Re: CVS commit: src/sys/arch/mips/mips
skrll@ wrote: > > (i.e. reverting removed lines is not "fix" but workaround). > > Not sure what you mean here, but as bad cache aliases can happen my > change is valid. Well, I have been waitng proper description about the new UVM design for VIPT cache systems (i.e. what's the "right" thing) for >5 years, as mentioned in the PR... If it doesn't work as expected, I think it would be better to remove all useless COLORMATCH code. --- Izumi Tsutsui
Re: CVS commit: src/sys/arch/mips/mips
On 06/30/16 15:59, Izumi Tsutsui wrote: skrll@ wrote: Module Name:src Committed By: skrll Date: Thu Jun 30 12:57:35 UTC 2016 Modified Files: src/sys/arch/mips/mips: pmap.c Log Message: Fix MIPS3_NO_PV_UNCACHED alias handling by looping through the pv_list looking for bad aliases and removing the bad entries. That is, revert to the code before the matt-mips64 merge. : Fixes the following two PRs : Hmm. This means current implementation of UVM_KMF_COLORMATCH and UVM_FLAG_COLORMATCH is completely broken Probably :) (i.e. reverting removed lines is not "fix" but workaround). Not sure what you mean here, but as bad cache aliases can happen my change is valid. See PR/45746 --- Izumi Tsutsui Nick
Re: CVS commit: src/sys/arch/mips/mips
skrll@ wrote: > Module Name: src > Committed By: skrll > Date: Thu Jun 30 12:57:35 UTC 2016 > > Modified Files: > src/sys/arch/mips/mips: pmap.c > > Log Message: > Fix MIPS3_NO_PV_UNCACHED alias handling by looping through the pv_list > looking for bad aliases and removing the bad entries. That is, revert > to the code before the matt-mips64 merge. : > Fixes the following two PRs : Hmm. This means current implementation of UVM_KMF_COLORMATCH and UVM_FLAG_COLORMATCH is completely broken (i.e. reverting removed lines is not "fix" but workaround). See PR/45746 --- Izumi Tsutsui