cbaee87f2e
At btrfs_orphan_cleanup(), if we can't find an inode (btrfs_iget() returns
an -ENOENT error pointer), we proceed with 'ret' set to -ENOENT and the
inode pointer set to ERR_PTR(-ENOENT). Later when we proceed to the body
of the following if statement:
if (ret == -ENOENT || inode->i_nlink) {
(...)
trans = btrfs_start_transaction(root, 1);
if (IS_ERR(trans)) {
ret = PTR_ERR(trans);
iput(inode);
goto out;
}
(...)
ret = btrfs_del_orphan_item(trans, root,
found_key.objectid);
btrfs_end_transaction(trans);
if (ret) {
iput(inode);
goto out;
}
continue;
}
If we get an error from btrfs_start_transaction() or from the call to
btrfs_del_orphan_item() we end calling iput() against an inode pointer
that has a value of ERR_PTR(-ENOENT), resulting in a crash with the
following trace:
[876.667] BUG: kernel NULL pointer dereference, address: 0000000000000096
[876.667] #PF: supervisor read access in kernel mode
[876.667] #PF: error_code(0x0000) - not-present page
[876.667] PGD 0 P4D 0
[876.668] Oops: 0000 [#1] PREEMPT SMP PTI
[876.668] CPU: 0 PID: 2356187 Comm: mount Tainted: G W 6.4.0-rc6-btrfs-next-134+ #1
[876.668] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.16.2-0-gea1b7a073390-prebuilt.qemu.org 04/01/2014
[876.668] RIP: 0010:iput+0xa/0x20
[876.668] Code: ff ff ff 66 (...)
[876.669] RSP: 0018:ffffafa9c0c9f9d0 EFLAGS: 00010282
[876.669] RAX: ffffffffffffffe4 RBX: 000000000009453b RCX: 0000000000000000
[876.669] RDX: 0000000000000001 RSI: ffffafa9c0c9f930 RDI: fffffffffffffffe
[876.669] RBP: ffff95c612f3b800 R08: 0000000000000001 R09: ffffffffffffffe4
[876.670] R10: 00018f2a71010000 R11: 000000000ead96e3 R12: ffff95cb7d6909a0
[876.670] R13: fffffffffffffffe R14: ffff95c60f477000 R15: 00000000ffffffe4
[876.670] FS: 00007f5fbe30a840(0000) GS:ffff95ccdfa00000(0000) knlGS:0000000000000000
[876.670] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[876.671] CR2: 0000000000000096 CR3: 000000055e9f6004 CR4: 0000000000370ef0
[876.671] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[876.671] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[876.672] Call Trace:
[876.744] <TASK>
[876.744] ? __die_body+0x1b/0x60
[876.744] ? page_fault_oops+0x15d/0x450
[876.745] ? __kmem_cache_alloc_node+0x47/0x410
[876.745] ? do_user_addr_fault+0x65/0x8a0
[876.745] ? exc_page_fault+0x74/0x170
[876.746] ? asm_exc_page_fault+0x22/0x30
[876.746] ? iput+0xa/0x20
[876.746] btrfs_orphan_cleanup+0x221/0x330 [btrfs]
[876.746] btrfs_lookup_dentry+0x58f/0x5f0 [btrfs]
[876.747] btrfs_lookup+0xe/0x30 [btrfs]
[876.747] __lookup_slow+0x82/0x130
[876.785] walk_component+0xe5/0x160
[876.786] path_lookupat.isra.0+0x6e/0x150
[876.786] filename_lookup+0xcf/0x1a0
[876.786] ? mod_objcg_state+0xd2/0x360
[876.786] ? obj_cgroup_charge+0xf5/0x110
[876.787] ? should_failslab+0xa/0x20
[876.787] ? kmem_cache_alloc+0x47/0x450
[876.787] vfs_path_lookup+0x51/0x90
[876.788] mount_subtree+0x8d/0x130
[876.788] btrfs_mount+0x149/0x410 [btrfs]
[876.788] ? __kmem_cache_alloc_node+0x47/0x410
[876.788] ? vfs_parse_fs_param+0xc0/0x110
[876.789] legacy_get_tree+0x24/0x50
[876.834] vfs_get_tree+0x22/0xd0
[876.852] path_mount+0x2d8/0x9c0
[876.852] do_mount+0x79/0x90
[876.852] __x64_sys_mount+0x8e/0xd0
[876.853] do_syscall_64+0x38/0x90
[876.899] entry_SYSCALL_64_after_hwframe+0x72/0xdc
[876.958] RIP: 0033:0x7f5fbe50b76a
[876.959] Code: 48 8b 0d a9 (...)
[876.959] RSP: 002b:00007fff01925798 EFLAGS: 00000246 ORIG_RAX: 00000000000000a5
[876.959] RAX: ffffffffffffffda RBX: 00007f5fbe694264 RCX: 00007f5fbe50b76a
[876.960] RDX: 0000561bde6c8720 RSI: 0000561bde6bdec0 RDI: 0000561bde6c31a0
[876.960] RBP: 0000561bde6bdc70 R08: 0000000000000000 R09: 0000000000000001
[876.960] R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
[876.960] R13: 0000561bde6c31a0 R14: 0000561bde6c8720 R15: 0000561bde6bdc70
[876.960] </TASK>
So fix this by setting 'inode' to NULL whenever we get an error from
btrfs_iget(), and to make the code simpler, stop testing for 'ret' being
-ENOENT to check if we have an inode - instead test for 'inode' being NULL
or not. Having a NULL 'inode' prevents any iput() call from crashing, as
iput() ignores NULL inode pointers. Also, stop testing for a NULL return
value from btrfs_iget() with PTR_ERR_OR_ZERO(), because btrfs_iget() never
returns NULL - in case an inode is not found, it returns ERR_PTR(-ENOENT),
and in case of memory allocation failure, it returns ERR_PTR(-ENOMEM).
We also don't need the extra iput() calls on the error branches for the
btrfs_start_transaction() and btrfs_del_orphan_item() calls, as we have
already called iput() before, so remove them.
Fixes:
|
||
---|---|---|
.. | ||
9p | ||
adfs | ||
affs | ||
afs | ||
autofs | ||
befs | ||
bfs | ||
btrfs | ||
cachefiles | ||
ceph | ||
coda | ||
configfs | ||
cramfs | ||
crypto | ||
debugfs | ||
devpts | ||
dlm | ||
ecryptfs | ||
efivarfs | ||
efs | ||
erofs | ||
exfat | ||
exportfs | ||
ext2 | ||
ext4 | ||
f2fs | ||
fat | ||
freevxfs | ||
fscache | ||
fuse | ||
gfs2 | ||
hfs | ||
hfsplus | ||
hostfs | ||
hpfs | ||
hugetlbfs | ||
iomap | ||
isofs | ||
jbd2 | ||
jffs2 | ||
jfs | ||
kernfs | ||
lockd | ||
minix | ||
netfs | ||
nfs | ||
nfs_common | ||
nfsd | ||
nilfs2 | ||
nls | ||
notify | ||
ntfs | ||
ntfs3 | ||
ocfs2 | ||
omfs | ||
openpromfs | ||
orangefs | ||
overlayfs | ||
proc | ||
pstore | ||
qnx4 | ||
qnx6 | ||
quota | ||
ramfs | ||
reiserfs | ||
romfs | ||
smb | ||
squashfs | ||
sysfs | ||
sysv | ||
tracefs | ||
ubifs | ||
udf | ||
ufs | ||
unicode | ||
vboxsf | ||
verity | ||
xfs | ||
zonefs | ||
aio.c | ||
anon_inodes.c | ||
attr.c | ||
bad_inode.c | ||
binfmt_elf_fdpic.c | ||
binfmt_elf_test.c | ||
binfmt_elf.c | ||
binfmt_flat.c | ||
binfmt_misc.c | ||
binfmt_script.c | ||
buffer.c | ||
char_dev.c | ||
compat_binfmt_elf.c | ||
coredump.c | ||
d_path.c | ||
dax.c | ||
dcache.c | ||
direct-io.c | ||
drop_caches.c | ||
eventfd.c | ||
eventpoll.c | ||
exec.c | ||
fcntl.c | ||
fhandle.c | ||
file_table.c | ||
file.c | ||
filesystems.c | ||
fs_context.c | ||
fs_parser.c | ||
fs_pin.c | ||
fs_struct.c | ||
fs_types.c | ||
fs-writeback.c | ||
fsopen.c | ||
init.c | ||
inode.c | ||
internal.h | ||
ioctl.c | ||
Kconfig | ||
Kconfig.binfmt | ||
kernel_read_file.c | ||
libfs.c | ||
locks.c | ||
Makefile | ||
mbcache.c | ||
mnt_idmapping.c | ||
mount.h | ||
mpage.c | ||
namei.c | ||
namespace.c | ||
no-block.c | ||
nsfs.c | ||
open.c | ||
pipe.c | ||
pnode.c | ||
pnode.h | ||
posix_acl.c | ||
proc_namespace.c | ||
read_write.c | ||
readdir.c | ||
remap_range.c | ||
select.c | ||
seq_file.c | ||
signalfd.c | ||
splice.c | ||
stack.c | ||
stat.c | ||
statfs.c | ||
super.c | ||
sync.c | ||
sysctls.c | ||
timerfd.c | ||
userfaultfd.c | ||
utimes.c | ||
xattr.c |