Skip to contentSkip to conversation

x86/resctrl: fix min_cbm_bits for AMD

PATCH In review Stephane Eranian 2 participants 4 messages May 16, 2022

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 x86/resctrl: fix min_cbm_bits for AMD

AMD supports cbm with no bits set as reflected in rdt_init_res_defs_amd() by:

	r->cache.arch_has_empty_bitmaps = true;

However given the unified code in cbm_validate(), checking for

    val == 0 && !arch_has_empty_bitmaps

is not enough because you also have in cbm_validate():

	if ((zero_bit - first_bit) < r->cache.min_cbm_bits)

Default value for if r->cache.min_cbm_bits = 1

Leading to:

$ cd /sys/fs/resctrl
$ mkdir foo
$ cd foo
$ echo L3:0=0 > schemata
-bash: echo: write error: Invalid argument

Patch initializes r->cache.min_cbm_bits to 0 for AMD.

---
 arch/x86/kernel/cpu/resctrl/core.c | 1 +
 1 file changed, 1 insertion(+)
arch/x86/kernel/cpu/resctrl/core.c +1 -0
index bb1c3f5f60c8..14782361ebb7 100644
@@ -897,6 +897,7 @@ static __init void rdt_init_res_defs_amd(void)
 			r->cache.arch_has_sparse_bitmaps = true;
 			r->cache.arch_has_empty_bitmaps = true;
 			r->cache.arch_has_per_cpu_cfg = true;
+			r->cache.min_cbm_bits = 0;
 		} else if (r->rid == RDT_RESOURCE_MBA) {
 			hw_res->msr_base = MSR_IA32_MBA_BW_BASE;
 			hw_res->msr_update = mba_wrmsr_amd;
Signature · 1 line
2.36.0.550.gb090851708-goog
maintainer
#1 lore raw

+ x86 maintainers

Hi Stephane,

Thank you very much for catching this. While the fix is onto something I would prefer the fix to be obvious and not a side effect of bit checking in an empty bitmap.

When resubmitting, please ensure the subject starts with an uppercase letter.

Also, when resubmitting, could you please add [email protected]? The resctrl patches are routed upstream by the x86 architecture maintainers.

Quoted 1 line · Stephane Eranian

AMD supports cbm with no bits set as reflected in rdt_init_res_defs_amd() by:

Please watch the line lengths. Getting a pass from scripts/checkpatch.pl would help.

Quoted 3 lines
	r->cache.arch_has_empty_bitmaps = true;

With the above the needed information is present. Changing min_cbm_bits is not required.

Quoted 6 lines

However given the unified code in cbm_validate(), checking for

    val == 0 && !arch_has_empty_bitmaps

is not enough because you also have in cbm_validate():

	if ((zero_bit - first_bit) < r->cache.min_cbm_bits)

You are correct, but that relies on checking of bits in a bitmap where no bits are set so this fix relies on the failure cases of the earlier find_first_bit() and find_next_zero_bit() to be used. I find that it obscures the scenario being handled.

The code should be clear and to that end I think it would be simpler to explicitly check that an empty bitmap is provided and not search for bits at all when that is the case.

Something like this before the bit parsing starts:

	if (r->cache.arch_has_empty_bitmaps && val == 0)
		goto done;

	/* Skip bit parsing */

done:

	*data = val;
	return true;
Quoted 5 lines

Default value for if r->cache.min_cbm_bits = 1

Leading to:

Quoted 8 lines
$ cd /sys/fs/resctrl
$ mkdir foo
$ cd foo
$ echo L3:0=0 > schemata
-bash: echo: write error: Invalid argument

Patch initializes r->cache.min_cbm_bits to 0 for AMD.

Seems like a Fixes tag is needed.

Fixes 316e7f901f5a ("x86/resctrl: Add struct rdt_cache::arch_has_{sparse, empty}_bitmaps")
Quoted 10 lines
---
 arch/x86/kernel/cpu/resctrl/core.c | 1 +
 1 file changed, 1 insertion(+)
arch/x86/kernel/cpu/resctrl/core.c +0 -0
index bb1c3f5f60c8..14782361ebb7 100644

arch/x86/kernel/cpu/resctrl/core.c static __init void rdt_init_res_defs_amd(void) · line 903

Quoted 7 lines
arch/x86/kernel/cpu/resctrl/core.c +1 -0
@@ -897,6 +897,7 @@ static __init void rdt_init_res_defs_amd(void)
 			r->cache.arch_has_sparse_bitmaps = true;
 			r->cache.arch_has_empty_bitmaps = true;
 			r->cache.arch_has_per_cpu_cfg = true;
+			r->cache.min_cbm_bits = 0;
 		} else if (r->rid == RDT_RESOURCE_MBA) {
 			hw_res->msr_base = MSR_IA32_MBA_BW_BASE;
 			hw_res->msr_update = mba_wrmsr_amd;

Reinette

#2 lore raw
Quoted 9 lines · Reinette Chatre

+ x86 maintainers

Hi Stephane,

Thank you very much for catching this. While the fix is onto something I would prefer the fix to be obvious and not a side effect of bit checking in an empty bitmap.

Are you asking for me to add a comment on the modified line or are you asking for a change in cbm_validate()? There, I could add an empty_bitmask check to that if.

Quoted 2 lines

When resubmitting, please ensure the subject starts with an uppercase letter.

ok.

Quoted 3 lines

Also, when resubmitting, could you please add [email protected]? The resctrl patches are routed upstream by the x86 architecture maintainers.

ok.

Quoted 75 lines · Stephane Eranian
Quoted 1 line

AMD supports cbm with no bits set as reflected in rdt_init_res_defs_amd() by:

Please watch the line lengths. Getting a pass from scripts/checkpatch.pl would help.

Quoted 3 lines
      r->cache.arch_has_empty_bitmaps = true;

With the above the needed information is present. Changing min_cbm_bits is not required.

Quoted 6 lines

However given the unified code in cbm_validate(), checking for

    val == 0 && !arch_has_empty_bitmaps

is not enough because you also have in cbm_validate():

      if ((zero_bit - first_bit) < r->cache.min_cbm_bits)

You are correct, but that relies on checking of bits in a bitmap where no bits are set so this fix relies on the failure cases of the earlier find_first_bit() and find_next_zero_bit() to be used. I find that it obscures the scenario being handled.

The code should be clear and to that end I think it would be simpler to explicitly check that an empty bitmap is provided and not search for bits at all when that is the case.

Something like this before the bit parsing starts:

        if (r->cache.arch_has_empty_bitmaps && val == 0)
                goto done;

        /* Skip bit parsing */

done:

        *data = val;
        return true;
Quoted 5 lines

Default value for if r->cache.min_cbm_bits = 1

Leading to:

Quoted 8 lines
$ cd /sys/fs/resctrl
$ mkdir foo
$ cd foo
$ echo L3:0=0 > schemata
-bash: echo: write error: Invalid argument

Patch initializes r->cache.min_cbm_bits to 0 for AMD.

Seems like a Fixes tag is needed.

Fixes 316e7f901f5a ("x86/resctrl: Add struct rdt_cache::arch_has_{sparse, empty}_bitmaps")
Quoted 10 lines
---
 arch/x86/kernel/cpu/resctrl/core.c | 1 +
 1 file changed, 1 insertion(+)
arch/x86/kernel/cpu/resctrl/core.c +0 -0
index bb1c3f5f60c8..14782361ebb7 100644

arch/x86/kernel/cpu/resctrl/core.c static __init void rdt_init_res_defs_amd(void) · line 903

Quoted 7 lines
arch/x86/kernel/cpu/resctrl/core.c +1 -0
@@ -897,6 +897,7 @@ static __init void rdt_init_res_defs_amd(void)
                      r->cache.arch_has_sparse_bitmaps = true;
                      r->cache.arch_has_empty_bitmaps = true;
                      r->cache.arch_has_per_cpu_cfg = true;
+                     r->cache.min_cbm_bits = 0;
              } else if (r->rid == RDT_RESOURCE_MBA) {
                      hw_res->msr_base = MSR_IA32_MBA_BW_BASE;
                      hw_res->msr_update = mba_wrmsr_amd;

Reinette

maintainer
#3 lore raw

Hi Stephane,

Quoted 2 lines · Stephane Eranian

On Mon, May 16, 2022 at 9:35 AM Reinette Chatre
<[email protected]> wrote:

...

Quoted 7 lines
Quoted 4 lines

Thank you very much for catching this. While the fix is onto something I would prefer the fix to be obvious and not a side effect of bit checking in an empty bitmap.

Are you asking for me to add a comment on the modified line or are you asking for a change in cbm_validate()? There, I could add an empty_bitmask check to that if.

Please check my original response for inline comments to your patch.

I am asking for a change in cbm_validate() and my response did include some sample code copied below.

Quoted 10 lines
Quoted 10 lines

Something like this before the bit parsing starts:

        if (r->cache.arch_has_empty_bitmaps && val == 0)
                goto done;

        /* Skip bit parsing */

done:

        *data = val;
        return true;

Reinette