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  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, 30 Apr 2012 10:52:30 +0900
From:	Minchan Kim <>
To:	Andrew Morton <>
	Neil Brown <>,
	Artem Bityutskiy <>,
	David Woodhouse <>,
	Theodore Ts'o <>,
	Adrian Hunter <>,
	Steven Whitehouse <>,
	"David S. Miller" <>,
	James Morris <>,
	Alexander Viro <>,
	Sage Weil <>
Subject: Re: [RFC] vmalloc: add warning in __vmalloc

On 04/28/2012 07:00 AM, Andrew Morton wrote:

> On Fri, 27 Apr 2012 17:42:24 +0900
> Minchan Kim <> wrote:
>> Now there are several places to use __vmalloc with GFP_ATOMIC,
>> GFP_NOIO, GFP_NOFS but unfortunately __vmalloc calls map_vm_area
>> which calls alloc_pages with GFP_KERNEL to allocate page tables.
>> It means it's possible to happen deadlock.
>> I don't know why it doesn't have reported until now.
>> Firstly, I tried passing gfp_t to lower functions to support __vmalloc
>> with such flags but other mm guys don't want and decided that
>> all of caller should be fixed.
>> To begin with, let's listen other's opinion whether they can fix it
>> by other approach without calling __vmalloc with such flags.
>> So this patch adds warning to detect and to be fixed hopely.
>> I Cced related maintainers.
>> If I miss someone, please Cced them.
>> side-note:
>>   I added WARN_ON instead of WARN_ONCE to detect all of callers
>>   and each WARN_ON for each flag to detect to use any flag easily.
>>   After we fix all of caller or reduce such caller, we can merge
>>   a warning with WARN_ONCE.
> Just WARN_ONCE, please.  If that exposes some sort of calamity then we
> can reconsider.


>> ...
>> --- a/mm/vmalloc.c
>> +++ b/mm/vmalloc.c
>> @@ -1700,6 +1700,15 @@ static void *__vmalloc_node(unsigned long size, unsigned long align,
>>  			    gfp_t gfp_mask, pgprot_t prot,
>>  			    int node, void *caller)
>>  {
>> +	/*
>> +	 * This function calls map_vm_area so that it allocates
>> +	 * page table with GFP_KERNEL so caller should avoid using
>> +	 * GFP_NOIO, GFP_NOFS and !__GFP_WAIT.
>> +	 */
>> +	WARN_ON(!(gfp_mask & __GFP_WAIT));
>> +	WARN_ON(!(gfp_mask & __GFP_IO));
>> +	WARN_ON(!(gfp_mask & __GFP_FS));
>> +
>>  	return __vmalloc_node_range(size, align, VMALLOC_START, VMALLOC_END,
>>  				gfp_mask, prot, node, caller);
>>  }
> This seems strange.  There are many entry points to this code and the
> patch appears to go into a randomly-chosen middle point in the various
> call chains and sticks a check in there.  Why was __vmalloc_node()
> chosen?  Does this provide full coverage or all entry points?

I think it covers all of caller with calls __vmalloc with gfp_flags.
Only exception is __vmalloc_node_range which is called by module_alloc
but it surely calls __vmalloc_node_range with GFP_KERNEL so it's no problem now.
If you want to catch potential use of __vmalloc_node_range in future, I can move it
to it. 

> Also, the patch won't warn in the most problematic cases such as
> vmalloc() being called from a __GFP_NOFS context.  Presumably there are

I agree but this patch's goal is just to prevent calling __vmalloc with GFP_ATOMIC, GFP_NOFS and
GFP_NOIO. We should consider vmalloc on __GFP_NOFS context as another problem and maybe
reclaimfs lockdep would be a good start point.

> might_sleep() warnings somewhere on the allocation path which will

> catch vmalloc() being called from atomic contexts.


> I'm not sure what to do about that - we don't have machinery in place
> to be able to detect when a GFP_KERNEL allocation is deadlockable. 
> Perhaps a lot of hacking on lockdep might get us this - we'd need to
> teach lockdep about which locks prohibit FS entry, which locks prevent
> IO entry, etc.  And there are secret locks such as ext3/4
> journal_start(), and bitlocks and lock_page().  eek.


> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to  For more info on Linux MM,
> see: .
> Fight unfair telecom internet charges in Canada: sign
> Don't email: <a href=mailto:""> </a>

Kind regards,
Minchan Kim
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
More majordomo info at
Please read the FAQ at

Powered by blists - more mailing lists