| Message ID | 20240619144215.3273989-1-zhao1.liu@intel.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | target/i386/cpu: Misc Cleanup on host-cache-info | expand |
Hi Igor, Just a gentle poke and what do you think about this minor series? Thanks, Zhao On Wed, Jun 19, 2024 at 10:42:12PM +0800, Zhao Liu wrote: > Date: Wed, 19 Jun 2024 22:42:12 +0800 > From: Zhao Liu <zhao1.liu@intel.com> > Subject: [PATCH 0/3] target/i386/cpu: Misc Cleanup on host-cache-info > X-Mailer: git-send-email 2.34.1 > > Hi, > > This series is mainly to addresss Igor's comment about if one check in > host-cache-info could be removed [1], i.e., whether Guest's cache > topology should be self-consistent (able to correspond to Guest's CPU > topology level, as we currently do with the Guest cache topo). > > I originally thought (in the mail thread with Igor) that host-cache-info > should allow Guest and Host to have the same topology level information, > e.g. if Host shares cache on core level, then via host-cache-info, Guest > should also share on core level. > > But in practice, I gave up on this idea, because in the cache info > passthrough case, it should be possible for Guest to get the original > Host cache info (including the original threads sharing cache) without > further modifying the info to Guest. > > Therefore, I simply added the comment in PATCH 3 to hopefully illustrate > the need for such a check. > > Hope my explanation is clear enough so that my poor English doesn't > bother you! > > [1]: https://lore.kernel.org/qemu-devel/20240527170317.14520a2f@imammedo.users.ipa.redhat.com/ > > Thanks and Best Regards, > Zhao > --- > Zhao Liu (3): > target/i386/cpu: Use hex mask to check for valid cache CPUID leaf > target/i386/cpu: Check guest_thread_ids_per_pkg for host-cache-info > case > target/i386/cpu: Add comment about adjusting the Guest cache topo for > host-cache-info > > target/i386/cpu.c | 24 +++++++++++++++++++----- > 1 file changed, 19 insertions(+), 5 deletions(-) > > -- > 2.34.1 > >
Hi Igor, Just a friendly ping :-) May I ask if you are satisfied with the clarification in this series? Thanks, Zhao On Mon, Jul 08, 2024 at 10:23:26AM +0800, Zhao Liu wrote: > Date: Mon, 8 Jul 2024 10:23:26 +0800 > From: Zhao Liu <zhao1.liu@intel.com> > Subject: Re: [PATCH 0/3] target/i386/cpu: Misc Cleanup on host-cache-info > > Hi Igor, > > Just a gentle poke and what do you think about this minor series? > > Thanks, > Zhao > > On Wed, Jun 19, 2024 at 10:42:12PM +0800, Zhao Liu wrote: > > Date: Wed, 19 Jun 2024 22:42:12 +0800 > > From: Zhao Liu <zhao1.liu@intel.com> > > Subject: [PATCH 0/3] target/i386/cpu: Misc Cleanup on host-cache-info > > X-Mailer: git-send-email 2.34.1 > > > > Hi, > > > > This series is mainly to addresss Igor's comment about if one check in > > host-cache-info could be removed [1], i.e., whether Guest's cache > > topology should be self-consistent (able to correspond to Guest's CPU > > topology level, as we currently do with the Guest cache topo). > > > > I originally thought (in the mail thread with Igor) that host-cache-info > > should allow Guest and Host to have the same topology level information, > > e.g. if Host shares cache on core level, then via host-cache-info, Guest > > should also share on core level. > > > > But in practice, I gave up on this idea, because in the cache info > > passthrough case, it should be possible for Guest to get the original > > Host cache info (including the original threads sharing cache) without > > further modifying the info to Guest. > > > > Therefore, I simply added the comment in PATCH 3 to hopefully illustrate > > the need for such a check. > > > > Hope my explanation is clear enough so that my poor English doesn't > > bother you! > > > > [1]: https://lore.kernel.org/qemu-devel/20240527170317.14520a2f@imammedo.users.ipa.redhat.com/ > > > > Thanks and Best Regards, > > Zhao > > --- > > Zhao Liu (3): > > target/i386/cpu: Use hex mask to check for valid cache CPUID leaf > > target/i386/cpu: Check guest_thread_ids_per_pkg for host-cache-info > > case > > target/i386/cpu: Add comment about adjusting the Guest cache topo for > > host-cache-info > > > > target/i386/cpu.c | 24 +++++++++++++++++++----- > > 1 file changed, 19 insertions(+), 5 deletions(-) > > > > -- > > 2.34.1 > > > > >
Hi, This series is mainly to addresss Igor's comment about if one check in host-cache-info could be removed [1], i.e., whether Guest's cache topology should be self-consistent (able to correspond to Guest's CPU topology level, as we currently do with the Guest cache topo). I originally thought (in the mail thread with Igor) that host-cache-info should allow Guest and Host to have the same topology level information, e.g. if Host shares cache on core level, then via host-cache-info, Guest should also share on core level. But in practice, I gave up on this idea, because in the cache info passthrough case, it should be possible for Guest to get the original Host cache info (including the original threads sharing cache) without further modifying the info to Guest. Therefore, I simply added the comment in PATCH 3 to hopefully illustrate the need for such a check. Hope my explanation is clear enough so that my poor English doesn't bother you! [1]: https://lore.kernel.org/qemu-devel/20240527170317.14520a2f@imammedo.users.ipa.redhat.com/ Thanks and Best Regards, Zhao --- Zhao Liu (3): target/i386/cpu: Use hex mask to check for valid cache CPUID leaf target/i386/cpu: Check guest_thread_ids_per_pkg for host-cache-info case target/i386/cpu: Add comment about adjusting the Guest cache topo for host-cache-info target/i386/cpu.c | 24 +++++++++++++++++++----- 1 file changed, 19 insertions(+), 5 deletions(-)