Re: [PATCH] [RFC] xfs: work around unlikely() profiler glitch

[Arnd Bergmann]

On Wed, Jan 25, 2017 at 4:09 PM, Christoph Hellwig <hch@xxxxxxxxxxxxx> wrote:
> On Wed, Jan 25, 2017 at 03:08:10PM +0100, Arnd Bergmann wrote:
>> diff --git a/fs/xfs/libxfs/xfs_bmap.c b/fs/xfs/libxfs/xfs_bmap.c
>> index d22f7930eb75..dca3ddd737d4 100644
>> --- a/fs/xfs/libxfs/xfs_bmap.c
>> +++ b/fs/xfs/libxfs/xfs_bmap.c
>> @@ -3629,7 +3629,7 @@ xfs_bmap_btalloc(
>>               align = xfs_get_cowextsz_hint(ap->ip);
>>       else if (xfs_alloc_is_userdata(ap->datatype))
>>               align = xfs_get_extsz_hint(ap->ip);
>> -     if (unlikely(align)) {
>> +     if (unlikely_notrace(align)) {
>>               error = xfs_bmap_extsize_align(mp, &ap->got, &ap->prev,
>>                                               align, 0, ap->eof, 0, ap->conv,
>>                                               &ap->offset, &ap->length);
>
> The unlikely calls on align in xfs_bmap_btalloc should simply be
> removed.  They aren't actually unlikely for many workloads.  I have
> a patch in my queue that I can expedite based on your report.

That would defines help, thanks!

I also noticed that my patch wouldn't work, as unlikely_notrace() is not
defined unless we are actually tracing, so while it fixes some rare
configurations, it breaks all the configurations that matter.

    Arnd