linux-kernel - Re: [BUG] mm/vmalloc: ___might_sleep is called under a spinlock in __purge_vmap_area

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 for Android: free password hash cracker in your pocket

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

Message-ID: <56ef9c88-ba54-ce01-15dc-7b661b64ab8b@163.com>
Date:   Mon, 9 Oct 2017 12:48:22 +0800
From:   Jia-Ju Bai <baijiaju1990@....com>
To:     "Kirill A. Shutemov" <kirill@...temov.name>
Cc:     akpm@...ux-foundation.org, mhocko@...e.com, mingo@...nel.org,
        catalin.marinas@....com, labbott@...hat.com, thgarnie@...gle.com,
        kirill.shutemov@...ux.intel.com, aryabinin@...tuozzo.com,
        ard.biesheuvel@...aro.org, zijun_hu@....com, linux-mm@...ck.org,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [BUG] mm/vmalloc: ___might_sleep is called under a spinlock in
 __purge_vmap_area_lazy

Thanks for your reply and explanation :)
I will improve my analysis.

Thanks,
Jia-Ju Bai

On 2017/10/9 12:10, Kirill A. Shutemov wrote:
> On Mon, Oct 09, 2017 at 12:00:33PM +0800, Jia-Ju Bai wrote:
>> The ___might_sleep is called under a spinlock, and the function call graph
>> is:
>> __purge_vmap_area_lazy (acquire the spinlock)
>>    cond_resched_lock
>>      ___might_sleep
>>
>> In this situation, ___might_sleep may prints error log message because a
>> spinlock is held.
>> A possible fix is to remove ___might_sleep in cond_resched_lock.
>>
>> This bug is found by my static analysis tool and my code review.
> This analysis doesn't makes sense.
>
> The point of cond_resched_lock() is that it drops the lock, if resched is
> required.
>
> ___might_sleep() is called with preempt_offset equal to
> PREEMPT_LOCK_OFFSET, so it won't report error if it's the only lock we
> hold.
>