Thread (2 messages) 2 messages, 1 author, 2021-09-14

Re: [PATCH] btrfs: unlock the original extent buffer when error happens in __btrfs_cow_block()

From: Qu Wenruo <hidden>
Date: 2021-09-14 06:24:52


On 2021/9/14 下午1:55, Qu Wenruo wrote:
[BUG]
There is a very detailed bug report that injected ENOMEM error could
leave a tree block locked while we return to user-space:

   BTRFS info (device loop0): enabling ssd optimizations
   FAULT_INJECTION: forcing a failure.
   name failslab, interval 1, probability 0, space 0, times 0
   CPU: 0 PID: 7579 Comm: syz-executor Not tainted 5.15.0-rc1 #16
   Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS
   rel-1.12.0-59-gc9ba5276e321-prebuilt.qemu.org 04/01/2014
   Call Trace:
    __dump_stack lib/dump_stack.c:88 [inline]
    dump_stack_lvl+0x8d/0xcf lib/dump_stack.c:106
    fail_dump lib/fault-inject.c:52 [inline]
    should_fail+0x13c/0x160 lib/fault-inject.c:146
    should_failslab+0x5/0x10 mm/slab_common.c:1328
    slab_pre_alloc_hook.constprop.99+0x4e/0xc0 mm/slab.h:494
    slab_alloc_node mm/slub.c:3120 [inline]
    slab_alloc mm/slub.c:3214 [inline]
    kmem_cache_alloc+0x44/0x280 mm/slub.c:3219
    btrfs_alloc_delayed_extent_op fs/btrfs/delayed-ref.h:299 [inline]
    btrfs_alloc_tree_block+0x38c/0x670 fs/btrfs/extent-tree.c:4833
    __btrfs_cow_block+0x16f/0x7d0 fs/btrfs/ctree.c:415
    btrfs_cow_block+0x12a/0x300 fs/btrfs/ctree.c:570
    btrfs_search_slot+0x6b0/0xee0 fs/btrfs/ctree.c:1768
    btrfs_insert_empty_items+0x80/0xf0 fs/btrfs/ctree.c:3905
    btrfs_new_inode+0x311/0xa60 fs/btrfs/inode.c:6530
    btrfs_create+0x12b/0x270 fs/btrfs/inode.c:6783
    lookup_open+0x660/0x780 fs/namei.c:3282
    open_last_lookups fs/namei.c:3352 [inline]
    path_openat+0x465/0xe20 fs/namei.c:3557
    do_filp_open+0xe3/0x170 fs/namei.c:3588
    do_sys_openat2+0x357/0x4a0 fs/open.c:1200
    do_sys_open+0x87/0xd0 fs/open.c:1216
    do_syscall_x64 arch/x86/entry/common.c:50 [inline]
    do_syscall_64+0x34/0xb0 arch/x86/entry/common.c:80
    entry_SYSCALL_64_after_hwframe+0x44/0xae
   RIP: 0033:0x46ae99
   Code: f7 d8 64 89 02 b8 ff ff ff ff c3 66 0f 1f 44 00 00 48 89 f8 48
   89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d
   01 f0 ff ff 73 01 c3 48 c7 c1 bc ff ff ff f7 d8 64 89 01 48
   RSP: 002b:00007f46711b9c48 EFLAGS: 00000246 ORIG_RAX: 0000000000000055
   RAX: ffffffffffffffda RBX: 000000000078c0a0 RCX: 000000000046ae99
   RDX: 0000000000000000 RSI: 00000000000000a1 RDI: 0000000020005800
   RBP: 00007f46711b9c80 R08: 0000000000000000 R09: 0000000000000000
   R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000017
   R13: 0000000000000000 R14: 000000000078c0a0 R15: 00007ffc129da6e0

   ================================================
   WARNING: lock held when returning to user space!
   5.15.0-rc1 #16 Not tainted
   ------------------------------------------------
   syz-executor/7579 is leaving the kernel with locks still held!
   1 lock held by syz-executor/7579:
    #0: ffff888104b73da8 (btrfs-tree-01/1){+.+.}-{3:3}, at:
   __btrfs_tree_lock+0x2e/0x1a0 fs/btrfs/locking.c:112

[CAUSE]
In __btrfs_cow_block() we could have a case where buf == *cow_ret, this
is the common call pattern in btrfs_search_slow().

In that case, before we return we should unlock the original buffer.

As in the btrfs_search_slot() call site:

			if (last_level)
				err = btrfs_cow_block(trans, root, b, NULL, 0,
						      &b,
						      BTRFS_NESTING_COW);
			else
				err = btrfs_cow_block(trans, root, b,
						      p->nodes[level + 1],
						      p->slots[level + 1], &b,
						      BTRFS_NESTING_COW);

btrfs_search_slot() expects btrfs_cow_block() to unlock the original
extent buffer @b.

As btrfs_search_slot() only puts the cowed tree block into path @p, thus
if btrfs_cow_block() fails, there will be no one to unlock extent buffer
@b.

[FIX]
Add unlock_orig check for all error paths in __btrfs_cow_block().
The patch is causing btrfs/010 to hang, it looks like there are some 
non-error path that we shouldn't unlock the original buf.

Will update the fix soon.

Thanks,
Qu
quoted hunk ↗ jump to hunk
Reported-by: Hao Sun <redacted>
Link: https://lore.kernel.org/linux-btrfs/CACkBjsZ9O6Zr0KK1yGn=1rQi6Crh1yeCRdTSBxx9R99L4xdn-Q@mail.gmail.com/ (local)
Signed-off-by: Qu Wenruo <redacted>
---
  fs/btrfs/ctree.c | 11 ++++++++++-
  1 file changed, 10 insertions(+), 1 deletion(-)
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 84627cbd5b5b..5cbbeb8384c7 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -415,8 +415,11 @@ static noinline int __btrfs_cow_block(struct btrfs_trans_handle *trans,
  	cow = btrfs_alloc_tree_block(trans, root, parent_start,
  				     root->root_key.objectid, &disk_key, level,
  				     search_start, empty_size, nest);
-	if (IS_ERR(cow))
+	if (IS_ERR(cow)) {
+		if (unlock_orig)
+			btrfs_tree_unlock(buf);
  		return PTR_ERR(cow);
+	}
  
  	/* cow is set to blocking by btrfs_init_new_buffer */
  
@@ -436,6 +439,8 @@ static noinline int __btrfs_cow_block(struct btrfs_trans_handle *trans,
  	ret = update_ref_for_cow(trans, root, buf, cow, &last_ref);
  	if (ret) {
  		btrfs_tree_unlock(cow);
+		if (unlock_orig)
+			btrfs_tree_unlock(buf);
  		free_extent_buffer(cow);
  		btrfs_abort_transaction(trans, ret);
  		return ret;
@@ -445,6 +450,8 @@ static noinline int __btrfs_cow_block(struct btrfs_trans_handle *trans,
  		ret = btrfs_reloc_cow_block(trans, root, buf, cow);
  		if (ret) {
  			btrfs_tree_unlock(cow);
+			if (unlock_orig)
+				btrfs_tree_unlock(buf);
  			free_extent_buffer(cow);
  			btrfs_abort_transaction(trans, ret);
  			return ret;
@@ -479,6 +486,8 @@ static noinline int __btrfs_cow_block(struct btrfs_trans_handle *trans,
  			ret = btrfs_tree_mod_log_free_eb(buf);
  			if (ret) {
  				btrfs_tree_unlock(cow);
+				if (unlock_orig)
+					btrfs_tree_unlock(buf);
  				free_extent_buffer(cow);
  				btrfs_abort_transaction(trans, ret);
  				return ret;
  
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help