| Message ID | 1469207641-20662-1-git-send-email-boris.ostrovsky@oracle.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Fri, Jul 22, 2016 at 01:14:01PM -0400, Boris Ostrovsky wrote: > Current code incorrectly adds 1 to full register instead of > incrementing the field in bits 15:12. > > Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com> Acked-by: Wei Liu <wei.liu2@citrix.com> I trust your expertise in this field. :-) > --- > tools/libxc/xc_cpuid_x86.c | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/tools/libxc/xc_cpuid_x86.c b/tools/libxc/xc_cpuid_x86.c > index 84f4e08..fbbac9e 100644 > --- a/tools/libxc/xc_cpuid_x86.c > +++ b/tools/libxc/xc_cpuid_x86.c > @@ -331,7 +331,8 @@ static void amd_xc_cpuid_policy(xc_interface *xch, > * ECX[15:12] is ApicIdCoreSize: ECX[7:0] is NumberOfCores (minus one). > * Update to reflect vLAPIC_ID = vCPU_ID * 2. > */ > - regs[2] = ((regs[2] & 0xf000u) + 1) | ((regs[2] & 0xffu) << 1) | 1u; > + regs[2] = ((regs[2] + (1u << 12)) & 0xf000u) | > + ((regs[2] & 0xffu) << 1) | 1u; > break; > > case 0x8000000a: { > -- > 1.8.3.1 >
On 07/22/2016 01:38 PM, Wei Liu wrote: > On Fri, Jul 22, 2016 at 01:14:01PM -0400, Boris Ostrovsky wrote: >> Current code incorrectly adds 1 to full register instead of >> incrementing the field in bits 15:12. >> >> Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com> > Acked-by: Wei Liu <wei.liu2@citrix.com> > > I trust your expertise in this field. :-) Just in this field? The field is only 4 bits! ;-) But this actually fixes a regression that was triggered by recent hvmloader change on one particular processor that we have in the test farm. -boris > >> --- >> tools/libxc/xc_cpuid_x86.c | 3 ++- >> 1 file changed, 2 insertions(+), 1 deletion(-) >> >> diff --git a/tools/libxc/xc_cpuid_x86.c b/tools/libxc/xc_cpuid_x86.c >> index 84f4e08..fbbac9e 100644 >> --- a/tools/libxc/xc_cpuid_x86.c >> +++ b/tools/libxc/xc_cpuid_x86.c >> @@ -331,7 +331,8 @@ static void amd_xc_cpuid_policy(xc_interface *xch, >> * ECX[15:12] is ApicIdCoreSize: ECX[7:0] is NumberOfCores (minus one). >> * Update to reflect vLAPIC_ID = vCPU_ID * 2. >> */ >> - regs[2] = ((regs[2] & 0xf000u) + 1) | ((regs[2] & 0xffu) << 1) | 1u; >> + regs[2] = ((regs[2] + (1u << 12)) & 0xf000u) | >> + ((regs[2] & 0xffu) << 1) | 1u; >> break; >> >> case 0x8000000a: { >> -- >> 1.8.3.1 >>
On Fri, Jul 22, 2016 at 01:45:05PM -0400, Boris Ostrovsky wrote: > On 07/22/2016 01:38 PM, Wei Liu wrote: > > On Fri, Jul 22, 2016 at 01:14:01PM -0400, Boris Ostrovsky wrote: > >> Current code incorrectly adds 1 to full register instead of > >> incrementing the field in bits 15:12. > >> > >> Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com> > > Acked-by: Wei Liu <wei.liu2@citrix.com> > > > > I trust your expertise in this field. :-) > > > Just in this field? The field is only 4 bits! ;-) > No, it's only one bit! :-P > But this actually fixes a regression that was triggered by recent > hvmloader change on one particular processor that we have in the test farm. Ian, this is a backport candidate and needs to go back as far as possible. The bogus calculation was introduced in 2008 (!). Wei. > > -boris > > > > >> --- > >> tools/libxc/xc_cpuid_x86.c | 3 ++- > >> 1 file changed, 2 insertions(+), 1 deletion(-) > >> > >> diff --git a/tools/libxc/xc_cpuid_x86.c b/tools/libxc/xc_cpuid_x86.c > >> index 84f4e08..fbbac9e 100644 > >> --- a/tools/libxc/xc_cpuid_x86.c > >> +++ b/tools/libxc/xc_cpuid_x86.c > >> @@ -331,7 +331,8 @@ static void amd_xc_cpuid_policy(xc_interface *xch, > >> * ECX[15:12] is ApicIdCoreSize: ECX[7:0] is NumberOfCores (minus one). > >> * Update to reflect vLAPIC_ID = vCPU_ID * 2. > >> */ > >> - regs[2] = ((regs[2] & 0xf000u) + 1) | ((regs[2] & 0xffu) << 1) | 1u; > >> + regs[2] = ((regs[2] + (1u << 12)) & 0xf000u) | > >> + ((regs[2] & 0xffu) << 1) | 1u; > >> break; > >> > >> case 0x8000000a: { > >> -- > >> 1.8.3.1 > >> >
Wei Liu writes ("Re: [PATCH] tools/libxc: Properly increment ApicIdCoreSize field on AMD"):
> On Fri, Jul 22, 2016 at 01:45:05PM -0400, Boris Ostrovsky wrote:
> > On 07/22/2016 01:38 PM, Wei Liu wrote:
> > But this actually fixes a regression that was triggered by recent
> > hvmloader change on one particular processor that we have in the test farm.
>
> Ian, this is a backport candidate and needs to go back as far as
> possible. The bogus calculation was introduced in 2008 (!).
Thanks, but:
Not queued for backport because the patch is not in staging.
Thanks,
Ian.
On Mon, Jul 25, 2016 at 12:33:37PM +0100, Ian Jackson wrote: > Wei Liu writes ("Re: [PATCH] tools/libxc: Properly increment ApicIdCoreSize field on AMD"): > > On Fri, Jul 22, 2016 at 01:45:05PM -0400, Boris Ostrovsky wrote: > > > On 07/22/2016 01:38 PM, Wei Liu wrote: > > > But this actually fixes a regression that was triggered by recent > > > hvmloader change on one particular processor that we have in the test farm. > > > > Ian, this is a backport candidate and needs to go back as far as > > possible. The bogus calculation was introduced in 2008 (!). > > Thanks, but: > > Not queued for backport because the patch is not in staging. > Pushed to staging.
Wei Liu writes ("Re: [PATCH] tools/libxc: Properly increment ApicIdCoreSize field on AMD"):
> On Mon, Jul 25, 2016 at 12:33:37PM +0100, Ian Jackson wrote:
> > Wei Liu writes ("Re: [PATCH] tools/libxc: Properly increment ApicIdCoreSize field on AMD"):
> > > Ian, this is a backport candidate and needs to go back as far as
> > > possible. The bogus calculation was introduced in 2008 (!).
>
> Pushed to staging.
Thanks, queued.
Ian.
diff --git a/tools/libxc/xc_cpuid_x86.c b/tools/libxc/xc_cpuid_x86.c index 84f4e08..fbbac9e 100644 --- a/tools/libxc/xc_cpuid_x86.c +++ b/tools/libxc/xc_cpuid_x86.c @@ -331,7 +331,8 @@ static void amd_xc_cpuid_policy(xc_interface *xch, * ECX[15:12] is ApicIdCoreSize: ECX[7:0] is NumberOfCores (minus one). * Update to reflect vLAPIC_ID = vCPU_ID * 2. */ - regs[2] = ((regs[2] & 0xf000u) + 1) | ((regs[2] & 0xffu) << 1) | 1u; + regs[2] = ((regs[2] + (1u << 12)) & 0xf000u) | + ((regs[2] & 0xffu) << 1) | 1u; break; case 0x8000000a: {
Current code incorrectly adds 1 to full register instead of incrementing the field in bits 15:12. Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com> --- tools/libxc/xc_cpuid_x86.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)