Re: CMA alignment question

[Weijie Yang <weijie.yang.kh@gmail.com>]

On Wed, Nov 5, 2014 at 12:18 PM, Gregory Fong <gregory.0xf0@gmail.com> wrote:
> On Tue, Nov 4, 2014 at 2:27 PM, Michal Nazarewicz <mina86@mina86.com> wrote:
>> On Tue, Nov 04 2014, Gregory Fong wrote:
>>> The alignment in cma_alloc() is done w.r.t. the bitmap.  This is a
>>> problem when, for example:
>>>
>>> - a device requires 16M (order 12) alignment
>>> - the CMA region is not 16 M aligned

I think the device driver should ensure that situation could not occur,
by assign suitable alignment parameter in cma_declare_contiguous().

>>> In such a case, can result with the CMA region starting at, say,
>>> 0x2f800000 but any allocation you make from there will be aligned from
>>> there.  Requesting an allocation of 32 M with 16 M alignment, will
>>> result in an allocation from 0x2f800000 to 0x31800000, which doesn't
>>> work very well if your strange device requires 16M alignment.
>>>
>>> This doesn't have the behavior I would expect, which would be for the
>>> allocation to be aligned w.r.t. the start of memory.  I realize that
>>> aligning the CMA region is an option, but don't see why cma_alloc()
>>> aligns to the start of the CMA region.  Is there a good reason for
>>> having cma_alloc() alignment work this way?
>>
>> No, it's a bug.  The alignment should indicate alignment of physical
>> address not position in CMA region.
>>
>
> Ah, now I see that Marek submitted this patch from you back in 2011
> that would have allowed the bitmap lib to support an alignment offset:
> http://thread.gmane.org/gmane.linux.kernel/1121103/focus=1121100
>
> Any idea why this didn't make it into the later changesets?  If not,
> I'll resubmit it and to use it to fix this bug.
>
> Thanks,
> Gregory
> --
> 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/

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>