[Kernel 6.9.3 BUG] general protection fault in jfs_ioc_trim (bad page state)
From: Zhang Zhiyu
Date: Mon Jun 10 2024 - 13:45:04 EST
Hi upstream maintainers and community,
We found a Linux kernel bug with our modified Syzkaller.
===================== Meta info =====================
Found version: 6.8
Affected files: fs/jfs/ioctl.c:131, fs/jfs/super.c:194
Reproduced version: 6.8 ~ 6.9.3 ~ 6.10-rc3 (with or without sanitizers)
Kernel config (e.g. for 6.9.3 with sanitizers disabled):
https://drive.google.com/file/d/1pJS6B2cMVX1XYIBUgSwpcG-vsb8MCQ-g/view?usp=sharing
Syz reproducer:
https://drive.google.com/file/d/1nz0jTx3-9FYGJ9ePyNmp46IuO-10WGZQ/view?usp=sharing
C reproducer: https://drive.google.com/file/d/1vqXhsSPDQX1l5DkaJYI92dEYwCnbDCqs/view?usp=sharing
Reproduction console output:
https://drive.google.com/file/d/1daq8GayqgVavEdofVLQQhyeLdU-jPm6H/view?usp=sharing
================== How to reproduce ==================
1. Create a image with syzkaller create-image.sh and boot it (or using
an existing one is ok)
2. Compile the repro.c with gcc and scp to qemu vm
3. Execute the repro and observe the bug backtrace looping.
======================= Report ======================
[copy from console output]
[?2004hroot@syzkaller:~# ./bd08f46
[?2004l
[ 282.651266][ T30] audit: type=1400 audit(1718036000.960:9): avc:
denied { execmem } for pid=7748 comm="bd08f46"
scontext=system_u:system_r:kernel_t:s0
tcontext=system_u:system_r:kernel_t:s0 tclass=process permissive=1
[ 282.836121][ T7749] loop0: detected capacity change from 0 to 32768
[ 282.927686][ T7748] BUG: Bad page state in process bd08f46 pfn:5e2ab
[ 282.928954][ T7748] page: refcount:0 mapcount:0
mapping:0000000000000000 index:0x3 pfn:0x5e2ab
[ 282.930357][ T7748] flags:
0x4fff1800000820c(referenced|uptodate|workingset|private|node=1|zone=1|lastcpupid=0x7ff)
[ 282.935368][ T7748] page_type: 0xffffffff()
[ 282.936424][ T7748] raw: 04fff1800000820c dead000000000100
dead000000000122 0000000000000000
[ 282.938473][ T7748] raw: 0000000000000003 ffff8880614fd558
00000000ffffffff 0000000000000000
[ 282.940469][ T7748] page dumped because: PAGE_FLAGS_CHECK_AT_FREE flag(s) set
[ 282.941989][ T7748] page_owner tracks the page as allocated
[ 282.943817][ T7748] page last allocated via order 0, migratetype
Unmovable, gfp_mask 0x40c40(GFP_NOFS|__GFP_COMP), pid 7749, tgid 7749
(bd08f46), ts 282878748769, free_ts 282874052762
[ 282.946592][ T7748] post_alloc_hook+0x15c/0x180
[ 282.947449][ T7748] get_page_from_freelist+0x19d2/0x1b40
[ 282.948414][ T7748] __alloc_pages+0x117/0x2d0
[ 282.949224][ T7748] alloc_pages_mpol+0x266/0x380
[ 282.950079][ T7748] folio_alloc+0xd7/0x100
[ 282.950832][ T7748] do_read_cache_folio+0x8e/0x550
[ 282.951711][ T7748] do_read_cache_page+0x31/0x150
[ 282.952644][ T7748] __get_metapage+0x16f/0x8f0
[ 282.953472][ T7748] dbAdjCtl+0xa2/0x400
[ 282.954196][ T7748] dbAllocDmapLev+0x1f2/0x2d0
[ 282.955018][ T7748] dbAllocCtl+0x8c/0x4e0
[ 282.955754][ T7748] dbAllocAG+0xe5/0x590
[ 282.956480][ T7748] dbDiscardAG+0x1e4/0x500
[ 282.957257][ T7748] jfs_ioc_trim+0x161/0x1b0
[ 282.958055][ T7748] jfs_ioctl+0x13a/0x1c0
[ 282.958776][ T7748] __se_sys_ioctl+0xc0/0x130
[ 282.959591][ T7748] page last free pid 7749 tgid 7749 stack trace:
[ 282.960700][ T7748] free_unref_page_prepare+0x433/0x4d0
[ 282.961651][ T7748] free_unref_page+0x35/0x250
[ 282.965020][ T7748] lmLogShutdown+0x1a2/0x2c0
[ 282.965827][ T7748] lmLogClose+0xf9/0x200
[ 282.966574][ T7748] jfs_remount+0x1e6/0x2b0
[ 282.967358][ T7748] reconfigure_super+0x1da/0x3e0
[ 282.968225][ T7748] __se_sys_fsconfig+0x6a6/0x860
[ 282.969094][ T7748] do_syscall_64+0xf8/0x240
[ 282.969897][ T7748] entry_SYSCALL_64_after_hwframe+0x77/0x7f
[ 282.971238][ T7748] Modules linked in:
[ 282.971914][ T7748] CPU: 1 PID: 7748 Comm: bd08f46 Not tainted 6.9.3 #1
[ 282.973065][ T7748] Hardware name: QEMU Standard PC (i440FX + PIIX,
1996), BIOS 1.13.0-1ubuntu1.1 04/01/2014
[ 282.974753][ T7748] Call Trace:
[ 282.975319][ T7748] <TASK>
[ 282.975821][ T7748] dump_stack_lvl+0x12e/0x1e0
[ 282.976645][ T7748] bad_page+0xde/0xf0
[ 282.977339][ T7748] free_unref_page_prepare+0x4a9/0x4d0
[ 282.978283][ T7748] free_unref_folios+0x116/0x580
[ 282.979127][ T7748] folios_put_refs+0x417/0x480
[ 282.979970][ T7748] truncate_inode_pages_range+0x1ed/0x9d0
[ 282.980977][ T7748] ? debug_object_active_state+0x53/0x150
[ 282.981988][ T7748] dbUnmount+0x60/0x90
[ 282.982710][ T7748] jfs_umount+0xe3/0x1a0
[ 282.983454][ T7748] jfs_put_super+0x63/0xf0
[ 282.984226][ T7748] ? __pfx_jfs_put_super+0x10/0x10
[ 282.985124][ T7748] generic_shutdown_super+0x9b/0x180
[ 282.986024][ T7748] kill_block_super+0x1e/0x50
[ 282.986850][ T7748] deactivate_locked_super+0x5f/0xd0
[ 282.987764][ T7748] cleanup_mnt+0x1a0/0x210
[ 282.988541][ T7748] task_work_run+0x106/0x150
[ 282.989314][ T7748] syscall_exit_to_user_mode+0x170/0x370
[ 282.990290][ T7748] do_syscall_64+0x105/0x240
[ 282.991096][ T7748] entry_SYSCALL_64_after_hwframe+0x77/0x7f
[ 282.992106][ T7748] RIP: 0033:0x7fdf212d8aa7
[ 282.992875][ T7748] Code: 23 0d 00 f7 d8 64 89 01 48 83 c8 ff c3 66
0f 1f 44 00 00 31 f6 e9 09 00 00 00 66 0f 1f 84 00 00 00 00 00 b8 a6
00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d b9 23 0d 00 f7 d8
64 89 01 48
[ 282.996131][ T7748] RSP: 002b:00007ffecb179c38 EFLAGS: 00000202
ORIG_RAX: 00000000000000a6
[ 282.997553][ T7748] RAX: 0000000000000000 RBX: 0000000000000000
RCX: 00007fdf212d8aa7
[ 282.998903][ T7748] RDX: 0000000000000009 RSI: 0000000000000009
RDI: 00007ffecb179ce0
[ 283.000244][ T7748] RBP: 00007ffecb17ad20 R08: 00007fdf2136c040
R09: 00007ffecb179ad0
[ 283.001579][ T7748] R10: 00007fdf2136bfc0 R11: 0000000000000202
R12: 00000000004012c0
[ 283.002925][ T7748] R13: 0000000000000000 R14: 0000000000000000
R15: 0000000000000000
[ 283.004275][ T7748] </TASK>
====================== Analysis ======================
gdb debug screenshots:
1. gdb backtrace:
https://drive.google.com/file/d/14KDOYjUtVxe-ugYPc_MTmlCGHydOQav1/view?usp=sharing
2. gdb print page->flags:
https://drive.google.com/file/d/1tMiAkb1DjJ-YkaEyPkMWAVnt1AgrMx58/view?usp=sharing
Possible root cause: A page was marked as allocated by page_owner
during page releasing in jfs_umount by mistake, which leads to the
badpage check through free_page_prepare.
Unfortunately, we did not find more information about the root cause.
Hope the attached files would be of help for addressing this bug. And
please let me know for anything I can help.
Best,
Zhiyu Zhang