linux-kernel - Re: Sleeping BUG in khugepaged for i586

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Mon, 5 Jun 2017 14:44:01 -0700
From:   Andrew Morton <akpm@...ux-foundation.org>
To:     Larry Finger <Larry.Finger@...inger.net>
Cc:     LKML <linux-kernel@...r.kernel.org>, linux-mm@...ck.org
Subject: Re: Sleeping BUG in khugepaged for i586

On Sat, 3 Jun 2017 14:24:26 -0500 Larry Finger <Larry.Finger@...inger.net> wrote:

> I recently turned on locking diagnostics for a Dell Latitude D600 laptop, which 
> requires a 32-bit kernel. In the log I found the following:
> 
> BUG: sleeping function called from invalid context at mm/khugepaged.c:655
> in_atomic(): 1, irqs_disabled(): 0, pid: 20, name: khugepaged
> 1 lock held by khugepaged/20:
>   #0:  (&mm->mmap_sem){++++++}, at: [<c03d6609>] 
> collapse_huge_page.isra.47+0x439/0x1240
> CPU: 0 PID: 20 Comm: khugepaged Tainted: G        W 
> 4.12.0-rc1-wl-12125-g952a068 #80
> Hardware name: Dell Computer Corporation Latitude D600 
> /03U652, BIOS A05 05/29/2003
> Call Trace:
>   dump_stack+0x76/0xb2
>   ___might_sleep+0x174/0x230
>   collapse_huge_page.isra.47+0xacf/0x1240
>   khugepaged_scan_mm_slot+0x41e/0xc00
>   ? _raw_spin_lock+0x46/0x50
>   khugepaged+0x277/0x4f0
>   ? prepare_to_wait_event+0xe0/0xe0
>   kthread+0xeb/0x120
>   ? khugepaged_scan_mm_slot+0xc00/0xc00
>   ? kthread_create_on_node+0x30/0x30
>   ret_from_fork+0x21/0x30
> 
> I have no idea when this problem was introduced. Of course, I will test any 
> proposed fixes.
> 

Odd.  There's nothing wrong with cond_resched() while holding mmap_sem.
It looks like khugepaged forgot to do a spin_unlock somewhere and we
leaked a preempt_count.