dma-debug: check_for_stack background?

dma-debug: check_for_stack background?

[Stefan Richter]

Hi,

it took a while but I finally noticed that the DMA debug facility warns 
about mappings of on-stack memory.  What's the problem with that?  Are 
there architectures which don't support such mappings, or is it a 
question of caution not to DMA-map the stack?
-- 
Stefan Richter
-=====-==--= -==- -===-
http://arcgraph.de/sr/

Re: dma-debug: check_for_stack background?

[Ben Dooks]

On Sun, Jun 14, 2009 at 08:56:29PM +0200, Stefan Richter wrote:
> Hi,
>
> it took a while but I finally noticed that the DMA debug facility warns  
> about mappings of on-stack memory.  What's the problem with that?  Are  
> there architectures which don't support such mappings, or is it a  
> question of caution not to DMA-map the stack?

I'm not sure if it is possible on all architectures, however the
DMA mapping documentation prohibts the use of stack memory.

-- 
Ben (ben@fluff.org, http://www.fluff.org/)

  'a smiley only costs 4 bytes'

Re: dma-debug: check_for_stack background?

[Stefan Richter]

Ben Dooks wrote:
> On Sun, Jun 14, 2009 at 08:56:29PM +0200, Stefan Richter wrote:
>> the DMA debug facility warns  
>> about mappings of on-stack memory.  What's the problem with that?
...
> I'm not sure if it is possible on all architectures, however the
> DMA mapping documentation prohibts the use of stack memory.

Ah, it has all been documented in plain sight for ages.  Thanks for the 
pointer.
-- 
Stefan Richter
-=====-==--= -==- -===-
http://arcgraph.de/sr/

Re: dma-debug: check_for_stack background?

[Joerg Roedel]

On Sun, Jun 14, 2009 at 11:43:42PM +0200, Stefan Richter wrote:
> Ben Dooks wrote:
>> On Sun, Jun 14, 2009 at 08:56:29PM +0200, Stefan Richter wrote:
>>> the DMA debug facility warns  about mappings of on-stack memory.  
>>> What's the problem with that?
> ...
>> I'm not sure if it is possible on all architectures, however the
>> DMA mapping documentation prohibts the use of stack memory.
>
> Ah, it has all been documented in plain sight for ages.  Thanks for the  
> pointer.

Problem is that the stack might go away or could be otherwise reused
while the DMA is in-flight. This could cause really hard to debug
problems.

	Joerg

-- 
           | Advanced Micro Devices GmbH
 Operating | Karl-Hammerschmidt-Str. 34, 85609 Dornach bei München
 System    | 
 Research  | Geschäftsführer: Thomas M. McCoy, Giuliano Meroni
 Center    | Sitz: Dornach, Gemeinde Aschheim, Landkreis München
           | Registergericht München, HRB Nr. 43632

Re: dma-debug: check_for_stack background?

[FUJITA Tomonori]

On Mon, 15 Jun 2009 10:13:55 +0200
Joerg Roedel <joerg.roedel@amd.com> wrote:

> On Sun, Jun 14, 2009 at 11:43:42PM +0200, Stefan Richter wrote:
> > Ben Dooks wrote:
> >> On Sun, Jun 14, 2009 at 08:56:29PM +0200, Stefan Richter wrote:
> >>> the DMA debug facility warns  about mappings of on-stack memory.  
> >>> What's the problem with that?
> > ...
> >> I'm not sure if it is possible on all architectures, however the
> >> DMA mapping documentation prohibts the use of stack memory.
> >
> > Ah, it has all been documented in plain sight for ages.  Thanks for the  
> > pointer.
> 
> Problem is that the stack might go away or could be otherwise reused
> while the DMA is in-flight. This could cause really hard to debug
> problems.

Not true. Even if you use the stack buffer for DMA synchronously, it
doesn't work (we had the bad code like starting dma and waiting for
the completion).

On some architectures, we can't use the stack buffer for DMA. That's
why the DMA mapping documentation prohibits the use of stack memory
for DMA.

Re: dma-debug: check_for_stack background?

[Ming Lei]

2009/6/16 FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>:
> On Mon, 15 Jun 2009 10:13:55 +0200
> Joerg Roedel <joerg.roedel@amd.com> wrote:
>
>> On Sun, Jun 14, 2009 at 11:43:42PM +0200, Stefan Richter wrote:
>> > Ben Dooks wrote:
>> >> On Sun, Jun 14, 2009 at 08:56:29PM +0200, Stefan Richter wrote:
>> >>> the DMA debug facility warns  about mappings of on-stack memory.
>> >>> What's the problem with that?
>> > ...
>> >> I'm not sure if it is possible on all architectures, however the
>> >> DMA mapping documentation prohibts the use of stack memory.
>> >
>> > Ah, it has all been documented in plain sight for ages.  Thanks for the
>> > pointer.
>>
>> Problem is that the stack might go away or could be otherwise reused
>> while the DMA is in-flight. This could cause really hard to debug
>> problems.
>
> Not true. Even if you use the stack buffer for DMA synchronously, it
> doesn't work (we had the bad code like starting dma and waiting for
> the completion).

Yes, there is cache alignement problem for stack buffer also.

>
> On some architectures, we can't use the stack buffer for DMA. That's
> why the DMA mapping documentation prohibits the use of stack memory
> for DMA.
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>



-- 
Lei Ming

Re: dma-debug: check_for_stack background?

[Joerg Roedel]

On Tue, Jun 16, 2009 at 03:16:37PM +0900, FUJITA Tomonori wrote:
> On Mon, 15 Jun 2009 10:13:55 +0200
> Joerg Roedel <joerg.roedel@amd.com> wrote:
> 
> > On Sun, Jun 14, 2009 at 11:43:42PM +0200, Stefan Richter wrote:
> > > Ben Dooks wrote:
> > >> On Sun, Jun 14, 2009 at 08:56:29PM +0200, Stefan Richter wrote:
> > >>> the DMA debug facility warns  about mappings of on-stack memory.  
> > >>> What's the problem with that?
> > > ...
> > >> I'm not sure if it is possible on all architectures, however the
> > >> DMA mapping documentation prohibts the use of stack memory.
> > >
> > > Ah, it has all been documented in plain sight for ages.  Thanks for the  
> > > pointer.
> > 
> > Problem is that the stack might go away or could be otherwise reused
> > while the DMA is in-flight. This could cause really hard to debug
> > problems.
> 
> Not true. Even if you use the stack buffer for DMA synchronously, it
> doesn't work (we had the bad code like starting dma and waiting for
> the completion).

I've not written that this is the only problem. I know that there may be
others like the one you experienced. But what I wrote could also happen
and is a problem :)

Joerg

-- 
           | Advanced Micro Devices GmbH
 Operating | Karl-Hammerschmidt-Str. 34, 85609 Dornach bei München
 System    | 
 Research  | Geschäftsführer: Thomas M. McCoy, Giuliano Meroni
 Center    | Sitz: Dornach, Gemeinde Aschheim, Landkreis München
           | Registergericht München, HRB Nr. 43632