On 05/24/2012 09:31 AM, Xiao Guangrong wrote: > On 05/23/2012 07:37 PM, Avi Kivity wrote: > >> On 05/23/2012 11:51 AM, Xiao Guangrong wrote: >>> Changlog: >>> This are some changes from Marcelo's review: >>> - drop SPTE_WRITE_PROTECT bit, now, only one bit is needed to do >>> lockless update. >>> >>> - always atomicly update spte if it can be updated out of mmu-lock. >>> >>> - flod the judgement of tlb flush into mmu_spte_update(), make the >>> code easilyer audited. >>> >>> Performance result: >>> - autoest migration test (smp2.Fedora.16.64.migrate.with_autotest.dbench.unix): >>> before after >>> smp2.Fedora.16.64.migrate.unix 93 91 +2.1% >>> smp2.Fedora.16.64.migrate.with_autotest.dbench.unix 218 188 +13.7% >>> >>> - the benchmark attached is used to measure the resuming time >>> after dirty-log >>> before after >>> Run 10 times, Avg time: 512466818 ns. 269231261 ns +47.5% >> >> Still scary (esp. indirect sptes), but looks pretty good. > > > Hmm, i do not have a better way to solve the ABA problem on indirect spte now. :( > How about only allow fast page fault to work for direct spte? > I'll certainly be more comfortable with that, at least to start with. -- error compiling committee.c: too many arguments to function -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
