Skip to contentSkip to conversation

x86/mwait-idle: (mostly) SPR support

PATCH Superseded Jan Beulich 2 participants 16 messages Apr 26, 2022 – Apr 28, 2022

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

cover x86/mwait-idle: (mostly) SPR support

Fresh from Linux 5.18-rc (and with adjustments to address issues noticed while porting), except for the 1st patch.

1: switch to asm/intel-family.h naming
2: add SPR support
3: add 'preferred_cstates' module argument
4: add core C6 optimization for SPR

Jan

#1 lore raw

PATCH 1/4 x86/mwait-idle: switch to asm/intel-family.h naming

This brings us (back) closer to the original Linux source.

While touching mwait_idle_state_table_update() also drop a stray leading blank.

xen/arch/x86/cpu/mwait-idle.c +48 -46
@@ -61,6 +61,7 @@
 #include <xen/trace.h>
 #include <asm/cpuidle.h>
 #include <asm/hpet.h>
+#include <asm/intel-family.h>
 #include <asm/mwait.h>
 #include <asm/msr.h>
 #include <asm/spec_ctrl.h>
@@ -996,48 +997,49 @@ static const struct idle_cpu idle_cpu_sn
 };
 
 #define ICPU(model, cpu) \
-	{ X86_VENDOR_INTEL, 6, model, X86_FEATURE_ALWAYS, &idle_cpu_##cpu}
+	{ X86_VENDOR_INTEL, 6, INTEL_FAM6_ ## model, X86_FEATURE_ALWAYS, \
+	  &idle_cpu_ ## cpu}
 
 static const struct x86_cpu_id intel_idle_ids[] __initconstrel = {
-	ICPU(0x1a, nehalem),
-	ICPU(0x1e, nehalem),
-	ICPU(0x1f, nehalem),
-	ICPU(0x25, nehalem),
-	ICPU(0x2c, nehalem),
-	ICPU(0x2e, nehalem),
-	ICPU(0x2f, nehalem),
-	ICPU(0x1c, atom),
-	ICPU(0x26, lincroft),
-	ICPU(0x2a, snb),
-	ICPU(0x2d, snb),
-	ICPU(0x36, atom),
-	ICPU(0x37, byt),
-	ICPU(0x4a, tangier),
-	ICPU(0x4c, cht),
-	ICPU(0x3a, ivb),
-	ICPU(0x3e, ivt),
-	ICPU(0x3c, hsw),
-	ICPU(0x3f, hsw),
-	ICPU(0x45, hsw),
-	ICPU(0x46, hsw),
-	ICPU(0x4d, avn),
-	ICPU(0x3d, bdw),
-	ICPU(0x47, bdw),
-	ICPU(0x4f, bdw),
-	ICPU(0x56, bdw),
-	ICPU(0x4e, skl),
-	ICPU(0x5e, skl),
-	ICPU(0x8e, skl),
-	ICPU(0x9e, skl),
-	ICPU(0x55, skx),
-	ICPU(0x6a, icx),
-	ICPU(0x6c, icx),
-	ICPU(0x57, knl),
-	ICPU(0x85, knl),
-	ICPU(0x5c, bxt),
-	ICPU(0x7a, bxt),
-	ICPU(0x5f, dnv),
-	ICPU(0x86, snr),
+	ICPU(NEHALEM_EP,		nehalem),
+	ICPU(NEHALEM,			nehalem),
+	ICPU(NEHALEM_G,			nehalem),
+	ICPU(WESTMERE,			nehalem),
+	ICPU(WESTMERE_EP,		nehalem),
+	ICPU(NEHALEM_EX,		nehalem),
+	ICPU(WESTMERE_EX,		nehalem),
+	ICPU(ATOM_BONNELL,		atom),
+	ICPU(ATOM_BONNELL_MID,		lincroft),
+	ICPU(SANDYBRIDGE,		snb),
+	ICPU(SANDYBRIDGE_X,		snb),
+	ICPU(ATOM_SALTWELL,		atom),
+	ICPU(ATOM_SILVERMONT,		byt),
+	ICPU(ATOM_SILVERMONT_MID,	tangier),
+	ICPU(ATOM_AIRMONT,		cht),
+	ICPU(IVYBRIDGE,			ivb),
+	ICPU(IVYBRIDGE_X,		ivt),
+	ICPU(HASWELL,			hsw),
+	ICPU(HASWELL_X,			hsw),
+	ICPU(HASWELL_L,			hsw),
+	ICPU(HASWELL_G,			hsw),
+	ICPU(ATOM_SILVERMONT_D,		avn),
+	ICPU(BROADWELL,			bdw),
+	ICPU(BROADWELL_G,		bdw),
+	ICPU(BROADWELL_X,		bdw),
+	ICPU(BROADWELL_D,		bdw),
+	ICPU(SKYLAKE_L,			skl),
+	ICPU(SKYLAKE,			skl),
+	ICPU(KABYLAKE_L,		skl),
+	ICPU(KABYLAKE,			skl),
+	ICPU(SKYLAKE_X,			skx),
+	ICPU(ICELAKE_X,			icx),
+	ICPU(ICELAKE_D,			icx),
+	ICPU(XEON_PHI_KNL,		knl),
+	ICPU(XEON_PHI_KNM,		knl),
+	ICPU(ATOM_GOLDMONT,		bxt),
+	ICPU(ATOM_GOLDMONT_PLUS,	bxt),
+	ICPU(ATOM_GOLDMONT_D,		dnv),
+	ICPU(ATOM_TREMONT_D,		snr),
 	{}
 };
 
@@ -1208,20 +1210,20 @@ static void __init skx_idle_state_table_
 static void __init mwait_idle_state_table_update(void)
 {
 	switch (boot_cpu_data.x86_model) {
-	case 0x3e: /* IVT */
+	case INTEL_FAM6_IVYBRIDGE_X:
 		ivt_idle_state_table_update();
 		break;
-	case 0x5c: /* BXT */
-	case 0x7a:
+	case INTEL_FAM6_ATOM_GOLDMONT:
+	case INTEL_FAM6_ATOM_GOLDMONT_PLUS:
 		bxt_idle_state_table_update();
 		break;
-	case 0x5e: /* SKL-H */
+	case INTEL_FAM6_SKYLAKE:
 		sklh_idle_state_table_update();
 		break;
-	case 0x55: /* SKL-X */
+	case INTEL_FAM6_SKYLAKE_X:
 		skx_idle_state_table_update();
 		break;
- 	}
+	}
 }
 
 static int __init mwait_idle_probe(void)
#3 lore raw

PATCH 2/4 mwait-idle: add SPR support

In reply to Jan Beulich

From: Artem Bityutskiy <[email protected]>

Add Sapphire Rapids Xeon support.

Up until very recently, the C1 and C1E C-states were independent, but this has changed in some new chips, including Sapphire Rapids Xeon (SPR). In these chips the C1 and C1E states cannot be enabled at the same time. The "C1E promotion" bit in 'MSR_IA32_POWER_CTL' also has its semantics changed a bit.

Here are the C1, C1E, and "C1E promotion" bit rules on Xeons before SPR.

1. If C1E promotion bit is disabled.
a. C1 requests end up with C1 C-state.
b. C1E requests end up with C1E C-state.
2. If C1E promotion bit is enabled.
a. C1 requests end up with C1E C-state.
b. C1E requests end up with C1E C-state.

Here are the C1, C1E, and "C1E promotion" bit rules on Sapphire Rapids Xeon.
1. If C1E promotion bit is disabled.
a. C1 requests end up with C1 C-state.
b. C1E requests end up with C1 C-state.
2. If C1E promotion bit is enabled.
a. C1 requests end up with C1E C-state.
b. C1E requests end up with C1E C-state.

Before SPR Xeon, the 'intel_idle' driver was disabling C1E promotion and was exposing C1 and C1E as independent C-states. But on SPR, C1 and C1E cannot be enabled at the same time.

This patch adds both C1 and C1E states. However, C1E is marked as with the "CPUIDLE_FLAG_UNUSABLE" flag, which means that in won't be registered by default. The C1E promotion bit will be cleared, which means that by default only C1 and C6 will be registered on SPR.

The next patch will add an option for enabling C1E and disabling C1 on SPR.

Origin: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git 9edf3c0ffef0

xen/arch/x86/cpu/mwait-idle.c +38 -0
@@ -586,6 +586,38 @@ static const struct cpuidle_state icx_cs
        {}
 };
 
+/*
+ * On Sapphire Rapids Xeon C1 has to be disabled if C1E is enabled, and vice
+ * versa. On SPR C1E is enabled only if "C1E promotion" bit is set in
+ * MSR_IA32_POWER_CTL. But in this case there effectively no C1, because C1
+ * requests are promoted to C1E. If the "C1E promotion" bit is cleared, then
+ * both C1 and C1E requests end up with C1, so there is effectively no C1E.
+ *
+ * By default we enable C1 and disable C1E by marking it with
+ * 'CPUIDLE_FLAG_DISABLED'.
+ */
+static struct cpuidle_state __read_mostly spr_cstates[] = {
+	{
+		.name = "C1",
+		.flags = MWAIT2flg(0x00),
+		.exit_latency = 1,
+		.target_residency = 1,
+	},
+	{
+		.name = "C1E",
+		.flags = MWAIT2flg(0x01) | CPUIDLE_FLAG_DISABLED,
+		.exit_latency = 2,
+		.target_residency = 4,
+	},
+	{
+		.name = "C6",
+		.flags = MWAIT2flg(0x20) | CPUIDLE_FLAG_TLB_FLUSHED,
+		.exit_latency = 290,
+		.target_residency = 800,
+	},
+	{}
+};
+
 static const struct cpuidle_state atom_cstates[] = {
 	{
 		.name = "C1E",
@@ -972,6 +1004,11 @@ static const struct idle_cpu idle_cpu_ic
        .disable_promotion_to_c1e = true,
 };
 
+static struct idle_cpu __read_mostly idle_cpu_spr = {
+	.state_table = spr_cstates,
+	.disable_promotion_to_c1e = true,
+};
+
 static const struct idle_cpu idle_cpu_avn = {
 	.state_table = avn_cstates,
 	.disable_promotion_to_c1e = true,
@@ -1034,6 +1071,7 @@ static const struct x86_cpu_id intel_idl
 	ICPU(SKYLAKE_X,			skx),
 	ICPU(ICELAKE_X,			icx),
 	ICPU(ICELAKE_D,			icx),
+	ICPU(SAPPHIRERAPIDS_X,		spr),
 	ICPU(XEON_PHI_KNL,		knl),
 	ICPU(XEON_PHI_KNM,		knl),
 	ICPU(ATOM_GOLDMONT,		bxt),
Acked-by on v1 2/4
#4 lore raw

Re: [PATCH 2/4] mwait-idle: add SPR support

Quoted 35 lines · Jan Beulich

From: Artem Bityutskiy <[email protected]>

Add Sapphire Rapids Xeon support.

Up until very recently, the C1 and C1E C-states were independent, but this has changed in some new chips, including Sapphire Rapids Xeon (SPR). In these chips the C1 and C1E states cannot be enabled at the same time. The "C1E promotion" bit in 'MSR_IA32_POWER_CTL' also has its semantics changed a bit.

Here are the C1, C1E, and "C1E promotion" bit rules on Xeons before SPR.

1. If C1E promotion bit is disabled.
a. C1 requests end up with C1 C-state.
b. C1E requests end up with C1E C-state.
2. If C1E promotion bit is enabled.
a. C1 requests end up with C1E C-state.
b. C1E requests end up with C1E C-state.

Here are the C1, C1E, and "C1E promotion" bit rules on Sapphire Rapids Xeon.
1. If C1E promotion bit is disabled.
a. C1 requests end up with C1 C-state.
b. C1E requests end up with C1 C-state.
2. If C1E promotion bit is enabled.
a. C1 requests end up with C1E C-state.
b. C1E requests end up with C1E C-state.

Before SPR Xeon, the 'intel_idle' driver was disabling C1E promotion and was exposing C1 and C1E as independent C-states. But on SPR, C1 and C1E cannot be enabled at the same time.

This patch adds both C1 and C1E states. However, C1E is marked as with the "CPUIDLE_FLAG_UNUSABLE" flag, which means that in won't be registered by default. The C1E promotion bit will be cleared, which means that by default only C1 and C6 will be registered on SPR.

Quoted 6 lines

The next patch will add an option for enabling C1E and disabling C1 on SPR.

Origin: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git 9edf3c0ffef0

Thanks, Roger.

#5 lore raw

PATCH 3/4 mwait-idle: add 'preferred_cstates' module argument

In reply to Jan Beulich

From: Artem Bityutskiy <[email protected]>

On Sapphire Rapids Xeon (SPR) the C1 and C1E states are basically mutually exclusive - only one of them can be enabled. By default, 'intel_idle' driver enables C1 and disables C1E. However, some users prefer to use C1E instead of C1, because it saves more energy.

This patch adds a new module parameter ('preferred_cstates') for enabling C1E and disabling C1. Here is the idea behind it.

1. This option has effect only for "mutually exclusive" C-states like C1 and C1E on SPR.
2. It does not have any effect on independent C-states, which do not require other C-states to be disabled (most states on most platforms as of today).
3. For mutually exclusive C-states, the 'intel_idle' driver always has a reasonable default, such as enabling C1 on SPR by default. On other platforms, the default may be different.
4. Users can override the default using the 'preferred_cstates' parameter.
5. The parameter accepts the preferred C-states bit-mask, similarly to the existing 'states_off' parameter.
6. This parameter is not limited to C1/C1E, and leaves room for supporting other mutually exclusive C-states, if they come in the future.

Today 'intel_idle' can only be compiled-in, which means that on SPR, in order to disable C1 and enable C1E, users should boot with the following kernel
argument: intel_idle.preferred_cstates=4

Origin: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git da0e58c038e6

Enable C1E (if requested) not only on the BSP's socket / package.

docs/misc/xen-command-line.pandoc +6 -0
@@ -1884,6 +1884,12 @@ paging controls access to usermode addre
 ### ple_window (Intel)
 > `= <integer>`
 
+### preferred-cstates (x86)
+> `= <integer>`
+
+This is a mask of C-states which are to be use preferably.  This option is
+applicable only oh hardware were certain C-states are exlusive of one another.
+
 ### psr (Intel)
 > `= List of ( cmt:<boolean> | rmid_max:<integer> | cat:<boolean> | cos_max:<integer> | cdp:<boolean> )`
 
xen/arch/x86/cpu/mwait-idle.c +47 -0

83 lines not on this page · raw message

#6 lore raw

Re: [PATCH 3/4] mwait-idle: add 'preferred_cstates' module argument

Quoted 27 lines · Jan Beulich

From: Artem Bityutskiy <[email protected]>

On Sapphire Rapids Xeon (SPR) the C1 and C1E states are basically mutually exclusive - only one of them can be enabled. By default, 'intel_idle' driver enables C1 and disables C1E. However, some users prefer to use C1E instead of C1, because it saves more energy.

This patch adds a new module parameter ('preferred_cstates') for enabling C1E and disabling C1. Here is the idea behind it.

1. This option has effect only for "mutually exclusive" C-states like C1 and C1E on SPR.
2. It does not have any effect on independent C-states, which do not require other C-states to be disabled (most states on most platforms as of today).
3. For mutually exclusive C-states, the 'intel_idle' driver always has a reasonable default, such as enabling C1 on SPR by default. On other platforms, the default may be different.
4. Users can override the default using the 'preferred_cstates' parameter.
5. The parameter accepts the preferred C-states bit-mask, similarly to the existing 'states_off' parameter.
6. This parameter is not limited to C1/C1E, and leaves room for supporting other mutually exclusive C-states, if they come in the future.

Today 'intel_idle' can only be compiled-in, which means that on SPR, in order to disable C1 and enable C1E, users should boot with the following kernel
argument: intel_idle.preferred_cstates=4

Quoted 5 lines

Origin: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git da0e58c038e6

Enable C1E (if requested) not only on the BSP's socket / package.

Maybe we should also add a note here that the command line option for Xen is preferred-cstates instead of intel_idle.preferred_cstates?

I think this is a bad interface however, we should have a more generic option (ie: cstate-mode = 'performance | powersave') so that users don't have to fiddle with model specific C state masks.

Quoted 40 lines
docs/misc/xen-command-line.pandoc +6 -0
@@ -1884,6 +1884,12 @@ paging controls access to usermode addre
 ### ple_window (Intel)
 > `= <integer>`
 
+### preferred-cstates (x86)
+> `= <integer>`
+
+This is a mask of C-states which are to be use preferably.  This option is
+applicable only oh hardware were certain C-states are exlusive of one another.
+
 ### psr (Intel)
 > `= List of ( cmt:<boolean> | rmid_max:<integer> | cat:<boolean> | cos_max:<integer> | cdp:<boolean> )`
 
xen/arch/x86/cpu/mwait-idle.c +12 -0
@@ -82,6 +82,18 @@ boolean_param("mwait-idle", opt_mwait_id
 
 static unsigned int mwait_substates;
 
+/*
+ * Some platforms come with mutually exclusive C-states, so that if one is
+ * enabled, the other C-states must not be used. Example: C1 and C1E on
+ * Sapphire Rapids platform. This parameter allows for selecting the
+ * preferred C-states among the groups of mutually exclusive C-states - the
+ * selected C-states will be registered, the other C-states from the mutually
+ * exclusive group won't be registered. If the platform has no mutually
+ * exclusive C-states, this parameter has no effect.
+ */
+static unsigned int __ro_after_init preferred_states_mask;
+integer_param("preferred-cstates", preferred_states_mask);
+
 #define LAPIC_TIMER_ALWAYS_RELIABLE 0xFFFFFFFF
 /* Reliable LAPIC Timer States, bit 1 for C1 etc. Default to only C1. */
 static unsigned int lapic_timer_reliable_states = (1 << 1);

xen/arch/x86/cpu/mwait-idle.c struct idle_cpu { · line 111

Quoted 4 lines
xen/arch/x86/cpu/mwait-idle.c +1 -0
@@ -96,6 +108,7 @@ struct idle_cpu {
 	unsigned long auto_demotion_disable_flags;
 	bool byt_auto_demotion_disable_flag;
 	bool disable_promotion_to_c1e;
+	bool enable_promotion_to_c1e;

I'm confused by those fields, shouldn't we just have:
promotion_to_c1e = true | false?

As one field is the negation of the other:
enable_promotion_to_c1e = !disable_promotion_to_c1e

I know this is code from Linux, but would like to understand why two fields are needed.

Thanks, Roger.

#7 lore raw

Re: [PATCH 3/4] mwait-idle: add 'preferred_cstates' module argument

Quoted 32 lines · Roger Pau Monné
Quoted 31 lines · Jan Beulich

From: Artem Bityutskiy <[email protected]>

On Sapphire Rapids Xeon (SPR) the C1 and C1E states are basically mutually exclusive - only one of them can be enabled. By default, 'intel_idle' driver enables C1 and disables C1E. However, some users prefer to use C1E instead of C1, because it saves more energy.

This patch adds a new module parameter ('preferred_cstates') for enabling C1E and disabling C1. Here is the idea behind it.

1. This option has effect only for "mutually exclusive" C-states like C1 and C1E on SPR.
2. It does not have any effect on independent C-states, which do not require other C-states to be disabled (most states on most platforms as of today).
3. For mutually exclusive C-states, the 'intel_idle' driver always has a reasonable default, such as enabling C1 on SPR by default. On other platforms, the default may be different.
4. Users can override the default using the 'preferred_cstates' parameter.
5. The parameter accepts the preferred C-states bit-mask, similarly to the existing 'states_off' parameter.
6. This parameter is not limited to C1/C1E, and leaves room for supporting other mutually exclusive C-states, if they come in the future.

Today 'intel_idle' can only be compiled-in, which means that on SPR, in order to disable C1 and enable C1E, users should boot with the following kernel
argument: intel_idle.preferred_cstates=4

Origin: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git da0e58c038e6

Quoted 8 lines
Quoted 1 line

Enable C1E (if requested) not only on the BSP's socket / package.

Maybe we should also add a note here that the command line option for Xen is preferred-cstates instead of intel_idle.preferred_cstates?

I think this is a bad interface however, we should have a more generic option (ie: cstate-mode = 'performance | powersave') so that users don't have to fiddle with model specific C state masks.

Performance vs powersave doesn't cover it imo, especially if down the road more states would appear which can be controlled this way. I don't think there's a way around providing _some_ way to control things one a per-state level. When porting this over, I too didn't like this interface very much, but I had no good replacement idea.

Quoted 30 lines
Quoted 22 lines
xen/arch/x86/cpu/mwait-idle.c +12 -0
@@ -82,6 +82,18 @@ boolean_param("mwait-idle", opt_mwait_id
 
 static unsigned int mwait_substates;
 
+/*
+ * Some platforms come with mutually exclusive C-states, so that if one is
+ * enabled, the other C-states must not be used. Example: C1 and C1E on
+ * Sapphire Rapids platform. This parameter allows for selecting the
+ * preferred C-states among the groups of mutually exclusive C-states - the
+ * selected C-states will be registered, the other C-states from the mutually
+ * exclusive group won't be registered. If the platform has no mutually
+ * exclusive C-states, this parameter has no effect.
+ */
+static unsigned int __ro_after_init preferred_states_mask;
+integer_param("preferred-cstates", preferred_states_mask);
+
 #define LAPIC_TIMER_ALWAYS_RELIABLE 0xFFFFFFFF
 /* Reliable LAPIC Timer States, bit 1 for C1 etc. Default to only C1. */
 static unsigned int lapic_timer_reliable_states = (1 << 1);

xen/arch/x86/cpu/mwait-idle.c struct idle_cpu { · line 111

Quoted 4 lines
xen/arch/x86/cpu/mwait-idle.c +1 -0
@@ -96,6 +108,7 @@ struct idle_cpu {
 	unsigned long auto_demotion_disable_flags;
 	bool byt_auto_demotion_disable_flag;
 	bool disable_promotion_to_c1e;
+	bool enable_promotion_to_c1e;

I'm confused by those fields, shouldn't we just have:
promotion_to_c1e = true | false?

Quoted 5 lines

As one field is the negation of the other:
enable_promotion_to_c1e = !disable_promotion_to_c1e

I know this is code from Linux, but would like to understand why two fields are needed.

This really is a tristate; Linux is now changing their global variable to an enum, but we don't have an equivalent of that global variable.

Jan

#8 lore raw

Re: [PATCH 3/4] mwait-idle: add 'preferred_cstates' module argument

Quoted 42 lines · Jan Beulich
Quoted 32 lines · Roger Pau Monné
Quoted 31 lines · Jan Beulich

From: Artem Bityutskiy <[email protected]>

On Sapphire Rapids Xeon (SPR) the C1 and C1E states are basically mutually exclusive - only one of them can be enabled. By default, 'intel_idle' driver enables C1 and disables C1E. However, some users prefer to use C1E instead of C1, because it saves more energy.

This patch adds a new module parameter ('preferred_cstates') for enabling C1E and disabling C1. Here is the idea behind it.

1. This option has effect only for "mutually exclusive" C-states like C1 and C1E on SPR.
2. It does not have any effect on independent C-states, which do not require other C-states to be disabled (most states on most platforms as of today).
3. For mutually exclusive C-states, the 'intel_idle' driver always has a reasonable default, such as enabling C1 on SPR by default. On other platforms, the default may be different.
4. Users can override the default using the 'preferred_cstates' parameter.
5. The parameter accepts the preferred C-states bit-mask, similarly to the existing 'states_off' parameter.
6. This parameter is not limited to C1/C1E, and leaves room for supporting other mutually exclusive C-states, if they come in the future.

Today 'intel_idle' can only be compiled-in, which means that on SPR, in order to disable C1 and enable C1E, users should boot with the following kernel
argument: intel_idle.preferred_cstates=4

Origin: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git da0e58c038e6

Quoted 8 lines
Quoted 1 line

Enable C1E (if requested) not only on the BSP's socket / package.

Maybe we should also add a note here that the command line option for Xen is preferred-cstates instead of intel_idle.preferred_cstates?

I think this is a bad interface however, we should have a more generic option (ie: cstate-mode = 'performance | powersave') so that users don't have to fiddle with model specific C state masks.

Quoted 5 lines

Performance vs powersave doesn't cover it imo, especially if down the road more states would appear which can be controlled this way. I don't think there's a way around providing _some_ way to control things one a per-state level. When porting this over, I too didn't like this interface very much, but I had no good replacement idea.

I think it's fine to have this more fine grained control of C states, but it doesn't seem practical from a user (or distro) PoV. But then I also wonder how much of a difference this will make regarding power consumption.

Quoted 30 lines
Quoted 30 lines
Quoted 22 lines
xen/arch/x86/cpu/mwait-idle.c +12 -0
@@ -82,6 +82,18 @@ boolean_param("mwait-idle", opt_mwait_id
 
 static unsigned int mwait_substates;
 
+/*
+ * Some platforms come with mutually exclusive C-states, so that if one is
+ * enabled, the other C-states must not be used. Example: C1 and C1E on
+ * Sapphire Rapids platform. This parameter allows for selecting the
+ * preferred C-states among the groups of mutually exclusive C-states - the
+ * selected C-states will be registered, the other C-states from the mutually
+ * exclusive group won't be registered. If the platform has no mutually
+ * exclusive C-states, this parameter has no effect.
+ */
+static unsigned int __ro_after_init preferred_states_mask;
+integer_param("preferred-cstates", preferred_states_mask);
+
 #define LAPIC_TIMER_ALWAYS_RELIABLE 0xFFFFFFFF
 /* Reliable LAPIC Timer States, bit 1 for C1 etc. Default to only C1. */
 static unsigned int lapic_timer_reliable_states = (1 << 1);

xen/arch/x86/cpu/mwait-idle.c struct idle_cpu { · line 111

Quoted 4 lines
xen/arch/x86/cpu/mwait-idle.c +1 -0
@@ -96,6 +108,7 @@ struct idle_cpu {
 	unsigned long auto_demotion_disable_flags;
 	bool byt_auto_demotion_disable_flag;
 	bool disable_promotion_to_c1e;
+	bool enable_promotion_to_c1e;

I'm confused by those fields, shouldn't we just have:
promotion_to_c1e = true | false?

Quoted 8 lines
Quoted 5 lines

As one field is the negation of the other:
enable_promotion_to_c1e = !disable_promotion_to_c1e

I know this is code from Linux, but would like to understand why two fields are needed.

This really is a tristate; Linux is now changing their global variable to an enum, but we don't have an equivalent of that global variable.

So it would be: leave default, disable C1E promotion, enable C1E promotion.

And Linux is leaving the {disable,enable}_promotion_to_c1e in idle_cpu?

I guess there's not much we can do unless we want to diverge from upstream.

Thanks, Roger.

#9 lore raw

Re: [PATCH 3/4] mwait-idle: add 'preferred_cstates' module argument

Quoted 39 lines · Roger Pau Monné
Quoted 38 lines · Jan Beulich
Quoted 36 lines · Roger Pau Monné
Quoted 22 lines · Jan Beulich
xen/arch/x86/cpu/mwait-idle.c +12 -0
@@ -82,6 +82,18 @@ boolean_param("mwait-idle", opt_mwait_id
 
 static unsigned int mwait_substates;
 
+/*
+ * Some platforms come with mutually exclusive C-states, so that if one is
+ * enabled, the other C-states must not be used. Example: C1 and C1E on
+ * Sapphire Rapids platform. This parameter allows for selecting the
+ * preferred C-states among the groups of mutually exclusive C-states - the
+ * selected C-states will be registered, the other C-states from the mutually
+ * exclusive group won't be registered. If the platform has no mutually
+ * exclusive C-states, this parameter has no effect.
+ */
+static unsigned int __ro_after_init preferred_states_mask;
+integer_param("preferred-cstates", preferred_states_mask);
+
 #define LAPIC_TIMER_ALWAYS_RELIABLE 0xFFFFFFFF
 /* Reliable LAPIC Timer States, bit 1 for C1 etc. Default to only C1. */
 static unsigned int lapic_timer_reliable_states = (1 << 1);

xen/arch/x86/cpu/mwait-idle.c struct idle_cpu { · line 111

Quoted 4 lines
xen/arch/x86/cpu/mwait-idle.c +1 -0
@@ -96,6 +108,7 @@ struct idle_cpu {
 	unsigned long auto_demotion_disable_flags;
 	bool byt_auto_demotion_disable_flag;
 	bool disable_promotion_to_c1e;
+	bool enable_promotion_to_c1e;

I'm confused by those fields, shouldn't we just have:
promotion_to_c1e = true | false?

As one field is the negation of the other:
enable_promotion_to_c1e = !disable_promotion_to_c1e

I know this is code from Linux, but would like to understand why two fields are needed.

Quoted 8 lines
Quoted 2 lines

This really is a tristate; Linux is now changing their global variable to an enum, but we don't have an equivalent of that global variable.

So it would be: leave default, disable C1E promotion, enable C1E promotion.

And Linux is leaving the {disable,enable}_promotion_to_c1e in idle_cpu?

Iirc they only have disable_promotion_to_c1e there (as a struct field) and keep it, but they convert the similarly named file-scope variable to a tristate.

Quoted 2 lines

I guess there's not much we can do unless we want to diverge from upstream.

We've diverged some from Linux here already - as said, for example we don't have their file-scope variable. I could convert our struct field to an enum, but that would be larger code churn for (I think) little gain.

Jan

#10 lore raw

Re: [PATCH 3/4] mwait-idle: add 'preferred_cstates' module argument

Quoted 53 lines · Jan Beulich
Quoted 39 lines · Roger Pau Monné
Quoted 38 lines · Jan Beulich
Quoted 36 lines · Roger Pau Monné
Quoted 22 lines · Jan Beulich
xen/arch/x86/cpu/mwait-idle.c +12 -0
@@ -82,6 +82,18 @@ boolean_param("mwait-idle", opt_mwait_id
 
 static unsigned int mwait_substates;
 
+/*
+ * Some platforms come with mutually exclusive C-states, so that if one is
+ * enabled, the other C-states must not be used. Example: C1 and C1E on
+ * Sapphire Rapids platform. This parameter allows for selecting the
+ * preferred C-states among the groups of mutually exclusive C-states - the
+ * selected C-states will be registered, the other C-states from the mutually
+ * exclusive group won't be registered. If the platform has no mutually
+ * exclusive C-states, this parameter has no effect.
+ */
+static unsigned int __ro_after_init preferred_states_mask;
+integer_param("preferred-cstates", preferred_states_mask);
+
 #define LAPIC_TIMER_ALWAYS_RELIABLE 0xFFFFFFFF
 /* Reliable LAPIC Timer States, bit 1 for C1 etc. Default to only C1. */
 static unsigned int lapic_timer_reliable_states = (1 << 1);

xen/arch/x86/cpu/mwait-idle.c struct idle_cpu { · line 111

Quoted 4 lines
xen/arch/x86/cpu/mwait-idle.c +1 -0
@@ -96,6 +108,7 @@ struct idle_cpu {
 	unsigned long auto_demotion_disable_flags;
 	bool byt_auto_demotion_disable_flag;
 	bool disable_promotion_to_c1e;
+	bool enable_promotion_to_c1e;

I'm confused by those fields, shouldn't we just have:
promotion_to_c1e = true | false?

As one field is the negation of the other:
enable_promotion_to_c1e = !disable_promotion_to_c1e

I know this is code from Linux, but would like to understand why two fields are needed.

Quoted 8 lines
Quoted 2 lines

This really is a tristate; Linux is now changing their global variable to an enum, but we don't have an equivalent of that global variable.

So it would be: leave default, disable C1E promotion, enable C1E promotion.

And Linux is leaving the {disable,enable}_promotion_to_c1e in idle_cpu?

Iirc they only have disable_promotion_to_c1e there (as a struct field) and keep it, but they convert the similarly named file-scope variable to a tristate.

Quoted 7 lines
Quoted 2 lines

I guess there's not much we can do unless we want to diverge from upstream.

We've diverged some from Linux here already - as said, for example we don't have their file-scope variable. I could convert our struct field to an enum, but that would be larger code churn for (I think) little gain.

Hm, OK, could gaining the file scope variable would make sense in order to reduce divergences? Or are the other roadblocks there?

I think this is ugly, but would make sense as long as it allows us to keep closer to upstream.

Thanks, Roger.

#11 lore raw

Re: [PATCH 3/4] mwait-idle: add 'preferred_cstates' module argument

Quoted 57 lines · Roger Pau Monné
Quoted 56 lines · Jan Beulich
Quoted 39 lines · Roger Pau Monné
Quoted 38 lines · Jan Beulich
Quoted 36 lines · Roger Pau Monné
Quoted 22 lines · Jan Beulich
xen/arch/x86/cpu/mwait-idle.c +12 -0
@@ -82,6 +82,18 @@ boolean_param("mwait-idle", opt_mwait_id
 
 static unsigned int mwait_substates;
 
+/*
+ * Some platforms come with mutually exclusive C-states, so that if one is
+ * enabled, the other C-states must not be used. Example: C1 and C1E on
+ * Sapphire Rapids platform. This parameter allows for selecting the
+ * preferred C-states among the groups of mutually exclusive C-states - the
+ * selected C-states will be registered, the other C-states from the mutually
+ * exclusive group won't be registered. If the platform has no mutually
+ * exclusive C-states, this parameter has no effect.
+ */
+static unsigned int __ro_after_init preferred_states_mask;
+integer_param("preferred-cstates", preferred_states_mask);
+
 #define LAPIC_TIMER_ALWAYS_RELIABLE 0xFFFFFFFF
 /* Reliable LAPIC Timer States, bit 1 for C1 etc. Default to only C1. */
 static unsigned int lapic_timer_reliable_states = (1 << 1);

xen/arch/x86/cpu/mwait-idle.c struct idle_cpu { · line 111

Quoted 4 lines
xen/arch/x86/cpu/mwait-idle.c +1 -0
@@ -96,6 +108,7 @@ struct idle_cpu {
 	unsigned long auto_demotion_disable_flags;
 	bool byt_auto_demotion_disable_flag;
 	bool disable_promotion_to_c1e;
+	bool enable_promotion_to_c1e;

I'm confused by those fields, shouldn't we just have:
promotion_to_c1e = true | false?

As one field is the negation of the other:
enable_promotion_to_c1e = !disable_promotion_to_c1e

I know this is code from Linux, but would like to understand why two fields are needed.

Quoted 8 lines
Quoted 2 lines

This really is a tristate; Linux is now changing their global variable to an enum, but we don't have an equivalent of that global variable.

So it would be: leave default, disable C1E promotion, enable C1E promotion.

And Linux is leaving the {disable,enable}_promotion_to_c1e in idle_cpu?

Iirc they only have disable_promotion_to_c1e there (as a struct field) and keep it, but they convert the similarly named file-scope variable to a tristate.

Quoted 2 lines

I guess there's not much we can do unless we want to diverge from upstream.

Quoted 7 lines
Quoted 4 lines

We've diverged some from Linux here already - as said, for example we don't have their file-scope variable. I could convert our struct field to an enum, but that would be larger code churn for (I think) little gain.

Hm, OK, could gaining the file scope variable would make sense in order to reduce divergences? Or are the other roadblocks there?

I don't recall. It might have originated from a change I decided to not port over, or I might have dropped it while porting. To be honest I'm not keen on putting time into researching this, the more that I would generally try to avoid static variables.

What I would be willing to put time in is making a more user friendly command line option, but as said - I can't think of any good alternative (except perhaps "preferred-cstates=c1e" or "cstates=preferred:c1e", with internal translation of the strings into a bit mask, as long as (a) you would think that's an improvement and (b) the further divergence from Linux is not deemed a problem).

Jan

Quoted 5 lines

I think this is ugly, but would make sense as long as it allows us to keep closer to upstream.

Thanks, Roger.

#12 lore raw

Re: [PATCH 3/4] mwait-idle: add 'preferred_cstates' module argument

Quoted 71 lines · Jan Beulich
Quoted 57 lines · Roger Pau Monné
Quoted 56 lines · Jan Beulich
Quoted 39 lines · Roger Pau Monné
Quoted 38 lines · Jan Beulich
Quoted 36 lines · Roger Pau Monné
Quoted 22 lines · Jan Beulich
xen/arch/x86/cpu/mwait-idle.c +12 -0
@@ -82,6 +82,18 @@ boolean_param("mwait-idle", opt_mwait_id
 
 static unsigned int mwait_substates;
 
+/*
+ * Some platforms come with mutually exclusive C-states, so that if one is
+ * enabled, the other C-states must not be used. Example: C1 and C1E on
+ * Sapphire Rapids platform. This parameter allows for selecting the
+ * preferred C-states among the groups of mutually exclusive C-states - the
+ * selected C-states will be registered, the other C-states from the mutually
+ * exclusive group won't be registered. If the platform has no mutually
+ * exclusive C-states, this parameter has no effect.
+ */
+static unsigned int __ro_after_init preferred_states_mask;
+integer_param("preferred-cstates", preferred_states_mask);
+
 #define LAPIC_TIMER_ALWAYS_RELIABLE 0xFFFFFFFF
 /* Reliable LAPIC Timer States, bit 1 for C1 etc. Default to only C1. */
 static unsigned int lapic_timer_reliable_states = (1 << 1);

xen/arch/x86/cpu/mwait-idle.c struct idle_cpu { · line 111

Quoted 4 lines
xen/arch/x86/cpu/mwait-idle.c +1 -0
@@ -96,6 +108,7 @@ struct idle_cpu {
 	unsigned long auto_demotion_disable_flags;
 	bool byt_auto_demotion_disable_flag;
 	bool disable_promotion_to_c1e;
+	bool enable_promotion_to_c1e;

I'm confused by those fields, shouldn't we just have:
promotion_to_c1e = true | false?

As one field is the negation of the other:
enable_promotion_to_c1e = !disable_promotion_to_c1e

I know this is code from Linux, but would like to understand why two fields are needed.

Quoted 8 lines
Quoted 2 lines

This really is a tristate; Linux is now changing their global variable to an enum, but we don't have an equivalent of that global variable.

So it would be: leave default, disable C1E promotion, enable C1E promotion.

And Linux is leaving the {disable,enable}_promotion_to_c1e in idle_cpu?

Iirc they only have disable_promotion_to_c1e there (as a struct field) and keep it, but they convert the similarly named file-scope variable to a tristate.

Quoted 2 lines

I guess there's not much we can do unless we want to diverge from upstream.

Quoted 7 lines
Quoted 4 lines

We've diverged some from Linux here already - as said, for example we don't have their file-scope variable. I could convert our struct field to an enum, but that would be larger code churn for (I think) little gain.

Hm, OK, could gaining the file scope variable would make sense in order to reduce divergences? Or are the other roadblocks there?

I don't recall. It might have originated from a change I decided to not port over, or I might have dropped it while porting. To be honest I'm not keen on putting time into researching this, the more that I would generally try to avoid static variables.

Quoted 6 lines

What I would be willing to put time in is making a more user friendly command line option, but as said - I can't think of any good alternative (except perhaps "preferred-cstates=c1e" or "cstates=preferred:c1e", with internal translation of the strings into a bit mask, as long as (a) you would think that's an improvement and (b) the further divergence from Linux is not deemed a problem).

I think (b) won't be a problem as long as internally the user option is translated into a bitmask.

Regarding (a) I do think it would be helpful to express this in a more user friendly way, I'm not sure whether it would make sense to keep Linux format also for compatibility reasons if users already have a bitmask and want to use the same parameter for Xen and Linux, ie:

preferred-cstates = <string of c1e,c1,...> | <integer bitmask>

What I think we should fix is the naming of the two booleans:

bool disable_promotion_to_c1e;
bool enable_promotion_to_c1e;

I would rather translated this into an enum, as right now it's confusing IMO.

Thanks, Roger.

#13 lore raw

Re: [PATCH 3/4] mwait-idle: add 'preferred_cstates' module argument

Quoted 82 lines · Roger Pau Monné
Quoted 71 lines · Jan Beulich
Quoted 57 lines · Roger Pau Monné
Quoted 56 lines · Jan Beulich
Quoted 39 lines · Roger Pau Monné
Quoted 38 lines · Jan Beulich
Quoted 36 lines · Roger Pau Monné
Quoted 22 lines · Jan Beulich
xen/arch/x86/cpu/mwait-idle.c +12 -0
@@ -82,6 +82,18 @@ boolean_param("mwait-idle", opt_mwait_id
 
 static unsigned int mwait_substates;
 
+/*
+ * Some platforms come with mutually exclusive C-states, so that if one is
+ * enabled, the other C-states must not be used. Example: C1 and C1E on
+ * Sapphire Rapids platform. This parameter allows for selecting the
+ * preferred C-states among the groups of mutually exclusive C-states - the
+ * selected C-states will be registered, the other C-states from the mutually
+ * exclusive group won't be registered. If the platform has no mutually
+ * exclusive C-states, this parameter has no effect.
+ */
+static unsigned int __ro_after_init preferred_states_mask;
+integer_param("preferred-cstates", preferred_states_mask);
+
 #define LAPIC_TIMER_ALWAYS_RELIABLE 0xFFFFFFFF
 /* Reliable LAPIC Timer States, bit 1 for C1 etc. Default to only C1. */
 static unsigned int lapic_timer_reliable_states = (1 << 1);

xen/arch/x86/cpu/mwait-idle.c struct idle_cpu { · line 111

Quoted 4 lines
xen/arch/x86/cpu/mwait-idle.c +1 -0
@@ -96,6 +108,7 @@ struct idle_cpu {
 	unsigned long auto_demotion_disable_flags;
 	bool byt_auto_demotion_disable_flag;
 	bool disable_promotion_to_c1e;
+	bool enable_promotion_to_c1e;

I'm confused by those fields, shouldn't we just have:
promotion_to_c1e = true | false?

As one field is the negation of the other:
enable_promotion_to_c1e = !disable_promotion_to_c1e

I know this is code from Linux, but would like to understand why two fields are needed.

Quoted 8 lines
Quoted 2 lines

This really is a tristate; Linux is now changing their global variable to an enum, but we don't have an equivalent of that global variable.

So it would be: leave default, disable C1E promotion, enable C1E promotion.

And Linux is leaving the {disable,enable}_promotion_to_c1e in idle_cpu?

Iirc they only have disable_promotion_to_c1e there (as a struct field) and keep it, but they convert the similarly named file-scope variable to a tristate.

Quoted 2 lines

I guess there's not much we can do unless we want to diverge from upstream.

Quoted 7 lines
Quoted 4 lines

We've diverged some from Linux here already - as said, for example we don't have their file-scope variable. I could convert our struct field to an enum, but that would be larger code churn for (I think) little gain.

Hm, OK, could gaining the file scope variable would make sense in order to reduce divergences? Or are the other roadblocks there?

I don't recall. It might have originated from a change I decided to not port over, or I might have dropped it while porting. To be honest I'm not keen on putting time into researching this, the more that I would generally try to avoid static variables.

Quoted 6 lines

What I would be willing to put time in is making a more user friendly command line option, but as said - I can't think of any good alternative (except perhaps "preferred-cstates=c1e" or "cstates=preferred:c1e", with internal translation of the strings into a bit mask, as long as (a) you would think that's an improvement and (b) the further divergence from Linux is not deemed a problem).

I think (b) won't be a problem as long as internally the user option is translated into a bitmask.

Quoted 6 lines

Regarding (a) I do think it would be helpful to express this in a more user friendly way, I'm not sure whether it would make sense to keep Linux format also for compatibility reasons if users already have a bitmask and want to use the same parameter for Xen and Linux, ie:

preferred-cstates = <string of c1e,c1,...> | <integer bitmask>

Yes, I would have been planning to accept both (but probably reject mixing of both).

Quoted 7 lines

What I think we should fix is the naming of the two booleans:

bool disable_promotion_to_c1e;
bool enable_promotion_to_c1e;

I would rather translated this into an enum, as right now it's confusing IMO.

Okay, I can certainly do that. The more that I did consider doing so already, breaking ties simply based on the "less code churn" argument.

Jan

#14 lore raw

PATCH 4/4 mwait-idle: add core C6 optimization for SPR

In reply to Jan Beulich

From: Artem Bityutskiy <[email protected]>

Add a Sapphire Rapids Xeon C6 optimization, similar to what we have for Sky Lake
Xeon: if package C6 is disabled, adjust C6 exit latency and target residency to match core C6 values, instead of using the default package C6 values.

Origin: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git 3a9cf77b60dc

Make sure a contradictory "preferred-cstates" wouldn't cause bypassing of the added logic.

xen/arch/x86/cpu/mwait-idle.c +18 -5
@@ -1267,12 +1267,12 @@ static void __init skx_idle_state_table_
  */
 static void __init spr_idle_state_table_update(void)
 {
-	/* Check if user prefers C1E over C1. */
-	if (preferred_states_mask & BIT(2, U)) {
-		if (preferred_states_mask & BIT(1, U))
-			/* Both can't be enabled, stick to the defaults. */
-			return;
+	uint64_t msr;
 
+	/* Check if user prefers C1E over C1. */
+	if (preferred_states_mask & BIT(2, U) &&
+	    /* Both can't be enabled, stick to the defaults. */
+	    !(preferred_states_mask & BIT(1, U))) {
 		spr_cstates[0].flags |= CPUIDLE_FLAG_DISABLED;
 		spr_cstates[1].flags &= ~CPUIDLE_FLAG_DISABLED;
 
@@ -1280,6 +1280,19 @@ static void __init spr_idle_state_table_
 		idle_cpu_spr.disable_promotion_to_c1e = false;
 		idle_cpu_spr.enable_promotion_to_c1e = true;
 	}
+
+	/*
+	 * By default, the C6 state assumes the worst-case scenario of package
+	 * C6. However, if PC6 is disabled, we update the numbers to match
+	 * core C6.
+	 */
+	rdmsrl(MSR_PKG_CST_CONFIG_CONTROL, msr);
+
+	/* Limit value 2 and above allow for PC6. */
+	if ((msr & 0x7) < 2) {
+		spr_cstates[2].exit_latency = 190;
+		spr_cstates[2].target_residency = 600;
+	}
 }
 
 /*
Acked-by on v1 4/4
#15 lore raw

Re: [PATCH 4/4] mwait-idle: add core C6 optimization for SPR

Quoted 6 lines · Jan Beulich

From: Artem Bityutskiy <[email protected]>

Add a Sapphire Rapids Xeon C6 optimization, similar to what we have for Sky Lake
Xeon: if package C6 is disabled, adjust C6 exit latency and target residency to match core C6 values, instead of using the default package C6 values.

Quoted 8 lines

Origin: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git 3a9cf77b60dc

Make sure a contradictory "preferred-cstates" wouldn't cause bypassing of the added logic.

Thanks, Roger.