xfs: abort intent items when recovery intents fail
authorLong Li <leo.lilong@huawei.com>
Mon, 31 Jul 2023 12:46:18 +0000 (20:46 +0800)
committerChandan Babu R <chandanbabu@kernel.org>
Mon, 13 Nov 2023 03:38:34 +0000 (09:08 +0530)
When recovering intents, we capture newly created intent items as part of
committing recovered intent items.  If intent recovery fails at a later
point, we forget to remove those newly created intent items from the AIL
and hang:

    [root@localhost ~]# cat /proc/539/stack
    [<0>] xfs_ail_push_all_sync+0x174/0x230
    [<0>] xfs_unmount_flush_inodes+0x8d/0xd0
    [<0>] xfs_mountfs+0x15f7/0x1e70
    [<0>] xfs_fs_fill_super+0x10ec/0x1b20
    [<0>] get_tree_bdev+0x3c8/0x730
    [<0>] vfs_get_tree+0x89/0x2c0
    [<0>] path_mount+0xecf/0x1800
    [<0>] do_mount+0xf3/0x110
    [<0>] __x64_sys_mount+0x154/0x1f0
    [<0>] do_syscall_64+0x39/0x80
    [<0>] entry_SYSCALL_64_after_hwframe+0x63/0xcd

When newly created intent items fail to commit via transaction, intent
recovery hasn't created done items for these newly created intent items,
so the capture structure is the sole owner of the captured intent items.
We must release them explicitly or else they leak:

unreferenced object 0xffff888016719108 (size 432):
  comm "mount", pid 529, jiffies 4294706839 (age 144.463s)
  hex dump (first 32 bytes):
    08 91 71 16 80 88 ff ff 08 91 71 16 80 88 ff ff  ..q.......q.....
    18 91 71 16 80 88 ff ff 18 91 71 16 80 88 ff ff  ..q.......q.....
  backtrace:
    [<ffffffff8230c68f>] xfs_efi_init+0x18f/0x1d0
    [<ffffffff8230c720>] xfs_extent_free_create_intent+0x50/0x150
    [<ffffffff821b671a>] xfs_defer_create_intents+0x16a/0x340
    [<ffffffff821bac3e>] xfs_defer_ops_capture_and_commit+0x8e/0xad0
    [<ffffffff82322bb9>] xfs_cui_item_recover+0x819/0x980
    [<ffffffff823289b6>] xlog_recover_process_intents+0x246/0xb70
    [<ffffffff8233249a>] xlog_recover_finish+0x8a/0x9a0
    [<ffffffff822eeafb>] xfs_log_mount_finish+0x2bb/0x4a0
    [<ffffffff822c0f4f>] xfs_mountfs+0x14bf/0x1e70
    [<ffffffff822d1f80>] xfs_fs_fill_super+0x10d0/0x1b20
    [<ffffffff81a21fa2>] get_tree_bdev+0x3d2/0x6d0
    [<ffffffff81a1ee09>] vfs_get_tree+0x89/0x2c0
    [<ffffffff81a9f35f>] path_mount+0xecf/0x1800
    [<ffffffff81a9fd83>] do_mount+0xf3/0x110
    [<ffffffff81aa00e4>] __x64_sys_mount+0x154/0x1f0
    [<ffffffff83968739>] do_syscall_64+0x39/0x80

Fix the problem above by abort intent items that don't have a done item
when recovery intents fail.

Fixes: e6fff81e4870 ("xfs: proper replay of deferred ops queued during log recovery")
Signed-off-by: Long Li <leo.lilong@huawei.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Chandan Babu R <chandanbabu@kernel.org>
fs/xfs/libxfs/xfs_defer.c
fs/xfs/libxfs/xfs_defer.h
fs/xfs/xfs_log_recover.c

index 88388e12f8e7cbfad1cb5d2a36c55c956e77647c..f71679ce23b95da4d07536581d15711eafa7fd46 100644 (file)
@@ -763,12 +763,13 @@ xfs_defer_ops_capture(
 
 /* Release all resources that we used to capture deferred ops. */
 void
-xfs_defer_ops_capture_free(
+xfs_defer_ops_capture_abort(
        struct xfs_mount                *mp,
        struct xfs_defer_capture        *dfc)
 {
        unsigned short                  i;
 
+       xfs_defer_pending_abort(mp, &dfc->dfc_dfops);
        xfs_defer_cancel_list(mp, &dfc->dfc_dfops);
 
        for (i = 0; i < dfc->dfc_held.dr_bufs; i++)
@@ -809,7 +810,7 @@ xfs_defer_ops_capture_and_commit(
        /* Commit the transaction and add the capture structure to the list. */
        error = xfs_trans_commit(tp);
        if (error) {
-               xfs_defer_ops_capture_free(mp, dfc);
+               xfs_defer_ops_capture_abort(mp, dfc);
                return error;
        }
 
index 114a3a4930a3c47080a1e3e38bf06f3b17cb1bda..8788ad5f6a731fbe0fa20faa641eeaa8679a0a98 100644 (file)
@@ -121,7 +121,7 @@ int xfs_defer_ops_capture_and_commit(struct xfs_trans *tp,
                struct list_head *capture_list);
 void xfs_defer_ops_continue(struct xfs_defer_capture *d, struct xfs_trans *tp,
                struct xfs_defer_resources *dres);
-void xfs_defer_ops_capture_free(struct xfs_mount *mp,
+void xfs_defer_ops_capture_abort(struct xfs_mount *mp,
                struct xfs_defer_capture *d);
 void xfs_defer_resources_rele(struct xfs_defer_resources *dres);
 
index 13b94d2e605bd9ddb852f76eeeea0bd956a595e0..a1e18b24971a28eedcf79b4d43b16778b42133f8 100644 (file)
@@ -2511,7 +2511,7 @@ xlog_abort_defer_ops(
 
        list_for_each_entry_safe(dfc, next, capture_list, dfc_list) {
                list_del_init(&dfc->dfc_list);
-               xfs_defer_ops_capture_free(mp, dfc);
+               xfs_defer_ops_capture_abort(mp, dfc);
        }
 }