On 2019-11-08 18:54, Damien Le Moal wrote:
> On 2019/11/09 4:06, Bart Van Assche wrote:
>> On 11/7/19 5:57 PM, Damien Le Moal wrote:
>>> - buf = vzalloc(bufsize);
>>> - if (buf)
>>> - *buflen = bufsize;
>>> + while (bufsize >= SECTOR_SIZE) {
>>> + buf = vzalloc(bufsize);
>>> + if (buf) {
>>> + *buflen = bufsize;
>>> + return buf;
>>> + }
>>> + bufsize >>= 1;
>>> + }
>>
>> Hi Damien,
>>
>> Has it been considered to pass the __GFP_NORETRY flag to this vzalloc()
>> call?
>
> Do you mean using
>
> __vmalloc(bufsize,
> GFP_KERNEL | __GFP_ZERO | __GFP_NORETRY, PAGE_KERNEL);
>
> instead of vzalloc() ? (since we cannot pass GFP flags to vzalloc()...)
>
> Note that this is called with GFP_NOIO set for the caller context in the
> case of revalidate zones, and default to GFP_KERNEL for
> blkdev_report_zones() unless the caller also tweaks the context memalloc
> flags.
Hi Damien,
Yes, that's what I meant. The following comment from mm/util.c explains
why __GFP_RETRY should be used if it is OK for an allocation to fail:
/*
* We want to attempt a large physically contiguous block first because
* it is less likely to fragment multiple larger blocks and therefore
* contribute to a long term fragmentation less than vmalloc fallback.
* However make sure that larger requests are not too disruptive - no
* OOM killer and no allocation failure warnings as we have a fallback.
*/
Thanks,
Bart.
