachefs: fix misplaced newline in __bch2_inode_unpacked_to_text() before: u64s 18 type inode_v3 0:1879048192:U32_MAX len 0 ver 0: mode=40700 flags= (15300000) journal_seq=4 bi_size=0 bi_sectors=0 bi_version=0bi_atime=227064388944 ... after: u64s 18 type inode_v3 0:1879048192:U32_MAX len 0 ver 0: mode=40700 flags= (15300000) journal_seq=4 bi_size=0 bi_sectors=0 bi_version=0 bi_atime=227064388944 ... Signed-off-by: Thomas Bertschinger Signed-off-by: Kent Overstreet --- fs/bcachefs/inode.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/bcachefs/inode.c b/fs/bcachefs/inode.c index 2b5e06770ab3..ca4a066e9a54 100644 --- a/fs/bcachefs/inode.c +++ b/fs/bcachefs/inode.c @@ -552,8 +552,8 @@ static void __bch2_inode_unpacked_to_text(struct printbuf *out, prt_printf(out, "bi_sectors=%llu", inode->bi_sectors); prt_newline(out); - prt_newline(out); prt_printf(out, "bi_version=%llu", inode->bi_version); + prt_newline(out); #define x(_name, _bits) \ prt_printf(out, #_name "=%llu", (u64) inode->_name); \ -- cgit From 4bd02d3fb33d8a46e73085b8d47d21c0ccb3de9d Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Thu, 28 Mar 2024 01:20:36 -0400 Subject: bcachefs: fix mount error path Signed-off-by: Kent Overstreet --- fs/bcachefs/fs.c | 1 + 1 file changed, 1 insertion(+) (limited to 'fs') diff --git a/fs/bcachefs/fs.c b/fs/bcachefs/fs.c index 0ccee05f6887..b5ea9fa1259d 100644 --- a/fs/bcachefs/fs.c +++ b/fs/bcachefs/fs.c @@ -1997,6 +1997,7 @@ out: return dget(sb->s_root); err_put_super: + __bch2_fs_stop(c); deactivate_locked_super(sb); return ERR_PTR(bch2_err_class(ret)); } -- cgit From aa6e130e3c2965a5c26a4033ff63b5dc9549bd76 Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Sun, 24 Mar 2024 19:52:03 -0400 Subject: bcachefs: Add an assertion for trying to evict btree root Signed-off-by: Kent Overstreet --- fs/bcachefs/btree_cache.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'fs') diff --git a/fs/bcachefs/btree_cache.c b/fs/bcachefs/btree_cache.c index 562561a9a510..8ff21b2e1463 100644 --- a/fs/bcachefs/btree_cache.c +++ b/fs/bcachefs/btree_cache.c @@ -1134,6 +1134,8 @@ void bch2_btree_node_evict(struct btree_trans *trans, const struct bkey_i *k) b = btree_cache_find(bc, k); if (!b) return; + + BUG_ON(b == btree_node_root(trans->c, b)); wait_on_io: /* not allowed to wait on io with btree locks held: */ -- cgit From 63332394c7e1f4f26e8e5b1387212016aaa7eae2 Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Thu, 21 Mar 2024 20:16:23 -0400 Subject: bcachefs: Move snapshot table size to struct snapshot_table We need to add bounds checking for snapshot table accesses - it turns out there are cases where we do need to use the snapshots table before fsck checks have completed (and indeed, fsck may not have been run). Signed-off-by: Kent Overstreet --- fs/bcachefs/bcachefs.h | 1 - fs/bcachefs/snapshot.c | 25 ++++++++++++++----------- fs/bcachefs/snapshot.h | 6 +++++- fs/bcachefs/subvolume_types.h | 2 ++ 4 files changed, 21 insertions(+), 13 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/bcachefs.h b/fs/bcachefs/bcachefs.h index 799aa32b6b4d..a9ade0b1f78c 100644 --- a/fs/bcachefs/bcachefs.h +++ b/fs/bcachefs/bcachefs.h @@ -810,7 +810,6 @@ struct bch_fs { /* snapshot.c: */ struct snapshot_table __rcu *snapshots; - size_t snapshot_table_size; struct mutex snapshot_table_lock; struct rw_semaphore snapshot_create_lock; diff --git a/fs/bcachefs/snapshot.c b/fs/bcachefs/snapshot.c index 39debe814bf3..9cd71e613dc9 100644 --- a/fs/bcachefs/snapshot.c +++ b/fs/bcachefs/snapshot.c @@ -151,36 +151,39 @@ out: static noinline struct snapshot_t *__snapshot_t_mut(struct bch_fs *c, u32 id) { size_t idx = U32_MAX - id; - size_t new_size; struct snapshot_table *new, *old; - new_size = max(16UL, roundup_pow_of_two(idx + 1)); + size_t new_bytes = kmalloc_size_roundup(struct_size(new, s, idx + 1)); + size_t new_size = (new_bytes - sizeof(*new)) / sizeof(new->s[0]); - new = kvzalloc(struct_size(new, s, new_size), GFP_KERNEL); + new = kvzalloc(new_bytes, GFP_KERNEL); if (!new) return NULL; + new->nr = new_size; + old = rcu_dereference_protected(c->snapshots, true); if (old) - memcpy(new->s, - rcu_dereference_protected(c->snapshots, true)->s, - sizeof(new->s[0]) * c->snapshot_table_size); + memcpy(new->s, old->s, sizeof(old->s[0]) * old->nr); rcu_assign_pointer(c->snapshots, new); - c->snapshot_table_size = new_size; - kvfree_rcu_mightsleep(old); + kvfree_rcu(old, rcu); - return &rcu_dereference_protected(c->snapshots, true)->s[idx]; + return &rcu_dereference_protected(c->snapshots, + lockdep_is_held(&c->snapshot_table_lock))->s[idx]; } static inline struct snapshot_t *snapshot_t_mut(struct bch_fs *c, u32 id) { size_t idx = U32_MAX - id; + struct snapshot_table *table = + rcu_dereference_protected(c->snapshots, + lockdep_is_held(&c->snapshot_table_lock)); lockdep_assert_held(&c->snapshot_table_lock); - if (likely(idx < c->snapshot_table_size)) - return &rcu_dereference_protected(c->snapshots, true)->s[idx]; + if (likely(table && idx < table->nr)) + return &table->s[idx]; return __snapshot_t_mut(c, id); } diff --git a/fs/bcachefs/snapshot.h b/fs/bcachefs/snapshot.h index 7c66ffc06385..8538b7fddfed 100644 --- a/fs/bcachefs/snapshot.h +++ b/fs/bcachefs/snapshot.h @@ -33,7 +33,11 @@ int bch2_mark_snapshot(struct btree_trans *, enum btree_id, unsigned, static inline struct snapshot_t *__snapshot_t(struct snapshot_table *t, u32 id) { - return &t->s[U32_MAX - id]; + u32 idx = U32_MAX - id; + + return likely(t && idx < t->nr) + ? &t->s[idx] + : NULL; } static inline const struct snapshot_t *snapshot_t(struct bch_fs *c, u32 id) diff --git a/fs/bcachefs/subvolume_types.h b/fs/bcachefs/subvolume_types.h index ae644adfc391..9b10c8947828 100644 --- a/fs/bcachefs/subvolume_types.h +++ b/fs/bcachefs/subvolume_types.h @@ -20,6 +20,8 @@ struct snapshot_t { }; struct snapshot_table { + struct rcu_head rcu; + size_t nr; #ifndef RUST_BINDGEN DECLARE_FLEX_ARRAY(struct snapshot_t, s); #else -- cgit From ec9cc18fc2e65b08c588e01f24aaeb71551a7132 Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Fri, 22 Mar 2024 16:29:23 -0400 Subject: bcachefs: Add checks for invalid snapshot IDs Previously, we assumed that keys were consistent with the snapshots btree - but that's not correct as fsck may not have been run or may not be complete. This adds checks and error handling when using the in-memory snapshots table (that mirrors the snapshots btree). Signed-off-by: Kent Overstreet --- fs/bcachefs/btree_trans_commit.c | 2 +- fs/bcachefs/data_update.c | 9 +++++++ fs/bcachefs/errcode.h | 3 ++- fs/bcachefs/snapshot.c | 8 ++++-- fs/bcachefs/snapshot.h | 57 ++++++++++++++++------------------------ 5 files changed, 40 insertions(+), 39 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/btree_trans_commit.c b/fs/bcachefs/btree_trans_commit.c index 30d69a6d133e..96669fede7d3 100644 --- a/fs/bcachefs/btree_trans_commit.c +++ b/fs/bcachefs/btree_trans_commit.c @@ -318,7 +318,7 @@ static inline void btree_insert_entry_checks(struct btree_trans *trans, !(i->flags & BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE) && test_bit(JOURNAL_REPLAY_DONE, &trans->c->journal.flags) && i->k->k.p.snapshot && - bch2_snapshot_is_internal_node(trans->c, i->k->k.p.snapshot)); + bch2_snapshot_is_internal_node(trans->c, i->k->k.p.snapshot) > 0); } static __always_inline int bch2_trans_journal_res_get(struct btree_trans *trans, diff --git a/fs/bcachefs/data_update.c b/fs/bcachefs/data_update.c index 4150feca42a2..b564404dad71 100644 --- a/fs/bcachefs/data_update.c +++ b/fs/bcachefs/data_update.c @@ -14,6 +14,7 @@ #include "move.h" #include "nocow_locking.h" #include "rebalance.h" +#include "snapshot.h" #include "subvolume.h" #include "trace.h" @@ -509,6 +510,14 @@ int bch2_data_update_init(struct btree_trans *trans, unsigned ptrs_locked = 0; int ret = 0; + /* + * fs is corrupt we have a key for a snapshot node that doesn't exist, + * and we have to check for this because we go rw before repairing the + * snapshots table - just skip it, we can move it later. + */ + if (unlikely(k.k->p.snapshot && !bch2_snapshot_equiv(c, k.k->p.snapshot))) + return -BCH_ERR_data_update_done; + bch2_bkey_buf_init(&m->k); bch2_bkey_buf_reassemble(&m->k, c, k); m->btree_id = btree_id; diff --git a/fs/bcachefs/errcode.h b/fs/bcachefs/errcode.h index af25d8ec60f2..01a79fa3eacb 100644 --- a/fs/bcachefs/errcode.h +++ b/fs/bcachefs/errcode.h @@ -252,7 +252,8 @@ x(BCH_ERR_nopromote, nopromote_in_flight) \ x(BCH_ERR_nopromote, nopromote_no_writes) \ x(BCH_ERR_nopromote, nopromote_enomem) \ - x(0, need_inode_lock) + x(0, need_inode_lock) \ + x(0, invalid_snapshot_node) enum bch_errcode { BCH_ERR_START = 2048, diff --git a/fs/bcachefs/snapshot.c b/fs/bcachefs/snapshot.c index 9cd71e613dc9..4e074136c490 100644 --- a/fs/bcachefs/snapshot.c +++ b/fs/bcachefs/snapshot.c @@ -93,8 +93,10 @@ static int bch2_snapshot_tree_create(struct btree_trans *trans, static bool __bch2_snapshot_is_ancestor_early(struct snapshot_table *t, u32 id, u32 ancestor) { - while (id && id < ancestor) - id = __snapshot_t(t, id)->parent; + while (id && id < ancestor) { + const struct snapshot_t *s = __snapshot_t(t, id); + id = s ? s->parent : 0; + } return id == ancestor; } @@ -110,6 +112,8 @@ static bool bch2_snapshot_is_ancestor_early(struct bch_fs *c, u32 id, u32 ancest static inline u32 get_ancestor_below(struct snapshot_table *t, u32 id, u32 ancestor) { const struct snapshot_t *s = __snapshot_t(t, id); + if (!s) + return 0; if (s->skip[2] <= ancestor) return s->skip[2]; diff --git a/fs/bcachefs/snapshot.h b/fs/bcachefs/snapshot.h index 8538b7fddfed..331f20fd8d03 100644 --- a/fs/bcachefs/snapshot.h +++ b/fs/bcachefs/snapshot.h @@ -48,7 +48,8 @@ static inline const struct snapshot_t *snapshot_t(struct bch_fs *c, u32 id) static inline u32 bch2_snapshot_tree(struct bch_fs *c, u32 id) { rcu_read_lock(); - id = snapshot_t(c, id)->tree; + const struct snapshot_t *s = snapshot_t(c, id); + id = s ? s->tree : 0; rcu_read_unlock(); return id; @@ -56,7 +57,8 @@ static inline u32 bch2_snapshot_tree(struct bch_fs *c, u32 id) static inline u32 __bch2_snapshot_parent_early(struct bch_fs *c, u32 id) { - return snapshot_t(c, id)->parent; + const struct snapshot_t *s = snapshot_t(c, id); + return s ? s->parent : 0; } static inline u32 bch2_snapshot_parent_early(struct bch_fs *c, u32 id) @@ -70,19 +72,19 @@ static inline u32 bch2_snapshot_parent_early(struct bch_fs *c, u32 id) static inline u32 __bch2_snapshot_parent(struct bch_fs *c, u32 id) { -#ifdef CONFIG_BCACHEFS_DEBUG - u32 parent = snapshot_t(c, id)->parent; + const struct snapshot_t *s = snapshot_t(c, id); + if (!s) + return 0; - if (parent && - snapshot_t(c, id)->depth != snapshot_t(c, parent)->depth + 1) + u32 parent = s->parent; + if (IS_ENABLED(CONFIG_BCACHEFS_DEBU) && + parent && + s->depth != snapshot_t(c, parent)->depth + 1) panic("id %u depth=%u parent %u depth=%u\n", id, snapshot_t(c, id)->depth, parent, snapshot_t(c, parent)->depth); return parent; -#else - return snapshot_t(c, id)->parent; -#endif } static inline u32 bch2_snapshot_parent(struct bch_fs *c, u32 id) @@ -120,7 +122,8 @@ static inline u32 bch2_snapshot_root(struct bch_fs *c, u32 id) static inline u32 __bch2_snapshot_equiv(struct bch_fs *c, u32 id) { - return snapshot_t(c, id)->equiv; + const struct snapshot_t *s = snapshot_t(c, id); + return s ? s->equiv : 0; } static inline u32 bch2_snapshot_equiv(struct bch_fs *c, u32 id) @@ -137,38 +140,22 @@ static inline bool bch2_snapshot_is_equiv(struct bch_fs *c, u32 id) return id == bch2_snapshot_equiv(c, id); } -static inline bool bch2_snapshot_is_internal_node(struct bch_fs *c, u32 id) +static inline int bch2_snapshot_is_internal_node(struct bch_fs *c, u32 id) { - const struct snapshot_t *s; - bool ret; - rcu_read_lock(); - s = snapshot_t(c, id); - ret = s->children[0]; + const struct snapshot_t *s = snapshot_t(c, id); + int ret = s ? s->children[0] : -BCH_ERR_invalid_snapshot_node; rcu_read_unlock(); return ret; } -static inline u32 bch2_snapshot_is_leaf(struct bch_fs *c, u32 id) +static inline int bch2_snapshot_is_leaf(struct bch_fs *c, u32 id) { - return !bch2_snapshot_is_internal_node(c, id); -} - -static inline u32 bch2_snapshot_sibling(struct bch_fs *c, u32 id) -{ - const struct snapshot_t *s; - u32 parent = __bch2_snapshot_parent(c, id); - - if (!parent) - return 0; - - s = snapshot_t(c, __bch2_snapshot_parent(c, id)); - if (id == s->children[0]) - return s->children[1]; - if (id == s->children[1]) - return s->children[0]; - return 0; + int ret = bch2_snapshot_is_internal_node(c, id); + if (ret < 0) + return ret; + return !ret; } static inline u32 bch2_snapshot_depth(struct bch_fs *c, u32 parent) @@ -253,7 +240,7 @@ static inline int bch2_key_has_snapshot_overwrites(struct btree_trans *trans, struct bpos pos) { if (!btree_type_has_snapshots(id) || - bch2_snapshot_is_leaf(trans->c, pos.snapshot)) + bch2_snapshot_is_leaf(trans->c, pos.snapshot) > 0) return 0; return __bch2_key_has_snapshot_overwrites(trans, id, pos); -- cgit From 57339b24a0eda5433751e7e0f4a8ea1e23315f60 Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Sat, 23 Mar 2024 19:29:55 -0400 Subject: bcachefs: Don't do extent merging before journal replay is finished We don't normally do extent updates this early in recovery, but some of the repair paths have to and when we do, we don't want to do anything that requires the snapshots table. Signed-off-by: Kent Overstreet --- fs/bcachefs/btree_update.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'fs') diff --git a/fs/bcachefs/btree_update.c b/fs/bcachefs/btree_update.c index a4b40c1656a5..8e47e260eba5 100644 --- a/fs/bcachefs/btree_update.c +++ b/fs/bcachefs/btree_update.c @@ -38,6 +38,9 @@ static noinline int extent_front_merge(struct btree_trans *trans, struct bkey_i *update; int ret; + if (unlikely(trans->journal_replay_not_finished)) + return 0; + update = bch2_bkey_make_mut_noupdate(trans, k); ret = PTR_ERR_OR_ZERO(update); if (ret) @@ -69,6 +72,9 @@ static noinline int extent_back_merge(struct btree_trans *trans, struct bch_fs *c = trans->c; int ret; + if (unlikely(trans->journal_replay_not_finished)) + return 0; + ret = bch2_key_has_snapshot_overwrites(trans, iter->btree_id, insert->k.p) ?: bch2_key_has_snapshot_overwrites(trans, iter->btree_id, k.k->p); if (ret < 0) -- cgit From 36f9ef109b1c6935928d09a3e73d744291f71545 Mon Sep 17 00:00:00 2001 From: Hongbo Li Date: Mon, 25 Mar 2024 10:50:48 +0800 Subject: bcachefs: fix trans->mem realloc in __bch2_trans_kmalloc The old code doesn't consider the mem alloced from mempool when call krealloc on trans->mem. Also in bch2_trans_put, using mempool_free to free trans->mem by condition "trans->mem_bytes == BTREE_TRANS_MEM_MAX" is inaccurate when trans->mem was allocated by krealloc function. Instead, we use used_mempool stuff to record the situation, and realloc or free the trans->mem in elegant way. Also, after krealloc failed in __bch2_trans_kmalloc, the old data should be copied to the new buffer when alloc from mempool_alloc. Fixes: 31403dca5bb1 ("bcachefs: optimize __bch2_trans_get(), kill DEBUG_TRANSACTIONS") Signed-off-by: Hongbo Li Signed-off-by: Kent Overstreet --- fs/bcachefs/btree_iter.c | 33 ++++++++++++++++++++++++++++++--- 1 file changed, 30 insertions(+), 3 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/btree_iter.c b/fs/bcachefs/btree_iter.c index 51bcdc6c6d1c..ae4a28a33ad1 100644 --- a/fs/bcachefs/btree_iter.c +++ b/fs/bcachefs/btree_iter.c @@ -2790,6 +2790,31 @@ void *__bch2_trans_kmalloc(struct btree_trans *trans, size_t size) struct btree_transaction_stats *s = btree_trans_stats(trans); s->max_mem = max(s->max_mem, new_bytes); + if (trans->used_mempool) { + if (trans->mem_bytes >= new_bytes) + goto out_change_top; + + /* No more space from mempool item, need malloc new one */ + new_mem = kmalloc(new_bytes, GFP_NOWAIT|__GFP_NOWARN); + if (unlikely(!new_mem)) { + bch2_trans_unlock(trans); + + new_mem = kmalloc(new_bytes, GFP_KERNEL); + if (!new_mem) + return ERR_PTR(-BCH_ERR_ENOMEM_trans_kmalloc); + + ret = bch2_trans_relock(trans); + if (ret) { + kfree(new_mem); + return ERR_PTR(ret); + } + } + memcpy(new_mem, trans->mem, trans->mem_top); + trans->used_mempool = false; + mempool_free(trans->mem, &c->btree_trans_mem_pool); + goto out_new_mem; + } + new_mem = krealloc(trans->mem, new_bytes, GFP_NOWAIT|__GFP_NOWARN); if (unlikely(!new_mem)) { bch2_trans_unlock(trans); @@ -2798,6 +2823,8 @@ void *__bch2_trans_kmalloc(struct btree_trans *trans, size_t size) if (!new_mem && new_bytes <= BTREE_TRANS_MEM_MAX) { new_mem = mempool_alloc(&c->btree_trans_mem_pool, GFP_KERNEL); new_bytes = BTREE_TRANS_MEM_MAX; + memcpy(new_mem, trans->mem, trans->mem_top); + trans->used_mempool = true; kfree(trans->mem); } @@ -2811,7 +2838,7 @@ void *__bch2_trans_kmalloc(struct btree_trans *trans, size_t size) if (ret) return ERR_PTR(ret); } - +out_new_mem: trans->mem = new_mem; trans->mem_bytes = new_bytes; @@ -2819,7 +2846,7 @@ void *__bch2_trans_kmalloc(struct btree_trans *trans, size_t size) trace_and_count(c, trans_restart_mem_realloced, trans, _RET_IP_, new_bytes); return ERR_PTR(btree_trans_restart(trans, BCH_ERR_transaction_restart_mem_realloced)); } - +out_change_top: p = trans->mem + trans->mem_top; trans->mem_top += size; memset(p, 0, size); @@ -3093,7 +3120,7 @@ void bch2_trans_put(struct btree_trans *trans) if (paths_allocated != trans->_paths_allocated) kvfree_rcu_mightsleep(paths_allocated); - if (trans->mem_bytes == BTREE_TRANS_MEM_MAX) + if (trans->used_mempool) mempool_free(trans->mem, &c->btree_trans_mem_pool); else kfree(trans->mem); -- cgit From 048f47e83fc315499dc1943176b3ebe1a55574fb Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Mon, 25 Mar 2024 13:51:29 -0400 Subject: bcachefs: btree_and_journal_iter now respects trans->journal_replay_not_finished btree_and_journal_iter is now safe to use at runtime, not just during recovery before journal keys have been freed. Signed-off-by: Kent Overstreet --- fs/bcachefs/btree_journal_iter.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/btree_journal_iter.c b/fs/bcachefs/btree_journal_iter.c index 50e04356d72c..0272ef0d7310 100644 --- a/fs/bcachefs/btree_journal_iter.c +++ b/fs/bcachefs/btree_journal_iter.c @@ -363,7 +363,7 @@ static void btree_and_journal_iter_prefetch(struct btree_and_journal_iter *_iter struct bkey_s_c bch2_btree_and_journal_iter_peek(struct btree_and_journal_iter *iter) { - struct bkey_s_c btree_k, journal_k, ret; + struct bkey_s_c btree_k, journal_k = bkey_s_c_null, ret; if (iter->prefetch && iter->journal.level) btree_and_journal_iter_prefetch(iter); @@ -375,9 +375,10 @@ again: bpos_lt(btree_k.k->p, iter->pos)) bch2_journal_iter_advance_btree(iter); - while ((journal_k = bch2_journal_iter_peek(&iter->journal)).k && - bpos_lt(journal_k.k->p, iter->pos)) - bch2_journal_iter_advance(&iter->journal); + if (iter->trans->journal_replay_not_finished) + while ((journal_k = bch2_journal_iter_peek(&iter->journal)).k && + bpos_lt(journal_k.k->p, iter->pos)) + bch2_journal_iter_advance(&iter->journal); ret = journal_k.k && (!btree_k.k || bpos_le(journal_k.k->p, btree_k.k->p)) @@ -435,7 +436,9 @@ void bch2_btree_and_journal_iter_init_node_iter(struct btree_trans *trans, bch2_btree_node_iter_init_from_start(&node_iter, b); __bch2_btree_and_journal_iter_init_node_iter(trans, iter, b, node_iter, b->data->min_key); - list_add(&iter->journal.list, &trans->c->journal_iters); + if (trans->journal_replay_not_finished && + !test_bit(BCH_FS_may_go_rw, &trans->c->flags)) + list_add(&iter->journal.list, &trans->c->journal_iters); } /* sort and dedup all keys in the journal: */ -- cgit From 40cb26233a060aeb936de7ea1f6ac2659ed9951c Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Tue, 26 Mar 2024 17:14:43 -0400 Subject: bcachefs: Be careful about btree node splits during journal replay Don't pick a pivot that's going to be deleted. Signed-off-by: Kent Overstreet --- fs/bcachefs/btree_journal_iter.c | 16 ++++++++++++++++ fs/bcachefs/btree_journal_iter.h | 4 ++-- fs/bcachefs/btree_update_interior.c | 9 ++++++++- 3 files changed, 26 insertions(+), 3 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/btree_journal_iter.c b/fs/bcachefs/btree_journal_iter.c index 0272ef0d7310..1f588264575d 100644 --- a/fs/bcachefs/btree_journal_iter.c +++ b/fs/bcachefs/btree_journal_iter.c @@ -261,6 +261,22 @@ int bch2_journal_key_delete(struct bch_fs *c, enum btree_id id, return bch2_journal_key_insert(c, id, level, &whiteout); } +bool bch2_key_deleted_in_journal(struct btree_trans *trans, enum btree_id btree, + unsigned level, struct bpos pos) +{ + struct journal_keys *keys = &trans->c->journal_keys; + size_t idx = bch2_journal_key_search(keys, btree, level, pos); + + if (!trans->journal_replay_not_finished) + return false; + + return (idx < keys->size && + keys->data[idx].btree_id == btree && + keys->data[idx].level == level && + bpos_eq(keys->data[idx].k->k.p, pos) && + bkey_deleted(&keys->data[idx].k->k)); +} + void bch2_journal_key_overwritten(struct bch_fs *c, enum btree_id btree, unsigned level, struct bpos pos) { diff --git a/fs/bcachefs/btree_journal_iter.h b/fs/bcachefs/btree_journal_iter.h index c9d19da3ea04..09cd53091a05 100644 --- a/fs/bcachefs/btree_journal_iter.h +++ b/fs/bcachefs/btree_journal_iter.h @@ -40,8 +40,8 @@ int bch2_journal_key_insert(struct bch_fs *, enum btree_id, unsigned, struct bkey_i *); int bch2_journal_key_delete(struct bch_fs *, enum btree_id, unsigned, struct bpos); -void bch2_journal_key_overwritten(struct bch_fs *, enum btree_id, - unsigned, struct bpos); +bool bch2_key_deleted_in_journal(struct btree_trans *, enum btree_id, unsigned, struct bpos); +void bch2_journal_key_overwritten(struct bch_fs *, enum btree_id, unsigned, struct bpos); void bch2_btree_and_journal_iter_advance(struct btree_and_journal_iter *); struct bkey_s_c bch2_btree_and_journal_iter_peek(struct btree_and_journal_iter *); diff --git a/fs/bcachefs/btree_update_interior.c b/fs/bcachefs/btree_update_interior.c index b2f5f2e50f7e..b45c7ced15d8 100644 --- a/fs/bcachefs/btree_update_interior.c +++ b/fs/bcachefs/btree_update_interior.c @@ -1380,9 +1380,16 @@ static void __btree_split_node(struct btree_update *as, if (bkey_deleted(k)) continue; + uk = bkey_unpack_key(b, k); + + if (b->c.level && + u64s < n1_u64s && + u64s + k->u64s >= n1_u64s && + bch2_key_deleted_in_journal(trans, b->c.btree_id, b->c.level, uk.p)) + n1_u64s += k->u64s; + i = u64s >= n1_u64s; u64s += k->u64s; - uk = bkey_unpack_key(b, k); if (!i) n1_pos = uk.p; bch2_bkey_format_add_key(&format[i], &uk); -- cgit From 79032b078173f87a13f8618cdab710798be67314 Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Sat, 23 Mar 2024 19:29:19 -0400 Subject: bcachefs: Improved topology repair checks Consolidate bch2_gc_check_topology() and btree_node_interior_verify(), and replace them with an improved version, bch2_btree_node_check_topology(). This checks that children of an interior node correctly span the full range of the parent node with no overlaps. Also, ensure that topology repairs at runtime are always a fatal error; in particular, this adds a check in btree_iter_down() - if we don't find a key while walking down the btree that's indicative of a topology error and should be flagged as such, not a null ptr deref. Some checks in btree_update_interior.c remaining BUG_ONS(), because we already checked the node for topology errors when starting the update, and the assertions indicate that we _just_ corrupted the btree node - i.e. the problem can't be that existing on disk corruption, they indicate an actual algorithmic bug. In the future, we'll be annotating the fsck errors list with which recovery pass corrects them; the open coded "run explicit recovery pass or fatal error" in bch2_btree_node_check_topology() will in the future be done for every fsck_err() call. Signed-off-by: Kent Overstreet --- fs/bcachefs/btree_cache.c | 3 +- fs/bcachefs/btree_gc.c | 133 ++++------------------------------- fs/bcachefs/btree_io.c | 2 +- fs/bcachefs/btree_iter.c | 19 ++++- fs/bcachefs/btree_update_interior.c | 134 +++++++++++++++++++++++++----------- fs/bcachefs/btree_update_interior.h | 2 + fs/bcachefs/error.h | 6 ++ fs/bcachefs/sb-errors_types.h | 3 +- 8 files changed, 137 insertions(+), 165 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/btree_cache.c b/fs/bcachefs/btree_cache.c index 8ff21b2e1463..84474324dba9 100644 --- a/fs/bcachefs/btree_cache.c +++ b/fs/bcachefs/btree_cache.c @@ -808,7 +808,8 @@ static noinline void btree_bad_header(struct bch_fs *c, struct btree *b) prt_printf(&buf, "\nmax "); bch2_bpos_to_text(&buf, b->data->max_key); - bch2_fs_inconsistent(c, "%s", buf.buf); + bch2_fs_topology_error(c, "%s", buf.buf); + printbuf_exit(&buf); } diff --git a/fs/bcachefs/btree_gc.c b/fs/bcachefs/btree_gc.c index bdaed29f084a..6d32ac9be243 100644 --- a/fs/bcachefs/btree_gc.c +++ b/fs/bcachefs/btree_gc.c @@ -70,90 +70,6 @@ static inline void gc_pos_set(struct bch_fs *c, struct gc_pos new_pos) __gc_pos_set(c, new_pos); } -/* - * Missing: if an interior btree node is empty, we need to do something - - * perhaps just kill it - */ -static int bch2_gc_check_topology(struct bch_fs *c, - struct btree *b, - struct bkey_buf *prev, - struct bkey_buf cur, - bool is_last) -{ - struct bpos node_start = b->data->min_key; - struct bpos node_end = b->data->max_key; - struct bpos expected_start = bkey_deleted(&prev->k->k) - ? node_start - : bpos_successor(prev->k->k.p); - struct printbuf buf1 = PRINTBUF, buf2 = PRINTBUF; - int ret = 0; - - if (cur.k->k.type == KEY_TYPE_btree_ptr_v2) { - struct bkey_i_btree_ptr_v2 *bp = bkey_i_to_btree_ptr_v2(cur.k); - - if (!bpos_eq(expected_start, bp->v.min_key)) { - bch2_topology_error(c); - - if (bkey_deleted(&prev->k->k)) { - prt_printf(&buf1, "start of node: "); - bch2_bpos_to_text(&buf1, node_start); - } else { - bch2_bkey_val_to_text(&buf1, c, bkey_i_to_s_c(prev->k)); - } - bch2_bkey_val_to_text(&buf2, c, bkey_i_to_s_c(cur.k)); - - if (__fsck_err(c, - FSCK_CAN_FIX| - FSCK_CAN_IGNORE| - FSCK_NO_RATELIMIT, - btree_node_topology_bad_min_key, - "btree node with incorrect min_key at btree %s level %u:\n" - " prev %s\n" - " cur %s", - bch2_btree_id_str(b->c.btree_id), b->c.level, - buf1.buf, buf2.buf) && should_restart_for_topology_repair(c)) { - bch_info(c, "Halting mark and sweep to start topology repair pass"); - ret = bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_check_topology); - goto err; - } else { - set_bit(BCH_FS_initial_gc_unfixed, &c->flags); - } - } - } - - if (is_last && !bpos_eq(cur.k->k.p, node_end)) { - bch2_topology_error(c); - - printbuf_reset(&buf1); - printbuf_reset(&buf2); - - bch2_bkey_val_to_text(&buf1, c, bkey_i_to_s_c(cur.k)); - bch2_bpos_to_text(&buf2, node_end); - - if (__fsck_err(c, FSCK_CAN_FIX|FSCK_CAN_IGNORE|FSCK_NO_RATELIMIT, - btree_node_topology_bad_max_key, - "btree node with incorrect max_key at btree %s level %u:\n" - " %s\n" - " expected %s", - bch2_btree_id_str(b->c.btree_id), b->c.level, - buf1.buf, buf2.buf) && - should_restart_for_topology_repair(c)) { - bch_info(c, "Halting mark and sweep to start topology repair pass"); - ret = bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_check_topology); - goto err; - } else { - set_bit(BCH_FS_initial_gc_unfixed, &c->flags); - } - } - - bch2_bkey_buf_copy(prev, c, cur.k); -err: -fsck_err: - printbuf_exit(&buf2); - printbuf_exit(&buf1); - return ret; -} - static void btree_ptr_to_v2(struct btree *b, struct bkey_i_btree_ptr_v2 *dst) { switch (b->key.k.type) { @@ -445,6 +361,7 @@ again: prev = NULL; if (ret == DROP_PREV_NODE) { + bch_info(c, "dropped prev node"); bch2_btree_node_evict(trans, prev_k.k); ret = bch2_journal_key_delete(c, b->c.btree_id, b->c.level, prev_k.k->k.p); @@ -841,42 +758,30 @@ err: static int btree_gc_mark_node(struct btree_trans *trans, struct btree *b, bool initial) { - struct bch_fs *c = trans->c; struct btree_node_iter iter; struct bkey unpacked; struct bkey_s_c k; - struct bkey_buf prev, cur; int ret = 0; + ret = bch2_btree_node_check_topology(trans, b); + if (ret) + return ret; + if (!btree_node_type_needs_gc(btree_node_type(b))) return 0; bch2_btree_node_iter_init_from_start(&iter, b); - bch2_bkey_buf_init(&prev); - bch2_bkey_buf_init(&cur); - bkey_init(&prev.k->k); while ((k = bch2_btree_node_iter_peek_unpack(&iter, b, &unpacked)).k) { ret = bch2_gc_mark_key(trans, b->c.btree_id, b->c.level, false, &k, initial); if (ret) - break; + return ret; bch2_btree_node_iter_advance(&iter, b); - - if (b->c.level) { - bch2_bkey_buf_reassemble(&cur, c, k); - - ret = bch2_gc_check_topology(c, b, &prev, cur, - bch2_btree_node_iter_end(&iter)); - if (ret) - break; - } } - bch2_bkey_buf_exit(&cur, c); - bch2_bkey_buf_exit(&prev, c); - return ret; + return 0; } static int bch2_gc_btree(struct btree_trans *trans, enum btree_id btree_id, @@ -925,14 +830,16 @@ static int bch2_gc_btree_init_recurse(struct btree_trans *trans, struct btree *b struct bch_fs *c = trans->c; struct btree_and_journal_iter iter; struct bkey_s_c k; - struct bkey_buf cur, prev; + struct bkey_buf cur; struct printbuf buf = PRINTBUF; int ret = 0; + ret = bch2_btree_node_check_topology(trans, b); + if (ret) + return ret; + bch2_btree_and_journal_iter_init_node_iter(trans, &iter, b); - bch2_bkey_buf_init(&prev); bch2_bkey_buf_init(&cur); - bkey_init(&prev.k->k); while ((k = bch2_btree_and_journal_iter_peek(&iter)).k) { BUG_ON(bpos_lt(k.k->p, b->data->min_key)); @@ -943,20 +850,7 @@ static int bch2_gc_btree_init_recurse(struct btree_trans *trans, struct btree *b if (ret) goto fsck_err; - if (b->c.level) { - bch2_bkey_buf_reassemble(&cur, c, k); - k = bkey_i_to_s_c(cur.k); - - bch2_btree_and_journal_iter_advance(&iter); - - ret = bch2_gc_check_topology(c, b, - &prev, cur, - !bch2_btree_and_journal_iter_peek(&iter).k); - if (ret) - goto fsck_err; - } else { - bch2_btree_and_journal_iter_advance(&iter); - } + bch2_btree_and_journal_iter_advance(&iter); } if (b->c.level > target_depth) { @@ -1015,7 +909,6 @@ static int bch2_gc_btree_init_recurse(struct btree_trans *trans, struct btree *b } fsck_err: bch2_bkey_buf_exit(&cur, c); - bch2_bkey_buf_exit(&prev, c); bch2_btree_and_journal_iter_exit(&iter); printbuf_exit(&buf); return ret; diff --git a/fs/bcachefs/btree_io.c b/fs/bcachefs/btree_io.c index 34df8ccc5fec..b85e2cd61c38 100644 --- a/fs/bcachefs/btree_io.c +++ b/fs/bcachefs/btree_io.c @@ -1657,7 +1657,7 @@ void bch2_btree_node_read(struct btree_trans *trans, struct btree *b, prt_str(&buf, "btree node read error: no device to read from\n at "); bch2_btree_pos_to_text(&buf, c, b); - bch_err(c, "%s", buf.buf); + bch_err_ratelimited(c, "%s", buf.buf); if (c->recovery_passes_explicit & BIT_ULL(BCH_RECOVERY_PASS_check_topology) && c->curr_recovery_pass > BCH_RECOVERY_PASS_check_topology) diff --git a/fs/bcachefs/btree_iter.c b/fs/bcachefs/btree_iter.c index ae4a28a33ad1..2a211a4bebd1 100644 --- a/fs/bcachefs/btree_iter.c +++ b/fs/bcachefs/btree_iter.c @@ -927,8 +927,22 @@ static __always_inline int btree_path_down(struct btree_trans *trans, if (ret) goto err; } else { - bch2_bkey_buf_unpack(&tmp, c, l->b, - bch2_btree_node_iter_peek(&l->iter, l->b)); + struct bkey_packed *k = bch2_btree_node_iter_peek(&l->iter, l->b); + if (!k) { + struct printbuf buf = PRINTBUF; + + prt_str(&buf, "node not found at pos "); + bch2_bpos_to_text(&buf, path->pos); + prt_str(&buf, " within parent node "); + bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&l->b->key)); + + bch2_fs_fatal_error(c, "%s", buf.buf); + printbuf_exit(&buf); + ret = -BCH_ERR_btree_need_topology_repair; + goto err; + } + + bch2_bkey_buf_unpack(&tmp, c, l->b, k); if ((flags & BTREE_ITER_PREFETCH) && c->opts.btree_node_prefetch) { @@ -962,7 +976,6 @@ err: return ret; } - static int bch2_btree_path_traverse_all(struct btree_trans *trans) { struct bch_fs *c = trans->c; diff --git a/fs/bcachefs/btree_update_interior.c b/fs/bcachefs/btree_update_interior.c index b45c7ced15d8..aae7a2687eee 100644 --- a/fs/bcachefs/btree_update_interior.c +++ b/fs/bcachefs/btree_update_interior.c @@ -2,6 +2,7 @@ #include "bcachefs.h" #include "alloc_foreground.h" +#include "bkey_buf.h" #include "bkey_methods.h" #include "btree_cache.h" #include "btree_gc.h" @@ -18,6 +19,7 @@ #include "journal.h" #include "journal_reclaim.h" #include "keylist.h" +#include "recovery.h" #include "replicas.h" #include "super-io.h" #include "trace.h" @@ -44,56 +46,103 @@ static btree_path_idx_t get_unlocked_mut_path(struct btree_trans *trans, return path_idx; } -/* Debug code: */ - /* * Verify that child nodes correctly span parent node's range: */ -static void btree_node_interior_verify(struct bch_fs *c, struct btree *b) +int bch2_btree_node_check_topology(struct btree_trans *trans, struct btree *b) { -#ifdef CONFIG_BCACHEFS_DEBUG - struct bpos next_node = b->data->min_key; - struct btree_node_iter iter; + struct bch_fs *c = trans->c; + struct bpos node_min = b->key.k.type == KEY_TYPE_btree_ptr_v2 + ? bkey_i_to_btree_ptr_v2(&b->key)->v.min_key + : b->data->min_key; + struct btree_and_journal_iter iter; struct bkey_s_c k; - struct bkey_s_c_btree_ptr_v2 bp; - struct bkey unpacked; - struct printbuf buf1 = PRINTBUF, buf2 = PRINTBUF; + struct printbuf buf = PRINTBUF; + struct bkey_buf prev; + int ret = 0; - BUG_ON(!b->c.level); + BUG_ON(b->key.k.type == KEY_TYPE_btree_ptr_v2 && + !bpos_eq(bkey_i_to_btree_ptr_v2(&b->key)->v.min_key, + b->data->min_key)); - if (!test_bit(JOURNAL_REPLAY_DONE, &c->journal.flags)) - return; + if (!b->c.level) + return 0; - bch2_btree_node_iter_init_from_start(&iter, b); + bch2_bkey_buf_init(&prev); + bkey_init(&prev.k->k); + bch2_btree_and_journal_iter_init_node_iter(trans, &iter, b); - while (1) { - k = bch2_btree_node_iter_peek_unpack(&iter, b, &unpacked); + while ((k = bch2_btree_and_journal_iter_peek(&iter)).k) { if (k.k->type != KEY_TYPE_btree_ptr_v2) - break; - bp = bkey_s_c_to_btree_ptr_v2(k); + goto out; - if (!bpos_eq(next_node, bp.v->min_key)) { - bch2_dump_btree_node(c, b); - bch2_bpos_to_text(&buf1, next_node); - bch2_bpos_to_text(&buf2, bp.v->min_key); - panic("expected next min_key %s got %s\n", buf1.buf, buf2.buf); - } + struct bkey_s_c_btree_ptr_v2 bp = bkey_s_c_to_btree_ptr_v2(k); - bch2_btree_node_iter_advance(&iter, b); + struct bpos expected_min = bkey_deleted(&prev.k->k) + ? node_min + : bpos_successor(prev.k->k.p); - if (bch2_btree_node_iter_end(&iter)) { - if (!bpos_eq(k.k->p, b->key.k.p)) { - bch2_dump_btree_node(c, b); - bch2_bpos_to_text(&buf1, b->key.k.p); - bch2_bpos_to_text(&buf2, k.k->p); - panic("expected end %s got %s\n", buf1.buf, buf2.buf); - } - break; + if (!bpos_eq(expected_min, bp.v->min_key)) { + bch2_topology_error(c); + + printbuf_reset(&buf); + prt_str(&buf, "end of prev node doesn't match start of next node\n"), + prt_printf(&buf, " in btree %s level %u node ", + bch2_btree_id_str(b->c.btree_id), b->c.level); + bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&b->key)); + prt_str(&buf, "\n prev "); + bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(prev.k)); + prt_str(&buf, "\n next "); + bch2_bkey_val_to_text(&buf, c, k); + + need_fsck_err(c, btree_node_topology_bad_min_key, "%s", buf.buf); + goto topology_repair; } - next_node = bpos_successor(k.k->p); + bch2_bkey_buf_reassemble(&prev, c, k); + bch2_btree_and_journal_iter_advance(&iter); + } + + if (bkey_deleted(&prev.k->k)) { + bch2_topology_error(c); + + printbuf_reset(&buf); + prt_str(&buf, "empty interior node\n"); + prt_printf(&buf, " in btree %s level %u node ", + bch2_btree_id_str(b->c.btree_id), b->c.level); + bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&b->key)); + + need_fsck_err(c, btree_node_topology_empty_interior_node, "%s", buf.buf); + goto topology_repair; + } else if (!bpos_eq(prev.k->k.p, b->key.k.p)) { + bch2_topology_error(c); + + printbuf_reset(&buf); + prt_str(&buf, "last child node doesn't end at end of parent node\n"); + prt_printf(&buf, " in btree %s level %u node ", + bch2_btree_id_str(b->c.btree_id), b->c.level); + bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&b->key)); + prt_str(&buf, "\n last key "); + bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(prev.k)); + + need_fsck_err(c, btree_node_topology_bad_max_key, "%s", buf.buf); + goto topology_repair; + } +out: +fsck_err: + bch2_btree_and_journal_iter_exit(&iter); + bch2_bkey_buf_exit(&prev, c); + printbuf_exit(&buf); + return ret; +topology_repair: + if ((c->recovery_passes_explicit & BIT_ULL(BCH_RECOVERY_PASS_check_topology)) && + c->curr_recovery_pass > BCH_RECOVERY_PASS_check_topology) { + bch2_inconsistent_error(c); + ret = -BCH_ERR_btree_need_topology_repair; + } else { + ret = bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_check_topology); } -#endif + goto out; } /* Calculate ideal packed bkey format for new btree nodes: */ @@ -1448,8 +1497,7 @@ static void __btree_split_node(struct btree_update *as, bch2_verify_btree_nr_keys(n[i]); - if (b->c.level) - btree_node_interior_verify(as->c, n[i]); + BUG_ON(bch2_btree_node_check_topology(trans, n[i])); } } @@ -1480,7 +1528,7 @@ static void btree_split_insert_keys(struct btree_update *as, __bch2_btree_insert_keys_interior(as, trans, path, b, node_iter, keys); - btree_node_interior_verify(as->c, b); + BUG_ON(bch2_btree_node_check_topology(trans, b)); } } @@ -1498,6 +1546,10 @@ static int btree_split(struct btree_update *as, struct btree_trans *trans, BUG_ON(!parent && (b != btree_node_root(c, b))); BUG_ON(parent && !btree_node_intent_locked(trans->paths + path, b->c.level + 1)); + ret = bch2_btree_node_check_topology(trans, b); + if (ret) + return ret; + bch2_btree_interior_update_will_free_node(as, b); if (b->nr.live_u64s > BTREE_SPLIT_THRESHOLD(c)) { @@ -1717,7 +1769,11 @@ static int bch2_btree_insert_node(struct btree_update *as, struct btree_trans *t goto split; } - btree_node_interior_verify(c, b); + ret = bch2_btree_node_check_topology(trans, b); + if (ret) { + bch2_btree_node_unlock_write(trans, path, b); + return ret; + } bch2_btree_insert_keys_interior(as, trans, path, b, keys); @@ -1735,7 +1791,7 @@ static int bch2_btree_insert_node(struct btree_update *as, struct btree_trans *t bch2_btree_node_unlock_write(trans, path, b); - btree_node_interior_verify(c, b); + BUG_ON(bch2_btree_node_check_topology(trans, b)); return 0; split: /* diff --git a/fs/bcachefs/btree_update_interior.h b/fs/bcachefs/btree_update_interior.h index f651dd48aaa0..7ac3e17b84fd 100644 --- a/fs/bcachefs/btree_update_interior.h +++ b/fs/bcachefs/btree_update_interior.h @@ -10,6 +10,8 @@ #define BTREE_UPDATE_JOURNAL_RES (BTREE_UPDATE_NODES_MAX * (BKEY_BTREE_PTR_U64s_MAX + 1)) +int bch2_btree_node_check_topology(struct btree_trans *, struct btree *); + /* * Tracks an in progress split/rewrite of a btree node and the update to the * parent node: diff --git a/fs/bcachefs/error.h b/fs/bcachefs/error.h index ae1d6674c512..36caedf72d89 100644 --- a/fs/bcachefs/error.h +++ b/fs/bcachefs/error.h @@ -32,6 +32,12 @@ bool bch2_inconsistent_error(struct bch_fs *); int bch2_topology_error(struct bch_fs *); +#define bch2_fs_topology_error(c, ...) \ +({ \ + bch_err(c, "btree topology error: " __VA_ARGS__); \ + bch2_topology_error(c); \ +}) + #define bch2_fs_inconsistent(c, ...) \ ({ \ bch_err(c, __VA_ARGS__); \ diff --git a/fs/bcachefs/sb-errors_types.h b/fs/bcachefs/sb-errors_types.h index 5178bf579f7c..8edbd9ef994c 100644 --- a/fs/bcachefs/sb-errors_types.h +++ b/fs/bcachefs/sb-errors_types.h @@ -265,7 +265,8 @@ x(subvol_children_bad, 257) \ x(subvol_loop, 258) \ x(subvol_unreachable, 259) \ - x(btree_node_bkey_bad_u64s, 260) + x(btree_node_bkey_bad_u64s, 260) \ + x(btree_node_topology_empty_interior_node, 261) enum bch_sb_error_id { #define x(t, n) BCH_FSCK_ERR_##t = n, -- cgit From bb66009958b277a9baffaa53d835661852550151 Mon Sep 17 00:00:00 2001 From: zhuxiaohui Date: Tue, 26 Mar 2024 20:03:45 +0800 Subject: bcachefs: add REQ_SYNC and REQ_IDLE in write dio when writing file with direct_IO on bcachefs, then performance is much lower than other fs due to write back throttle in block layer: wbt_wait+1 __rq_qos_throttle+32 blk_mq_submit_bio+394 submit_bio_noacct_nocheck+649 bch2_submit_wbio_replicas+538 __bch2_write+2539 bch2_direct_write+1663 bch2_write_iter+318 aio_write+355 io_submit_one+1224 __x64_sys_io_submit+169 do_syscall_64+134 entry_SYSCALL_64_after_hwframe+110 add set REQ_SYNC and REQ_IDLE in bio->bi_opf as standard dirct-io Signed-off-by: zhuxiaohui Signed-off-by: Kent Overstreet --- fs/bcachefs/fs-io-direct.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/fs-io-direct.c b/fs/bcachefs/fs-io-direct.c index 33cb6da3a5ad..f49e6c0f0f68 100644 --- a/fs/bcachefs/fs-io-direct.c +++ b/fs/bcachefs/fs-io-direct.c @@ -536,7 +536,7 @@ static __always_inline long bch2_dio_write_loop(struct dio_write *dio) if (likely(!dio->iter.count) || dio->op.error) break; - bio_reset(bio, NULL, REQ_OP_WRITE); + bio_reset(bio, NULL, REQ_OP_WRITE | REQ_SYNC | REQ_IDLE); } out: return bch2_dio_write_done(dio); @@ -618,7 +618,7 @@ ssize_t bch2_direct_write(struct kiocb *req, struct iov_iter *iter) bio = bio_alloc_bioset(NULL, bio_iov_vecs_to_alloc(iter, BIO_MAX_VECS), - REQ_OP_WRITE, + REQ_OP_WRITE | REQ_SYNC | REQ_IDLE, GFP_KERNEL, &c->dio_write_bioset); dio = container_of(bio, struct dio_write, op.wbio.bio); -- cgit From 805b535a8afbcd8073a03eb25aafd82cb816bff6 Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Mon, 25 Mar 2024 21:58:07 -0400 Subject: bcachefs: Check btree ptr min_key in .invalid Signed-off-by: Kent Overstreet --- fs/bcachefs/extents.c | 9 +++++++-- fs/bcachefs/sb-errors_types.h | 3 ++- 2 files changed, 9 insertions(+), 3 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/extents.c b/fs/bcachefs/extents.c index 61395b113df9..b2432d88cda6 100644 --- a/fs/bcachefs/extents.c +++ b/fs/bcachefs/extents.c @@ -189,13 +189,18 @@ int bch2_btree_ptr_v2_invalid(struct bch_fs *c, struct bkey_s_c k, enum bkey_invalid_flags flags, struct printbuf *err) { + struct bkey_s_c_btree_ptr_v2 bp = bkey_s_c_to_btree_ptr_v2(k); int ret = 0; - bkey_fsck_err_on(bkey_val_u64s(k.k) > BKEY_BTREE_PTR_VAL_U64s_MAX, c, err, - btree_ptr_v2_val_too_big, + bkey_fsck_err_on(bkey_val_u64s(k.k) > BKEY_BTREE_PTR_VAL_U64s_MAX, + c, err, btree_ptr_v2_val_too_big, "value too big (%zu > %zu)", bkey_val_u64s(k.k), BKEY_BTREE_PTR_VAL_U64s_MAX); + bkey_fsck_err_on(bpos_ge(bp.v->min_key, bp.k->p), + c, err, btree_ptr_v2_min_key_bad, + "min_key > key"); + ret = bch2_bkey_ptrs_invalid(c, k, flags, err); fsck_err: return ret; diff --git a/fs/bcachefs/sb-errors_types.h b/fs/bcachefs/sb-errors_types.h index 8edbd9ef994c..73e9634df8ff 100644 --- a/fs/bcachefs/sb-errors_types.h +++ b/fs/bcachefs/sb-errors_types.h @@ -266,7 +266,8 @@ x(subvol_loop, 258) \ x(subvol_unreachable, 259) \ x(btree_node_bkey_bad_u64s, 260) \ - x(btree_node_topology_empty_interior_node, 261) + x(btree_node_topology_empty_interior_node, 261) \ + x(btree_ptr_v2_min_key_bad, 262) enum bch_sb_error_id { #define x(t, n) BCH_FSCK_ERR_##t = n, -- cgit From 812a9297936a959c98a2e9e44a9a622bbe30b162 Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Tue, 26 Mar 2024 16:39:54 -0400 Subject: bcachefs: Fix btree node keys accounting in topology repair path When dropping keys now outside a now because we're changing the node min/max, we need to redo the node's accounting as well. Signed-off-by: Kent Overstreet --- fs/bcachefs/bset.c | 14 ++++++++++---- fs/bcachefs/bset.h | 2 ++ fs/bcachefs/btree_io.c | 1 + fs/bcachefs/btree_update_interior.c | 1 + 4 files changed, 14 insertions(+), 4 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/bset.c b/fs/bcachefs/bset.c index 3fd1085b6c61..3bb477840eab 100644 --- a/fs/bcachefs/bset.c +++ b/fs/bcachefs/bset.c @@ -134,18 +134,24 @@ void bch2_dump_btree_node_iter(struct btree *b, printbuf_exit(&buf); } -#ifdef CONFIG_BCACHEFS_DEBUG - -void __bch2_verify_btree_nr_keys(struct btree *b) +struct btree_nr_keys bch2_btree_node_count_keys(struct btree *b) { struct bset_tree *t; struct bkey_packed *k; - struct btree_nr_keys nr = { 0 }; + struct btree_nr_keys nr = {}; for_each_bset(b, t) bset_tree_for_each_key(b, t, k) if (!bkey_deleted(k)) btree_keys_account_key_add(&nr, t - b->set, k); + return nr; +} + +#ifdef CONFIG_BCACHEFS_DEBUG + +void __bch2_verify_btree_nr_keys(struct btree *b) +{ + struct btree_nr_keys nr = bch2_btree_node_count_keys(b); BUG_ON(memcmp(&nr, &b->nr, sizeof(nr))); } diff --git a/fs/bcachefs/bset.h b/fs/bcachefs/bset.h index 79c77baaa383..120a79fd456b 100644 --- a/fs/bcachefs/bset.h +++ b/fs/bcachefs/bset.h @@ -458,6 +458,8 @@ struct bkey_s_c bch2_btree_node_iter_peek_unpack(struct btree_node_iter *, /* Accounting: */ +struct btree_nr_keys bch2_btree_node_count_keys(struct btree *); + static inline void btree_keys_account_key(struct btree_nr_keys *n, unsigned bset, struct bkey_packed *k, diff --git a/fs/bcachefs/btree_io.c b/fs/bcachefs/btree_io.c index b85e2cd61c38..9c71e6fb9c41 100644 --- a/fs/bcachefs/btree_io.c +++ b/fs/bcachefs/btree_io.c @@ -654,6 +654,7 @@ void bch2_btree_node_drop_keys_outside_node(struct btree *b) */ bch2_bset_set_no_aux_tree(b, b->set); bch2_btree_build_aux_trees(b); + b->nr = bch2_btree_node_count_keys(b); struct bkey_s_c k; struct bkey unpacked; diff --git a/fs/bcachefs/btree_update_interior.c b/fs/bcachefs/btree_update_interior.c index aae7a2687eee..826cebe4bce0 100644 --- a/fs/bcachefs/btree_update_interior.c +++ b/fs/bcachefs/btree_update_interior.c @@ -1543,6 +1543,7 @@ static int btree_split(struct btree_update *as, struct btree_trans *trans, u64 start_time = local_clock(); int ret = 0; + bch2_verify_btree_nr_keys(b); BUG_ON(!parent && (b != btree_node_root(c, b))); BUG_ON(parent && !btree_node_intent_locked(trans->paths + path, b->c.level + 1)); -- cgit From 6f5869ffd9f111b81b95b73c6e54f07406591911 Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Tue, 26 Mar 2024 17:38:22 -0400 Subject: bcachefs: Fix use after free in bch2_check_fix_ptrs() Signed-off-by: Kent Overstreet --- fs/bcachefs/btree_gc.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/btree_gc.c b/fs/bcachefs/btree_gc.c index 6d32ac9be243..26e51af9acdc 100644 --- a/fs/bcachefs/btree_gc.c +++ b/fs/bcachefs/btree_gc.c @@ -691,12 +691,6 @@ found: } } - ret = bch2_journal_key_insert_take(c, btree_id, level, new); - if (ret) { - kfree(new); - goto err; - } - if (level) bch2_btree_node_update_key_early(trans, btree_id, level - 1, *k, new); @@ -710,6 +704,12 @@ found: bch_info(c, "new key %s", buf.buf); } + ret = bch2_journal_key_insert_take(c, btree_id, level, new); + if (ret) { + kfree(new); + goto err; + } + *k = bkey_i_to_s_c(new); } err: -- cgit From 83bb58539045b15653b61c6e8eb65f3f9c671cdf Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Tue, 26 Mar 2024 18:46:20 -0400 Subject: bcachefs: Fix repair path for missing indirect extents Signed-off-by: Kent Overstreet --- fs/bcachefs/reflink.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/reflink.c b/fs/bcachefs/reflink.c index c47c66c2b394..ff7864731a07 100644 --- a/fs/bcachefs/reflink.c +++ b/fs/bcachefs/reflink.c @@ -185,8 +185,7 @@ not_found: } else { bkey_error_init(update); update->k.p = p.k->p; - update->k.p.offset = next_idx; - update->k.size = next_idx - *idx; + update->k.size = p.k->size; set_bkey_val_u64s(&update->k, 0); } -- cgit From dcc1c04587aa9bc3515153f4c89cff73f2cb45b2 Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Tue, 26 Mar 2024 18:46:38 -0400 Subject: bcachefs: Fix use after free in check_root_trans() Signed-off-by: Kent Overstreet --- fs/bcachefs/fsck.c | 18 +++++++++++------- 1 file changed, 11 insertions(+), 7 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/fsck.c b/fs/bcachefs/fsck.c index 47d4eefaba7b..6d8367ab5ddd 100644 --- a/fs/bcachefs/fsck.c +++ b/fs/bcachefs/fsck.c @@ -2098,17 +2098,21 @@ static int check_root_trans(struct btree_trans *trans) if (mustfix_fsck_err_on(ret, c, root_subvol_missing, "root subvol missing")) { - struct bkey_i_subvolume root_subvol; + struct bkey_i_subvolume *root_subvol = + bch2_trans_kmalloc(trans, sizeof(*root_subvol)); + ret = PTR_ERR_OR_ZERO(root_subvol); + if (ret) + goto err; snapshot = U32_MAX; inum = BCACHEFS_ROOT_INO; - bkey_subvolume_init(&root_subvol.k_i); - root_subvol.k.p.offset = BCACHEFS_ROOT_SUBVOL; - root_subvol.v.flags = 0; - root_subvol.v.snapshot = cpu_to_le32(snapshot); - root_subvol.v.inode = cpu_to_le64(inum); - ret = bch2_btree_insert_trans(trans, BTREE_ID_subvolumes, &root_subvol.k_i, 0); + bkey_subvolume_init(&root_subvol->k_i); + root_subvol->k.p.offset = BCACHEFS_ROOT_SUBVOL; + root_subvol->v.flags = 0; + root_subvol->v.snapshot = cpu_to_le32(snapshot); + root_subvol->v.inode = cpu_to_le64(inum); + ret = bch2_btree_insert_trans(trans, BTREE_ID_subvolumes, &root_subvol->k_i, 0); bch_err_msg(c, ret, "writing root subvol"); if (ret) goto err; -- cgit From 47d2080e30b0b9fc636eba4e74f9e4bdc01543d7 Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Mon, 25 Mar 2024 19:26:05 -0400 Subject: bcachefs: Kill bch2_bkey_ptr_data_type() Remove some duplication, and inconsistency between check_fix_ptrs and the main ptr marking paths Signed-off-by: Kent Overstreet --- fs/bcachefs/backpointers.c | 6 ++--- fs/bcachefs/backpointers.h | 32 ++++++++++++++++++++----- fs/bcachefs/btree_gc.c | 59 +++++++++++++++++++++++++--------------------- fs/bcachefs/buckets.c | 12 ++++++---- fs/bcachefs/extents.h | 24 ------------------- 5 files changed, 67 insertions(+), 66 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/backpointers.c b/fs/bcachefs/backpointers.c index 21b3ae7df824..7537d78aa9db 100644 --- a/fs/bcachefs/backpointers.c +++ b/fs/bcachefs/backpointers.c @@ -29,8 +29,7 @@ static bool extent_matches_bp(struct bch_fs *c, if (p.ptr.cached) continue; - bch2_extent_ptr_to_bp(c, btree_id, level, k, p, - &bucket2, &bp2); + bch2_extent_ptr_to_bp(c, btree_id, level, k, p, entry, &bucket2, &bp2); if (bpos_eq(bucket, bucket2) && !memcmp(&bp, &bp2, sizeof(bp))) return true; @@ -507,8 +506,7 @@ static int check_extent_to_backpointers(struct btree_trans *trans, if (p.ptr.cached) continue; - bch2_extent_ptr_to_bp(c, btree, level, - k, p, &bucket_pos, &bp); + bch2_extent_ptr_to_bp(c, btree, level, k, p, entry, &bucket_pos, &bp); ret = check_bp_exists(trans, s, bucket_pos, bp, k); if (ret) diff --git a/fs/bcachefs/backpointers.h b/fs/bcachefs/backpointers.h index 327365a9feac..da012ca7daee 100644 --- a/fs/bcachefs/backpointers.h +++ b/fs/bcachefs/backpointers.h @@ -90,20 +90,40 @@ static inline int bch2_bucket_backpointer_mod(struct btree_trans *trans, return bch2_trans_update_buffered(trans, BTREE_ID_backpointers, &bp_k.k_i); } -static inline enum bch_data_type bkey_ptr_data_type(enum btree_id btree_id, unsigned level, - struct bkey_s_c k, struct extent_ptr_decoded p) +static inline enum bch_data_type bch2_bkey_ptr_data_type(struct bkey_s_c k, + struct extent_ptr_decoded p, + const union bch_extent_entry *entry) { - return level ? BCH_DATA_btree : - p.has_ec ? BCH_DATA_stripe : - BCH_DATA_user; + switch (k.k->type) { + case KEY_TYPE_btree_ptr: + case KEY_TYPE_btree_ptr_v2: + return BCH_DATA_btree; + case KEY_TYPE_extent: + case KEY_TYPE_reflink_v: + return p.has_ec ? BCH_DATA_stripe : BCH_DATA_user; + case KEY_TYPE_stripe: { + const struct bch_extent_ptr *ptr = &entry->ptr; + struct bkey_s_c_stripe s = bkey_s_c_to_stripe(k); + + BUG_ON(ptr < s.v->ptrs || + ptr >= s.v->ptrs + s.v->nr_blocks); + + return ptr >= s.v->ptrs + s.v->nr_blocks - s.v->nr_redundant + ? BCH_DATA_parity + : BCH_DATA_user; + } + default: + BUG(); + } } static inline void bch2_extent_ptr_to_bp(struct bch_fs *c, enum btree_id btree_id, unsigned level, struct bkey_s_c k, struct extent_ptr_decoded p, + const union bch_extent_entry *entry, struct bpos *bucket_pos, struct bch_backpointer *bp) { - enum bch_data_type data_type = bkey_ptr_data_type(btree_id, level, k, p); + enum bch_data_type data_type = bch2_bkey_ptr_data_type(k, p, entry); s64 sectors = level ? btree_sectors(c) : k.k->size; u32 bucket_offset; diff --git a/fs/bcachefs/btree_gc.c b/fs/bcachefs/btree_gc.c index 26e51af9acdc..0afefccf4e52 100644 --- a/fs/bcachefs/btree_gc.c +++ b/fs/bcachefs/btree_gc.c @@ -7,6 +7,7 @@ #include "bcachefs.h" #include "alloc_background.h" #include "alloc_foreground.h" +#include "backpointers.h" #include "bkey_methods.h" #include "bkey_buf.h" #include "btree_journal_iter.h" @@ -508,7 +509,7 @@ static int bch2_check_fix_ptrs(struct btree_trans *trans, enum btree_id btree_id bkey_for_each_ptr_decode(k->k, ptrs_c, p, entry_c) { struct bch_dev *ca = bch_dev_bkey_exists(c, p.ptr.dev); struct bucket *g = PTR_GC_BUCKET(ca, &p.ptr); - enum bch_data_type data_type = bch2_bkey_ptr_data_type(*k, &entry_c->ptr); + enum bch_data_type data_type = bch2_bkey_ptr_data_type(*k, p, entry_c); if (fsck_err_on(!g->gen_valid, c, ptr_to_missing_alloc_key, @@ -574,7 +575,8 @@ static int bch2_check_fix_ptrs(struct btree_trans *trans, enum btree_id btree_id continue; if (fsck_err_on(bucket_data_type(g->data_type) && - bucket_data_type(g->data_type) != data_type, c, + bucket_data_type(g->data_type) != + bucket_data_type(data_type), c, ptr_bucket_data_type_mismatch, "bucket %u:%zu different types of data in same bucket: %s, %s\n" "while marking %s", @@ -615,18 +617,13 @@ static int bch2_check_fix_ptrs(struct btree_trans *trans, enum btree_id btree_id } if (do_update) { - struct bkey_ptrs ptrs; - union bch_extent_entry *entry; - struct bch_extent_ptr *ptr; - struct bkey_i *new; - if (is_root) { bch_err(c, "cannot update btree roots yet"); ret = -EINVAL; goto err; } - new = kmalloc(bkey_bytes(k->k), GFP_KERNEL); + struct bkey_i *new = kmalloc(bkey_bytes(k->k), GFP_KERNEL); if (!new) { ret = -BCH_ERR_ENOMEM_gc_repair_key; bch_err_msg(c, ret, "allocating new key"); @@ -641,7 +638,7 @@ static int bch2_check_fix_ptrs(struct btree_trans *trans, enum btree_id btree_id * btree node isn't there anymore, the read path will * sort it out: */ - ptrs = bch2_bkey_ptrs(bkey_i_to_s(new)); + struct bkey_ptrs ptrs = bch2_bkey_ptrs(bkey_i_to_s(new)); bkey_for_each_ptr(ptrs, ptr) { struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev); struct bucket *g = PTR_GC_BUCKET(ca, ptr); @@ -649,19 +646,26 @@ static int bch2_check_fix_ptrs(struct btree_trans *trans, enum btree_id btree_id ptr->gen = g->gen; } } else { - bch2_bkey_drop_ptrs(bkey_i_to_s(new), ptr, ({ - struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev); - struct bucket *g = PTR_GC_BUCKET(ca, ptr); - enum bch_data_type data_type = bch2_bkey_ptr_data_type(*k, ptr); - - (ptr->cached && - (!g->gen_valid || gen_cmp(ptr->gen, g->gen) > 0)) || - (!ptr->cached && - gen_cmp(ptr->gen, g->gen) < 0) || - gen_cmp(g->gen, ptr->gen) > BUCKET_GC_GEN_MAX || - (g->data_type && - g->data_type != data_type); - })); + struct bkey_ptrs ptrs; + union bch_extent_entry *entry; +restart_drop_ptrs: + ptrs = bch2_bkey_ptrs(bkey_i_to_s(new)); + bkey_for_each_ptr_decode(bkey_i_to_s(new).k, ptrs, p, entry) { + struct bch_dev *ca = bch_dev_bkey_exists(c, p.ptr.dev); + struct bucket *g = PTR_GC_BUCKET(ca, &p.ptr); + enum bch_data_type data_type = bch2_bkey_ptr_data_type(bkey_i_to_s_c(new), p, entry); + + if ((p.ptr.cached && + (!g->gen_valid || gen_cmp(p.ptr.gen, g->gen) > 0)) || + (!p.ptr.cached && + gen_cmp(p.ptr.gen, g->gen) < 0) || + gen_cmp(g->gen, p.ptr.gen) > BUCKET_GC_GEN_MAX || + (g->data_type && + g->data_type != data_type)) { + bch2_bkey_drop_ptr(bkey_i_to_s(new), &entry->ptr); + goto restart_drop_ptrs; + } + } again: ptrs = bch2_bkey_ptrs(bkey_i_to_s(new)); bkey_extent_entry_for_each(ptrs, entry) { @@ -736,10 +740,6 @@ static int bch2_gc_mark_key(struct btree_trans *trans, enum btree_id btree_id, BUG_ON(bch2_journal_seq_verify && k->k->version.lo > atomic64_read(&c->journal.seq)); - ret = bch2_check_fix_ptrs(trans, btree_id, level, is_root, k); - if (ret) - goto err; - if (fsck_err_on(k->k->version.lo > atomic64_read(&c->key_version), c, bkey_version_in_future, "key version number higher than recorded: %llu > %llu", @@ -748,8 +748,13 @@ static int bch2_gc_mark_key(struct btree_trans *trans, enum btree_id btree_id, atomic64_set(&c->key_version, k->k->version.lo); } + ret = bch2_check_fix_ptrs(trans, btree_id, level, is_root, k); + if (ret) + goto err; + ret = commit_do(trans, NULL, NULL, 0, - bch2_key_trigger(trans, btree_id, level, old, unsafe_bkey_s_c_to_s(*k), BTREE_TRIGGER_GC)); + bch2_key_trigger(trans, btree_id, level, old, + unsafe_bkey_s_c_to_s(*k), BTREE_TRIGGER_GC)); fsck_err: err: bch_err_fn(c, ret); diff --git a/fs/bcachefs/buckets.c b/fs/bcachefs/buckets.c index 96edf2c34d43..941401a210f5 100644 --- a/fs/bcachefs/buckets.c +++ b/fs/bcachefs/buckets.c @@ -525,6 +525,7 @@ int bch2_mark_metadata_bucket(struct bch_fs *c, struct bch_dev *ca, "different types of data in same bucket: %s, %s", bch2_data_type_str(g->data_type), bch2_data_type_str(data_type))) { + BUG(); ret = -EIO; goto err; } @@ -628,6 +629,7 @@ int bch2_check_bucket_ref(struct btree_trans *trans, bch2_data_type_str(ptr_data_type), (printbuf_reset(&buf), bch2_bkey_val_to_text(&buf, c, k), buf.buf)); + BUG(); ret = -EIO; goto err; } @@ -815,14 +817,14 @@ static int __mark_pointer(struct btree_trans *trans, static int bch2_trigger_pointer(struct btree_trans *trans, enum btree_id btree_id, unsigned level, struct bkey_s_c k, struct extent_ptr_decoded p, - s64 *sectors, - unsigned flags) + const union bch_extent_entry *entry, + s64 *sectors, unsigned flags) { bool insert = !(flags & BTREE_TRIGGER_OVERWRITE); struct bpos bucket; struct bch_backpointer bp; - bch2_extent_ptr_to_bp(trans->c, btree_id, level, k, p, &bucket, &bp); + bch2_extent_ptr_to_bp(trans->c, btree_id, level, k, p, entry, &bucket, &bp); *sectors = insert ? bp.bucket_len : -((s64) bp.bucket_len); if (flags & BTREE_TRIGGER_TRANSACTIONAL) { @@ -851,7 +853,7 @@ static int bch2_trigger_pointer(struct btree_trans *trans, if (flags & BTREE_TRIGGER_GC) { struct bch_fs *c = trans->c; struct bch_dev *ca = bch_dev_bkey_exists(c, p.ptr.dev); - enum bch_data_type data_type = bkey_ptr_data_type(btree_id, level, k, p); + enum bch_data_type data_type = bch2_bkey_ptr_data_type(k, p, entry); percpu_down_read(&c->mark_lock); struct bucket *g = PTR_GC_BUCKET(ca, &p.ptr); @@ -979,7 +981,7 @@ static int __trigger_extent(struct btree_trans *trans, bkey_for_each_ptr_decode(k.k, ptrs, p, entry) { s64 disk_sectors; - ret = bch2_trigger_pointer(trans, btree_id, level, k, p, &disk_sectors, flags); + ret = bch2_trigger_pointer(trans, btree_id, level, k, p, entry, &disk_sectors, flags); if (ret < 0) return ret; diff --git a/fs/bcachefs/extents.h b/fs/bcachefs/extents.h index fd2669cdd76f..3fd0169b98c1 100644 --- a/fs/bcachefs/extents.h +++ b/fs/bcachefs/extents.h @@ -596,30 +596,6 @@ static inline struct bch_devs_list bch2_bkey_cached_devs(struct bkey_s_c k) return ret; } -static inline unsigned bch2_bkey_ptr_data_type(struct bkey_s_c k, const struct bch_extent_ptr *ptr) -{ - switch (k.k->type) { - case KEY_TYPE_btree_ptr: - case KEY_TYPE_btree_ptr_v2: - return BCH_DATA_btree; - case KEY_TYPE_extent: - case KEY_TYPE_reflink_v: - return BCH_DATA_user; - case KEY_TYPE_stripe: { - struct bkey_s_c_stripe s = bkey_s_c_to_stripe(k); - - BUG_ON(ptr < s.v->ptrs || - ptr >= s.v->ptrs + s.v->nr_blocks); - - return ptr >= s.v->ptrs + s.v->nr_blocks - s.v->nr_redundant - ? BCH_DATA_parity - : BCH_DATA_user; - } - default: - BUG(); - } -} - unsigned bch2_bkey_nr_ptrs(struct bkey_s_c); unsigned bch2_bkey_nr_ptrs_allocated(struct bkey_s_c); unsigned bch2_bkey_nr_ptrs_fully_allocated(struct bkey_s_c); -- cgit From 7f9e5080366726084eb765a5d689bdf502e7e2ed Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Thu, 14 Mar 2024 19:39:26 -0400 Subject: bcachefs: Fix bch2_btree_increase_depth() When we haven't yet allocated any btree nodes for a given btree, we first need to call the regular split path to allocate one. Signed-off-by: Kent Overstreet --- fs/bcachefs/btree_update_interior.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/btree_update_interior.c b/fs/bcachefs/btree_update_interior.c index 826cebe4bce0..9c4325786ba8 100644 --- a/fs/bcachefs/btree_update_interior.c +++ b/fs/bcachefs/btree_update_interior.c @@ -1882,9 +1882,12 @@ int bch2_btree_increase_depth(struct btree_trans *trans, btree_path_idx_t path, { struct bch_fs *c = trans->c; struct btree *b = bch2_btree_id_root(c, trans->paths[path].btree_id)->b; + + if (btree_node_fake(b)) + return bch2_btree_split_leaf(trans, path, flags); + struct btree_update *as = - bch2_btree_update_start(trans, trans->paths + path, - b->c.level, true, flags); + bch2_btree_update_start(trans, trans->paths + path, b->c.level, true, flags); if (IS_ERR(as)) return PTR_ERR(as); -- cgit From 11d5568d3e04a2e6734d1eccc394cfcf5ca8523c Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Thu, 28 Mar 2024 01:41:03 -0400 Subject: bcachefs: fix backpointer for missing alloc key msg Signed-off-by: Kent Overstreet --- fs/bcachefs/backpointers.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/bcachefs/backpointers.c b/fs/bcachefs/backpointers.c index 7537d78aa9db..762c8ddfc5e7 100644 --- a/fs/bcachefs/backpointers.c +++ b/fs/bcachefs/backpointers.c @@ -382,7 +382,7 @@ static int bch2_check_btree_backpointer(struct btree_trans *trans, struct btree_ backpointer_to_missing_alloc, "backpointer for nonexistent alloc key: %llu:%llu:0\n%s", alloc_iter.pos.inode, alloc_iter.pos.offset, - (bch2_bkey_val_to_text(&buf, c, alloc_k), buf.buf))) { + (bch2_bkey_val_to_text(&buf, c, k), buf.buf))) { ret = bch2_btree_delete_at(trans, bp_iter, 0); goto out; } -- cgit From d2554263adcb4041f3608cb7476f102fda036ccc Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Sat, 23 Mar 2024 20:07:46 -0400 Subject: bcachefs: Split out recovery_passes.c We've grown a fair amount of code for managing recovery passes; tracking which ones we're running, which ones need to be run, and flagging in the superblock which ones need to be run on the next recovery. So it's worth splitting out into its own file, this code is pretty different from the code in recovery.c. Signed-off-by: Kent Overstreet --- fs/bcachefs/Makefile | 1 + fs/bcachefs/bcachefs.h | 2 +- fs/bcachefs/btree_gc.c | 2 +- fs/bcachefs/btree_update_interior.c | 2 +- fs/bcachefs/chardev.c | 2 +- fs/bcachefs/error.c | 2 +- fs/bcachefs/fsck.c | 2 +- fs/bcachefs/recovery.c | 249 +----------------------------------- fs/bcachefs/recovery.h | 32 +---- fs/bcachefs/recovery_passes.c | 203 +++++++++++++++++++++++++++++ fs/bcachefs/recovery_passes.h | 16 +++ fs/bcachefs/recovery_passes_types.h | 69 ++++++++++ fs/bcachefs/recovery_types.h | 68 ---------- fs/bcachefs/sb-downgrade.c | 2 +- fs/bcachefs/subvolume.c | 72 +++++++++++ fs/bcachefs/subvolume.h | 3 + fs/bcachefs/super-io.c | 2 +- 17 files changed, 378 insertions(+), 351 deletions(-) create mode 100644 fs/bcachefs/recovery_passes.c create mode 100644 fs/bcachefs/recovery_passes.h create mode 100644 fs/bcachefs/recovery_passes_types.h delete mode 100644 fs/bcachefs/recovery_types.h (limited to 'fs') diff --git a/fs/bcachefs/Makefile b/fs/bcachefs/Makefile index b02796c8a595..f6d86eb4b976 100644 --- a/fs/bcachefs/Makefile +++ b/fs/bcachefs/Makefile @@ -67,6 +67,7 @@ bcachefs-y := \ quota.o \ rebalance.o \ recovery.o \ + recovery_passes.o \ reflink.o \ replicas.o \ sb-clean.o \ diff --git a/fs/bcachefs/bcachefs.h b/fs/bcachefs/bcachefs.h index a9ade0b1f78c..963162a627cd 100644 --- a/fs/bcachefs/bcachefs.h +++ b/fs/bcachefs/bcachefs.h @@ -209,7 +209,7 @@ #include "fifo.h" #include "nocow_locking_types.h" #include "opts.h" -#include "recovery_types.h" +#include "recovery_passes_types.h" #include "sb-errors_types.h" #include "seqmutex.h" #include "time_stats.h" diff --git a/fs/bcachefs/btree_gc.c b/fs/bcachefs/btree_gc.c index 0afefccf4e52..e5d2c6daa663 100644 --- a/fs/bcachefs/btree_gc.c +++ b/fs/bcachefs/btree_gc.c @@ -25,7 +25,7 @@ #include "journal.h" #include "keylist.h" #include "move.h" -#include "recovery.h" +#include "recovery_passes.h" #include "reflink.h" #include "replicas.h" #include "super-io.h" diff --git a/fs/bcachefs/btree_update_interior.c b/fs/bcachefs/btree_update_interior.c index 9c4325786ba8..983bb27298cc 100644 --- a/fs/bcachefs/btree_update_interior.c +++ b/fs/bcachefs/btree_update_interior.c @@ -19,7 +19,7 @@ #include "journal.h" #include "journal_reclaim.h" #include "keylist.h" -#include "recovery.h" +#include "recovery_passes.h" #include "replicas.h" #include "super-io.h" #include "trace.h" diff --git a/fs/bcachefs/chardev.c b/fs/bcachefs/chardev.c index 38defa19d52d..cbfa6459bdbc 100644 --- a/fs/bcachefs/chardev.c +++ b/fs/bcachefs/chardev.c @@ -7,7 +7,7 @@ #include "chardev.h" #include "journal.h" #include "move.h" -#include "recovery.h" +#include "recovery_passes.h" #include "replicas.h" #include "super.h" #include "super-io.h" diff --git a/fs/bcachefs/error.c b/fs/bcachefs/error.c index 043431206799..f942a3947496 100644 --- a/fs/bcachefs/error.c +++ b/fs/bcachefs/error.c @@ -1,7 +1,7 @@ // SPDX-License-Identifier: GPL-2.0 #include "bcachefs.h" #include "error.h" -#include "recovery.h" +#include "recovery_passes.h" #include "super.h" #include "thread_with_file.h" diff --git a/fs/bcachefs/fsck.c b/fs/bcachefs/fsck.c index 6d8367ab5ddd..aca6fae409cd 100644 --- a/fs/bcachefs/fsck.c +++ b/fs/bcachefs/fsck.c @@ -12,7 +12,7 @@ #include "fsck.h" #include "inode.h" #include "keylist.h" -#include "recovery.h" +#include "recovery_passes.h" #include "snapshot.h" #include "super.h" #include "xattr.h" diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c index 03f9d6afe467..f9e2b011f6d4 100644 --- a/fs/bcachefs/recovery.c +++ b/fs/bcachefs/recovery.c @@ -1,35 +1,30 @@ // SPDX-License-Identifier: GPL-2.0 #include "bcachefs.h" -#include "backpointers.h" -#include "bkey_buf.h" #include "alloc_background.h" -#include "btree_gc.h" +#include "bkey_buf.h" #include "btree_journal_iter.h" #include "btree_update.h" #include "btree_update_interior.h" #include "btree_io.h" #include "buckets.h" #include "dirent.h" -#include "ec.h" #include "errcode.h" #include "error.h" #include "fs-common.h" -#include "fsck.h" #include "journal_io.h" #include "journal_reclaim.h" #include "journal_seq_blacklist.h" -#include "lru.h" #include "logged_ops.h" #include "move.h" #include "quota.h" #include "rebalance.h" #include "recovery.h" +#include "recovery_passes.h" #include "replicas.h" #include "sb-clean.h" #include "sb-downgrade.h" #include "snapshot.h" -#include "subvolume.h" #include "super-io.h" #include @@ -186,7 +181,7 @@ static int journal_sort_seq_cmp(const void *_l, const void *_r) return cmp_int(l->journal_seq, r->journal_seq); } -static int bch2_journal_replay(struct bch_fs *c) +int bch2_journal_replay(struct bch_fs *c) { struct journal_keys *keys = &c->journal_keys; DARRAY(struct journal_key *) keys_sorted = { 0 }; @@ -471,150 +466,6 @@ fsck_err: return ret; } -static int bch2_initialize_subvolumes(struct bch_fs *c) -{ - struct bkey_i_snapshot_tree root_tree; - struct bkey_i_snapshot root_snapshot; - struct bkey_i_subvolume root_volume; - int ret; - - bkey_snapshot_tree_init(&root_tree.k_i); - root_tree.k.p.offset = 1; - root_tree.v.master_subvol = cpu_to_le32(1); - root_tree.v.root_snapshot = cpu_to_le32(U32_MAX); - - bkey_snapshot_init(&root_snapshot.k_i); - root_snapshot.k.p.offset = U32_MAX; - root_snapshot.v.flags = 0; - root_snapshot.v.parent = 0; - root_snapshot.v.subvol = cpu_to_le32(BCACHEFS_ROOT_SUBVOL); - root_snapshot.v.tree = cpu_to_le32(1); - SET_BCH_SNAPSHOT_SUBVOL(&root_snapshot.v, true); - - bkey_subvolume_init(&root_volume.k_i); - root_volume.k.p.offset = BCACHEFS_ROOT_SUBVOL; - root_volume.v.flags = 0; - root_volume.v.snapshot = cpu_to_le32(U32_MAX); - root_volume.v.inode = cpu_to_le64(BCACHEFS_ROOT_INO); - - ret = bch2_btree_insert(c, BTREE_ID_snapshot_trees, &root_tree.k_i, NULL, 0) ?: - bch2_btree_insert(c, BTREE_ID_snapshots, &root_snapshot.k_i, NULL, 0) ?: - bch2_btree_insert(c, BTREE_ID_subvolumes, &root_volume.k_i, NULL, 0); - bch_err_fn(c, ret); - return ret; -} - -static int __bch2_fs_upgrade_for_subvolumes(struct btree_trans *trans) -{ - struct btree_iter iter; - struct bkey_s_c k; - struct bch_inode_unpacked inode; - int ret; - - k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_inodes, - SPOS(0, BCACHEFS_ROOT_INO, U32_MAX), 0); - ret = bkey_err(k); - if (ret) - return ret; - - if (!bkey_is_inode(k.k)) { - bch_err(trans->c, "root inode not found"); - ret = -BCH_ERR_ENOENT_inode; - goto err; - } - - ret = bch2_inode_unpack(k, &inode); - BUG_ON(ret); - - inode.bi_subvol = BCACHEFS_ROOT_SUBVOL; - - ret = bch2_inode_write(trans, &iter, &inode); -err: - bch2_trans_iter_exit(trans, &iter); - return ret; -} - -/* set bi_subvol on root inode */ -noinline_for_stack -static int bch2_fs_upgrade_for_subvolumes(struct bch_fs *c) -{ - int ret = bch2_trans_do(c, NULL, NULL, BCH_TRANS_COMMIT_lazy_rw, - __bch2_fs_upgrade_for_subvolumes(trans)); - bch_err_fn(c, ret); - return ret; -} - -const char * const bch2_recovery_passes[] = { -#define x(_fn, ...) #_fn, - BCH_RECOVERY_PASSES() -#undef x - NULL -}; - -static int bch2_check_allocations(struct bch_fs *c) -{ - return bch2_gc(c, true, c->opts.norecovery); -} - -static int bch2_set_may_go_rw(struct bch_fs *c) -{ - struct journal_keys *keys = &c->journal_keys; - - /* - * After we go RW, the journal keys buffer can't be modified (except for - * setting journal_key->overwritten: it will be accessed by multiple - * threads - */ - move_gap(keys, keys->nr); - - set_bit(BCH_FS_may_go_rw, &c->flags); - - if (keys->nr || c->opts.fsck || !c->sb.clean) - return bch2_fs_read_write_early(c); - return 0; -} - -struct recovery_pass_fn { - int (*fn)(struct bch_fs *); - unsigned when; -}; - -static struct recovery_pass_fn recovery_pass_fns[] = { -#define x(_fn, _id, _when) { .fn = bch2_##_fn, .when = _when }, - BCH_RECOVERY_PASSES() -#undef x -}; - -u64 bch2_recovery_passes_to_stable(u64 v) -{ - static const u8 map[] = { -#define x(n, id, ...) [BCH_RECOVERY_PASS_##n] = BCH_RECOVERY_PASS_STABLE_##n, - BCH_RECOVERY_PASSES() -#undef x - }; - - u64 ret = 0; - for (unsigned i = 0; i < ARRAY_SIZE(map); i++) - if (v & BIT_ULL(i)) - ret |= BIT_ULL(map[i]); - return ret; -} - -u64 bch2_recovery_passes_from_stable(u64 v) -{ - static const u8 map[] = { -#define x(n, id, ...) [BCH_RECOVERY_PASS_STABLE_##n] = BCH_RECOVERY_PASS_##n, - BCH_RECOVERY_PASSES() -#undef x - }; - - u64 ret = 0; - for (unsigned i = 0; i < ARRAY_SIZE(map); i++) - if (v & BIT_ULL(i)) - ret |= BIT_ULL(map[i]); - return ret; -} - static bool check_version_upgrade(struct bch_fs *c) { unsigned latest_version = bcachefs_metadata_version_current; @@ -687,96 +538,6 @@ static bool check_version_upgrade(struct bch_fs *c) return false; } -u64 bch2_fsck_recovery_passes(void) -{ - u64 ret = 0; - - for (unsigned i = 0; i < ARRAY_SIZE(recovery_pass_fns); i++) - if (recovery_pass_fns[i].when & PASS_FSCK) - ret |= BIT_ULL(i); - return ret; -} - -static bool should_run_recovery_pass(struct bch_fs *c, enum bch_recovery_pass pass) -{ - struct recovery_pass_fn *p = recovery_pass_fns + pass; - - if (c->opts.norecovery && pass > BCH_RECOVERY_PASS_snapshots_read) - return false; - if (c->recovery_passes_explicit & BIT_ULL(pass)) - return true; - if ((p->when & PASS_FSCK) && c->opts.fsck) - return true; - if ((p->when & PASS_UNCLEAN) && !c->sb.clean) - return true; - if (p->when & PASS_ALWAYS) - return true; - return false; -} - -static int bch2_run_recovery_pass(struct bch_fs *c, enum bch_recovery_pass pass) -{ - struct recovery_pass_fn *p = recovery_pass_fns + pass; - int ret; - - if (!(p->when & PASS_SILENT)) - bch2_print(c, KERN_INFO bch2_log_msg(c, "%s..."), - bch2_recovery_passes[pass]); - ret = p->fn(c); - if (ret) - return ret; - if (!(p->when & PASS_SILENT)) - bch2_print(c, KERN_CONT " done\n"); - - return 0; -} - -static int bch2_run_recovery_passes(struct bch_fs *c) -{ - int ret = 0; - - while (c->curr_recovery_pass < ARRAY_SIZE(recovery_pass_fns)) { - if (should_run_recovery_pass(c, c->curr_recovery_pass)) { - unsigned pass = c->curr_recovery_pass; - - ret = bch2_run_recovery_pass(c, c->curr_recovery_pass); - if (bch2_err_matches(ret, BCH_ERR_restart_recovery) || - (ret && c->curr_recovery_pass < pass)) - continue; - if (ret) - break; - - c->recovery_passes_complete |= BIT_ULL(c->curr_recovery_pass); - } - c->curr_recovery_pass++; - c->recovery_pass_done = max(c->recovery_pass_done, c->curr_recovery_pass); - } - - return ret; -} - -int bch2_run_online_recovery_passes(struct bch_fs *c) -{ - int ret = 0; - - for (unsigned i = 0; i < ARRAY_SIZE(recovery_pass_fns); i++) { - struct recovery_pass_fn *p = recovery_pass_fns + i; - - if (!(p->when & PASS_ONLINE)) - continue; - - ret = bch2_run_recovery_pass(c, i); - if (bch2_err_matches(ret, BCH_ERR_restart_recovery)) { - i = c->curr_recovery_pass; - continue; - } - if (ret) - break; - } - - return ret; -} - int bch2_fs_recovery(struct bch_fs *c) { struct bch_sb_field_clean *clean = NULL; @@ -1155,7 +916,7 @@ int bch2_fs_initialize(struct bch_fs *c) } mutex_unlock(&c->sb_lock); - c->curr_recovery_pass = ARRAY_SIZE(recovery_pass_fns); + c->curr_recovery_pass = BCH_RECOVERY_PASS_NR; set_bit(BCH_FS_may_go_rw, &c->flags); for (unsigned i = 0; i < BTREE_ID_NR; i++) @@ -1230,7 +991,7 @@ int bch2_fs_initialize(struct bch_fs *c) if (ret) goto err; - c->recovery_pass_done = ARRAY_SIZE(recovery_pass_fns) - 1; + c->recovery_pass_done = BCH_RECOVERY_PASS_NR - 1; if (enabled_qtypes(c)) { ret = bch2_fs_quota_read(c); diff --git a/fs/bcachefs/recovery.h b/fs/bcachefs/recovery.h index 4e9d24719b2e..3962fd87b50d 100644 --- a/fs/bcachefs/recovery.h +++ b/fs/bcachefs/recovery.h @@ -2,37 +2,7 @@ #ifndef _BCACHEFS_RECOVERY_H #define _BCACHEFS_RECOVERY_H -extern const char * const bch2_recovery_passes[]; - -u64 bch2_recovery_passes_to_stable(u64 v); -u64 bch2_recovery_passes_from_stable(u64 v); - -/* - * For when we need to rewind recovery passes and run a pass we skipped: - */ -static inline int bch2_run_explicit_recovery_pass(struct bch_fs *c, - enum bch_recovery_pass pass) -{ - if (c->recovery_passes_explicit & BIT_ULL(pass)) - return 0; - - bch_info(c, "running explicit recovery pass %s (%u), currently at %s (%u)", - bch2_recovery_passes[pass], pass, - bch2_recovery_passes[c->curr_recovery_pass], c->curr_recovery_pass); - - c->recovery_passes_explicit |= BIT_ULL(pass); - - if (c->curr_recovery_pass >= pass) { - c->curr_recovery_pass = pass; - c->recovery_passes_complete &= (1ULL << pass) >> 1; - return -BCH_ERR_restart_recovery; - } else { - return 0; - } -} - -int bch2_run_online_recovery_passes(struct bch_fs *); -u64 bch2_fsck_recovery_passes(void); +int bch2_journal_replay(struct bch_fs *); int bch2_fs_recovery(struct bch_fs *); int bch2_fs_initialize(struct bch_fs *); diff --git a/fs/bcachefs/recovery_passes.c b/fs/bcachefs/recovery_passes.c new file mode 100644 index 000000000000..bab586dc395a --- /dev/null +++ b/fs/bcachefs/recovery_passes.c @@ -0,0 +1,203 @@ +// SPDX-License-Identifier: GPL-2.0 + +#include "bcachefs.h" +#include "alloc_background.h" +#include "backpointers.h" +#include "btree_gc.h" +#include "ec.h" +#include "fsck.h" +#include "inode.h" +#include "journal.h" +#include "lru.h" +#include "logged_ops.h" +#include "rebalance.h" +#include "recovery.h" +#include "recovery_passes.h" +#include "snapshot.h" +#include "subvolume.h" +#include "super.h" + +const char * const bch2_recovery_passes[] = { +#define x(_fn, ...) #_fn, + BCH_RECOVERY_PASSES() +#undef x + NULL +}; + +static int bch2_check_allocations(struct bch_fs *c) +{ + return bch2_gc(c, true, c->opts.norecovery); +} + +static int bch2_set_may_go_rw(struct bch_fs *c) +{ + struct journal_keys *keys = &c->journal_keys; + + /* + * After we go RW, the journal keys buffer can't be modified (except for + * setting journal_key->overwritten: it will be accessed by multiple + * threads + */ + move_gap(keys, keys->nr); + + set_bit(BCH_FS_may_go_rw, &c->flags); + + if (keys->nr || c->opts.fsck || !c->sb.clean) + return bch2_fs_read_write_early(c); + return 0; +} + +struct recovery_pass_fn { + int (*fn)(struct bch_fs *); + unsigned when; +}; + +static struct recovery_pass_fn recovery_pass_fns[] = { +#define x(_fn, _id, _when) { .fn = bch2_##_fn, .when = _when }, + BCH_RECOVERY_PASSES() +#undef x +}; + +u64 bch2_recovery_passes_to_stable(u64 v) +{ + static const u8 map[] = { +#define x(n, id, ...) [BCH_RECOVERY_PASS_##n] = BCH_RECOVERY_PASS_STABLE_##n, + BCH_RECOVERY_PASSES() +#undef x + }; + + u64 ret = 0; + for (unsigned i = 0; i < ARRAY_SIZE(map); i++) + if (v & BIT_ULL(i)) + ret |= BIT_ULL(map[i]); + return ret; +} + +u64 bch2_recovery_passes_from_stable(u64 v) +{ + static const u8 map[] = { +#define x(n, id, ...) [BCH_RECOVERY_PASS_STABLE_##n] = BCH_RECOVERY_PASS_##n, + BCH_RECOVERY_PASSES() +#undef x + }; + + u64 ret = 0; + for (unsigned i = 0; i < ARRAY_SIZE(map); i++) + if (v & BIT_ULL(i)) + ret |= BIT_ULL(map[i]); + return ret; +} + +/* + * For when we need to rewind recovery passes and run a pass we skipped: + */ +int bch2_run_explicit_recovery_pass(struct bch_fs *c, + enum bch_recovery_pass pass) +{ + if (c->recovery_passes_explicit & BIT_ULL(pass)) + return 0; + + bch_info(c, "running explicit recovery pass %s (%u), currently at %s (%u)", + bch2_recovery_passes[pass], pass, + bch2_recovery_passes[c->curr_recovery_pass], c->curr_recovery_pass); + + c->recovery_passes_explicit |= BIT_ULL(pass); + + if (c->curr_recovery_pass >= pass) { + c->curr_recovery_pass = pass; + c->recovery_passes_complete &= (1ULL << pass) >> 1; + return -BCH_ERR_restart_recovery; + } else { + return 0; + } +} + +u64 bch2_fsck_recovery_passes(void) +{ + u64 ret = 0; + + for (unsigned i = 0; i < ARRAY_SIZE(recovery_pass_fns); i++) + if (recovery_pass_fns[i].when & PASS_FSCK) + ret |= BIT_ULL(i); + return ret; +} + +static bool should_run_recovery_pass(struct bch_fs *c, enum bch_recovery_pass pass) +{ + struct recovery_pass_fn *p = recovery_pass_fns + pass; + + if (c->opts.norecovery && pass > BCH_RECOVERY_PASS_snapshots_read) + return false; + if (c->recovery_passes_explicit & BIT_ULL(pass)) + return true; + if ((p->when & PASS_FSCK) && c->opts.fsck) + return true; + if ((p->when & PASS_UNCLEAN) && !c->sb.clean) + return true; + if (p->when & PASS_ALWAYS) + return true; + return false; +} + +static int bch2_run_recovery_pass(struct bch_fs *c, enum bch_recovery_pass pass) +{ + struct recovery_pass_fn *p = recovery_pass_fns + pass; + int ret; + + if (!(p->when & PASS_SILENT)) + bch2_print(c, KERN_INFO bch2_log_msg(c, "%s..."), + bch2_recovery_passes[pass]); + ret = p->fn(c); + if (ret) + return ret; + if (!(p->when & PASS_SILENT)) + bch2_print(c, KERN_CONT " done\n"); + + return 0; +} + +int bch2_run_online_recovery_passes(struct bch_fs *c) +{ + int ret = 0; + + for (unsigned i = 0; i < ARRAY_SIZE(recovery_pass_fns); i++) { + struct recovery_pass_fn *p = recovery_pass_fns + i; + + if (!(p->when & PASS_ONLINE)) + continue; + + ret = bch2_run_recovery_pass(c, i); + if (bch2_err_matches(ret, BCH_ERR_restart_recovery)) { + i = c->curr_recovery_pass; + continue; + } + if (ret) + break; + } + + return ret; +} + +int bch2_run_recovery_passes(struct bch_fs *c) +{ + int ret = 0; + + while (c->curr_recovery_pass < ARRAY_SIZE(recovery_pass_fns)) { + if (should_run_recovery_pass(c, c->curr_recovery_pass)) { + unsigned pass = c->curr_recovery_pass; + + ret = bch2_run_recovery_pass(c, c->curr_recovery_pass); + if (bch2_err_matches(ret, BCH_ERR_restart_recovery) || + (ret && c->curr_recovery_pass < pass)) + continue; + if (ret) + break; + + c->recovery_passes_complete |= BIT_ULL(c->curr_recovery_pass); + } + c->curr_recovery_pass++; + c->recovery_pass_done = max(c->recovery_pass_done, c->curr_recovery_pass); + } + + return ret; +} diff --git a/fs/bcachefs/recovery_passes.h b/fs/bcachefs/recovery_passes.h new file mode 100644 index 000000000000..abefa67749eb --- /dev/null +++ b/fs/bcachefs/recovery_passes.h @@ -0,0 +1,16 @@ +#ifndef _BCACHEFS_RECOVERY_PASSES_H +#define _BCACHEFS_RECOVERY_PASSES_H + +extern const char * const bch2_recovery_passes[]; + +u64 bch2_recovery_passes_to_stable(u64 v); +u64 bch2_recovery_passes_from_stable(u64 v); + +u64 bch2_fsck_recovery_passes(void); + +int bch2_run_explicit_recovery_pass(struct bch_fs *, enum bch_recovery_pass); + +int bch2_run_online_recovery_passes(struct bch_fs *); +int bch2_run_recovery_passes(struct bch_fs *); + +#endif /* _BCACHEFS_RECOVERY_PASSES_H */ diff --git a/fs/bcachefs/recovery_passes_types.h b/fs/bcachefs/recovery_passes_types.h new file mode 100644 index 000000000000..09c1ab032a67 --- /dev/null +++ b/fs/bcachefs/recovery_passes_types.h @@ -0,0 +1,69 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +#ifndef _BCACHEFS_RECOVERY_PASSES_TYPES_H +#define _BCACHEFS_RECOVERY_PASSES_TYPES_H + +#define PASS_SILENT BIT(0) +#define PASS_FSCK BIT(1) +#define PASS_UNCLEAN BIT(2) +#define PASS_ALWAYS BIT(3) +#define PASS_ONLINE BIT(4) + +/* + * Passes may be reordered, but the second field is a persistent identifier and + * must never change: + */ +#define BCH_RECOVERY_PASSES() \ + x(check_topology, 4, 0) \ + x(alloc_read, 0, PASS_ALWAYS) \ + x(stripes_read, 1, PASS_ALWAYS) \ + x(initialize_subvolumes, 2, 0) \ + x(snapshots_read, 3, PASS_ALWAYS) \ + x(check_allocations, 5, PASS_FSCK) \ + x(trans_mark_dev_sbs, 6, PASS_ALWAYS|PASS_SILENT) \ + x(fs_journal_alloc, 7, PASS_ALWAYS|PASS_SILENT) \ + x(set_may_go_rw, 8, PASS_ALWAYS|PASS_SILENT) \ + x(journal_replay, 9, PASS_ALWAYS) \ + x(check_alloc_info, 10, PASS_ONLINE|PASS_FSCK) \ + x(check_lrus, 11, PASS_ONLINE|PASS_FSCK) \ + x(check_btree_backpointers, 12, PASS_ONLINE|PASS_FSCK) \ + x(check_backpointers_to_extents, 13, PASS_ONLINE|PASS_FSCK) \ + x(check_extents_to_backpointers, 14, PASS_ONLINE|PASS_FSCK) \ + x(check_alloc_to_lru_refs, 15, PASS_ONLINE|PASS_FSCK) \ + x(fs_freespace_init, 16, PASS_ALWAYS|PASS_SILENT) \ + x(bucket_gens_init, 17, 0) \ + x(check_snapshot_trees, 18, PASS_ONLINE|PASS_FSCK) \ + x(check_snapshots, 19, PASS_ONLINE|PASS_FSCK) \ + x(check_subvols, 20, PASS_ONLINE|PASS_FSCK) \ + x(check_subvol_children, 35, PASS_ONLINE|PASS_FSCK) \ + x(delete_dead_snapshots, 21, PASS_ONLINE|PASS_FSCK) \ + x(fs_upgrade_for_subvolumes, 22, 0) \ + x(resume_logged_ops, 23, PASS_ALWAYS) \ + x(check_inodes, 24, PASS_FSCK) \ + x(check_extents, 25, PASS_FSCK) \ + x(check_indirect_extents, 26, PASS_FSCK) \ + x(check_dirents, 27, PASS_FSCK) \ + x(check_xattrs, 28, PASS_FSCK) \ + x(check_root, 29, PASS_ONLINE|PASS_FSCK) \ + x(check_subvolume_structure, 36, PASS_ONLINE|PASS_FSCK) \ + x(check_directory_structure, 30, PASS_ONLINE|PASS_FSCK) \ + x(check_nlinks, 31, PASS_FSCK) \ + x(delete_dead_inodes, 32, PASS_FSCK|PASS_UNCLEAN) \ + x(fix_reflink_p, 33, 0) \ + x(set_fs_needs_rebalance, 34, 0) \ + +/* We normally enumerate recovery passes in the order we run them: */ +enum bch_recovery_pass { +#define x(n, id, when) BCH_RECOVERY_PASS_##n, + BCH_RECOVERY_PASSES() +#undef x + BCH_RECOVERY_PASS_NR +}; + +/* But we also need stable identifiers that can be used in the superblock */ +enum bch_recovery_pass_stable { +#define x(n, id, when) BCH_RECOVERY_PASS_STABLE_##n = id, + BCH_RECOVERY_PASSES() +#undef x +}; + +#endif /* _BCACHEFS_RECOVERY_PASSES_TYPES_H */ diff --git a/fs/bcachefs/recovery_types.h b/fs/bcachefs/recovery_types.h deleted file mode 100644 index 4959e95e7c74..000000000000 --- a/fs/bcachefs/recovery_types.h +++ /dev/null @@ -1,68 +0,0 @@ -/* SPDX-License-Identifier: GPL-2.0 */ -#ifndef _BCACHEFS_RECOVERY_TYPES_H -#define _BCACHEFS_RECOVERY_TYPES_H - -#define PASS_SILENT BIT(0) -#define PASS_FSCK BIT(1) -#define PASS_UNCLEAN BIT(2) -#define PASS_ALWAYS BIT(3) -#define PASS_ONLINE BIT(4) - -/* - * Passes may be reordered, but the second field is a persistent identifier and - * must never change: - */ -#define BCH_RECOVERY_PASSES() \ - x(check_topology, 4, 0) \ - x(alloc_read, 0, PASS_ALWAYS) \ - x(stripes_read, 1, PASS_ALWAYS) \ - x(initialize_subvolumes, 2, 0) \ - x(snapshots_read, 3, PASS_ALWAYS) \ - x(check_allocations, 5, PASS_FSCK) \ - x(trans_mark_dev_sbs, 6, PASS_ALWAYS|PASS_SILENT) \ - x(fs_journal_alloc, 7, PASS_ALWAYS|PASS_SILENT) \ - x(set_may_go_rw, 8, PASS_ALWAYS|PASS_SILENT) \ - x(journal_replay, 9, PASS_ALWAYS) \ - x(check_alloc_info, 10, PASS_ONLINE|PASS_FSCK) \ - x(check_lrus, 11, PASS_ONLINE|PASS_FSCK) \ - x(check_btree_backpointers, 12, PASS_ONLINE|PASS_FSCK) \ - x(check_backpointers_to_extents, 13, PASS_ONLINE|PASS_FSCK) \ - x(check_extents_to_backpointers, 14, PASS_ONLINE|PASS_FSCK) \ - x(check_alloc_to_lru_refs, 15, PASS_ONLINE|PASS_FSCK) \ - x(fs_freespace_init, 16, PASS_ALWAYS|PASS_SILENT) \ - x(bucket_gens_init, 17, 0) \ - x(check_snapshot_trees, 18, PASS_ONLINE|PASS_FSCK) \ - x(check_snapshots, 19, PASS_ONLINE|PASS_FSCK) \ - x(check_subvols, 20, PASS_ONLINE|PASS_FSCK) \ - x(check_subvol_children, 35, PASS_ONLINE|PASS_FSCK) \ - x(delete_dead_snapshots, 21, PASS_ONLINE|PASS_FSCK) \ - x(fs_upgrade_for_subvolumes, 22, 0) \ - x(resume_logged_ops, 23, PASS_ALWAYS) \ - x(check_inodes, 24, PASS_FSCK) \ - x(check_extents, 25, PASS_FSCK) \ - x(check_indirect_extents, 26, PASS_FSCK) \ - x(check_dirents, 27, PASS_FSCK) \ - x(check_xattrs, 28, PASS_FSCK) \ - x(check_root, 29, PASS_ONLINE|PASS_FSCK) \ - x(check_subvolume_structure, 36, PASS_ONLINE|PASS_FSCK) \ - x(check_directory_structure, 30, PASS_ONLINE|PASS_FSCK) \ - x(check_nlinks, 31, PASS_FSCK) \ - x(delete_dead_inodes, 32, PASS_FSCK|PASS_UNCLEAN) \ - x(fix_reflink_p, 33, 0) \ - x(set_fs_needs_rebalance, 34, 0) \ - -/* We normally enumerate recovery passes in the order we run them: */ -enum bch_recovery_pass { -#define x(n, id, when) BCH_RECOVERY_PASS_##n, - BCH_RECOVERY_PASSES() -#undef x -}; - -/* But we also need stable identifiers that can be used in the superblock */ -enum bch_recovery_pass_stable { -#define x(n, id, when) BCH_RECOVERY_PASS_STABLE_##n = id, - BCH_RECOVERY_PASSES() -#undef x -}; - -#endif /* _BCACHEFS_RECOVERY_TYPES_H */ diff --git a/fs/bcachefs/sb-downgrade.c b/fs/bcachefs/sb-downgrade.c index e4396cb0bacb..d6f81179c3a2 100644 --- a/fs/bcachefs/sb-downgrade.c +++ b/fs/bcachefs/sb-downgrade.c @@ -7,7 +7,7 @@ #include "bcachefs.h" #include "darray.h" -#include "recovery.h" +#include "recovery_passes.h" #include "sb-downgrade.h" #include "sb-errors.h" #include "super-io.h" diff --git a/fs/bcachefs/subvolume.c b/fs/bcachefs/subvolume.c index ce7aed121942..88a79c823276 100644 --- a/fs/bcachefs/subvolume.c +++ b/fs/bcachefs/subvolume.c @@ -595,6 +595,78 @@ err: return ret; } +int bch2_initialize_subvolumes(struct bch_fs *c) +{ + struct bkey_i_snapshot_tree root_tree; + struct bkey_i_snapshot root_snapshot; + struct bkey_i_subvolume root_volume; + int ret; + + bkey_snapshot_tree_init(&root_tree.k_i); + root_tree.k.p.offset = 1; + root_tree.v.master_subvol = cpu_to_le32(1); + root_tree.v.root_snapshot = cpu_to_le32(U32_MAX); + + bkey_snapshot_init(&root_snapshot.k_i); + root_snapshot.k.p.offset = U32_MAX; + root_snapshot.v.flags = 0; + root_snapshot.v.parent = 0; + root_snapshot.v.subvol = cpu_to_le32(BCACHEFS_ROOT_SUBVOL); + root_snapshot.v.tree = cpu_to_le32(1); + SET_BCH_SNAPSHOT_SUBVOL(&root_snapshot.v, true); + + bkey_subvolume_init(&root_volume.k_i); + root_volume.k.p.offset = BCACHEFS_ROOT_SUBVOL; + root_volume.v.flags = 0; + root_volume.v.snapshot = cpu_to_le32(U32_MAX); + root_volume.v.inode = cpu_to_le64(BCACHEFS_ROOT_INO); + + ret = bch2_btree_insert(c, BTREE_ID_snapshot_trees, &root_tree.k_i, NULL, 0) ?: + bch2_btree_insert(c, BTREE_ID_snapshots, &root_snapshot.k_i, NULL, 0) ?: + bch2_btree_insert(c, BTREE_ID_subvolumes, &root_volume.k_i, NULL, 0); + bch_err_fn(c, ret); + return ret; +} + +static int __bch2_fs_upgrade_for_subvolumes(struct btree_trans *trans) +{ + struct btree_iter iter; + struct bkey_s_c k; + struct bch_inode_unpacked inode; + int ret; + + k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_inodes, + SPOS(0, BCACHEFS_ROOT_INO, U32_MAX), 0); + ret = bkey_err(k); + if (ret) + return ret; + + if (!bkey_is_inode(k.k)) { + bch_err(trans->c, "root inode not found"); + ret = -BCH_ERR_ENOENT_inode; + goto err; + } + + ret = bch2_inode_unpack(k, &inode); + BUG_ON(ret); + + inode.bi_subvol = BCACHEFS_ROOT_SUBVOL; + + ret = bch2_inode_write(trans, &iter, &inode); +err: + bch2_trans_iter_exit(trans, &iter); + return ret; +} + +/* set bi_subvol on root inode */ +int bch2_fs_upgrade_for_subvolumes(struct bch_fs *c) +{ + int ret = bch2_trans_do(c, NULL, NULL, BCH_TRANS_COMMIT_lazy_rw, + __bch2_fs_upgrade_for_subvolumes(trans)); + bch_err_fn(c, ret); + return ret; +} + int bch2_fs_subvolumes_init(struct bch_fs *c) { INIT_WORK(&c->snapshot_delete_work, bch2_delete_dead_snapshots_work); diff --git a/fs/bcachefs/subvolume.h b/fs/bcachefs/subvolume.h index 903c05162c06..d2015d549bd2 100644 --- a/fs/bcachefs/subvolume.h +++ b/fs/bcachefs/subvolume.h @@ -37,6 +37,9 @@ void bch2_delete_dead_snapshots_async(struct bch_fs *); int bch2_subvolume_unlink(struct btree_trans *, u32); int bch2_subvolume_create(struct btree_trans *, u64, u32, u32, u32 *, u32 *, bool); +int bch2_initialize_subvolumes(struct bch_fs *); +int bch2_fs_upgrade_for_subvolumes(struct bch_fs *); + int bch2_fs_subvolumes_init(struct bch_fs *); #endif /* _BCACHEFS_SUBVOLUME_H */ diff --git a/fs/bcachefs/super-io.c b/fs/bcachefs/super-io.c index ad28e370b640..e15f8b1f30c2 100644 --- a/fs/bcachefs/super-io.c +++ b/fs/bcachefs/super-io.c @@ -8,7 +8,7 @@ #include "journal.h" #include "journal_sb.h" #include "journal_seq_blacklist.h" -#include "recovery.h" +#include "recovery_passes.h" #include "replicas.h" #include "quota.h" #include "sb-clean.h" -- cgit From e5aa80464155287cc309d18c1c93962357e3e393 Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Sat, 23 Mar 2024 19:30:58 -0400 Subject: bcachefs: Add error messages to logged ops fns Signed-off-by: Kent Overstreet --- fs/bcachefs/io_misc.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'fs') diff --git a/fs/bcachefs/io_misc.c b/fs/bcachefs/io_misc.c index 1baf78594cca..82f9170dab3f 100644 --- a/fs/bcachefs/io_misc.c +++ b/fs/bcachefs/io_misc.c @@ -264,6 +264,7 @@ static int __bch2_resume_logged_op_truncate(struct btree_trans *trans, ret = 0; err: bch2_logged_op_finish(trans, op_k); + bch_err_fn(c, ret); return ret; } @@ -476,6 +477,7 @@ case LOGGED_OP_FINSERT_finish: break; } err: + bch_err_fn(c, ret); bch2_logged_op_finish(trans, op_k); bch2_trans_iter_exit(trans, &iter); return ret; -- cgit From af855a5f5e74cf0ef1166759fca937ce692b4aac Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Sat, 23 Mar 2024 19:31:15 -0400 Subject: bcachefs: Resume logged ops after fsck Finishing logged ops requires the filesystem to be in a reasonably consistent state - and other fsck passes don't require it to have completed, so just run it last. Signed-off-by: Kent Overstreet --- fs/bcachefs/recovery_passes_types.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/bcachefs/recovery_passes_types.h b/fs/bcachefs/recovery_passes_types.h index 09c1ab032a67..f30521285706 100644 --- a/fs/bcachefs/recovery_passes_types.h +++ b/fs/bcachefs/recovery_passes_types.h @@ -37,7 +37,6 @@ x(check_subvol_children, 35, PASS_ONLINE|PASS_FSCK) \ x(delete_dead_snapshots, 21, PASS_ONLINE|PASS_FSCK) \ x(fs_upgrade_for_subvolumes, 22, 0) \ - x(resume_logged_ops, 23, PASS_ALWAYS) \ x(check_inodes, 24, PASS_FSCK) \ x(check_extents, 25, PASS_FSCK) \ x(check_indirect_extents, 26, PASS_FSCK) \ @@ -47,6 +46,7 @@ x(check_subvolume_structure, 36, PASS_ONLINE|PASS_FSCK) \ x(check_directory_structure, 30, PASS_ONLINE|PASS_FSCK) \ x(check_nlinks, 31, PASS_FSCK) \ + x(resume_logged_ops, 23, PASS_ALWAYS) \ x(delete_dead_inodes, 32, PASS_FSCK|PASS_UNCLEAN) \ x(fix_reflink_p, 33, 0) \ x(set_fs_needs_rebalance, 34, 0) \ -- cgit From 4fe0eeeae477328cbd26af1e6f81a94e2080ffa8 Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Thu, 28 Mar 2024 02:36:10 -0400 Subject: bcachefs: Flush journal immediately after replay if we did early repair Signed-off-by: Kent Overstreet --- fs/bcachefs/recovery.c | 20 ++++++++++++++++++++ 1 file changed, 20 insertions(+) (limited to 'fs') diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c index f9e2b011f6d4..dbedb5fd1dde 100644 --- a/fs/bcachefs/recovery.c +++ b/fs/bcachefs/recovery.c @@ -189,6 +189,7 @@ int bch2_journal_replay(struct bch_fs *c) u64 start_seq = c->journal_replay_seq_start; u64 end_seq = c->journal_replay_seq_start; struct btree_trans *trans = bch2_trans_get(c); + bool immediate_flush = false; int ret = 0; if (keys->nr) { @@ -210,6 +211,13 @@ int bch2_journal_replay(struct bch_fs *c) darray_for_each(*keys, k) { cond_resched(); + /* + * k->allocated means the key wasn't read in from the journal, + * rather it was from early repair code + */ + if (k->allocated) + immediate_flush = true; + /* Skip fastpath if we're low on space in the journal */ ret = c->journal.watermark ? -1 : commit_do(trans, NULL, NULL, @@ -269,6 +277,12 @@ int bch2_journal_replay(struct bch_fs *c) bch2_journal_set_replay_done(j); + /* if we did any repair, flush it immediately */ + if (immediate_flush) { + bch2_journal_flush_all_pins(&c->journal); + ret = bch2_journal_meta(&c->journal); + } + if (keys->nr) bch2_journal_log_msg(c, "journal replay finished"); err: @@ -778,6 +792,12 @@ use_clean: clear_bit(BCH_FS_fsck_running, &c->flags); + /* fsync if we fixed errors */ + if (test_bit(BCH_FS_errors_fixed, &c->flags)) { + bch2_journal_flush_all_pins(&c->journal); + bch2_journal_meta(&c->journal); + } + /* If we fixed errors, verify that fs is actually clean now: */ if (IS_ENABLED(CONFIG_BCACHEFS_DEBUG) && test_bit(BCH_FS_errors_fixed, &c->flags) && -- cgit From 0a34c058fca84b10002228a1724e2e613e4dc3cc Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Sat, 30 Mar 2024 18:57:53 -0400 Subject: bcachefs: Ensure bch_sb_field_ext always exists This makes bch_sb_field_ext more consistent with the rest of -o nochanges - we don't want to be varying other codepaths based on -o nochanges, since it's used for testing in dry run mode; also fixes some potential null ptr derefs. Signed-off-by: Kent Overstreet --- fs/bcachefs/recovery.c | 25 ++++++++----------------- fs/bcachefs/super.c | 8 ++++++++ 2 files changed, 16 insertions(+), 17 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c index dbedb5fd1dde..e3b06430d606 100644 --- a/fs/bcachefs/recovery.c +++ b/fs/bcachefs/recovery.c @@ -592,16 +592,9 @@ int bch2_fs_recovery(struct bch_fs *c) if (!c->opts.nochanges) { mutex_lock(&c->sb_lock); + struct bch_sb_field_ext *ext = bch2_sb_field_get(c->disk_sb.sb, ext); bool write_sb = false; - struct bch_sb_field_ext *ext = - bch2_sb_field_get_minsize(&c->disk_sb, ext, sizeof(*ext) / sizeof(u64)); - if (!ext) { - ret = -BCH_ERR_ENOSPC_sb; - mutex_unlock(&c->sb_lock); - goto err; - } - if (BCH_SB_HAS_TOPOLOGY_ERRORS(c->disk_sb.sb)) { ext->recovery_passes_required[0] |= cpu_to_le64(bch2_recovery_passes_to_stable(BIT_ULL(BCH_RECOVERY_PASS_check_topology))); @@ -832,6 +825,7 @@ use_clean: } mutex_lock(&c->sb_lock); + struct bch_sb_field_ext *ext = bch2_sb_field_get(c->disk_sb.sb, ext); bool write_sb = false; if (BCH_SB_VERSION_UPGRADE_COMPLETE(c->disk_sb.sb) != le16_to_cpu(c->disk_sb.sb->version)) { @@ -845,15 +839,12 @@ use_clean: write_sb = true; } - if (!test_bit(BCH_FS_error, &c->flags)) { - struct bch_sb_field_ext *ext = bch2_sb_field_get(c->disk_sb.sb, ext); - if (ext && - (!bch2_is_zero(ext->recovery_passes_required, sizeof(ext->recovery_passes_required)) || - !bch2_is_zero(ext->errors_silent, sizeof(ext->errors_silent)))) { - memset(ext->recovery_passes_required, 0, sizeof(ext->recovery_passes_required)); - memset(ext->errors_silent, 0, sizeof(ext->errors_silent)); - write_sb = true; - } + if (!test_bit(BCH_FS_error, &c->flags) && + (!bch2_is_zero(ext->recovery_passes_required, sizeof(ext->recovery_passes_required)) || + !bch2_is_zero(ext->errors_silent, sizeof(ext->errors_silent)))) { + memset(ext->recovery_passes_required, 0, sizeof(ext->recovery_passes_required)); + memset(ext->errors_silent, 0, sizeof(ext->errors_silent)); + write_sb = true; } if (c->opts.fsck && diff --git a/fs/bcachefs/super.c b/fs/bcachefs/super.c index 1ad6e5cd9476..89ee5d3ec119 100644 --- a/fs/bcachefs/super.c +++ b/fs/bcachefs/super.c @@ -1015,8 +1015,16 @@ int bch2_fs_start(struct bch_fs *c) for_each_online_member(c, ca) bch2_members_v2_get_mut(c->disk_sb.sb, ca->dev_idx)->last_mount = cpu_to_le64(now); + struct bch_sb_field_ext *ext = + bch2_sb_field_get_minsize(&c->disk_sb, ext, sizeof(*ext) / sizeof(u64)); mutex_unlock(&c->sb_lock); + if (!ext) { + bch_err(c, "insufficient space in superblock for sb_field_ext"); + ret = -BCH_ERR_ENOSPC_sb; + goto err; + } + for_each_rw_member(c, ca) bch2_dev_allocator_add(c, ca); bch2_recalc_capacity(c); -- cgit From 060ff30a8596b649a80c19935758000dde7855fe Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Fri, 29 Mar 2024 20:43:39 -0400 Subject: bcachefs: bch2_run_explicit_recovery_pass_persistent() Flag that we need to run a recovery pass and run it - persistenly, so if we crash it'll still get run. Signed-off-by: Kent Overstreet --- fs/bcachefs/recovery_passes.c | 18 ++++++++++++++++++ fs/bcachefs/recovery_passes.h | 1 + 2 files changed, 19 insertions(+) (limited to 'fs') diff --git a/fs/bcachefs/recovery_passes.c b/fs/bcachefs/recovery_passes.c index bab586dc395a..fb22cce10f66 100644 --- a/fs/bcachefs/recovery_passes.c +++ b/fs/bcachefs/recovery_passes.c @@ -16,6 +16,7 @@ #include "snapshot.h" #include "subvolume.h" #include "super.h" +#include "super-io.h" const char * const bch2_recovery_passes[] = { #define x(_fn, ...) #_fn, @@ -112,6 +113,23 @@ int bch2_run_explicit_recovery_pass(struct bch_fs *c, } } +int bch2_run_explicit_recovery_pass_persistent(struct bch_fs *c, + enum bch_recovery_pass pass) +{ + __le64 s = cpu_to_le64(bch2_recovery_passes_to_stable(BIT_ULL(pass))); + + mutex_lock(&c->sb_lock); + struct bch_sb_field_ext *ext = bch2_sb_field_get(c->disk_sb.sb, ext); + + if (!(ext->recovery_passes_required[0] & s)) { + ext->recovery_passes_required[0] |= s; + bch2_write_super(c); + } + mutex_unlock(&c->sb_lock); + + return bch2_run_explicit_recovery_pass(c, pass); +} + u64 bch2_fsck_recovery_passes(void) { u64 ret = 0; diff --git a/fs/bcachefs/recovery_passes.h b/fs/bcachefs/recovery_passes.h index abefa67749eb..99b464e127b8 100644 --- a/fs/bcachefs/recovery_passes.h +++ b/fs/bcachefs/recovery_passes.h @@ -9,6 +9,7 @@ u64 bch2_recovery_passes_from_stable(u64 v); u64 bch2_fsck_recovery_passes(void); int bch2_run_explicit_recovery_pass(struct bch_fs *, enum bch_recovery_pass); +int bch2_run_explicit_recovery_pass_persistent(struct bch_fs *, enum bch_recovery_pass); int bch2_run_online_recovery_passes(struct bch_fs *); int bch2_run_recovery_passes(struct bch_fs *); -- cgit From 13c1e583f9179ad7953dc71ebb2f12e613b9d052 Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Thu, 28 Mar 2024 21:34:14 -0400 Subject: bcachefs: Improve -o norecovery; opts.recovery_pass_limit This adds opts.recovery_pass_limit, and redoes -o norecovery to make use of it; this fixes some issues with -o norecovery so it can be safely used for data recovery. Norecovery means "don't do journal replay"; it's an important data recovery tool when we're getting stuck in journal replay. When using it this way we need to make sure we don't free journal keys after startup, so we continue to overlay them: thus it needs to imply retain_recovery_info, as well as nochanges. recovery_pass_limit is an explicit option for telling recovery to exit after a specific recovery pass; this is a much cleaner way of implementing -o norecovery, as well as being a useful debug feature in its own right. Signed-off-by: Kent Overstreet --- fs/bcachefs/opts.c | 4 ++++ fs/bcachefs/opts.h | 7 ++++++- fs/bcachefs/recovery.c | 10 ++++------ fs/bcachefs/recovery_passes.c | 12 ++++++++---- fs/bcachefs/snapshot.c | 2 +- fs/bcachefs/super.c | 5 +++-- 6 files changed, 26 insertions(+), 14 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/opts.c b/fs/bcachefs/opts.c index 08ea0cfc4aef..e1800c4119b5 100644 --- a/fs/bcachefs/opts.c +++ b/fs/bcachefs/opts.c @@ -7,6 +7,7 @@ #include "disk_groups.h" #include "error.h" #include "opts.h" +#include "recovery_passes.h" #include "super-io.h" #include "util.h" @@ -205,6 +206,9 @@ const struct bch_option bch2_opt_table[] = { #define OPT_STR(_choices) .type = BCH_OPT_STR, \ .min = 0, .max = ARRAY_SIZE(_choices), \ .choices = _choices +#define OPT_STR_NOLIMIT(_choices) .type = BCH_OPT_STR, \ + .min = 0, .max = U64_MAX, \ + .choices = _choices #define OPT_FN(_fn) .type = BCH_OPT_FN, .fn = _fn #define x(_name, _bits, _flags, _type, _sb_opt, _default, _hint, _help) \ diff --git a/fs/bcachefs/opts.h b/fs/bcachefs/opts.h index 136083c11f3a..084247c13bd8 100644 --- a/fs/bcachefs/opts.h +++ b/fs/bcachefs/opts.h @@ -362,7 +362,12 @@ enum fsck_err_opts { OPT_FS|OPT_MOUNT, \ OPT_BOOL(), \ BCH2_NO_SB_OPT, false, \ - NULL, "Don't replay the journal") \ + NULL, "Exit recovery immediately prior to journal replay")\ + x(recovery_pass_last, u8, \ + OPT_FS|OPT_MOUNT, \ + OPT_STR_NOLIMIT(bch2_recovery_passes), \ + BCH2_NO_SB_OPT, 0, \ + NULL, "Exit recovery after specified pass") \ x(keep_journal, u8, \ 0, \ OPT_BOOL(), \ diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c index e3b06430d606..e8b434009293 100644 --- a/fs/bcachefs/recovery.c +++ b/fs/bcachefs/recovery.c @@ -269,7 +269,8 @@ int bch2_journal_replay(struct bch_fs *c) bch2_trans_put(trans); trans = NULL; - if (!c->opts.keep_journal) + if (!c->opts.keep_journal && + c->recovery_pass_done >= BCH_RECOVERY_PASS_journal_replay) bch2_journal_keys_put_initial(c); replay_now_at(j, j->replay_journal_seq_end); @@ -584,11 +585,8 @@ int bch2_fs_recovery(struct bch_fs *c) goto err; } - if (c->opts.fsck && c->opts.norecovery) { - bch_err(c, "cannot select both norecovery and fsck"); - ret = -EINVAL; - goto err; - } + if (c->opts.norecovery) + c->opts.recovery_pass_last = BCH_RECOVERY_PASS_journal_replay - 1; if (!c->opts.nochanges) { mutex_lock(&c->sb_lock); diff --git a/fs/bcachefs/recovery_passes.c b/fs/bcachefs/recovery_passes.c index fb22cce10f66..0089d9456b10 100644 --- a/fs/bcachefs/recovery_passes.c +++ b/fs/bcachefs/recovery_passes.c @@ -27,7 +27,7 @@ const char * const bch2_recovery_passes[] = { static int bch2_check_allocations(struct bch_fs *c) { - return bch2_gc(c, true, c->opts.norecovery); + return bch2_gc(c, true, false); } static int bch2_set_may_go_rw(struct bch_fs *c) @@ -144,8 +144,6 @@ static bool should_run_recovery_pass(struct bch_fs *c, enum bch_recovery_pass pa { struct recovery_pass_fn *p = recovery_pass_fns + pass; - if (c->opts.norecovery && pass > BCH_RECOVERY_PASS_snapshots_read) - return false; if (c->recovery_passes_explicit & BIT_ULL(pass)) return true; if ((p->when & PASS_FSCK) && c->opts.fsck) @@ -201,6 +199,10 @@ int bch2_run_recovery_passes(struct bch_fs *c) int ret = 0; while (c->curr_recovery_pass < ARRAY_SIZE(recovery_pass_fns)) { + if (c->opts.recovery_pass_last && + c->curr_recovery_pass > c->opts.recovery_pass_last) + break; + if (should_run_recovery_pass(c, c->curr_recovery_pass)) { unsigned pass = c->curr_recovery_pass; @@ -213,8 +215,10 @@ int bch2_run_recovery_passes(struct bch_fs *c) c->recovery_passes_complete |= BIT_ULL(c->curr_recovery_pass); } - c->curr_recovery_pass++; + c->recovery_pass_done = max(c->recovery_pass_done, c->curr_recovery_pass); + + c->curr_recovery_pass++; } return ret; diff --git a/fs/bcachefs/snapshot.c b/fs/bcachefs/snapshot.c index 4e074136c490..4577ee7939a2 100644 --- a/fs/bcachefs/snapshot.c +++ b/fs/bcachefs/snapshot.c @@ -131,7 +131,7 @@ bool __bch2_snapshot_is_ancestor(struct bch_fs *c, u32 id, u32 ancestor) rcu_read_lock(); struct snapshot_table *t = rcu_dereference(c->snapshots); - if (unlikely(c->recovery_pass_done <= BCH_RECOVERY_PASS_check_snapshots)) { + if (unlikely(c->recovery_pass_done < BCH_RECOVERY_PASS_check_snapshots)) { ret = __bch2_snapshot_is_ancestor_early(t, id, ancestor); goto out; } diff --git a/fs/bcachefs/super.c b/fs/bcachefs/super.c index 89ee5d3ec119..bc026a77eb99 100644 --- a/fs/bcachefs/super.c +++ b/fs/bcachefs/super.c @@ -365,7 +365,7 @@ void bch2_fs_read_only(struct bch_fs *c) !test_bit(BCH_FS_emergency_ro, &c->flags) && test_bit(BCH_FS_started, &c->flags) && test_bit(BCH_FS_clean_shutdown, &c->flags) && - !c->opts.norecovery) { + c->recovery_pass_done >= BCH_RECOVERY_PASS_journal_replay) { BUG_ON(c->journal.last_empty_seq != journal_cur_seq(&c->journal)); BUG_ON(atomic_read(&c->btree_cache.dirty)); BUG_ON(atomic_long_read(&c->btree_key_cache.nr_dirty)); @@ -510,7 +510,8 @@ err: int bch2_fs_read_write(struct bch_fs *c) { - if (c->opts.norecovery) + if (c->opts.recovery_pass_last && + c->opts.recovery_pass_last < BCH_RECOVERY_PASS_journal_replay) return -BCH_ERR_erofs_norecovery; if (c->opts.nochanges) -- cgit From cecfed9b446da5fba9d73e6448c9f0d1ff5d95ff Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Sun, 31 Mar 2024 22:34:45 -0400 Subject: bcachefs: Logged op errors should be ignored If something is wrong with a logged op, we just want to delete it - there's nothing to repair. Signed-off-by: Kent Overstreet --- fs/bcachefs/logged_ops.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/logged_ops.c b/fs/bcachefs/logged_ops.c index 9fac838d123e..b82f8209041f 100644 --- a/fs/bcachefs/logged_ops.c +++ b/fs/bcachefs/logged_ops.c @@ -37,7 +37,6 @@ static int resume_logged_op(struct btree_trans *trans, struct btree_iter *iter, const struct bch_logged_op_fn *fn = logged_op_fn(k.k->type); struct bkey_buf sk; u32 restart_count = trans->restart_count; - int ret; if (!fn) return 0; @@ -45,11 +44,11 @@ static int resume_logged_op(struct btree_trans *trans, struct btree_iter *iter, bch2_bkey_buf_init(&sk); bch2_bkey_buf_reassemble(&sk, c, k); - ret = drop_locks_do(trans, (bch2_fs_lazy_rw(c), 0)) ?: - fn->resume(trans, sk.k) ?: trans_was_restarted(trans, restart_count); + fn->resume(trans, sk.k); bch2_bkey_buf_exit(&sk, c); - return ret; + + return trans_was_restarted(trans, restart_count); } int bch2_resume_logged_ops(struct bch_fs *c) -- cgit From 8ce1db8091b23f5d2a0dd1dabe8007954114cb68 Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Mon, 1 Apr 2024 00:00:32 -0400 Subject: bcachefs: Fix remove_dirent() We were missing an iter_traverse(). Signed-off-by: Kent Overstreet --- fs/bcachefs/fsck.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/fsck.c b/fs/bcachefs/fsck.c index aca6fae409cd..0a47b0a473d8 100644 --- a/fs/bcachefs/fsck.c +++ b/fs/bcachefs/fsck.c @@ -158,9 +158,10 @@ static int __remove_dirent(struct btree_trans *trans, struct bpos pos) bch2_trans_iter_init(trans, &iter, BTREE_ID_dirents, pos, BTREE_ITER_INTENT); - ret = bch2_hash_delete_at(trans, bch2_dirent_hash_desc, - &dir_hash_info, &iter, - BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE); + ret = bch2_btree_iter_traverse(&iter) ?: + bch2_hash_delete_at(trans, bch2_dirent_hash_desc, + &dir_hash_info, &iter, + BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE); bch2_trans_iter_exit(trans, &iter); err: bch_err_fn(c, ret); -- cgit From eab3a3ce2dea1a4013a3a553722b85f55a76ac2d Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Sat, 30 Mar 2024 01:00:50 -0400 Subject: bcachefs: Fix overlapping extent repair overlapping extent repair was colliding with extent past end of inode checks - don't update "extent ends at" until we know we have an extent. Signed-off-by: Kent Overstreet --- fs/bcachefs/fsck.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/fsck.c b/fs/bcachefs/fsck.c index 0a47b0a473d8..cbb8b43e419f 100644 --- a/fs/bcachefs/fsck.c +++ b/fs/bcachefs/fsck.c @@ -1372,10 +1372,6 @@ static int check_overlapping_extents(struct btree_trans *trans, goto err; } - ret = extent_ends_at(c, extent_ends, seen, k); - if (ret) - goto err; - extent_ends->last_pos = k.k->p; err: return ret; @@ -1505,6 +1501,12 @@ static int check_extent(struct btree_trans *trans, struct btree_iter *iter, i->seen_this_pos = true; } + + if (k.k->type != KEY_TYPE_whiteout) { + ret = extent_ends_at(c, extent_ends, s, k); + if (ret) + goto err; + } out: err: fsck_err: -- cgit From b3c7fd35c03c17a950737fb56a06b730a7962d28 Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Sat, 30 Mar 2024 15:59:57 -0400 Subject: bcachefs: On emergency shutdown, print out current journal sequence number Signed-off-by: Kent Overstreet --- fs/bcachefs/error.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/bcachefs/error.c b/fs/bcachefs/error.c index f942a3947496..82a6656c941c 100644 --- a/fs/bcachefs/error.c +++ b/fs/bcachefs/error.c @@ -1,6 +1,7 @@ // SPDX-License-Identifier: GPL-2.0 #include "bcachefs.h" #include "error.h" +#include "journal.h" #include "recovery_passes.h" #include "super.h" #include "thread_with_file.h" @@ -16,7 +17,8 @@ bool bch2_inconsistent_error(struct bch_fs *c) return false; case BCH_ON_ERROR_ro: if (bch2_fs_emergency_read_only(c)) - bch_err(c, "inconsistency detected - emergency read only"); + bch_err(c, "inconsistency detected - emergency read only at journal seq %llu", + journal_cur_seq(&c->journal)); return true; case BCH_ON_ERROR_panic: panic(bch2_fmt(c, "panic after error")); -- cgit From ba947ecd39ea0e6a6f6f1101f99611fc30943bcb Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Mon, 1 Apr 2024 19:16:19 -0400 Subject: bcachefs: Fix btree node reserve Sign error when checking the watermark - oops. Signed-off-by: Kent Overstreet --- fs/bcachefs/btree_update_interior.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/bcachefs/btree_update_interior.c b/fs/bcachefs/btree_update_interior.c index 983bb27298cc..29aee215384a 100644 --- a/fs/bcachefs/btree_update_interior.c +++ b/fs/bcachefs/btree_update_interior.c @@ -303,7 +303,7 @@ static struct btree *__bch2_btree_node_alloc(struct btree_trans *trans, struct open_buckets obs = { .nr = 0 }; struct bch_devs_list devs_have = (struct bch_devs_list) { 0 }; enum bch_watermark watermark = flags & BCH_WATERMARK_MASK; - unsigned nr_reserve = watermark > BCH_WATERMARK_reclaim + unsigned nr_reserve = watermark < BCH_WATERMARK_reclaim ? BTREE_NODE_RESERVE : 0; int ret; -- cgit From e2a316b3cc45a1198f3feb18707403bb7f0cbc15 Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Mon, 1 Apr 2024 19:20:36 -0400 Subject: bcachefs: BCH_WATERMARK_interior_updates This adds a new watermark, higher priority than BCH_WATERMARK_reclaim, for interior btree updates. We've seen a deadlock where journal replay triggers a ton of btree node merges, and these use up all available open buckets and then interior updates get stuck. One cause of this is that we're currently lacking btree node merging on write buffer btrees - that needs to be fixed as well. Signed-off-by: Kent Overstreet --- fs/bcachefs/alloc_foreground.c | 4 +++- fs/bcachefs/alloc_types.h | 3 ++- fs/bcachefs/btree_io.c | 2 +- fs/bcachefs/btree_trans_commit.c | 3 ++- fs/bcachefs/btree_update_interior.c | 6 +++--- fs/bcachefs/buckets.h | 1 + 6 files changed, 12 insertions(+), 7 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/alloc_foreground.c b/fs/bcachefs/alloc_foreground.c index 214b15c84d1f..a1fc30adf912 100644 --- a/fs/bcachefs/alloc_foreground.c +++ b/fs/bcachefs/alloc_foreground.c @@ -188,8 +188,10 @@ long bch2_bucket_alloc_new_fs(struct bch_dev *ca) static inline unsigned open_buckets_reserved(enum bch_watermark watermark) { switch (watermark) { - case BCH_WATERMARK_reclaim: + case BCH_WATERMARK_interior_updates: return 0; + case BCH_WATERMARK_reclaim: + return OPEN_BUCKETS_COUNT / 6; case BCH_WATERMARK_btree: case BCH_WATERMARK_btree_copygc: return OPEN_BUCKETS_COUNT / 4; diff --git a/fs/bcachefs/alloc_types.h b/fs/bcachefs/alloc_types.h index b91b7a461056..c2226e947c41 100644 --- a/fs/bcachefs/alloc_types.h +++ b/fs/bcachefs/alloc_types.h @@ -22,7 +22,8 @@ struct bucket_alloc_state { x(copygc) \ x(btree) \ x(btree_copygc) \ - x(reclaim) + x(reclaim) \ + x(interior_updates) enum bch_watermark { #define x(name) BCH_WATERMARK_##name, diff --git a/fs/bcachefs/btree_io.c b/fs/bcachefs/btree_io.c index 9c71e6fb9c41..f3f27bb85a5b 100644 --- a/fs/bcachefs/btree_io.c +++ b/fs/bcachefs/btree_io.c @@ -1861,7 +1861,7 @@ static void btree_node_write_work(struct work_struct *work) } else { ret = bch2_trans_do(c, NULL, NULL, 0, bch2_btree_node_update_key_get_iter(trans, b, &wbio->key, - BCH_WATERMARK_reclaim| + BCH_WATERMARK_interior_updates| BCH_TRANS_COMMIT_journal_reclaim| BCH_TRANS_COMMIT_no_enospc| BCH_TRANS_COMMIT_no_check_rw, diff --git a/fs/bcachefs/btree_trans_commit.c b/fs/bcachefs/btree_trans_commit.c index 96669fede7d3..aa9da4970740 100644 --- a/fs/bcachefs/btree_trans_commit.c +++ b/fs/bcachefs/btree_trans_commit.c @@ -887,6 +887,7 @@ int bch2_trans_commit_error(struct btree_trans *trans, unsigned flags, int ret, unsigned long trace_ip) { struct bch_fs *c = trans->c; + enum bch_watermark watermark = flags & BCH_WATERMARK_MASK; switch (ret) { case -BCH_ERR_btree_insert_btree_node_full: @@ -905,7 +906,7 @@ int bch2_trans_commit_error(struct btree_trans *trans, unsigned flags, * flag */ if ((flags & BCH_TRANS_COMMIT_journal_reclaim) && - (flags & BCH_WATERMARK_MASK) != BCH_WATERMARK_reclaim) { + watermark < BCH_WATERMARK_reclaim) { ret = -BCH_ERR_journal_reclaim_would_deadlock; break; } diff --git a/fs/bcachefs/btree_update_interior.c b/fs/bcachefs/btree_update_interior.c index 29aee215384a..9fd2dd0f4682 100644 --- a/fs/bcachefs/btree_update_interior.c +++ b/fs/bcachefs/btree_update_interior.c @@ -687,7 +687,7 @@ static void btree_update_nodes_written(struct btree_update *as) * which may require allocations as well. */ ret = commit_do(trans, &as->disk_res, &journal_seq, - BCH_WATERMARK_reclaim| + BCH_WATERMARK_interior_updates| BCH_TRANS_COMMIT_no_enospc| BCH_TRANS_COMMIT_no_check_rw| BCH_TRANS_COMMIT_journal_reclaim, @@ -1121,7 +1121,7 @@ bch2_btree_update_start(struct btree_trans *trans, struct btree_path *path, unsigned journal_flags = watermark|JOURNAL_RES_GET_CHECK; if ((flags & BCH_TRANS_COMMIT_journal_reclaim) && - watermark != BCH_WATERMARK_reclaim) + watermark < BCH_WATERMARK_reclaim) journal_flags |= JOURNAL_RES_GET_NONBLOCK; ret = drop_locks_do(trans, @@ -1217,7 +1217,7 @@ bch2_btree_update_start(struct btree_trans *trans, struct btree_path *path, */ if (bch2_err_matches(ret, ENOSPC) && (flags & BCH_TRANS_COMMIT_journal_reclaim) && - watermark != BCH_WATERMARK_reclaim) { + watermark < BCH_WATERMARK_reclaim) { ret = -BCH_ERR_journal_reclaim_would_deadlock; goto err; } diff --git a/fs/bcachefs/buckets.h b/fs/bcachefs/buckets.h index 6387e039f789..00aaf4bb5139 100644 --- a/fs/bcachefs/buckets.h +++ b/fs/bcachefs/buckets.h @@ -226,6 +226,7 @@ static inline u64 bch2_dev_buckets_reserved(struct bch_dev *ca, enum bch_waterma fallthrough; case BCH_WATERMARK_btree_copygc: case BCH_WATERMARK_reclaim: + case BCH_WATERMARK_interior_updates: break; } -- cgit From c42cd606e4f004e9ba36a05b9adb9e4eead5834a Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Tue, 2 Apr 2024 01:03:58 -0400 Subject: bcachefs: fix nocow lock deadlock Signed-off-by: Kent Overstreet --- fs/bcachefs/data_update.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/data_update.c b/fs/bcachefs/data_update.c index b564404dad71..34731ee0217f 100644 --- a/fs/bcachefs/data_update.c +++ b/fs/bcachefs/data_update.c @@ -580,8 +580,7 @@ int bch2_data_update_init(struct btree_trans *trans, move_ctxt_wait_event(ctxt, (locked = bch2_bucket_nocow_trylock(&c->nocow_locks, PTR_BUCKET_POS(c, &p.ptr), 0)) || - (!atomic_read(&ctxt->read_sectors) && - !atomic_read(&ctxt->write_sectors))); + list_empty(&ctxt->ios)); if (!locked) bch2_bucket_nocow_lock(&c->nocow_locks, -- cgit From 97ca7c1f93bbac6982717a7055cd727813c45e61 Mon Sep 17 00:00:00 2001 From: Guenter Roeck Date: Sun, 25 Feb 2024 08:29:25 -0800 Subject: mean_and_variance: Drop always failing tests mean_and_variance_test_2 and mean_and_variance_test_4 always fail. The input parameters to those tests are identical to the input parameters to tests 1 and 3, yet the expected result for tests 2 and 4 is different for the mean and stddev tests. That will always fail. Expected mean_and_variance_get_mean(mv) == mean[i], but mean_and_variance_get_mean(mv) == 22 (0x16) mean[i] == 10 (0xa) Drop the bad tests. Fixes: 65bc41090720 ("mean and variance: More tests") Closes: https://lore.kernel.org/lkml/065b94eb-6a24-4248-b7d7-d3212efb4787@roeck-us.net/ Cc: Kent Overstreet Signed-off-by: Guenter Roeck Signed-off-by: Kent Overstreet --- fs/bcachefs/mean_and_variance_test.c | 28 +--------------------------- 1 file changed, 1 insertion(+), 27 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/mean_and_variance_test.c b/fs/bcachefs/mean_and_variance_test.c index db63b3f3b338..4c298e74723d 100644 --- a/fs/bcachefs/mean_and_variance_test.c +++ b/fs/bcachefs/mean_and_variance_test.c @@ -136,20 +136,8 @@ static void mean_and_variance_test_1(struct kunit *test) d, mean, stddev, weighted_mean, weighted_stddev); } -static void mean_and_variance_test_2(struct kunit *test) -{ - s64 d[] = { 100, 10, 10, 10, 10, 10, 10 }; - s64 mean[] = { 10, 10, 10, 10, 10, 10, 10 }; - s64 stddev[] = { 9, 9, 9, 9, 9, 9, 9 }; - s64 weighted_mean[] = { 32, 27, 22, 19, 17, 15, 14 }; - s64 weighted_stddev[] = { 38, 35, 31, 27, 24, 21, 18 }; - - do_mean_and_variance_test(test, 10, 6, ARRAY_SIZE(d), 2, - d, mean, stddev, weighted_mean, weighted_stddev); -} - /* Test behaviour where we switch from one steady state to another: */ -static void mean_and_variance_test_3(struct kunit *test) +static void mean_and_variance_test_2(struct kunit *test) { s64 d[] = { 100, 100, 100, 100, 100 }; s64 mean[] = { 22, 32, 40, 46, 50 }; @@ -161,18 +149,6 @@ static void mean_and_variance_test_3(struct kunit *test) d, mean, stddev, weighted_mean, weighted_stddev); } -static void mean_and_variance_test_4(struct kunit *test) -{ - s64 d[] = { 100, 100, 100, 100, 100 }; - s64 mean[] = { 10, 11, 12, 13, 14 }; - s64 stddev[] = { 9, 13, 15, 17, 19 }; - s64 weighted_mean[] = { 32, 49, 61, 71, 78 }; - s64 weighted_stddev[] = { 38, 44, 44, 41, 38 }; - - do_mean_and_variance_test(test, 10, 6, ARRAY_SIZE(d), 2, - d, mean, stddev, weighted_mean, weighted_stddev); -} - static void mean_and_variance_fast_divpow2(struct kunit *test) { s64 i; @@ -230,8 +206,6 @@ static struct kunit_case mean_and_variance_test_cases[] = { KUNIT_CASE(mean_and_variance_weighted_advanced_test), KUNIT_CASE(mean_and_variance_test_1), KUNIT_CASE(mean_and_variance_test_2), - KUNIT_CASE(mean_and_variance_test_3), - KUNIT_CASE(mean_and_variance_test_4), {} }; -- cgit From e0319af2b6cdfa7c39edf73dcb813b7ff1261fa5 Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Tue, 2 Apr 2024 16:42:27 -0400 Subject: bcachefs: Improve bch2_btree_update_to_text() Print out the mode as a string, and also print out the btree and watermark. Signed-off-by: Kent Overstreet --- fs/bcachefs/btree_update_interior.c | 41 +++++++++++++++++++++++++------------ fs/bcachefs/btree_update_interior.h | 24 ++++++++++++++-------- 2 files changed, 43 insertions(+), 22 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/btree_update_interior.c b/fs/bcachefs/btree_update_interior.c index 9fd2dd0f4682..2b46e6ad248a 100644 --- a/fs/bcachefs/btree_update_interior.c +++ b/fs/bcachefs/btree_update_interior.c @@ -26,6 +26,13 @@ #include +const char * const bch2_btree_update_modes[] = { +#define x(t) #t, + BCH_WATERMARKS() +#undef x + NULL +}; + static int bch2_btree_insert_node(struct btree_update *, struct btree_trans *, btree_path_idx_t, struct btree *, struct keylist *); static void bch2_btree_update_add_new_node(struct btree_update *, struct btree *); @@ -846,11 +853,11 @@ static void btree_update_updated_node(struct btree_update *as, struct btree *b) mutex_lock(&c->btree_interior_update_lock); list_add_tail(&as->unwritten_list, &c->btree_interior_updates_unwritten); - BUG_ON(as->mode != BTREE_INTERIOR_NO_UPDATE); + BUG_ON(as->mode != BTREE_UPDATE_none); BUG_ON(!btree_node_dirty(b)); BUG_ON(!b->c.level); - as->mode = BTREE_INTERIOR_UPDATING_NODE; + as->mode = BTREE_UPDATE_node; as->b = b; set_btree_node_write_blocked(b); @@ -873,7 +880,7 @@ static void btree_update_reparent(struct btree_update *as, lockdep_assert_held(&c->btree_interior_update_lock); child->b = NULL; - child->mode = BTREE_INTERIOR_UPDATING_AS; + child->mode = BTREE_UPDATE_update; bch2_journal_pin_copy(&c->journal, &as->journal, &child->journal, bch2_update_reparent_journal_pin_flush); @@ -884,7 +891,7 @@ static void btree_update_updated_root(struct btree_update *as, struct btree *b) struct bkey_i *insert = &b->key; struct bch_fs *c = as->c; - BUG_ON(as->mode != BTREE_INTERIOR_NO_UPDATE); + BUG_ON(as->mode != BTREE_UPDATE_none); BUG_ON(as->journal_u64s + jset_u64s(insert->k.u64s) > ARRAY_SIZE(as->journal_entries)); @@ -898,7 +905,7 @@ static void btree_update_updated_root(struct btree_update *as, struct btree *b) mutex_lock(&c->btree_interior_update_lock); list_add_tail(&as->unwritten_list, &c->btree_interior_updates_unwritten); - as->mode = BTREE_INTERIOR_UPDATING_ROOT; + as->mode = BTREE_UPDATE_root; mutex_unlock(&c->btree_interior_update_lock); } @@ -1076,7 +1083,7 @@ static void bch2_btree_update_done(struct btree_update *as, struct btree_trans * struct bch_fs *c = as->c; u64 start_time = as->start_time; - BUG_ON(as->mode == BTREE_INTERIOR_NO_UPDATE); + BUG_ON(as->mode == BTREE_UPDATE_none); if (as->took_gc_lock) up_read(&as->c->gc_lock); @@ -1172,7 +1179,8 @@ bch2_btree_update_start(struct btree_trans *trans, struct btree_path *path, as->c = c; as->start_time = start_time; as->ip_started = _RET_IP_; - as->mode = BTREE_INTERIOR_NO_UPDATE; + as->mode = BTREE_UPDATE_none; + as->watermark = watermark; as->took_gc_lock = true; as->btree_id = path->btree_id; as->update_level = update_level; @@ -2509,18 +2517,25 @@ void bch2_btree_root_alloc(struct bch_fs *c, enum btree_id id) bch2_trans_run(c, __bch2_btree_root_alloc(trans, id)); } +static void bch2_btree_update_to_text(struct printbuf *out, struct btree_update *as) +{ + prt_printf(out, "%ps: btree=%s watermark=%s mode=%s nodes_written=%u cl.remaining=%u journal_seq=%llu\n", + (void *) as->ip_started, + bch2_btree_id_str(as->btree_id), + bch2_watermarks[as->watermark], + bch2_btree_update_modes[as->mode], + as->nodes_written, + closure_nr_remaining(&as->cl), + as->journal.seq); +} + void bch2_btree_updates_to_text(struct printbuf *out, struct bch_fs *c) { struct btree_update *as; mutex_lock(&c->btree_interior_update_lock); list_for_each_entry(as, &c->btree_interior_update_list, list) - prt_printf(out, "%ps: mode=%u nodes_written=%u cl.remaining=%u journal_seq=%llu\n", - (void *) as->ip_started, - as->mode, - as->nodes_written, - closure_nr_remaining(&as->cl), - as->journal.seq); + bch2_btree_update_to_text(out, as); mutex_unlock(&c->btree_interior_update_lock); } diff --git a/fs/bcachefs/btree_update_interior.h b/fs/bcachefs/btree_update_interior.h index 7ac3e17b84fd..1d41d3f4b107 100644 --- a/fs/bcachefs/btree_update_interior.h +++ b/fs/bcachefs/btree_update_interior.h @@ -12,6 +12,18 @@ int bch2_btree_node_check_topology(struct btree_trans *, struct btree *); +#define BTREE_UPDATE_MODES() \ + x(none) \ + x(node) \ + x(root) \ + x(update) + +enum btree_update_mode { +#define x(n) BTREE_UPDATE_##n, + BTREE_UPDATE_MODES() +#undef x +}; + /* * Tracks an in progress split/rewrite of a btree node and the update to the * parent node: @@ -39,14 +51,8 @@ struct btree_update { struct list_head list; struct list_head unwritten_list; - /* What kind of update are we doing? */ - enum { - BTREE_INTERIOR_NO_UPDATE, - BTREE_INTERIOR_UPDATING_NODE, - BTREE_INTERIOR_UPDATING_ROOT, - BTREE_INTERIOR_UPDATING_AS, - } mode; - + enum btree_update_mode mode; + enum bch_watermark watermark; unsigned nodes_written:1; unsigned took_gc_lock:1; @@ -56,7 +62,7 @@ struct btree_update { struct disk_reservation disk_res; /* - * BTREE_INTERIOR_UPDATING_NODE: + * BTREE_UPDATE_node: * The update that made the new nodes visible was a regular update to an * existing interior node - @b. We can't write out the update to @b * until the new nodes we created are finished writing, so we block @b -- cgit From 7ee88737ab802ac832f978d6e6258571fe08d870 Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Tue, 2 Apr 2024 18:30:14 -0400 Subject: bcachefs: Check for bad needs_discard before doing discard In the discard worker, we were failing to validate the bucket state - meaning a corrupt needs_discard btree could cause us to discard a bucket that we shouldn't. If check_alloc_info hasn't run yet we just want to bail out, otherwise it's a filesystem inconsistent error. Signed-off-by: Kent Overstreet --- fs/bcachefs/alloc_background.c | 47 +++++++++++++++++++++++------------------- 1 file changed, 26 insertions(+), 21 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/alloc_background.c b/fs/bcachefs/alloc_background.c index 893e38f9db80..4ff56fa4d539 100644 --- a/fs/bcachefs/alloc_background.c +++ b/fs/bcachefs/alloc_background.c @@ -1713,34 +1713,37 @@ static int bch2_discard_one_bucket(struct btree_trans *trans, if (ret) goto out; - if (BCH_ALLOC_V4_NEED_INC_GEN(&a->v)) { - a->v.gen++; - SET_BCH_ALLOC_V4_NEED_INC_GEN(&a->v, false); - goto write; - } - - if (a->v.journal_seq > c->journal.flushed_seq_ondisk) { - if (c->curr_recovery_pass > BCH_RECOVERY_PASS_check_alloc_info) { - bch2_trans_inconsistent(trans, - "clearing need_discard but journal_seq %llu > flushed_seq %llu\n" - "%s", - a->v.journal_seq, - c->journal.flushed_seq_ondisk, - (bch2_bkey_val_to_text(&buf, c, k), buf.buf)); + if (a->v.dirty_sectors) { + if (bch2_trans_inconsistent_on(c->curr_recovery_pass > BCH_RECOVERY_PASS_check_alloc_info, + trans, "attempting to discard bucket with dirty data\n%s", + (bch2_bkey_val_to_text(&buf, c, k), buf.buf))) ret = -EIO; - } goto out; } if (a->v.data_type != BCH_DATA_need_discard) { - if (c->curr_recovery_pass > BCH_RECOVERY_PASS_check_alloc_info) { - bch2_trans_inconsistent(trans, - "bucket incorrectly set in need_discard btree\n" - "%s", - (bch2_bkey_val_to_text(&buf, c, k), buf.buf)); - ret = -EIO; + if (data_type_is_empty(a->v.data_type) && + BCH_ALLOC_V4_NEED_INC_GEN(&a->v)) { + a->v.gen++; + SET_BCH_ALLOC_V4_NEED_INC_GEN(&a->v, false); + goto write; } + if (bch2_trans_inconsistent_on(c->curr_recovery_pass > BCH_RECOVERY_PASS_check_alloc_info, + trans, "bucket incorrectly set in need_discard btree\n" + "%s", + (bch2_bkey_val_to_text(&buf, c, k), buf.buf))) + ret = -EIO; + goto out; + } + + if (a->v.journal_seq > c->journal.flushed_seq_ondisk) { + if (bch2_trans_inconsistent_on(c->curr_recovery_pass > BCH_RECOVERY_PASS_check_alloc_info, + trans, "clearing need_discard but journal_seq %llu > flushed_seq %llu\n%s", + a->v.journal_seq, + c->journal.flushed_seq_ondisk, + (bch2_bkey_val_to_text(&buf, c, k), buf.buf))) + ret = -EIO; goto out; } @@ -1835,6 +1838,7 @@ static int bch2_clear_bucket_needs_discard(struct btree_trans *trans, struct bpo if (ret) goto err; + BUG_ON(a->v.dirty_sectors); SET_BCH_ALLOC_V4_NEED_DISCARD(&a->v, false); a->v.data_type = alloc_data_type(a->v, a->v.data_type); @@ -1942,6 +1946,7 @@ static int invalidate_one_bucket(struct btree_trans *trans, goto out; BUG_ON(a->v.data_type != BCH_DATA_cached); + BUG_ON(a->v.dirty_sectors); if (!a->v.cached_sectors) bch_err(c, "invalidating empty bucket, confused"); -- cgit From fa14b50460baba40c8b1138c517fb8cf04464292 Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Tue, 2 Apr 2024 18:57:05 -0400 Subject: bcachefs: ratelimit informational fsck errors Signed-off-by: Kent Overstreet --- fs/bcachefs/fsck.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/fsck.c b/fs/bcachefs/fsck.c index cbb8b43e419f..b704fb0dda95 100644 --- a/fs/bcachefs/fsck.c +++ b/fs/bcachefs/fsck.c @@ -1131,8 +1131,8 @@ static int check_i_sectors_notnested(struct btree_trans *trans, struct inode_wal i->count = count2; if (i->count != count2) { - bch_err(c, "fsck counted i_sectors wrong for inode %llu:%u: got %llu should be %llu", - w->last_pos.inode, i->snapshot, i->count, count2); + bch_err_ratelimited(c, "fsck counted i_sectors wrong for inode %llu:%u: got %llu should be %llu", + w->last_pos.inode, i->snapshot, i->count, count2); return -BCH_ERR_internal_fsck_err; } @@ -1587,8 +1587,8 @@ static int check_subdir_count_notnested(struct btree_trans *trans, struct inode_ return count2; if (i->count != count2) { - bch_err(c, "fsck counted subdirectories wrong: got %llu should be %llu", - i->count, count2); + bch_err_ratelimited(c, "fsck counted subdirectories wrong for inum %llu:%u: got %llu should be %llu", + w->last_pos.inode, i->snapshot, i->count, count2); i->count = count2; if (i->inode.bi_nlink == i->count) continue; -- cgit From 0200ceed3042222a1f78b3e79ec71a5a52977e3a Mon Sep 17 00:00:00 2001 From: Colin Ian King Date: Thu, 29 Feb 2024 22:51:38 +0000 Subject: vboxsf: remove redundant variable out_len The variable out_len is being used to accumulate the number of bytes but it is not being used for any other purpose. The variable is redundant and can be removed. Cleans up clang scan build warning: fs/vboxsf/utils.c:443:9: warning: variable 'out_len' set but not used [-Wunused-but-set-variable] Signed-off-by: Colin Ian King Link: https://lore.kernel.org/r/20240229225138.351909-1-colin.i.king@gmail.com Reviewed-by: Hans de Goede Signed-off-by: Hans de Goede --- fs/vboxsf/utils.c | 3 --- 1 file changed, 3 deletions(-) (limited to 'fs') diff --git a/fs/vboxsf/utils.c b/fs/vboxsf/utils.c index 72ac9320e6a3..9515bbf0b54c 100644 --- a/fs/vboxsf/utils.c +++ b/fs/vboxsf/utils.c @@ -440,7 +440,6 @@ int vboxsf_nlscpy(struct vboxsf_sbi *sbi, char *name, size_t name_bound_len, { const char *in; char *out; - size_t out_len; size_t out_bound_len; size_t in_bound_len; @@ -448,7 +447,6 @@ int vboxsf_nlscpy(struct vboxsf_sbi *sbi, char *name, size_t name_bound_len, in_bound_len = utf8_len; out = name; - out_len = 0; /* Reserve space for terminating 0 */ out_bound_len = name_bound_len - 1; @@ -469,7 +467,6 @@ int vboxsf_nlscpy(struct vboxsf_sbi *sbi, char *name, size_t name_bound_len, out += nb; out_bound_len -= nb; - out_len += nb; } *out = 0; -- cgit From de3f64b738af57e2732b91a0774facc675b75b54 Mon Sep 17 00:00:00 2001 From: Christophe JAILLET Date: Wed, 1 Nov 2023 11:49:48 +0100 Subject: vboxsf: Avoid an spurious warning if load_nls_xxx() fails If an load_nls_xxx() function fails a few lines above, the 'sbi->bdi_id' is still 0. So, in the error handling path, we will call ida_simple_remove(..., 0) which is not allocated yet. In order to prevent a spurious "ida_free called for id=0 which is not allocated." message, tweak the error handling path and add a new label. Fixes: 0fd169576648 ("fs: Add VirtualBox guest shared folder (vboxsf) support") Signed-off-by: Christophe JAILLET Link: https://lore.kernel.org/r/d09eaaa4e2e08206c58a1a27ca9b3e81dc168773.1698835730.git.christophe.jaillet@wanadoo.fr Reviewed-by: Hans de Goede Signed-off-by: Hans de Goede --- fs/vboxsf/super.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/vboxsf/super.c b/fs/vboxsf/super.c index cabe8ac4fefc..1ed6b1a76bf9 100644 --- a/fs/vboxsf/super.c +++ b/fs/vboxsf/super.c @@ -151,7 +151,7 @@ static int vboxsf_fill_super(struct super_block *sb, struct fs_context *fc) if (!sbi->nls) { vbg_err("vboxsf: Count not load '%s' nls\n", nls_name); err = -EINVAL; - goto fail_free; + goto fail_destroy_idr; } } @@ -224,6 +224,7 @@ fail_free: ida_simple_remove(&vboxsf_bdi_ida, sbi->bdi_id); if (sbi->nls) unload_nls(sbi->nls); +fail_destroy_idr: idr_destroy(&sbi->ino_idr); kfree(sbi); return err; -- cgit From 0141d68f86d78953fb4c3983d666e92f7df4a43d Mon Sep 17 00:00:00 2001 From: Christophe JAILLET Date: Wed, 1 Nov 2023 11:49:49 +0100 Subject: vboxsf: Remove usage of the deprecated ida_simple_xx() API ida_alloc() and ida_free() should be preferred to the deprecated ida_simple_get() and ida_simple_remove(). This is less verbose. Signed-off-by: Christophe JAILLET Link: https://lore.kernel.org/r/b3c057c86b73f0309a6362031d21f4d7ebb60587.1698835730.git.christophe.jaillet@wanadoo.fr Reviewed-by: Hans de Goede Signed-off-by: Hans de Goede --- fs/vboxsf/super.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'fs') diff --git a/fs/vboxsf/super.c b/fs/vboxsf/super.c index 1ed6b1a76bf9..ffb1d565da39 100644 --- a/fs/vboxsf/super.c +++ b/fs/vboxsf/super.c @@ -155,7 +155,7 @@ static int vboxsf_fill_super(struct super_block *sb, struct fs_context *fc) } } - sbi->bdi_id = ida_simple_get(&vboxsf_bdi_ida, 0, 0, GFP_KERNEL); + sbi->bdi_id = ida_alloc(&vboxsf_bdi_ida, GFP_KERNEL); if (sbi->bdi_id < 0) { err = sbi->bdi_id; goto fail_free; @@ -221,7 +221,7 @@ fail_unmap: vboxsf_unmap_folder(sbi->root); fail_free: if (sbi->bdi_id >= 0) - ida_simple_remove(&vboxsf_bdi_ida, sbi->bdi_id); + ida_free(&vboxsf_bdi_ida, sbi->bdi_id); if (sbi->nls) unload_nls(sbi->nls); fail_destroy_idr: @@ -269,7 +269,7 @@ static void vboxsf_put_super(struct super_block *sb) vboxsf_unmap_folder(sbi->root); if (sbi->bdi_id >= 0) - ida_simple_remove(&vboxsf_bdi_ida, sbi->bdi_id); + ida_free(&vboxsf_bdi_ida, sbi->bdi_id); if (sbi->nls) unload_nls(sbi->nls); -- cgit From 1ece2c43b88660ddbdf8ecb772e9c41ed9cda3dd Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Tue, 19 Mar 2024 12:32:04 -0400 Subject: vboxsf: explicitly deny setlease attempts vboxsf does not break leases on its own, so it can't properly handle the case where the hypervisor changes the data. Don't allow file leases on vboxsf. Signed-off-by: Jeff Layton Link: https://lore.kernel.org/r/20240319-setlease-v1-1-5997d67e04b3@kernel.org Reviewed-by: Hans de Goede Signed-off-by: Hans de Goede --- fs/vboxsf/file.c | 1 + 1 file changed, 1 insertion(+) (limited to 'fs') diff --git a/fs/vboxsf/file.c b/fs/vboxsf/file.c index 2307f8037efc..118dedef8ebe 100644 --- a/fs/vboxsf/file.c +++ b/fs/vboxsf/file.c @@ -218,6 +218,7 @@ const struct file_operations vboxsf_reg_fops = { .release = vboxsf_file_release, .fsync = noop_fsync, .splice_read = filemap_splice_read, + .setlease = simple_nosetlease, }; const struct inode_operations vboxsf_reg_iops = { -- cgit From 701b38995e5bdd2a293936c55782140423827fb1 Mon Sep 17 00:00:00 2001 From: Roberto Sassu Date: Wed, 3 Apr 2024 09:57:29 +0200 Subject: security: Place security_path_post_mknod() where the original IMA call was Commit 08abce60d63f ("security: Introduce path_post_mknod hook") introduced security_path_post_mknod(), to replace the IMA-specific call to ima_post_path_mknod(). For symmetry with security_path_mknod(), security_path_post_mknod() was called after a successful mknod operation, for any file type, rather than only for regular files at the time there was the IMA call. However, as reported by VFS maintainers, successful mknod operation does not mean that the dentry always has an inode attached to it (for example, not for FIFOs on a SAMBA mount). If that condition happens, the kernel crashes when security_path_post_mknod() attempts to verify if the inode associated to the dentry is private. Move security_path_post_mknod() where the ima_post_path_mknod() call was, which is obviously correct from IMA/EVM perspective. IMA/EVM are the only in-kernel users, and only need to inspect regular files. Reported-by: Steve French Closes: https://lore.kernel.org/linux-kernel/CAH2r5msAVzxCUHHG8VKrMPUKQHmBpE6K9_vjhgDa1uAvwx4ppw@mail.gmail.com/ Suggested-by: Al Viro Fixes: 08abce60d63f ("security: Introduce path_post_mknod hook") Signed-off-by: Roberto Sassu Reviewed-by: Christian Brauner Reviewed-by: Mimi Zohar Acked-by: Paul Moore Signed-off-by: Linus Torvalds --- fs/namei.c | 7 ++----- security/security.c | 4 ++-- 2 files changed, 4 insertions(+), 7 deletions(-) (limited to 'fs') diff --git a/fs/namei.c b/fs/namei.c index ceb9ddf8dfdd..c5b2a25be7d0 100644 --- a/fs/namei.c +++ b/fs/namei.c @@ -4050,6 +4050,8 @@ retry: case 0: case S_IFREG: error = vfs_create(idmap, path.dentry->d_inode, dentry, mode, true); + if (!error) + security_path_post_mknod(idmap, dentry); break; case S_IFCHR: case S_IFBLK: error = vfs_mknod(idmap, path.dentry->d_inode, @@ -4060,11 +4062,6 @@ retry: dentry, mode, 0); break; } - - if (error) - goto out2; - - security_path_post_mknod(idmap, dentry); out2: done_path_create(&path, dentry); if (retry_estale(error, lookup_flags)) { diff --git a/security/security.c b/security/security.c index 7e118858b545..0a9a0ac3f266 100644 --- a/security/security.c +++ b/security/security.c @@ -1793,11 +1793,11 @@ int security_path_mknod(const struct path *dir, struct dentry *dentry, EXPORT_SYMBOL(security_path_mknod); /** - * security_path_post_mknod() - Update inode security field after file creation + * security_path_post_mknod() - Update inode security after reg file creation * @idmap: idmap of the mount * @dentry: new file * - * Update inode security field after a file has been created. + * Update inode security field after a regular file has been created. */ void security_path_post_mknod(struct mnt_idmap *idmap, struct dentry *dentry) { -- cgit From 27fcec6c27caec05a512d2f1be7f855d8899cb8d Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Sat, 30 Mar 2024 22:25:45 -0400 Subject: bcachefs: Clear recovery_passes_required as they complete without errors Signed-off-by: Kent Overstreet --- fs/bcachefs/recovery.c | 4 +--- fs/bcachefs/recovery_passes.c | 41 ++++++++++++++++++++++++++++++++--------- fs/bcachefs/util.h | 10 ++++++++++ 3 files changed, 43 insertions(+), 12 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c index e8b434009293..88b794d5dbd0 100644 --- a/fs/bcachefs/recovery.c +++ b/fs/bcachefs/recovery.c @@ -838,9 +838,7 @@ use_clean: } if (!test_bit(BCH_FS_error, &c->flags) && - (!bch2_is_zero(ext->recovery_passes_required, sizeof(ext->recovery_passes_required)) || - !bch2_is_zero(ext->errors_silent, sizeof(ext->errors_silent)))) { - memset(ext->recovery_passes_required, 0, sizeof(ext->recovery_passes_required)); + !bch2_is_zero(ext->errors_silent, sizeof(ext->errors_silent))) { memset(ext->errors_silent, 0, sizeof(ext->errors_silent)); write_sb = true; } diff --git a/fs/bcachefs/recovery_passes.c b/fs/bcachefs/recovery_passes.c index 0089d9456b10..b066c5155b74 100644 --- a/fs/bcachefs/recovery_passes.c +++ b/fs/bcachefs/recovery_passes.c @@ -59,18 +59,23 @@ static struct recovery_pass_fn recovery_pass_fns[] = { #undef x }; -u64 bch2_recovery_passes_to_stable(u64 v) -{ - static const u8 map[] = { +static const u8 passes_to_stable_map[] = { #define x(n, id, ...) [BCH_RECOVERY_PASS_##n] = BCH_RECOVERY_PASS_STABLE_##n, BCH_RECOVERY_PASSES() #undef x - }; +}; + +static enum bch_recovery_pass_stable bch2_recovery_pass_to_stable(enum bch_recovery_pass pass) +{ + return passes_to_stable_map[pass]; +} +u64 bch2_recovery_passes_to_stable(u64 v) +{ u64 ret = 0; - for (unsigned i = 0; i < ARRAY_SIZE(map); i++) + for (unsigned i = 0; i < ARRAY_SIZE(passes_to_stable_map); i++) if (v & BIT_ULL(i)) - ret |= BIT_ULL(map[i]); + ret |= BIT_ULL(passes_to_stable_map[i]); return ret; } @@ -116,13 +121,13 @@ int bch2_run_explicit_recovery_pass(struct bch_fs *c, int bch2_run_explicit_recovery_pass_persistent(struct bch_fs *c, enum bch_recovery_pass pass) { - __le64 s = cpu_to_le64(bch2_recovery_passes_to_stable(BIT_ULL(pass))); + enum bch_recovery_pass_stable s = bch2_recovery_pass_to_stable(pass); mutex_lock(&c->sb_lock); struct bch_sb_field_ext *ext = bch2_sb_field_get(c->disk_sb.sb, ext); - if (!(ext->recovery_passes_required[0] & s)) { - ext->recovery_passes_required[0] |= s; + if (!test_bit_le64(s, ext->recovery_passes_required)) { + __set_bit_le64(s, ext->recovery_passes_required); bch2_write_super(c); } mutex_unlock(&c->sb_lock); @@ -130,6 +135,21 @@ int bch2_run_explicit_recovery_pass_persistent(struct bch_fs *c, return bch2_run_explicit_recovery_pass(c, pass); } +static void bch2_clear_recovery_pass_required(struct bch_fs *c, + enum bch_recovery_pass pass) +{ + enum bch_recovery_pass_stable s = bch2_recovery_pass_to_stable(pass); + + mutex_lock(&c->sb_lock); + struct bch_sb_field_ext *ext = bch2_sb_field_get(c->disk_sb.sb, ext); + + if (test_bit_le64(s, ext->recovery_passes_required)) { + __clear_bit_le64(s, ext->recovery_passes_required); + bch2_write_super(c); + } + mutex_unlock(&c->sb_lock); +} + u64 bch2_fsck_recovery_passes(void) { u64 ret = 0; @@ -218,6 +238,9 @@ int bch2_run_recovery_passes(struct bch_fs *c) c->recovery_pass_done = max(c->recovery_pass_done, c->curr_recovery_pass); + if (!test_bit(BCH_FS_error, &c->flags)) + bch2_clear_recovery_pass_required(c, c->curr_recovery_pass); + c->curr_recovery_pass++; } diff --git a/fs/bcachefs/util.h b/fs/bcachefs/util.h index 175aee3074c7..4577c0789a3a 100644 --- a/fs/bcachefs/util.h +++ b/fs/bcachefs/util.h @@ -797,4 +797,14 @@ static inline void __set_bit_le64(size_t bit, __le64 *addr) addr[bit / 64] |= cpu_to_le64(BIT_ULL(bit % 64)); } +static inline void __clear_bit_le64(size_t bit, __le64 *addr) +{ + addr[bit / 64] &= !cpu_to_le64(BIT_ULL(bit % 64)); +} + +static inline bool test_bit_le64(size_t bit, __le64 *addr) +{ + return (addr[bit / 64] & cpu_to_le64(BIT_ULL(bit % 64))) != 0; +} + #endif /* _BCACHEFS_UTIL_H */ -- cgit From bdbf953b3c9036e284e28c9484dda5c81b2a45fa Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Tue, 19 Mar 2024 18:56:26 -0400 Subject: bcachefs: bch2_shoot_down_journal_keys() Signed-off-by: Kent Overstreet --- fs/bcachefs/btree_journal_iter.c | 19 +++++++++++++++++++ fs/bcachefs/btree_journal_iter.h | 4 ++++ fs/bcachefs/recovery.c | 22 ++++++++++++---------- 3 files changed, 35 insertions(+), 10 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/btree_journal_iter.c b/fs/bcachefs/btree_journal_iter.c index 1f588264575d..5cbcbfe85235 100644 --- a/fs/bcachefs/btree_journal_iter.c +++ b/fs/bcachefs/btree_journal_iter.c @@ -567,3 +567,22 @@ int bch2_journal_keys_sort(struct bch_fs *c) bch_verbose(c, "Journal keys: %zu read, %zu after sorting and compacting", nr_read, keys->nr); return 0; } + +void bch2_shoot_down_journal_keys(struct bch_fs *c, enum btree_id btree, + unsigned level_min, unsigned level_max, + struct bpos start, struct bpos end) +{ + struct journal_keys *keys = &c->journal_keys; + size_t dst = 0; + + move_gap(keys, keys->nr); + + darray_for_each(*keys, i) + if (!(i->btree_id == btree && + i->level >= level_min && + i->level <= level_max && + bpos_ge(i->k->k.p, start) && + bpos_le(i->k->k.p, end))) + keys->data[dst++] = *i; + keys->nr = keys->gap = dst; +} diff --git a/fs/bcachefs/btree_journal_iter.h b/fs/bcachefs/btree_journal_iter.h index 09cd53091a05..af25046ebcaa 100644 --- a/fs/bcachefs/btree_journal_iter.h +++ b/fs/bcachefs/btree_journal_iter.h @@ -66,4 +66,8 @@ void bch2_journal_entries_free(struct bch_fs *); int bch2_journal_keys_sort(struct bch_fs *); +void bch2_shoot_down_journal_keys(struct bch_fs *, enum btree_id, + unsigned, unsigned, + struct bpos, struct bpos); + #endif /* _BCACHEFS_BTREE_JOURNAL_ITER_H */ diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c index 88b794d5dbd0..7cbf30e3ef73 100644 --- a/fs/bcachefs/recovery.c +++ b/fs/bcachefs/recovery.c @@ -47,7 +47,7 @@ static bool btree_id_is_alloc(enum btree_id id) } /* for -o reconstruct_alloc: */ -static void do_reconstruct_alloc(struct bch_fs *c) +static void bch2_reconstruct_alloc(struct bch_fs *c) { bch2_journal_log_msg(c, "dropping alloc info"); bch_info(c, "dropping and reconstructing all alloc info"); @@ -82,15 +82,17 @@ static void do_reconstruct_alloc(struct bch_fs *c) c->recovery_passes_explicit |= bch2_recovery_passes_from_stable(le64_to_cpu(ext->recovery_passes_required[0])); - struct journal_keys *keys = &c->journal_keys; - size_t src, dst; - - move_gap(keys, keys->nr); - for (src = 0, dst = 0; src < keys->nr; src++) - if (!btree_id_is_alloc(keys->data[src].btree_id)) - keys->data[dst++] = keys->data[src]; - keys->nr = keys->gap = dst; + bch2_shoot_down_journal_keys(c, BTREE_ID_alloc, + 0, BTREE_MAX_DEPTH, POS_MIN, SPOS_MAX); + bch2_shoot_down_journal_keys(c, BTREE_ID_backpointers, + 0, BTREE_MAX_DEPTH, POS_MIN, SPOS_MAX); + bch2_shoot_down_journal_keys(c, BTREE_ID_need_discard, + 0, BTREE_MAX_DEPTH, POS_MIN, SPOS_MAX); + bch2_shoot_down_journal_keys(c, BTREE_ID_freespace, + 0, BTREE_MAX_DEPTH, POS_MIN, SPOS_MAX); + bch2_shoot_down_journal_keys(c, BTREE_ID_bucket_gens, + 0, BTREE_MAX_DEPTH, POS_MIN, SPOS_MAX); } /* @@ -731,7 +733,7 @@ use_clean: c->journal_replay_seq_end = blacklist_seq - 1; if (c->opts.reconstruct_alloc) - do_reconstruct_alloc(c); + bch2_reconstruct_alloc(c); zero_out_btree_mem_ptr(&c->journal_keys); -- cgit From ca1e02f7e9a1352b3f7b04de821ae74c9e07df74 Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Fri, 22 Mar 2024 19:26:33 -0400 Subject: bcachefs: Etyzinger cleanups Pull out eytzinger.c and kill eytzinger_cmp_fn. We now provide eytzinger0_sort and eytzinger0_sort_r, which use the standard cmp_func_t and cmp_r_func_t callbacks. Signed-off-by: Kent Overstreet --- fs/bcachefs/Makefile | 1 + fs/bcachefs/eytzinger.c | 234 ++++++++++++++++++++++++++++++++++++ fs/bcachefs/eytzinger.h | 63 ++++++---- fs/bcachefs/journal_seq_blacklist.c | 3 +- fs/bcachefs/replicas.c | 19 +-- fs/bcachefs/util.c | 143 ---------------------- fs/bcachefs/util.h | 4 - 7 files changed, 285 insertions(+), 182 deletions(-) create mode 100644 fs/bcachefs/eytzinger.c (limited to 'fs') diff --git a/fs/bcachefs/Makefile b/fs/bcachefs/Makefile index f6d86eb4b976..0933493a322c 100644 --- a/fs/bcachefs/Makefile +++ b/fs/bcachefs/Makefile @@ -37,6 +37,7 @@ bcachefs-y := \ error.o \ extents.o \ extent_update.o \ + eytzinger.o \ fs.o \ fs-common.o \ fs-ioctl.o \ diff --git a/fs/bcachefs/eytzinger.c b/fs/bcachefs/eytzinger.c new file mode 100644 index 000000000000..4ce5e957a6e9 --- /dev/null +++ b/fs/bcachefs/eytzinger.c @@ -0,0 +1,234 @@ +// SPDX-License-Identifier: GPL-2.0 + +#include "eytzinger.h" + +/** + * is_aligned - is this pointer & size okay for word-wide copying? + * @base: pointer to data + * @size: size of each element + * @align: required alignment (typically 4 or 8) + * + * Returns true if elements can be copied using word loads and stores. + * The size must be a multiple of the alignment, and the base address must + * be if we do not have CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS. + * + * For some reason, gcc doesn't know to optimize "if (a & mask || b & mask)" + * to "if ((a | b) & mask)", so we do that by hand. + */ +__attribute_const__ __always_inline +static bool is_aligned(const void *base, size_t size, unsigned char align) +{ + unsigned char lsbits = (unsigned char)size; + + (void)base; +#ifndef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS + lsbits |= (unsigned char)(uintptr_t)base; +#endif + return (lsbits & (align - 1)) == 0; +} + +/** + * swap_words_32 - swap two elements in 32-bit chunks + * @a: pointer to the first element to swap + * @b: pointer to the second element to swap + * @n: element size (must be a multiple of 4) + * + * Exchange the two objects in memory. This exploits base+index addressing, + * which basically all CPUs have, to minimize loop overhead computations. + * + * For some reason, on x86 gcc 7.3.0 adds a redundant test of n at the + * bottom of the loop, even though the zero flag is still valid from the + * subtract (since the intervening mov instructions don't alter the flags). + * Gcc 8.1.0 doesn't have that problem. + */ +static void swap_words_32(void *a, void *b, size_t n) +{ + do { + u32 t = *(u32 *)(a + (n -= 4)); + *(u32 *)(a + n) = *(u32 *)(b + n); + *(u32 *)(b + n) = t; + } while (n); +} + +/** + * swap_words_64 - swap two elements in 64-bit chunks + * @a: pointer to the first element to swap + * @b: pointer to the second element to swap + * @n: element size (must be a multiple of 8) + * + * Exchange the two objects in memory. This exploits base+index + * addressing, which basically all CPUs have, to minimize loop overhead + * computations. + * + * We'd like to use 64-bit loads if possible. If they're not, emulating + * one requires base+index+4 addressing which x86 has but most other + * processors do not. If CONFIG_64BIT, we definitely have 64-bit loads, + * but it's possible to have 64-bit loads without 64-bit pointers (e.g. + * x32 ABI). Are there any cases the kernel needs to worry about? + */ +static void swap_words_64(void *a, void *b, size_t n) +{ + do { +#ifdef CONFIG_64BIT + u64 t = *(u64 *)(a + (n -= 8)); + *(u64 *)(a + n) = *(u64 *)(b + n); + *(u64 *)(b + n) = t; +#else + /* Use two 32-bit transfers to avoid base+index+4 addressing */ + u32 t = *(u32 *)(a + (n -= 4)); + *(u32 *)(a + n) = *(u32 *)(b + n); + *(u32 *)(b + n) = t; + + t = *(u32 *)(a + (n -= 4)); + *(u32 *)(a + n) = *(u32 *)(b + n); + *(u32 *)(b + n) = t; +#endif + } while (n); +} + +/** + * swap_bytes - swap two elements a byte at a time + * @a: pointer to the first element to swap + * @b: pointer to the second element to swap + * @n: element size + * + * This is the fallback if alignment doesn't allow using larger chunks. + */ +static void swap_bytes(void *a, void *b, size_t n) +{ + do { + char t = ((char *)a)[--n]; + ((char *)a)[n] = ((char *)b)[n]; + ((char *)b)[n] = t; + } while (n); +} + +/* + * The values are arbitrary as long as they can't be confused with + * a pointer, but small integers make for the smallest compare + * instructions. + */ +#define SWAP_WORDS_64 (swap_r_func_t)0 +#define SWAP_WORDS_32 (swap_r_func_t)1 +#define SWAP_BYTES (swap_r_func_t)2 +#define SWAP_WRAPPER (swap_r_func_t)3 + +struct wrapper { + cmp_func_t cmp; + swap_func_t swap; +}; + +/* + * The function pointer is last to make tail calls most efficient if the + * compiler decides not to inline this function. + */ +static void do_swap(void *a, void *b, size_t size, swap_r_func_t swap_func, const void *priv) +{ + if (swap_func == SWAP_WRAPPER) { + ((const struct wrapper *)priv)->swap(a, b, (int)size); + return; + } + + if (swap_func == SWAP_WORDS_64) + swap_words_64(a, b, size); + else if (swap_func == SWAP_WORDS_32) + swap_words_32(a, b, size); + else if (swap_func == SWAP_BYTES) + swap_bytes(a, b, size); + else + swap_func(a, b, (int)size, priv); +} + +#define _CMP_WRAPPER ((cmp_r_func_t)0L) + +static int do_cmp(const void *a, const void *b, cmp_r_func_t cmp, const void *priv) +{ + if (cmp == _CMP_WRAPPER) + return ((const struct wrapper *)priv)->cmp(a, b); + return cmp(a, b, priv); +} + +static inline int eytzinger0_do_cmp(void *base, size_t n, size_t size, + cmp_r_func_t cmp_func, const void *priv, + size_t l, size_t r) +{ + return do_cmp(base + inorder_to_eytzinger0(l, n) * size, + base + inorder_to_eytzinger0(r, n) * size, + cmp_func, priv); +} + +static inline void eytzinger0_do_swap(void *base, size_t n, size_t size, + swap_r_func_t swap_func, const void *priv, + size_t l, size_t r) +{ + do_swap(base + inorder_to_eytzinger0(l, n) * size, + base + inorder_to_eytzinger0(r, n) * size, + size, swap_func, priv); +} + +void eytzinger0_sort_r(void *base, size_t n, size_t size, + cmp_r_func_t cmp_func, + swap_r_func_t swap_func, + const void *priv) +{ + int i, c, r; + + /* called from 'sort' without swap function, let's pick the default */ + if (swap_func == SWAP_WRAPPER && !((struct wrapper *)priv)->swap) + swap_func = NULL; + + if (!swap_func) { + if (is_aligned(base, size, 8)) + swap_func = SWAP_WORDS_64; + else if (is_aligned(base, size, 4)) + swap_func = SWAP_WORDS_32; + else + swap_func = SWAP_BYTES; + } + + /* heapify */ + for (i = n / 2 - 1; i >= 0; --i) { + for (r = i; r * 2 + 1 < n; r = c) { + c = r * 2 + 1; + + if (c + 1 < n && + eytzinger0_do_cmp(base, n, size, cmp_func, priv, c, c + 1) < 0) + c++; + + if (eytzinger0_do_cmp(base, n, size, cmp_func, priv, r, c) >= 0) + break; + + eytzinger0_do_swap(base, n, size, swap_func, priv, r, c); + } + } + + /* sort */ + for (i = n - 1; i > 0; --i) { + eytzinger0_do_swap(base, n, size, swap_func, priv, 0, i); + + for (r = 0; r * 2 + 1 < i; r = c) { + c = r * 2 + 1; + + if (c + 1 < i && + eytzinger0_do_cmp(base, n, size, cmp_func, priv, c, c + 1) < 0) + c++; + + if (eytzinger0_do_cmp(base, n, size, cmp_func, priv, r, c) >= 0) + break; + + eytzinger0_do_swap(base, n, size, swap_func, priv, r, c); + } + } +} + +void eytzinger0_sort(void *base, size_t n, size_t size, + cmp_func_t cmp_func, + swap_func_t swap_func) +{ + struct wrapper w = { + .cmp = cmp_func, + .swap = swap_func, + }; + + return eytzinger0_sort_r(base, n, size, _CMP_WRAPPER, SWAP_WRAPPER, &w); +} diff --git a/fs/bcachefs/eytzinger.h b/fs/bcachefs/eytzinger.h index b04750dbf870..ee0e2df33322 100644 --- a/fs/bcachefs/eytzinger.h +++ b/fs/bcachefs/eytzinger.h @@ -5,23 +5,33 @@ #include #include -#include "util.h" +#ifdef EYTZINGER_DEBUG +#define EYTZINGER_BUG_ON(cond) BUG_ON(cond) +#else +#define EYTZINGER_BUG_ON(cond) +#endif /* * Traversal for trees in eytzinger layout - a full binary tree layed out in an - * array - */ - -/* - * One based indexing version: + * array. * - * With one based indexing each level of the tree starts at a power of two - - * good for cacheline alignment: + * Consider using an eytzinger tree any time you would otherwise be doing binary + * search over an array. Binary search is a worst case scenario for branch + * prediction and prefetching, but in an eytzinger tree every node's children + * are adjacent in memory, thus we can prefetch children before knowing the + * result of the comparison, assuming multiple nodes fit on a cacheline. + * + * Two variants are provided, for one based indexing and zero based indexing. + * + * Zero based indexing is more convenient, but one based indexing has better + * alignment and thus better performance because each new level of the tree + * starts at a power of two, and thus if element 0 was cacheline aligned, each + * new level will be as well. */ static inline unsigned eytzinger1_child(unsigned i, unsigned child) { - EBUG_ON(child > 1); + EYTZINGER_BUG_ON(child > 1); return (i << 1) + child; } @@ -58,7 +68,7 @@ static inline unsigned eytzinger1_last(unsigned size) static inline unsigned eytzinger1_next(unsigned i, unsigned size) { - EBUG_ON(i > size); + EYTZINGER_BUG_ON(i > size); if (eytzinger1_right_child(i) <= size) { i = eytzinger1_right_child(i); @@ -74,7 +84,7 @@ static inline unsigned eytzinger1_next(unsigned i, unsigned size) static inline unsigned eytzinger1_prev(unsigned i, unsigned size) { - EBUG_ON(i > size); + EYTZINGER_BUG_ON(i > size); if (eytzinger1_left_child(i) <= size) { i = eytzinger1_left_child(i) + 1; @@ -101,7 +111,7 @@ static inline unsigned __eytzinger1_to_inorder(unsigned i, unsigned size, unsigned shift = __fls(size) - b; int s; - EBUG_ON(!i || i > size); + EYTZINGER_BUG_ON(!i || i > size); i ^= 1U << b; i <<= 1; @@ -126,7 +136,7 @@ static inline unsigned __inorder_to_eytzinger1(unsigned i, unsigned size, unsigned shift; int s; - EBUG_ON(!i || i > size); + EYTZINGER_BUG_ON(!i || i > size); /* * sign bit trick: @@ -164,7 +174,7 @@ static inline unsigned inorder_to_eytzinger1(unsigned i, unsigned size) static inline unsigned eytzinger0_child(unsigned i, unsigned child) { - EBUG_ON(child > 1); + EYTZINGER_BUG_ON(child > 1); return (i << 1) + 1 + child; } @@ -231,11 +241,9 @@ static inline unsigned inorder_to_eytzinger0(unsigned i, unsigned size) (_i) != -1; \ (_i) = eytzinger0_next((_i), (_size))) -typedef int (*eytzinger_cmp_fn)(const void *l, const void *r, size_t size); - /* return greatest node <= @search, or -1 if not found */ static inline ssize_t eytzinger0_find_le(void *base, size_t nr, size_t size, - eytzinger_cmp_fn cmp, const void *search) + cmp_func_t cmp, const void *search) { unsigned i, n = 0; @@ -244,21 +252,24 @@ static inline ssize_t eytzinger0_find_le(void *base, size_t nr, size_t size, do { i = n; - n = eytzinger0_child(i, cmp(search, base + i * size, size) >= 0); + n = eytzinger0_child(i, cmp(base + i * size, search) <= 0); } while (n < nr); if (n & 1) { /* @i was greater than @search, return previous node: */ - - if (i == eytzinger0_first(nr)) - return -1; - return eytzinger0_prev(i, nr); } else { return i; } } +static inline ssize_t eytzinger0_find_gt(void *base, size_t nr, size_t size, + cmp_func_t cmp, const void *search) +{ + ssize_t idx = eytzinger0_find_le(base, nr, size, cmp, search); + return eytzinger0_next(idx, size); +} + #define eytzinger0_find(base, nr, size, _cmp, search) \ ({ \ void *_base = (base); \ @@ -269,13 +280,13 @@ static inline ssize_t eytzinger0_find_le(void *base, size_t nr, size_t size, int _res; \ \ while (_i < _nr && \ - (_res = _cmp(_search, _base + _i * _size, _size))) \ + (_res = _cmp(_search, _base + _i * _size))) \ _i = eytzinger0_child(_i, _res > 0); \ _i; \ }) -void eytzinger0_sort(void *, size_t, size_t, - int (*cmp_func)(const void *, const void *, size_t), - void (*swap_func)(void *, void *, size_t)); +void eytzinger0_sort_r(void *, size_t, size_t, + cmp_r_func_t, swap_r_func_t, const void *); +void eytzinger0_sort(void *, size_t, size_t, cmp_func_t, swap_func_t); #endif /* _EYTZINGER_H */ diff --git a/fs/bcachefs/journal_seq_blacklist.c b/fs/bcachefs/journal_seq_blacklist.c index b5303874fc35..37a024e034d4 100644 --- a/fs/bcachefs/journal_seq_blacklist.c +++ b/fs/bcachefs/journal_seq_blacklist.c @@ -95,8 +95,7 @@ out: return ret ?: bch2_blacklist_table_initialize(c); } -static int journal_seq_blacklist_table_cmp(const void *_l, - const void *_r, size_t size) +static int journal_seq_blacklist_table_cmp(const void *_l, const void *_r) { const struct journal_seq_blacklist_table_entry *l = _l; const struct journal_seq_blacklist_table_entry *r = _r; diff --git a/fs/bcachefs/replicas.c b/fs/bcachefs/replicas.c index cc2672c12031..678b9c20e251 100644 --- a/fs/bcachefs/replicas.c +++ b/fs/bcachefs/replicas.c @@ -6,12 +6,15 @@ #include "replicas.h" #include "super-io.h" +#include + static int bch2_cpu_replicas_to_sb_replicas(struct bch_fs *, struct bch_replicas_cpu *); /* Some (buggy!) compilers don't allow memcmp to be passed as a pointer */ -static int bch2_memcmp(const void *l, const void *r, size_t size) +static int bch2_memcmp(const void *l, const void *r, const void *priv) { + size_t size = (size_t) priv; return memcmp(l, r, size); } @@ -39,7 +42,8 @@ void bch2_replicas_entry_sort(struct bch_replicas_entry_v1 *e) static void bch2_cpu_replicas_sort(struct bch_replicas_cpu *r) { - eytzinger0_sort(r->entries, r->nr, r->entry_size, bch2_memcmp, NULL); + eytzinger0_sort_r(r->entries, r->nr, r->entry_size, + bch2_memcmp, NULL, (void *)(size_t)r->entry_size); } static void bch2_replicas_entry_v0_to_text(struct printbuf *out, @@ -228,7 +232,7 @@ static inline int __replicas_entry_idx(struct bch_replicas_cpu *r, verify_replicas_entry(search); -#define entry_cmp(_l, _r, size) memcmp(_l, _r, entry_size) +#define entry_cmp(_l, _r) memcmp(_l, _r, entry_size) idx = eytzinger0_find(r->entries, r->nr, r->entry_size, entry_cmp, search); #undef entry_cmp @@ -824,10 +828,11 @@ static int bch2_cpu_replicas_validate(struct bch_replicas_cpu *cpu_r, { unsigned i; - sort_cmp_size(cpu_r->entries, - cpu_r->nr, - cpu_r->entry_size, - bch2_memcmp, NULL); + sort_r(cpu_r->entries, + cpu_r->nr, + cpu_r->entry_size, + bch2_memcmp, NULL, + (void *)(size_t)cpu_r->entry_size); for (i = 0; i < cpu_r->nr; i++) { struct bch_replicas_entry_v1 *e = diff --git a/fs/bcachefs/util.c b/fs/bcachefs/util.c index 216fadf16928..92c6ad75e702 100644 --- a/fs/bcachefs/util.c +++ b/fs/bcachefs/util.c @@ -707,149 +707,6 @@ void memcpy_from_bio(void *dst, struct bio *src, struct bvec_iter src_iter) } } -static int alignment_ok(const void *base, size_t align) -{ - return IS_ENABLED(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) || - ((unsigned long)base & (align - 1)) == 0; -} - -static void u32_swap(void *a, void *b, size_t size) -{ - u32 t = *(u32 *)a; - *(u32 *)a = *(u32 *)b; - *(u32 *)b = t; -} - -static void u64_swap(void *a, void *b, size_t size) -{ - u64 t = *(u64 *)a; - *(u64 *)a = *(u64 *)b; - *(u64 *)b = t; -} - -static void generic_swap(void *a, void *b, size_t size) -{ - char t; - - do { - t = *(char *)a; - *(char *)a++ = *(char *)b; - *(char *)b++ = t; - } while (--size > 0); -} - -static inline int do_cmp(void *base, size_t n, size_t size, - int (*cmp_func)(const void *, const void *, size_t), - size_t l, size_t r) -{ - return cmp_func(base + inorder_to_eytzinger0(l, n) * size, - base + inorder_to_eytzinger0(r, n) * size, - size); -} - -static inline void do_swap(void *base, size_t n, size_t size, - void (*swap_func)(void *, void *, size_t), - size_t l, size_t r) -{ - swap_func(base + inorder_to_eytzinger0(l, n) * size, - base + inorder_to_eytzinger0(r, n) * size, - size); -} - -void eytzinger0_sort(void *base, size_t n, size_t size, - int (*cmp_func)(const void *, const void *, size_t), - void (*swap_func)(void *, void *, size_t)) -{ - int i, c, r; - - if (!swap_func) { - if (size == 4 && alignment_ok(base, 4)) - swap_func = u32_swap; - else if (size == 8 && alignment_ok(base, 8)) - swap_func = u64_swap; - else - swap_func = generic_swap; - } - - /* heapify */ - for (i = n / 2 - 1; i >= 0; --i) { - for (r = i; r * 2 + 1 < n; r = c) { - c = r * 2 + 1; - - if (c + 1 < n && - do_cmp(base, n, size, cmp_func, c, c + 1) < 0) - c++; - - if (do_cmp(base, n, size, cmp_func, r, c) >= 0) - break; - - do_swap(base, n, size, swap_func, r, c); - } - } - - /* sort */ - for (i = n - 1; i > 0; --i) { - do_swap(base, n, size, swap_func, 0, i); - - for (r = 0; r * 2 + 1 < i; r = c) { - c = r * 2 + 1; - - if (c + 1 < i && - do_cmp(base, n, size, cmp_func, c, c + 1) < 0) - c++; - - if (do_cmp(base, n, size, cmp_func, r, c) >= 0) - break; - - do_swap(base, n, size, swap_func, r, c); - } - } -} - -void sort_cmp_size(void *base, size_t num, size_t size, - int (*cmp_func)(const void *, const void *, size_t), - void (*swap_func)(void *, void *, size_t size)) -{ - /* pre-scale counters for performance */ - int i = (num/2 - 1) * size, n = num * size, c, r; - - if (!swap_func) { - if (size == 4 && alignment_ok(base, 4)) - swap_func = u32_swap; - else if (size == 8 && alignment_ok(base, 8)) - swap_func = u64_swap; - else - swap_func = generic_swap; - } - - /* heapify */ - for ( ; i >= 0; i -= size) { - for (r = i; r * 2 + size < n; r = c) { - c = r * 2 + size; - if (c < n - size && - cmp_func(base + c, base + c + size, size) < 0) - c += size; - if (cmp_func(base + r, base + c, size) >= 0) - break; - swap_func(base + r, base + c, size); - } - } - - /* sort */ - for (i = n - size; i > 0; i -= size) { - swap_func(base, base + i, size); - for (r = 0; r * 2 + size < i; r = c) { - c = r * 2 + size; - if (c < i - size && - cmp_func(base + c, base + c + size, size) < 0) - c += size; - if (cmp_func(base + r, base + c, size) >= 0) - break; - swap_func(base + r, base + c, size); - } - } -} - #if 0 void eytzinger1_test(void) { diff --git a/fs/bcachefs/util.h b/fs/bcachefs/util.h index 4577c0789a3a..b7e7c29278fc 100644 --- a/fs/bcachefs/util.h +++ b/fs/bcachefs/util.h @@ -631,10 +631,6 @@ static inline void memset_u64s_tail(void *s, int c, unsigned bytes) memset(s + bytes, c, rem); } -void sort_cmp_size(void *base, size_t num, size_t size, - int (*cmp_func)(const void *, const void *, size_t), - void (*swap_func)(void *, void *, size_t)); - /* just the memmove, doesn't update @_nr */ #define __array_insert_item(_array, _nr, _pos) \ memmove(&(_array)[(_pos) + 1], \ -- cgit From f2f61f4192de536fa36bebe49dc2c241213b53ee Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Thu, 14 Mar 2024 22:17:40 -0400 Subject: bcachefs: bch2_btree_root_alloc() -> bch2_btree_root_alloc_fake() Signed-off-by: Kent Overstreet --- fs/bcachefs/btree_update_interior.c | 8 ++++---- fs/bcachefs/btree_update_interior.h | 2 +- fs/bcachefs/recovery.c | 4 ++-- 3 files changed, 7 insertions(+), 7 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/btree_update_interior.c b/fs/bcachefs/btree_update_interior.c index 2b46e6ad248a..32397b99752f 100644 --- a/fs/bcachefs/btree_update_interior.c +++ b/fs/bcachefs/btree_update_interior.c @@ -2466,7 +2466,7 @@ void bch2_btree_set_root_for_read(struct bch_fs *c, struct btree *b) bch2_btree_set_root_inmem(c, b); } -static int __bch2_btree_root_alloc(struct btree_trans *trans, enum btree_id id) +static int __bch2_btree_root_alloc_fake(struct btree_trans *trans, enum btree_id id, unsigned level) { struct bch_fs *c = trans->c; struct closure cl; @@ -2485,7 +2485,7 @@ static int __bch2_btree_root_alloc(struct btree_trans *trans, enum btree_id id) set_btree_node_fake(b); set_btree_node_need_rewrite(b); - b->c.level = 0; + b->c.level = level; b->c.btree_id = id; bkey_btree_ptr_init(&b->key); @@ -2512,9 +2512,9 @@ static int __bch2_btree_root_alloc(struct btree_trans *trans, enum btree_id id) return 0; } -void bch2_btree_root_alloc(struct bch_fs *c, enum btree_id id) +void bch2_btree_root_alloc_fake(struct bch_fs *c, enum btree_id id, unsigned level) { - bch2_trans_run(c, __bch2_btree_root_alloc(trans, id)); + bch2_trans_run(c, __bch2_btree_root_alloc_fake(trans, id, level)); } static void bch2_btree_update_to_text(struct printbuf *out, struct btree_update *as) diff --git a/fs/bcachefs/btree_update_interior.h b/fs/bcachefs/btree_update_interior.h index 1d41d3f4b107..88dcf5a22a3b 100644 --- a/fs/bcachefs/btree_update_interior.h +++ b/fs/bcachefs/btree_update_interior.h @@ -171,7 +171,7 @@ int bch2_btree_node_update_key_get_iter(struct btree_trans *, struct btree *, struct bkey_i *, unsigned, bool); void bch2_btree_set_root_for_read(struct bch_fs *, struct btree *); -void bch2_btree_root_alloc(struct bch_fs *, enum btree_id); +void bch2_btree_root_alloc_fake(struct bch_fs *, enum btree_id, unsigned); static inline unsigned btree_update_reserve_required(struct bch_fs *c, struct btree *b) diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c index 7cbf30e3ef73..f234de0ac834 100644 --- a/fs/bcachefs/recovery.c +++ b/fs/bcachefs/recovery.c @@ -476,7 +476,7 @@ static int read_btree_roots(struct bch_fs *c) if (!r->b) { r->alive = false; r->level = 0; - bch2_btree_root_alloc(c, i); + bch2_btree_root_alloc_fake(c, i, 0); } } fsck_err: @@ -929,7 +929,7 @@ int bch2_fs_initialize(struct bch_fs *c) set_bit(BCH_FS_may_go_rw, &c->flags); for (unsigned i = 0; i < BTREE_ID_NR; i++) - bch2_btree_root_alloc(c, i); + bch2_btree_root_alloc_fake(c, i, 0); for_each_member_device(c, ca) bch2_dev_usage_init(ca); -- cgit From b268aa4e7fb8be3c50e25a09008fb2feed2cd345 Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Sun, 10 Mar 2024 16:18:41 -0400 Subject: bcachefs: Don't skip fake btree roots in fsck When a btree root is unreadable, we might still have keys fro the journal to walk and mark. Signed-off-by: Kent Overstreet --- fs/bcachefs/btree_gc.c | 3 --- 1 file changed, 3 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/btree_gc.c b/fs/bcachefs/btree_gc.c index e5d2c6daa663..9db9c8a5beaa 100644 --- a/fs/bcachefs/btree_gc.c +++ b/fs/bcachefs/btree_gc.c @@ -931,9 +931,6 @@ static int bch2_gc_btree_init(struct btree_trans *trans, b = bch2_btree_id_root(c, btree_id)->b; - if (btree_node_fake(b)) - return 0; - six_lock_read(&b->c.lock, NULL, NULL); printbuf_reset(&buf); bch2_bpos_to_text(&buf, b->data->min_key); -- cgit From 4409b8081d1624af814a9cda781ad9cdda3973cb Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Mon, 11 Mar 2024 23:11:46 -0400 Subject: bcachefs: Repair pass for scanning for btree nodes If a btree root or interior btree node goes bad, we're going to lose a lot of data, unless we can recover the nodes that it pointed to by scanning. Fortunately btree node headers are fully self describing, and additionally the magic number is xored with the filesytem UUID, so we can do so safely. This implements the scanning - next patch will rework topology repair to make use of the found nodes. Signed-off-by: Kent Overstreet --- fs/bcachefs/Makefile | 1 + fs/bcachefs/bcachefs.h | 3 + fs/bcachefs/btree_node_scan.c | 495 ++++++++++++++++++++++++++++++++++++ fs/bcachefs/btree_node_scan.h | 11 + fs/bcachefs/btree_node_scan_types.h | 30 +++ fs/bcachefs/extents.c | 52 ++-- fs/bcachefs/extents.h | 1 + fs/bcachefs/opts.h | 4 +- fs/bcachefs/recovery.c | 54 ++-- fs/bcachefs/recovery_passes.c | 1 + fs/bcachefs/recovery_passes_types.h | 1 + fs/bcachefs/super.c | 3 + 12 files changed, 605 insertions(+), 51 deletions(-) create mode 100644 fs/bcachefs/btree_node_scan.c create mode 100644 fs/bcachefs/btree_node_scan.h create mode 100644 fs/bcachefs/btree_node_scan_types.h (limited to 'fs') diff --git a/fs/bcachefs/Makefile b/fs/bcachefs/Makefile index 0933493a322c..66ca0bbee639 100644 --- a/fs/bcachefs/Makefile +++ b/fs/bcachefs/Makefile @@ -17,6 +17,7 @@ bcachefs-y := \ btree_journal_iter.o \ btree_key_cache.o \ btree_locking.o \ + btree_node_scan.o \ btree_trans_commit.o \ btree_update.o \ btree_update_interior.o \ diff --git a/fs/bcachefs/bcachefs.h b/fs/bcachefs/bcachefs.h index 963162a627cd..93a61dbaa3d8 100644 --- a/fs/bcachefs/bcachefs.h +++ b/fs/bcachefs/bcachefs.h @@ -456,6 +456,7 @@ enum bch_time_stats { #include "alloc_types.h" #include "btree_types.h" +#include "btree_node_scan_types.h" #include "btree_write_buffer_types.h" #include "buckets_types.h" #include "buckets_waiting_for_journal_types.h" @@ -1103,6 +1104,8 @@ struct bch_fs { struct journal_keys journal_keys; struct list_head journal_iters; + struct find_btree_nodes found_btree_nodes; + u64 last_bucket_seq_cleanup; u64 counters_on_mount[BCH_COUNTER_NR]; diff --git a/fs/bcachefs/btree_node_scan.c b/fs/bcachefs/btree_node_scan.c new file mode 100644 index 000000000000..3f33be7e5e5c --- /dev/null +++ b/fs/bcachefs/btree_node_scan.c @@ -0,0 +1,495 @@ +// SPDX-License-Identifier: GPL-2.0 + +#include "bcachefs.h" +#include "btree_cache.h" +#include "btree_io.h" +#include "btree_journal_iter.h" +#include "btree_node_scan.h" +#include "btree_update_interior.h" +#include "buckets.h" +#include "error.h" +#include "journal_io.h" +#include "recovery_passes.h" + +#include +#include + +struct find_btree_nodes_worker { + struct closure *cl; + struct find_btree_nodes *f; + struct bch_dev *ca; +}; + +static void found_btree_node_to_text(struct printbuf *out, struct bch_fs *c, const struct found_btree_node *n) +{ + prt_printf(out, "%s l=%u seq=%u cookie=%llx ", bch2_btree_id_str(n->btree_id), n->level, n->seq, n->cookie); + bch2_bpos_to_text(out, n->min_key); + prt_str(out, "-"); + bch2_bpos_to_text(out, n->max_key); + + if (n->range_updated) + prt_str(out, " range updated"); + if (n->overwritten) + prt_str(out, " overwritten"); + + for (unsigned i = 0; i < n->nr_ptrs; i++) { + prt_char(out, ' '); + bch2_extent_ptr_to_text(out, c, n->ptrs + i); + } +} + +static void found_btree_nodes_to_text(struct printbuf *out, struct bch_fs *c, found_btree_nodes nodes) +{ + printbuf_indent_add(out, 2); + darray_for_each(nodes, i) { + found_btree_node_to_text(out, c, i); + prt_newline(out); + } + printbuf_indent_sub(out, 2); +} + +static void found_btree_node_to_key(struct bkey_i *k, const struct found_btree_node *f) +{ + struct bkey_i_btree_ptr_v2 *bp = bkey_btree_ptr_v2_init(k); + + set_bkey_val_u64s(&bp->k, sizeof(struct bch_btree_ptr_v2) / sizeof(u64) + f->nr_ptrs); + bp->k.p = f->max_key; + bp->v.seq = cpu_to_le64(f->cookie); + bp->v.sectors_written = 0; + bp->v.flags = 0; + bp->v.min_key = f->min_key; + SET_BTREE_PTR_RANGE_UPDATED(&bp->v, f->range_updated); + memcpy(bp->v.start, f->ptrs, sizeof(struct bch_extent_ptr) * f->nr_ptrs); +} + +static bool found_btree_node_is_readable(struct btree_trans *trans, + const struct found_btree_node *f) +{ + struct { __BKEY_PADDED(k, BKEY_BTREE_PTR_VAL_U64s_MAX); } k; + + found_btree_node_to_key(&k.k, f); + + struct btree *b = bch2_btree_node_get_noiter(trans, &k.k, f->btree_id, f->level, false); + bool ret = !IS_ERR_OR_NULL(b); + if (ret) + six_unlock_read(&b->c.lock); + + /* + * We might update this node's range; if that happens, we need the node + * to be re-read so the read path can trim keys that are no longer in + * this node + */ + if (b != btree_node_root(trans->c, b)) + bch2_btree_node_evict(trans, &k.k); + return ret; +} + +static int found_btree_node_cmp_cookie(const void *_l, const void *_r) +{ + const struct found_btree_node *l = _l; + const struct found_btree_node *r = _r; + + return cmp_int(l->btree_id, r->btree_id) ?: + cmp_int(l->level, r->level) ?: + cmp_int(l->cookie, r->cookie); +} + +/* + * Given two found btree nodes, if their sequence numbers are equal, take the + * one that's readable: + */ +static int found_btree_node_cmp_time(const struct found_btree_node *l, + const struct found_btree_node *r) +{ + return cmp_int(l->seq, r->seq); +} + +static int found_btree_node_cmp_pos(const void *_l, const void *_r) +{ + const struct found_btree_node *l = _l; + const struct found_btree_node *r = _r; + + return cmp_int(l->btree_id, r->btree_id) ?: + -cmp_int(l->level, r->level) ?: + bpos_cmp(l->min_key, r->min_key) ?: + -found_btree_node_cmp_time(l, r); +} + +static void try_read_btree_node(struct find_btree_nodes *f, struct bch_dev *ca, + struct bio *bio, struct btree_node *bn, u64 offset) +{ + struct bch_fs *c = container_of(f, struct bch_fs, found_btree_nodes); + + bio_reset(bio, ca->disk_sb.bdev, REQ_OP_READ); + bio->bi_iter.bi_sector = offset; + bch2_bio_map(bio, bn, PAGE_SIZE); + + submit_bio_wait(bio); + if (bch2_dev_io_err_on(bio->bi_status, ca, BCH_MEMBER_ERROR_read, + "IO error in try_read_btree_node() at %llu: %s", + offset, bch2_blk_status_to_str(bio->bi_status))) + return; + + if (le64_to_cpu(bn->magic) != bset_magic(c)) + return; + + rcu_read_lock(); + struct found_btree_node n = { + .btree_id = BTREE_NODE_ID(bn), + .level = BTREE_NODE_LEVEL(bn), + .seq = BTREE_NODE_SEQ(bn), + .cookie = le64_to_cpu(bn->keys.seq), + .min_key = bn->min_key, + .max_key = bn->max_key, + .nr_ptrs = 1, + .ptrs[0].type = 1 << BCH_EXTENT_ENTRY_ptr, + .ptrs[0].offset = offset, + .ptrs[0].dev = ca->dev_idx, + .ptrs[0].gen = *bucket_gen(ca, sector_to_bucket(ca, offset)), + }; + rcu_read_unlock(); + + if (bch2_trans_run(c, found_btree_node_is_readable(trans, &n))) { + mutex_lock(&f->lock); + if (BSET_BIG_ENDIAN(&bn->keys) != CPU_BIG_ENDIAN) { + bch_err(c, "try_read_btree_node() can't handle endian conversion"); + f->ret = -EINVAL; + goto unlock; + } + + if (darray_push(&f->nodes, n)) + f->ret = -ENOMEM; +unlock: + mutex_unlock(&f->lock); + } +} + +static int read_btree_nodes_worker(void *p) +{ + struct find_btree_nodes_worker *w = p; + struct bch_fs *c = container_of(w->f, struct bch_fs, found_btree_nodes); + struct bch_dev *ca = w->ca; + void *buf = (void *) __get_free_page(GFP_KERNEL); + struct bio *bio = bio_alloc(NULL, 1, 0, GFP_KERNEL); + unsigned long last_print = jiffies; + + if (!buf || !bio) { + bch_err(c, "read_btree_nodes_worker: error allocating bio/buf"); + w->f->ret = -ENOMEM; + goto err; + } + + for (u64 bucket = ca->mi.first_bucket; bucket < ca->mi.nbuckets; bucket++) + for (unsigned bucket_offset = 0; + bucket_offset + btree_sectors(c) <= ca->mi.bucket_size; + bucket_offset += btree_sectors(c)) { + if (time_after(jiffies, last_print + HZ * 30)) { + u64 cur_sector = bucket * ca->mi.bucket_size + bucket_offset; + u64 end_sector = ca->mi.nbuckets * ca->mi.bucket_size; + + bch_info(ca, "%s: %2u%% done", __func__, + (unsigned) div64_u64(cur_sector * 100, end_sector)); + last_print = jiffies; + } + + try_read_btree_node(w->f, ca, bio, buf, + bucket * ca->mi.bucket_size + bucket_offset); + } +err: + bio_put(bio); + free_page((unsigned long) buf); + percpu_ref_get(&ca->io_ref); + closure_put(w->cl); + kfree(w); + return 0; +} + +static int read_btree_nodes(struct find_btree_nodes *f) +{ + struct bch_fs *c = container_of(f, struct bch_fs, found_btree_nodes); + struct closure cl; + int ret = 0; + + closure_init_stack(&cl); + + for_each_online_member(c, ca) { + struct find_btree_nodes_worker *w = kmalloc(sizeof(*w), GFP_KERNEL); + struct task_struct *t; + + if (!w) { + percpu_ref_put(&ca->io_ref); + ret = -ENOMEM; + goto err; + } + + percpu_ref_get(&ca->io_ref); + closure_get(&cl); + w->cl = &cl; + w->f = f; + w->ca = ca; + + t = kthread_run(read_btree_nodes_worker, w, "read_btree_nodes/%s", ca->name); + ret = IS_ERR_OR_NULL(t); + if (ret) { + percpu_ref_put(&ca->io_ref); + closure_put(&cl); + f->ret = ret; + bch_err(c, "error starting kthread: %i", ret); + break; + } + } +err: + closure_sync(&cl); + return f->ret ?: ret; +} + +static void bubble_up(struct found_btree_node *n, struct found_btree_node *end) +{ + while (n + 1 < end && + found_btree_node_cmp_pos(n, n + 1) > 0) { + swap(n[0], n[1]); + n++; + } +} + +static int handle_overwrites(struct bch_fs *c, + struct found_btree_node *start, + struct found_btree_node *end) +{ + struct found_btree_node *n; +again: + for (n = start + 1; + n < end && + n->btree_id == start->btree_id && + n->level == start->level && + bpos_lt(n->min_key, start->max_key); + n++) { + int cmp = found_btree_node_cmp_time(start, n); + + if (cmp > 0) { + if (bpos_cmp(start->max_key, n->max_key) >= 0) + n->overwritten = true; + else { + n->range_updated = true; + n->min_key = bpos_successor(start->max_key); + n->range_updated = true; + bubble_up(n, end); + goto again; + } + } else if (cmp < 0) { + BUG_ON(bpos_cmp(n->min_key, start->min_key) <= 0); + + start->max_key = bpos_predecessor(n->min_key); + start->range_updated = true; + } else { + struct printbuf buf = PRINTBUF; + + prt_str(&buf, "overlapping btree nodes with same seq! halting\n "); + found_btree_node_to_text(&buf, c, start); + prt_str(&buf, "\n "); + found_btree_node_to_text(&buf, c, n); + bch_err(c, "%s", buf.buf); + printbuf_exit(&buf); + return -1; + } + } + + return 0; +} + +int bch2_scan_for_btree_nodes(struct bch_fs *c) +{ + struct find_btree_nodes *f = &c->found_btree_nodes; + struct printbuf buf = PRINTBUF; + size_t dst; + int ret = 0; + + if (f->nodes.nr) + return 0; + + mutex_init(&f->lock); + + ret = read_btree_nodes(f); + if (ret) + return ret; + + if (!f->nodes.nr) { + bch_err(c, "%s: no btree nodes found", __func__); + ret = -EINVAL; + goto err; + } + + if (0 && c->opts.verbose) { + printbuf_reset(&buf); + prt_printf(&buf, "%s: nodes found:\n", __func__); + found_btree_nodes_to_text(&buf, c, f->nodes); + bch2_print_string_as_lines(KERN_INFO, buf.buf); + } + + sort(f->nodes.data, f->nodes.nr, sizeof(f->nodes.data[0]), found_btree_node_cmp_cookie, NULL); + + dst = 0; + darray_for_each(f->nodes, i) { + struct found_btree_node *prev = dst ? f->nodes.data + dst - 1 : NULL; + + if (prev && + prev->cookie == i->cookie) { + if (prev->nr_ptrs == ARRAY_SIZE(prev->ptrs)) { + bch_err(c, "%s: found too many replicas for btree node", __func__); + ret = -EINVAL; + goto err; + } + prev->ptrs[prev->nr_ptrs++] = i->ptrs[0]; + } else { + f->nodes.data[dst++] = *i; + } + } + f->nodes.nr = dst; + + sort(f->nodes.data, f->nodes.nr, sizeof(f->nodes.data[0]), found_btree_node_cmp_pos, NULL); + + if (0 && c->opts.verbose) { + printbuf_reset(&buf); + prt_printf(&buf, "%s: nodes after merging replicas:\n", __func__); + found_btree_nodes_to_text(&buf, c, f->nodes); + bch2_print_string_as_lines(KERN_INFO, buf.buf); + } + + dst = 0; + darray_for_each(f->nodes, i) { + if (i->overwritten) + continue; + + ret = handle_overwrites(c, i, &darray_top(f->nodes)); + if (ret) + goto err; + + BUG_ON(i->overwritten); + f->nodes.data[dst++] = *i; + } + f->nodes.nr = dst; + + if (c->opts.verbose) { + printbuf_reset(&buf); + prt_printf(&buf, "%s: nodes found after overwrites:\n", __func__); + found_btree_nodes_to_text(&buf, c, f->nodes); + bch2_print_string_as_lines(KERN_INFO, buf.buf); + } + + eytzinger0_sort(f->nodes.data, f->nodes.nr, sizeof(f->nodes.data[0]), found_btree_node_cmp_pos, NULL); +err: + printbuf_exit(&buf); + return ret; +} + +static int found_btree_node_range_start_cmp(const void *_l, const void *_r) +{ + const struct found_btree_node *l = _l; + const struct found_btree_node *r = _r; + + return cmp_int(l->btree_id, r->btree_id) ?: + -cmp_int(l->level, r->level) ?: + bpos_cmp(l->max_key, r->min_key); +} + +#define for_each_found_btree_node_in_range(_f, _search, _idx) \ + for (size_t _idx = eytzinger0_find_gt((_f)->nodes.data, (_f)->nodes.nr, \ + sizeof((_f)->nodes.data[0]), \ + found_btree_node_range_start_cmp, &search); \ + _idx < (_f)->nodes.nr && \ + (_f)->nodes.data[_idx].btree_id == _search.btree_id && \ + (_f)->nodes.data[_idx].level == _search.level && \ + bpos_lt((_f)->nodes.data[_idx].min_key, _search.max_key); \ + _idx = eytzinger0_next(_idx, (_f)->nodes.nr)) + +bool bch2_btree_node_is_stale(struct bch_fs *c, struct btree *b) +{ + struct find_btree_nodes *f = &c->found_btree_nodes; + + struct found_btree_node search = { + .btree_id = b->c.btree_id, + .level = b->c.level, + .min_key = b->data->min_key, + .max_key = b->key.k.p, + }; + + for_each_found_btree_node_in_range(f, search, idx) + if (f->nodes.data[idx].seq > BTREE_NODE_SEQ(b->data)) + return true; + return false; +} + +bool bch2_btree_has_scanned_nodes(struct bch_fs *c, enum btree_id btree) +{ + struct found_btree_node search = { + .btree_id = btree, + .level = 0, + .min_key = POS_MIN, + .max_key = SPOS_MAX, + }; + + for_each_found_btree_node_in_range(&c->found_btree_nodes, search, idx) + return true; + return false; +} + +int bch2_get_scanned_nodes(struct bch_fs *c, enum btree_id btree, + unsigned level, struct bpos node_min, struct bpos node_max) +{ + struct find_btree_nodes *f = &c->found_btree_nodes; + + int ret = bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_scan_for_btree_nodes); + if (ret) + return ret; + + if (c->opts.verbose) { + struct printbuf buf = PRINTBUF; + + prt_printf(&buf, "recovering %s l=%u ", bch2_btree_id_str(btree), level); + bch2_bpos_to_text(&buf, node_min); + prt_str(&buf, " - "); + bch2_bpos_to_text(&buf, node_max); + + bch_info(c, "%s(): %s", __func__, buf.buf); + printbuf_exit(&buf); + } + + struct found_btree_node search = { + .btree_id = btree, + .level = level, + .min_key = node_min, + .max_key = node_max, + }; + + for_each_found_btree_node_in_range(f, search, idx) { + struct found_btree_node n = f->nodes.data[idx]; + + n.range_updated |= bpos_lt(n.min_key, node_min); + n.min_key = bpos_max(n.min_key, node_min); + + n.range_updated |= bpos_gt(n.max_key, node_max); + n.max_key = bpos_min(n.max_key, node_max); + + struct { __BKEY_PADDED(k, BKEY_BTREE_PTR_VAL_U64s_MAX); } tmp; + + found_btree_node_to_key(&tmp.k, &n); + + struct printbuf buf = PRINTBUF; + bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&tmp.k)); + bch_verbose(c, "%s(): recovering %s", __func__, buf.buf); + printbuf_exit(&buf); + + BUG_ON(bch2_bkey_invalid(c, bkey_i_to_s_c(&tmp.k), BKEY_TYPE_btree, 0, NULL)); + + ret = bch2_journal_key_insert(c, btree, level + 1, &tmp.k); + if (ret) + return ret; + } + + return 0; +} + +void bch2_find_btree_nodes_exit(struct find_btree_nodes *f) +{ + darray_exit(&f->nodes); +} diff --git a/fs/bcachefs/btree_node_scan.h b/fs/bcachefs/btree_node_scan.h new file mode 100644 index 000000000000..08687b209787 --- /dev/null +++ b/fs/bcachefs/btree_node_scan.h @@ -0,0 +1,11 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +#ifndef _BCACHEFS_BTREE_NODE_SCAN_H +#define _BCACHEFS_BTREE_NODE_SCAN_H + +int bch2_scan_for_btree_nodes(struct bch_fs *); +bool bch2_btree_node_is_stale(struct bch_fs *, struct btree *); +bool bch2_btree_has_scanned_nodes(struct bch_fs *, enum btree_id); +int bch2_get_scanned_nodes(struct bch_fs *, enum btree_id, unsigned, struct bpos, struct bpos); +void bch2_find_btree_nodes_exit(struct find_btree_nodes *); + +#endif /* _BCACHEFS_BTREE_NODE_SCAN_H */ diff --git a/fs/bcachefs/btree_node_scan_types.h b/fs/bcachefs/btree_node_scan_types.h new file mode 100644 index 000000000000..abb7b27d556a --- /dev/null +++ b/fs/bcachefs/btree_node_scan_types.h @@ -0,0 +1,30 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +#ifndef _BCACHEFS_BTREE_NODE_SCAN_TYPES_H +#define _BCACHEFS_BTREE_NODE_SCAN_TYPES_H + +#include "darray.h" + +struct found_btree_node { + bool range_updated:1; + bool overwritten:1; + u8 btree_id; + u8 level; + u32 seq; + u64 cookie; + + struct bpos min_key; + struct bpos max_key; + + unsigned nr_ptrs; + struct bch_extent_ptr ptrs[BCH_REPLICAS_MAX]; +}; + +typedef DARRAY(struct found_btree_node) found_btree_nodes; + +struct find_btree_nodes { + int ret; + struct mutex lock; + found_btree_nodes nodes; +}; + +#endif /* _BCACHEFS_BTREE_NODE_SCAN_TYPES_H */ diff --git a/fs/bcachefs/extents.c b/fs/bcachefs/extents.c index b2432d88cda6..0e3ca99fbd2d 100644 --- a/fs/bcachefs/extents.c +++ b/fs/bcachefs/extents.c @@ -978,6 +978,31 @@ bool bch2_extent_normalize(struct bch_fs *c, struct bkey_s k) return bkey_deleted(k.k); } +void bch2_extent_ptr_to_text(struct printbuf *out, struct bch_fs *c, const struct bch_extent_ptr *ptr) +{ + struct bch_dev *ca = c && ptr->dev < c->sb.nr_devices && c->devs[ptr->dev] + ? bch_dev_bkey_exists(c, ptr->dev) + : NULL; + + if (!ca) { + prt_printf(out, "ptr: %u:%llu gen %u%s", ptr->dev, + (u64) ptr->offset, ptr->gen, + ptr->cached ? " cached" : ""); + } else { + u32 offset; + u64 b = sector_to_bucket_and_offset(ca, ptr->offset, &offset); + + prt_printf(out, "ptr: %u:%llu:%u gen %u", + ptr->dev, b, offset, ptr->gen); + if (ptr->cached) + prt_str(out, " cached"); + if (ptr->unwritten) + prt_str(out, " unwritten"); + if (ca && ptr_stale(ca, ptr)) + prt_printf(out, " stale"); + } +} + void bch2_bkey_ptrs_to_text(struct printbuf *out, struct bch_fs *c, struct bkey_s_c k) { @@ -993,31 +1018,10 @@ void bch2_bkey_ptrs_to_text(struct printbuf *out, struct bch_fs *c, prt_printf(out, " "); switch (__extent_entry_type(entry)) { - case BCH_EXTENT_ENTRY_ptr: { - const struct bch_extent_ptr *ptr = entry_to_ptr(entry); - struct bch_dev *ca = c && ptr->dev < c->sb.nr_devices && c->devs[ptr->dev] - ? bch_dev_bkey_exists(c, ptr->dev) - : NULL; - - if (!ca) { - prt_printf(out, "ptr: %u:%llu gen %u%s", ptr->dev, - (u64) ptr->offset, ptr->gen, - ptr->cached ? " cached" : ""); - } else { - u32 offset; - u64 b = sector_to_bucket_and_offset(ca, ptr->offset, &offset); - - prt_printf(out, "ptr: %u:%llu:%u gen %u", - ptr->dev, b, offset, ptr->gen); - if (ptr->cached) - prt_str(out, " cached"); - if (ptr->unwritten) - prt_str(out, " unwritten"); - if (ca && ptr_stale(ca, ptr)) - prt_printf(out, " stale"); - } + case BCH_EXTENT_ENTRY_ptr: + bch2_extent_ptr_to_text(out, c, entry_to_ptr(entry)); break; - } + case BCH_EXTENT_ENTRY_crc32: case BCH_EXTENT_ENTRY_crc64: case BCH_EXTENT_ENTRY_crc128: { diff --git a/fs/bcachefs/extents.h b/fs/bcachefs/extents.h index 3fd0169b98c1..528e817eacbd 100644 --- a/fs/bcachefs/extents.h +++ b/fs/bcachefs/extents.h @@ -676,6 +676,7 @@ bch2_extent_has_ptr(struct bkey_s_c, struct extent_ptr_decoded, struct bkey_s); void bch2_extent_ptr_set_cached(struct bkey_s, struct bch_extent_ptr *); bool bch2_extent_normalize(struct bch_fs *, struct bkey_s); +void bch2_extent_ptr_to_text(struct printbuf *out, struct bch_fs *, const struct bch_extent_ptr *); void bch2_bkey_ptrs_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c); int bch2_bkey_ptrs_invalid(struct bch_fs *, struct bkey_s_c, diff --git a/fs/bcachefs/opts.h b/fs/bcachefs/opts.h index 084247c13bd8..1ac4135cca1c 100644 --- a/fs/bcachefs/opts.h +++ b/fs/bcachefs/opts.h @@ -368,11 +368,11 @@ enum fsck_err_opts { OPT_STR_NOLIMIT(bch2_recovery_passes), \ BCH2_NO_SB_OPT, 0, \ NULL, "Exit recovery after specified pass") \ - x(keep_journal, u8, \ + x(retain_recovery_info, u8, \ 0, \ OPT_BOOL(), \ BCH2_NO_SB_OPT, false, \ - NULL, "Don't free journal entries/keys after startup")\ + NULL, "Don't free journal entries/keys, scanned btree nodes after startup")\ x(read_entire_journal, u8, \ 0, \ OPT_BOOL(), \ diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c index f234de0ac834..24671020f22b 100644 --- a/fs/bcachefs/recovery.c +++ b/fs/bcachefs/recovery.c @@ -4,6 +4,7 @@ #include "alloc_background.h" #include "bkey_buf.h" #include "btree_journal_iter.h" +#include "btree_node_scan.h" #include "btree_update.h" #include "btree_update_interior.h" #include "btree_io.h" @@ -271,7 +272,7 @@ int bch2_journal_replay(struct bch_fs *c) bch2_trans_put(trans); trans = NULL; - if (!c->opts.keep_journal && + if (!c->opts.retain_recovery_info && c->recovery_pass_done >= BCH_RECOVERY_PASS_journal_replay) bch2_journal_keys_put_initial(c); @@ -435,10 +436,9 @@ static int journal_replay_early(struct bch_fs *c, static int read_btree_roots(struct bch_fs *c) { - unsigned i; int ret = 0; - for (i = 0; i < btree_id_nr_alive(c); i++) { + for (unsigned i = 0; i < btree_id_nr_alive(c); i++) { struct btree_root *r = bch2_btree_id_root(c, i); if (!r->alive) @@ -447,33 +447,36 @@ static int read_btree_roots(struct bch_fs *c) if (btree_id_is_alloc(i) && c->opts.reconstruct_alloc) continue; - if (r->error) { - __fsck_err(c, - btree_id_is_alloc(i) - ? FSCK_CAN_IGNORE : 0, - btree_root_bkey_invalid, - "invalid btree root %s", - bch2_btree_id_str(i)); - if (i == BTREE_ID_alloc) + if (mustfix_fsck_err_on((ret = r->error), + c, btree_root_bkey_invalid, + "invalid btree root %s", + bch2_btree_id_str(i)) || + mustfix_fsck_err_on((ret = r->error = bch2_btree_root_read(c, i, &r->key, r->level)), + c, btree_root_read_error, + "error reading btree root %s l=%u: %s", + bch2_btree_id_str(i), r->level, bch2_err_str(ret))) { + if (btree_id_is_alloc(i)) { + c->recovery_passes_explicit |= BIT_ULL(BCH_RECOVERY_PASS_check_allocations); + c->recovery_passes_explicit |= BIT_ULL(BCH_RECOVERY_PASS_check_alloc_info); + c->recovery_passes_explicit |= BIT_ULL(BCH_RECOVERY_PASS_check_lrus); + c->recovery_passes_explicit |= BIT_ULL(BCH_RECOVERY_PASS_check_extents_to_backpointers); + c->recovery_passes_explicit |= BIT_ULL(BCH_RECOVERY_PASS_check_alloc_to_lru_refs); c->sb.compat &= ~(1ULL << BCH_COMPAT_alloc_info); - } + r->error = 0; + } else if (!(c->recovery_passes_explicit & BIT_ULL(BCH_RECOVERY_PASS_scan_for_btree_nodes))) { + bch_info(c, "will run btree node scan"); + c->recovery_passes_explicit |= BIT_ULL(BCH_RECOVERY_PASS_scan_for_btree_nodes); + c->recovery_passes_explicit |= BIT_ULL(BCH_RECOVERY_PASS_check_topology); + } - ret = bch2_btree_root_read(c, i, &r->key, r->level); - if (ret) { - fsck_err(c, - btree_root_read_error, - "error reading btree root %s", - bch2_btree_id_str(i)); - if (btree_id_is_alloc(i)) - c->sb.compat &= ~(1ULL << BCH_COMPAT_alloc_info); ret = 0; } } - for (i = 0; i < BTREE_ID_NR; i++) { + for (unsigned i = 0; i < BTREE_ID_NR; i++) { struct btree_root *r = bch2_btree_id_root(c, i); - if (!r->b) { + if (!r->b && !r->error) { r->alive = false; r->level = 0; bch2_btree_root_alloc_fake(c, i, 0); @@ -653,7 +656,7 @@ int bch2_fs_recovery(struct bch_fs *c) goto err; } - if (!c->sb.clean || c->opts.fsck || c->opts.keep_journal) { + if (!c->sb.clean || c->opts.fsck || c->opts.retain_recovery_info) { struct genradix_iter iter; struct journal_replay **i; @@ -883,9 +886,10 @@ use_clean: out: bch2_flush_fsck_errs(c); - if (!c->opts.keep_journal && - test_bit(JOURNAL_REPLAY_DONE, &c->journal.flags)) + if (!c->opts.retain_recovery_info) { bch2_journal_keys_put_initial(c); + bch2_find_btree_nodes_exit(&c->found_btree_nodes); + } kfree(clean); if (!ret && diff --git a/fs/bcachefs/recovery_passes.c b/fs/bcachefs/recovery_passes.c index b066c5155b74..cb501460d615 100644 --- a/fs/bcachefs/recovery_passes.c +++ b/fs/bcachefs/recovery_passes.c @@ -4,6 +4,7 @@ #include "alloc_background.h" #include "backpointers.h" #include "btree_gc.h" +#include "btree_node_scan.h" #include "ec.h" #include "fsck.h" #include "inode.h" diff --git a/fs/bcachefs/recovery_passes_types.h b/fs/bcachefs/recovery_passes_types.h index f30521285706..840542cfd65b 100644 --- a/fs/bcachefs/recovery_passes_types.h +++ b/fs/bcachefs/recovery_passes_types.h @@ -13,6 +13,7 @@ * must never change: */ #define BCH_RECOVERY_PASSES() \ + x(scan_for_btree_nodes, 37, 0) \ x(check_topology, 4, 0) \ x(alloc_read, 0, PASS_ALWAYS) \ x(stripes_read, 1, PASS_ALWAYS) \ diff --git a/fs/bcachefs/super.c b/fs/bcachefs/super.c index bc026a77eb99..ed63018f21be 100644 --- a/fs/bcachefs/super.c +++ b/fs/bcachefs/super.c @@ -15,6 +15,7 @@ #include "btree_gc.h" #include "btree_journal_iter.h" #include "btree_key_cache.h" +#include "btree_node_scan.h" #include "btree_update_interior.h" #include "btree_io.h" #include "btree_write_buffer.h" @@ -536,6 +537,7 @@ static void __bch2_fs_free(struct bch_fs *c) for (i = 0; i < BCH_TIME_STAT_NR; i++) bch2_time_stats_exit(&c->times[i]); + bch2_find_btree_nodes_exit(&c->found_btree_nodes); bch2_free_pending_node_rewrites(c); bch2_fs_sb_errors_exit(c); bch2_fs_counters_exit(c); @@ -560,6 +562,7 @@ static void __bch2_fs_free(struct bch_fs *c) bch2_io_clock_exit(&c->io_clock[READ]); bch2_fs_compress_exit(c); bch2_journal_keys_put_initial(c); + bch2_find_btree_nodes_exit(&c->found_btree_nodes); BUG_ON(atomic_read(&c->journal_keys.ref)); bch2_fs_btree_write_buffer_exit(c); percpu_free_rwsem(&c->mark_lock); -- cgit From 43f5ea4646b2271a9a5af3729dfdf644d69b3282 Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Sat, 16 Mar 2024 22:45:30 -0400 Subject: bcachefs: Topology repair now uses nodes found by scanning to fill holes With the new btree node scan code, we can now recover from corrupt btree roots - simply create a new fake root at depth 1, and then insert all the leaves we found. If the root wasn't corrupt but there's corruption elsewhere in the btree, we can fill in holes as needed with the newest version of a given node(s) from the scan; we also check if a given btree node is older than what we found from the scan. Signed-off-by: Kent Overstreet --- fs/bcachefs/btree_gc.c | 303 +++++++++++++++++++++++++++--------------- fs/bcachefs/sb-errors_types.h | 3 +- 2 files changed, 199 insertions(+), 107 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/btree_gc.c b/fs/bcachefs/btree_gc.c index 9db9c8a5beaa..6280da1244b5 100644 --- a/fs/bcachefs/btree_gc.c +++ b/fs/bcachefs/btree_gc.c @@ -13,6 +13,7 @@ #include "btree_journal_iter.h" #include "btree_key_cache.h" #include "btree_locking.h" +#include "btree_node_scan.h" #include "btree_update_interior.h" #include "btree_io.h" #include "btree_gc.h" @@ -41,6 +42,7 @@ #define DROP_THIS_NODE 10 #define DROP_PREV_NODE 11 +#define DID_FILL_FROM_SCAN 12 static struct bkey_s unsafe_bkey_s_c_to_s(struct bkey_s_c k) { @@ -129,6 +131,17 @@ static int set_node_min(struct bch_fs *c, struct btree *b, struct bpos new_min) struct bkey_i_btree_ptr_v2 *new; int ret; + if (c->opts.verbose) { + struct printbuf buf = PRINTBUF; + + bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&b->key)); + prt_str(&buf, " -> "); + bch2_bpos_to_text(&buf, new_min); + + bch_info(c, "%s(): %s", __func__, buf.buf); + printbuf_exit(&buf); + } + new = kmalloc_array(BKEY_BTREE_PTR_U64s_MAX, sizeof(u64), GFP_KERNEL); if (!new) return -BCH_ERR_ENOMEM_gc_repair_key; @@ -154,6 +167,17 @@ static int set_node_max(struct bch_fs *c, struct btree *b, struct bpos new_max) struct bkey_i_btree_ptr_v2 *new; int ret; + if (c->opts.verbose) { + struct printbuf buf = PRINTBUF; + + bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&b->key)); + prt_str(&buf, " -> "); + bch2_bpos_to_text(&buf, new_max); + + bch_info(c, "%s(): %s", __func__, buf.buf); + printbuf_exit(&buf); + } + ret = bch2_journal_key_delete(c, b->c.btree_id, b->c.level + 1, b->key.k.p); if (ret) return ret; @@ -185,127 +209,138 @@ static int set_node_max(struct bch_fs *c, struct btree *b, struct bpos new_max) return 0; } -static int btree_repair_node_boundaries(struct bch_fs *c, struct btree *b, - struct btree *prev, struct btree *cur) +static int btree_check_node_boundaries(struct bch_fs *c, struct btree *b, + struct btree *prev, struct btree *cur, + struct bpos *pulled_from_scan) { struct bpos expected_start = !prev ? b->data->min_key : bpos_successor(prev->key.k.p); - struct printbuf buf1 = PRINTBUF, buf2 = PRINTBUF; + struct printbuf buf = PRINTBUF; int ret = 0; - if (!prev) { - prt_printf(&buf1, "start of node: "); - bch2_bpos_to_text(&buf1, b->data->min_key); - } else { - bch2_bkey_val_to_text(&buf1, c, bkey_i_to_s_c(&prev->key)); + BUG_ON(b->key.k.type == KEY_TYPE_btree_ptr_v2 && + !bpos_eq(bkey_i_to_btree_ptr_v2(&b->key)->v.min_key, + b->data->min_key)); + + if (bpos_eq(expected_start, cur->data->min_key)) + return 0; + + prt_printf(&buf, " at btree %s level %u:\n parent: ", + bch2_btree_id_str(b->c.btree_id), b->c.level); + bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&b->key)); + + if (prev) { + prt_printf(&buf, "\n prev: "); + bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&prev->key)); } - bch2_bkey_val_to_text(&buf2, c, bkey_i_to_s_c(&cur->key)); - - if (prev && - bpos_gt(expected_start, cur->data->min_key) && - BTREE_NODE_SEQ(cur->data) > BTREE_NODE_SEQ(prev->data)) { - /* cur overwrites prev: */ - - if (mustfix_fsck_err_on(bpos_ge(prev->data->min_key, - cur->data->min_key), c, - btree_node_topology_overwritten_by_next_node, - "btree node overwritten by next node at btree %s level %u:\n" - " node %s\n" - " next %s", - bch2_btree_id_str(b->c.btree_id), b->c.level, - buf1.buf, buf2.buf)) { - ret = DROP_PREV_NODE; - goto out; - } + prt_str(&buf, "\n next: "); + bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&cur->key)); - if (mustfix_fsck_err_on(!bpos_eq(prev->key.k.p, - bpos_predecessor(cur->data->min_key)), c, - btree_node_topology_bad_max_key, - "btree node with incorrect max_key at btree %s level %u:\n" - " node %s\n" - " next %s", - bch2_btree_id_str(b->c.btree_id), b->c.level, - buf1.buf, buf2.buf)) - ret = set_node_max(c, prev, - bpos_predecessor(cur->data->min_key)); - } else { - /* prev overwrites cur: */ - - if (mustfix_fsck_err_on(bpos_ge(expected_start, - cur->data->max_key), c, - btree_node_topology_overwritten_by_prev_node, - "btree node overwritten by prev node at btree %s level %u:\n" - " prev %s\n" - " node %s", - bch2_btree_id_str(b->c.btree_id), b->c.level, - buf1.buf, buf2.buf)) { - ret = DROP_THIS_NODE; - goto out; - } + if (bpos_lt(expected_start, cur->data->min_key)) { /* gap */ + if (b->c.level == 1 && + bpos_lt(*pulled_from_scan, cur->data->min_key)) { + ret = bch2_get_scanned_nodes(c, b->c.btree_id, 0, + expected_start, + bpos_predecessor(cur->data->min_key)); + if (ret) + goto err; - if (mustfix_fsck_err_on(!bpos_eq(expected_start, cur->data->min_key), c, - btree_node_topology_bad_min_key, - "btree node with incorrect min_key at btree %s level %u:\n" - " prev %s\n" - " node %s", - bch2_btree_id_str(b->c.btree_id), b->c.level, - buf1.buf, buf2.buf)) - ret = set_node_min(c, cur, expected_start); + *pulled_from_scan = cur->data->min_key; + ret = DID_FILL_FROM_SCAN; + } else { + if (mustfix_fsck_err(c, btree_node_topology_bad_min_key, + "btree node with incorrect min_key%s", buf.buf)) + ret = set_node_min(c, cur, expected_start); + } + } else { /* overlap */ + if (prev && BTREE_NODE_SEQ(cur->data) > BTREE_NODE_SEQ(prev->data)) { /* cur overwrites prev */ + if (bpos_ge(prev->data->min_key, cur->data->min_key)) { /* fully? */ + if (mustfix_fsck_err(c, btree_node_topology_overwritten_by_next_node, + "btree node overwritten by next node%s", buf.buf)) + ret = DROP_PREV_NODE; + } else { + if (mustfix_fsck_err(c, btree_node_topology_bad_max_key, + "btree node with incorrect max_key%s", buf.buf)) + ret = set_node_max(c, prev, + bpos_predecessor(cur->data->min_key)); + } + } else { + if (bpos_ge(expected_start, cur->data->max_key)) { /* fully? */ + if (mustfix_fsck_err(c, btree_node_topology_overwritten_by_prev_node, + "btree node overwritten by prev node%s", buf.buf)) + ret = DROP_THIS_NODE; + } else { + if (mustfix_fsck_err(c, btree_node_topology_bad_min_key, + "btree node with incorrect min_key%s", buf.buf)) + ret = set_node_min(c, cur, expected_start); + } + } } -out: +err: fsck_err: - printbuf_exit(&buf2); - printbuf_exit(&buf1); + printbuf_exit(&buf); return ret; } static int btree_repair_node_end(struct bch_fs *c, struct btree *b, - struct btree *child) + struct btree *child, struct bpos *pulled_from_scan) { - struct printbuf buf1 = PRINTBUF, buf2 = PRINTBUF; + struct printbuf buf = PRINTBUF; int ret = 0; - bch2_bkey_val_to_text(&buf1, c, bkey_i_to_s_c(&child->key)); - bch2_bpos_to_text(&buf2, b->key.k.p); + if (bpos_eq(child->key.k.p, b->key.k.p)) + return 0; - if (mustfix_fsck_err_on(!bpos_eq(child->key.k.p, b->key.k.p), c, - btree_node_topology_bad_max_key, - "btree node with incorrect max_key at btree %s level %u:\n" - " %s\n" - " expected %s", - bch2_btree_id_str(b->c.btree_id), b->c.level, - buf1.buf, buf2.buf)) { - ret = set_node_max(c, child, b->key.k.p); - if (ret) - goto err; + prt_printf(&buf, "at btree %s level %u:\n parent: ", + bch2_btree_id_str(b->c.btree_id), b->c.level); + bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&b->key)); + + prt_str(&buf, "\n child: "); + bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&child->key)); + + if (mustfix_fsck_err(c, btree_node_topology_bad_max_key, + "btree node with incorrect max_key%s", buf.buf)) { + if (b->c.level == 1 && + bpos_lt(*pulled_from_scan, b->key.k.p)) { + ret = bch2_get_scanned_nodes(c, b->c.btree_id, 0, + bpos_successor(child->key.k.p), b->key.k.p); + if (ret) + goto err; + + *pulled_from_scan = b->key.k.p; + ret = DID_FILL_FROM_SCAN; + } else { + ret = set_node_max(c, child, b->key.k.p); + } } err: fsck_err: - printbuf_exit(&buf2); - printbuf_exit(&buf1); + printbuf_exit(&buf); return ret; } -static int bch2_btree_repair_topology_recurse(struct btree_trans *trans, struct btree *b) +static int bch2_btree_repair_topology_recurse(struct btree_trans *trans, struct btree *b, + struct bpos *pulled_from_scan) { struct bch_fs *c = trans->c; struct btree_and_journal_iter iter; struct bkey_s_c k; struct bkey_buf prev_k, cur_k; struct btree *prev = NULL, *cur = NULL; - bool have_child, dropped_children = false; + bool have_child, new_pass = false; struct printbuf buf = PRINTBUF; int ret = 0; if (!b->c.level) return 0; -again: - prev = NULL; - have_child = dropped_children = false; + bch2_bkey_buf_init(&prev_k); bch2_bkey_buf_init(&cur_k); +again: + cur = prev = NULL; + have_child = new_pass = false; bch2_btree_and_journal_iter_init_node_iter(trans, &iter, b); iter.prefetch = true; @@ -332,9 +367,10 @@ again: b->c.level - 1, buf.buf)) { bch2_btree_node_evict(trans, cur_k.k); - ret = bch2_journal_key_delete(c, b->c.btree_id, - b->c.level, cur_k.k->k.p); cur = NULL; + ret = bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_scan_for_btree_nodes) ?: + bch2_journal_key_delete(c, b->c.btree_id, + b->c.level, cur_k.k->k.p); if (ret) break; continue; @@ -344,7 +380,23 @@ again: if (ret) break; - ret = btree_repair_node_boundaries(c, b, prev, cur); + if (bch2_btree_node_is_stale(c, cur)) { + bch_info(c, "btree node %s older than nodes found by scanning", buf.buf); + six_unlock_read(&cur->c.lock); + bch2_btree_node_evict(trans, cur_k.k); + ret = bch2_journal_key_delete(c, b->c.btree_id, + b->c.level, cur_k.k->k.p); + cur = NULL; + if (ret) + break; + continue; + } + + ret = btree_check_node_boundaries(c, b, prev, cur, pulled_from_scan); + if (ret == DID_FILL_FROM_SCAN) { + new_pass = true; + ret = 0; + } if (ret == DROP_THIS_NODE) { six_unlock_read(&cur->c.lock); @@ -370,8 +422,6 @@ again: break; bch2_btree_and_journal_iter_exit(&iter); - bch2_bkey_buf_exit(&prev_k, c); - bch2_bkey_buf_exit(&cur_k, c); goto again; } else if (ret) break; @@ -383,7 +433,11 @@ again: if (!ret && !IS_ERR_OR_NULL(prev)) { BUG_ON(cur); - ret = btree_repair_node_end(c, b, prev); + ret = btree_repair_node_end(c, b, prev, pulled_from_scan); + if (ret == DID_FILL_FROM_SCAN) { + new_pass = true; + ret = 0; + } } if (!IS_ERR_OR_NULL(prev)) @@ -397,6 +451,10 @@ again: goto err; bch2_btree_and_journal_iter_exit(&iter); + + if (new_pass) + goto again; + bch2_btree_and_journal_iter_init_node_iter(trans, &iter, b); iter.prefetch = true; @@ -413,7 +471,7 @@ again: if (ret) goto err; - ret = bch2_btree_repair_topology_recurse(trans, cur); + ret = bch2_btree_repair_topology_recurse(trans, cur, pulled_from_scan); six_unlock_read(&cur->c.lock); cur = NULL; @@ -421,7 +479,7 @@ again: bch2_btree_node_evict(trans, cur_k.k); ret = bch2_journal_key_delete(c, b->c.btree_id, b->c.level, cur_k.k->k.p); - dropped_children = true; + new_pass = true; } if (ret) @@ -448,12 +506,14 @@ fsck_err: six_unlock_read(&cur->c.lock); bch2_btree_and_journal_iter_exit(&iter); - bch2_bkey_buf_exit(&prev_k, c); - bch2_bkey_buf_exit(&cur_k, c); - if (!ret && dropped_children) + if (!ret && new_pass) goto again; + BUG_ON(!ret && bch2_btree_node_check_topology(trans, b)); + + bch2_bkey_buf_exit(&prev_k, c); + bch2_bkey_buf_exit(&cur_k, c); printbuf_exit(&buf); return ret; } @@ -461,32 +521,63 @@ fsck_err: int bch2_check_topology(struct bch_fs *c) { struct btree_trans *trans = bch2_trans_get(c); - struct btree *b; - unsigned i; + struct bpos pulled_from_scan = POS_MIN; int ret = 0; - for (i = 0; i < btree_id_nr_alive(c) && !ret; i++) { + for (unsigned i = 0; i < btree_id_nr_alive(c) && !ret; i++) { struct btree_root *r = bch2_btree_id_root(c, i); + bool reconstructed_root = false; - if (!r->alive) - continue; + if (r->error) { + ret = bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_scan_for_btree_nodes); + if (ret) + break; +reconstruct_root: + bch_info(c, "btree root %s unreadable, must recover from scan", bch2_btree_id_str(i)); - b = r->b; - if (btree_node_fake(b)) - continue; + r->alive = false; + r->error = 0; + + if (!bch2_btree_has_scanned_nodes(c, i)) { + mustfix_fsck_err(c, btree_root_unreadable_and_scan_found_nothing, + "no nodes found for btree %s, continue?", bch2_btree_id_str(i)); + bch2_btree_root_alloc_fake(c, i, 0); + } else { + bch2_btree_root_alloc_fake(c, i, 1); + ret = bch2_get_scanned_nodes(c, i, 0, POS_MIN, SPOS_MAX); + if (ret) + break; + } + + bch2_shoot_down_journal_keys(c, i, 1, BTREE_MAX_DEPTH, POS_MIN, SPOS_MAX); + reconstructed_root = true; + } + + struct btree *b = r->b; btree_node_lock_nopath_nofail(trans, &b->c, SIX_LOCK_read); - ret = bch2_btree_repair_topology_recurse(trans, b); + ret = bch2_btree_repair_topology_recurse(trans, b, &pulled_from_scan); six_unlock_read(&b->c.lock); if (ret == DROP_THIS_NODE) { - bch_err(c, "empty btree root - repair unimplemented"); - ret = -BCH_ERR_fsck_repair_unimplemented; + bch2_btree_node_hash_remove(&c->btree_cache, b); + mutex_lock(&c->btree_cache.lock); + list_move(&b->list, &c->btree_cache.freeable); + mutex_unlock(&c->btree_cache.lock); + + r->b = NULL; + + if (!reconstructed_root) + goto reconstruct_root; + + bch_err(c, "empty btree root %s", bch2_btree_id_str(i)); + bch2_btree_root_alloc_fake(c, i, 0); + r->alive = false; + ret = 0; } } - +fsck_err: bch2_trans_put(trans); - return ret; } diff --git a/fs/bcachefs/sb-errors_types.h b/fs/bcachefs/sb-errors_types.h index 73e9634df8ff..2fec03a24c95 100644 --- a/fs/bcachefs/sb-errors_types.h +++ b/fs/bcachefs/sb-errors_types.h @@ -267,7 +267,8 @@ x(subvol_unreachable, 259) \ x(btree_node_bkey_bad_u64s, 260) \ x(btree_node_topology_empty_interior_node, 261) \ - x(btree_ptr_v2_min_key_bad, 262) + x(btree_ptr_v2_min_key_bad, 262) \ + x(btree_root_unreadable_and_scan_found_nothing, 263) enum bch_sb_error_id { #define x(t, n) BCH_FSCK_ERR_##t = n, -- cgit From 55936afe11077a84d9e1c5068169af328bbf2811 Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Fri, 15 Mar 2024 23:03:42 -0400 Subject: bcachefs: Flag btrees with missing data We need this to know when we should attempt to reconstruct the snapshots btree Signed-off-by: Kent Overstreet --- fs/bcachefs/bcachefs.h | 1 + fs/bcachefs/bcachefs_format.h | 1 + fs/bcachefs/btree_io.c | 13 +++++++++---- fs/bcachefs/recovery.c | 23 +++++++++++++++++++++++ fs/bcachefs/recovery.h | 2 ++ fs/bcachefs/super-io.c | 9 ++++++++- 6 files changed, 44 insertions(+), 5 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/bcachefs.h b/fs/bcachefs/bcachefs.h index 93a61dbaa3d8..d1a0e5478503 100644 --- a/fs/bcachefs/bcachefs.h +++ b/fs/bcachefs/bcachefs.h @@ -797,6 +797,7 @@ struct bch_fs { u64 features; u64 compat; unsigned long errors_silent[BITS_TO_LONGS(BCH_SB_ERR_MAX)]; + u64 btrees_lost_data; } sb; diff --git a/fs/bcachefs/bcachefs_format.h b/fs/bcachefs/bcachefs_format.h index bff8750ac0d7..63102992d955 100644 --- a/fs/bcachefs/bcachefs_format.h +++ b/fs/bcachefs/bcachefs_format.h @@ -818,6 +818,7 @@ struct bch_sb_field_ext { struct bch_sb_field field; __le64 recovery_passes_required[2]; __le64 errors_silent[8]; + __le64 btrees_lost_data; }; struct bch_sb_field_downgrade_entry { diff --git a/fs/bcachefs/btree_io.c b/fs/bcachefs/btree_io.c index f3f27bb85a5b..d7de82ac3893 100644 --- a/fs/bcachefs/btree_io.c +++ b/fs/bcachefs/btree_io.c @@ -1264,10 +1264,12 @@ out: return retry_read; fsck_err: if (ret == -BCH_ERR_btree_node_read_err_want_retry || - ret == -BCH_ERR_btree_node_read_err_must_retry) + ret == -BCH_ERR_btree_node_read_err_must_retry) { retry_read = 1; - else + } else { set_btree_node_read_error(b); + bch2_btree_lost_data(c, b->c.btree_id); + } goto out; } @@ -1328,6 +1330,7 @@ start: if (!can_retry) { set_btree_node_read_error(b); + bch2_btree_lost_data(c, b->c.btree_id); break; } } @@ -1527,9 +1530,10 @@ fsck_err: ret = -1; } - if (ret) + if (ret) { set_btree_node_read_error(b); - else if (*saw_error) + bch2_btree_lost_data(c, b->c.btree_id); + } else if (*saw_error) bch2_btree_node_rewrite_async(c, b); for (i = 0; i < ra->nr; i++) { @@ -1665,6 +1669,7 @@ void bch2_btree_node_read(struct btree_trans *trans, struct btree *b, bch2_fatal_error(c); set_btree_node_read_error(b); + bch2_btree_lost_data(c, b->c.btree_id); clear_btree_node_read_in_flight(b); wake_up_bit(&b->flags, BTREE_NODE_read_in_flight); printbuf_exit(&buf); diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c index 24671020f22b..b3c67ae3d3b2 100644 --- a/fs/bcachefs/recovery.c +++ b/fs/bcachefs/recovery.c @@ -33,6 +33,20 @@ #define QSTR(n) { { { .len = strlen(n) } }, .name = n } +void bch2_btree_lost_data(struct bch_fs *c, enum btree_id btree) +{ + u64 b = BIT_ULL(btree); + + if (!(c->sb.btrees_lost_data & b)) { + bch_err(c, "flagging btree %s lost data", bch2_btree_id_str(btree)); + + mutex_lock(&c->sb_lock); + bch2_sb_field_get(c->disk_sb.sb, ext)->btrees_lost_data |= cpu_to_le64(b); + bch2_write_super(c); + mutex_unlock(&c->sb_lock); + } +} + static bool btree_id_is_alloc(enum btree_id id) { switch (id) { @@ -470,6 +484,7 @@ static int read_btree_roots(struct bch_fs *c) } ret = 0; + bch2_btree_lost_data(c, i); } } @@ -848,6 +863,14 @@ use_clean: write_sb = true; } + if (c->opts.fsck && + !test_bit(BCH_FS_error, &c->flags) && + c->recovery_pass_done == BCH_RECOVERY_PASS_NR - 1 && + ext->btrees_lost_data) { + ext->btrees_lost_data = 0; + write_sb = true; + } + if (c->opts.fsck && !test_bit(BCH_FS_error, &c->flags) && !test_bit(BCH_FS_errors_not_fixed, &c->flags)) { diff --git a/fs/bcachefs/recovery.h b/fs/bcachefs/recovery.h index 3962fd87b50d..4bf818de1f2f 100644 --- a/fs/bcachefs/recovery.h +++ b/fs/bcachefs/recovery.h @@ -2,6 +2,8 @@ #ifndef _BCACHEFS_RECOVERY_H #define _BCACHEFS_RECOVERY_H +void bch2_btree_lost_data(struct bch_fs *, enum btree_id); + int bch2_journal_replay(struct bch_fs *); int bch2_fs_recovery(struct bch_fs *); diff --git a/fs/bcachefs/super-io.c b/fs/bcachefs/super-io.c index e15f8b1f30c2..e0aa3655b63b 100644 --- a/fs/bcachefs/super-io.c +++ b/fs/bcachefs/super-io.c @@ -527,9 +527,11 @@ static void bch2_sb_update(struct bch_fs *c) memset(c->sb.errors_silent, 0, sizeof(c->sb.errors_silent)); struct bch_sb_field_ext *ext = bch2_sb_field_get(src, ext); - if (ext) + if (ext) { le_bitvector_to_cpu(c->sb.errors_silent, (void *) ext->errors_silent, sizeof(c->sb.errors_silent) * 8); + c->sb.btrees_lost_data = le64_to_cpu(ext->btrees_lost_data); + } for_each_member_device(c, ca) { struct bch_member m = bch2_sb_member_get(src, ca->dev_idx); @@ -1162,6 +1164,11 @@ static void bch2_sb_ext_to_text(struct printbuf *out, struct bch_sb *sb, kfree(errors_silent); } + + prt_printf(out, "Btrees with missing data:"); + prt_tab(out); + prt_bitflags(out, __bch2_btree_ids, le64_to_cpu(e->btrees_lost_data)); + prt_newline(out); } static const struct bch_sb_field_ops bch_sb_field_ops_ext = { -- cgit From a292be3b68f3fdad6cff50c716174f51b119efd1 Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Wed, 27 Mar 2024 22:50:19 -0400 Subject: bcachefs: Reconstruct missing snapshot nodes When the snapshots btree is going, we'll have to delete huge amounts of data - unless we can reconstruct it by looking at the keys that refer to it. Signed-off-by: Kent Overstreet --- fs/bcachefs/bcachefs.h | 1 + fs/bcachefs/recovery.c | 1 + fs/bcachefs/recovery_passes_types.h | 1 + fs/bcachefs/sb-errors_types.h | 3 +- fs/bcachefs/snapshot.c | 173 +++++++++++++++++++++++++++++++++++- fs/bcachefs/snapshot.h | 26 +++++- 6 files changed, 199 insertions(+), 6 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/bcachefs.h b/fs/bcachefs/bcachefs.h index d1a0e5478503..a31a5f706929 100644 --- a/fs/bcachefs/bcachefs.h +++ b/fs/bcachefs/bcachefs.h @@ -615,6 +615,7 @@ struct bch_dev { */ #define BCH_FS_FLAGS() \ + x(new_fs) \ x(started) \ x(may_go_rw) \ x(rw) \ diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c index b3c67ae3d3b2..b76c16152579 100644 --- a/fs/bcachefs/recovery.c +++ b/fs/bcachefs/recovery.c @@ -938,6 +938,7 @@ int bch2_fs_initialize(struct bch_fs *c) int ret; bch_notice(c, "initializing new filesystem"); + set_bit(BCH_FS_new_fs, &c->flags); mutex_lock(&c->sb_lock); c->disk_sb.sb->compat[0] |= cpu_to_le64(1ULL << BCH_COMPAT_extents_above_btree_updates_done); diff --git a/fs/bcachefs/recovery_passes_types.h b/fs/bcachefs/recovery_passes_types.h index 840542cfd65b..773aea9a0080 100644 --- a/fs/bcachefs/recovery_passes_types.h +++ b/fs/bcachefs/recovery_passes_types.h @@ -32,6 +32,7 @@ x(check_alloc_to_lru_refs, 15, PASS_ONLINE|PASS_FSCK) \ x(fs_freespace_init, 16, PASS_ALWAYS|PASS_SILENT) \ x(bucket_gens_init, 17, 0) \ + x(reconstruct_snapshots, 38, 0) \ x(check_snapshot_trees, 18, PASS_ONLINE|PASS_FSCK) \ x(check_snapshots, 19, PASS_ONLINE|PASS_FSCK) \ x(check_subvols, 20, PASS_ONLINE|PASS_FSCK) \ diff --git a/fs/bcachefs/sb-errors_types.h b/fs/bcachefs/sb-errors_types.h index 2fec03a24c95..2f8f4d2388b0 100644 --- a/fs/bcachefs/sb-errors_types.h +++ b/fs/bcachefs/sb-errors_types.h @@ -268,7 +268,8 @@ x(btree_node_bkey_bad_u64s, 260) \ x(btree_node_topology_empty_interior_node, 261) \ x(btree_ptr_v2_min_key_bad, 262) \ - x(btree_root_unreadable_and_scan_found_nothing, 263) + x(btree_root_unreadable_and_scan_found_nothing, 263) \ + x(snapshot_node_missing, 264) enum bch_sb_error_id { #define x(t, n) BCH_FSCK_ERR_##t = n, diff --git a/fs/bcachefs/snapshot.c b/fs/bcachefs/snapshot.c index 4577ee7939a2..0e806f04f3d7 100644 --- a/fs/bcachefs/snapshot.c +++ b/fs/bcachefs/snapshot.c @@ -8,6 +8,7 @@ #include "errcode.h" #include "error.h" #include "fs.h" +#include "recovery_passes.h" #include "snapshot.h" #include @@ -574,6 +575,13 @@ static int check_snapshot_tree(struct btree_trans *trans, u32 subvol_id; ret = bch2_snapshot_tree_master_subvol(trans, root_id, &subvol_id); + bch_err_fn(c, ret); + + if (bch2_err_matches(ret, ENOENT)) { /* nothing to be done here */ + ret = 0; + goto err; + } + if (ret) goto err; @@ -731,7 +739,6 @@ static int check_snapshot(struct btree_trans *trans, u32 parent_id = bch2_snapshot_parent_early(c, k.k->p.offset); u32 real_depth; struct printbuf buf = PRINTBUF; - bool should_have_subvol; u32 i, id; int ret = 0; @@ -777,7 +784,7 @@ static int check_snapshot(struct btree_trans *trans, } } - should_have_subvol = BCH_SNAPSHOT_SUBVOL(&s) && + bool should_have_subvol = BCH_SNAPSHOT_SUBVOL(&s) && !BCH_SNAPSHOT_DELETED(&s); if (should_have_subvol) { @@ -879,6 +886,154 @@ int bch2_check_snapshots(struct bch_fs *c) return ret; } +static int check_snapshot_exists(struct btree_trans *trans, u32 id) +{ + struct bch_fs *c = trans->c; + + if (bch2_snapshot_equiv(c, id)) + return 0; + + u32 tree_id; + int ret = bch2_snapshot_tree_create(trans, id, 0, &tree_id); + if (ret) + return ret; + + struct bkey_i_snapshot *snapshot = bch2_trans_kmalloc(trans, sizeof(*snapshot)); + ret = PTR_ERR_OR_ZERO(snapshot); + if (ret) + return ret; + + bkey_snapshot_init(&snapshot->k_i); + snapshot->k.p = POS(0, id); + snapshot->v.tree = cpu_to_le32(tree_id); + snapshot->v.btime.lo = cpu_to_le64(bch2_current_time(c)); + + return bch2_btree_insert_trans(trans, BTREE_ID_snapshots, &snapshot->k_i, 0) ?: + bch2_mark_snapshot(trans, BTREE_ID_snapshots, 0, + bkey_s_c_null, bkey_i_to_s(&snapshot->k_i), 0) ?: + bch2_snapshot_set_equiv(trans, bkey_i_to_s_c(&snapshot->k_i)); +} + +/* Figure out which snapshot nodes belong in the same tree: */ +struct snapshot_tree_reconstruct { + enum btree_id btree; + struct bpos cur_pos; + snapshot_id_list cur_ids; + DARRAY(snapshot_id_list) trees; +}; + +static void snapshot_tree_reconstruct_exit(struct snapshot_tree_reconstruct *r) +{ + darray_for_each(r->trees, i) + darray_exit(i); + darray_exit(&r->trees); + darray_exit(&r->cur_ids); +} + +static inline bool same_snapshot(struct snapshot_tree_reconstruct *r, struct bpos pos) +{ + return r->btree == BTREE_ID_inodes + ? r->cur_pos.offset == pos.offset + : r->cur_pos.inode == pos.inode; +} + +static inline bool snapshot_id_lists_have_common(snapshot_id_list *l, snapshot_id_list *r) +{ + darray_for_each(*l, i) + if (snapshot_list_has_id(r, *i)) + return true; + return false; +} + +static void snapshot_id_list_to_text(struct printbuf *out, snapshot_id_list *s) +{ + bool first = true; + darray_for_each(*s, i) { + if (!first) + prt_char(out, ' '); + first = false; + prt_printf(out, "%u", *i); + } +} + +static int snapshot_tree_reconstruct_next(struct bch_fs *c, struct snapshot_tree_reconstruct *r) +{ + if (r->cur_ids.nr) { + darray_for_each(r->trees, i) + if (snapshot_id_lists_have_common(i, &r->cur_ids)) { + int ret = snapshot_list_merge(c, i, &r->cur_ids); + if (ret) + return ret; + goto out; + } + darray_push(&r->trees, r->cur_ids); + darray_init(&r->cur_ids); + } +out: + r->cur_ids.nr = 0; + return 0; +} + +static int get_snapshot_trees(struct bch_fs *c, struct snapshot_tree_reconstruct *r, struct bpos pos) +{ + if (!same_snapshot(r, pos)) + snapshot_tree_reconstruct_next(c, r); + r->cur_pos = pos; + return snapshot_list_add_nodup(c, &r->cur_ids, pos.snapshot); +} + +int bch2_reconstruct_snapshots(struct bch_fs *c) +{ + struct btree_trans *trans = bch2_trans_get(c); + struct printbuf buf = PRINTBUF; + struct snapshot_tree_reconstruct r = {}; + int ret = 0; + + for (unsigned btree = 0; btree < BTREE_ID_NR; btree++) { + if (btree_type_has_snapshots(btree)) { + r.btree = btree; + + ret = for_each_btree_key(trans, iter, btree, POS_MIN, + BTREE_ITER_ALL_SNAPSHOTS|BTREE_ITER_PREFETCH, k, ({ + get_snapshot_trees(c, &r, k.k->p); + })); + if (ret) + goto err; + + snapshot_tree_reconstruct_next(c, &r); + } + } + + darray_for_each(r.trees, t) { + printbuf_reset(&buf); + snapshot_id_list_to_text(&buf, t); + + darray_for_each(*t, id) { + if (fsck_err_on(!bch2_snapshot_equiv(c, *id), + c, snapshot_node_missing, + "snapshot node %u from tree %s missing", *id, buf.buf)) { + if (t->nr > 1) { + bch_err(c, "cannot reconstruct snapshot trees with multiple nodes"); + ret = -BCH_ERR_fsck_repair_unimplemented; + goto err; + } + + ret = commit_do(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc, + check_snapshot_exists(trans, *id)); + if (ret) + goto err; + } + } + } +fsck_err: +err: + bch2_trans_put(trans); + snapshot_tree_reconstruct_exit(&r); + printbuf_exit(&buf); + bch_err_fn(c, ret); + return ret; +} + /* * Mark a snapshot as deleted, for future cleanup: */ @@ -1689,6 +1844,20 @@ int bch2_snapshots_read(struct bch_fs *c) POS_MIN, 0, k, (set_is_ancestor_bitmap(c, k.k->p.offset), 0))); bch_err_fn(c, ret); + + /* + * It's important that we check if we need to reconstruct snapshots + * before going RW, so we mark that pass as required in the superblock - + * otherwise, we could end up deleting keys with missing snapshot nodes + * instead + */ + BUG_ON(!test_bit(BCH_FS_new_fs, &c->flags) && + test_bit(BCH_FS_may_go_rw, &c->flags)); + + if (bch2_err_matches(ret, EIO) || + (c->sb.btrees_lost_data & BIT_ULL(BTREE_ID_snapshots))) + ret = bch2_run_explicit_recovery_pass_persistent(c, BCH_RECOVERY_PASS_reconstruct_snapshots); + return ret; } diff --git a/fs/bcachefs/snapshot.h b/fs/bcachefs/snapshot.h index 331f20fd8d03..b7d2fed37c4f 100644 --- a/fs/bcachefs/snapshot.h +++ b/fs/bcachefs/snapshot.h @@ -209,15 +209,34 @@ static inline bool snapshot_list_has_ancestor(struct bch_fs *c, snapshot_id_list static inline int snapshot_list_add(struct bch_fs *c, snapshot_id_list *s, u32 id) { - int ret; - BUG_ON(snapshot_list_has_id(s, id)); - ret = darray_push(s, id); + int ret = darray_push(s, id); if (ret) bch_err(c, "error reallocating snapshot_id_list (size %zu)", s->size); return ret; } +static inline int snapshot_list_add_nodup(struct bch_fs *c, snapshot_id_list *s, u32 id) +{ + int ret = snapshot_list_has_id(s, id) + ? 0 + : darray_push(s, id); + if (ret) + bch_err(c, "error reallocating snapshot_id_list (size %zu)", s->size); + return ret; +} + +static inline int snapshot_list_merge(struct bch_fs *c, snapshot_id_list *dst, snapshot_id_list *src) +{ + darray_for_each(*src, i) { + int ret = snapshot_list_add_nodup(c, dst, *i); + if (ret) + return ret; + } + + return 0; +} + int bch2_snapshot_lookup(struct btree_trans *trans, u32 id, struct bch_snapshot *s); int bch2_snapshot_get_subvol(struct btree_trans *, u32, @@ -229,6 +248,7 @@ int bch2_snapshot_node_create(struct btree_trans *, u32, int bch2_check_snapshot_trees(struct bch_fs *); int bch2_check_snapshots(struct bch_fs *); +int bch2_reconstruct_snapshots(struct bch_fs *); int bch2_snapshot_node_set_deleted(struct btree_trans *, u32); void bch2_delete_dead_snapshots_work(struct work_struct *); -- cgit From 4c02e63dadc7f5a92732ce3e267d0511749f60fb Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Sat, 30 Mar 2024 18:43:00 -0400 Subject: bcachefs: Check for extents that point to same space In backpointer repair, if we get a missing backpointer - but there's already a backpointer that points to an existing extent - we've got multiple extents that point to the same space and need to decide which to keep. Signed-off-by: Kent Overstreet --- fs/bcachefs/backpointers.c | 173 ++++++++++++++++++++++++++++++++++++++++-- fs/bcachefs/sb-errors_types.h | 3 +- 2 files changed, 168 insertions(+), 8 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/backpointers.c b/fs/bcachefs/backpointers.c index 762c8ddfc5e7..114328acde72 100644 --- a/fs/bcachefs/backpointers.c +++ b/fs/bcachefs/backpointers.c @@ -8,6 +8,7 @@ #include "btree_update.h" #include "btree_update_interior.h" #include "btree_write_buffer.h" +#include "checksum.h" #include "error.h" #include @@ -418,6 +419,84 @@ struct extents_to_bp_state { struct bkey_buf last_flushed; }; +static int drop_dev_and_update(struct btree_trans *trans, enum btree_id btree, + struct bkey_s_c extent, unsigned dev) +{ + struct bkey_i *n = bch2_bkey_make_mut_noupdate(trans, extent); + int ret = PTR_ERR_OR_ZERO(n); + if (ret) + return ret; + + bch2_bkey_drop_device(bkey_i_to_s(n), dev); + return bch2_btree_insert_trans(trans, btree, n, 0); +} + +static int check_extent_checksum(struct btree_trans *trans, + enum btree_id btree, struct bkey_s_c extent, + enum btree_id o_btree, struct bkey_s_c extent2, unsigned dev) +{ + struct bch_fs *c = trans->c; + struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(extent); + const union bch_extent_entry *entry; + struct extent_ptr_decoded p; + struct printbuf buf = PRINTBUF; + void *data_buf = NULL; + struct bio *bio = NULL; + size_t bytes; + int ret = 0; + + if (bkey_is_btree_ptr(extent.k)) + return false; + + bkey_for_each_ptr_decode(extent.k, ptrs, p, entry) + if (p.ptr.dev == dev) + goto found; + BUG(); +found: + if (!p.crc.csum_type) + return false; + + bytes = p.crc.compressed_size << 9; + + struct bch_dev *ca = bch_dev_bkey_exists(c, dev); + if (!bch2_dev_get_ioref(ca, READ)) + return false; + + data_buf = kvmalloc(bytes, GFP_KERNEL); + if (!data_buf) { + ret = -ENOMEM; + goto err; + } + + bio = bio_alloc(ca->disk_sb.bdev, 1, REQ_OP_READ, GFP_KERNEL); + bio->bi_iter.bi_sector = p.ptr.offset; + bch2_bio_map(bio, data_buf, bytes); + ret = submit_bio_wait(bio); + if (ret) + goto err; + + prt_str(&buf, "extents pointing to same space, but first extent checksum bad:"); + prt_printf(&buf, "\n %s ", bch2_btree_id_str(btree)); + bch2_bkey_val_to_text(&buf, c, extent); + prt_printf(&buf, "\n %s ", bch2_btree_id_str(o_btree)); + bch2_bkey_val_to_text(&buf, c, extent2); + + struct nonce nonce = extent_nonce(extent.k->version, p.crc); + struct bch_csum csum = bch2_checksum(c, p.crc.csum_type, nonce, data_buf, bytes); + if (fsck_err_on(bch2_crc_cmp(csum, p.crc.csum), + c, dup_backpointer_to_bad_csum_extent, + "%s", buf.buf)) + ret = drop_dev_and_update(trans, btree, extent, dev) ?: 1; +fsck_err: +err: + if (bio) + bio_put(bio); + kvfree(data_buf); + percpu_ref_put(&ca->io_ref); + printbuf_exit(&buf); + return ret; +} + static int check_bp_exists(struct btree_trans *trans, struct extents_to_bp_state *s, struct bpos bucket, @@ -425,7 +504,8 @@ static int check_bp_exists(struct btree_trans *trans, struct bkey_s_c orig_k) { struct bch_fs *c = trans->c; - struct btree_iter bp_iter = { NULL }; + struct btree_iter bp_iter = {}; + struct btree_iter other_extent_iter = {}; struct printbuf buf = PRINTBUF; struct bkey_s_c bp_k; struct bkey_buf tmp; @@ -433,13 +513,19 @@ static int check_bp_exists(struct btree_trans *trans, bch2_bkey_buf_init(&tmp); + if (!bch2_dev_bucket_exists(c, bucket)) { + prt_str(&buf, "extent for nonexistent device:bucket "); + bch2_bpos_to_text(&buf, bucket); + prt_str(&buf, "\n "); + bch2_bkey_val_to_text(&buf, c, orig_k); + bch_err(c, "%s", buf.buf); + return -BCH_ERR_fsck_repair_unimplemented; + } + if (bpos_lt(bucket, s->bucket_start) || bpos_gt(bucket, s->bucket_end)) return 0; - if (!bch2_dev_bucket_exists(c, bucket)) - goto missing; - bp_k = bch2_bkey_get_iter(trans, &bp_iter, BTREE_ID_backpointers, bucket_pos_to_bp(c, bucket, bp.bucket_offset), 0); @@ -465,21 +551,94 @@ static int check_bp_exists(struct btree_trans *trans, ret = -BCH_ERR_transaction_restart_write_buffer_flush; goto out; } - goto missing; + + goto check_existing_bp; } out: err: fsck_err: + bch2_trans_iter_exit(trans, &other_extent_iter); bch2_trans_iter_exit(trans, &bp_iter); bch2_bkey_buf_exit(&tmp, c); printbuf_exit(&buf); return ret; +check_existing_bp: + /* Do we have a backpointer for a different extent? */ + if (bp_k.k->type != KEY_TYPE_backpointer) + goto missing; + + struct bch_backpointer other_bp = *bkey_s_c_to_backpointer(bp_k).v; + + struct bkey_s_c other_extent = + bch2_backpointer_get_key(trans, &other_extent_iter, bp_k.k->p, other_bp, 0); + ret = bkey_err(other_extent); + if (ret == -BCH_ERR_backpointer_to_overwritten_btree_node) + ret = 0; + if (ret) + goto err; + + if (!other_extent.k) + goto missing; + + if (bch2_extents_match(orig_k, other_extent)) { + printbuf_reset(&buf); + prt_printf(&buf, "duplicate versions of same extent, deleting smaller\n "); + bch2_bkey_val_to_text(&buf, c, orig_k); + prt_str(&buf, "\n "); + bch2_bkey_val_to_text(&buf, c, other_extent); + bch_err(c, "%s", buf.buf); + + if (other_extent.k->size <= orig_k.k->size) { + ret = drop_dev_and_update(trans, other_bp.btree_id, other_extent, bucket.inode); + if (ret) + goto err; + goto out; + } else { + ret = drop_dev_and_update(trans, bp.btree_id, orig_k, bucket.inode); + if (ret) + goto err; + goto missing; + } + } + + ret = check_extent_checksum(trans, other_bp.btree_id, other_extent, bp.btree_id, orig_k, bucket.inode); + if (ret < 0) + goto err; + if (ret) { + ret = 0; + goto missing; + } + + ret = check_extent_checksum(trans, bp.btree_id, orig_k, other_bp.btree_id, other_extent, bucket.inode); + if (ret < 0) + goto err; + if (ret) { + ret = 0; + goto out; + } + + printbuf_reset(&buf); + prt_printf(&buf, "duplicate extents pointing to same space on dev %llu\n ", bucket.inode); + bch2_bkey_val_to_text(&buf, c, orig_k); + prt_str(&buf, "\n "); + bch2_bkey_val_to_text(&buf, c, other_extent); + bch_err(c, "%s", buf.buf); + ret = -BCH_ERR_fsck_repair_unimplemented; + goto err; missing: + printbuf_reset(&buf); prt_printf(&buf, "missing backpointer for btree=%s l=%u ", bch2_btree_id_str(bp.btree_id), bp.level); bch2_bkey_val_to_text(&buf, c, orig_k); - prt_printf(&buf, "\nbp pos "); - bch2_bpos_to_text(&buf, bp_iter.pos); + prt_printf(&buf, "\n got: "); + bch2_bkey_val_to_text(&buf, c, bp_k); + + struct bkey_i_backpointer n_bp_k; + bkey_backpointer_init(&n_bp_k.k_i); + n_bp_k.k.p = bucket_pos_to_bp(trans->c, bucket, bp.bucket_offset); + n_bp_k.v = bp; + prt_printf(&buf, "\n want: "); + bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&n_bp_k.k_i)); if (fsck_err(c, ptr_to_missing_backpointer, "%s", buf.buf)) ret = bch2_bucket_backpointer_mod(trans, bucket, bp, orig_k, true); diff --git a/fs/bcachefs/sb-errors_types.h b/fs/bcachefs/sb-errors_types.h index 2f8f4d2388b0..d7d609131030 100644 --- a/fs/bcachefs/sb-errors_types.h +++ b/fs/bcachefs/sb-errors_types.h @@ -269,7 +269,8 @@ x(btree_node_topology_empty_interior_node, 261) \ x(btree_ptr_v2_min_key_bad, 262) \ x(btree_root_unreadable_and_scan_found_nothing, 263) \ - x(snapshot_node_missing, 264) + x(snapshot_node_missing, 264) \ + x(dup_backpointer_to_bad_csum_extent, 265) enum bch_sb_error_id { #define x(t, n) BCH_FSCK_ERR_##t = n, -- cgit From cc0532900bcf1896a81dcdd30873ffa6c4f6926b Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Sun, 31 Mar 2024 02:03:03 -0400 Subject: bcachefs: Subvolume reconstruction We can now recreate missing subvolumes from dirents and/or inodes. Signed-off-by: Kent Overstreet --- fs/bcachefs/fsck.c | 167 +++++++++++++++++++++++++++++++++++++++++++++++------ 1 file changed, 148 insertions(+), 19 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/fsck.c b/fs/bcachefs/fsck.c index b704fb0dda95..c1edc5647ba0 100644 --- a/fs/bcachefs/fsck.c +++ b/fs/bcachefs/fsck.c @@ -63,9 +63,7 @@ static int subvol_lookup(struct btree_trans *trans, u32 subvol, u32 *snapshot, u64 *inum) { struct bch_subvolume s; - int ret; - - ret = bch2_subvolume_get(trans, subvol, false, 0, &s); + int ret = bch2_subvolume_get(trans, subvol, false, 0, &s); *snapshot = le32_to_cpu(s.snapshot); *inum = le64_to_cpu(s.inode); @@ -170,7 +168,8 @@ err: /* Get lost+found, create if it doesn't exist: */ static int lookup_lostfound(struct btree_trans *trans, u32 snapshot, - struct bch_inode_unpacked *lostfound) + struct bch_inode_unpacked *lostfound, + u64 reattaching_inum) { struct bch_fs *c = trans->c; struct qstr lostfound_str = QSTR("lost+found"); @@ -185,19 +184,36 @@ static int lookup_lostfound(struct btree_trans *trans, u32 snapshot, return ret; subvol_inum root_inum = { .subvol = le32_to_cpu(st.master_subvol) }; - u32 subvol_snapshot; - ret = subvol_lookup(trans, le32_to_cpu(st.master_subvol), - &subvol_snapshot, &root_inum.inum); - bch_err_msg(c, ret, "looking up root subvol"); + struct bch_subvolume subvol; + ret = bch2_subvolume_get(trans, le32_to_cpu(st.master_subvol), + false, 0, &subvol); + bch_err_msg(c, ret, "looking up root subvol %u for snapshot %u", + le32_to_cpu(st.master_subvol), snapshot); if (ret) return ret; + if (!subvol.inode) { + struct btree_iter iter; + struct bkey_i_subvolume *subvol = bch2_bkey_get_mut_typed(trans, &iter, + BTREE_ID_subvolumes, POS(0, le32_to_cpu(st.master_subvol)), + 0, subvolume); + ret = PTR_ERR_OR_ZERO(subvol); + if (ret) + return ret; + + subvol->v.inode = cpu_to_le64(reattaching_inum); + bch2_trans_iter_exit(trans, &iter); + } + + root_inum.inum = le64_to_cpu(subvol.inode); + struct bch_inode_unpacked root_inode; struct bch_hash_info root_hash_info; u32 root_inode_snapshot = snapshot; ret = lookup_inode(trans, root_inum.inum, &root_inode, &root_inode_snapshot); - bch_err_msg(c, ret, "looking up root inode"); + bch_err_msg(c, ret, "looking up root inode %llu for subvol %u", + root_inum.inum, le32_to_cpu(st.master_subvol)); if (ret) return ret; @@ -293,7 +309,7 @@ static int reattach_inode(struct btree_trans *trans, snprintf(name_buf, sizeof(name_buf), "%llu", inode->bi_inum); } - ret = lookup_lostfound(trans, dirent_snapshot, &lostfound); + ret = lookup_lostfound(trans, dirent_snapshot, &lostfound, inode->bi_inum); if (ret) return ret; @@ -364,6 +380,85 @@ static int reattach_subvol(struct btree_trans *trans, struct bkey_s_c_subvolume return ret; } +static int reconstruct_subvol(struct btree_trans *trans, u32 snapshotid, u32 subvolid, u64 inum) +{ + struct bch_fs *c = trans->c; + + if (!bch2_snapshot_is_leaf(c, snapshotid)) { + bch_err(c, "need to reconstruct subvol, but have interior node snapshot"); + return -BCH_ERR_fsck_repair_unimplemented; + } + + /* + * If inum isn't set, that means we're being called from check_dirents, + * not check_inodes - the root of this subvolume doesn't exist or we + * would have found it there: + */ + if (!inum) { + struct btree_iter inode_iter = {}; + struct bch_inode_unpacked new_inode; + u64 cpu = raw_smp_processor_id(); + + bch2_inode_init_early(c, &new_inode); + bch2_inode_init_late(&new_inode, bch2_current_time(c), 0, 0, S_IFDIR|0755, 0, NULL); + + new_inode.bi_subvol = subvolid; + + int ret = bch2_inode_create(trans, &inode_iter, &new_inode, snapshotid, cpu) ?: + bch2_btree_iter_traverse(&inode_iter) ?: + bch2_inode_write(trans, &inode_iter, &new_inode); + bch2_trans_iter_exit(trans, &inode_iter); + if (ret) + return ret; + + inum = new_inode.bi_inum; + } + + bch_info(c, "reconstructing subvol %u with root inode %llu", subvolid, inum); + + struct bkey_i_subvolume *new_subvol = bch2_trans_kmalloc(trans, sizeof(*new_subvol)); + int ret = PTR_ERR_OR_ZERO(new_subvol); + if (ret) + return ret; + + bkey_subvolume_init(&new_subvol->k_i); + new_subvol->k.p.offset = subvolid; + new_subvol->v.snapshot = cpu_to_le32(snapshotid); + new_subvol->v.inode = cpu_to_le64(inum); + ret = bch2_btree_insert_trans(trans, BTREE_ID_subvolumes, &new_subvol->k_i, 0); + if (ret) + return ret; + + struct btree_iter iter; + struct bkey_i_snapshot *s = bch2_bkey_get_mut_typed(trans, &iter, + BTREE_ID_snapshots, POS(0, snapshotid), + 0, snapshot); + ret = PTR_ERR_OR_ZERO(s); + bch_err_msg(c, ret, "getting snapshot %u", snapshotid); + if (ret) + return ret; + + u32 snapshot_tree = le32_to_cpu(s->v.tree); + + s->v.subvol = cpu_to_le32(subvolid); + SET_BCH_SNAPSHOT_SUBVOL(&s->v, true); + bch2_trans_iter_exit(trans, &iter); + + struct bkey_i_snapshot_tree *st = bch2_bkey_get_mut_typed(trans, &iter, + BTREE_ID_snapshot_trees, POS(0, snapshot_tree), + 0, snapshot_tree); + ret = PTR_ERR_OR_ZERO(st); + bch_err_msg(c, ret, "getting snapshot tree %u", snapshot_tree); + if (ret) + return ret; + + if (!st->v.master_subvol) + st->v.master_subvol = cpu_to_le32(subvolid); + + bch2_trans_iter_exit(trans, &iter); + return 0; +} + struct snapshots_seen_entry { u32 id; u32 equiv; @@ -1065,6 +1160,11 @@ static int check_inode(struct btree_trans *trans, if (ret && !bch2_err_matches(ret, ENOENT)) goto err; + if (ret && (c->sb.btrees_lost_data & BIT_ULL(BTREE_ID_subvolumes))) { + ret = reconstruct_subvol(trans, k.k->p.snapshot, u.bi_subvol, u.bi_inum); + goto do_update; + } + if (fsck_err_on(ret, c, inode_bi_subvol_missing, "inode %llu:%u bi_subvol points to missing subvolume %u", @@ -1082,7 +1182,7 @@ static int check_inode(struct btree_trans *trans, do_update = true; } } - +do_update: if (do_update) { ret = __bch2_fsck_write_inode(trans, &u, iter->pos.snapshot); bch_err_msg(c, ret, "in fsck updating inode"); @@ -1785,6 +1885,7 @@ static int check_dirent_to_subvol(struct btree_trans *trans, struct btree_iter * u32 parent_subvol = le32_to_cpu(d.v->d_parent_subvol); u32 target_subvol = le32_to_cpu(d.v->d_child_subvol); u32 parent_snapshot; + u32 new_parent_subvol = 0; u64 parent_inum; struct printbuf buf = PRINTBUF; int ret = 0; @@ -1793,6 +1894,27 @@ static int check_dirent_to_subvol(struct btree_trans *trans, struct btree_iter * if (ret && !bch2_err_matches(ret, ENOENT)) return ret; + if (ret || + (!ret && !bch2_snapshot_is_ancestor(c, parent_snapshot, d.k->p.snapshot))) { + int ret2 = find_snapshot_subvol(trans, d.k->p.snapshot, &new_parent_subvol); + if (ret2 && !bch2_err_matches(ret, ENOENT)) + return ret2; + } + + if (ret && + !new_parent_subvol && + (c->sb.btrees_lost_data & BIT_ULL(BTREE_ID_subvolumes))) { + /* + * Couldn't find a subvol for dirent's snapshot - but we lost + * subvols, so we need to reconstruct: + */ + ret = reconstruct_subvol(trans, d.k->p.snapshot, parent_subvol, 0); + if (ret) + return ret; + + parent_snapshot = d.k->p.snapshot; + } + if (fsck_err_on(ret, c, dirent_to_missing_parent_subvol, "dirent parent_subvol points to missing subvolume\n%s", (bch2_bkey_val_to_text(&buf, c, d.s_c), buf.buf)) || @@ -1801,10 +1923,10 @@ static int check_dirent_to_subvol(struct btree_trans *trans, struct btree_iter * "dirent not visible in parent_subvol (not an ancestor of subvol snap %u)\n%s", parent_snapshot, (bch2_bkey_val_to_text(&buf, c, d.s_c), buf.buf))) { - u32 new_parent_subvol; - ret = find_snapshot_subvol(trans, d.k->p.snapshot, &new_parent_subvol); - if (ret) - goto err; + if (!new_parent_subvol) { + bch_err(c, "could not find a subvol for snapshot %u", d.k->p.snapshot); + return -BCH_ERR_fsck_repair_unimplemented; + } struct bkey_i_dirent *new_dirent = bch2_bkey_make_mut_typed(trans, iter, &d.s_c, 0, dirent); ret = PTR_ERR_OR_ZERO(new_dirent); @@ -1850,9 +1972,16 @@ static int check_dirent_to_subvol(struct btree_trans *trans, struct btree_iter * ret = lookup_inode(trans, target_inum, &subvol_root, &target_snapshot); if (ret && !bch2_err_matches(ret, ENOENT)) - return ret; + goto err; - if (fsck_err_on(parent_subvol != subvol_root.bi_parent_subvol, + if (ret) { + bch_err(c, "subvol %u points to missing inode root %llu", target_subvol, target_inum); + ret = -BCH_ERR_fsck_repair_unimplemented; + ret = 0; + goto err; + } + + if (fsck_err_on(!ret && parent_subvol != subvol_root.bi_parent_subvol, c, inode_bi_parent_wrong, "subvol root %llu has wrong bi_parent_subvol: got %u, should be %u", target_inum, @@ -1860,13 +1989,13 @@ static int check_dirent_to_subvol(struct btree_trans *trans, struct btree_iter * subvol_root.bi_parent_subvol = parent_subvol; ret = __bch2_fsck_write_inode(trans, &subvol_root, target_snapshot); if (ret) - return ret; + goto err; } ret = check_dirent_target(trans, iter, d, &subvol_root, target_snapshot); if (ret) - return ret; + goto err; out: err: fsck_err: -- cgit From 09d4c2acbf4c864fef0f520bbcba256c9a19102e Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Mon, 1 Apr 2024 00:00:56 -0400 Subject: bcachefs: reconstruct_inode() If an inode is missing, but corresponding extents and dirent still exist, it's well worth recreating it - this does so. Signed-off-by: Kent Overstreet --- fs/bcachefs/fsck.c | 52 ++++++++++++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 50 insertions(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/bcachefs/fsck.c b/fs/bcachefs/fsck.c index c1edc5647ba0..8e2010212cc3 100644 --- a/fs/bcachefs/fsck.c +++ b/fs/bcachefs/fsck.c @@ -459,6 +459,33 @@ static int reconstruct_subvol(struct btree_trans *trans, u32 snapshotid, u32 sub return 0; } +static int reconstruct_inode(struct btree_trans *trans, u32 snapshot, u64 inum, u64 size, unsigned mode) +{ + struct bch_fs *c = trans->c; + struct bch_inode_unpacked new_inode; + + bch2_inode_init_early(c, &new_inode); + bch2_inode_init_late(&new_inode, bch2_current_time(c), 0, 0, mode|0755, 0, NULL); + new_inode.bi_size = size; + new_inode.bi_inum = inum; + + return __bch2_fsck_write_inode(trans, &new_inode, snapshot); +} + +static int reconstruct_reg_inode(struct btree_trans *trans, u32 snapshot, u64 inum) +{ + struct btree_iter iter = {}; + + bch2_trans_iter_init(trans, &iter, BTREE_ID_extents, SPOS(inum, U64_MAX, snapshot), 0); + struct bkey_s_c k = bch2_btree_iter_peek_prev(&iter); + bch2_trans_iter_exit(trans, &iter); + int ret = bkey_err(k); + if (ret) + return ret; + + return reconstruct_inode(trans, snapshot, inum, k.k->p.offset << 9, S_IFREG); +} + struct snapshots_seen_entry { u32 id; u32 equiv; @@ -1535,6 +1562,17 @@ static int check_extent(struct btree_trans *trans, struct btree_iter *iter, goto err; if (k.k->type != KEY_TYPE_whiteout) { + if (!i && (c->sb.btrees_lost_data & BIT_ULL(BTREE_ID_inodes))) { + ret = reconstruct_reg_inode(trans, k.k->p.snapshot, k.k->p.inode) ?: + bch2_trans_commit(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc); + if (ret) + goto err; + + inode->last_pos.inode--; + ret = -BCH_ERR_transaction_restart_nested; + goto err; + } + if (fsck_err_on(!i, c, extent_in_missing_inode, "extent in missing inode:\n %s", (printbuf_reset(&buf), @@ -2012,7 +2050,6 @@ static int check_dirent(struct btree_trans *trans, struct btree_iter *iter, struct snapshots_seen *s) { struct bch_fs *c = trans->c; - struct bkey_s_c_dirent d; struct inode_walker_entry *i; struct printbuf buf = PRINTBUF; struct bpos equiv; @@ -2051,6 +2088,17 @@ static int check_dirent(struct btree_trans *trans, struct btree_iter *iter, *hash_info = bch2_hash_info_init(c, &dir->inodes.data[0].inode); dir->first_this_inode = false; + if (!i && (c->sb.btrees_lost_data & BIT_ULL(BTREE_ID_inodes))) { + ret = reconstruct_inode(trans, k.k->p.snapshot, k.k->p.inode, 0, S_IFDIR) ?: + bch2_trans_commit(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc); + if (ret) + goto err; + + dir->last_pos.inode--; + ret = -BCH_ERR_transaction_restart_nested; + goto err; + } + if (fsck_err_on(!i, c, dirent_in_missing_dir_inode, "dirent in nonexisting directory:\n%s", (printbuf_reset(&buf), @@ -2085,7 +2133,7 @@ static int check_dirent(struct btree_trans *trans, struct btree_iter *iter, if (k.k->type != KEY_TYPE_dirent) goto out; - d = bkey_s_c_to_dirent(k); + struct bkey_s_c_dirent d = bkey_s_c_to_dirent(k); if (d.v->d_type == DT_SUBVOL) { ret = check_dirent_to_subvol(trans, iter, d); -- cgit