From 1ad2cfe0a57031505df682dc1e26922d9d43737f Mon Sep 17 00:00:00 2001 From: "Darrick J. Wong" Date: Mon, 31 May 2021 11:31:57 -0700 Subject: xfs: move the quotaoff dqrele inode walk into xfs_icache.c The only external caller of xfs_inode_walk* happens in quotaoff, when we want to walk all the incore inodes to detach the dquots. Move this code to xfs_icache.c so that we can hide xfs_inode_walk as the starting step in more cleanups of inode walks. Signed-off-by: Darrick J. Wong Reviewed-by: Dave Chinner --- fs/xfs/xfs_icache.c | 65 +++++++++++++++++++++++++++++++++++++++++++++++- fs/xfs/xfs_icache.h | 8 +++--- fs/xfs/xfs_qm.h | 1 - fs/xfs/xfs_qm_syscalls.c | 54 ++-------------------------------------- 4 files changed, 71 insertions(+), 57 deletions(-) diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c index 3c81daca0e9a..e2edbcf7a528 100644 --- a/fs/xfs/xfs_icache.c +++ b/fs/xfs/xfs_icache.c @@ -26,6 +26,18 @@ #include +/* + * Private inode cache walk flags for struct xfs_eofblocks. Must not coincide + * with XFS_EOF_FLAGS_*. + */ +#define XFS_ICWALK_FLAG_DROP_UDQUOT (1U << 31) +#define XFS_ICWALK_FLAG_DROP_GDQUOT (1U << 30) +#define XFS_ICWALK_FLAG_DROP_PDQUOT (1U << 29) + +#define XFS_ICWALK_PRIVATE_FLAGS (XFS_ICWALK_FLAG_DROP_UDQUOT | \ + XFS_ICWALK_FLAG_DROP_GDQUOT | \ + XFS_ICWALK_FLAG_DROP_PDQUOT) + /* * Allocate and initialise an xfs_inode. */ @@ -890,7 +902,7 @@ xfs_inode_walk_get_perag( * Call the @execute function on all incore inodes matching the radix tree * @tag. */ -int +static int xfs_inode_walk( struct xfs_mount *mp, int iter_flags, @@ -915,7 +927,58 @@ xfs_inode_walk( } } return last_error; + BUILD_BUG_ON(XFS_ICWALK_PRIVATE_FLAGS & XFS_EOF_FLAGS_VALID); +} + +#ifdef CONFIG_XFS_QUOTA +/* Drop this inode's dquots. */ +static int +xfs_dqrele_inode( + struct xfs_inode *ip, + void *priv) +{ + struct xfs_eofblocks *eofb = priv; + + xfs_ilock(ip, XFS_ILOCK_EXCL); + if (eofb->eof_flags & XFS_ICWALK_FLAG_DROP_UDQUOT) { + xfs_qm_dqrele(ip->i_udquot); + ip->i_udquot = NULL; + } + if (eofb->eof_flags & XFS_ICWALK_FLAG_DROP_GDQUOT) { + xfs_qm_dqrele(ip->i_gdquot); + ip->i_gdquot = NULL; + } + if (eofb->eof_flags & XFS_ICWALK_FLAG_DROP_PDQUOT) { + xfs_qm_dqrele(ip->i_pdquot); + ip->i_pdquot = NULL; + } + xfs_iunlock(ip, XFS_ILOCK_EXCL); + return 0; +} + +/* + * Detach all dquots from incore inodes if we can. The caller must already + * have dropped the relevant XFS_[UGP]QUOTA_ACTIVE flags so that dquots will + * not get reattached. + */ +int +xfs_dqrele_all_inodes( + struct xfs_mount *mp, + unsigned int qflags) +{ + struct xfs_eofblocks eofb = { .eof_flags = 0 }; + + if (qflags & XFS_UQUOTA_ACCT) + eofb.eof_flags |= XFS_ICWALK_FLAG_DROP_UDQUOT; + if (qflags & XFS_GQUOTA_ACCT) + eofb.eof_flags |= XFS_ICWALK_FLAG_DROP_GDQUOT; + if (qflags & XFS_PQUOTA_ACCT) + eofb.eof_flags |= XFS_ICWALK_FLAG_DROP_PDQUOT; + + return xfs_inode_walk(mp, XFS_INODE_WALK_INEW_WAIT, xfs_dqrele_inode, + &eofb, XFS_ICI_NO_TAG); } +#endif /* CONFIG_XFS_QUOTA */ /* * Grab the inode for reclaim exclusively. diff --git a/fs/xfs/xfs_icache.h b/fs/xfs/xfs_icache.h index d1fddb152420..d9baa6df1121 100644 --- a/fs/xfs/xfs_icache.h +++ b/fs/xfs/xfs_icache.h @@ -68,9 +68,11 @@ void xfs_inode_clear_cowblocks_tag(struct xfs_inode *ip); void xfs_blockgc_worker(struct work_struct *work); -int xfs_inode_walk(struct xfs_mount *mp, int iter_flags, - int (*execute)(struct xfs_inode *ip, void *args), - void *args, int tag); +#ifdef CONFIG_XFS_QUOTA +int xfs_dqrele_all_inodes(struct xfs_mount *mp, unsigned int qflags); +#else +# define xfs_dqrele_all_inodes(mp, qflags) (0) +#endif int xfs_icache_inode_is_allocated(struct xfs_mount *mp, struct xfs_trans *tp, xfs_ino_t ino, bool *inuse); diff --git a/fs/xfs/xfs_qm.h b/fs/xfs/xfs_qm.h index e3dabab44097..ebbb484c49dc 100644 --- a/fs/xfs/xfs_qm.h +++ b/fs/xfs/xfs_qm.h @@ -142,7 +142,6 @@ extern void xfs_qm_destroy_quotainfo(struct xfs_mount *); /* dquot stuff */ extern void xfs_qm_dqpurge_all(struct xfs_mount *, uint); -extern void xfs_qm_dqrele_all_inodes(struct xfs_mount *, uint); /* quota ops */ extern int xfs_qm_scall_trunc_qfiles(struct xfs_mount *, uint); diff --git a/fs/xfs/xfs_qm_syscalls.c b/fs/xfs/xfs_qm_syscalls.c index 11f1e2fbf22f..13a56e1ea15c 100644 --- a/fs/xfs/xfs_qm_syscalls.c +++ b/fs/xfs/xfs_qm_syscalls.c @@ -201,7 +201,8 @@ xfs_qm_scall_quotaoff( * depend on the quota inodes (and other things) being valid as long as * we keep the lock(s). */ - xfs_qm_dqrele_all_inodes(mp, flags); + error = xfs_dqrele_all_inodes(mp, flags); + ASSERT(!error); /* * Next we make the changes in the quota flag in the mount struct. @@ -747,54 +748,3 @@ xfs_qm_scall_getquota_next( xfs_qm_dqput(dqp); return error; } - -STATIC int -xfs_dqrele_inode( - struct xfs_inode *ip, - void *args) -{ - uint *flags = args; - - /* skip quota inodes */ - if (ip == ip->i_mount->m_quotainfo->qi_uquotaip || - ip == ip->i_mount->m_quotainfo->qi_gquotaip || - ip == ip->i_mount->m_quotainfo->qi_pquotaip) { - ASSERT(ip->i_udquot == NULL); - ASSERT(ip->i_gdquot == NULL); - ASSERT(ip->i_pdquot == NULL); - return 0; - } - - xfs_ilock(ip, XFS_ILOCK_EXCL); - if ((*flags & XFS_UQUOTA_ACCT) && ip->i_udquot) { - xfs_qm_dqrele(ip->i_udquot); - ip->i_udquot = NULL; - } - if ((*flags & XFS_GQUOTA_ACCT) && ip->i_gdquot) { - xfs_qm_dqrele(ip->i_gdquot); - ip->i_gdquot = NULL; - } - if ((*flags & XFS_PQUOTA_ACCT) && ip->i_pdquot) { - xfs_qm_dqrele(ip->i_pdquot); - ip->i_pdquot = NULL; - } - xfs_iunlock(ip, XFS_ILOCK_EXCL); - return 0; -} - - -/* - * Go thru all the inodes in the file system, releasing their dquots. - * - * Note that the mount structure gets modified to indicate that quotas are off - * AFTER this, in the case of quotaoff. - */ -void -xfs_qm_dqrele_all_inodes( - struct xfs_mount *mp, - uint flags) -{ - ASSERT(mp->m_quotainfo); - xfs_inode_walk(mp, XFS_INODE_WALK_INEW_WAIT, xfs_dqrele_inode, - &flags, XFS_ICI_NO_TAG); -} -- cgit v1.2.3 From 3ea06d73e3c02ee2952a62bf92abc18f9c98aba1 Mon Sep 17 00:00:00 2001 From: "Darrick J. Wong" Date: Mon, 31 May 2021 11:31:57 -0700 Subject: xfs: detach inode dquots at the end of inactivation Once we're done with inactivating an inode, we're finished updating metadata for that inode. This means that we can detach the dquots at the end and not have to wait for reclaim to do it for us. Signed-off-by: Darrick J. Wong Reviewed-by: Dave Chinner --- fs/xfs/xfs_icache.c | 2 +- fs/xfs/xfs_inode.c | 22 +++++++++++----------- 2 files changed, 12 insertions(+), 12 deletions(-) diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c index e2edbcf7a528..dfa0ec7d02b8 100644 --- a/fs/xfs/xfs_icache.c +++ b/fs/xfs/xfs_icache.c @@ -1093,7 +1093,7 @@ reclaim: * unlocked after the lookup before we go ahead and free it. */ xfs_ilock(ip, XFS_ILOCK_EXCL); - xfs_qm_dqdetach(ip); + ASSERT(!ip->i_udquot && !ip->i_gdquot && !ip->i_pdquot); xfs_iunlock(ip, XFS_ILOCK_EXCL); ASSERT(xfs_inode_clean(ip)); diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c index e4c2da4566f1..51972549e73c 100644 --- a/fs/xfs/xfs_inode.c +++ b/fs/xfs/xfs_inode.c @@ -1716,7 +1716,7 @@ xfs_inactive( */ if (VFS_I(ip)->i_mode == 0) { ASSERT(ip->i_df.if_broot_bytes == 0); - return; + goto out; } mp = ip->i_mount; @@ -1724,11 +1724,11 @@ xfs_inactive( /* If this is a read-only mount, don't do this (would generate I/O) */ if (mp->m_flags & XFS_MOUNT_RDONLY) - return; + goto out; /* Metadata inodes require explicit resource cleanup. */ if (xfs_is_metadata_inode(ip)) - return; + goto out; /* Try to clean out the cow blocks if there are any. */ if (xfs_inode_has_cow_data(ip)) @@ -1747,7 +1747,7 @@ xfs_inactive( if (xfs_can_free_eofblocks(ip, true)) xfs_free_eofblocks(ip); - return; + goto out; } if (S_ISREG(VFS_I(ip)->i_mode) && @@ -1757,14 +1757,14 @@ xfs_inactive( error = xfs_qm_dqattach(ip); if (error) - return; + goto out; if (S_ISLNK(VFS_I(ip)->i_mode)) error = xfs_inactive_symlink(ip); else if (truncate) error = xfs_inactive_truncate(ip); if (error) - return; + goto out; /* * If there are attributes associated with the file then blow them away @@ -1774,7 +1774,7 @@ xfs_inactive( if (XFS_IFORK_Q(ip)) { error = xfs_attr_inactive(ip); if (error) - return; + goto out; } ASSERT(!ip->i_afp); @@ -1783,12 +1783,12 @@ xfs_inactive( /* * Free the inode. */ - error = xfs_inactive_ifree(ip); - if (error) - return; + xfs_inactive_ifree(ip); +out: /* - * Release the dquots held by inode, if any. + * We're done making metadata updates for this inode, so we can release + * the attached dquots. */ xfs_qm_dqdetach(ip); } -- cgit v1.2.3 From df60019739d8850b865d313053d30aa93dc38a65 Mon Sep 17 00:00:00 2001 From: "Darrick J. Wong" Date: Tue, 1 Jun 2021 13:29:41 -0700 Subject: xfs: move the inode walk functions further down Move the inode walk functions further down in the file to limit the forward declarations to the two walk functions as we add new code that uses the inode walks. We'll clean them out later (i.e. after the deferred inode inactivation series). Signed-off-by: Darrick J. Wong Reviewed-by: Dave Chinner --- fs/xfs/xfs_icache.c | 401 +++++++++++++++++++++++++++------------------------- 1 file changed, 206 insertions(+), 195 deletions(-) diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c index dfa0ec7d02b8..55c55e449cab 100644 --- a/fs/xfs/xfs_icache.c +++ b/fs/xfs/xfs_icache.c @@ -26,6 +26,13 @@ #include +static int xfs_inode_walk(struct xfs_mount *mp, int iter_flags, + int (*execute)(struct xfs_inode *ip, void *args), + void *args, int tag); +static int xfs_inode_walk_ag(struct xfs_perag *pag, int iter_flags, + int (*execute)(struct xfs_inode *ip, void *args), + void *args, int tag); + /* * Private inode cache walk flags for struct xfs_eofblocks. Must not coincide * with XFS_EOF_FLAGS_*. @@ -732,204 +739,12 @@ xfs_icache_inode_is_allocated( * radix tree lookups to a minimum. The batch size is a trade off between * lookup reduction and stack usage. This is in the reclaim path, so we can't * be too greedy. + * + * XXX: This will be moved closer to xfs_inode_walk* once we get rid of the + * separate reclaim walk functions. */ #define XFS_LOOKUP_BATCH 32 -/* - * Decide if the given @ip is eligible to be a part of the inode walk, and - * grab it if so. Returns true if it's ready to go or false if we should just - * ignore it. - */ -STATIC bool -xfs_inode_walk_ag_grab( - struct xfs_inode *ip, - int flags) -{ - struct inode *inode = VFS_I(ip); - bool newinos = !!(flags & XFS_INODE_WALK_INEW_WAIT); - - ASSERT(rcu_read_lock_held()); - - /* Check for stale RCU freed inode */ - spin_lock(&ip->i_flags_lock); - if (!ip->i_ino) - goto out_unlock_noent; - - /* avoid new or reclaimable inodes. Leave for reclaim code to flush */ - if ((!newinos && __xfs_iflags_test(ip, XFS_INEW)) || - __xfs_iflags_test(ip, XFS_IRECLAIMABLE | XFS_IRECLAIM)) - goto out_unlock_noent; - spin_unlock(&ip->i_flags_lock); - - /* nothing to sync during shutdown */ - if (XFS_FORCED_SHUTDOWN(ip->i_mount)) - return false; - - /* If we can't grab the inode, it must on it's way to reclaim. */ - if (!igrab(inode)) - return false; - - /* inode is valid */ - return true; - -out_unlock_noent: - spin_unlock(&ip->i_flags_lock); - return false; -} - -/* - * For a given per-AG structure @pag, grab, @execute, and rele all incore - * inodes with the given radix tree @tag. - */ -STATIC int -xfs_inode_walk_ag( - struct xfs_perag *pag, - int iter_flags, - int (*execute)(struct xfs_inode *ip, void *args), - void *args, - int tag) -{ - struct xfs_mount *mp = pag->pag_mount; - uint32_t first_index; - int last_error = 0; - int skipped; - bool done; - int nr_found; - -restart: - done = false; - skipped = 0; - first_index = 0; - nr_found = 0; - do { - struct xfs_inode *batch[XFS_LOOKUP_BATCH]; - int error = 0; - int i; - - rcu_read_lock(); - - if (tag == XFS_ICI_NO_TAG) - nr_found = radix_tree_gang_lookup(&pag->pag_ici_root, - (void **)batch, first_index, - XFS_LOOKUP_BATCH); - else - nr_found = radix_tree_gang_lookup_tag( - &pag->pag_ici_root, - (void **) batch, first_index, - XFS_LOOKUP_BATCH, tag); - - if (!nr_found) { - rcu_read_unlock(); - break; - } - - /* - * Grab the inodes before we drop the lock. if we found - * nothing, nr == 0 and the loop will be skipped. - */ - for (i = 0; i < nr_found; i++) { - struct xfs_inode *ip = batch[i]; - - if (done || !xfs_inode_walk_ag_grab(ip, iter_flags)) - batch[i] = NULL; - - /* - * 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. - * - * Because we may see inodes that are from the wrong AG - * due to RCU freeing and reallocation, only update the - * index if it lies in this AG. It was a race that lead - * us to see this inode, so another lookup from the - * same index will not find it again. - */ - if (XFS_INO_TO_AGNO(mp, ip->i_ino) != pag->pag_agno) - continue; - first_index = XFS_INO_TO_AGINO(mp, ip->i_ino + 1); - if (first_index < XFS_INO_TO_AGINO(mp, ip->i_ino)) - done = true; - } - - /* unlock now we've grabbed the inodes. */ - rcu_read_unlock(); - - for (i = 0; i < nr_found; i++) { - if (!batch[i]) - continue; - if ((iter_flags & XFS_INODE_WALK_INEW_WAIT) && - xfs_iflags_test(batch[i], XFS_INEW)) - xfs_inew_wait(batch[i]); - error = execute(batch[i], args); - xfs_irele(batch[i]); - if (error == -EAGAIN) { - skipped++; - continue; - } - if (error && last_error != -EFSCORRUPTED) - last_error = error; - } - - /* bail out if the filesystem is corrupted. */ - if (error == -EFSCORRUPTED) - break; - - cond_resched(); - - } while (nr_found && !done); - - if (skipped) { - delay(1); - goto restart; - } - return last_error; -} - -/* Fetch the next (possibly tagged) per-AG structure. */ -static inline struct xfs_perag * -xfs_inode_walk_get_perag( - struct xfs_mount *mp, - xfs_agnumber_t agno, - int tag) -{ - if (tag == XFS_ICI_NO_TAG) - return xfs_perag_get(mp, agno); - return xfs_perag_get_tag(mp, agno, tag); -} - -/* - * Call the @execute function on all incore inodes matching the radix tree - * @tag. - */ -static int -xfs_inode_walk( - struct xfs_mount *mp, - int iter_flags, - int (*execute)(struct xfs_inode *ip, void *args), - void *args, - int tag) -{ - struct xfs_perag *pag; - int error = 0; - int last_error = 0; - xfs_agnumber_t ag; - - ag = 0; - while ((pag = xfs_inode_walk_get_perag(mp, ag, tag))) { - ag = pag->pag_agno + 1; - error = xfs_inode_walk_ag(pag, iter_flags, execute, args, tag); - xfs_perag_put(pag); - if (error) { - last_error = error; - if (error == -EFSCORRUPTED) - break; - } - } - return last_error; - BUILD_BUG_ON(XFS_ICWALK_PRIVATE_FLAGS & XFS_EOF_FLAGS_VALID); -} - #ifdef CONFIG_XFS_QUOTA /* Drop this inode's dquots. */ static int @@ -1649,6 +1464,48 @@ xfs_blockgc_start( xfs_blockgc_queue(pag); } +/* + * Decide if the given @ip is eligible to be a part of the inode walk, and + * grab it if so. Returns true if it's ready to go or false if we should just + * ignore it. + */ +static bool +xfs_inode_walk_ag_grab( + struct xfs_inode *ip, + int flags) +{ + struct inode *inode = VFS_I(ip); + bool newinos = !!(flags & XFS_INODE_WALK_INEW_WAIT); + + ASSERT(rcu_read_lock_held()); + + /* Check for stale RCU freed inode */ + spin_lock(&ip->i_flags_lock); + if (!ip->i_ino) + goto out_unlock_noent; + + /* avoid new or reclaimable inodes. Leave for reclaim code to flush */ + if ((!newinos && __xfs_iflags_test(ip, XFS_INEW)) || + __xfs_iflags_test(ip, XFS_IRECLAIMABLE | XFS_IRECLAIM)) + goto out_unlock_noent; + spin_unlock(&ip->i_flags_lock); + + /* nothing to sync during shutdown */ + if (XFS_FORCED_SHUTDOWN(ip->i_mount)) + return false; + + /* If we can't grab the inode, it must on it's way to reclaim. */ + if (!igrab(inode)) + return false; + + /* inode is valid */ + return true; + +out_unlock_noent: + spin_unlock(&ip->i_flags_lock); + return false; +} + /* Scan one incore inode for block preallocations that we can remove. */ static int xfs_blockgc_scan_inode( @@ -1769,3 +1626,157 @@ xfs_blockgc_free_quota( xfs_inode_dquot(ip, XFS_DQTYPE_GROUP), xfs_inode_dquot(ip, XFS_DQTYPE_PROJ), eof_flags); } + +/* XFS Inode Cache Walking Code */ + +/* + * For a given per-AG structure @pag, grab, @execute, and rele all incore + * inodes with the given radix tree @tag. + */ +static int +xfs_inode_walk_ag( + struct xfs_perag *pag, + int iter_flags, + int (*execute)(struct xfs_inode *ip, void *args), + void *args, + int tag) +{ + struct xfs_mount *mp = pag->pag_mount; + uint32_t first_index; + int last_error = 0; + int skipped; + bool done; + int nr_found; + +restart: + done = false; + skipped = 0; + first_index = 0; + nr_found = 0; + do { + struct xfs_inode *batch[XFS_LOOKUP_BATCH]; + int error = 0; + int i; + + rcu_read_lock(); + + if (tag == XFS_ICI_NO_TAG) + nr_found = radix_tree_gang_lookup(&pag->pag_ici_root, + (void **)batch, first_index, + XFS_LOOKUP_BATCH); + else + nr_found = radix_tree_gang_lookup_tag( + &pag->pag_ici_root, + (void **) batch, first_index, + XFS_LOOKUP_BATCH, tag); + + if (!nr_found) { + rcu_read_unlock(); + break; + } + + /* + * Grab the inodes before we drop the lock. if we found + * nothing, nr == 0 and the loop will be skipped. + */ + for (i = 0; i < nr_found; i++) { + struct xfs_inode *ip = batch[i]; + + if (done || !xfs_inode_walk_ag_grab(ip, iter_flags)) + batch[i] = NULL; + + /* + * 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. + * + * Because we may see inodes that are from the wrong AG + * due to RCU freeing and reallocation, only update the + * index if it lies in this AG. It was a race that lead + * us to see this inode, so another lookup from the + * same index will not find it again. + */ + if (XFS_INO_TO_AGNO(mp, ip->i_ino) != pag->pag_agno) + continue; + first_index = XFS_INO_TO_AGINO(mp, ip->i_ino + 1); + if (first_index < XFS_INO_TO_AGINO(mp, ip->i_ino)) + done = true; + } + + /* unlock now we've grabbed the inodes. */ + rcu_read_unlock(); + + for (i = 0; i < nr_found; i++) { + if (!batch[i]) + continue; + if ((iter_flags & XFS_INODE_WALK_INEW_WAIT) && + xfs_iflags_test(batch[i], XFS_INEW)) + xfs_inew_wait(batch[i]); + error = execute(batch[i], args); + xfs_irele(batch[i]); + if (error == -EAGAIN) { + skipped++; + continue; + } + if (error && last_error != -EFSCORRUPTED) + last_error = error; + } + + /* bail out if the filesystem is corrupted. */ + if (error == -EFSCORRUPTED) + break; + + cond_resched(); + + } while (nr_found && !done); + + if (skipped) { + delay(1); + goto restart; + } + return last_error; +} + +/* Fetch the next (possibly tagged) per-AG structure. */ +static inline struct xfs_perag * +xfs_inode_walk_get_perag( + struct xfs_mount *mp, + xfs_agnumber_t agno, + int tag) +{ + if (tag == XFS_ICI_NO_TAG) + return xfs_perag_get(mp, agno); + return xfs_perag_get_tag(mp, agno, tag); +} + +/* + * Call the @execute function on all incore inodes matching the radix tree + * @tag. + */ +static int +xfs_inode_walk( + struct xfs_mount *mp, + int iter_flags, + int (*execute)(struct xfs_inode *ip, void *args), + void *args, + int tag) +{ + struct xfs_perag *pag; + int error = 0; + int last_error = 0; + xfs_agnumber_t agno = 0; + + while ((pag = xfs_inode_walk_get_perag(mp, agno, tag))) { + agno = pag->pag_agno + 1; + error = xfs_inode_walk_ag(pag, iter_flags, execute, args, tag); + xfs_perag_put(pag); + if (error) { + last_error = error; + if (error == -EFSCORRUPTED) + break; + } + } + return last_error; + BUILD_BUG_ON(XFS_ICWALK_PRIVATE_FLAGS & XFS_EOF_FLAGS_VALID); +} -- cgit v1.2.3 From c1115c0cba2b82e71ec77e794c684ac87160fcf6 Mon Sep 17 00:00:00 2001 From: "Darrick J. Wong" Date: Tue, 1 Jun 2021 22:41:25 -0700 Subject: xfs: rename xfs_inode_walk functions to xfs_icwalk Shorten the prefix so that all the incore inode cache walk code has "xfs_icwalk" in the name somewhere. Signed-off-by: Darrick J. Wong Reviewed-by: Dave Chinner --- fs/xfs/xfs_icache.c | 22 +++++++++++----------- 1 file changed, 11 insertions(+), 11 deletions(-) diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c index 55c55e449cab..d5ecd4cd3ef5 100644 --- a/fs/xfs/xfs_icache.c +++ b/fs/xfs/xfs_icache.c @@ -26,10 +26,10 @@ #include -static int xfs_inode_walk(struct xfs_mount *mp, int iter_flags, +static int xfs_icwalk(struct xfs_mount *mp, int iter_flags, int (*execute)(struct xfs_inode *ip, void *args), void *args, int tag); -static int xfs_inode_walk_ag(struct xfs_perag *pag, int iter_flags, +static int xfs_icwalk_ag(struct xfs_perag *pag, int iter_flags, int (*execute)(struct xfs_inode *ip, void *args), void *args, int tag); @@ -740,7 +740,7 @@ xfs_icache_inode_is_allocated( * lookup reduction and stack usage. This is in the reclaim path, so we can't * be too greedy. * - * XXX: This will be moved closer to xfs_inode_walk* once we get rid of the + * XXX: This will be moved closer to xfs_icwalk* once we get rid of the * separate reclaim walk functions. */ #define XFS_LOOKUP_BATCH 32 @@ -790,7 +790,7 @@ xfs_dqrele_all_inodes( if (qflags & XFS_PQUOTA_ACCT) eofb.eof_flags |= XFS_ICWALK_FLAG_DROP_PDQUOT; - return xfs_inode_walk(mp, XFS_INODE_WALK_INEW_WAIT, xfs_dqrele_inode, + return xfs_icwalk(mp, XFS_INODE_WALK_INEW_WAIT, xfs_dqrele_inode, &eofb, XFS_ICI_NO_TAG); } #endif /* CONFIG_XFS_QUOTA */ @@ -1538,7 +1538,7 @@ xfs_blockgc_worker( if (!sb_start_write_trylock(mp->m_super)) return; - error = xfs_inode_walk_ag(pag, 0, xfs_blockgc_scan_inode, NULL, + error = xfs_icwalk_ag(pag, 0, xfs_blockgc_scan_inode, NULL, XFS_ICI_BLOCKGC_TAG); if (error) xfs_info(mp, "AG %u preallocation gc worker failed, err=%d", @@ -1557,7 +1557,7 @@ xfs_blockgc_free_space( { trace_xfs_blockgc_free_space(mp, eofb, _RET_IP_); - return xfs_inode_walk(mp, 0, xfs_blockgc_scan_inode, eofb, + return xfs_icwalk(mp, 0, xfs_blockgc_scan_inode, eofb, XFS_ICI_BLOCKGC_TAG); } @@ -1634,7 +1634,7 @@ xfs_blockgc_free_quota( * inodes with the given radix tree @tag. */ static int -xfs_inode_walk_ag( +xfs_icwalk_ag( struct xfs_perag *pag, int iter_flags, int (*execute)(struct xfs_inode *ip, void *args), @@ -1740,7 +1740,7 @@ restart: /* Fetch the next (possibly tagged) per-AG structure. */ static inline struct xfs_perag * -xfs_inode_walk_get_perag( +xfs_icwalk_get_perag( struct xfs_mount *mp, xfs_agnumber_t agno, int tag) @@ -1755,7 +1755,7 @@ xfs_inode_walk_get_perag( * @tag. */ static int -xfs_inode_walk( +xfs_icwalk( struct xfs_mount *mp, int iter_flags, int (*execute)(struct xfs_inode *ip, void *args), @@ -1767,9 +1767,9 @@ xfs_inode_walk( int last_error = 0; xfs_agnumber_t agno = 0; - while ((pag = xfs_inode_walk_get_perag(mp, agno, tag))) { + while ((pag = xfs_icwalk_get_perag(mp, agno, tag))) { agno = pag->pag_agno + 1; - error = xfs_inode_walk_ag(pag, iter_flags, execute, args, tag); + error = xfs_icwalk_ag(pag, iter_flags, execute, args, tag); xfs_perag_put(pag); if (error) { last_error = error; -- cgit v1.2.3 From c809d7e948a131cba8fdf9fbd0b50e1f59255f50 Mon Sep 17 00:00:00 2001 From: "Darrick J. Wong" Date: Tue, 1 Jun 2021 13:49:52 -0700 Subject: xfs: pass the goal of the incore inode walk to xfs_inode_walk() As part of removing the indirect calls and radix tag implementation details from the incore inode walk loop, create an enum to represent the goal of the inode iteration. More immediately, this separate removes the need for the "ICI_NOTAG" define which makes little sense. Signed-off-by: Darrick J. Wong Reviewed-by: Dave Chinner --- fs/xfs/xfs_icache.c | 55 +++++++++++++++++++++++++++++++++++++++++------------ fs/xfs/xfs_icache.h | 9 --------- 2 files changed, 43 insertions(+), 21 deletions(-) diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c index d5ecd4cd3ef5..c6d956406033 100644 --- a/fs/xfs/xfs_icache.c +++ b/fs/xfs/xfs_icache.c @@ -26,12 +26,40 @@ #include +/* Radix tree tags for incore inode tree. */ + +/* inode is to be reclaimed */ +#define XFS_ICI_RECLAIM_TAG 0 +/* Inode has speculative preallocations (posteof or cow) to clean. */ +#define XFS_ICI_BLOCKGC_TAG 1 + +/* + * The goal for walking incore inodes. These can correspond with incore inode + * radix tree tags when convenient. Avoid existing XFS_IWALK namespace. + */ +enum xfs_icwalk_goal { + /* Goals that are not related to tags; these must be < 0. */ + XFS_ICWALK_DQRELE = -1, + + /* Goals directly associated with tagged inodes. */ + XFS_ICWALK_BLOCKGC = XFS_ICI_BLOCKGC_TAG, +}; + +#define XFS_ICWALK_NULL_TAG (-1U) + +/* Compute the inode radix tree tag for this goal. */ +static inline unsigned int +xfs_icwalk_tag(enum xfs_icwalk_goal goal) +{ + return goal < 0 ? XFS_ICWALK_NULL_TAG : goal; +} + static int xfs_icwalk(struct xfs_mount *mp, int iter_flags, int (*execute)(struct xfs_inode *ip, void *args), - void *args, int tag); + void *args, enum xfs_icwalk_goal goal); static int xfs_icwalk_ag(struct xfs_perag *pag, int iter_flags, int (*execute)(struct xfs_inode *ip, void *args), - void *args, int tag); + void *args, enum xfs_icwalk_goal goal); /* * Private inode cache walk flags for struct xfs_eofblocks. Must not coincide @@ -791,7 +819,7 @@ xfs_dqrele_all_inodes( eofb.eof_flags |= XFS_ICWALK_FLAG_DROP_PDQUOT; return xfs_icwalk(mp, XFS_INODE_WALK_INEW_WAIT, xfs_dqrele_inode, - &eofb, XFS_ICI_NO_TAG); + &eofb, XFS_ICWALK_DQRELE); } #endif /* CONFIG_XFS_QUOTA */ @@ -1539,7 +1567,7 @@ xfs_blockgc_worker( if (!sb_start_write_trylock(mp->m_super)) return; error = xfs_icwalk_ag(pag, 0, xfs_blockgc_scan_inode, NULL, - XFS_ICI_BLOCKGC_TAG); + XFS_ICWALK_BLOCKGC); if (error) xfs_info(mp, "AG %u preallocation gc worker failed, err=%d", pag->pag_agno, error); @@ -1558,7 +1586,7 @@ xfs_blockgc_free_space( trace_xfs_blockgc_free_space(mp, eofb, _RET_IP_); return xfs_icwalk(mp, 0, xfs_blockgc_scan_inode, eofb, - XFS_ICI_BLOCKGC_TAG); + XFS_ICWALK_BLOCKGC); } /* @@ -1639,7 +1667,7 @@ xfs_icwalk_ag( int iter_flags, int (*execute)(struct xfs_inode *ip, void *args), void *args, - int tag) + enum xfs_icwalk_goal goal) { struct xfs_mount *mp = pag->pag_mount; uint32_t first_index; @@ -1655,12 +1683,13 @@ restart: nr_found = 0; do { struct xfs_inode *batch[XFS_LOOKUP_BATCH]; + unsigned int tag = xfs_icwalk_tag(goal); int error = 0; int i; rcu_read_lock(); - if (tag == XFS_ICI_NO_TAG) + if (tag == XFS_ICWALK_NULL_TAG) nr_found = radix_tree_gang_lookup(&pag->pag_ici_root, (void **)batch, first_index, XFS_LOOKUP_BATCH); @@ -1743,9 +1772,11 @@ static inline struct xfs_perag * xfs_icwalk_get_perag( struct xfs_mount *mp, xfs_agnumber_t agno, - int tag) + enum xfs_icwalk_goal goal) { - if (tag == XFS_ICI_NO_TAG) + unsigned int tag = xfs_icwalk_tag(goal); + + if (tag == XFS_ICWALK_NULL_TAG) return xfs_perag_get(mp, agno); return xfs_perag_get_tag(mp, agno, tag); } @@ -1760,16 +1791,16 @@ xfs_icwalk( int iter_flags, int (*execute)(struct xfs_inode *ip, void *args), void *args, - int tag) + enum xfs_icwalk_goal goal) { struct xfs_perag *pag; int error = 0; int last_error = 0; xfs_agnumber_t agno = 0; - while ((pag = xfs_icwalk_get_perag(mp, agno, tag))) { + while ((pag = xfs_icwalk_get_perag(mp, agno, goal))) { agno = pag->pag_agno + 1; - error = xfs_icwalk_ag(pag, iter_flags, execute, args, tag); + error = xfs_icwalk_ag(pag, iter_flags, execute, args, goal); xfs_perag_put(pag); if (error) { last_error = error; diff --git a/fs/xfs/xfs_icache.h b/fs/xfs/xfs_icache.h index d9baa6df1121..c4274c45d914 100644 --- a/fs/xfs/xfs_icache.h +++ b/fs/xfs/xfs_icache.h @@ -17,15 +17,6 @@ struct xfs_eofblocks { __u64 eof_min_file_size; }; -/* - * tags for inode radix tree - */ -#define XFS_ICI_NO_TAG (-1) /* special flag for an untagged lookup - in xfs_inode_walk */ -#define XFS_ICI_RECLAIM_TAG 0 /* inode is to be reclaimed */ -/* Inode has speculative preallocations (posteof or cow) to clean. */ -#define XFS_ICI_BLOCKGC_TAG 1 - /* * Flags for xfs_iget() */ -- cgit v1.2.3 From b9baaef42f764db7089a19c82d2b783aef836437 Mon Sep 17 00:00:00 2001 From: "Darrick J. Wong" Date: Mon, 31 May 2021 11:31:58 -0700 Subject: xfs: separate the dqrele_all inode grab logic from xfs_inode_walk_ag_grab Disentangle the dqrele_all inode grab code from the "generic" inode walk grabbing code, and and use the opportunity to document why the dqrele grab function does what it does. Since xfs_inode_walk_ag_grab is now only used for blockgc, rename it to reflect that. Ultimately, there will be four reasons to perform a walk of incore inodes: quotaoff dquote releasing (dqrele), garbage collection of speculative preallocations (blockgc), reclamation of incore inodes (reclaim), and deferred inactivation (inodegc). Each of these four have their own slightly different criteria for deciding if they want to handle an inode, so it makes more sense to have four cohesive igrab functions than one confusing parameteric grab function like we do now. Signed-off-by: Darrick J. Wong Reviewed-by: Dave Chinner --- fs/xfs/xfs_icache.c | 71 +++++++++++++++++++++++++++++++++++++++++++++++++---- 1 file changed, 66 insertions(+), 5 deletions(-) diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c index c6d956406033..45979791313f 100644 --- a/fs/xfs/xfs_icache.c +++ b/fs/xfs/xfs_icache.c @@ -774,6 +774,44 @@ xfs_icache_inode_is_allocated( #define XFS_LOOKUP_BATCH 32 #ifdef CONFIG_XFS_QUOTA +/* Decide if we want to grab this inode to drop its dquots. */ +static bool +xfs_dqrele_igrab( + struct xfs_inode *ip) +{ + bool ret = false; + + ASSERT(rcu_read_lock_held()); + + /* Check for stale RCU freed inode */ + spin_lock(&ip->i_flags_lock); + if (!ip->i_ino) + goto out_unlock; + + /* + * Skip inodes that are anywhere in the reclaim machinery because we + * drop dquots before tagging an inode for reclamation. + */ + if (ip->i_flags & (XFS_IRECLAIM | XFS_IRECLAIMABLE)) + goto out_unlock; + + /* + * The inode looks alive; try to grab a VFS reference so that it won't + * get destroyed. If we got the reference, return true to say that + * we grabbed the inode. + * + * If we can't get the reference, then we know the inode had its VFS + * state torn down and hasn't yet entered the reclaim machinery. Since + * we also know that dquots are detached from an inode before it enters + * reclaim, we can skip the inode. + */ + ret = igrab(VFS_I(ip)) != NULL; + +out_unlock: + spin_unlock(&ip->i_flags_lock); + return ret; +} + /* Drop this inode's dquots. */ static int xfs_dqrele_inode( @@ -821,6 +859,8 @@ xfs_dqrele_all_inodes( return xfs_icwalk(mp, XFS_INODE_WALK_INEW_WAIT, xfs_dqrele_inode, &eofb, XFS_ICWALK_DQRELE); } +#else +# define xfs_dqrele_igrab(ip) (false) #endif /* CONFIG_XFS_QUOTA */ /* @@ -1493,12 +1533,12 @@ xfs_blockgc_start( } /* - * Decide if the given @ip is eligible to be a part of the inode walk, and - * grab it if so. Returns true if it's ready to go or false if we should just - * ignore it. + * Decide if the given @ip is eligible for garbage collection of speculative + * preallocations, and grab it if so. Returns true if it's ready to go or + * false if we should just ignore it. */ static bool -xfs_inode_walk_ag_grab( +xfs_blockgc_igrab( struct xfs_inode *ip, int flags) { @@ -1657,6 +1697,27 @@ xfs_blockgc_free_quota( /* XFS Inode Cache Walking Code */ +/* + * Decide if we want to grab this inode in anticipation of doing work towards + * the goal. If selected, the VFS must hold a reference to this inode, which + * will be released after processing. + */ +static inline bool +xfs_icwalk_igrab( + enum xfs_icwalk_goal goal, + struct xfs_inode *ip, + int iter_flags) +{ + switch (goal) { + case XFS_ICWALK_DQRELE: + return xfs_dqrele_igrab(ip); + case XFS_ICWALK_BLOCKGC: + return xfs_blockgc_igrab(ip, iter_flags); + default: + return false; + } +} + /* * For a given per-AG structure @pag, grab, @execute, and rele all incore * inodes with the given radix tree @tag. @@ -1711,7 +1772,7 @@ restart: for (i = 0; i < nr_found; i++) { struct xfs_inode *ip = batch[i]; - if (done || !xfs_inode_walk_ag_grab(ip, iter_flags)) + if (done || !xfs_icwalk_igrab(goal, ip, iter_flags)) batch[i] = NULL; /* -- cgit v1.2.3 From 9d2793ceecb9fd711f70a860685b71129cac5dc9 Mon Sep 17 00:00:00 2001 From: "Darrick J. Wong" Date: Mon, 31 May 2021 11:31:59 -0700 Subject: xfs: move xfs_inew_wait call into xfs_dqrele_inode Move the INEW wait into xfs_dqrele_inode so that we can drop the iter_flags parameter in the next patch. Signed-off-by: Darrick J. Wong Reviewed-by: Dave Chinner --- fs/xfs/xfs_icache.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c index 45979791313f..5f52948f9cfa 100644 --- a/fs/xfs/xfs_icache.c +++ b/fs/xfs/xfs_icache.c @@ -820,6 +820,9 @@ xfs_dqrele_inode( { struct xfs_eofblocks *eofb = priv; + if (xfs_iflags_test(ip, XFS_INEW)) + xfs_inew_wait(ip); + xfs_ilock(ip, XFS_ILOCK_EXCL); if (eofb->eof_flags & XFS_ICWALK_FLAG_DROP_UDQUOT) { xfs_qm_dqrele(ip->i_udquot); @@ -856,8 +859,7 @@ xfs_dqrele_all_inodes( if (qflags & XFS_PQUOTA_ACCT) eofb.eof_flags |= XFS_ICWALK_FLAG_DROP_PDQUOT; - return xfs_icwalk(mp, XFS_INODE_WALK_INEW_WAIT, xfs_dqrele_inode, - &eofb, XFS_ICWALK_DQRELE); + return xfs_icwalk(mp, 0, xfs_dqrele_inode, &eofb, XFS_ICWALK_DQRELE); } #else # define xfs_dqrele_igrab(ip) (false) -- cgit v1.2.3 From 7fdff52623b4df9c9ae665fe8bb727978c29414e Mon Sep 17 00:00:00 2001 From: "Darrick J. Wong" Date: Mon, 31 May 2021 11:31:59 -0700 Subject: xfs: remove iter_flags parameter from xfs_inode_walk_* The sole iter_flags is XFS_INODE_WALK_INEW_WAIT, and there are no users. Remove the flag, and the parameter, and all the code that used it. Signed-off-by: Darrick J. Wong Reviewed-by: Dave Chinner --- fs/xfs/xfs_icache.c | 33 ++++++++++++--------------------- fs/xfs/xfs_icache.h | 5 ----- 2 files changed, 12 insertions(+), 26 deletions(-) diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c index 5f52948f9cfa..b5ce9580934f 100644 --- a/fs/xfs/xfs_icache.c +++ b/fs/xfs/xfs_icache.c @@ -54,10 +54,10 @@ xfs_icwalk_tag(enum xfs_icwalk_goal goal) return goal < 0 ? XFS_ICWALK_NULL_TAG : goal; } -static int xfs_icwalk(struct xfs_mount *mp, int iter_flags, +static int xfs_icwalk(struct xfs_mount *mp, int (*execute)(struct xfs_inode *ip, void *args), void *args, enum xfs_icwalk_goal goal); -static int xfs_icwalk_ag(struct xfs_perag *pag, int iter_flags, +static int xfs_icwalk_ag(struct xfs_perag *pag, int (*execute)(struct xfs_inode *ip, void *args), void *args, enum xfs_icwalk_goal goal); @@ -282,7 +282,7 @@ xfs_inode_clear_reclaim_tag( xfs_perag_clear_reclaim_tag(pag); } -static void +static inline void xfs_inew_wait( struct xfs_inode *ip) { @@ -859,7 +859,7 @@ xfs_dqrele_all_inodes( if (qflags & XFS_PQUOTA_ACCT) eofb.eof_flags |= XFS_ICWALK_FLAG_DROP_PDQUOT; - return xfs_icwalk(mp, 0, xfs_dqrele_inode, &eofb, XFS_ICWALK_DQRELE); + return xfs_icwalk(mp, xfs_dqrele_inode, &eofb, XFS_ICWALK_DQRELE); } #else # define xfs_dqrele_igrab(ip) (false) @@ -1541,11 +1541,9 @@ xfs_blockgc_start( */ static bool xfs_blockgc_igrab( - struct xfs_inode *ip, - int flags) + struct xfs_inode *ip) { struct inode *inode = VFS_I(ip); - bool newinos = !!(flags & XFS_INODE_WALK_INEW_WAIT); ASSERT(rcu_read_lock_held()); @@ -1555,8 +1553,7 @@ xfs_blockgc_igrab( goto out_unlock_noent; /* avoid new or reclaimable inodes. Leave for reclaim code to flush */ - if ((!newinos && __xfs_iflags_test(ip, XFS_INEW)) || - __xfs_iflags_test(ip, XFS_IRECLAIMABLE | XFS_IRECLAIM)) + if (__xfs_iflags_test(ip, XFS_INEW | XFS_IRECLAIMABLE | XFS_IRECLAIM)) goto out_unlock_noent; spin_unlock(&ip->i_flags_lock); @@ -1608,7 +1605,7 @@ xfs_blockgc_worker( if (!sb_start_write_trylock(mp->m_super)) return; - error = xfs_icwalk_ag(pag, 0, xfs_blockgc_scan_inode, NULL, + error = xfs_icwalk_ag(pag, xfs_blockgc_scan_inode, NULL, XFS_ICWALK_BLOCKGC); if (error) xfs_info(mp, "AG %u preallocation gc worker failed, err=%d", @@ -1627,7 +1624,7 @@ xfs_blockgc_free_space( { trace_xfs_blockgc_free_space(mp, eofb, _RET_IP_); - return xfs_icwalk(mp, 0, xfs_blockgc_scan_inode, eofb, + return xfs_icwalk(mp, xfs_blockgc_scan_inode, eofb, XFS_ICWALK_BLOCKGC); } @@ -1707,14 +1704,13 @@ xfs_blockgc_free_quota( static inline bool xfs_icwalk_igrab( enum xfs_icwalk_goal goal, - struct xfs_inode *ip, - int iter_flags) + struct xfs_inode *ip) { switch (goal) { case XFS_ICWALK_DQRELE: return xfs_dqrele_igrab(ip); case XFS_ICWALK_BLOCKGC: - return xfs_blockgc_igrab(ip, iter_flags); + return xfs_blockgc_igrab(ip); default: return false; } @@ -1727,7 +1723,6 @@ xfs_icwalk_igrab( static int xfs_icwalk_ag( struct xfs_perag *pag, - int iter_flags, int (*execute)(struct xfs_inode *ip, void *args), void *args, enum xfs_icwalk_goal goal) @@ -1774,7 +1769,7 @@ restart: for (i = 0; i < nr_found; i++) { struct xfs_inode *ip = batch[i]; - if (done || !xfs_icwalk_igrab(goal, ip, iter_flags)) + if (done || !xfs_icwalk_igrab(goal, ip)) batch[i] = NULL; /* @@ -1802,9 +1797,6 @@ restart: for (i = 0; i < nr_found; i++) { if (!batch[i]) continue; - if ((iter_flags & XFS_INODE_WALK_INEW_WAIT) && - xfs_iflags_test(batch[i], XFS_INEW)) - xfs_inew_wait(batch[i]); error = execute(batch[i], args); xfs_irele(batch[i]); if (error == -EAGAIN) { @@ -1851,7 +1843,6 @@ xfs_icwalk_get_perag( static int xfs_icwalk( struct xfs_mount *mp, - int iter_flags, int (*execute)(struct xfs_inode *ip, void *args), void *args, enum xfs_icwalk_goal goal) @@ -1863,7 +1854,7 @@ xfs_icwalk( while ((pag = xfs_icwalk_get_perag(mp, agno, goal))) { agno = pag->pag_agno + 1; - error = xfs_icwalk_ag(pag, iter_flags, execute, args, goal); + error = xfs_icwalk_ag(pag, execute, args, goal); xfs_perag_put(pag); if (error) { last_error = error; diff --git a/fs/xfs/xfs_icache.h b/fs/xfs/xfs_icache.h index c4274c45d914..3ec00f1fea86 100644 --- a/fs/xfs/xfs_icache.h +++ b/fs/xfs/xfs_icache.h @@ -25,11 +25,6 @@ struct xfs_eofblocks { #define XFS_IGET_DONTCACHE 0x4 #define XFS_IGET_INCORE 0x8 /* don't read from disk or reinit */ -/* - * flags for AG inode iterator - */ -#define XFS_INODE_WALK_INEW_WAIT 0x1 /* wait on new inodes */ - int xfs_iget(struct xfs_mount *mp, struct xfs_trans *tp, xfs_ino_t ino, uint flags, uint lock_flags, xfs_inode_t **ipp); -- cgit v1.2.3 From f427cf5c6236acdf72b4d8564b2e18937c4cc8d8 Mon Sep 17 00:00:00 2001 From: "Darrick J. Wong" Date: Mon, 31 May 2021 11:32:00 -0700 Subject: xfs: remove indirect calls from xfs_inode_walk{,_ag} It turns out that there is a 1:1 mapping between the execute and goal parameters that are passed to xfs_inode_walk_ag: xfs_blockgc_scan_inode <=> XFS_ICWALK_BLOCKGC xfs_dqrele_inode <=> XFS_ICWALK_DQRELE Because of this exact correspondence, we don't need the execute function pointer and can replace it with a direct call. For the price of a forward static declaration, we can eliminate the indirect function call. This likely has a negligible impact on performance (since the execute function runs transactions), but it also simplifies the function signature. Signed-off-by: Darrick J. Wong Reviewed-by: Dave Chinner --- fs/xfs/xfs_icache.c | 60 ++++++++++++++++++++++++++++++++--------------------- 1 file changed, 36 insertions(+), 24 deletions(-) diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c index b5ce9580934f..5ca5bd2ee5ae 100644 --- a/fs/xfs/xfs_icache.c +++ b/fs/xfs/xfs_icache.c @@ -55,11 +55,9 @@ xfs_icwalk_tag(enum xfs_icwalk_goal goal) } static int xfs_icwalk(struct xfs_mount *mp, - int (*execute)(struct xfs_inode *ip, void *args), - void *args, enum xfs_icwalk_goal goal); + enum xfs_icwalk_goal goal, void *args); static int xfs_icwalk_ag(struct xfs_perag *pag, - int (*execute)(struct xfs_inode *ip, void *args), - void *args, enum xfs_icwalk_goal goal); + enum xfs_icwalk_goal goal, void *args); /* * Private inode cache walk flags for struct xfs_eofblocks. Must not coincide @@ -859,10 +857,11 @@ xfs_dqrele_all_inodes( if (qflags & XFS_PQUOTA_ACCT) eofb.eof_flags |= XFS_ICWALK_FLAG_DROP_PDQUOT; - return xfs_icwalk(mp, xfs_dqrele_inode, &eofb, XFS_ICWALK_DQRELE); + return xfs_icwalk(mp, XFS_ICWALK_DQRELE, &eofb); } #else # define xfs_dqrele_igrab(ip) (false) +# define xfs_dqrele_inode(ip, priv) (0) #endif /* CONFIG_XFS_QUOTA */ /* @@ -1605,8 +1604,7 @@ xfs_blockgc_worker( if (!sb_start_write_trylock(mp->m_super)) return; - error = xfs_icwalk_ag(pag, xfs_blockgc_scan_inode, NULL, - XFS_ICWALK_BLOCKGC); + error = xfs_icwalk_ag(pag, XFS_ICWALK_BLOCKGC, NULL); if (error) xfs_info(mp, "AG %u preallocation gc worker failed, err=%d", pag->pag_agno, error); @@ -1624,8 +1622,7 @@ xfs_blockgc_free_space( { trace_xfs_blockgc_free_space(mp, eofb, _RET_IP_); - return xfs_icwalk(mp, xfs_blockgc_scan_inode, eofb, - XFS_ICWALK_BLOCKGC); + return xfs_icwalk(mp, XFS_ICWALK_BLOCKGC, eofb); } /* @@ -1716,16 +1713,36 @@ xfs_icwalk_igrab( } } +/* Process an inode and release it. Return -EAGAIN to skip an inode. */ +static inline int +xfs_icwalk_process_inode( + enum xfs_icwalk_goal goal, + struct xfs_inode *ip, + void *args) +{ + int error; + + switch (goal) { + case XFS_ICWALK_DQRELE: + error = xfs_dqrele_inode(ip, args); + break; + case XFS_ICWALK_BLOCKGC: + error = xfs_blockgc_scan_inode(ip, args); + break; + } + xfs_irele(ip); + return error; +} + /* - * For a given per-AG structure @pag, grab, @execute, and rele all incore - * inodes with the given radix tree @tag. + * For a given per-AG structure @pag and a goal, grab qualifying inodes and + * process them in some manner. */ static int xfs_icwalk_ag( struct xfs_perag *pag, - int (*execute)(struct xfs_inode *ip, void *args), - void *args, - enum xfs_icwalk_goal goal) + enum xfs_icwalk_goal goal, + void *args) { struct xfs_mount *mp = pag->pag_mount; uint32_t first_index; @@ -1797,8 +1814,7 @@ restart: for (i = 0; i < nr_found; i++) { if (!batch[i]) continue; - error = execute(batch[i], args); - xfs_irele(batch[i]); + error = xfs_icwalk_process_inode(goal, batch[i], args); if (error == -EAGAIN) { skipped++; continue; @@ -1836,16 +1852,12 @@ xfs_icwalk_get_perag( return xfs_perag_get_tag(mp, agno, tag); } -/* - * Call the @execute function on all incore inodes matching the radix tree - * @tag. - */ +/* Walk all incore inodes to achieve a given goal. */ static int xfs_icwalk( struct xfs_mount *mp, - int (*execute)(struct xfs_inode *ip, void *args), - void *args, - enum xfs_icwalk_goal goal) + enum xfs_icwalk_goal goal, + void *args) { struct xfs_perag *pag; int error = 0; @@ -1854,7 +1866,7 @@ xfs_icwalk( while ((pag = xfs_icwalk_get_perag(mp, agno, goal))) { agno = pag->pag_agno + 1; - error = xfs_icwalk_ag(pag, execute, args, goal); + error = xfs_icwalk_ag(pag, goal, args); xfs_perag_put(pag); if (error) { last_error = error; -- cgit v1.2.3 From d20d5edcf941e70e03cdbda2f8df93e3969c31a2 Mon Sep 17 00:00:00 2001 From: "Darrick J. Wong" Date: Tue, 1 Jun 2021 23:01:44 -0700 Subject: xfs: clean up inode state flag tests in xfs_blockgc_igrab Clean up the definition of which inode states are not eligible for speculative preallocation garbage collecting by creating a private #define. The deferred inactivation patchset will add two new entries to the set of flags-to-ignore, so we want the definition not to end up a cluttered mess. Signed-off-by: Darrick J. Wong Reviewed-by: Dave Chinner --- fs/xfs/xfs_icache.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c index 5ca5bd2ee5ae..94dba5c1b98d 100644 --- a/fs/xfs/xfs_icache.c +++ b/fs/xfs/xfs_icache.c @@ -1533,6 +1533,10 @@ xfs_blockgc_start( xfs_blockgc_queue(pag); } +/* Don't try to run block gc on an inode that's in any of these states. */ +#define XFS_BLOCKGC_NOGRAB_IFLAGS (XFS_INEW | \ + XFS_IRECLAIMABLE | \ + XFS_IRECLAIM) /* * Decide if the given @ip is eligible for garbage collection of speculative * preallocations, and grab it if so. Returns true if it's ready to go or @@ -1551,8 +1555,7 @@ xfs_blockgc_igrab( if (!ip->i_ino) goto out_unlock_noent; - /* avoid new or reclaimable inodes. Leave for reclaim code to flush */ - if (__xfs_iflags_test(ip, XFS_INEW | XFS_IRECLAIMABLE | XFS_IRECLAIM)) + if (ip->i_flags & XFS_BLOCKGC_NOGRAB_IFLAGS) goto out_unlock_noent; spin_unlock(&ip->i_flags_lock); -- cgit v1.2.3 From 594ab00b760f1722b800c45d37adc21eecf42dc1 Mon Sep 17 00:00:00 2001 From: "Darrick J. Wong" Date: Mon, 31 May 2021 11:32:00 -0700 Subject: xfs: make the icwalk processing functions clean up the grab state Soon we're going to be adding two new callers to the incore inode walk code: reclaim of incore inodes, and (later) inactivation of inodes. Both states operate on inodes that no longer have any VFS state, so we need to move the xfs_irele calls into the processing functions. In other words, icwalk processing functions are responsible for cleaning up whatever state changes are made by the corresponding icwalk igrab function that picked the inode for processing. Signed-off-by: Darrick J. Wong Reviewed-by: Dave Chinner --- fs/xfs/xfs_icache.c | 20 +++++++++++--------- 1 file changed, 11 insertions(+), 9 deletions(-) diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c index 94dba5c1b98d..806faa8df7e9 100644 --- a/fs/xfs/xfs_icache.c +++ b/fs/xfs/xfs_icache.c @@ -811,7 +811,7 @@ out_unlock: } /* Drop this inode's dquots. */ -static int +static void xfs_dqrele_inode( struct xfs_inode *ip, void *priv) @@ -835,7 +835,7 @@ xfs_dqrele_inode( ip->i_pdquot = NULL; } xfs_iunlock(ip, XFS_ILOCK_EXCL); - return 0; + xfs_irele(ip); } /* @@ -861,7 +861,7 @@ xfs_dqrele_all_inodes( } #else # define xfs_dqrele_igrab(ip) (false) -# define xfs_dqrele_inode(ip, priv) (0) +# define xfs_dqrele_inode(ip, priv) ((void)0) #endif /* CONFIG_XFS_QUOTA */ /* @@ -1592,6 +1592,7 @@ xfs_blockgc_scan_inode( unlock: if (lockflags) xfs_iunlock(ip, lockflags); + xfs_irele(ip); return error; } @@ -1698,8 +1699,7 @@ xfs_blockgc_free_quota( /* * Decide if we want to grab this inode in anticipation of doing work towards - * the goal. If selected, the VFS must hold a reference to this inode, which - * will be released after processing. + * the goal. */ static inline bool xfs_icwalk_igrab( @@ -1716,24 +1716,26 @@ xfs_icwalk_igrab( } } -/* Process an inode and release it. Return -EAGAIN to skip an inode. */ +/* + * Process an inode. Each processing function must handle any state changes + * made by the icwalk igrab function. Return -EAGAIN to skip an inode. + */ static inline int xfs_icwalk_process_inode( enum xfs_icwalk_goal goal, struct xfs_inode *ip, void *args) { - int error; + int error = 0; switch (goal) { case XFS_ICWALK_DQRELE: - error = xfs_dqrele_inode(ip, args); + xfs_dqrele_inode(ip, args); break; case XFS_ICWALK_BLOCKGC: error = xfs_blockgc_scan_inode(ip, args); break; } - xfs_irele(ip); return error; } -- cgit v1.2.3 From 919a4ddb68413056ecb7c71d9d5465bb54c8032b Mon Sep 17 00:00:00 2001 From: "Darrick J. Wong" Date: Mon, 31 May 2021 11:32:01 -0700 Subject: xfs: fix radix tree tag signs Radix tree tags are supposed to be unsigned ints, so fix the callers. Signed-off-by: Darrick J. Wong Reviewed-by: Dave Chinner --- fs/xfs/libxfs/xfs_sb.c | 2 +- fs/xfs/libxfs/xfs_sb.h | 4 ++-- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/fs/xfs/libxfs/xfs_sb.c b/fs/xfs/libxfs/xfs_sb.c index dfbbcbd448c1..300d0a1a8049 100644 --- a/fs/xfs/libxfs/xfs_sb.c +++ b/fs/xfs/libxfs/xfs_sb.c @@ -61,7 +61,7 @@ struct xfs_perag * xfs_perag_get_tag( struct xfs_mount *mp, xfs_agnumber_t first, - int tag) + unsigned int tag) { struct xfs_perag *pag; int found; diff --git a/fs/xfs/libxfs/xfs_sb.h b/fs/xfs/libxfs/xfs_sb.h index f79f9dc632b6..e5f1c2d879eb 100644 --- a/fs/xfs/libxfs/xfs_sb.h +++ b/fs/xfs/libxfs/xfs_sb.h @@ -17,8 +17,8 @@ struct xfs_perag; * perag get/put wrappers for ref counting */ extern struct xfs_perag *xfs_perag_get(struct xfs_mount *, xfs_agnumber_t); -extern struct xfs_perag *xfs_perag_get_tag(struct xfs_mount *, xfs_agnumber_t, - int tag); +struct xfs_perag *xfs_perag_get_tag(struct xfs_mount *mp, xfs_agnumber_t agno, + unsigned int tag); extern void xfs_perag_put(struct xfs_perag *pag); extern int xfs_initialize_perag_data(struct xfs_mount *, xfs_agnumber_t); -- cgit v1.2.3 From 9d5ee837595134f91bb2d66f571f498c3b8ab148 Mon Sep 17 00:00:00 2001 From: "Darrick J. Wong" Date: Mon, 31 May 2021 11:32:01 -0700 Subject: xfs: pass struct xfs_eofblocks to the inode scan callback Pass a pointer to the actual eofb structure around the inode scanner functions instead of a void pointer, now that none of the functions is used as a callback. Signed-off-by: Darrick J. Wong Reviewed-by: Christoph Hellwig Reviewed-by: Dave Chinner --- fs/xfs/xfs_icache.c | 34 +++++++++++++++------------------- 1 file changed, 15 insertions(+), 19 deletions(-) diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c index 806faa8df7e9..0c40c39a5f9f 100644 --- a/fs/xfs/xfs_icache.c +++ b/fs/xfs/xfs_icache.c @@ -55,9 +55,9 @@ xfs_icwalk_tag(enum xfs_icwalk_goal goal) } static int xfs_icwalk(struct xfs_mount *mp, - enum xfs_icwalk_goal goal, void *args); + enum xfs_icwalk_goal goal, struct xfs_eofblocks *eofb); static int xfs_icwalk_ag(struct xfs_perag *pag, - enum xfs_icwalk_goal goal, void *args); + enum xfs_icwalk_goal goal, struct xfs_eofblocks *eofb); /* * Private inode cache walk flags for struct xfs_eofblocks. Must not coincide @@ -814,10 +814,8 @@ out_unlock: static void xfs_dqrele_inode( struct xfs_inode *ip, - void *priv) + struct xfs_eofblocks *eofb) { - struct xfs_eofblocks *eofb = priv; - if (xfs_iflags_test(ip, XFS_INEW)) xfs_inew_wait(ip); @@ -1232,10 +1230,9 @@ xfs_reclaim_worker( STATIC int xfs_inode_free_eofblocks( struct xfs_inode *ip, - void *args, + struct xfs_eofblocks *eofb, unsigned int *lockflags) { - struct xfs_eofblocks *eofb = args; bool wait; wait = eofb && (eofb->eof_flags & XFS_EOF_FLAGS_SYNC); @@ -1439,10 +1436,9 @@ xfs_prep_free_cowblocks( STATIC int xfs_inode_free_cowblocks( struct xfs_inode *ip, - void *args, + struct xfs_eofblocks *eofb, unsigned int *lockflags) { - struct xfs_eofblocks *eofb = args; bool wait; int ret = 0; @@ -1579,16 +1575,16 @@ out_unlock_noent: static int xfs_blockgc_scan_inode( struct xfs_inode *ip, - void *args) + struct xfs_eofblocks *eofb) { unsigned int lockflags = 0; int error; - error = xfs_inode_free_eofblocks(ip, args, &lockflags); + error = xfs_inode_free_eofblocks(ip, eofb, &lockflags); if (error) goto unlock; - error = xfs_inode_free_cowblocks(ip, args, &lockflags); + error = xfs_inode_free_cowblocks(ip, eofb, &lockflags); unlock: if (lockflags) xfs_iunlock(ip, lockflags); @@ -1724,16 +1720,16 @@ static inline int xfs_icwalk_process_inode( enum xfs_icwalk_goal goal, struct xfs_inode *ip, - void *args) + struct xfs_eofblocks *eofb) { int error = 0; switch (goal) { case XFS_ICWALK_DQRELE: - xfs_dqrele_inode(ip, args); + xfs_dqrele_inode(ip, eofb); break; case XFS_ICWALK_BLOCKGC: - error = xfs_blockgc_scan_inode(ip, args); + error = xfs_blockgc_scan_inode(ip, eofb); break; } return error; @@ -1747,7 +1743,7 @@ static int xfs_icwalk_ag( struct xfs_perag *pag, enum xfs_icwalk_goal goal, - void *args) + struct xfs_eofblocks *eofb) { struct xfs_mount *mp = pag->pag_mount; uint32_t first_index; @@ -1819,7 +1815,7 @@ restart: for (i = 0; i < nr_found; i++) { if (!batch[i]) continue; - error = xfs_icwalk_process_inode(goal, batch[i], args); + error = xfs_icwalk_process_inode(goal, batch[i], eofb); if (error == -EAGAIN) { skipped++; continue; @@ -1862,7 +1858,7 @@ static int xfs_icwalk( struct xfs_mount *mp, enum xfs_icwalk_goal goal, - void *args) + struct xfs_eofblocks *eofb) { struct xfs_perag *pag; int error = 0; @@ -1871,7 +1867,7 @@ xfs_icwalk( while ((pag = xfs_icwalk_get_perag(mp, agno, goal))) { agno = pag->pag_agno + 1; - error = xfs_icwalk_ag(pag, goal, args); + error = xfs_icwalk_ag(pag, goal, eofb); xfs_perag_put(pag); if (error) { last_error = error; -- cgit v1.2.3 From f1bc5c5630f90b83b339e8970dcf6d03abba5bd5 Mon Sep 17 00:00:00 2001 From: "Darrick J. Wong" Date: Mon, 31 May 2021 11:32:02 -0700 Subject: xfs: merge xfs_reclaim_inodes_ag into xfs_inode_walk_ag Merge these two inode walk loops together, since they're pretty similar now. Signed-off-by: Darrick J. Wong Reviewed-by: Dave Chinner --- fs/xfs/xfs_icache.c | 162 ++++++++++++++++------------------------------------ fs/xfs/xfs_icache.h | 1 + fs/xfs/xfs_trace.h | 5 +- 3 files changed, 53 insertions(+), 115 deletions(-) diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c index 0c40c39a5f9f..1223921fb01c 100644 --- a/fs/xfs/xfs_icache.c +++ b/fs/xfs/xfs_icache.c @@ -43,6 +43,7 @@ enum xfs_icwalk_goal { /* Goals directly associated with tagged inodes. */ XFS_ICWALK_BLOCKGC = XFS_ICI_BLOCKGC_TAG, + XFS_ICWALK_RECLAIM = XFS_ICI_RECLAIM_TAG, }; #define XFS_ICWALK_NULL_TAG (-1U) @@ -67,9 +68,13 @@ static int xfs_icwalk_ag(struct xfs_perag *pag, #define XFS_ICWALK_FLAG_DROP_GDQUOT (1U << 30) #define XFS_ICWALK_FLAG_DROP_PDQUOT (1U << 29) +/* Stop scanning after icw_scan_limit inodes. */ +#define XFS_ICWALK_FLAG_SCAN_LIMIT (1U << 28) + #define XFS_ICWALK_PRIVATE_FLAGS (XFS_ICWALK_FLAG_DROP_UDQUOT | \ XFS_ICWALK_FLAG_DROP_GDQUOT | \ - XFS_ICWALK_FLAG_DROP_PDQUOT) + XFS_ICWALK_FLAG_DROP_PDQUOT | \ + XFS_ICWALK_FLAG_SCAN_LIMIT) /* * Allocate and initialise an xfs_inode. @@ -760,17 +765,6 @@ xfs_icache_inode_is_allocated( return 0; } -/* - * The inode lookup is done in batches to keep the amount of lock traffic and - * radix tree lookups to a minimum. The batch size is a trade off between - * lookup reduction and stack usage. This is in the reclaim path, so we can't - * be too greedy. - * - * XXX: This will be moved closer to xfs_icwalk* once we get rid of the - * separate reclaim walk functions. - */ -#define XFS_LOOKUP_BATCH 32 - #ifdef CONFIG_XFS_QUOTA /* Decide if we want to grab this inode to drop its dquots. */ static bool @@ -880,7 +874,7 @@ xfs_dqrele_all_inodes( * Return true if we grabbed it, false otherwise. */ static bool -xfs_reclaim_inode_grab( +xfs_reclaim_igrab( struct xfs_inode *ip) { ASSERT(rcu_read_lock_held()); @@ -990,108 +984,13 @@ out: xfs_iflags_clear(ip, XFS_IRECLAIM); } -/* - * Walk the AGs and reclaim the inodes in them. Even if the filesystem is - * corrupted, we still want to try to reclaim all the inodes. If we don't, - * then a shut down during filesystem unmount reclaim walk leak all the - * unreclaimed inodes. - * - * Returns non-zero if any AGs or inodes were skipped in the reclaim pass - * so that callers that want to block until all dirty inodes are written back - * and reclaimed can sanely loop. - */ -static void -xfs_reclaim_inodes_ag( - struct xfs_mount *mp, - int *nr_to_scan) -{ - struct xfs_perag *pag; - xfs_agnumber_t ag = 0; - - while ((pag = xfs_perag_get_tag(mp, ag, XFS_ICI_RECLAIM_TAG))) { - unsigned long first_index = 0; - int done = 0; - int nr_found = 0; - - ag = pag->pag_agno + 1; - - first_index = READ_ONCE(pag->pag_ici_reclaim_cursor); - do { - struct xfs_inode *batch[XFS_LOOKUP_BATCH]; - int i; - - rcu_read_lock(); - nr_found = radix_tree_gang_lookup_tag( - &pag->pag_ici_root, - (void **)batch, first_index, - XFS_LOOKUP_BATCH, - XFS_ICI_RECLAIM_TAG); - if (!nr_found) { - done = 1; - rcu_read_unlock(); - break; - } - - /* - * Grab the inodes before we drop the lock. if we found - * nothing, nr == 0 and the loop will be skipped. - */ - for (i = 0; i < nr_found; i++) { - struct xfs_inode *ip = batch[i]; - - if (done || !xfs_reclaim_inode_grab(ip)) - batch[i] = NULL; - - /* - * 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. - * - * Because we may see inodes that are from the - * wrong AG due to RCU freeing and - * reallocation, only update the index if it - * lies in this AG. It was a race that lead us - * to see this inode, so another lookup from - * the same index will not find it again. - */ - if (XFS_INO_TO_AGNO(mp, ip->i_ino) != - pag->pag_agno) - continue; - first_index = XFS_INO_TO_AGINO(mp, ip->i_ino + 1); - if (first_index < XFS_INO_TO_AGINO(mp, ip->i_ino)) - done = 1; - } - - /* unlock now we've grabbed the inodes. */ - rcu_read_unlock(); - - for (i = 0; i < nr_found; i++) { - if (batch[i]) - xfs_reclaim_inode(batch[i], pag); - } - - *nr_to_scan -= XFS_LOOKUP_BATCH; - cond_resched(); - } while (nr_found && !done && *nr_to_scan > 0); - - if (done) - first_index = 0; - WRITE_ONCE(pag->pag_ici_reclaim_cursor, first_index); - xfs_perag_put(pag); - } -} - void xfs_reclaim_inodes( struct xfs_mount *mp) { - int nr_to_scan = INT_MAX; - while (radix_tree_tagged(&mp->m_perag_tree, XFS_ICI_RECLAIM_TAG)) { xfs_ail_push_all_sync(mp->m_ail); - xfs_reclaim_inodes_ag(mp, &nr_to_scan); + xfs_icwalk(mp, XFS_ICWALK_RECLAIM, NULL); } } @@ -1107,11 +1006,16 @@ xfs_reclaim_inodes_nr( struct xfs_mount *mp, int nr_to_scan) { + struct xfs_eofblocks eofb = { + .eof_flags = XFS_ICWALK_FLAG_SCAN_LIMIT, + .icw_scan_limit = nr_to_scan, + }; + /* kick background reclaimer and push the AIL */ xfs_reclaim_work_queue(mp); xfs_ail_push_all(mp->m_ail); - xfs_reclaim_inodes_ag(mp, &nr_to_scan); + xfs_icwalk(mp, XFS_ICWALK_RECLAIM, &eofb); return 0; } @@ -1221,9 +1125,8 @@ xfs_reclaim_worker( { struct xfs_mount *mp = container_of(to_delayed_work(work), struct xfs_mount, m_reclaim_work); - int nr_to_scan = INT_MAX; - xfs_reclaim_inodes_ag(mp, &nr_to_scan); + xfs_icwalk(mp, XFS_ICWALK_RECLAIM, NULL); xfs_reclaim_work_queue(mp); } @@ -1693,6 +1596,15 @@ xfs_blockgc_free_quota( /* XFS Inode Cache Walking Code */ +/* + * The inode lookup is done in batches to keep the amount of lock traffic and + * radix tree lookups to a minimum. The batch size is a trade off between + * lookup reduction and stack usage. This is in the reclaim path, so we can't + * be too greedy. + */ +#define XFS_LOOKUP_BATCH 32 + + /* * Decide if we want to grab this inode in anticipation of doing work towards * the goal. @@ -1707,6 +1619,8 @@ xfs_icwalk_igrab( return xfs_dqrele_igrab(ip); case XFS_ICWALK_BLOCKGC: return xfs_blockgc_igrab(ip); + case XFS_ICWALK_RECLAIM: + return xfs_reclaim_igrab(ip); default: return false; } @@ -1720,6 +1634,7 @@ static inline int xfs_icwalk_process_inode( enum xfs_icwalk_goal goal, struct xfs_inode *ip, + struct xfs_perag *pag, struct xfs_eofblocks *eofb) { int error = 0; @@ -1731,6 +1646,9 @@ xfs_icwalk_process_inode( case XFS_ICWALK_BLOCKGC: error = xfs_blockgc_scan_inode(ip, eofb); break; + case XFS_ICWALK_RECLAIM: + xfs_reclaim_inode(ip, pag); + break; } return error; } @@ -1755,7 +1673,10 @@ xfs_icwalk_ag( restart: done = false; skipped = 0; - first_index = 0; + if (goal == XFS_ICWALK_RECLAIM) + first_index = READ_ONCE(pag->pag_ici_reclaim_cursor); + else + first_index = 0; nr_found = 0; do { struct xfs_inode *batch[XFS_LOOKUP_BATCH]; @@ -1776,6 +1697,7 @@ restart: XFS_LOOKUP_BATCH, tag); if (!nr_found) { + done = true; rcu_read_unlock(); break; } @@ -1815,7 +1737,8 @@ restart: for (i = 0; i < nr_found; i++) { if (!batch[i]) continue; - error = xfs_icwalk_process_inode(goal, batch[i], eofb); + error = xfs_icwalk_process_inode(goal, batch[i], pag, + eofb); if (error == -EAGAIN) { skipped++; continue; @@ -1830,8 +1753,19 @@ restart: cond_resched(); + if (eofb && (eofb->eof_flags & XFS_ICWALK_FLAG_SCAN_LIMIT)) { + eofb->icw_scan_limit -= XFS_LOOKUP_BATCH; + if (eofb->icw_scan_limit <= 0) + break; + } } while (nr_found && !done); + if (goal == XFS_ICWALK_RECLAIM) { + if (done) + first_index = 0; + WRITE_ONCE(pag->pag_ici_reclaim_cursor, first_index); + } + if (skipped) { delay(1); goto restart; diff --git a/fs/xfs/xfs_icache.h b/fs/xfs/xfs_icache.h index 3ec00f1fea86..b6ab1067c52b 100644 --- a/fs/xfs/xfs_icache.h +++ b/fs/xfs/xfs_icache.h @@ -15,6 +15,7 @@ struct xfs_eofblocks { kgid_t eof_gid; prid_t eof_prid; __u64 eof_min_file_size; + int icw_scan_limit; }; /* diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h index 808ae337b222..1377b1e24e1d 100644 --- a/fs/xfs/xfs_trace.h +++ b/fs/xfs/xfs_trace.h @@ -3898,6 +3898,7 @@ DECLARE_EVENT_CLASS(xfs_eofblocks_class, __field(uint32_t, gid) __field(prid_t, prid) __field(__u64, min_file_size) + __field(int, scan_limit) __field(unsigned long, caller_ip) ), TP_fast_assign( @@ -3909,15 +3910,17 @@ DECLARE_EVENT_CLASS(xfs_eofblocks_class, eofb->eof_gid) : 0; __entry->prid = eofb ? eofb->eof_prid : 0; __entry->min_file_size = eofb ? eofb->eof_min_file_size : 0; + __entry->scan_limit = eofb ? eofb->icw_scan_limit : 0; __entry->caller_ip = caller_ip; ), - TP_printk("dev %d:%d flags 0x%x uid %u gid %u prid %u minsize %llu caller %pS", + TP_printk("dev %d:%d flags 0x%x uid %u gid %u prid %u minsize %llu scan_limit %d caller %pS", MAJOR(__entry->dev), MINOR(__entry->dev), __entry->flags, __entry->uid, __entry->gid, __entry->prid, __entry->min_file_size, + __entry->scan_limit, (char *)__entry->caller_ip) ); #define DEFINE_EOFBLOCKS_EVENT(name) \ -- cgit v1.2.3 From c076ae7a9361b87624900c722012a837fee0b1b3 Mon Sep 17 00:00:00 2001 From: "Darrick J. Wong" Date: Mon, 31 May 2021 11:32:02 -0700 Subject: xfs: refactor per-AG inode tagging functions In preparation for adding another incore inode tree tag, refactor the code that sets and clears tags from the per-AG inode tree and the tree of per-AG structures, and remove the open-coded versions used by the blockgc code. Note: For reclaim, we now rely on the radix tree tags instead of the reclaimable inode count more heavily than we used to. The conversion should be fine, but the logic isn't 100% identical. Signed-off-by: Darrick J. Wong Reviewed-by: Dave Chinner --- fs/xfs/xfs_icache.c | 158 +++++++++++++++++++++++++--------------------------- fs/xfs/xfs_icache.h | 2 +- fs/xfs/xfs_super.c | 2 +- fs/xfs/xfs_trace.h | 6 +- 4 files changed, 80 insertions(+), 88 deletions(-) diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c index 1223921fb01c..396cc54ca03f 100644 --- a/fs/xfs/xfs_icache.c +++ b/fs/xfs/xfs_icache.c @@ -207,46 +207,94 @@ xfs_reclaim_work_queue( rcu_read_unlock(); } -static void -xfs_perag_set_reclaim_tag( +/* + * Background scanning to trim preallocated space. This is queued based on the + * 'speculative_prealloc_lifetime' tunable (5m by default). + */ +static inline void +xfs_blockgc_queue( struct xfs_perag *pag) +{ + rcu_read_lock(); + if (radix_tree_tagged(&pag->pag_ici_root, XFS_ICI_BLOCKGC_TAG)) + queue_delayed_work(pag->pag_mount->m_gc_workqueue, + &pag->pag_blockgc_work, + msecs_to_jiffies(xfs_blockgc_secs * 1000)); + rcu_read_unlock(); +} + +/* Set a tag on both the AG incore inode tree and the AG radix tree. */ +static void +xfs_perag_set_inode_tag( + struct xfs_perag *pag, + xfs_agino_t agino, + unsigned int tag) { struct xfs_mount *mp = pag->pag_mount; + bool was_tagged; lockdep_assert_held(&pag->pag_ici_lock); - if (pag->pag_ici_reclaimable++) + + was_tagged = radix_tree_tagged(&pag->pag_ici_root, tag); + radix_tree_tag_set(&pag->pag_ici_root, agino, tag); + + if (tag == XFS_ICI_RECLAIM_TAG) + pag->pag_ici_reclaimable++; + + if (was_tagged) return; - /* propagate the reclaim tag up into the perag radix tree */ + /* propagate the tag up into the perag radix tree */ spin_lock(&mp->m_perag_lock); - radix_tree_tag_set(&mp->m_perag_tree, pag->pag_agno, - XFS_ICI_RECLAIM_TAG); + radix_tree_tag_set(&mp->m_perag_tree, pag->pag_agno, tag); spin_unlock(&mp->m_perag_lock); - /* schedule periodic background inode reclaim */ - xfs_reclaim_work_queue(mp); + /* start background work */ + switch (tag) { + case XFS_ICI_RECLAIM_TAG: + xfs_reclaim_work_queue(mp); + break; + case XFS_ICI_BLOCKGC_TAG: + xfs_blockgc_queue(pag); + break; + } - trace_xfs_perag_set_reclaim(mp, pag->pag_agno, -1, _RET_IP_); + trace_xfs_perag_set_inode_tag(mp, pag->pag_agno, tag, _RET_IP_); } +/* Clear a tag on both the AG incore inode tree and the AG radix tree. */ static void -xfs_perag_clear_reclaim_tag( - struct xfs_perag *pag) +xfs_perag_clear_inode_tag( + struct xfs_perag *pag, + xfs_agino_t agino, + unsigned int tag) { struct xfs_mount *mp = pag->pag_mount; lockdep_assert_held(&pag->pag_ici_lock); - if (--pag->pag_ici_reclaimable) + + /* + * Reclaim can signal (with a null agino) that it cleared its own tag + * by removing the inode from the radix tree. + */ + if (agino != NULLAGINO) + radix_tree_tag_clear(&pag->pag_ici_root, agino, tag); + else + ASSERT(tag == XFS_ICI_RECLAIM_TAG); + + if (tag == XFS_ICI_RECLAIM_TAG) + pag->pag_ici_reclaimable--; + + if (radix_tree_tagged(&pag->pag_ici_root, tag)) return; - /* clear the reclaim tag from the perag radix tree */ + /* clear the tag from the perag radix tree */ spin_lock(&mp->m_perag_lock); - radix_tree_tag_clear(&mp->m_perag_tree, pag->pag_agno, - XFS_ICI_RECLAIM_TAG); + radix_tree_tag_clear(&mp->m_perag_tree, pag->pag_agno, tag); spin_unlock(&mp->m_perag_lock); - trace_xfs_perag_clear_reclaim(mp, pag->pag_agno, -1, _RET_IP_); -} + trace_xfs_perag_clear_inode_tag(mp, pag->pag_agno, tag, _RET_IP_); +} /* * We set the inode flag atomically with the radix tree tag. @@ -254,7 +302,7 @@ xfs_perag_clear_reclaim_tag( * can go away. */ void -xfs_inode_set_reclaim_tag( +xfs_inode_mark_reclaimable( struct xfs_inode *ip) { struct xfs_mount *mp = ip->i_mount; @@ -264,9 +312,8 @@ xfs_inode_set_reclaim_tag( spin_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_perag_set_reclaim_tag(pag); + xfs_perag_set_inode_tag(pag, XFS_INO_TO_AGINO(mp, ip->i_ino), + XFS_ICI_RECLAIM_TAG); __xfs_iflags_set(ip, XFS_IRECLAIMABLE); spin_unlock(&ip->i_flags_lock); @@ -274,17 +321,6 @@ xfs_inode_set_reclaim_tag( xfs_perag_put(pag); } -STATIC void -xfs_inode_clear_reclaim_tag( - struct xfs_perag *pag, - xfs_ino_t ino) -{ - radix_tree_tag_clear(&pag->pag_ici_root, - XFS_INO_TO_AGINO(pag->pag_mount, ino), - XFS_ICI_RECLAIM_TAG); - xfs_perag_clear_reclaim_tag(pag); -} - static inline void xfs_inew_wait( struct xfs_inode *ip) @@ -483,7 +519,9 @@ xfs_iget_cache_hit( */ ip->i_flags &= ~XFS_IRECLAIM_RESET_FLAGS; ip->i_flags |= XFS_INEW; - xfs_inode_clear_reclaim_tag(pag, ip->i_ino); + xfs_perag_clear_inode_tag(pag, + XFS_INO_TO_AGINO(pag->pag_mount, ino), + XFS_ICI_RECLAIM_TAG); inode->i_state = I_NEW; ip->i_sick = 0; ip->i_checked = 0; @@ -957,7 +995,7 @@ reclaim: if (!radix_tree_delete(&pag->pag_ici_root, XFS_INO_TO_AGINO(ip->i_mount, ino))) ASSERT(0); - xfs_perag_clear_reclaim_tag(pag); + xfs_perag_clear_inode_tag(pag, NULLAGINO, XFS_ICI_RECLAIM_TAG); spin_unlock(&pag->pag_ici_lock); /* @@ -1173,22 +1211,6 @@ xfs_inode_free_eofblocks( return 0; } -/* - * Background scanning to trim preallocated space. This is queued based on the - * 'speculative_prealloc_lifetime' tunable (5m by default). - */ -static inline void -xfs_blockgc_queue( - struct xfs_perag *pag) -{ - rcu_read_lock(); - if (radix_tree_tagged(&pag->pag_ici_root, XFS_ICI_BLOCKGC_TAG)) - queue_delayed_work(pag->pag_mount->m_gc_workqueue, - &pag->pag_blockgc_work, - msecs_to_jiffies(xfs_blockgc_secs * 1000)); - rcu_read_unlock(); -} - static void xfs_blockgc_set_iflag( struct xfs_inode *ip, @@ -1196,7 +1218,6 @@ xfs_blockgc_set_iflag( { struct xfs_mount *mp = ip->i_mount; struct xfs_perag *pag; - int tagged; ASSERT((iflag & ~(XFS_IEOFBLOCKS | XFS_ICOWBLOCKS)) == 0); @@ -1213,24 +1234,8 @@ xfs_blockgc_set_iflag( pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, ip->i_ino)); spin_lock(&pag->pag_ici_lock); - tagged = radix_tree_tagged(&pag->pag_ici_root, XFS_ICI_BLOCKGC_TAG); - radix_tree_tag_set(&pag->pag_ici_root, - XFS_INO_TO_AGINO(ip->i_mount, ip->i_ino), - XFS_ICI_BLOCKGC_TAG); - if (!tagged) { - /* propagate the blockgc tag up into the perag radix tree */ - spin_lock(&ip->i_mount->m_perag_lock); - radix_tree_tag_set(&ip->i_mount->m_perag_tree, - XFS_INO_TO_AGNO(ip->i_mount, ip->i_ino), - XFS_ICI_BLOCKGC_TAG); - spin_unlock(&ip->i_mount->m_perag_lock); - - /* kick off background trimming */ - xfs_blockgc_queue(pag); - - trace_xfs_perag_set_blockgc(ip->i_mount, pag->pag_agno, -1, - _RET_IP_); - } + xfs_perag_set_inode_tag(pag, XFS_INO_TO_AGINO(mp, ip->i_ino), + XFS_ICI_BLOCKGC_TAG); spin_unlock(&pag->pag_ici_lock); xfs_perag_put(pag); @@ -1266,19 +1271,8 @@ xfs_blockgc_clear_iflag( pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, ip->i_ino)); spin_lock(&pag->pag_ici_lock); - radix_tree_tag_clear(&pag->pag_ici_root, - XFS_INO_TO_AGINO(ip->i_mount, ip->i_ino), - XFS_ICI_BLOCKGC_TAG); - if (!radix_tree_tagged(&pag->pag_ici_root, XFS_ICI_BLOCKGC_TAG)) { - /* clear the blockgc tag from the perag radix tree */ - spin_lock(&ip->i_mount->m_perag_lock); - radix_tree_tag_clear(&ip->i_mount->m_perag_tree, - XFS_INO_TO_AGNO(ip->i_mount, ip->i_ino), - XFS_ICI_BLOCKGC_TAG); - spin_unlock(&ip->i_mount->m_perag_lock); - trace_xfs_perag_clear_blockgc(ip->i_mount, pag->pag_agno, -1, - _RET_IP_); - } + xfs_perag_clear_inode_tag(pag, XFS_INO_TO_AGINO(mp, ip->i_ino), + XFS_ICI_BLOCKGC_TAG); spin_unlock(&pag->pag_ici_lock); xfs_perag_put(pag); diff --git a/fs/xfs/xfs_icache.h b/fs/xfs/xfs_icache.h index b6ab1067c52b..191620a069af 100644 --- a/fs/xfs/xfs_icache.h +++ b/fs/xfs/xfs_icache.h @@ -39,7 +39,7 @@ void xfs_reclaim_inodes(struct xfs_mount *mp); int xfs_reclaim_inodes_count(struct xfs_mount *mp); long xfs_reclaim_inodes_nr(struct xfs_mount *mp, int nr_to_scan); -void xfs_inode_set_reclaim_tag(struct xfs_inode *ip); +void xfs_inode_mark_reclaimable(struct xfs_inode *ip); int xfs_blockgc_free_dquots(struct xfs_mount *mp, struct xfs_dquot *udqp, struct xfs_dquot *gdqp, struct xfs_dquot *pdqp, diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c index a2dab05332ac..db61e9cdc013 100644 --- a/fs/xfs/xfs_super.c +++ b/fs/xfs/xfs_super.c @@ -667,7 +667,7 @@ xfs_fs_destroy_inode( * reclaim path handles this more efficiently than we can here, so * simply let background reclaim tear down all inodes. */ - xfs_inode_set_reclaim_tag(ip); + xfs_inode_mark_reclaimable(ip); } static void diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h index 1377b1e24e1d..0171d93239a2 100644 --- a/fs/xfs/xfs_trace.h +++ b/fs/xfs/xfs_trace.h @@ -153,10 +153,8 @@ DEFINE_EVENT(xfs_perag_class, name, \ DEFINE_PERAG_REF_EVENT(xfs_perag_get); DEFINE_PERAG_REF_EVENT(xfs_perag_get_tag); DEFINE_PERAG_REF_EVENT(xfs_perag_put); -DEFINE_PERAG_REF_EVENT(xfs_perag_set_reclaim); -DEFINE_PERAG_REF_EVENT(xfs_perag_clear_reclaim); -DEFINE_PERAG_REF_EVENT(xfs_perag_set_blockgc); -DEFINE_PERAG_REF_EVENT(xfs_perag_clear_blockgc); +DEFINE_PERAG_REF_EVENT(xfs_perag_set_inode_tag); +DEFINE_PERAG_REF_EVENT(xfs_perag_clear_inode_tag); DECLARE_EVENT_CLASS(xfs_ag_class, TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t agno), -- cgit v1.2.3