Changeset
target/i386/cpu.c | 10 +++++++++-
1 file changed, 9 insertions(+), 1 deletion(-)
Git apply log
Switched to a new branch '20180416163953.31188-1-ehabkost@redhat.com'
Applying: i386: Don't automatically enable FEAT_KVM_HINTS bits
To https://github.com/patchew-project/qemu
 * [new tag]         patchew/20180416163953.31188-1-ehabkost@redhat.com -> patchew/20180416163953.31188-1-ehabkost@redhat.com
Test passed: checkpatch

loading

Test passed: docker-mingw@fedora

loading

Test passed: docker-build@min-glib

loading

Test passed: s390x

loading

[Qemu-devel] [PULL 0/1] i386: Don't automatically enable FEAT_KVM_HINTS bits
Posted by Eduardo Habkost, 3 days ago
Last remaining fix for -rc4.

The following changes since commit 042f6a31af3d38eefc6ec995cce1d762c41d4515:

  Merge remote-tracking branch 'remotes/maxreitz/tags/pull-block-2018-04-16' into staging (2018-04-16 15:30:54 +0100)

are available in the Git repository at:

  git://github.com/ehabkost/qemu.git tags/x86-next-pull-request

for you to fetch changes up to 0d914f39a77954bf5472b16130aeefe573a068f5:

  i386: Don't automatically enable FEAT_KVM_HINTS bits (2018-04-16 13:36:52 -0300)

----------------------------------------------------------------
i386: Don't automatically enable FEAT_KVM_HINTS bits

Bug fix for "-cpu host" with newer kernels.

----------------------------------------------------------------

Eduardo Habkost (1):
  i386: Don't automatically enable FEAT_KVM_HINTS bits

 target/i386/cpu.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

-- 
2.14.3


Re: [Qemu-devel] [PULL 0/1] i386: Don't automatically enable FEAT_KVM_HINTS bits
Posted by Peter Maydell, 3 days ago
On 16 April 2018 at 17:39, Eduardo Habkost <ehabkost@redhat.com> wrote:
> Last remaining fix for -rc4.
>
> The following changes since commit 042f6a31af3d38eefc6ec995cce1d762c41d4515:
>
>   Merge remote-tracking branch 'remotes/maxreitz/tags/pull-block-2018-04-16' into staging (2018-04-16 15:30:54 +0100)
>
> are available in the Git repository at:
>
>   git://github.com/ehabkost/qemu.git tags/x86-next-pull-request
>
> for you to fetch changes up to 0d914f39a77954bf5472b16130aeefe573a068f5:
>
>   i386: Don't automatically enable FEAT_KVM_HINTS bits (2018-04-16 13:36:52 -0300)
>
> ----------------------------------------------------------------
> i386: Don't automatically enable FEAT_KVM_HINTS bits
>
> Bug fix for "-cpu host" with newer kernels.
>
> ----------------------------------------------------------------

Is this a regression since 2.11 ?

thanks
-- PMM

Re: [Qemu-devel] [PULL 0/1] i386: Don't automatically enable FEAT_KVM_HINTS bits
Posted by Eduardo Habkost, 3 days ago
On Mon, Apr 16, 2018 at 05:57:35PM +0100, Peter Maydell wrote:
> On 16 April 2018 at 17:39, Eduardo Habkost <ehabkost@redhat.com> wrote:
> > Last remaining fix for -rc4.
> >
> > The following changes since commit 042f6a31af3d38eefc6ec995cce1d762c41d4515:
> >
> >   Merge remote-tracking branch 'remotes/maxreitz/tags/pull-block-2018-04-16' into staging (2018-04-16 15:30:54 +0100)
> >
> > are available in the Git repository at:
> >
> >   git://github.com/ehabkost/qemu.git tags/x86-next-pull-request
> >
> > for you to fetch changes up to 0d914f39a77954bf5472b16130aeefe573a068f5:
> >
> >   i386: Don't automatically enable FEAT_KVM_HINTS bits (2018-04-16 13:36:52 -0300)
> >
> > ----------------------------------------------------------------
> > i386: Don't automatically enable FEAT_KVM_HINTS bits
> >
> > Bug fix for "-cpu host" with newer kernels.
> >
> > ----------------------------------------------------------------
> 
> Is this a regression since 2.11 ?

Yes.

-- 
Eduardo

Re: [Qemu-devel] [PULL 0/1] i386: Don't automatically enable FEAT_KVM_HINTS bits
Posted by Eduardo Habkost, 3 days ago
On Mon, Apr 16, 2018 at 02:42:52PM -0300, Eduardo Habkost wrote:
> On Mon, Apr 16, 2018 at 05:57:35PM +0100, Peter Maydell wrote:
> > On 16 April 2018 at 17:39, Eduardo Habkost <ehabkost@redhat.com> wrote:
> > > Last remaining fix for -rc4.
> > >
> > > The following changes since commit 042f6a31af3d38eefc6ec995cce1d762c41d4515:
> > >
> > >   Merge remote-tracking branch 'remotes/maxreitz/tags/pull-block-2018-04-16' into staging (2018-04-16 15:30:54 +0100)
> > >
> > > are available in the Git repository at:
> > >
> > >   git://github.com/ehabkost/qemu.git tags/x86-next-pull-request
> > >
> > > for you to fetch changes up to 0d914f39a77954bf5472b16130aeefe573a068f5:
> > >
> > >   i386: Don't automatically enable FEAT_KVM_HINTS bits (2018-04-16 13:36:52 -0300)
> > >
> > > ----------------------------------------------------------------
> > > i386: Don't automatically enable FEAT_KVM_HINTS bits
> > >
> > > Bug fix for "-cpu host" with newer kernels.
> > >
> > > ----------------------------------------------------------------
> > 
> > Is this a regression since 2.11 ?
> 
> Yes.

For reference, it was introduced by:

commit be7773268d98176489483a315d3e2323cb0615b9
Author: Wanpeng Li <wanpengli@tencent.com>
Date:   Fri Feb 9 06:15:25 2018 -0800

    target-i386: add KVM_HINTS_DEDICATED performance hint
    
    Add KVM_HINTS_DEDICATED performance hint, guest checks this feature bit
    to determine if they run on dedicated vCPUs, allowing optimizations such
    as usage of qspinlocks.

Sorry for not including a "Fixes:" tag in the commit message.  If
you think it's worth it, I can redo the pull request with it.

-- 
Eduardo

Re: [Qemu-devel] [PULL 0/1] i386: Don't automatically enable FEAT_KVM_HINTS bits
Posted by Peter Maydell, 3 days ago
On 16 April 2018 at 18:46, Eduardo Habkost <ehabkost@redhat.com> wrote:
> On Mon, Apr 16, 2018 at 02:42:52PM -0300, Eduardo Habkost wrote:
>> On Mon, Apr 16, 2018 at 05:57:35PM +0100, Peter Maydell wrote:
>> > On 16 April 2018 at 17:39, Eduardo Habkost <ehabkost@redhat.com> wrote:
>> > > Last remaining fix for -rc4.
>> > >
>> > > The following changes since commit 042f6a31af3d38eefc6ec995cce1d762c41d4515:
>> > >
>> > >   Merge remote-tracking branch 'remotes/maxreitz/tags/pull-block-2018-04-16' into staging (2018-04-16 15:30:54 +0100)
>> > >
>> > > are available in the Git repository at:
>> > >
>> > >   git://github.com/ehabkost/qemu.git tags/x86-next-pull-request
>> > >
>> > > for you to fetch changes up to 0d914f39a77954bf5472b16130aeefe573a068f5:
>> > >
>> > >   i386: Don't automatically enable FEAT_KVM_HINTS bits (2018-04-16 13:36:52 -0300)
>> > >
>> > > ----------------------------------------------------------------
>> > > i386: Don't automatically enable FEAT_KVM_HINTS bits
>> > >
>> > > Bug fix for "-cpu host" with newer kernels.
>> > >
>> > > ----------------------------------------------------------------
>> >
>> > Is this a regression since 2.11 ?
>>
>> Yes.
>
> For reference, it was introduced by:
>
> commit be7773268d98176489483a315d3e2323cb0615b9
> Author: Wanpeng Li <wanpengli@tencent.com>
> Date:   Fri Feb 9 06:15:25 2018 -0800
>
>     target-i386: add KVM_HINTS_DEDICATED performance hint
>
>     Add KVM_HINTS_DEDICATED performance hint, guest checks this feature bit
>     to determine if they run on dedicated vCPUs, allowing optimizations such
>     as usage of qspinlocks.
>
> Sorry for not including a "Fixes:" tag in the commit message.  If
> you think it's worth it, I can redo the pull request with it.

No, that's fine. I'm just checking that we're all on the same page
about the level of fix that goes into a late-stage rc. Particularly
as we get past rc2 or so, it's definitely helpful to me if pull
request cover letters include justifications for why the changes
should go in.

thanks
-- PMM

Re: [Qemu-devel] [PULL 0/1] i386: Don't automatically enable FEAT_KVM_HINTS bits
Posted by Peter Maydell, 2 days ago
On 16 April 2018 at 17:39, Eduardo Habkost <ehabkost@redhat.com> wrote:
> Last remaining fix for -rc4.
>
> The following changes since commit 042f6a31af3d38eefc6ec995cce1d762c41d4515:
>
>   Merge remote-tracking branch 'remotes/maxreitz/tags/pull-block-2018-04-16' into staging (2018-04-16 15:30:54 +0100)
>
> are available in the Git repository at:
>
>   git://github.com/ehabkost/qemu.git tags/x86-next-pull-request
>
> for you to fetch changes up to 0d914f39a77954bf5472b16130aeefe573a068f5:
>
>   i386: Don't automatically enable FEAT_KVM_HINTS bits (2018-04-16 13:36:52 -0300)
>
> ----------------------------------------------------------------
> i386: Don't automatically enable FEAT_KVM_HINTS bits
>
> Bug fix for "-cpu host" with newer kernels.
>
> ----------------------------------------------------------------
>
> Eduardo Habkost (1):
>   i386: Don't automatically enable FEAT_KVM_HINTS bits
>

Applied, thanks.

-- PMM

[Qemu-devel] [PULL 1/1] i386: Don't automatically enable FEAT_KVM_HINTS bits
Posted by Eduardo Habkost, 3 days ago
The assumption in the cpu->max_features code is that anything
enabled on GET_SUPPORTED_CPUID should be enabled on "-cpu host".
This shouldn't be the case for FEAT_KVM_HINTS.

This adds a new FeatureWordInfo::no_autoenable_flags field, that
can be used to prevent FEAT_KVM_HINTS bits to be enabled
automatically.

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Message-Id: <20180410211534.26079-1-ehabkost@redhat.com>
Tested-by: Wanpeng Li <wanpengli@tencent.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
---
 target/i386/cpu.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/target/i386/cpu.c b/target/i386/cpu.c
index 1a6b082b6f..a20fe26573 100644
--- a/target/i386/cpu.c
+++ b/target/i386/cpu.c
@@ -295,6 +295,8 @@ typedef struct FeatureWordInfo {
     uint32_t tcg_features; /* Feature flags supported by TCG */
     uint32_t unmigratable_flags; /* Feature flags known to be unmigratable */
     uint32_t migratable_flags; /* Feature flags known to be migratable */
+    /* Features that shouldn't be auto-enabled by "-cpu host" */
+    uint32_t no_autoenable_flags;
 } FeatureWordInfo;
 
 static FeatureWordInfo feature_word_info[FEATURE_WORDS] = {
@@ -400,6 +402,11 @@ static FeatureWordInfo feature_word_info[FEATURE_WORDS] = {
         },
         .cpuid_eax = KVM_CPUID_FEATURES, .cpuid_reg = R_EDX,
         .tcg_features = TCG_KVM_FEATURES,
+        /*
+         * KVM hints aren't auto-enabled by -cpu host, they need to be
+         * explicitly enabled in the command-line.
+         */
+        .no_autoenable_flags = ~0U,
     },
     [FEAT_HYPERV_EAX] = {
         .feat_names = {
@@ -4062,7 +4069,8 @@ static void x86_cpu_expand_features(X86CPU *cpu, Error **errp)
              */
             env->features[w] |=
                 x86_cpu_get_supported_feature_word(w, cpu->migratable) &
-                ~env->user_features[w];
+                ~env->user_features[w] & \
+                ~feature_word_info[w].no_autoenable_flags;
         }
     }
 
-- 
2.14.3