[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] RE: [Xen-ia64-devel] [IPF-ia64] with Cset 10690, creating a VTImake xen0 hang
Hi Alex, When VTI-domain does IO operation, an event is sent to dom0, then Qemu is woken up to handle this request. Process switch on dom0 may call itr_d, if task_struct of qemu is not covered by TR. But the emulation of itr_d is very slow, due to purge vhpt. This is the root cause of slowness of vti-domain boot on your side, and vti-domain can't boot on my side. I think purging vhpt in emulating itr_d also slow down dom0 and domU, So the question is, is purging vhpt in emulating itr_d necessary? Thanks, Anthony >-----Original Message----- >From: xen-ia64-devel-bounces@xxxxxxxxxxxxxxxxxxx >[mailto:xen-ia64-devel-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of Zhang, >Xiantao >Sent: 2006年7月12日 10:56 >To: Atsushi SAKAI; Alex Williamson >Cc: Isaku Yamahata; xen-ia64-devel@xxxxxxxxxxxxxxxxxxx >Subject: RE: [Xen-ia64-devel] [IPF-ia64] with Cset 10690,creating a VTImake >xen0 hang > >Hi Alex, > Our QA team reported that xen0 would hang after creating VTi domain > @CSet >10688-10694. Should we reverse the logic of vcpu_itr_d first ? Maybe we can >find out the real bottleneck later.:) >Thanks & Best Regards >-Xiantao > >OTC,Intel Corporation > >> -----Original Message----- >> From: Atsushi SAKAI [mailto:sakaia@xxxxxxxxxxxxxx] >> Sent: 2006年7月11日 21:46 >> To: Alex Williamson; Zhang, Xiantao >> Cc: Isaku Yamahata; xen-ia64-devel@xxxxxxxxxxxxxxxxxxx >> Subject: Re: [Xen-ia64-devel] [IPF-ia64] with Cset 10690, creating a VTImake >> xen0 hang >> >> Hi, Alex >> >> Sorry for late. >> >> I found your problem(boot time difference w/ PAL_HALT_LIGHT emulation patch) >> occurred in SMP(credit). >> But, it does not occurred in UP, SMP(bvt) and SMP(credit w/ affinity). >> >> I think the emulation of pal_halt_light for domU >> does not good work for DomVTI boot up >> under credit scheduling w/o affinity. >> >> And consider the Xiantao survey, >> qemu make heavy I/O operations at the boot up. >> >> Consider the above two conditions, >> I think credit scheduler algorithm does not consider >> the block state.(caused by pal_halt_light emulation) >> So I want to switch off the vcpu migration at heavy load >> >> >> I planned as follows. >> >> 1)In the short term, >> I want to avoid this problem by >> HALT the PAL_HALT_LIGHT emulation while DomVTI boot up. >> or >> Lock VCPUs migrations while DomVTI boot up. >> (when Credit scheduler runs) >> >> 2)In the long term, >> I will make a patch to avoid this problem. >> (Consider the heavy io w/ vcpu migration) >> >> N.B. >> I checked under CS:10559.(original patch made) >> >> Thanks, >> Atsushi SAKAI >> >> >> >> >> >> >On Tue, 2006-07-11 at 19:42 +0800, Zhang, Xiantao wrote: >> >> Hi Alex, >> >> Seems this issue was caused by Cset 10688. In vcpu_itr_d, the current >> >> logic purges vhpt with cpu_flush_vhpt_range but it is very heavy to >> >> xen0. When creating VTi domain @ early stage, IO operation is very >> >> excessive, so qemu was scheduled out and in very frequently and this >> >> logic was executed every time. In addition, cpu_flush_vhpt_range using >> >> identity map to purge vhpt may cause more tlb miss due to no TR map. >> >> If remove vcpu_flush_tlb_vhpt_range logic although it definitely >> >> needed, seems VTi becomes healthy. Maybe potential bugs exist there.:) >> > >> > Thanks for investigating Xiantao. Isaku, any thoughts on how to >> >regain VTI performance? Thanks, >> > >> > Alex >> > >> >-- >> >Alex Williamson HP Open Source & Linux Org. >> > >> > >> >_______________________________________________ >> >Xen-ia64-devel mailing list >> >Xen-ia64-devel@xxxxxxxxxxxxxxxxxxx >> >http://lists.xensource.com/xen-ia64-devel >> > >> >> >> > >_______________________________________________ >Xen-ia64-devel mailing list >Xen-ia64-devel@xxxxxxxxxxxxxxxxxxx >http://lists.xensource.com/xen-ia64-devel _______________________________________________ Xen-ia64-devel mailing list Xen-ia64-devel@xxxxxxxxxxxxxxxxxxx http://lists.xensource.com/xen-ia64-devel
|
![]() |
Lists.xenproject.org is hosted with RackSpace, monitoring our |