#include "xfs_inode_item.h"
#include "xfs_rw.h"
+#include <linux/kthread.h>
+#include <linux/freezer.h>
+
/*
- * xfs_sync flushes any pending I/O to file system vfsp.
- *
- * This routine is called by vfs_sync() to make sure that things make it
- * out to disk eventually, on sync() system calls to flush out everything,
- * and when the file system is unmounted. For the vfs_sync() case, all
- * we really need to do is sync out the log to make all of our meta-data
- * updates permanent (except for timestamps). For calls from pflushd(),
- * dirty pages are kept moving by calling pdflush() on the inodes
- * containing them. We also flush the inodes that we can lock without
- * sleeping and the superblock if we can lock it without sleeping from
- * vfs_sync() so that items at the tail of the log are always moving out.
- *
- * Flags:
- * SYNC_BDFLUSH - We're being called from vfs_sync() so we don't want
- * to sleep if we can help it. All we really need
- * to do is ensure that the log is synced at least
- * periodically. We also push the inodes and
- * superblock if we can lock them without sleeping
- * and they are not pinned.
- * SYNC_ATTR - We need to flush the inodes. If SYNC_BDFLUSH is not
- * set, then we really want to lock each inode and flush
- * it.
- * SYNC_WAIT - All the flushes that take place in this call should
- * be synchronous.
- * SYNC_DELWRI - This tells us to push dirty pages associated with
- * inodes. SYNC_WAIT and SYNC_BDFLUSH are used to
- * determine if they should be flushed sync, async, or
- * delwri.
- * SYNC_CLOSE - This flag is passed when the system is being
- * unmounted. We should sync and invalidate everything.
- * SYNC_FSDATA - This indicates that the caller would like to make
- * sure the superblock is safe on disk. We can ensure
- * this by simply making sure the log gets flushed
- * if SYNC_BDFLUSH is set, and by actually writing it
- * out otherwise.
- * SYNC_IOWAIT - The caller wants us to wait for all data I/O to complete
- * before we return (including direct I/O). Forms the drain
- * side of the write barrier needed to safely quiesce the
- * filesystem.
- *
+ * Sync all the inodes in the given AG according to the
+ * direction given by the flags.
*/
-int
-xfs_sync(
+STATIC int
+xfs_sync_inodes_ag(
xfs_mount_t *mp,
+ int ag,
int flags)
{
- int error;
+ xfs_perag_t *pag = &mp->m_perag[ag];
+ int nr_found;
+ uint32_t first_index = 0;
+ int error = 0;
+ int last_error = 0;
+ int fflag = XFS_B_ASYNC;
+
+ if (flags & SYNC_DELWRI)
+ fflag = XFS_B_DELWRI;
+ if (flags & SYNC_WAIT)
+ fflag = 0; /* synchronous overrides all */
+
+ do {
+ struct inode *inode;
+ xfs_inode_t *ip = NULL;
+ int lock_flags = XFS_ILOCK_SHARED;
+
+ /*
+ * use a gang lookup to find the next inode in the tree
+ * as the tree is sparse and a gang lookup walks to find
+ * the number of objects requested.
+ */
+ read_lock(&pag->pag_ici_lock);
+ nr_found = radix_tree_gang_lookup(&pag->pag_ici_root,
+ (void**)&ip, first_index, 1);
+
+ if (!nr_found) {
+ read_unlock(&pag->pag_ici_lock);
+ break;
+ }
+
+ /*
+ * Update the index for the next lookup. Catch overflows
+ * into the next AG range which can occur if we have inodes
+ * in the last block of the AG and we are currently
+ * pointing to the last inode.
+ */
+ first_index = XFS_INO_TO_AGINO(mp, ip->i_ino + 1);
+ if (first_index < XFS_INO_TO_AGINO(mp, ip->i_ino)) {
+ read_unlock(&pag->pag_ici_lock);
+ break;
+ }
+
+ /* nothing to sync during shutdown */
+ if (XFS_FORCED_SHUTDOWN(mp)) {
+ read_unlock(&pag->pag_ici_lock);
+ return 0;
+ }
+
+ /*
+ * If we can't get a reference on the inode, it must be
+ * in reclaim. Leave it for the reclaim code to flush.
+ */
+ inode = VFS_I(ip);
+ if (!igrab(inode)) {
+ read_unlock(&pag->pag_ici_lock);
+ continue;
+ }
+ read_unlock(&pag->pag_ici_lock);
+
+ /* avoid new or bad inodes */
+ if (is_bad_inode(inode) ||
+ xfs_iflags_test(ip, XFS_INEW)) {
+ IRELE(ip);
+ continue;
+ }
- /*
- * Get the Quota Manager to flush the dquots.
- *
- * If XFS quota support is not enabled or this filesystem
- * instance does not use quotas XFS_QM_DQSYNC will always
- * return zero.
- */
- error = XFS_QM_DQSYNC(mp, flags);
- if (error) {
/*
- * If we got an IO error, we will be shutting down.
- * So, there's nothing more for us to do here.
+ * If we have to flush data or wait for I/O completion
+ * we need to hold the iolock.
*/
- ASSERT(error != EIO || XFS_FORCED_SHUTDOWN(mp));
- if (XFS_FORCED_SHUTDOWN(mp))
+ if ((flags & SYNC_DELWRI) && VN_DIRTY(inode)) {
+ xfs_ilock(ip, XFS_IOLOCK_SHARED);
+ lock_flags |= XFS_IOLOCK_SHARED;
+ error = xfs_flush_pages(ip, 0, -1, fflag, FI_NONE);
+ if (flags & SYNC_IOWAIT)
+ xfs_ioend_wait(ip);
+ }
+ xfs_ilock(ip, XFS_ILOCK_SHARED);
+
+ if ((flags & SYNC_ATTR) && !xfs_inode_clean(ip)) {
+ if (flags & SYNC_WAIT) {
+ xfs_iflock(ip);
+ if (!xfs_inode_clean(ip))
+ error = xfs_iflush(ip, XFS_IFLUSH_SYNC);
+ else
+ xfs_ifunlock(ip);
+ } else if (xfs_iflock_nowait(ip)) {
+ if (!xfs_inode_clean(ip))
+ error = xfs_iflush(ip, XFS_IFLUSH_DELWRI);
+ else
+ xfs_ifunlock(ip);
+ }
+ }
+ xfs_iput(ip, lock_flags);
+
+ if (error)
+ last_error = error;
+ /*
+ * bail out if the filesystem is corrupted.
+ */
+ if (error == EFSCORRUPTED)
return XFS_ERROR(error);
- }
- if (flags & SYNC_IOWAIT)
- xfs_filestream_flush(mp);
+ } while (nr_found);
- return xfs_syncsub(mp, flags, NULL);
+ return last_error;
}
-/*
- * xfs sync routine for internal use
- *
- * This routine supports all of the flags defined for the generic vfs_sync
- * interface as explained above under xfs_sync.
- *
- */
int
xfs_sync_inodes(
xfs_mount_t *mp,
- int flags,
- int *bypassed)
+ int flags)
{
- xfs_inode_t *ip = NULL;
- struct inode *vp = NULL;
int error;
int last_error;
- uint64_t fflag;
- uint lock_flags;
- uint base_lock_flags;
- boolean_t mount_locked;
- boolean_t vnode_refed;
- int preempt;
- xfs_iptr_t *ipointer;
-#ifdef DEBUG
- boolean_t ipointer_in = B_FALSE;
-
-#define IPOINTER_SET ipointer_in = B_TRUE
-#define IPOINTER_CLR ipointer_in = B_FALSE
-#else
-#define IPOINTER_SET
-#define IPOINTER_CLR
-#endif
-
-
-/* Insert a marker record into the inode list after inode ip. The list
- * must be locked when this is called. After the call the list will no
- * longer be locked.
- */
-#define IPOINTER_INSERT(ip, mp) { \
- ASSERT(ipointer_in == B_FALSE); \
- ipointer->ip_mnext = ip->i_mnext; \
- ipointer->ip_mprev = ip; \
- ip->i_mnext = (xfs_inode_t *)ipointer; \
- ipointer->ip_mnext->i_mprev = (xfs_inode_t *)ipointer; \
- preempt = 0; \
- XFS_MOUNT_IUNLOCK(mp); \
- mount_locked = B_FALSE; \
- IPOINTER_SET; \
- }
-
-/* Remove the marker from the inode list. If the marker was the only item
- * in the list then there are no remaining inodes and we should zero out
- * the whole list. If we are the current head of the list then move the head
- * past us.
- */
-#define IPOINTER_REMOVE(ip, mp) { \
- ASSERT(ipointer_in == B_TRUE); \
- if (ipointer->ip_mnext != (xfs_inode_t *)ipointer) { \
- ip = ipointer->ip_mnext; \
- ip->i_mprev = ipointer->ip_mprev; \
- ipointer->ip_mprev->i_mnext = ip; \
- if (mp->m_inodes == (xfs_inode_t *)ipointer) { \
- mp->m_inodes = ip; \
- } \
- } else { \
- ASSERT(mp->m_inodes == (xfs_inode_t *)ipointer); \
- mp->m_inodes = NULL; \
- ip = NULL; \
- } \
- IPOINTER_CLR; \
- }
-
-#define XFS_PREEMPT_MASK 0x7f
+ int i;
+ int lflags = XFS_LOG_FORCE;
- ASSERT(!(flags & SYNC_BDFLUSH));
-
- if (bypassed)
- *bypassed = 0;
if (mp->m_flags & XFS_MOUNT_RDONLY)
return 0;
error = 0;
last_error = 0;
- preempt = 0;
- /* Allocate a reference marker */
- ipointer = (xfs_iptr_t *)kmem_zalloc(sizeof(xfs_iptr_t), KM_SLEEP);
+ if (flags & SYNC_WAIT)
+ lflags |= XFS_LOG_SYNC;
- fflag = XFS_B_ASYNC; /* default is don't wait */
+ for (i = 0; i < mp->m_sb.sb_agcount; i++) {
+ if (!mp->m_perag[i].pag_ici_init)
+ continue;
+ error = xfs_sync_inodes_ag(mp, i, flags);
+ if (error)
+ last_error = error;
+ if (error == EFSCORRUPTED)
+ break;
+ }
if (flags & SYNC_DELWRI)
- fflag = XFS_B_DELWRI;
- if (flags & SYNC_WAIT)
- fflag = 0; /* synchronous overrides all */
+ xfs_log_force(mp, 0, lflags);
- base_lock_flags = XFS_ILOCK_SHARED;
- if (flags & (SYNC_DELWRI | SYNC_CLOSE)) {
- /*
- * We need the I/O lock if we're going to call any of
- * the flush/inval routines.
- */
- base_lock_flags |= XFS_IOLOCK_SHARED;
+ return XFS_ERROR(last_error);
+}
+
+STATIC int
+xfs_commit_dummy_trans(
+ struct xfs_mount *mp,
+ uint log_flags)
+{
+ struct xfs_inode *ip = mp->m_rootip;
+ struct xfs_trans *tp;
+ int error;
+
+ /*
+ * Put a dummy transaction in the log to tell recovery
+ * that all others are OK.
+ */
+ tp = xfs_trans_alloc(mp, XFS_TRANS_DUMMY1);
+ error = xfs_trans_reserve(tp, 0, XFS_ICHANGE_LOG_RES(mp), 0, 0, 0);
+ if (error) {
+ xfs_trans_cancel(tp, 0);
+ return error;
}
- XFS_MOUNT_ILOCK(mp);
+ xfs_ilock(ip, XFS_ILOCK_EXCL);
- ip = mp->m_inodes;
+ xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
+ xfs_trans_ihold(tp, ip);
+ xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
+ /* XXX(hch): ignoring the error here.. */
+ error = xfs_trans_commit(tp, 0);
- mount_locked = B_TRUE;
- vnode_refed = B_FALSE;
+ xfs_iunlock(ip, XFS_ILOCK_EXCL);
- IPOINTER_CLR;
+ xfs_log_force(mp, 0, log_flags);
+ return 0;
+}
- do {
- ASSERT(ipointer_in == B_FALSE);
- ASSERT(vnode_refed == B_FALSE);
+int
+xfs_sync_fsdata(
+ struct xfs_mount *mp,
+ int flags)
+{
+ struct xfs_buf *bp;
+ struct xfs_buf_log_item *bip;
+ int error = 0;
- lock_flags = base_lock_flags;
+ /*
+ * If this is xfssyncd() then only sync the superblock if we can
+ * lock it without sleeping and it is not pinned.
+ */
+ if (flags & SYNC_BDFLUSH) {
+ ASSERT(!(flags & SYNC_WAIT));
- /*
- * There were no inodes in the list, just break out
- * of the loop.
- */
- if (ip == NULL) {
- break;
- }
+ bp = xfs_getsb(mp, XFS_BUF_TRYLOCK);
+ if (!bp)
+ goto out;
+
+ bip = XFS_BUF_FSPRIVATE(bp, struct xfs_buf_log_item *);
+ if (!bip || !xfs_buf_item_dirty(bip) || XFS_BUF_ISPINNED(bp))
+ goto out_brelse;
+ } else {
+ bp = xfs_getsb(mp, 0);
/*
- * We found another sync thread marker - skip it
+ * If the buffer is pinned then push on the log so we won't
+ * get stuck waiting in the write for someone, maybe
+ * ourselves, to flush the log.
+ *
+ * Even though we just pushed the log above, we did not have
+ * the superblock buffer locked at that point so it can
+ * become pinned in between there and here.
*/
- if (ip->i_mount == NULL) {
- ip = ip->i_mnext;
- continue;
- }
+ if (XFS_BUF_ISPINNED(bp))
+ xfs_log_force(mp, 0, XFS_LOG_FORCE);
+ }
- vp = VFS_I(ip);
- /*
- * If the vnode is gone then this is being torn down,
- * call reclaim if it is flushed, else let regular flush
- * code deal with it later in the loop.
- */
+ if (flags & SYNC_WAIT)
+ XFS_BUF_UNASYNC(bp);
+ else
+ XFS_BUF_ASYNC(bp);
- if (vp == NULL) {
- /* Skip ones already in reclaim */
- if (ip->i_flags & XFS_IRECLAIM) {
- ip = ip->i_mnext;
- continue;
- }
- if (xfs_ilock_nowait(ip, XFS_ILOCK_EXCL) == 0) {
- ip = ip->i_mnext;
- } else if ((xfs_ipincount(ip) == 0) &&
- xfs_iflock_nowait(ip)) {
- IPOINTER_INSERT(ip, mp);
-
- xfs_finish_reclaim(ip, 1,
- XFS_IFLUSH_DELWRI_ELSE_ASYNC);
-
- XFS_MOUNT_ILOCK(mp);
- mount_locked = B_TRUE;
- IPOINTER_REMOVE(ip, mp);
- } else {
- xfs_iunlock(ip, XFS_ILOCK_EXCL);
- ip = ip->i_mnext;
- }
- continue;
- }
+ return xfs_bwrite(mp, bp);
- if (VN_BAD(vp)) {
- ip = ip->i_mnext;
- continue;
- }
+ out_brelse:
+ xfs_buf_relse(bp);
+ out:
+ return error;
+}
- if (XFS_FORCED_SHUTDOWN(mp) && !(flags & SYNC_CLOSE)) {
- XFS_MOUNT_IUNLOCK(mp);
- kmem_free(ipointer);
- return 0;
- }
+/*
+ * When remounting a filesystem read-only or freezing the filesystem, we have
+ * two phases to execute. This first phase is syncing the data before we
+ * quiesce the filesystem, and the second is flushing all the inodes out after
+ * we've waited for all the transactions created by the first phase to
+ * complete. The second phase ensures that the inodes are written to their
+ * location on disk rather than just existing in transactions in the log. This
+ * means after a quiesce there is no log replay required to write the inodes to
+ * disk (this is the main difference between a sync and a quiesce).
+ */
+/*
+ * First stage of freeze - no writers will make progress now we are here,
+ * so we flush delwri and delalloc buffers here, then wait for all I/O to
+ * complete. Data is frozen at that point. Metadata is not frozen,
+ * transactions can still occur here so don't bother flushing the buftarg
+ * because it'll just get dirty again.
+ */
+int
+xfs_quiesce_data(
+ struct xfs_mount *mp)
+{
+ int error;
- /*
- * Try to lock without sleeping. We're out of order with
- * the inode list lock here, so if we fail we need to drop
- * the mount lock and try again. If we're called from
- * bdflush() here, then don't bother.
- *
- * The inode lock here actually coordinates with the
- * almost spurious inode lock in xfs_ireclaim() to prevent
- * the vnode we handle here without a reference from
- * being freed while we reference it. If we lock the inode
- * while it's on the mount list here, then the spurious inode
- * lock in xfs_ireclaim() after the inode is pulled from
- * the mount list will sleep until we release it here.
- * This keeps the vnode from being freed while we reference
- * it.
- */
- if (xfs_ilock_nowait(ip, lock_flags) == 0) {
- if (vp == NULL) {
- ip = ip->i_mnext;
- continue;
- }
+ /* push non-blocking */
+ xfs_sync_inodes(mp, SYNC_DELWRI|SYNC_BDFLUSH);
+ XFS_QM_DQSYNC(mp, SYNC_BDFLUSH);
+ xfs_filestream_flush(mp);
- vp = vn_grab(vp);
- if (vp == NULL) {
- ip = ip->i_mnext;
- continue;
- }
+ /* push and block */
+ xfs_sync_inodes(mp, SYNC_DELWRI|SYNC_WAIT|SYNC_IOWAIT);
+ XFS_QM_DQSYNC(mp, SYNC_WAIT);
- IPOINTER_INSERT(ip, mp);
- xfs_ilock(ip, lock_flags);
+ /* write superblock and hoover up shutdown errors */
+ error = xfs_sync_fsdata(mp, 0);
- ASSERT(vp == VFS_I(ip));
- ASSERT(ip->i_mount == mp);
+ /* flush data-only devices */
+ if (mp->m_rtdev_targp)
+ XFS_bflush(mp->m_rtdev_targp);
- vnode_refed = B_TRUE;
+ return error;
+}
+
+STATIC void
+xfs_quiesce_fs(
+ struct xfs_mount *mp)
+{
+ int count = 0, pincount;
+
+ xfs_flush_buftarg(mp->m_ddev_targp, 0);
+ xfs_reclaim_inodes(mp, 0, XFS_IFLUSH_DELWRI_ELSE_ASYNC);
+
+ /*
+ * This loop must run at least twice. The first instance of the loop
+ * will flush most meta data but that will generate more meta data
+ * (typically directory updates). Which then must be flushed and
+ * logged before we can write the unmount record.
+ */
+ do {
+ xfs_sync_inodes(mp, SYNC_ATTR|SYNC_WAIT);
+ pincount = xfs_flush_buftarg(mp->m_ddev_targp, 1);
+ if (!pincount) {
+ delay(50);
+ count++;
}
+ } while (count < 2);
+}
- /* From here on in the loop we may have a marker record
- * in the inode list.
- */
+/*
+ * Second stage of a quiesce. The data is already synced, now we have to take
+ * care of the metadata. New transactions are already blocked, so we need to
+ * wait for any remaining transactions to drain out before proceding.
+ */
+void
+xfs_quiesce_attr(
+ struct xfs_mount *mp)
+{
+ int error = 0;
- /*
- * If we have to flush data or wait for I/O completion
- * we need to drop the ilock that we currently hold.
- * If we need to drop the lock, insert a marker if we
- * have not already done so.
- */
- if ((flags & (SYNC_CLOSE|SYNC_IOWAIT)) ||
- ((flags & SYNC_DELWRI) && VN_DIRTY(vp))) {
- if (mount_locked) {
- IPOINTER_INSERT(ip, mp);
- }
- xfs_iunlock(ip, XFS_ILOCK_SHARED);
+ /* wait for all modifications to complete */
+ while (atomic_read(&mp->m_active_trans) > 0)
+ delay(100);
- if (flags & SYNC_CLOSE) {
- /* Shutdown case. Flush and invalidate. */
- if (XFS_FORCED_SHUTDOWN(mp))
- xfs_tosspages(ip, 0, -1,
- FI_REMAPF);
- else
- error = xfs_flushinval_pages(ip,
- 0, -1, FI_REMAPF);
- } else if ((flags & SYNC_DELWRI) && VN_DIRTY(vp)) {
- error = xfs_flush_pages(ip, 0,
- -1, fflag, FI_NONE);
- }
+ /* flush inodes and push all remaining buffers out to disk */
+ xfs_quiesce_fs(mp);
- /*
- * When freezing, we need to wait ensure all I/O (including direct
- * I/O) is complete to ensure no further data modification can take
- * place after this point
- */
- if (flags & SYNC_IOWAIT)
- vn_iowait(ip);
+ /*
+ * Just warn here till VFS can correctly support
+ * read-only remount without racing.
+ */
+ WARN_ON(atomic_read(&mp->m_active_trans) != 0);
+
+ /* Push the superblock and write an unmount record */
+ error = xfs_log_sbcount(mp, 1);
+ if (error)
+ xfs_fs_cmn_err(CE_WARN, mp,
+ "xfs_attr_quiesce: failed to log sb changes. "
+ "Frozen image may not be consistent.");
+ xfs_log_unmount_write(mp);
+ xfs_unmountfs_writesb(mp);
+}
- xfs_ilock(ip, XFS_ILOCK_SHARED);
- }
+/*
+ * Enqueue a work item to be picked up by the vfs xfssyncd thread.
+ * Doing this has two advantages:
+ * - It saves on stack space, which is tight in certain situations
+ * - It can be used (with care) as a mechanism to avoid deadlocks.
+ * Flushing while allocating in a full filesystem requires both.
+ */
+STATIC void
+xfs_syncd_queue_work(
+ struct xfs_mount *mp,
+ void *data,
+ void (*syncer)(struct xfs_mount *, void *))
+{
+ struct bhv_vfs_sync_work *work;
+
+ work = kmem_alloc(sizeof(struct bhv_vfs_sync_work), KM_SLEEP);
+ INIT_LIST_HEAD(&work->w_list);
+ work->w_syncer = syncer;
+ work->w_data = data;
+ work->w_mount = mp;
+ spin_lock(&mp->m_sync_lock);
+ list_add_tail(&work->w_list, &mp->m_sync_list);
+ spin_unlock(&mp->m_sync_lock);
+ wake_up_process(mp->m_sync_task);
+}
- if ((flags & SYNC_ATTR) &&
- (ip->i_update_core ||
- (ip->i_itemp && ip->i_itemp->ili_format.ilf_fields))) {
- if (mount_locked)
- IPOINTER_INSERT(ip, mp);
+/*
+ * Flush delayed allocate data, attempting to free up reserved space
+ * from existing allocations. At this point a new allocation attempt
+ * has failed with ENOSPC and we are in the process of scratching our
+ * heads, looking about for more room...
+ */
+STATIC void
+xfs_flush_inode_work(
+ struct xfs_mount *mp,
+ void *arg)
+{
+ struct inode *inode = arg;
+ filemap_flush(inode->i_mapping);
+ iput(inode);
+}
- if (flags & SYNC_WAIT) {
- xfs_iflock(ip);
- error = xfs_iflush(ip, XFS_IFLUSH_SYNC);
-
- /*
- * If we can't acquire the flush lock, then the inode
- * is already being flushed so don't bother waiting.
- *
- * If we can lock it then do a delwri flush so we can
- * combine multiple inode flushes in each disk write.
- */
- } else if (xfs_iflock_nowait(ip)) {
- error = xfs_iflush(ip, XFS_IFLUSH_DELWRI);
- } else if (bypassed) {
- (*bypassed)++;
- }
- }
+void
+xfs_flush_inode(
+ xfs_inode_t *ip)
+{
+ struct inode *inode = VFS_I(ip);
- if (lock_flags != 0) {
- xfs_iunlock(ip, lock_flags);
- }
+ igrab(inode);
+ xfs_syncd_queue_work(ip->i_mount, inode, xfs_flush_inode_work);
+ delay(msecs_to_jiffies(500));
+}
- if (vnode_refed) {
- /*
- * If we had to take a reference on the vnode
- * above, then wait until after we've unlocked
- * the inode to release the reference. This is
- * because we can be already holding the inode
- * lock when IRELE() calls xfs_inactive().
- *
- * Make sure to drop the mount lock before calling
- * IRELE() so that we don't trip over ourselves if
- * we have to go for the mount lock again in the
- * inactive code.
- */
- if (mount_locked) {
- IPOINTER_INSERT(ip, mp);
- }
+/*
+ * This is the "bigger hammer" version of xfs_flush_inode_work...
+ * (IOW, "If at first you don't succeed, use a Bigger Hammer").
+ */
+STATIC void
+xfs_flush_device_work(
+ struct xfs_mount *mp,
+ void *arg)
+{
+ struct inode *inode = arg;
+ sync_blockdev(mp->m_super->s_bdev);
+ iput(inode);
+}
- IRELE(ip);
+void
+xfs_flush_device(
+ xfs_inode_t *ip)
+{
+ struct inode *inode = VFS_I(ip);
- vnode_refed = B_FALSE;
- }
+ igrab(inode);
+ xfs_syncd_queue_work(ip->i_mount, inode, xfs_flush_device_work);
+ delay(msecs_to_jiffies(500));
+ xfs_log_force(ip->i_mount, (xfs_lsn_t)0, XFS_LOG_FORCE|XFS_LOG_SYNC);
+}
- if (error) {
- last_error = error;
- }
+/*
+ * Every sync period we need to unpin all items, reclaim inodes, sync
+ * quota and write out the superblock. We might need to cover the log
+ * to indicate it is idle.
+ */
+STATIC void
+xfs_sync_worker(
+ struct xfs_mount *mp,
+ void *unused)
+{
+ int error;
+
+ if (!(mp->m_flags & XFS_MOUNT_RDONLY)) {
+ xfs_log_force(mp, (xfs_lsn_t)0, XFS_LOG_FORCE);
+ xfs_reclaim_inodes(mp, 0, XFS_IFLUSH_DELWRI_ELSE_ASYNC);
+ /* dgc: errors ignored here */
+ error = XFS_QM_DQSYNC(mp, SYNC_BDFLUSH);
+ error = xfs_sync_fsdata(mp, SYNC_BDFLUSH);
+ if (xfs_log_need_covered(mp))
+ error = xfs_commit_dummy_trans(mp, XFS_LOG_FORCE);
+ }
+ mp->m_sync_seq++;
+ wake_up(&mp->m_wait_single_sync_task);
+}
+STATIC int
+xfssyncd(
+ void *arg)
+{
+ struct xfs_mount *mp = arg;
+ long timeleft;
+ bhv_vfs_sync_work_t *work, *n;
+ LIST_HEAD (tmp);
+
+ set_freezable();
+ timeleft = xfs_syncd_centisecs * msecs_to_jiffies(10);
+ for (;;) {
+ timeleft = schedule_timeout_interruptible(timeleft);
+ /* swsusp */
+ try_to_freeze();
+ if (kthread_should_stop() && list_empty(&mp->m_sync_list))
+ break;
+
+ spin_lock(&mp->m_sync_lock);
/*
- * bail out if the filesystem is corrupted.
+ * We can get woken by laptop mode, to do a sync -
+ * that's the (only!) case where the list would be
+ * empty with time remaining.
*/
- if (error == EFSCORRUPTED) {
- if (!mount_locked) {
- XFS_MOUNT_ILOCK(mp);
- IPOINTER_REMOVE(ip, mp);
- }
- XFS_MOUNT_IUNLOCK(mp);
- ASSERT(ipointer_in == B_FALSE);
- kmem_free(ipointer);
- return XFS_ERROR(error);
+ if (!timeleft || list_empty(&mp->m_sync_list)) {
+ if (!timeleft)
+ timeleft = xfs_syncd_centisecs *
+ msecs_to_jiffies(10);
+ INIT_LIST_HEAD(&mp->m_sync_work.w_list);
+ list_add_tail(&mp->m_sync_work.w_list,
+ &mp->m_sync_list);
}
-
- /* Let other threads have a chance at the mount lock
- * if we have looped many times without dropping the
- * lock.
- */
- if ((++preempt & XFS_PREEMPT_MASK) == 0) {
- if (mount_locked) {
- IPOINTER_INSERT(ip, mp);
- }
+ list_for_each_entry_safe(work, n, &mp->m_sync_list, w_list)
+ list_move(&work->w_list, &tmp);
+ spin_unlock(&mp->m_sync_lock);
+
+ list_for_each_entry_safe(work, n, &tmp, w_list) {
+ (*work->w_syncer)(mp, work->w_data);
+ list_del(&work->w_list);
+ if (work == &mp->m_sync_work)
+ continue;
+ kmem_free(work);
}
+ }
- if (mount_locked == B_FALSE) {
- XFS_MOUNT_ILOCK(mp);
- mount_locked = B_TRUE;
- IPOINTER_REMOVE(ip, mp);
- continue;
- }
+ return 0;
+}
+
+int
+xfs_syncd_init(
+ struct xfs_mount *mp)
+{
+ mp->m_sync_work.w_syncer = xfs_sync_worker;
+ mp->m_sync_work.w_mount = mp;
+ mp->m_sync_task = kthread_run(xfssyncd, mp, "xfssyncd");
+ if (IS_ERR(mp->m_sync_task))
+ return -PTR_ERR(mp->m_sync_task);
+ return 0;
+}
+
+void
+xfs_syncd_stop(
+ struct xfs_mount *mp)
+{
+ kthread_stop(mp->m_sync_task);
+}
- ASSERT(ipointer_in == B_FALSE);
- ip = ip->i_mnext;
+int
+xfs_reclaim_inode(
+ xfs_inode_t *ip,
+ int locked,
+ int sync_mode)
+{
+ xfs_perag_t *pag = xfs_get_perag(ip->i_mount, ip->i_ino);
- } while (ip != mp->m_inodes);
+ /* The hash lock here protects a thread in xfs_iget_core from
+ * racing with us on linking the inode back with a vnode.
+ * Once we have the XFS_IRECLAIM flag set it will not touch
+ * us.
+ */
+ write_lock(&pag->pag_ici_lock);
+ spin_lock(&ip->i_flags_lock);
+ if (__xfs_iflags_test(ip, XFS_IRECLAIM) ||
+ !__xfs_iflags_test(ip, XFS_IRECLAIMABLE)) {
+ spin_unlock(&ip->i_flags_lock);
+ write_unlock(&pag->pag_ici_lock);
+ if (locked) {
+ xfs_ifunlock(ip);
+ xfs_iunlock(ip, XFS_ILOCK_EXCL);
+ }
+ return 1;
+ }
+ __xfs_iflags_set(ip, XFS_IRECLAIM);
+ spin_unlock(&ip->i_flags_lock);
+ write_unlock(&pag->pag_ici_lock);
+ xfs_put_perag(ip->i_mount, pag);
- XFS_MOUNT_IUNLOCK(mp);
+ /*
+ * If the inode is still dirty, then flush it out. If the inode
+ * is not in the AIL, then it will be OK to flush it delwri as
+ * long as xfs_iflush() does not keep any references to the inode.
+ * We leave that decision up to xfs_iflush() since it has the
+ * knowledge of whether it's OK to simply do a delwri flush of
+ * the inode or whether we need to wait until the inode is
+ * pulled from the AIL.
+ * We get the flush lock regardless, though, just to make sure
+ * we don't free it while it is being flushed.
+ */
+ if (!locked) {
+ xfs_ilock(ip, XFS_ILOCK_EXCL);
+ xfs_iflock(ip);
+ }
- ASSERT(ipointer_in == B_FALSE);
+ /*
+ * In the case of a forced shutdown we rely on xfs_iflush() to
+ * wait for the inode to be unpinned before returning an error.
+ */
+ if (!is_bad_inode(VFS_I(ip)) && xfs_iflush(ip, sync_mode) == 0) {
+ /* synchronize with xfs_iflush_done */
+ xfs_iflock(ip);
+ xfs_ifunlock(ip);
+ }
- kmem_free(ipointer);
- return XFS_ERROR(last_error);
+ xfs_iunlock(ip, XFS_ILOCK_EXCL);
+ xfs_ireclaim(ip);
+ return 0;
}
/*
- * xfs sync routine for internal use
- *
- * This routine supports all of the flags defined for the generic vfs_sync
- * interface as explained above under xfs_sync.
- *
+ * We set the inode flag atomically with the radix tree tag.
+ * Once we get tag lookups on the radix tree, this inode flag
+ * can go away.
*/
-int
-xfs_syncsub(
+void
+xfs_inode_set_reclaim_tag(
+ xfs_inode_t *ip)
+{
+ xfs_mount_t *mp = ip->i_mount;
+ xfs_perag_t *pag = xfs_get_perag(mp, ip->i_ino);
+
+ read_lock(&pag->pag_ici_lock);
+ spin_lock(&ip->i_flags_lock);
+ radix_tree_tag_set(&pag->pag_ici_root,
+ XFS_INO_TO_AGINO(mp, ip->i_ino), XFS_ICI_RECLAIM_TAG);
+ __xfs_iflags_set(ip, XFS_IRECLAIMABLE);
+ spin_unlock(&ip->i_flags_lock);
+ read_unlock(&pag->pag_ici_lock);
+ xfs_put_perag(mp, pag);
+}
+
+void
+__xfs_inode_clear_reclaim_tag(
xfs_mount_t *mp,
- int flags,
- int *bypassed)
+ xfs_perag_t *pag,
+ xfs_inode_t *ip)
{
- int error = 0;
- int last_error = 0;
- uint log_flags = XFS_LOG_FORCE;
- xfs_buf_t *bp;
- xfs_buf_log_item_t *bip;
+ radix_tree_tag_clear(&pag->pag_ici_root,
+ XFS_INO_TO_AGINO(mp, ip->i_ino), XFS_ICI_RECLAIM_TAG);
+}
- /*
- * Sync out the log. This ensures that the log is periodically
- * flushed even if there is not enough activity to fill it up.
- */
- if (flags & SYNC_WAIT)
- log_flags |= XFS_LOG_SYNC;
+void
+xfs_inode_clear_reclaim_tag(
+ xfs_inode_t *ip)
+{
+ xfs_mount_t *mp = ip->i_mount;
+ xfs_perag_t *pag = xfs_get_perag(mp, ip->i_ino);
+
+ read_lock(&pag->pag_ici_lock);
+ spin_lock(&ip->i_flags_lock);
+ __xfs_inode_clear_reclaim_tag(mp, pag, ip);
+ spin_unlock(&ip->i_flags_lock);
+ read_unlock(&pag->pag_ici_lock);
+ xfs_put_perag(mp, pag);
+}
- xfs_log_force(mp, (xfs_lsn_t)0, log_flags);
- if (flags & (SYNC_ATTR|SYNC_DELWRI)) {
- if (flags & SYNC_BDFLUSH)
- xfs_finish_reclaim_all(mp, 1);
- else
- error = xfs_sync_inodes(mp, flags, bypassed);
- }
+STATIC void
+xfs_reclaim_inodes_ag(
+ xfs_mount_t *mp,
+ int ag,
+ int noblock,
+ int mode)
+{
+ xfs_inode_t *ip = NULL;
+ xfs_perag_t *pag = &mp->m_perag[ag];
+ int nr_found;
+ uint32_t first_index;
+ int skipped;
+
+restart:
+ first_index = 0;
+ skipped = 0;
+ do {
+ /*
+ * use a gang lookup to find the next inode in the tree
+ * as the tree is sparse and a gang lookup walks to find
+ * the number of objects requested.
+ */
+ read_lock(&pag->pag_ici_lock);
+ nr_found = radix_tree_gang_lookup_tag(&pag->pag_ici_root,
+ (void**)&ip, first_index, 1,
+ XFS_ICI_RECLAIM_TAG);
- /*
- * Flushing out dirty data above probably generated more
- * log activity, so if this isn't vfs_sync() then flush
- * the log again.
- */
- if (flags & SYNC_DELWRI) {
- xfs_log_force(mp, (xfs_lsn_t)0, log_flags);
- }
+ if (!nr_found) {
+ read_unlock(&pag->pag_ici_lock);
+ break;
+ }
- if (flags & SYNC_FSDATA) {
/*
- * If this is vfs_sync() then only sync the superblock
- * if we can lock it without sleeping and it is not pinned.
+ * Update the index for the next lookup. Catch overflows
+ * into the next AG range which can occur if we have inodes
+ * in the last block of the AG and we are currently
+ * pointing to the last inode.
*/
- if (flags & SYNC_BDFLUSH) {
- bp = xfs_getsb(mp, XFS_BUF_TRYLOCK);
- if (bp != NULL) {
- bip = XFS_BUF_FSPRIVATE(bp,xfs_buf_log_item_t*);
- if ((bip != NULL) &&
- xfs_buf_item_dirty(bip)) {
- if (!(XFS_BUF_ISPINNED(bp))) {
- XFS_BUF_ASYNC(bp);
- error = xfs_bwrite(mp, bp);
- } else {
- xfs_buf_relse(bp);
- }
- } else {
- xfs_buf_relse(bp);
- }
- }
- } else {
- bp = xfs_getsb(mp, 0);
- /*
- * If the buffer is pinned then push on the log so
- * we won't get stuck waiting in the write for
- * someone, maybe ourselves, to flush the log.
- * Even though we just pushed the log above, we
- * did not have the superblock buffer locked at
- * that point so it can become pinned in between
- * there and here.
- */
- if (XFS_BUF_ISPINNED(bp))
- xfs_log_force(mp, (xfs_lsn_t)0, XFS_LOG_FORCE);
- if (flags & SYNC_WAIT)
- XFS_BUF_UNASYNC(bp);
- else
- XFS_BUF_ASYNC(bp);
- error = xfs_bwrite(mp, bp);
+ first_index = XFS_INO_TO_AGINO(mp, ip->i_ino + 1);
+ if (first_index < XFS_INO_TO_AGINO(mp, ip->i_ino)) {
+ read_unlock(&pag->pag_ici_lock);
+ break;
}
- if (error) {
- last_error = error;
+
+ /* ignore if already under reclaim */
+ if (xfs_iflags_test(ip, XFS_IRECLAIM)) {
+ read_unlock(&pag->pag_ici_lock);
+ continue;
}
- }
- /*
- * Now check to see if the log needs a "dummy" transaction.
- */
- if (!(flags & SYNC_REMOUNT) && xfs_log_need_covered(mp)) {
- xfs_trans_t *tp;
- xfs_inode_t *ip;
+ if (noblock) {
+ if (!xfs_ilock_nowait(ip, XFS_ILOCK_EXCL)) {
+ read_unlock(&pag->pag_ici_lock);
+ continue;
+ }
+ if (xfs_ipincount(ip) ||
+ !xfs_iflock_nowait(ip)) {
+ xfs_iunlock(ip, XFS_ILOCK_EXCL);
+ read_unlock(&pag->pag_ici_lock);
+ continue;
+ }
+ }
+ read_unlock(&pag->pag_ici_lock);
/*
- * Put a dummy transaction in the log to tell
- * recovery that all others are OK.
+ * hmmm - this is an inode already in reclaim. Do
+ * we even bother catching it here?
*/
- tp = xfs_trans_alloc(mp, XFS_TRANS_DUMMY1);
- if ((error = xfs_trans_reserve(tp, 0,
- XFS_ICHANGE_LOG_RES(mp),
- 0, 0, 0))) {
- xfs_trans_cancel(tp, 0);
- return error;
- }
-
- ip = mp->m_rootip;
- xfs_ilock(ip, XFS_ILOCK_EXCL);
+ if (xfs_reclaim_inode(ip, noblock, mode))
+ skipped++;
+ } while (nr_found);
- xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
- xfs_trans_ihold(tp, ip);
- xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
- error = xfs_trans_commit(tp, 0);
- xfs_iunlock(ip, XFS_ILOCK_EXCL);
- xfs_log_force(mp, (xfs_lsn_t)0, log_flags);
+ if (skipped) {
+ delay(1);
+ goto restart;
}
+ return;
- /*
- * When shutting down, we need to insure that the AIL is pushed
- * to disk or the filesystem can appear corrupt from the PROM.
- */
- if ((flags & (SYNC_CLOSE|SYNC_WAIT)) == (SYNC_CLOSE|SYNC_WAIT)) {
- XFS_bflush(mp->m_ddev_targp);
- if (mp->m_rtdev_targp) {
- XFS_bflush(mp->m_rtdev_targp);
- }
- }
+}
- return XFS_ERROR(last_error);
+int
+xfs_reclaim_inodes(
+ xfs_mount_t *mp,
+ int noblock,
+ int mode)
+{
+ int i;
+
+ for (i = 0; i < mp->m_sb.sb_agcount; i++) {
+ if (!mp->m_perag[i].pag_ici_init)
+ continue;
+ xfs_reclaim_inodes_ag(mp, i, noblock, mode);
+ }
+ return 0;
}
+
+