Re: BUG in 3.19.0-rc3+

From: Konstantin Khlebnikov
Date: Sun Jan 11 2015 - 05:28:46 EST

On Sun, Jan 11, 2015 at 12:52 PM, Oded Gabbay <oded.gabbay@xxxxxxx> wrote:
>
>
> On 01/11/2015 11:37 AM, Konstantin Khlebnikov wrote:
>> On Sun, Jan 11, 2015 at 11:16 AM, Chris Clayton
>> <chris2553@xxxxxxxxxxxxxx> wrote:
>>> Hi,
>>>
>>> I've done the bisect and the outcome is below, but, because I almost always forget to mention it, I'll say here that I
>>> am running a 32 bit user space on a 64 bit kernel.
>>>
>>> On 01/10/15 20:17, Chris Clayton wrote:
>>>> Hi,
>>>>
>>>> I'm getting a bug a BUG report from a kernel built from a pull (earlier today) of the current development kernel
>>>> (running git describe gives v3.19-rc3-169-geb74926). So that I have useable wireless networking, I have also applied the
>>>> latest seven iwlwifi patches from the wireless-drivers git tree. Prior to today's pull, I was not seeing anything
>>>> unusual in dmesg.
>>>>
>>>> The BUG reported is as follows:
>>>>
>>>> Jan 10 19:41:32 laptop kernel: ------------[ cut here ]------------
>>>> Jan 10 19:41:32 laptop kernel: kernel BUG at mm/rmap.c:399!
>>>> Jan 10 19:41:32 laptop kernel: invalid opcode: 0000 [#1] PREEMPT SMP
>>>> Jan 10 19:41:32 laptop kernel: Modules linked in: rfcomm snd_hda_codec_via iwlmvm coretemp snd_hda_codec_hdmi
>>>> snd_hda_codec_generic snd_hda_intel mac80211 hwmon snd_hda_controller x86_pkg_temp_thermal acpi_cpufreq iwlwifi cfg80211
>>>> snd_hda_codec snd_hwdep
>>>> Jan 10 19:41:32 laptop kernel: CPU: 1 PID: 353 Comm: fc-cache Not tainted 3.19.0-rc3+ #42
>>>> Jan 10 19:41:32 laptop kernel: Hardware name: Notebook W65_67SZ /W65_67SZ
>>>> , BIOS 1.03.05 02/26/2014
>>>> Jan 10 19:41:32 laptop kernel: task: ffff8800da98c5c0 ti: ffff880408dd4000 task.ti: ffff880408dd4000
>>>> Jan 10 19:41:32 laptop kernel: RIP: 0010:[<ffffffff810ef7ea>] [<ffffffff810ef7ea>] unlink_anon_vmas+0x17a/0x200
>>>> Jan 10 19:41:33 laptop kernel: RSP: 0018:ffff880408dd7d88 EFLAGS: 00010286
>>>> Jan 10 19:41:33 laptop kernel: RAX: ffff88040b79e150 RBX: ffff88040b79e140 RCX: 00000000ffffffff
>>>> Jan 10 19:41:33 laptop kernel: RDX: ffffffff00000001 RSI: ffff880409f04360 RDI: ffff880409f04320
>>>> Jan 10 19:41:33 laptop kernel: RBP: ffff88040cb13278 R08: 0000000000000000 R09: ffff88040d801c00
>>>> Jan 10 19:41:33 laptop kernel: R10: ffff88041fa546e0 R11: ffff88040b79e160 R12: ffff880409f04320
>>>> Jan 10 19:41:33 laptop kernel: R13: ffff88040cb13278 R14: ffff88040cb13288 R15: ffff88040cb13210
>>>> Jan 10 19:41:33 laptop kernel: FS: 0000000000000000(0000) GS:ffff88041fa40000(0000) knlGS:0000000000000000
>>>> Jan 10 19:41:33 laptop kernel: CS: 0010 DS: 002b ES: 002b CR0: 0000000080050033
>>>> Jan 10 19:41:33 laptop kernel: CR2: 00000000f722c8d4 CR3: 00000004082a8000 CR4: 00000000001407e0
>>>> Jan 10 19:41:33 laptop kernel: Stack:
>>>> Jan 10 19:41:33 laptop kernel: ffff88040d6cfbd8 ffff88040d6cfba0 ffff88040cecd160 ffff88040cb13210
>>>> Jan 10 19:41:33 laptop kernel: ffff88040cbbb630 00000000f7151000 ffff880408dd7e28 0000000000000000
>>>> Jan 10 19:41:33 laptop kernel: 0000000000000000 ffffffff810e3633 0000000000000000 0000000000000000
>>>> Jan 10 19:41:33 laptop kernel: Call Trace:
>>>> Jan 10 19:41:33 laptop kernel: [<ffffffff810e3633>] ? free_pgtables+0x83/0xf0
>>>> Jan 10 19:41:34 laptop kernel: [<ffffffff810ec3c3>] ? exit_mmap+0xc3/0x150
>>>> Jan 10 19:41:34 laptop kernel: [<ffffffff8103980d>] ? __do_page_fault+0x17d/0x4b0
>>>> Jan 10 19:41:34 laptop kernel: [<ffffffff81042a21>] ? mmput+0x21/0xc0
>>>> Jan 10 19:41:34 laptop kernel: [<ffffffff8104673d>] ? do_exit+0x26d/0xa50
>>>> Jan 10 19:41:34 laptop kernel: [<ffffffff8111fe89>] ? mntput_no_expire+0x9/0x140
>>>> Jan 10 19:41:34 laptop kernel: [<ffffffff8105ca1c>] ? task_work_run+0xbc/0xf0
>>>> Jan 10 19:41:34 laptop kernel: [<ffffffff81047d44>] ? do_group_exit+0x34/0xb0
>>>> Jan 10 19:41:34 laptop kernel: [<ffffffff81047dcf>] ? SyS_exit_group+0xf/0x10
>>>> Jan 10 19:41:34 laptop kernel: [<ffffffff815e0f9f>] ? sysenter_dispatch+0x7/0x1e
>>>> Jan 10 19:41:34 laptop kernel: Code: 00 ad de 48 89 43 18 e8 c5 f9 00 00 48 8b 45 10 48 8d 55 10 48 83 e8 10 49 39 d6 74
>>>> 54 48 8b 7d 08 48 89 eb 8b 57 34 85 d2 74 9e <0f> 0b 0f 1f 40 00 e8 6b fc ff ff eb 9a 66 0f 1f 84 00 00 00 00
>>>> Jan 10 19:41:34 laptop kernel: RIP [<ffffffff810ef7ea>] unlink_anon_vmas+0x17a/0x200
>>>> Jan 10 19:41:34 laptop kernel: RSP <ffff880408dd7d88>
>>>> Jan 10 19:41:34 laptop kernel: ---[ end trace 4aa713b2a9aa664b ]---
>>>> Jan 10 19:41:34 laptop kernel: Fixing recursive fault but reboot is needed!
>>>> Jan 10 19:41:34 laptop kernel: nf_conntrack version 0.5.0 (16384 buckets, 65536 max)
>>>
>>> [snip]
>>>
>>>>
>>>> I won't get time tonight, but I can bisect it tomorrow, so this is just a heads up in case the problem (and fix) jumps
>>>> out at anyone. Before I bisect I'll build and run a kernel without the iwlwifi patches.
>>>
>>> The bisect ended up at:
>>>
>>> 7a3ef208e662f4b63d43a23f61a64a129c525bbc is the first bad commit
>>> commit 7a3ef208e662f4b63d43a23f61a64a129c525bbc
>>> Author: Konstantin Khlebnikov <koct9i@xxxxxxxxx>
>>> Date: Thu Jan 8 14:32:15 2015 -0800
>>>
>>> mm: prevent endless growth of anon_vma hierarchy
>>>
>>> Constantly forking task causes unlimited grow of anon_vma chain. Each
>>> next child allocates new level of anon_vmas and links vma to all
>>> previous levels because pages might be inherited from any level.
>>>
>>> This patch adds heuristic which decides to reuse existing anon_vma
>>> instead of forking new one. It adds counter anon_vma->degree which
>>> counts linked vmas and directly descending anon_vmas and reuses anon_vma
>>> if counter is lower than two. As a result each anon_vma has either vma
>>> or at least two descending anon_vmas. In such trees half of nodes are
>>> leafs with alive vmas, thus count of anon_vmas is no more than two times
>>> bigger than count of vmas.
>>>
>>> This heuristic reuses anon_vmas as few as possible because each reuse
>>> adds false aliasing among vmas and rmap walker ought to scan more ptes
>>> when it searches where page is might be mapped.
>>>
>>> Link: http://lkml.kernel.org/r/20120816024610.GA5350@xxxxxxxxxxxxxxxxxxxxxxx
>>> Fixes: 5beb49305251 ("mm: change anon_vma linking to fix multi-process server scalability issue")
>>> [akpm@xxxxxxxxxxxxxxxxxxxx: fix typo, per Rik]
>>> Signed-off-by: Konstantin Khlebnikov <koct9i@xxxxxxxxx>
>>> Reported-by: Daniel Forrest <dan.forrest@xxxxxxxxxxxxx>
>>> Tested-by: Michal Hocko <mhocko@xxxxxxx>
>>> Tested-by: Jerome Marchand <jmarchan@xxxxxxxxxx>
>>> Reviewed-by: Michal Hocko <mhocko@xxxxxxx>
>>> Reviewed-by: Rik van Riel <riel@xxxxxxxxxx>
>>> Cc: <stable@xxxxxxxxxxxxxxx> [2.6.34+]
>>> Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
>>> Signed-off-by: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
>>>
>>> :040000 040000 ca27f69d02743e7347b19b1e07976732a49698d1 7104c9ec5eb200ee4a21548e15d2b71a5806e107 M include
>>> :040000 040000 5440efbda5ac44c2a2da7e068f40ee6f0d4c0c7e b76fd93bffebec1acdef5f1785eb578c5f4f6cc3 M mm
>>>
>>> I'm more than happy to provide additional diagnostics and/or test any patches, but please cc me as I'm not subscribed.
>>
>> Looks like degree (%edx) is 1 on anon-vma desruction.
>> Probably I've overlooked some weird conrner case in vma splitting/merging.
>>
>> Could you try this patch. It disables vma merging end eliminates half
>> of complicated paths.
>> As I see merging is optional, everything should work fine without it.
>>
>> --- a/mm/mmap.c
>> +++ b/mm/mmap.c
>> @@ -1048,7 +1048,7 @@ struct vm_area_struct *vma_merge(struct mm_struct *mm,
>> * We later require that vma->vm_flags == vm_flags,
>> * so this tests vma->vm_flags & VM_SPECIAL, too.
>> */
>> - if (vm_flags & VM_SPECIAL)
>> + if (1)
>> return NULL;
>>
>> if (prev)
>>
>>
>>
>>
>> Code from your oops.
>>
>> Code: 00 ad de 48 89 43 18 e8 c5 f9 00 00 48 8b 45 10 48 8d 55 10 48
>> 83 e8 10 49 39 d6 74 54 48 8b 7d 08 48 89 eb 8b 57 34 85 d2 74 9e <0f>
>> 0b 0f 1f 40 00 e8 6b fc ff ff eb 9a 66 0f 1f 84 00 00 00 00
>> All code
>> ========
>> 0: 00 ad de 48 89 43 add %ch,0x438948de(%rbp)
>> 6: 18 e8 sbb %ch,%al
>> 8: c5 f9 00 (bad)
>> b: 00 48 8b add %cl,-0x75(%rax)
>> e: 45 10 48 8d adc %r9b,-0x73(%r8)
>> 12: 55 push %rbp
>> 13: 10 48 83 adc %cl,-0x7d(%rax)
>> 16: e8 10 49 39 d6 callq 0xffffffffd639492b
>> 1b: 74 54 je 0x71
>> 1d: 48 8b 7d 08 mov 0x8(%rbp),%rdi
>> 21: 48 89 eb mov %rbp,%rbx
>> 24: 8b 57 34 mov 0x34(%rdi),%edx
>> 27: 85 d2 test %edx,%edx
>> 29: 74 9e je 0xffffffffffffffc9
>> 2b:* 0f 0b ud2 <-- trapping instruction
>> 2d: 0f 1f 40 00 nopl 0x0(%rax)
>> 31: e8 6b fc ff ff callq 0xfffffffffffffca1
>> 36: eb 9a jmp 0xffffffffffffffd2
>> 38: 66 data16
>> 39: 0f .byte 0xf
>> 3a: 1f (bad)
>> 3b: 84 00 test %al,(%rax)
>> 3d: 00 00 add %al,(%rax)
>> ...
>>
>> Code starting with the faulting instruction
>> ===========================================
>> 0: 0f 0b ud2
>> 2: 0f 1f 40 00 nopl 0x0(%rax)
>> 6: e8 6b fc ff ff callq 0xfffffffffffffc76
>> b: eb 9a jmp 0xffffffffffffffa7
>> d: 66 data16
>> e: 0f .byte 0xf
>> f: 1f (bad)
>> 10: 84 00 test %al,(%rax)
>> 12: 00 00 add %al,(%rax)
>>
>>
>> +Added Oded Gabbay <oded.gabbay@xxxxxxx> into cc, he's reported this
>> problem too.
>>
> Thanks for the fast reply.
>
> I applied the patch and tested it. I wasn't able to reproduce *my* problem,
> so you are definitely in the right direction :)

Ok. I've found something. Try patch from attachment.

>
> Oded
>
>>>
>>> In case it helps, I've attached the xz-compressed related config file.
>>>
>>> Chris
>>>
>>>>
>>>> I've attached the full kernel log file for that boot.
>>>>
>>>> Chris
>>>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
>> the body of a message to majordomo@xxxxxxxxxxxxxxx
>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>> Please read the FAQ at http://www.tux.org/lkml/
>>

Attachment: mm-fix-anon_vma-degree-counter-in-case-of-anon_vma-export
Description: Binary data