Hi!
On Wed 17-03-21 14:02:43, Greg Kroah-Hartman wrote:
On Wed, Mar 17, 2021 at 01:40:55PM +0100, Salvatore Bonaccorso wrote:
On Wed, Mar 17, 2021 at 11:11:04AM +0100, Greg KH wrote:
On Wed, Mar 17, 2021 at 11:21:23AM +0530, Rohit Keshri wrote:
Hello Team,
A flaw was found in the Linux kernel. A denial of service problem is identified if an extent tree is corrupted in a crafted ext4 filesystem in fs/ext4/extents.c in ext4_es_cache_extent. Fabricating an integer overflow, A local attacker with a special user privilege may cause a system crash problem which can lead to an availability threat.
Please include what kernel version things like this were "found in" and when it was fixed, otherwise you force everyone to go scramble just to find that this was reported in July of 2020 and fixed then in the 5.9 kernel release and has already been backported to all relevant stable kernel releases in August of last year.
I absolutely second this. Please include in report if the problem is already fixed and which commit fixed it. Because I've just spent 20 minutes this morning searching my mailboxes because I remembered I've been fixing a problem that very much resembled this report but it took me a while to find our bugzilla... And it was me who actually fixed the bug a few months ago so I knew what to look for. When I have to deal with similar situation for problems I never heard of before it's even bigger waste of time and in some cases I'm not even sure I found the right problem.
In other words, no one running an updated kernel version from kernel.org is vulnerable today, right? Are you saying that specific distro kernels are vulnerable to this? If so, which ones?
It might be missing in some stable trees from a quick check. I just checked the SUSE bug and it lists the following three relevant commits, whilst the last one seems the relevant one:
d176b1f62f24 "ext4: handle error of ext4_setup_system_zone() on remount" bf9a379d0980 "ext4: don't allow overlapping system zones" ce9f24cccdc0 "ext4: check journal inode extents more carefully"
ce9f24cccdc0 was indeed included in 5.9-rc2, and backported to
v5.7.18: 3b654d118548ef2bb212dca361a5d1d19707822d ext4: check journal inode extents more carefully v5.8.4: cfa678021a1bb6b5ce4aa45c865f2d3167646f89 ext4: check journal inode extents more carefully v5.9-rc2: ce9f24cccdc019229b70a5c15e2b09ad9c0ab5d1 ext4: check journal inode extents more carefully
Then though it has
G>
Fixes: 0a944e8a6c66 ("ext4: don't perform block validity checks on the journal inode")
and 0a944e8a6c66 itself was backported to some of the stable series as well, I found:
v3.16.85: 71bfaf9e30125ec5b408fd328e412abf3b23214d ext4: don't perform block validity checks on the journal inode v4.14.178: fc3293a80acc469fbabc91bfbf2e65dc84377dc7 ext4: don't perform block validity checks on the journal inode v4.19.73: 97fbf573460e56ddf172614f70cdfa2af03b20ea ext4: don't perform block validity checks on the journal inode v4.4.221: 571fa68cacdf5fa70a6fdb71bda051f822d3cfb6 ext4: don't perform block validity checks on the journal inode v4.9.221: 2130aae807893cff163404bf6f6f6a4906dd14a1 ext4: don't perform block validity checks on the journal inode v5.2-rc2: 0a944e8a6c66ca04c7afbaa17e22bf208a8b37f0 ext4: don't perform block validity checks on the journal inode
So in the current still supported stable series, in 4.9.221, 4.14.178 and 4.19.73.
I just tried the reproducer from https://bugzilla.suse.com/show_bug.cgi?id=1173485 on a system with 4.19.177 which so has not yet the 0a944e8a6c66 ("ext4: don't perform block validity checks on the journal inode") fix backported and it indeed causes:
[ 224.003978] ------------[ cut here ]------------ [ 224.005052] kernel BUG at fs/ext4/extents_status.c:762! [ 224.006659] invalid opcode: 0000 [#1] SMP PTI [ 224.008378] CPU: 0 PID: 594 Comm: mount Not tainted 4.19.0-15-amd64 #1 Debian 4.19.177-1 [ 224.011292] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.12.0-1 04/01/2014 [ 224.014043] RIP: 0010:ext4_es_cache_extent+0xfe/0x100 [ext4] [ 224.015770] Code: 48 8b 45 00 48 8b 7d 08 48 83 c5 18 48 89 e2 4c 89 e6 e8 a5 15 d8 d2 48 8b 45 00 48 85 c0 75 e4 e9 54 ff ff ff e8 a2 65 3f d2 <0f> 0b 0f 1f 44 00 00 41 55 49 89 fd 41 54 55 48 89 d5 53 89 f3 0f [ 224.019834] RSP: 0018:ffffa7a840b8f958 EFLAGS: 00010213 [ 224.021034] RAX: 07ffffffffffffff RBX: 0000000000007ffd RCX: 0000ffffffffffff [ 224.022658] RDX: 0000000000007fff RSI: 00000000ffffffff RDI: ffff940d78a78968 [ 224.024283] RBP: 0000000000007fff R08: 1000ffffffffffff R09: 00000000000002c9 [ 224.025913] R10: 00000000000002c9 R11: 0000000000000041 R12: ffff940d78a78968 [ 224.027549] R13: 00000000ffffffff R14: ffffffffffffffff R15: 00000000ffffffff [ 224.029179] FS: 00007fe8d33fb100(0000) GS:ffff940d7ba00000(0000) knlGS:0000000000000000 [ 224.031011] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 224.032354] CR2: 000056074c755878 CR3: 0000000135184001 CR4: 00000000001606f0 [ 224.033990] Call Trace: [ 224.034634] ext4_cache_extents+0x63/0xd0 [ext4] [ 224.035712] __read_extent_tree_block+0x111/0x160 [ext4] [ 224.036665] ? __kmalloc+0x180/0x220 [ 224.037343] ? ext4_find_extent+0x144/0x320 [ext4] [ 224.038247] ext4_find_extent+0x144/0x320 [ext4] [ 224.039144] ext4_ext_map_blocks+0x6a/0xda0 [ext4] [ 224.040063] ? schedule+0x28/0x80 [ 224.040747] ? __wait_on_bit+0x58/0x90 [ 224.041476] ext4_map_blocks+0x2e4/0x5f0 [ext4] [ 224.042350] ? ext4_data_block_valid+0x1d/0x20 [ext4] [ 224.043313] ? __ext4_ext_check+0x238/0x3a0 [ext4] [ 224.044261] _ext4_get_block+0x8e/0x110 [ext4] [ 224.045158] ? unlock_new_inode+0x4e/0x60 [ 224.045960] generic_block_bmap+0x4b/0x70 [ 224.046765] jbd2_journal_init_inode+0x11/0xb0 [jbd2] [ 224.047794] ext4_fill_super+0x3052/0x3c70 [ext4] [ 224.048730] ? bdev_name.isra.6+0x2a/0xa0 [ 224.049530] ? ext4_calculate_overhead+0x490/0x490 [ext4] [ 224.050555] ? snprintf+0x49/0x60 [ 224.051234] ? ext4_calculate_overhead+0x490/0x490 [ext4] [ 224.052309] ? mount_bdev+0x177/0x1b0 [ 224.053074] ? ext4_calculate_overhead+0x490/0x490 [ext4] [ 224.054146] mount_bdev+0x177/0x1b0 [ 224.054858] mount_fs+0x3e/0x150 [ 224.055533] vfs_kern_mount.part.36+0x54/0x120 [ 224.056394] do_mount+0x20e/0xcc0 [ 224.057022] ? _copy_from_user+0x37/0x60 [ 224.057781] ? memdup_user+0x4b/0x70 [ 224.058530] ksys_mount+0xb6/0xd0 [ 224.059231] __x64_sys_mount+0x21/0x30 [ 224.059949] do_syscall_64+0x53/0x110 [ 224.060631] entry_SYSCALL_64_after_hwframe+0x44/0xa9 [ 224.061531] RIP: 0033:0x7fe8d35f9fea [ 224.062255] Code: 48 8b 0d a9 0e 0c 00 f7 d8 64 89 01 48 83 c8 ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 49 89 ca b8 a5 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 76 0e 0c 00 f7 d8 64 89 01 48 [ 224.065741] RSP: 002b:00007ffcd2022e38 EFLAGS: 00000246 ORIG_RAX: 00000000000000a5 [ 224.067223] RAX: ffffffffffffffda RBX: 000056074c748fb0 RCX: 00007fe8d35f9fea [ 224.068620] RDX: 000056074c751050 RSI: 000056074c7491e0 RDI: 000056074c7491c0 [ 224.069987] RBP: 00007fe8d39471c4 R08: 0000000000000000 R09: 0000000000000000 [ 224.071399] R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000 [ 224.072764] R13: 0000000000000000 R14: 000056074c7491c0 R15: 000056074c751050 [ 224.074101] Modules linked in: loop sctp binfmt_misc crct10dif_pclmul crc32_pclmul ghash_clmulni_intel button virtio_console virtio_balloon evdev qemu_fw_cfg joydev pcspkr serio_raw nfsd auth_rpcgss nfs_acl lockd grace sunrpc ip_tables x_tables autofs4 ext4 crc16 mbcache jbd2 fscrypto ecb hid_generic usbhid hid btrfs xor zstd_decompress zstd_compress xxhash raid6_pq libcrc32c crc32c_generic dm_mod ata_generic virtio_net net_failover failover virtio_blk crc32c_intel ata_piix libata uhci_hcd ehci_pci ehci_hcd scsi_mod floppy aesni_intel usbcore psmouse aes_x86_64 crypto_simd cryptd glue_helper i2c_piix4 virtio_pci virtio_ring virtio usb_common [ 224.084374] ---[ end trace a93d957244af62ea ]--- [ 224.085298] RIP: 0010:ext4_es_cache_extent+0xfe/0x100 [ext4] [ 224.086402] Code: 48 8b 45 00 48 8b 7d 08 48 83 c5 18 48 89 e2 4c 89 e6 e8 a5 15 d8 d2 48 8b 45 00 48 85 c0 75 e4 e9 54 ff ff ff e8 a2 65 3f d2 <0f> 0b 0f 1f 44 00 00 41 55 49 89 fd 41 54 55 48 89 d5 53 89 f3 0f [ 224.089834] RSP: 0018:ffffa7a840b8f958 EFLAGS: 00010213 [ 224.090832] RAX: 07ffffffffffffff RBX: 0000000000007ffd RCX: 0000ffffffffffff [ 224.092181] RDX: 0000000000007fff RSI: 00000000ffffffff RDI: ffff940d78a78968 [ 224.093539] RBP: 0000000000007fff R08: 1000ffffffffffff R09: 00000000000002c9 [ 224.094899] R10: 00000000000002c9 R11: 0000000000000041 R12: ffff940d78a78968 [ 224.096264] R13: 00000000ffffffff R14: ffffffffffffffff R15: 00000000ffffffff [ 224.097586] FS: 00007fe8d33fb100(0000) GS:ffff940d7ba00000(0000) knlGS:0000000000000000 [ 224.099144] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 224.100284] CR2: 000056074c755878 CR3: 0000000135184001 CR4: 00000000001606f0
So likely the 4.9.y, 4.14.y and 4.19.y still have the bug? Is this correct? I only quickly skimmed over the report, but this seems to be the case.
It's hard to tell if those older kernels have this issue as the fix for this does not apply at all, and even SUSE didn't backport the change as it didn't seem relevant to them.
Yes, they do have the problem. In fact I'm just testing a backport to 4.4 kernel (which is as far as I can be bothered to backport this). We (as in SUSE) actually did backport the fixes to all "proactively maintained" branches at the time fixes were merged upstream. Which also included SLE15-SP1 based on 4.12 kernel.
But I'll gladly take backports if someone wants to provide them :)
I'll check whether the backports I have do apply to -stable branches and forward them to you.
Do distros consider "mounting untrusted ext4 filesystem images" a valid thing to worry about? If so, that is against what the ext4 developers have said in the past from what I recall, so that might be good to get straightened out, and maybe why SUSE didn't assign a CVE for this...
This is kind of a grey area. On one hand ext4 developers make it clear that mounting untrusted images is not safe. That's also why we (as in SUSE) didn't bother with CVE for the bug when it was originally found. On the other hand the reality is that especially for desktop / laptop users this can easily happen (most users have automount of plugged-in storage enabled in their desktop environment). So if ext4 developers learn about a problem, they try to fix it and in SUSE we tend to backport those fixes to all "community" distros that are still actively maintained. For server distros it's a judgement call as there mounting untrusted fs images is much less probable.
Honza