Re: [Xen-devel] [PATCHv10 4/4] gnttab: use per-VCPU maptrack free lists

2015-05-29 Thread Jan Beulich
>>> On 28.05.15 at 18:10, wrote: > On 28/05/15 16:39, Jan Beulich wrote: > On 26.05.15 at 20:00, wrote: >>> --- a/xen/common/grant_table.c >>> +++ b/xen/common/grant_table.c >>> @@ -57,7 +57,7 @@ integer_param("gnttab_max_frames", max_grant_frames); >>> * New options allow to set max_maptra

Re: [Xen-devel] [PATCHv10 4/4] gnttab: use per-VCPU maptrack free lists

2015-05-28 Thread David Vrabel
On 28/05/15 16:39, Jan Beulich wrote: On 26.05.15 at 20:00, wrote: >> --- a/xen/common/grant_table.c >> +++ b/xen/common/grant_table.c >> @@ -57,7 +57,7 @@ integer_param("gnttab_max_frames", max_grant_frames); >> * New options allow to set max_maptrack_frames and >> * map_grant_table_fram

Re: [Xen-devel] [PATCHv10 4/4] gnttab: use per-VCPU maptrack free lists

2015-05-28 Thread Jan Beulich
>>> On 26.05.15 at 20:00, wrote: > v10: > - Divide max_maptrack_frames evenly amongst the VCPUs. > - Increase default max_maptrack_frames to compensate. I'm not really happy about this, but also can't immediately suggest a better model (without removing maptrack frames from vCPU-s). I'd really li

[Xen-devel] [PATCHv10 4/4] gnttab: use per-VCPU maptrack free lists

2015-05-26 Thread David Vrabel
From: Malcolm Crossley Performance analysis of aggregate network throughput with many VMs shows that performance is signficantly limited by contention on the maptrack lock when obtaining/releasing maptrack handles from the free list. Instead of a single free list use a per-VCPU list. This avoids