Merge tag 'xfs-5.3-merge-12' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux

Pull xfs updates from Darrick Wong:
 "In this release there are a significant amounts of consolidations and
  cleanups in the log code; restructuring of the log to issue struct
  bios directly; new bulkstat ioctls to return v5 fs inode information
  (and fix all the padding problems of the old ioctl); the beginnings of
  multithreaded inode walks (e.g. quotacheck); and a reduction in memory
  usage in the online scrub code leading to reduced runtimes.

   - Refactor inode geometry calculation into a single structure instead
     of open-coding pieces everywhere.

   - Add online repair to build options.

   - Remove unnecessary function call flags and functions.

   - Claim maintainership of various loose xfs documentation and header
     files.

   - Use struct bio directly for log buffer IOs instead of struct
     xfs_buf.

   - Reduce log item boilerplate code requirements.

   - Merge log item code spread across too many files.

   - Further distinguish between log item commits and cancellations.

   - Various small cleanups to the ag small allocator.

   - Support cgroup-aware writeback

   - libxfs refactoring for mkfs cleanup

   - Remove unneeded #includes

   - Fix a memory allocation miscalculation in the new log bio code

   - Fix bisection problems

   - Fix a crash in ioend processing caused by tripping over freeing of
     preallocated transactions

   - Split out a generic inode walk mechanism from the bulkstat code,
     hook up all the internal users to use the walking code, then clean
     up bulkstat to serve only the bulkstat ioctls.

   - Add a multithreaded iwalk implementation to speed up quotacheck on
     fast storage with many CPUs.

   - Remove unnecessary return values in logging teardown functions.

   - Supplement the bstat and inogrp structures with new bulkstat and
     inumbers structures that have all the fields we need for v5
     filesystem features and none of the padding problems of their
     predecessors.

   - Wire up new ioctls that use the new structures with a much simpler
     bulk_ireq structure at the head instead of the pointerhappy mess we
     had before.

   - Enable userspace to constrain bulkstat returns to a single AG or a
     single special inode so that we can phase out a lot of geometry
     guesswork in userspace.

   - Reduce memory consumption and zeroing overhead in extended
     attribute scrub code.

   - Fix some behavioral regressions in the new bulkstat backend code.

   - Fix some behavioral regressions in the new log bio code"

* tag 'xfs-5.3-merge-12' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux: (100 commits)
  xfs: chain bios the right way around in xfs_rw_bdev
  xfs: bump INUMBERS cursor correctly in xfs_inumbers_walk
  xfs: don't update lastino for FSBULKSTAT_SINGLE
  xfs: online scrub needn't bother zeroing its temporary buffer
  xfs: only allocate memory for scrubbing attributes when we need it
  xfs: refactor attr scrub memory allocation function
  xfs: refactor extended attribute buffer pointer functions
  xfs: attribute scrub should use seen_enough to pass error values
  xfs: allow single bulkstat of special inodes
  xfs: specify AG in bulk req
  xfs: wire up the v5 inumbers ioctl
  xfs: wire up new v5 bulkstat ioctls
  xfs: introduce v5 inode group structure
  xfs: introduce new v5 bulkstat structure
  xfs: rename bulkstat functions
  xfs: remove various bulk request typedef usage
  fs: xfs: xfs_log: Change return type from int to void
  xfs: poll waiting for quotacheck
  xfs: multithreaded iwalk implementation
  xfs: refactor INUMBERS to use iwalk functions
  ...

1 files changed, 267 insertions, 377 deletions

diff --git a/fs/xfs/xfs_log.c b/fs/xfs/xfs_log.c
index 2466b0f5b6c4..00e9f5c388d3 100644
--- a/fs/xfs/xfs_log.c
+++ b/fs/xfs/xfs_log.c
@@ -16,11 +16,7 @@
 #include "xfs_trans_priv.h"
 #include "xfs_log.h"
 #include "xfs_log_priv.h"
-#include "xfs_log_recover.h"
-#include "xfs_inode.h"
 #include "xfs_trace.h"
-#include "xfs_fsops.h"
-#include "xfs_cksum.h"
 #include "xfs_sysfs.h"
 #include "xfs_sb.h"
 #include "xfs_health.h"
@@ -45,21 +41,14 @@ STATIC int
 xlog_space_left(
 	struct xlog		*log,
 	atomic64_t		*head);
-STATIC int
-xlog_sync(
-	struct xlog		*log,
-	struct xlog_in_core	*iclog);
 STATIC void
 xlog_dealloc_log(
 	struct xlog		*log);
 
 /* local state machine functions */
-STATIC void xlog_state_done_syncing(xlog_in_core_t *iclog, int);
-STATIC void
-xlog_state_do_callback(
-	struct xlog		*log,
-	int			aborted,
-	struct xlog_in_core	*iclog);
+STATIC void xlog_state_done_syncing(
+	struct xlog_in_core	*iclog,
+	bool			aborted);
 STATIC int
 xlog_state_get_iclog_space(
 	struct xlog		*log,
@@ -107,8 +96,7 @@ STATIC void
 xlog_verify_iclog(
 	struct xlog		*log,
 	struct xlog_in_core	*iclog,
-	int			count,
-	bool                    syncing);
+	int			count);
 STATIC void
 xlog_verify_tail_lsn(
 	struct xlog		*log,
@@ -117,7 +105,7 @@ xlog_verify_tail_lsn(
 #else
 #define xlog_verify_dest_ptr(a,b)
 #define xlog_verify_grant_tail(a)
-#define xlog_verify_iclog(a,b,c,d)
+#define xlog_verify_iclog(a,b,c)
 #define xlog_verify_tail_lsn(a,b,c)
 #endif
 
@@ -541,32 +529,6 @@ xfs_log_done(
 	return lsn;
 }
 
-/*
- * Attaches a new iclog I/O completion callback routine during
- * transaction commit.  If the log is in error state, a non-zero
- * return code is handed back and the caller is responsible for
- * executing the callback at an appropriate time.
- */
-int
-xfs_log_notify(
-	struct xlog_in_core	*iclog,
-	xfs_log_callback_t	*cb)
-{
-	int	abortflg;
-
-	spin_lock(&iclog->ic_callback_lock);
-	abortflg = (iclog->ic_state & XLOG_STATE_IOERROR);
-	if (!abortflg) {
-		ASSERT_ALWAYS((iclog->ic_state == XLOG_STATE_ACTIVE) ||
-			      (iclog->ic_state == XLOG_STATE_WANT_SYNC));
-		cb->cb_next = NULL;
-		*(iclog->ic_callback_tail) = cb;
-		iclog->ic_callback_tail = &(cb->cb_next);
-	}
-	spin_unlock(&iclog->ic_callback_lock);
-	return abortflg;
-}
-
 int
 xfs_log_release_iclog(
 	struct xfs_mount	*mp,
@@ -807,16 +769,12 @@ xfs_log_mount_finish(
  * The mount has failed. Cancel the recovery if it hasn't completed and destroy
  * the log.
  */
-int
+void
 xfs_log_mount_cancel(
 	struct xfs_mount	*mp)
 {
-	int			error;
-
-	error = xlog_recover_cancel(mp->m_log);
+	xlog_recover_cancel(mp->m_log);
 	xfs_log_unmount(mp);
-
-	return error;
 }
 
 /*
@@ -932,7 +890,7 @@ xfs_log_unmount_write(xfs_mount_t *mp)
 	 * Or, if we are doing a forced umount (typically because of IO errors).
 	 */
 	if (mp->m_flags & XFS_MOUNT_NORECOVERY ||
-	    xfs_readonly_buftarg(log->l_mp->m_logdev_targp)) {
+	    xfs_readonly_buftarg(log->l_targ)) {
 		ASSERT(mp->m_flags & XFS_MOUNT_RDONLY);
 		return 0;
 	}
@@ -1244,53 +1202,49 @@ xlog_space_left(
 }
 
 
-/*
- * Log function which is called when an io completes.
- *
- * The log manager needs its own routine, in order to control what
- * happens with the buffer after the write completes.
- */
 static void
-xlog_iodone(xfs_buf_t *bp)
+xlog_ioend_work(
+	struct work_struct	*work)
 {
-	struct xlog_in_core	*iclog = bp->b_log_item;
-	struct xlog		*l = iclog->ic_log;
-	int			aborted = 0;
+	struct xlog_in_core     *iclog =
+		container_of(work, struct xlog_in_core, ic_end_io_work);
+	struct xlog		*log = iclog->ic_log;
+	bool			aborted = false;
+	int			error;
+
+	error = blk_status_to_errno(iclog->ic_bio.bi_status);
+#ifdef DEBUG
+	/* treat writes with injected CRC errors as failed */
+	if (iclog->ic_fail_crc)
+		error = -EIO;
+#endif
 
 	/*
-	 * Race to shutdown the filesystem if we see an error or the iclog is in
-	 * IOABORT state. The IOABORT state is only set in DEBUG mode to inject
-	 * CRC errors into log recovery.
+	 * Race to shutdown the filesystem if we see an error.
 	 */
-	if (XFS_TEST_ERROR(bp->b_error, l->l_mp, XFS_ERRTAG_IODONE_IOERR) ||
-	    iclog->ic_state & XLOG_STATE_IOABORT) {
-		if (iclog->ic_state & XLOG_STATE_IOABORT)
-			iclog->ic_state &= ~XLOG_STATE_IOABORT;
-
-		xfs_buf_ioerror_alert(bp, __func__);
-		xfs_buf_stale(bp);
-		xfs_force_shutdown(l->l_mp, SHUTDOWN_LOG_IO_ERROR);
+	if (XFS_TEST_ERROR(error, log->l_mp, XFS_ERRTAG_IODONE_IOERR)) {
+		xfs_alert(log->l_mp, "log I/O error %d", error);
+		xfs_force_shutdown(log->l_mp, SHUTDOWN_LOG_IO_ERROR);
 		/*
 		 * This flag will be propagated to the trans-committed
 		 * callback routines to let them know that the log-commit
 		 * didn't succeed.
 		 */
-		aborted = XFS_LI_ABORTED;
+		aborted = true;
 	} else if (iclog->ic_state & XLOG_STATE_IOERROR) {
-		aborted = XFS_LI_ABORTED;
+		aborted = true;
 	}
 
-	/* log I/O is always issued ASYNC */
-	ASSERT(bp->b_flags & XBF_ASYNC);
 	xlog_state_done_syncing(iclog, aborted);
+	bio_uninit(&iclog->ic_bio);
 
 	/*
-	 * drop the buffer lock now that we are done. Nothing references
-	 * the buffer after this, so an unmount waiting on this lock can now
-	 * tear it down safely. As such, it is unsafe to reference the buffer
-	 * (bp) after the unlock as we could race with it being freed.
+	 * Drop the lock to signal that we are done. Nothing references the
+	 * iclog after this, so an unmount waiting on this lock can now tear it
+	 * down safely. As such, it is unsafe to reference the iclog after the
+	 * unlock as we could race with it being freed.
 	 */
-	xfs_buf_unlock(bp);
+	up(&iclog->ic_sema);
 }
 
 /*
@@ -1301,65 +1255,26 @@ xlog_iodone(xfs_buf_t *bp)
  * If the filesystem blocksize is too large, we may need to choose a
  * larger size since the directory code currently logs entire blocks.
  */
-
 STATIC void
 xlog_get_iclog_buffer_size(
 	struct xfs_mount	*mp,
 	struct xlog		*log)
 {
-	int size;
-	int xhdrs;
-
 	if (mp->m_logbufs <= 0)
-		log->l_iclog_bufs = XLOG_MAX_ICLOGS;
-	else
-		log->l_iclog_bufs = mp->m_logbufs;
+		mp->m_logbufs = XLOG_MAX_ICLOGS;
+	if (mp->m_logbsize <= 0)
+		mp->m_logbsize = XLOG_BIG_RECORD_BSIZE;
+
+	log->l_iclog_bufs = mp->m_logbufs;
+	log->l_iclog_size = mp->m_logbsize;
 
 	/*
-	 * Buffer size passed in from mount system call.
+	 * # headers = size / 32k - one header holds cycles from 32k of data.
 	 */
-	if (mp->m_logbsize > 0) {
-		size = log->l_iclog_size = mp->m_logbsize;
-		log->l_iclog_size_log = 0;
-		while (size != 1) {
-			log->l_iclog_size_log++;
-			size >>= 1;
-		}
-
-		if (xfs_sb_version_haslogv2(&mp->m_sb)) {
-			/* # headers = size / 32k
-			 * one header holds cycles from 32k of data
-			 */
-
-			xhdrs = mp->m_logbsize / XLOG_HEADER_CYCLE_SIZE;
-			if (mp->m_logbsize % XLOG_HEADER_CYCLE_SIZE)
-				xhdrs++;
-			log->l_iclog_hsize = xhdrs << BBSHIFT;
-			log->l_iclog_heads = xhdrs;
-		} else {
-			ASSERT(mp->m_logbsize <= XLOG_BIG_RECORD_BSIZE);
-			log->l_iclog_hsize = BBSIZE;
-			log->l_iclog_heads = 1;
-		}
-		goto done;
-	}
-
-	/* All machines use 32kB buffers by default. */
-	log->l_iclog_size = XLOG_BIG_RECORD_BSIZE;
-	log->l_iclog_size_log = XLOG_BIG_RECORD_BSHIFT;
-
-	/* the default log size is 16k or 32k which is one header sector */
-	log->l_iclog_hsize = BBSIZE;
-	log->l_iclog_heads = 1;
-
-done:
-	/* are we being asked to make the sizes selected above visible? */
-	if (mp->m_logbufs == 0)
-		mp->m_logbufs = log->l_iclog_bufs;
-	if (mp->m_logbsize == 0)
-		mp->m_logbsize = log->l_iclog_size;
-}	/* xlog_get_iclog_buffer_size */
-
+	log->l_iclog_heads =
+		DIV_ROUND_UP(mp->m_logbsize, XLOG_HEADER_CYCLE_SIZE);
+	log->l_iclog_hsize = log->l_iclog_heads << BBSHIFT;
+}
 
 void
 xfs_log_work_queue(
@@ -1422,7 +1337,6 @@ xlog_alloc_log(
 	xlog_rec_header_t	*head;
 	xlog_in_core_t		**iclogp;
 	xlog_in_core_t		*iclog, *prev_iclog=NULL;
-	xfs_buf_t		*bp;
 	int			i;
 	int			error = -ENOMEM;
 	uint			log2_size = 0;
@@ -1480,30 +1394,6 @@ xlog_alloc_log(
 
 	xlog_get_iclog_buffer_size(mp, log);
 
-	/*
-	 * Use a NULL block for the extra log buffer used during splits so that
-	 * it will trigger errors if we ever try to do IO on it without first
-	 * having set it up properly.
-	 */
-	error = -ENOMEM;
-	bp = xfs_buf_alloc(mp->m_logdev_targp, XFS_BUF_DADDR_NULL,
-			   BTOBB(log->l_iclog_size), XBF_NO_IOACCT);
-	if (!bp)
-		goto out_free_log;
-
-	/*
-	 * The iclogbuf buffer locks are held over IO but we are not going to do
-	 * IO yet.  Hence unlock the buffer so that the log IO path can grab it
-	 * when appropriately.
-	 */
-	ASSERT(xfs_buf_islocked(bp));
-	xfs_buf_unlock(bp);
-
-	/* use high priority wq for log I/O completion */
-	bp->b_ioend_wq = mp->m_log_workqueue;
-	bp->b_iodone = xlog_iodone;
-	log->l_xbuf = bp;
-
 	spin_lock_init(&log->l_icloglock);
 	init_waitqueue_head(&log->l_flush_wait);
 
@@ -1516,29 +1406,22 @@ xlog_alloc_log(
 	 * xlog_in_core_t in xfs_log_priv.h for details.
 	 */
 	ASSERT(log->l_iclog_size >= 4096);
-	for (i=0; i < log->l_iclog_bufs; i++) {
-		*iclogp = kmem_zalloc(sizeof(xlog_in_core_t), KM_MAYFAIL);
-		if (!*iclogp)
+	for (i = 0; i < log->l_iclog_bufs; i++) {
+		size_t bvec_size = howmany(log->l_iclog_size, PAGE_SIZE) *
+				sizeof(struct bio_vec);
+
+		iclog = kmem_zalloc(sizeof(*iclog) + bvec_size, KM_MAYFAIL);
+		if (!iclog)
 			goto out_free_iclog;
 
-		iclog = *iclogp;
+		*iclogp = iclog;
 		iclog->ic_prev = prev_iclog;
 		prev_iclog = iclog;
 
-		bp = xfs_buf_get_uncached(mp->m_logdev_targp,
-					  BTOBB(log->l_iclog_size),
-					  XBF_NO_IOACCT);
-		if (!bp)
+		iclog->ic_data = kmem_alloc_large(log->l_iclog_size,
+				KM_MAYFAIL);
+		if (!iclog->ic_data)
 			goto out_free_iclog;
-
-		ASSERT(xfs_buf_islocked(bp));
-		xfs_buf_unlock(bp);
-
-		/* use high priority wq for log I/O completion */
-		bp->b_ioend_wq = mp->m_log_workqueue;
-		bp->b_iodone = xlog_iodone;
-		iclog->ic_bp = bp;
-		iclog->ic_data = bp->b_addr;
 #ifdef DEBUG
 		log->l_iclog_bak[i] = &iclog->ic_header;
 #endif
@@ -1552,36 +1435,43 @@ xlog_alloc_log(
 		head->h_fmt = cpu_to_be32(XLOG_FMT);
 		memcpy(&head->h_fs_uuid, &mp->m_sb.sb_uuid, sizeof(uuid_t));
 
-		iclog->ic_size = BBTOB(bp->b_length) - log->l_iclog_hsize;
+		iclog->ic_size = log->l_iclog_size - log->l_iclog_hsize;
 		iclog->ic_state = XLOG_STATE_ACTIVE;
 		iclog->ic_log = log;
 		atomic_set(&iclog->ic_refcnt, 0);
 		spin_lock_init(&iclog->ic_callback_lock);
-		iclog->ic_callback_tail = &(iclog->ic_callback);
+		INIT_LIST_HEAD(&iclog->ic_callbacks);
 		iclog->ic_datap = (char *)iclog->ic_data + log->l_iclog_hsize;
 
 		init_waitqueue_head(&iclog->ic_force_wait);
 		init_waitqueue_head(&iclog->ic_write_wait);
+		INIT_WORK(&iclog->ic_end_io_work, xlog_ioend_work);
+		sema_init(&iclog->ic_sema, 1);
 
 		iclogp = &iclog->ic_next;
 	}
 	*iclogp = log->l_iclog;			/* complete ring */
 	log->l_iclog->ic_prev = prev_iclog;	/* re-write 1st prev ptr */
 
+	log->l_ioend_workqueue = alloc_workqueue("xfs-log/%s",
+			WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_HIGHPRI, 0,
+			mp->m_fsname);
+	if (!log->l_ioend_workqueue)
+		goto out_free_iclog;
+
 	error = xlog_cil_init(log);
 	if (error)
-		goto out_free_iclog;
+		goto out_destroy_workqueue;
 	return log;
 
+out_destroy_workqueue:
+	destroy_workqueue(log->l_ioend_workqueue);
 out_free_iclog:
 	for (iclog = log->l_iclog; iclog; iclog = prev_iclog) {
 		prev_iclog = iclog->ic_next;
-		if (iclog->ic_bp)
-			xfs_buf_free(iclog->ic_bp);
+		kmem_free(iclog->ic_data);
 		kmem_free(iclog);
 	}
-	spinlock_destroy(&log->l_icloglock);
-	xfs_buf_free(log->l_xbuf);
 out_free_log:
 	kmem_free(log);
 out:
@@ -1766,42 +1656,155 @@ xlog_cksum(
 	return xfs_end_cksum(crc);
 }
 
-/*
- * The bdstrat callback function for log bufs. This gives us a central
- * place to trap bufs in case we get hit by a log I/O error and need to
- * shutdown. Actually, in practice, even when we didn't get a log error,
- * we transition the iclogs to IOERROR state *after* flushing all existing
- * iclogs to disk. This is because we don't want anymore new transactions to be
- * started or completed afterwards.
- *
- * We lock the iclogbufs here so that we can serialise against IO completion
- * during unmount. We might be processing a shutdown triggered during unmount,
- * and that can occur asynchronously to the unmount thread, and hence we need to
- * ensure that completes before tearing down the iclogbufs. Hence we need to
- * hold the buffer lock across the log IO to acheive that.
- */
-STATIC int
-xlog_bdstrat(
-	struct xfs_buf		*bp)
+static void
+xlog_bio_end_io(
+	struct bio		*bio)
 {
-	struct xlog_in_core	*iclog = bp->b_log_item;
+	struct xlog_in_core	*iclog = bio->bi_private;
 
-	xfs_buf_lock(bp);
-	if (iclog->ic_state & XLOG_STATE_IOERROR) {
-		xfs_buf_ioerror(bp, -EIO);
-		xfs_buf_stale(bp);
-		xfs_buf_ioend(bp);
+	queue_work(iclog->ic_log->l_ioend_workqueue,
+		   &iclog->ic_end_io_work);
+}
+
+static void
+xlog_map_iclog_data(
+	struct bio		*bio,
+	void			*data,
+	size_t			count)
+{
+	do {
+		struct page	*page = kmem_to_page(data);
+		unsigned int	off = offset_in_page(data);
+		size_t		len = min_t(size_t, count, PAGE_SIZE - off);
+
+		WARN_ON_ONCE(bio_add_page(bio, page, len, off) != len);
+
+		data += len;
+		count -= len;
+	} while (count);
+}
+
+STATIC void
+xlog_write_iclog(
+	struct xlog		*log,
+	struct xlog_in_core	*iclog,
+	uint64_t		bno,
+	unsigned int		count,
+	bool			need_flush)
+{
+	ASSERT(bno < log->l_logBBsize);
+
+	/*
+	 * We lock the iclogbufs here so that we can serialise against I/O
+	 * completion during unmount.  We might be processing a shutdown
+	 * triggered during unmount, and that can occur asynchronously to the
+	 * unmount thread, and hence we need to ensure that completes before
+	 * tearing down the iclogbufs.  Hence we need to hold the buffer lock
+	 * across the log IO to archieve that.
+	 */
+	down(&iclog->ic_sema);
+	if (unlikely(iclog->ic_state & XLOG_STATE_IOERROR)) {
 		/*
 		 * It would seem logical to return EIO here, but we rely on
 		 * the log state machine to propagate I/O errors instead of
-		 * doing it here. Similarly, IO completion will unlock the
-		 * buffer, so we don't do it here.
+		 * doing it here.  We kick of the state machine and unlock
+		 * the buffer manually, the code needs to be kept in sync
+		 * with the I/O completion path.
 		 */
-		return 0;
+		xlog_state_done_syncing(iclog, XFS_LI_ABORTED);
+		up(&iclog->ic_sema);
+		return;
 	}
 
-	xfs_buf_submit(bp);
-	return 0;
+	iclog->ic_io_size = count;
+
+	bio_init(&iclog->ic_bio, iclog->ic_bvec, howmany(count, PAGE_SIZE));
+	bio_set_dev(&iclog->ic_bio, log->l_targ->bt_bdev);
+	iclog->ic_bio.bi_iter.bi_sector = log->l_logBBstart + bno;
+	iclog->ic_bio.bi_end_io = xlog_bio_end_io;
+	iclog->ic_bio.bi_private = iclog;
+	iclog->ic_bio.bi_opf = REQ_OP_WRITE | REQ_META | REQ_SYNC | REQ_FUA;
+	if (need_flush)
+		iclog->ic_bio.bi_opf |= REQ_PREFLUSH;
+
+	xlog_map_iclog_data(&iclog->ic_bio, iclog->ic_data, iclog->ic_io_size);
+	if (is_vmalloc_addr(iclog->ic_data))
+		flush_kernel_vmap_range(iclog->ic_data, iclog->ic_io_size);
+
+	/*
+	 * If this log buffer would straddle the end of the log we will have
+	 * to split it up into two bios, so that we can continue at the start.
+	 */
+	if (bno + BTOBB(count) > log->l_logBBsize) {
+		struct bio *split;
+
+		split = bio_split(&iclog->ic_bio, log->l_logBBsize - bno,
+				  GFP_NOIO, &fs_bio_set);
+		bio_chain(split, &iclog->ic_bio);
+		submit_bio(split);
+
+		/* restart at logical offset zero for the remainder */
+		iclog->ic_bio.bi_iter.bi_sector = log->l_logBBstart;
+	}
+
+	submit_bio(&iclog->ic_bio);
+}
+
+/*
+ * We need to bump cycle number for the part of the iclog that is
+ * written to the start of the log. Watch out for the header magic
+ * number case, though.
+ */
+static void
+xlog_split_iclog(
+	struct xlog		*log,
+	void			*data,
+	uint64_t		bno,
+	unsigned int		count)
+{
+	unsigned int		split_offset = BBTOB(log->l_logBBsize - bno);
+	unsigned int		i;
+
+	for (i = split_offset; i < count; i += BBSIZE) {
+		uint32_t cycle = get_unaligned_be32(data + i);
+
+		if (++cycle == XLOG_HEADER_MAGIC_NUM)
+			cycle++;
+		put_unaligned_be32(cycle, data + i);
+	}
+}
+
+static int
+xlog_calc_iclog_size(
+	struct xlog		*log,
+	struct xlog_in_core	*iclog,
+	uint32_t		*roundoff)
+{
+	uint32_t		count_init, count;
+	bool			use_lsunit;
+
+	use_lsunit = xfs_sb_version_haslogv2(&log->l_mp->m_sb) &&
+			log->l_mp->m_sb.sb_logsunit > 1;
+
+	/* Add for LR header */
+	count_init = log->l_iclog_hsize + iclog->ic_offset;
+
+	/* Round out the log write size */
+	if (use_lsunit) {
+		/* we have a v2 stripe unit to use */
+		count = XLOG_LSUNITTOB(log, XLOG_BTOLSUNIT(log, count_init));
+	} else {
+		count = BBTOB(BTOBB(count_init));
+	}
+
+	ASSERT(count >= count_init);
+	*roundoff = count - count_init;
+
+	if (use_lsunit)
+		ASSERT(*roundoff < log->l_mp->m_sb.sb_logsunit);
+	else
+		ASSERT(*roundoff < BBTOB(1));
+	return count;
 }
 
 /*
@@ -1824,46 +1827,23 @@ xlog_bdstrat(
  * log will require grabbing the lock though.
  *
  * The entire log manager uses a logical block numbering scheme.  Only
- * log_sync (and then only bwrite()) know about the fact that the log may
- * not start with block zero on a given device.  The log block start offset
- * is added immediately before calling bwrite().
+ * xlog_write_iclog knows about the fact that the log may not start with
+ * block zero on a given device.
  */
-
-STATIC int
+STATIC void
 xlog_sync(
 	struct xlog		*log,
 	struct xlog_in_core	*iclog)
 {
-	xfs_buf_t	*bp;
-	int		i;
-	uint		count;		/* byte count of bwrite */
-	uint		count_init;	/* initial count before roundup */
-	int		roundoff;       /* roundoff to BB or stripe */
-	int		split = 0;	/* split write into two regions */
-	int		error;
-	int		v2 = xfs_sb_version_haslogv2(&log->l_mp->m_sb);
-	int		size;
+	unsigned int		count;		/* byte count of bwrite */
+	unsigned int		roundoff;       /* roundoff to BB or stripe */
+	uint64_t		bno;
+	unsigned int		size;
+	bool			need_flush = true, split = false;
 
-	XFS_STATS_INC(log->l_mp, xs_log_writes);
 	ASSERT(atomic_read(&iclog->ic_refcnt) == 0);
 
-	/* Add for LR header */
-	count_init = log->l_iclog_hsize + iclog->ic_offset;
-
-	/* Round out the log write size */
-	if (v2 && log->l_mp->m_sb.sb_logsunit > 1) {
-		/* we have a v2 stripe unit to use */
-		count = XLOG_LSUNITTOB(log, XLOG_BTOLSUNIT(log, count_init));
-	} else {
-		count = BBTOB(BTOBB(count_init));
-	}
-	roundoff = count - count_init;
-	ASSERT(roundoff >= 0);
-	ASSERT((v2 && log->l_mp->m_sb.sb_logsunit > 1 && 
-                roundoff < log->l_mp->m_sb.sb_logsunit)
-		|| 
-		(log->l_mp->m_sb.sb_logsunit <= 1 && 
-		 roundoff < BBTOB(1)));
+	count = xlog_calc_iclog_size(log, iclog, &roundoff);
 
 	/* move grant heads by roundoff in sync */
 	xlog_grant_add_space(log, &log->l_reserve_head.grant, roundoff);
@@ -1874,41 +1854,19 @@ xlog_sync(
 
 	/* real byte length */
 	size = iclog->ic_offset;
-	if (v2)
+	if (xfs_sb_version_haslogv2(&log->l_mp->m_sb))
 		size += roundoff;
 	iclog->ic_header.h_len = cpu_to_be32(size);
 
-	bp = iclog->ic_bp;
-	XFS_BUF_SET_ADDR(bp, BLOCK_LSN(be64_to_cpu(iclog->ic_header.h_lsn)));
-
+	XFS_STATS_INC(log->l_mp, xs_log_writes);
 	XFS_STATS_ADD(log->l_mp, xs_log_blocks, BTOBB(count));
 
-	/* Do we need to split this write into 2 parts? */
-	if (XFS_BUF_ADDR(bp) + BTOBB(count) > log->l_logBBsize) {
-		char		*dptr;
-
-		split = count - (BBTOB(log->l_logBBsize - XFS_BUF_ADDR(bp)));
-		count = BBTOB(log->l_logBBsize - XFS_BUF_ADDR(bp));
-		iclog->ic_bwritecnt = 2;
+	bno = BLOCK_LSN(be64_to_cpu(iclog->ic_header.h_lsn));
 
-		/*
-		 * Bump the cycle numbers at the start of each block in the
-		 * part of the iclog that ends up in the buffer that gets
-		 * written to the start of the log.
-		 *
-		 * Watch out for the header magic number case, though.
-		 */
-		dptr = (char *)&iclog->ic_header + count;
-		for (i = 0; i < split; i += BBSIZE) {
-			uint32_t cycle = be32_to_cpu(*(__be32 *)dptr);
-			if (++cycle == XLOG_HEADER_MAGIC_NUM)
-				cycle++;
-			*(__be32 *)dptr = cpu_to_be32(cycle);
-
-			dptr += BBSIZE;
-		}
-	} else {
-		iclog->ic_bwritecnt = 1;
+	/* Do we need to split this write into 2 parts? */
+	if (bno + BTOBB(count) > log->l_logBBsize) {
+		xlog_split_iclog(log, &iclog->ic_header, bno, count);
+		split = true;
 	}
 
 	/* calculcate the checksum */
@@ -1921,18 +1879,15 @@ xlog_sync(
 	 * write on I/O completion and shutdown the fs. The subsequent mount
 	 * detects the bad CRC and attempts to recover.
 	 */
+#ifdef DEBUG
 	if (XFS_TEST_ERROR(false, log->l_mp, XFS_ERRTAG_LOG_BAD_CRC)) {
 		iclog->ic_header.h_crc &= cpu_to_le32(0xAAAAAAAA);
-		iclog->ic_state |= XLOG_STATE_IOABORT;
+		iclog->ic_fail_crc = true;
 		xfs_warn(log->l_mp,
 	"Intentionally corrupted log record at LSN 0x%llx. Shutdown imminent.",
 			 be64_to_cpu(iclog->ic_header.h_lsn));
 	}
-
-	bp->b_io_length = BTOBB(count);
-	bp->b_log_item = iclog;
-	bp->b_flags &= ~XBF_FLUSH;
-	bp->b_flags |= (XBF_ASYNC | XBF_SYNCIO | XBF_WRITE | XBF_FUA);
+#endif
 
 	/*
 	 * Flush the data device before flushing the log to make sure all meta
@@ -1942,50 +1897,14 @@ xlog_sync(
 	 * synchronously here; for an internal log we can simply use the block
 	 * layer state machine for preflushes.
 	 */
-	if (log->l_mp->m_logdev_targp != log->l_mp->m_ddev_targp)
+	if (log->l_targ != log->l_mp->m_ddev_targp || split) {
 		xfs_blkdev_issue_flush(log->l_mp->m_ddev_targp);
-	else
-		bp->b_flags |= XBF_FLUSH;
-
-	ASSERT(XFS_BUF_ADDR(bp) <= log->l_logBBsize-1);
-	ASSERT(XFS_BUF_ADDR(bp) + BTOBB(count) <= log->l_logBBsize);
-
-	xlog_verify_iclog(log, iclog, count, true);
-
-	/* account for log which doesn't start at block #0 */
-	XFS_BUF_SET_ADDR(bp, XFS_BUF_ADDR(bp) + log->l_logBBstart);
-
-	/*
-	 * Don't call xfs_bwrite here. We do log-syncs even when the filesystem
-	 * is shutting down.
-	 */
-	error = xlog_bdstrat(bp);
-	if (error) {
-		xfs_buf_ioerror_alert(bp, "xlog_sync");
-		return error;
+		need_flush = false;
 	}
-	if (split) {
-		bp = iclog->ic_log->l_xbuf;
-		XFS_BUF_SET_ADDR(bp, 0);	     /* logical 0 */
-		xfs_buf_associate_memory(bp,
-				(char *)&iclog->ic_header + count, split);
-		bp->b_log_item = iclog;
-		bp->b_flags &= ~XBF_FLUSH;
-		bp->b_flags |= (XBF_ASYNC | XBF_SYNCIO | XBF_WRITE | XBF_FUA);
-
-		ASSERT(XFS_BUF_ADDR(bp) <= log->l_logBBsize-1);
-		ASSERT(XFS_BUF_ADDR(bp) + BTOBB(count) <= log->l_logBBsize);
-
-		/* account for internal log which doesn't start at block #0 */
-		XFS_BUF_SET_ADDR(bp, XFS_BUF_ADDR(bp) + log->l_logBBstart);
-		error = xlog_bdstrat(bp);
-		if (error) {
-			xfs_buf_ioerror_alert(bp, "xlog_sync (split)");
-			return error;
-		}
-	}
-	return 0;
-}	/* xlog_sync */
+
+	xlog_verify_iclog(log, iclog, count);
+	xlog_write_iclog(log, iclog, bno, count, need_flush);
+}
 
 /*
  * Deallocate a log structure
@@ -2005,31 +1924,21 @@ xlog_dealloc_log(
 	 */
 	iclog = log->l_iclog;
 	for (i = 0; i < log->l_iclog_bufs; i++) {
-		xfs_buf_lock(iclog->ic_bp);
-		xfs_buf_unlock(iclog->ic_bp);
+		down(&iclog->ic_sema);
+		up(&iclog->ic_sema);
 		iclog = iclog->ic_next;
 	}
 
-	/*
-	 * Always need to ensure that the extra buffer does not point to memory
-	 * owned by another log buffer before we free it. Also, cycle the lock
-	 * first to ensure we've completed IO on it.
-	 */
-	xfs_buf_lock(log->l_xbuf);
-	xfs_buf_unlock(log->l_xbuf);
-	xfs_buf_set_empty(log->l_xbuf, BTOBB(log->l_iclog_size));
-	xfs_buf_free(log->l_xbuf);
-
 	iclog = log->l_iclog;
 	for (i = 0; i < log->l_iclog_bufs; i++) {
-		xfs_buf_free(iclog->ic_bp);
 		next_iclog = iclog->ic_next;
+		kmem_free(iclog->ic_data);
 		kmem_free(iclog);
 		iclog = next_iclog;
 	}
-	spinlock_destroy(&log->l_icloglock);
 
 	log->l_mp->m_log = NULL;
+	destroy_workqueue(log->l_ioend_workqueue);
 	kmem_free(log);
 }	/* xlog_dealloc_log */
 
@@ -2610,7 +2519,7 @@ xlog_state_clean_log(
 		if (iclog->ic_state == XLOG_STATE_DIRTY) {
 			iclog->ic_state	= XLOG_STATE_ACTIVE;
 			iclog->ic_offset       = 0;
-			ASSERT(iclog->ic_callback == NULL);
+			ASSERT(list_empty_careful(&iclog->ic_callbacks));
 			/*
 			 * If the number of ops in this iclog indicate it just
 			 * contains the dummy transaction, we can
@@ -2680,37 +2589,32 @@ xlog_state_clean_log(
 
 STATIC xfs_lsn_t
 xlog_get_lowest_lsn(
-	struct xlog	*log)
+	struct xlog		*log)
 {
-	xlog_in_core_t  *lsn_log;
-	xfs_lsn_t	lowest_lsn, lsn;
+	struct xlog_in_core	*iclog = log->l_iclog;
+	xfs_lsn_t		lowest_lsn = 0, lsn;
 
-	lsn_log = log->l_iclog;
-	lowest_lsn = 0;
 	do {
-	    if (!(lsn_log->ic_state & (XLOG_STATE_ACTIVE|XLOG_STATE_DIRTY))) {
-		lsn = be64_to_cpu(lsn_log->ic_header.h_lsn);
-		if ((lsn && !lowest_lsn) ||
-		    (XFS_LSN_CMP(lsn, lowest_lsn) < 0)) {
+		if (iclog->ic_state & (XLOG_STATE_ACTIVE | XLOG_STATE_DIRTY))
+			continue;
+
+		lsn = be64_to_cpu(iclog->ic_header.h_lsn);
+		if ((lsn && !lowest_lsn) || XFS_LSN_CMP(lsn, lowest_lsn) < 0)
 			lowest_lsn = lsn;
-		}
-	    }
-	    lsn_log = lsn_log->ic_next;
-	} while (lsn_log != log->l_iclog);
+	} while ((iclog = iclog->ic_next) != log->l_iclog);
+
 	return lowest_lsn;
 }
 
-
 STATIC void
 xlog_state_do_callback(
 	struct xlog		*log,
-	int			aborted,
+	bool			aborted,
 	struct xlog_in_core	*ciclog)
 {
 	xlog_in_core_t	   *iclog;
 	xlog_in_core_t	   *first_iclog;	/* used to know when we've
 						 * processed all iclogs once */
-	xfs_log_callback_t *cb, *cb_next;
 	int		   flushcnt = 0;
 	xfs_lsn_t	   lowest_lsn;
 	int		   ioerrors;	/* counter: iclogs with errors */
@@ -2821,7 +2725,7 @@ xlog_state_do_callback(
 				 */
 				ASSERT(XFS_LSN_CMP(atomic64_read(&log->l_last_sync_lsn),
 					be64_to_cpu(iclog->ic_header.h_lsn)) <= 0);
-				if (iclog->ic_callback)
+				if (!list_empty_careful(&iclog->ic_callbacks))
 					atomic64_set(&log->l_last_sync_lsn,
 						be64_to_cpu(iclog->ic_header.h_lsn));
 
@@ -2838,26 +2742,20 @@ xlog_state_do_callback(
 			 * callbacks being added.
 			 */
 			spin_lock(&iclog->ic_callback_lock);
-			cb = iclog->ic_callback;
-			while (cb) {
-				iclog->ic_callback_tail = &(iclog->ic_callback);
-				iclog->ic_callback = NULL;
-				spin_unlock(&iclog->ic_callback_lock);
+			while (!list_empty(&iclog->ic_callbacks)) {
+				LIST_HEAD(tmp);
 
-				/* perform callbacks in the order given */
-				for (; cb; cb = cb_next) {
-					cb_next = cb->cb_next;
-					cb->cb_func(cb->cb_arg, aborted);
-				}
+				list_splice_init(&iclog->ic_callbacks, &tmp);
+
+				spin_unlock(&iclog->ic_callback_lock);
+				xlog_cil_process_committed(&tmp, aborted);
 				spin_lock(&iclog->ic_callback_lock);
-				cb = iclog->ic_callback;
 			}
 
 			loopdidcallbacks++;
 			funcdidcallbacks++;
 
 			spin_lock(&log->l_icloglock);
-			ASSERT(iclog->ic_callback == NULL);
 			spin_unlock(&iclog->ic_callback_lock);
 			if (!(iclog->ic_state & XLOG_STATE_IOERROR))
 				iclog->ic_state = XLOG_STATE_DIRTY;
@@ -2943,18 +2841,16 @@ xlog_state_do_callback(
  */
 STATIC void
 xlog_state_done_syncing(
-	xlog_in_core_t	*iclog,
-	int		aborted)
+	struct xlog_in_core	*iclog,
+	bool			aborted)
 {
-	struct xlog	   *log = iclog->ic_log;
+	struct xlog		*log = iclog->ic_log;
 
 	spin_lock(&log->l_icloglock);
 
 	ASSERT(iclog->ic_state == XLOG_STATE_SYNCING ||
 	       iclog->ic_state == XLOG_STATE_IOERROR);
 	ASSERT(atomic_read(&iclog->ic_refcnt) == 0);
-	ASSERT(iclog->ic_bwritecnt == 1 || iclog->ic_bwritecnt == 2);
-
 
 	/*
 	 * If we got an error, either on the first buffer, or in the case of
@@ -2962,13 +2858,8 @@ xlog_state_done_syncing(
 	 * and none should ever be attempted to be written to disk
 	 * again.
 	 */
-	if (iclog->ic_state != XLOG_STATE_IOERROR) {
-		if (--iclog->ic_bwritecnt == 1) {
-			spin_unlock(&log->l_icloglock);
-			return;
-		}
+	if (iclog->ic_state != XLOG_STATE_IOERROR)
 		iclog->ic_state = XLOG_STATE_DONE_SYNC;
-	}
 
 	/*
 	 * Someone could be sleeping prior to writing out the next
@@ -3237,7 +3128,7 @@ xlog_state_release_iclog(
 	 * flags after this point.
 	 */
 	if (sync)
-		return xlog_sync(log, iclog);
+		xlog_sync(log, iclog);
 	return 0;
 }	/* xlog_state_release_iclog */
 
@@ -3828,8 +3719,7 @@ STATIC void
 xlog_verify_iclog(
 	struct xlog		*log,
 	struct xlog_in_core	*iclog,
-	int			count,
-	bool                    syncing)
+	int			count)
 {
 	xlog_op_header_t	*ophead;
 	xlog_in_core_t		*icptr;
@@ -3873,7 +3763,7 @@ xlog_verify_iclog(
 		/* clientid is only 1 byte */
 		p = &ophead->oh_clientid;
 		field_offset = p - base_ptr;
-		if (!syncing || (field_offset & 0x1ff)) {
+		if (field_offset & 0x1ff) {
 			clientid = ophead->oh_clientid;
 		} else {
 			idx = BTOBBT((char *)&ophead->oh_clientid - iclog->ic_datap);
@@ -3896,7 +3786,7 @@ xlog_verify_iclog(
 		/* check length */
 		p = &ophead->oh_len;
 		field_offset = p - base_ptr;
-		if (!syncing || (field_offset & 0x1ff)) {
+		if (field_offset & 0x1ff) {
 			op_len = be32_to_cpu(ophead->oh_len);
 		} else {
 			idx = BTOBBT((uintptr_t)&ophead->oh_len -
@@ -4033,7 +3923,7 @@ xfs_log_force_umount(
 	 * avoid races.
 	 */
 	wake_up_all(&log->l_cilp->xc_commit_wait);
-	xlog_state_do_callback(log, XFS_LI_ABORTED, NULL);
+	xlog_state_do_callback(log, true, NULL);
 
 #ifdef XFSERRORDEBUG
 	{