On 2010-10-22 15:04, Peter Zijlstra wrote:
> On Thu, 2010-10-21 at 20:07 +0200, Jens Axboe wrote:
>>
>>>> + do {
>>>> + bio = bio_alloc(GFP_NOIO, nvecs);
>>>> + nvecs >>= 1;
>>>> + } while (bio == NULL);
>>>
>>> This is surly bad. bio_alloc must be allowed to fail
>>> (Specially with GFP_NOIO). You should only loop down to
>>> 1 and then prepare to return -ENOMEM from this function
>>> and handle it properly in callers. (Or schedule and wait
>>> like below)
>>
>> Since __GFP_WAIT is set, it'll never return NULL. And as long as you
>> don't allocate more than 1 before doing you submit_bio(), it should be
>> OK in this case.
>
> __GFP_WAIT can return NULL, on OOM and when the size is over a magic
> threshold.
The memory allocator, yes, but a mempool backed allocation with
__GFP_WAIT cannot. That's the difference.
> I find it bad form to rely on any allocation not failing.
On that we can agree, but it's not always easily doable. I bet you would
not like the block layer tossing IOs if it can't get a bio or request
structure :-)
--
Jens Axboe
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
