1. Patchew
  2. linux
  3. [PATCH v4 0/8] x86/mce: Clean up some x86/mce code
  4. View patch

[PATCH v4 5/8] x86/mce: Convert family/model mixed checks to VFM-based checks

Qiuxu Zhuo posted 8 patches 1 week, 5 days ago
[PATCH v4 5/8] x86/mce: Convert family/model mixed checks to VFM-based checks
Posted by Qiuxu Zhuo 1 week, 5 days ago 
Convert family/model mixed checks to VFM-based checks to make
the code more compact.

Suggested-by: Sohil Mehta <sohil.mehta@intel.com>
Suggested-by: Dave Hansen <dave.hansen@intel.com>
Reviewed-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
---
Changes in v4:
  - No changes but rebased.

Changes in v3:
  - Newly added.

 arch/x86/kernel/cpu/mce/core.c | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)

diff --git a/arch/x86/kernel/cpu/mce/core.c b/arch/x86/kernel/cpu/mce/core.c
index 3855ec2ed0e0..d288cc7390f6 100644
--- a/arch/x86/kernel/cpu/mce/core.c
+++ b/arch/x86/kernel/cpu/mce/core.c
@@ -1954,6 +1954,10 @@ static void apply_quirks_intel(struct cpuinfo_x86 *c)
 {
 	struct mce_bank *mce_banks = this_cpu_ptr(mce_banks_array);
 
+	/* Older CPUs (prior to family 6) don't need quirks. */
+	if (c->x86_vfm < INTEL_PENTIUM_PRO)
+		return;
+
 	/*
 	 * SDM documents that on family 6 bank 0 should not be written
 	 * because it aliases to another special BIOS controlled
@@ -1962,22 +1966,21 @@ static void apply_quirks_intel(struct cpuinfo_x86 *c)
 	 * Don't ignore bank 0 completely because there could be a
 	 * valid event later, merely don't write CTL0.
 	 */
-	if (c->x86 == 6 && c->x86_model < 0x1A && this_cpu_read(mce_num_banks) > 0)
+	if (c->x86_vfm < INTEL_NEHALEM_EP && this_cpu_read(mce_num_banks) > 0)
 		mce_banks[0].init = false;
 
 	/*
 	 * All newer Intel systems support MCE broadcasting. Enable
 	 * synchronization with a one second timeout.
 	 */
-	if ((c->x86 > 6 || (c->x86 == 6 && c->x86_model >= 0xe)) &&
-	    mca_cfg.monarch_timeout < 0)
+	if (c->x86_vfm >= INTEL_CORE_YONAH && mca_cfg.monarch_timeout < 0)
 		mca_cfg.monarch_timeout = USEC_PER_SEC;
 
 	/*
 	 * There are also broken BIOSes on some Pentium M and
 	 * earlier systems:
 	 */
-	if (c->x86 == 6 && c->x86_model <= 13 && mca_cfg.bootlog < 0)
+	if (c->x86_vfm < INTEL_CORE_YONAH && mca_cfg.bootlog < 0)
 		mca_cfg.bootlog = 0;
 
 	if (c->x86_vfm == INTEL_SANDYBRIDGE_X)
-- 
2.17.1
Re: [PATCH v4 5/8] x86/mce: Convert family/model mixed checks to VFM-based checks
Posted by Yazen Ghannam 1 week, 4 days ago 
On Mon, Nov 11, 2024 at 02:04:25PM +0800, Qiuxu Zhuo wrote:
> Convert family/model mixed checks to VFM-based checks to make
> the code more compact.
> 
> Suggested-by: Sohil Mehta <sohil.mehta@intel.com>
> Suggested-by: Dave Hansen <dave.hansen@intel.com>
> Reviewed-by: Tony Luck <tony.luck@intel.com>
> Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
> ---
> Changes in v4:
>   - No changes but rebased.
> 
> Changes in v3:
>   - Newly added.
> 
>  arch/x86/kernel/cpu/mce/core.c | 11 +++++++----
>  1 file changed, 7 insertions(+), 4 deletions(-)
> 
> diff --git a/arch/x86/kernel/cpu/mce/core.c b/arch/x86/kernel/cpu/mce/core.c
> index 3855ec2ed0e0..d288cc7390f6 100644
> --- a/arch/x86/kernel/cpu/mce/core.c
> +++ b/arch/x86/kernel/cpu/mce/core.c
> @@ -1954,6 +1954,10 @@ static void apply_quirks_intel(struct cpuinfo_x86 *c)
>  {
>  	struct mce_bank *mce_banks = this_cpu_ptr(mce_banks_array);
>  
> +	/* Older CPUs (prior to family 6) don't need quirks. */
> +	if (c->x86_vfm < INTEL_PENTIUM_PRO)
> +		return;
> +

Is it possible for pre-"family 6" to get here?

Family 5 is "ancient" which has its own MCE init path. And I assume
anything older doesn't support MCE/MCA. Is this correct?

>  	/*
>  	 * SDM documents that on family 6 bank 0 should not be written
>  	 * because it aliases to another special BIOS controlled
> @@ -1962,22 +1966,21 @@ static void apply_quirks_intel(struct cpuinfo_x86 *c)
>  	 * Don't ignore bank 0 completely because there could be a
>  	 * valid event later, merely don't write CTL0.
>  	 */
> -	if (c->x86 == 6 && c->x86_model < 0x1A && this_cpu_read(mce_num_banks) > 0)
> +	if (c->x86_vfm < INTEL_NEHALEM_EP && this_cpu_read(mce_num_banks) > 0)

The "> 0" is not needed, since mce_num_banks is unsigned int. 

Otherwise, looks good.

Reviewed-by: Yazen Ghannam <yazen.ghannam@amd.com>

Thanks,
Yazen
RE: [PATCH v4 5/8] x86/mce: Convert family/model mixed checks to VFM-based checks
Posted by Zhuo, Qiuxu 1 week, 3 days ago 
Hi Yazen,

> From: Yazen Ghannam <yazen.ghannam@amd.com>
> [...]
> > @@ -1954,6 +1954,10 @@ static void apply_quirks_intel(struct
> > cpuinfo_x86 *c)  {
> >  	struct mce_bank *mce_banks = this_cpu_ptr(mce_banks_array);
> >
> > +	/* Older CPUs (prior to family 6) don't need quirks. */
> > +	if (c->x86_vfm < INTEL_PENTIUM_PRO)
> > +		return;
> > +
> 
> Is it possible for pre-"family 6" to get here?
> 
> Family 5 is "ancient" which has its own MCE init path. And I assume anything
> older doesn't support MCE/MCA. Is this correct?

Yes, there is an early return in __mcheck_cpu_ancient_init() for Family 5. 
However, this code explicitly indicates that "prior to families 6 don't need quirks"
and addresses concerns like:

   https://lore.kernel.org/all/dcfdba92-7004-413d-8011-12771636d11f@intel.com/

> >  	/*
> >  	 * SDM documents that on family 6 bank 0 should not be written
> >  	 * because it aliases to another special BIOS controlled @@ -1962,22
> > +1966,21 @@ static void apply_quirks_intel(struct cpuinfo_x86 *c)
> >  	 * Don't ignore bank 0 completely because there could be a
> >  	 * valid event later, merely don't write CTL0.
> >  	 */
> > -	if (c->x86 == 6 && c->x86_model < 0x1A &&
> this_cpu_read(mce_num_banks) > 0)
> > +	if (c->x86_vfm < INTEL_NEHALEM_EP &&
> this_cpu_read(mce_num_banks) >
> > +0)
> 
> The "> 0" is not needed, since mce_num_banks is unsigned int.

I don't get your point here. 
But it needs to check for the case where mce_num_banks == 0.

> Otherwise, looks good.
> 
> Reviewed-by: Yazen Ghannam <yazen.ghannam@amd.com>

Thanks!
-Qiuxu
Re: [PATCH v4 5/8] x86/mce: Convert family/model mixed checks to VFM-based checks
Posted by Yazen Ghannam 1 week, 3 days ago 
On Wed, Nov 13, 2024 at 12:10:31PM +0000, Zhuo, Qiuxu wrote:
> Hi Yazen,
> 
> > From: Yazen Ghannam <yazen.ghannam@amd.com>
> > [...]
> > > @@ -1954,6 +1954,10 @@ static void apply_quirks_intel(struct
> > > cpuinfo_x86 *c)  {
> > >  	struct mce_bank *mce_banks = this_cpu_ptr(mce_banks_array);
> > >
> > > +	/* Older CPUs (prior to family 6) don't need quirks. */
> > > +	if (c->x86_vfm < INTEL_PENTIUM_PRO)
> > > +		return;
> > > +
> > 
> > Is it possible for pre-"family 6" to get here?
> > 
> > Family 5 is "ancient" which has its own MCE init path. And I assume anything
> > older doesn't support MCE/MCA. Is this correct?
> 
> Yes, there is an early return in __mcheck_cpu_ancient_init() for Family 5. 
> However, this code explicitly indicates that "prior to families 6 don't need quirks"
> and addresses concerns like:
> 
>    https://lore.kernel.org/all/dcfdba92-7004-413d-8011-12771636d11f@intel.com/
>

Right, but my point is that this check would never be executed, since
the older systems would not get here during init. So this seems like
dead code.

> > >  	/*
> > >  	 * SDM documents that on family 6 bank 0 should not be written
> > >  	 * because it aliases to another special BIOS controlled @@ -1962,22
> > > +1966,21 @@ static void apply_quirks_intel(struct cpuinfo_x86 *c)
> > >  	 * Don't ignore bank 0 completely because there could be a
> > >  	 * valid event later, merely don't write CTL0.
> > >  	 */
> > > -	if (c->x86 == 6 && c->x86_model < 0x1A &&
> > this_cpu_read(mce_num_banks) > 0)
> > > +	if (c->x86_vfm < INTEL_NEHALEM_EP &&
> > this_cpu_read(mce_num_banks) >
> > > +0)
> > 
> > The "> 0" is not needed, since mce_num_banks is unsigned int.
> 
> I don't get your point here. 
> But it needs to check for the case where mce_num_banks == 0.
>

The check is "mce_num_banks > 0", and mce_num_banks is an unsigned int.
Therefore, the check is reduced to "mce_num_banks != 0". In this case,
you can just do "if (mce_num_banks)" to the same effect.

Thanks,
Yazen
RE: [PATCH v4 5/8] x86/mce: Convert family/model mixed checks to VFM-based checks
Posted by Zhuo, Qiuxu 1 week, 2 days ago 
Hi Yazen,

> From: Yazen Ghannam <yazen.ghannam@amd.com>
> [...]
> > > > +	if (c->x86_vfm < INTEL_NEHALEM_EP &&
> > > this_cpu_read(mce_num_banks) >
> > > > +0)
> > >
> > > The "> 0" is not needed, since mce_num_banks is unsigned int.
> >
> > I don't get your point here.
> > But it needs to check for the case where mce_num_banks == 0.
> >
> 
> The check is "mce_num_banks > 0", and mce_num_banks is an unsigned int.
> Therefore, the check is reduced to "mce_num_banks != 0". In this case, you
> can just do "if (mce_num_banks)" to the same effect.

I got you. OK, if nobody else objects, I'll update it in the next version.
[ Hope others won't blame this as over-optimization. ]

-Qiuxu
Re: [PATCH v4 5/8] x86/mce: Convert family/model mixed checks to VFM-based checks
Posted by Sohil Mehta 1 week, 4 days ago 
On 11/10/2024 10:04 PM, Qiuxu Zhuo wrote:
> Convert family/model mixed checks to VFM-based checks to make
> the code more compact.
> 
> Suggested-by: Sohil Mehta <sohil.mehta@intel.com>
> Suggested-by: Dave Hansen <dave.hansen@intel.com>
> Reviewed-by: Tony Luck <tony.luck@intel.com>
> Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
> ---

Reviewed-by: Sohil Mehta <sohil.mehta@intel.com>

Patchew 2.1-devel-b67e4c2

© 2016 - 2024 Red Hat, Inc.