From 954998b60caa8f2a3bf3abe490de6f08d283687a Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 4 Sep 2023 12:34:37 -0400 Subject: NFS: Fix error handling for O_DIRECT write scheduling If we fail to schedule a request for transmission, there are 2 possibilities: 1) Either we hit a fatal error, and we just want to drop the remaining requests on the floor. 2) We were asked to try again, in which case we should allow the outstanding RPC calls to complete, so that we can recoalesce requests and try again. Fixes: d600ad1f2bdb ("NFS41: pop some layoutget errors to application") Signed-off-by: Trond Myklebust Signed-off-by: Anna Schumaker --- fs/nfs/direct.c | 62 ++++++++++++++++++++++++++++++++++++++++++--------------- 1 file changed, 46 insertions(+), 16 deletions(-) diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index 47d892a1d363..ee88f0a6e7b8 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -528,10 +528,9 @@ nfs_direct_write_scan_commit_list(struct inode *inode, static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) { struct nfs_pageio_descriptor desc; - struct nfs_page *req, *tmp; + struct nfs_page *req; LIST_HEAD(reqs); struct nfs_commit_info cinfo; - LIST_HEAD(failed); nfs_init_cinfo_from_dreq(&cinfo, dreq); nfs_direct_write_scan_commit_list(dreq->inode, &reqs, &cinfo); @@ -549,27 +548,36 @@ static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) &nfs_direct_write_completion_ops); desc.pg_dreq = dreq; - list_for_each_entry_safe(req, tmp, &reqs, wb_list) { + while (!list_empty(&reqs)) { + req = nfs_list_entry(reqs.next); /* Bump the transmission count */ req->wb_nio++; if (!nfs_pageio_add_request(&desc, req)) { - nfs_list_move_request(req, &failed); spin_lock(&cinfo.inode->i_lock); - dreq->flags = 0; - if (desc.pg_error < 0) + if (dreq->error < 0) { + desc.pg_error = dreq->error; + } else if (desc.pg_error != -EAGAIN) { + dreq->flags = 0; + if (!desc.pg_error) + desc.pg_error = -EIO; dreq->error = desc.pg_error; - else - dreq->error = -EIO; + } else + dreq->flags = NFS_ODIRECT_RESCHED_WRITES; spin_unlock(&cinfo.inode->i_lock); + break; } nfs_release_request(req); } nfs_pageio_complete(&desc); - while (!list_empty(&failed)) { - req = nfs_list_entry(failed.next); + while (!list_empty(&reqs)) { + req = nfs_list_entry(reqs.next); nfs_list_remove_request(req); nfs_unlock_and_release_request(req); + if (desc.pg_error == -EAGAIN) + nfs_mark_request_commit(req, NULL, &cinfo, 0); + else + nfs_release_request(req); } if (put_dreq(dreq)) @@ -794,9 +802,11 @@ static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq, { struct nfs_pageio_descriptor desc; struct inode *inode = dreq->inode; + struct nfs_commit_info cinfo; ssize_t result = 0; size_t requested_bytes = 0; size_t wsize = max_t(size_t, NFS_SERVER(inode)->wsize, PAGE_SIZE); + bool defer = false; trace_nfs_direct_write_schedule_iovec(dreq); @@ -837,17 +847,37 @@ static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq, break; } - nfs_lock_request(req); - if (!nfs_pageio_add_request(&desc, req)) { - result = desc.pg_error; - nfs_unlock_and_release_request(req); - break; - } pgbase = 0; bytes -= req_len; requested_bytes += req_len; pos += req_len; dreq->bytes_left -= req_len; + + if (defer) { + nfs_mark_request_commit(req, NULL, &cinfo, 0); + continue; + } + + nfs_lock_request(req); + if (nfs_pageio_add_request(&desc, req)) + continue; + + /* Exit on hard errors */ + if (desc.pg_error < 0 && desc.pg_error != -EAGAIN) { + result = desc.pg_error; + nfs_unlock_and_release_request(req); + break; + } + + /* If the error is soft, defer remaining requests */ + nfs_init_cinfo_from_dreq(&cinfo, dreq); + spin_lock(&cinfo.inode->i_lock); + dreq->flags = NFS_ODIRECT_RESCHED_WRITES; + spin_unlock(&cinfo.inode->i_lock); + nfs_unlock_request(req); + nfs_mark_request_commit(req, NULL, &cinfo, 0); + desc.pg_error = 0; + defer = true; } nfs_direct_release_pages(pagevec, npages); kvfree(pagevec); -- cgit From 7c6339322ce0c6128acbe36aacc1eeb986dd7bf1 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 4 Sep 2023 12:34:38 -0400 Subject: NFS: Fix O_DIRECT locking issues The dreq fields are protected by the dreq->lock. Fixes: 954998b60caa ("NFS: Fix error handling for O_DIRECT write scheduling") Signed-off-by: Trond Myklebust Signed-off-by: Anna Schumaker --- fs/nfs/direct.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index ee88f0a6e7b8..e8a1645857dd 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -553,7 +553,7 @@ static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) /* Bump the transmission count */ req->wb_nio++; if (!nfs_pageio_add_request(&desc, req)) { - spin_lock(&cinfo.inode->i_lock); + spin_lock(&dreq->lock); if (dreq->error < 0) { desc.pg_error = dreq->error; } else if (desc.pg_error != -EAGAIN) { @@ -563,7 +563,7 @@ static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) dreq->error = desc.pg_error; } else dreq->flags = NFS_ODIRECT_RESCHED_WRITES; - spin_unlock(&cinfo.inode->i_lock); + spin_unlock(&dreq->lock); break; } nfs_release_request(req); @@ -871,9 +871,9 @@ static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq, /* If the error is soft, defer remaining requests */ nfs_init_cinfo_from_dreq(&cinfo, dreq); - spin_lock(&cinfo.inode->i_lock); + spin_lock(&dreq->lock); dreq->flags = NFS_ODIRECT_RESCHED_WRITES; - spin_unlock(&cinfo.inode->i_lock); + spin_unlock(&dreq->lock); nfs_unlock_request(req); nfs_mark_request_commit(req, NULL, &cinfo, 0); desc.pg_error = 0; -- cgit From 8982f7aff39fb526aba4441fff2525fcedd5e1a3 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 4 Sep 2023 12:34:39 -0400 Subject: NFS: More O_DIRECT accounting fixes for error paths If we hit a fatal error when retransmitting, we do need to record the removal of the request from the count of written bytes. Fixes: 031d73ed768a ("NFS: Fix O_DIRECT accounting of number of bytes read/written") Signed-off-by: Trond Myklebust Signed-off-by: Anna Schumaker --- fs/nfs/direct.c | 47 +++++++++++++++++++++++++++++++---------------- 1 file changed, 31 insertions(+), 16 deletions(-) diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index e8a1645857dd..a53e50123499 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -93,12 +93,10 @@ nfs_direct_handle_truncated(struct nfs_direct_req *dreq, dreq->max_count = dreq_len; if (dreq->count > dreq_len) dreq->count = dreq_len; - - if (test_bit(NFS_IOHDR_ERROR, &hdr->flags)) - dreq->error = hdr->error; - else /* Clear outstanding error if this is EOF */ - dreq->error = 0; } + + if (test_bit(NFS_IOHDR_ERROR, &hdr->flags) && !dreq->error) + dreq->error = hdr->error; } static void @@ -120,6 +118,18 @@ nfs_direct_count_bytes(struct nfs_direct_req *dreq, dreq->count = dreq_len; } +static void nfs_direct_truncate_request(struct nfs_direct_req *dreq, + struct nfs_page *req) +{ + loff_t offs = req_offset(req); + size_t req_start = (size_t)(offs - dreq->io_start); + + if (req_start < dreq->max_count) + dreq->max_count = req_start; + if (req_start < dreq->count) + dreq->count = req_start; +} + /** * nfs_swap_rw - NFS address space operation for swap I/O * @iocb: target I/O control block @@ -537,10 +547,6 @@ static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) nfs_direct_join_group(&reqs, dreq->inode); - dreq->count = 0; - dreq->max_count = 0; - list_for_each_entry(req, &reqs, wb_list) - dreq->max_count += req->wb_bytes; nfs_clear_pnfs_ds_commit_verifiers(&dreq->ds_cinfo); get_dreq(dreq); @@ -574,10 +580,14 @@ static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) req = nfs_list_entry(reqs.next); nfs_list_remove_request(req); nfs_unlock_and_release_request(req); - if (desc.pg_error == -EAGAIN) + if (desc.pg_error == -EAGAIN) { nfs_mark_request_commit(req, NULL, &cinfo, 0); - else + } else { + spin_lock(&dreq->lock); + nfs_direct_truncate_request(dreq, req); + spin_unlock(&dreq->lock); nfs_release_request(req); + } } if (put_dreq(dreq)) @@ -597,8 +607,6 @@ static void nfs_direct_commit_complete(struct nfs_commit_data *data) if (status < 0) { /* Errors in commit are fatal */ dreq->error = status; - dreq->max_count = 0; - dreq->count = 0; dreq->flags = NFS_ODIRECT_DONE; } else { status = dreq->error; @@ -609,7 +617,12 @@ static void nfs_direct_commit_complete(struct nfs_commit_data *data) while (!list_empty(&data->pages)) { req = nfs_list_entry(data->pages.next); nfs_list_remove_request(req); - if (status >= 0 && !nfs_write_match_verf(verf, req)) { + if (status < 0) { + spin_lock(&dreq->lock); + nfs_direct_truncate_request(dreq, req); + spin_unlock(&dreq->lock); + nfs_release_request(req); + } else if (!nfs_write_match_verf(verf, req)) { dreq->flags = NFS_ODIRECT_RESCHED_WRITES; /* * Despite the reboot, the write was successful, @@ -617,7 +630,7 @@ static void nfs_direct_commit_complete(struct nfs_commit_data *data) */ req->wb_nio = 0; nfs_mark_request_commit(req, NULL, &cinfo, 0); - } else /* Error or match */ + } else nfs_release_request(req); nfs_unlock_and_release_request(req); } @@ -670,6 +683,7 @@ static void nfs_direct_write_clear_reqs(struct nfs_direct_req *dreq) while (!list_empty(&reqs)) { req = nfs_list_entry(reqs.next); nfs_list_remove_request(req); + nfs_direct_truncate_request(dreq, req); nfs_release_request(req); nfs_unlock_and_release_request(req); } @@ -719,7 +733,8 @@ static void nfs_direct_write_completion(struct nfs_pgio_header *hdr) } nfs_direct_count_bytes(dreq, hdr); - if (test_bit(NFS_IOHDR_UNSTABLE_WRITES, &hdr->flags)) { + if (test_bit(NFS_IOHDR_UNSTABLE_WRITES, &hdr->flags) && + !test_bit(NFS_IOHDR_ERROR, &hdr->flags)) { if (!dreq->flags) dreq->flags = NFS_ODIRECT_DO_COMMIT; flags = dreq->flags; -- cgit From b193a78ddb5ee7dba074d3f28dc050069ba083c0 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 4 Sep 2023 12:34:40 -0400 Subject: NFS: Use the correct commit info in nfs_join_page_group() Ensure that nfs_clear_request_commit() updates the correct counters when it removes them from the commit list. Fixes: ed5d588fe47f ("NFS: Try to join page groups before an O_DIRECT retransmission") Signed-off-by: Trond Myklebust Signed-off-by: Anna Schumaker --- fs/nfs/direct.c | 8 +++++--- fs/nfs/write.c | 23 ++++++++++++----------- include/linux/nfs_page.h | 4 +++- 3 files changed, 20 insertions(+), 15 deletions(-) diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index a53e50123499..3391c8b97da5 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -498,7 +498,9 @@ static void nfs_direct_add_page_head(struct list_head *list, kref_get(&head->wb_kref); } -static void nfs_direct_join_group(struct list_head *list, struct inode *inode) +static void nfs_direct_join_group(struct list_head *list, + struct nfs_commit_info *cinfo, + struct inode *inode) { struct nfs_page *req, *subreq; @@ -520,7 +522,7 @@ static void nfs_direct_join_group(struct list_head *list, struct inode *inode) nfs_release_request(subreq); } } while ((subreq = subreq->wb_this_page) != req); - nfs_join_page_group(req, inode); + nfs_join_page_group(req, cinfo, inode); } } @@ -545,7 +547,7 @@ static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) nfs_init_cinfo_from_dreq(&cinfo, dreq); nfs_direct_write_scan_commit_list(dreq->inode, &reqs, &cinfo); - nfs_direct_join_group(&reqs, dreq->inode); + nfs_direct_join_group(&reqs, &cinfo, dreq->inode); nfs_clear_pnfs_ds_commit_verifiers(&dreq->ds_cinfo); get_dreq(dreq); diff --git a/fs/nfs/write.c b/fs/nfs/write.c index f4cca8f00c0c..8c1ee1a1a28f 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -59,7 +59,8 @@ static const struct nfs_pgio_completion_ops nfs_async_write_completion_ops; static const struct nfs_commit_completion_ops nfs_commit_completion_ops; static const struct nfs_rw_ops nfs_rw_write_ops; static void nfs_inode_remove_request(struct nfs_page *req); -static void nfs_clear_request_commit(struct nfs_page *req); +static void nfs_clear_request_commit(struct nfs_commit_info *cinfo, + struct nfs_page *req); static void nfs_init_cinfo_from_inode(struct nfs_commit_info *cinfo, struct inode *inode); static struct nfs_page * @@ -502,8 +503,8 @@ nfs_destroy_unlinked_subrequests(struct nfs_page *destroy_list, * the (former) group. All subrequests are removed from any write or commit * lists, unlinked from the group and destroyed. */ -void -nfs_join_page_group(struct nfs_page *head, struct inode *inode) +void nfs_join_page_group(struct nfs_page *head, struct nfs_commit_info *cinfo, + struct inode *inode) { struct nfs_page *subreq; struct nfs_page *destroy_list = NULL; @@ -533,7 +534,7 @@ nfs_join_page_group(struct nfs_page *head, struct inode *inode) * Commit list removal accounting is done after locks are dropped */ subreq = head; do { - nfs_clear_request_commit(subreq); + nfs_clear_request_commit(cinfo, subreq); subreq = subreq->wb_this_page; } while (subreq != head); @@ -566,8 +567,10 @@ static struct nfs_page *nfs_lock_and_join_requests(struct folio *folio) { struct inode *inode = folio_file_mapping(folio)->host; struct nfs_page *head; + struct nfs_commit_info cinfo; int ret; + nfs_init_cinfo_from_inode(&cinfo, inode); /* * A reference is taken only on the head request which acts as a * reference to the whole page group - the group will not be destroyed @@ -584,7 +587,7 @@ static struct nfs_page *nfs_lock_and_join_requests(struct folio *folio) return ERR_PTR(ret); } - nfs_join_page_group(head, inode); + nfs_join_page_group(head, &cinfo, inode); return head; } @@ -955,18 +958,16 @@ static void nfs_folio_clear_commit(struct folio *folio) } /* Called holding the request lock on @req */ -static void -nfs_clear_request_commit(struct nfs_page *req) +static void nfs_clear_request_commit(struct nfs_commit_info *cinfo, + struct nfs_page *req) { if (test_bit(PG_CLEAN, &req->wb_flags)) { struct nfs_open_context *ctx = nfs_req_openctx(req); struct inode *inode = d_inode(ctx->dentry); - struct nfs_commit_info cinfo; - nfs_init_cinfo_from_inode(&cinfo, inode); mutex_lock(&NFS_I(inode)->commit_mutex); - if (!pnfs_clear_request_commit(req, &cinfo)) { - nfs_request_remove_commit_list(req, &cinfo); + if (!pnfs_clear_request_commit(req, cinfo)) { + nfs_request_remove_commit_list(req, cinfo); } mutex_unlock(&NFS_I(inode)->commit_mutex); nfs_folio_clear_commit(nfs_page_to_folio(req)); diff --git a/include/linux/nfs_page.h b/include/linux/nfs_page.h index aa9f4c6ebe26..1c315f854ea8 100644 --- a/include/linux/nfs_page.h +++ b/include/linux/nfs_page.h @@ -157,7 +157,9 @@ extern void nfs_unlock_request(struct nfs_page *req); extern void nfs_unlock_and_release_request(struct nfs_page *); extern struct nfs_page *nfs_page_group_lock_head(struct nfs_page *req); extern int nfs_page_group_lock_subrequests(struct nfs_page *head); -extern void nfs_join_page_group(struct nfs_page *head, struct inode *inode); +extern void nfs_join_page_group(struct nfs_page *head, + struct nfs_commit_info *cinfo, + struct inode *inode); extern int nfs_page_group_lock(struct nfs_page *); extern void nfs_page_group_unlock(struct nfs_page *); extern bool nfs_page_group_sync_on_bit(struct nfs_page *, unsigned int); -- cgit From b11243f720ee5f9376861099019c8542969b6318 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 4 Sep 2023 12:34:41 -0400 Subject: NFS: More fixes for nfs_direct_write_reschedule_io() Ensure that all requests are put back onto the commit list so that they can be rescheduled. Fixes: 4daaeba93822 ("NFS: Fix nfs_direct_write_reschedule_io()") Signed-off-by: Trond Myklebust Signed-off-by: Anna Schumaker --- fs/nfs/direct.c | 17 +++++++++++------ 1 file changed, 11 insertions(+), 6 deletions(-) diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index 3391c8b97da5..f6c74f424691 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -780,18 +780,23 @@ static void nfs_write_sync_pgio_error(struct list_head *head, int error) static void nfs_direct_write_reschedule_io(struct nfs_pgio_header *hdr) { struct nfs_direct_req *dreq = hdr->dreq; + struct nfs_page *req; + struct nfs_commit_info cinfo; trace_nfs_direct_write_reschedule_io(dreq); + nfs_init_cinfo_from_dreq(&cinfo, dreq); spin_lock(&dreq->lock); - if (dreq->error == 0) { + if (dreq->error == 0) dreq->flags = NFS_ODIRECT_RESCHED_WRITES; - /* fake unstable write to let common nfs resend pages */ - hdr->verf.committed = NFS_UNSTABLE; - hdr->good_bytes = hdr->args.offset + hdr->args.count - - hdr->io_start; - } + set_bit(NFS_IOHDR_REDO, &hdr->flags); spin_unlock(&dreq->lock); + while (!list_empty(&hdr->pages)) { + req = nfs_list_entry(hdr->pages.next); + nfs_list_remove_request(req); + nfs_unlock_request(req); + nfs_mark_request_commit(req, NULL, &cinfo, 0); + } } static const struct nfs_pgio_completion_ops nfs_direct_write_completion_ops = { -- cgit From dd7d7ee3ba2a70d12d02defb478790cf57d5b87b Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 4 Sep 2023 12:43:58 -0400 Subject: NFS/pNFS: Report EINVAL errors from connect() to the server With IPv6, connect() can occasionally return EINVAL if a route is unavailable. If this happens during I/O to a data server, we want to report it using LAYOUTERROR as an inability to connect. Fixes: dd52128afdde ("NFSv4.1/pnfs Ensure flexfiles reports all connection related errors") Signed-off-by: Trond Myklebust Signed-off-by: Anna Schumaker --- fs/nfs/flexfilelayout/flexfilelayout.c | 1 + 1 file changed, 1 insertion(+) diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index 7deb3cd76abe..a1dc33864906 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -1235,6 +1235,7 @@ static void ff_layout_io_track_ds_error(struct pnfs_layout_segment *lseg, case -EPFNOSUPPORT: case -EPROTONOSUPPORT: case -EOPNOTSUPP: + case -EINVAL: case -ECONNREFUSED: case -ECONNRESET: case -EHOSTDOWN: -- cgit From 611fa42dfa9d2f3918ac5f4dd5705dfad81b323d Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 4 Sep 2023 12:50:09 -0400 Subject: SUNRPC: Mark the cred for revalidation if the server rejects it If the server rejects the credential as being stale, or bad, then we should mark it for revalidation before retransmitting. Fixes: 7f5667a5f8c4 ("SUNRPC: Clean up rpc_verify_header()") Signed-off-by: Trond Myklebust Signed-off-by: Anna Schumaker --- net/sunrpc/clnt.c | 1 + 1 file changed, 1 insertion(+) diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c index 8d75290f1a31..5c37621aa09a 100644 --- a/net/sunrpc/clnt.c +++ b/net/sunrpc/clnt.c @@ -2751,6 +2751,7 @@ out_msg_denied: case rpc_autherr_rejectedverf: case rpcsec_gsserr_credproblem: case rpcsec_gsserr_ctxproblem: + rpcauth_invalcred(task); if (!task->tk_cred_retry) break; task->tk_cred_retry--; -- cgit From e86fcf0820d914389b46658a5a7e8969c3af2d53 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 5 Sep 2023 21:03:28 -0400 Subject: Revert "SUNRPC: Fail faster on bad verifier" This reverts commit 0701214cd6e66585a999b132eb72ae0489beb724. The premise of this commit was incorrect. There are exactly 2 cases where rpcauth_checkverf() will return an error: 1) If there was an XDR decode problem (i.e. garbage data). 2) If gss_validate() had a problem verifying the RPCSEC_GSS MIC. In the second case, there are again 2 subcases: a) The GSS context expires, in which case gss_validate() will force a new context negotiation on retry by invalidating the cred. b) The sequence number check failed because an RPC call timed out, and the client retransmitted the request using a new sequence number, as required by RFC2203. In neither subcase is this a fatal error. Reported-by: Russell Cattelan Fixes: 0701214cd6e6 ("SUNRPC: Fail faster on bad verifier") Cc: stable@vger.kernel.org Signed-off-by: Trond Myklebust Signed-off-by: Anna Schumaker --- net/sunrpc/clnt.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c index 5c37621aa09a..edbcfdd84e1f 100644 --- a/net/sunrpc/clnt.c +++ b/net/sunrpc/clnt.c @@ -2725,7 +2725,7 @@ out_unparsable: out_verifier: trace_rpc_bad_verifier(task); - goto out_err; + goto out_garbage; out_msg_denied: error = -EACCES; -- cgit From 806a3bc421a115fbb287c1efce63a48c54ee804b Mon Sep 17 00:00:00 2001 From: Olga Kornievskaia Date: Wed, 30 Aug 2023 15:29:34 -0400 Subject: NFSv4.1: fix pnfs MDS=DS session trunking Currently, when GETDEVICEINFO returns multiple locations where each is a different IP but the server's identity is same as MDS, then nfs4_set_ds_client() finds the existing nfs_client structure which has the MDS's max_connect value (and if it's 1), then the 1st IP on the DS's list will get dropped due to MDS trunking rules. Other IPs would be added as they fall under the pnfs trunking rules. For the list of IPs the 1st goes thru calling nfs4_set_ds_client() which will eventually call nfs4_add_trunk() and call into rpc_clnt_test_and_add_xprt() which has the check for MDS trunking. The other IPs (after the 1st one), would call rpc_clnt_add_xprt() which doesn't go thru that check. nfs4_add_trunk() is called when MDS trunking is happening and it needs to enforce the usage of max_connect mount option of the 1st mount. However, this shouldn't be applied to pnfs flow. Instead, this patch proposed to treat MDS=DS as DS trunking and make sure that MDS's max_connect limit does not apply to the 1st IP returned in the GETDEVICEINFO list. It does so by marking the newly created client with a new flag NFS_CS_PNFS which then used to pass max_connect value to use into the rpc_clnt_test_and_add_xprt() instead of the existing rpc client's max_connect value set by the MDS connection. For example, mount was done without max_connect value set so MDS's rpc client has cl_max_connect=1. Upon calling into rpc_clnt_test_and_add_xprt() and using rpc client's value, the caller passes in max_connect value which is previously been set in the pnfs path (as a part of handling GETDEVICEINFO list of IPs) in nfs4_set_ds_client(). However, when NFS_CS_PNFS flag is not set and we know we are doing MDS trunking, comparing a new IP of the same server, we then set the max_connect value to the existing MDS's value and pass that into rpc_clnt_test_and_add_xprt(). Fixes: dc48e0abee24 ("SUNRPC enforce creation of no more than max_connect xprts") Signed-off-by: Olga Kornievskaia Signed-off-by: Anna Schumaker --- fs/nfs/nfs4client.c | 6 +++++- include/linux/nfs_fs_sb.h | 1 + net/sunrpc/clnt.c | 11 +++++++---- 3 files changed, 13 insertions(+), 5 deletions(-) diff --git a/fs/nfs/nfs4client.c b/fs/nfs/nfs4client.c index 27fb25567ce7..11e3a285594c 100644 --- a/fs/nfs/nfs4client.c +++ b/fs/nfs/nfs4client.c @@ -417,6 +417,8 @@ static void nfs4_add_trunk(struct nfs_client *clp, struct nfs_client *old) .net = old->cl_net, .servername = old->cl_hostname, }; + int max_connect = test_bit(NFS_CS_PNFS, &clp->cl_flags) ? + clp->cl_max_connect : old->cl_max_connect; if (clp->cl_proto != old->cl_proto) return; @@ -430,7 +432,7 @@ static void nfs4_add_trunk(struct nfs_client *clp, struct nfs_client *old) xprt_args.addrlen = clp_salen; rpc_clnt_add_xprt(old->cl_rpcclient, &xprt_args, - rpc_clnt_test_and_add_xprt, NULL); + rpc_clnt_test_and_add_xprt, &max_connect); } /** @@ -1010,6 +1012,8 @@ struct nfs_client *nfs4_set_ds_client(struct nfs_server *mds_srv, __set_bit(NFS_CS_NORESVPORT, &cl_init.init_flags); __set_bit(NFS_CS_DS, &cl_init.init_flags); + __set_bit(NFS_CS_PNFS, &cl_init.init_flags); + cl_init.max_connect = NFS_MAX_TRANSPORTS; /* * Set an authflavor equual to the MDS value. Use the MDS nfs_client * cl_ipaddr so as to use the same EXCHANGE_ID co_ownerid as the MDS diff --git a/include/linux/nfs_fs_sb.h b/include/linux/nfs_fs_sb.h index 20eeba8b009d..cd628c4b011e 100644 --- a/include/linux/nfs_fs_sb.h +++ b/include/linux/nfs_fs_sb.h @@ -48,6 +48,7 @@ struct nfs_client { #define NFS_CS_NOPING 6 /* - don't ping on connect */ #define NFS_CS_DS 7 /* - Server is a DS */ #define NFS_CS_REUSEPORT 8 /* - reuse src port on reconnect */ +#define NFS_CS_PNFS 9 /* - Server used for pnfs */ struct sockaddr_storage cl_addr; /* server identifier */ size_t cl_addrlen; char * cl_hostname; /* hostname of server */ diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c index edbcfdd84e1f..37b0b212b934 100644 --- a/net/sunrpc/clnt.c +++ b/net/sunrpc/clnt.c @@ -2908,19 +2908,22 @@ static const struct rpc_call_ops rpc_cb_add_xprt_call_ops = { * @clnt: pointer to struct rpc_clnt * @xps: pointer to struct rpc_xprt_switch, * @xprt: pointer struct rpc_xprt - * @dummy: unused + * @in_max_connect: pointer to the max_connect value for the passed in xprt transport */ int rpc_clnt_test_and_add_xprt(struct rpc_clnt *clnt, struct rpc_xprt_switch *xps, struct rpc_xprt *xprt, - void *dummy) + void *in_max_connect) { struct rpc_cb_add_xprt_calldata *data; struct rpc_task *task; + int max_connect = clnt->cl_max_connect; - if (xps->xps_nunique_destaddr_xprts + 1 > clnt->cl_max_connect) { + if (in_max_connect) + max_connect = *(int *)in_max_connect; + if (xps->xps_nunique_destaddr_xprts + 1 > max_connect) { rcu_read_lock(); pr_warn("SUNRPC: reached max allowed number (%d) did not add " - "transport to server: %s\n", clnt->cl_max_connect, + "transport to server: %s\n", max_connect, rpc_peeraddr2str(clnt, RPC_DISPLAY_ADDR)); rcu_read_unlock(); return -EINVAL; -- cgit From 4506f23e117161a20104c8fa04f33e1ca63c26af Mon Sep 17 00:00:00 2001 From: Olga Kornievskaia Date: Thu, 13 Jul 2023 15:54:16 -0400 Subject: NFSv4.1: fix zero value filehandle in post open getattr Currently, if the OPEN compound experiencing an error and needs to get the file attributes separately, it will send a stand alone GETATTR but it would use the filehandle from the results of the OPEN compound. In case of the CLAIM_FH OPEN, nfs_openres's fh is zero value. That generate a GETATTR that's sent with a zero value filehandle, and results in the server returning an error. Instead, for the CLAIM_FH OPEN, take the filehandle that was used in the PUTFH of the OPEN compound. Signed-off-by: Olga Kornievskaia Reviewed-by: Benjamin Coddington Signed-off-by: Anna Schumaker --- fs/nfs/nfs4proc.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 794343790ea8..3508d8238826 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -2703,8 +2703,12 @@ static int _nfs4_proc_open(struct nfs4_opendata *data, return status; } if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) { + struct nfs_fh *fh = &o_res->fh; + nfs4_sequence_free_slot(&o_res->seq_res); - nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, NULL); + if (o_arg->claim == NFS4_OPEN_CLAIM_FH) + fh = NFS_FH(d_inode(data->dentry)); + nfs4_proc_getattr(server, fh, o_res->f_attr, NULL); } return 0; } -- cgit From c656a4d5484ad99e97de549a9affc12a91d94963 Mon Sep 17 00:00:00 2001 From: Anna Schumaker Date: Fri, 15 Sep 2023 15:46:08 -0400 Subject: Revert "SUNRPC: clean up integer overflow check" This reverts commit e87cf8a28e7592bd19064e8181324ae26bc02932. This commit was added to silence a tautological comparison warning, but removing the 'len' value check before calling xdr_inline_decode() is really not what we want. Signed-off-by: Anna Schumaker --- include/linux/sunrpc/xdr.h | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/include/linux/sunrpc/xdr.h b/include/linux/sunrpc/xdr.h index 5b4fb3c791bc..896a6d2a9cf0 100644 --- a/include/linux/sunrpc/xdr.h +++ b/include/linux/sunrpc/xdr.h @@ -779,7 +779,9 @@ xdr_stream_decode_uint32_array(struct xdr_stream *xdr, if (unlikely(xdr_stream_decode_u32(xdr, &len) < 0)) return -EBADMSG; - p = xdr_inline_decode(xdr, size_mul(len, sizeof(*p))); + if (len > SIZE_MAX / sizeof(*p)) + return -EBADMSG; + p = xdr_inline_decode(xdr, len * sizeof(*p)); if (unlikely(!p)) return -EBADMSG; if (array == NULL) -- cgit From 993b5662f302628db4eb358d69b2720c88cbfaf0 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 13 Sep 2023 16:12:33 -0400 Subject: SUNRPC: Silence compiler complaints about tautological comparisons On 64-bit systems, the compiler will complain that the comparison between SIZE_MAX and the 32-bit unsigned int 'len' is unnecessary. Signed-off-by: Trond Myklebust Signed-off-by: Anna Schumaker --- include/linux/sunrpc/xdr.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/include/linux/sunrpc/xdr.h b/include/linux/sunrpc/xdr.h index 896a6d2a9cf0..2f8dc47f1eb0 100644 --- a/include/linux/sunrpc/xdr.h +++ b/include/linux/sunrpc/xdr.h @@ -779,7 +779,7 @@ xdr_stream_decode_uint32_array(struct xdr_stream *xdr, if (unlikely(xdr_stream_decode_u32(xdr, &len) < 0)) return -EBADMSG; - if (len > SIZE_MAX / sizeof(*p)) + if (U32_MAX >= SIZE_MAX / sizeof(*p) && len > SIZE_MAX / sizeof(*p)) return -EBADMSG; p = xdr_inline_decode(xdr, len * sizeof(*p)); if (unlikely(!p)) -- cgit