On Wed, Jun 07, 2023 at 08:45:32AM +1000, Dave Chinner wrote: > On Tue, Jun 06, 2023 at 04:24:05PM +0200, David Sterba wrote: > > On Thu, Jun 01, 2023 at 06:15:06PM -0700, syzbot wrote: > > > RIP: 0010:rep_movs_alternative+0x33/0xb0 arch/x86/lib/copy_user_64.S:56 > > > Code: 46 83 f9 08 73 21 85 c9 74 0f 8a 06 88 07 48 ff c7 48 ff c6 48 ff c9 75 f1 c3 66 66 2e 0f 1f 84 00 00 00 00 00 66 90 48 8b 06 <48> 89 07 48 83 c6 08 48 83 c7 08 83 e9 08 74 df 83 f9 08 73 e8 eb > > > RSP: 0018:ffffc9000becf728 EFLAGS: 00050206 > > > RAX: 0000000000000000 RBX: 0000000000000038 RCX: 0000000000000038 > > > RDX: fffff520017d9efb RSI: ffffc9000becf7a0 RDI: 0000000020000120 > > > RBP: 0000000020000120 R08: 0000000000000000 R09: fffff520017d9efa > > > R10: ffffc9000becf7d7 R11: 0000000000000001 R12: ffffc9000becf7a0 > > > R13: 0000000020000158 R14: 0000000000000000 R15: ffffc9000becf7a0 > > > copy_user_generic arch/x86/include/asm/uaccess_64.h:112 [inline] > > > raw_copy_to_user arch/x86/include/asm/uaccess_64.h:133 [inline] > > > _copy_to_user lib/usercopy.c:41 [inline] > > > _copy_to_user+0xab/0xc0 lib/usercopy.c:34 > > > copy_to_user include/linux/uaccess.h:191 [inline] > > > fiemap_fill_next_extent+0x217/0x370 fs/ioctl.c:144 > > > emit_fiemap_extent+0x18e/0x380 fs/btrfs/extent_io.c:2616 > > > fiemap_process_hole+0x516/0x610 fs/btrfs/extent_io.c:2874 > > > > and extent enumeration from FIEMAP, this would qualify as a stress on > > the inode > > FWIW, when I've seen this sort of hang on XFS in past times, it's > been caused by a corrupt extent list or a circular reference in a > btree that the fuzzing introduced. Hence FIEMAP just keeps going > around in circles and never gets out of the loop to drop the inode > lock.... Thanks for the info. The provided reproducer was able to get the VM stuck in a few hours so there is some problem. The generated image does not show any obvious problem so it's either lack of 'check' capability or the problem happens at run time.
