On Fri 25-04-25 15:29:41, Liebes Wang wrote: > Dear Linux maintainers and reviewers: > We are reporting a Linux kernel bug titled **kernel BUG in > zero_user_segments**, discovered using a modified version of Syzkaller. > > This bug seems to be duplicated as > https://syzkaller.appspot.com/bug?extid=78eeb671facb19832e95, but the test > case is much smaller, which may be helpful for analyzing the bug. > > Linux version: 9d7a0577c9db35c4cc52db90bc415ea248446472 > > The bisection log shows the first introduced commit is > 982bf37da09d078570650b691d9084f43805a5de > commit 982bf37da09d078570650b691d9084f43805a5de > Author: Zhang Yi <yi.zhang@xxxxxxxxxx> > Date: Fri Dec 20 09:16:31 2024 +0800 > > ext4: refactor ext4_punch_hole() > > The current implementation of ext4_punch_hole() contains complex > position calculations and stale error tags. To improve the code's > clarity and maintainability, it is essential to clean up the code and > improve its readability, this can be achieved by: a) simplifying and > renaming variables; b) eliminating unnecessary position calculations; > c) writing back all data in data=journal mode, and drop page cache from > the original offset to the end, rather than using aligned blocks, > d) renaming the stale error tags. > > Signed-off-by: Zhang Yi <yi.zhang@xxxxxxxxxx> > Reviewed-by: Jan Kara <jack@xxxxxxx> > Reviewed-by: Ojaswin Mujoo <ojaswin@xxxxxxxxxxxxx> > Link: > https://patch.msgid.link/20241220011637.1157197-5-yi.zhang@xxxxxxxxxxxxxxx > Signed-off-by: Theodore Ts'o <tytso@xxxxxxx> So there's something suspicious about this report. The stacktrace shows we've crashed in punch hole code (call from ioctl_preallocate()) but the reproducer actually never calls this. Anyway, the reported stack trace ends with truncate_inode_partial_folio() -> folio_zero_range() -> zero_user_segments(). The assertion that's failing is: BUG_ON(end1 > page_size(page) || end2 > page_size(page)); Now it seems that this assertion can indeed easily trigger when we have a large folio because truncate_inode_partial_folio() is called to zero out tail of the whole folio which can certainly be more than a page. Matthew, am I missing something (I guess I am because otherwise I'd expect we'd be crashing left and right) or is the folio conversion on this path indeed broken? Honza > > The test case, kernel config and full bisection log are attached. > > The report is (The full report is attached): > EXT4-fs (loop7): mounted filesystem 00000000-0000-0000-0000-000000000000 > r/w without journal. Quota mode: writeback. > EXT4-fs warning (device loop7): ext4_block_to_path:105: block 2147483648 > > max in inode 15 > ------------[ cut here ]------------ > kernel BUG at ./include/linux/highmem.h:275! > Oops: invalid opcode: 0000 [#1] SMP KASAN NOPTI > CPU: 0 UID: 0 PID: 6795 Comm: syz.7.479 Not tainted > 6.15.0-rc3-g9d7a0577c9db #1 PREEMPT(voluntary) > Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS > 1.13.0-1ubuntu1.1 04/01/2014 > RIP: 0010:zero_user_segments.constprop.0+0x10c/0x290 > include/linux/highmem.h:275 > Code: 0f b6 4b 40 ba 00 10 00 00 48 d3 e2 49 89 d7 e8 ba d5 e2 ff 4c 89 fe > 4c 89 ef e8 3f d0 e2 ff 4d 39 fd 76 08 e8 a5 d5 e2 ff 90 <0f> 0b e8 9d d5 > e2 ff be 08 00 00 00 48 89 df e8 a0 9c 1d 00 48 89 > RSP: 0018:ffff8881235ff678 EFLAGS: 00010216 > RAX: 000000000000025d RBX: ffffea00056071c0 RCX: ffffc90002e0b000 > RDX: 0000000000080000 RSI: ffffffff818f7b0b RDI: 0000000000000006 > RBP: 000000000040b000 R08: 0000000000000000 R09: fffff94000ac0e38 > R10: 0000000000001000 R11: 0000000000000000 R12: 0000000000000005 > R13: 000000000040b000 R14: 0000000000000000 R15: 0000000000001000 > FS: 00007fecef19d700(0000) GS:ffff888543948000(0000) knlGS:0000000000000000 > CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 > CR2: 00007f5e38b40008 CR3: 000000013ebaa001 CR4: 0000000000770ef0 > DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 > DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000600 > PKRU: 80000000 > Call Trace: > <TASK> > folio_zero_range include/linux/highmem.h:647 [inline] > truncate_inode_partial_folio+0x6da/0xbd0 mm/truncate.c:219 > truncate_inode_pages_range+0x3fc/0xcc0 mm/truncate.c:387 > ext4_truncate_page_cache_block_range+0xb3/0x5c0 fs/ext4/inode.c:3974 > ext4_punch_hole+0x2cd/0xec0 fs/ext4/inode.c:4049 > ext4_fallocate+0x128d/0x32c0 fs/ext4/extents.c:4766 > vfs_fallocate+0x3ed/0xd70 fs/open.c:338 > ioctl_preallocate+0x190/0x200 fs/ioctl.c:290 > file_ioctl fs/ioctl.c:333 [inline] > do_vfs_ioctl+0x149c/0x1850 fs/ioctl.c:885 > __do_sys_ioctl fs/ioctl.c:904 [inline] > __se_sys_ioctl fs/ioctl.c:892 [inline] > __x64_sys_ioctl+0x11f/0x200 fs/ioctl.c:892 > do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline] > do_syscall_64+0xc1/0x1d0 arch/x86/entry/syscall_64.c:94 > entry_SYSCALL_64_after_hwframe+0x77/0x7f > syz_mount_image$ext4(&(0x7f0000000400)='ext4\x00', &(0x7f00000001c0)='./file0\x00', 0x0, &(0x7f0000000280)={[{@journal_ioprio}, {@mb_optimize_scan}, {@data_err_ignore}, {@grpquota}, {@barrier}]}, 0x1, 0x3cb, &(0x7f00000026c0)="$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") > quotactl_fd$Q_SETINFO(0xffffffffffffffff, 0x2, 0x0, &(0x7f0000000080)={0x80000000000002, 0x80000000005, 0x1, 0x6}) > r0 = openat(0xffffffffffffff9c, &(0x7f0000000040)='./file1\x00', 0x42, 0x1ff) > ioctl$EXT4_IOC_CHECKPOINT(r0, 0x40305829, &(0x7f0000000080)=0x5) > r1 = openat(0xffffffffffffff9c, &(0x7f0000000040)='./file1\x00', 0x42, 0x1ff) > ioctl$EXT4_IOC_CHECKPOINT(r1, 0x40305829, &(0x7f0000000080)=0x5) -- Jan Kara <jack@xxxxxxxx> SUSE Labs, CR
