[syzbot] general protection fault in xas_create_range

[syzbot]

Hello,

syzbot found the following issue on:

HEAD commit: 56e337f2cf13 Revert "gpio: Revert regression in sysfs-gpio..
git tree: upstream
console output: https://syzkaller.appspot.com/x/log.txt?x=146069d9700000
kernel config: https://syzkaller.appspot.com/x/.config?x=70f75a89c7a0e6bc
dashboard link: https://syzkaller.appspot.com/bug?extid=0d2b0bf32ca5cfd09f2e
compiler: gcc (Debian 10.2.1-6) 10.2.1 20210110, GNU ld (GNU Binutils for Debian) 2.35.2

Unfortunately, I don't have any reproducer for this issue yet.

IMPORTANT: if you fix the issue, please add the following tag to the commit:
Reported-by: syzbot+0d2b0b...@syzkaller.appspotmail.com

general protection fault, probably for non-canonical address 0xdffffc0000000001: 0000 [#1] PREEMPT SMP KASAN
KASAN: null-ptr-deref in range [0x0000000000000008-0x000000000000000f]
CPU: 0 PID: 32 Comm: khugepaged Not tainted 5.17.0-rc8-syzkaller-00003-g56e337f2cf13 #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
RIP: 0010:xa_parent_locked include/linux/xarray.h:1207 [inline]
RIP: 0010:xas_create_range+0x2d9/0x6e0 lib/xarray.c:725
Code: 89 ee 49 83 ed 40 49 c1 ee 03 4c 89 6d 08 4c 8d 6d 12 4d 01 e6 e9 93 00 00 00 e8 d2 0f 72 fd 48 8d 7b 08 48 89 f8 48 c1 e8 03 <42> 80 3c 20 00 0f 85 fb 02 00 00 48 8b 04 24 4c 8b 7b 08 80 38 00
RSP: 0018:ffffc90001097a48 EFLAGS: 00010002
RAX: 0000000000000001 RBX: 0000000000000000 RCX: 0000000000000000
RDX: ffff8880139de240 RSI: ffffffff8405e5be RDI: 0000000000000008
RBP: ffffc90001097bb0 R08: 0000000000000000 R09: 0000000000000000
R10: ffffffff8405e679 R11: 0000000000000000 R12: dffffc0000000000
R13: ffffc90001097bc2 R14: fffff52000212f76 R15: 0000000000000001
FS: 0000000000000000(0000) GS:ffff8880b9c00000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007ffe25571c98 CR3: 0000000079206000 CR4: 00000000003506f0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
Call Trace:
<TASK>
collapse_file+0x2d3/0x3c40 mm/khugepaged.c:1678
khugepaged_scan_file mm/khugepaged.c:2073 [inline]
khugepaged_scan_mm_slot mm/khugepaged.c:2168 [inline]
khugepaged_do_scan mm/khugepaged.c:2252 [inline]
khugepaged+0x53c7/0x6c00 mm/khugepaged.c:2297
kthread+0x2e9/0x3a0 kernel/kthread.c:377
ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:295
</TASK>
Modules linked in:
---[ end trace 0000000000000000 ]---
RIP: 0010:xa_parent_locked include/linux/xarray.h:1207 [inline]
RIP: 0010:xas_create_range+0x2d9/0x6e0 lib/xarray.c:725
Code: 89 ee 49 83 ed 40 49 c1 ee 03 4c 89 6d 08 4c 8d 6d 12 4d 01 e6 e9 93 00 00 00 e8 d2 0f 72 fd 48 8d 7b 08 48 89 f8 48 c1 e8 03 <42> 80 3c 20 00 0f 85 fb 02 00 00 48 8b 04 24 4c 8b 7b 08 80 38 00
RSP: 0018:ffffc90001097a48 EFLAGS: 00010002
RAX: 0000000000000001 RBX: 0000000000000000 RCX: 0000000000000000
RDX: ffff8880139de240 RSI: ffffffff8405e5be RDI: 0000000000000008
RBP: ffffc90001097bb0 R08: 0000000000000000 R09: 0000000000000000
R10: ffffffff8405e679 R11: 0000000000000000 R12: dffffc0000000000
R13: ffffc90001097bc2 R14: fffff52000212f76 R15: 0000000000000001
FS: 0000000000000000(0000) GS:ffff8880b9c00000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007ffe25571c98 CR3: 0000000079206000 CR4: 00000000003506f0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
----------------
Code disassembly (best guess):
0: 89 ee mov %ebp,%esi
2: 49 83 ed 40 sub $0x40,%r13
6: 49 c1 ee 03 shr $0x3,%r14
a: 4c 89 6d 08 mov %r13,0x8(%rbp)
e: 4c 8d 6d 12 lea 0x12(%rbp),%r13
12: 4d 01 e6 add %r12,%r14
15: e9 93 00 00 00 jmpq 0xad
1a: e8 d2 0f 72 fd callq 0xfd720ff1
1f: 48 8d 7b 08 lea 0x8(%rbx),%rdi
23: 48 89 f8 mov %rdi,%rax
26: 48 c1 e8 03 shr $0x3,%rax
* 2a: 42 80 3c 20 00 cmpb $0x0,(%rax,%r12,1) <-- trapping instruction
2f: 0f 85 fb 02 00 00 jne 0x330
35: 48 8b 04 24 mov (%rsp),%rax
39: 4c 8b 7b 08 mov 0x8(%rbx),%r15
3d: 80 38 00 cmpb $0x0,(%rax)


---
This report is generated by a bot. It may contain errors.
See https://goo.gl/tpsmEJ for more information about syzbot.
syzbot engineers can be reached at syzk...@googlegroups.com.

syzbot will keep track of this issue. See:
https://goo.gl/tpsmEJ#status for how to communicate with syzbot.

[Matthew Wilcox]

On Mon, Mar 21, 2022 at 05:29:22AM -0700, syzbot wrote:
> Hello,
>
> syzbot found the following issue on:
>
> HEAD commit: 56e337f2cf13 Revert "gpio: Revert regression in sysfs-gpio..
> git tree: upstream
> console output: https://syzkaller.appspot.com/x/log.txt?x=146069d9700000
> kernel config: https://syzkaller.appspot.com/x/.config?x=70f75a89c7a0e6bc
> dashboard link: https://syzkaller.appspot.com/bug?extid=0d2b0bf32ca5cfd09f2e
> compiler: gcc (Debian 10.2.1-6) 10.2.1 20210110, GNU ld (GNU Binutils for Debian) 2.35.2
>
> Unfortunately, I don't have any reproducer for this issue yet.

I do! Thanks for the report; I thought deeply about what the problem
might be and wrote a test that produces this problem:

$ ./xarray
linux/../../../../include/linux/xarray.h:1208:9: runtime error: member access within null pointer of type 'const struct xa_node'
AddressSanitizer:DEADLYSIGNAL
=================================================================
==2288654==ERROR: AddressSanitizer: SEGV on unknown address 0x000000000008 (pc 0x5636348abdcb bp 0x000000000000 sp 0x7ffd64c7a570 T0)
==2288654==The signal is caused by a READ memory access.
==2288654==Hint: address points to the zero page.
#0 0x5636348abdcb in xa_parent_locked linux/../../../../include/linux/xarray.h:1208
#1 0x5636348abdcb in xas_create_range ../../../lib/xarray.c:725
#2 0x5636348b40d1 in check_create_range_5 ../../../lib/test_xarray.c:1474
#3 0x5636348d49da in check_create_range ../../../lib/test_xarray.c:1510
#4 0x5636348d7dc9 in xarray_checks ../../../lib/test_xarray.c:1808

I'll fix it. It's a race condition to reproduce it (khugepaged vs ...
actually, I'm not sure what it's racing against, but the other thing
also creates a THP before khugepaged gets the lock). There's no way
to write a reliable reproducer in userspace, but trivial to write the
test-case once you know what's happening.