Skip to contentSkip to conversation

x86/speculation: Disable IBRS when idle

PATCH Superseded Waiman Long 3 participants 8 messages Jul 10, 2023 – Jul 26, 2023

Subsystems (4)

v5 of 6 Superseded Latest is v6 →

lore
Share & tools
lore
#0 lore raw

v5 cover x86/speculation: Disable IBRS when idle

v5:
- Update comment in patch 1.
- Minor doc update and code twist in patch 4 as suggested by Peter and Randy.

v4:
- Add a new __update_spec_ctrl() helper in patch 1.
- Rebased to the latest linux kernel.

v3:
- Drop patches 1 ("x86/speculation: Provide a debugfs file to dump SPEC_CTRL MSRs") and 5 ("x86/idle: Disable IBRS entering mwait idle and enable it on wakeup") for now.
- Drop the MSR restoration code in ("x86/idle: Disable IBRS when cpu is offline") as native_play_dead() does not return.
- For patch ("intel_idle: Add ibrs_off module parameter to force disable IBRS"), change the name from "no_ibrs" to "ibrs_off" and document the new parameter in intel_idle.rst.

For Intel processors that need to turn on IBRS to protect against Spectre v2 and Retbleed, the IBRS bit in the SPEC_CTRL MSR affects the performance of the whole core even if only one thread is turning it on when running in the kernel. For user space heavy applications, the performance impact of occasionally turning IBRS on during syscalls shouldn't be significant. Unfortunately, that is not the case when the sibling thread is idling in the kernel. In that case, the performance impact can be significant.

When DPDK is running on an isolated CPU thread processing network packets in user space while its sibling thread is idle. The performance of the busy DPDK thread with IBRS on and off in the sibling idle thread are:

IBRS on IBRS off
------- --------
packets/second: 7.8M 10.4M
avg tsc cycles/packet: 282.26 209.86

This is a 25% performance degradation. The test system is a Intel Xeon 4114 CPU @ 2.20GHz.

Commit bf5835bcdb96 ("intel_idle: Disable IBRS during long idle") disables IBRS when the CPU enters long idle (C6 or below). However, there are existing users out there who have set "intel_idle.max_cstate=1" to decrease latency. Those users won't be able to benefit from this commit. This patch series extends this commit by providing a new "intel_idle.ibrs_off" module parameter to force disable IBRS even when "intel_idle.max_cstate=1" at the expense of increased IRQ response latency. It also includes a commit to allow the disabling of IBRS when a CPU becomes offline.
Waiman Long (4):
x86/speculation: Add __update_spec_ctrl() helper
x86/idle: Disable IBRS when cpu is offline
intel_idle: Use __update_spec_ctrl() in intel_idle_ibrs()
intel_idle: Add ibrs_off module parameter to force disable IBRS

 Documentation/admin-guide/pm/intel_idle.rst | 17 ++++++++++++++++-
 arch/x86/include/asm/nospec-branch.h        | 12 +++++++++++-
 arch/x86/kernel/smpboot.c                   |  8 ++++++++
 drivers/idle/intel_idle.c                   | 15 ++++++++++++---
 4 files changed, 47 insertions(+), 5 deletions(-)
Signature · 1 line
2.31.1
#1 lore raw

PATCH v5 1/4 x86/speculation: Add __update_spec_ctrl() helper

Add a new __update_spec_ctrl() helper which is a variant of update_spec_ctrl() that can be used in a noinstr function.

---
 arch/x86/include/asm/nospec-branch.h | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)
arch/x86/include/asm/nospec-branch.h +11 -1
index 55388c9f7601..06ceacfd1fe2 100644
@@ -9,7 +9,7 @@
 
 #include <asm/alternative.h>
 #include <asm/cpufeatures.h>
-#include <asm/msr-index.h>
+#include <asm/msr.h>
 #include <asm/unwind_hints.h>
 #include <asm/percpu.h>
 #include <asm/current.h>
@@ -488,6 +488,16 @@ DECLARE_PER_CPU(u64, x86_spec_ctrl_current);
 extern void update_spec_ctrl_cond(u64 val);
 extern u64 spec_ctrl_current(void);
 
+/*
+ * This can be used in noinstr function & should only be called in bare
+ * metal context.
+ */
+static __always_inline void __update_spec_ctrl(u64 val)
+{
+	__this_cpu_write(x86_spec_ctrl_current, val);
+	native_wrmsrl(MSR_IA32_SPEC_CTRL, val);
+}
+
 /*
  * With retpoline, we must use IBRS to restrict branch prediction
  * before calling into firmware.
Signature · 1 line
2.31.1
#2 lore raw

PATCH v5 2/4 x86/idle: Disable IBRS when cpu is offline

In reply to Waiman Long

Commit bf5835bcdb96 ("intel_idle: Disable IBRS during long idle") disables IBRS when the CPU enters long idle. However, when a CPU becomes offline, the IBRS bit is still set when X86_FEATURE_KERNEL_IBRS is enabled. That will impact the performance of a sibling CPU. Mitigate this performance impact by clearing all the mitigation bits in SPEC_CTRL MSR when offline. When the CPU is online again, it will be re-initialized and so restoring the SPEC_CTRL value isn't needed.

Add a comment to say that native_play_dead() is a __noreturn function, but it can't be marked as such to avoid confusion about the missing MSR restoration code.

---
 arch/x86/kernel/smpboot.c | 8 ++++++++
 1 file changed, 8 insertions(+)
arch/x86/kernel/smpboot.c +8 -0
index e1aa2cd7734b..68e2e044ab8b 100644
@@ -87,6 +87,7 @@
 #include <asm/hw_irq.h>
 #include <asm/stackprotector.h>
 #include <asm/sev.h>
+#include <asm/nospec-branch.h>
 
 /* representing HT siblings of each logical CPU */
 DEFINE_PER_CPU_READ_MOSTLY(cpumask_var_t, cpu_sibling_map);
@@ -1743,8 +1744,15 @@ void __noreturn hlt_play_dead(void)
 		native_halt();
 }
 
+/*
+ * native_play_dead() is essentially a __noreturn function, but it can't
+ * be marked as such as the compiler may complain about it.
+ */
 void native_play_dead(void)
 {
+	if (cpu_feature_enabled(X86_FEATURE_KERNEL_IBRS))
+		__update_spec_ctrl(0);
+
 	play_dead_common();
 	tboot_shutdown(TB_SHUTDOWN_WFS);
 
Signature · 1 line
2.31.1
#3 lore raw

PATCH v5 3/4 intel_idle: Use __update_spec_ctrl() in intel_idle_ibrs()

In reply to Waiman Long

When intel_idle_ibrs() is called, it modifies the SPEC_CTRL MSR to 0 in order disable IBRS. However, the new MSR value isn't reflected in x86_spec_ctrl_current which is at odd with the other code that keep track of its state in that percpu variable. Use the new __update_spec_ctrl() to have the x86_spec_ctrl_current percpu value properly updated.

---
 drivers/idle/intel_idle.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/idle/intel_idle.c +2 -2
index b930036edbbe..c9479f089037 100644
@@ -182,12 +182,12 @@ static __cpuidle int intel_idle_ibrs(struct cpuidle_device *dev,
 	int ret;
 
 	if (smt_active)
-		native_wrmsrl(MSR_IA32_SPEC_CTRL, 0);
+		__update_spec_ctrl(0);
 
 	ret = __intel_idle(dev, drv, index);
 
 	if (smt_active)
-		native_wrmsrl(MSR_IA32_SPEC_CTRL, spec_ctrl);
+		__update_spec_ctrl(spec_ctrl);
 
 	return ret;
 }
Signature · 1 line
2.31.1
#4 lore raw

PATCH v5 4/4 intel_idle: Add ibrs_off module parameter to force disable IBRS

In reply to Waiman Long

Commit bf5835bcdb96 ("intel_idle: Disable IBRS during long idle") disables IBRS when the cstate is 6 or lower. However, there are some use cases where a customer may want to use max_cstate=1 to lower latency. Such use cases will suffer from the performance degradation caused by the enabling of IBRS in the sibling idle thread. Add a "ibrs_off" module parameter to force disable IBRS and the CPUIDLE_FLAG_IRQ_ENABLE flag if set.

In the case of a Skylake server with max_cstate=1, this new ibrs_off option will likely increase the IRQ response latency as IRQ will now be disabled.

When running SPECjbb2015 with cstates set to C1 on a Skylake system.

First test when the kernel is booted with: "intel_idle.ibrs_off"
max-jOPS = 117828, critical-jOPS = 66047

Then retest when the kernel is booted without the "intel_idle.ibrs_off" added.
max-jOPS = 116408, critical-jOPS = 58958

That means booting with "intel_idle.ibrs_off" improves performance by:
max-jOPS: 1.2%, which could be considered noise range.
critical-jOPS: 12%, which is definitely a solid improvement.

The admin-guide/pm/intel_idle.rst file is updated to add a description about the new "ibrs_off" module parameter.

---
 Documentation/admin-guide/pm/intel_idle.rst | 17 ++++++++++++++++-
 drivers/idle/intel_idle.c                   | 11 ++++++++++-
 2 files changed, 26 insertions(+), 2 deletions(-)
Documentation/admin-guide/pm/intel_idle.rst +16 -1
index b799a43da62e..39bd6ecce7de 100644
@@ -170,7 +170,7 @@ and ``idle=nomwait``.  If any of them is present in the kernel command line, the
 ``MWAIT`` instruction is not allowed to be used, so the initialization of
 ``intel_idle`` will fail.
 
-Apart from that there are four module parameters recognized by ``intel_idle``
+Apart from that there are five module parameters recognized by ``intel_idle``
 itself that can be set via the kernel command line (they cannot be updated via
 sysfs, so that is the only way to change their values).
 
@@ -216,6 +216,21 @@ are ignored).
 The idle states disabled this way can be enabled (on a per-CPU basis) from user
 space via ``sysfs``.
 
+The ``ibrs_off`` module parameter is a boolean flag (defaults to
+false). If set, it is used to control if IBRS (Indirect Branch Restricted
+Speculation) should be turned off when the CPU enters an idle state.
+This flag does not affect CPUs that use Enhanced IBRS which can remain
+on with little performance impact.
+
+For some CPUs, IBRS will be selected as mitigation for Spectre v2 and Retbleed
+security vulnerabilities by default.  Leaving the IBRS mode on while idling may
+have a performance impact on its sibling CPU.  The IBRS mode will be turned off
+by default when the CPU enters into a deep idle state, but not in some
+shallower ones.  Setting the ``ibrs_off`` module parameter will force the IBRS
+mode to off when the CPU is in any one of the available idle states.  This may
+help performance of a sibling CPU at the expense of a slightly higher wakeup
+latency for the idle CPU.
+
 
 .. _intel-idle-core-and-package-idle-states:
 
drivers/idle/intel_idle.c +10 -1
index c9479f089037..e1b826344682 100644
@@ -69,6 +69,7 @@ static int max_cstate = CPUIDLE_STATE_MAX - 1;
 static unsigned int disabled_states_mask __read_mostly;
 static unsigned int preferred_states_mask __read_mostly;
 static bool force_irq_on __read_mostly;
+static bool ibrs_off __read_mostly;
 
 static struct cpuidle_device __percpu *intel_idle_cpuidle_devices;
 
@@ -1919,11 +1920,13 @@ static void state_update_enter_method(struct cpuidle_state *state, int cstate)
 	}
 
 	if (cpu_feature_enabled(X86_FEATURE_KERNEL_IBRS) &&
-			   state->flags & CPUIDLE_FLAG_IBRS) {
+			((state->flags & CPUIDLE_FLAG_IBRS) || ibrs_off)) {
 		/*
 		 * IBRS mitigation requires that C-states are entered
 		 * with interrupts disabled.
 		 */
+		if (ibrs_off && (state->flags & CPUIDLE_FLAG_IRQ_ENABLE))
+			state->flags &= ~CPUIDLE_FLAG_IRQ_ENABLE;
 		WARN_ON_ONCE(state->flags & CPUIDLE_FLAG_IRQ_ENABLE);
 		state->enter = intel_idle_ibrs;
 		return;
@@ -2346,3 +2349,9 @@ MODULE_PARM_DESC(preferred_cstates, "Mask of preferred idle states");
  * 'CPUIDLE_FLAG_INIT_XSTATE' and 'CPUIDLE_FLAG_IBRS' flags.
  */
 module_param(force_irq_on, bool, 0444);
+/*
+ * Force the disabling of IBRS when X86_FEATURE_KERNEL_IBRS is on and
+ * CPUIDLE_FLAG_IRQ_ENABLE isn't set.
+ */
+module_param(ibrs_off, bool, 0444);
+MODULE_PARM_DESC(ibrs_off, "Disable IBRS when idle");
Signature · 1 line
2.31.1
#5 lore raw

Re: [PATCH v5 4/4] intel_idle: Add ibrs_off module parameter to force disable IBRS

Quoted 28 lines · Waiman Long

Commit bf5835bcdb96 ("intel_idle: Disable IBRS during long idle") disables IBRS when the cstate is 6 or lower. However, there are some use cases where a customer may want to use max_cstate=1 to lower latency. Such use cases will suffer from the performance degradation caused by the enabling of IBRS in the sibling idle thread. Add a "ibrs_off" module parameter to force disable IBRS and the CPUIDLE_FLAG_IRQ_ENABLE flag if set.

In the case of a Skylake server with max_cstate=1, this new ibrs_off option will likely increase the IRQ response latency as IRQ will now be disabled.

When running SPECjbb2015 with cstates set to C1 on a Skylake system.

First test when the kernel is booted with: "intel_idle.ibrs_off"
max-jOPS = 117828, critical-jOPS = 66047

Then retest when the kernel is booted without the "intel_idle.ibrs_off" added.
max-jOPS = 116408, critical-jOPS = 58958

That means booting with "intel_idle.ibrs_off" improves performance by:
max-jOPS: 1.2%, which could be considered noise range.
critical-jOPS: 12%, which is definitely a solid improvement.

The admin-guide/pm/intel_idle.rst file is updated to add a description about the new "ibrs_off" module parameter.

Quoted 6 lines
---
  Documentation/admin-guide/pm/intel_idle.rst | 17 ++++++++++++++++-
  drivers/idle/intel_idle.c                   | 11 ++++++++++-
  2 files changed, 26 insertions(+), 2 deletions(-)

Ping! Is there further suggested changes for this patch series or is it good enough to be merged?

Thanks,
Longman

Quoted 73 lines
Documentation/admin-guide/pm/intel_idle.rst +16 -1
index b799a43da62e..39bd6ecce7de 100644
@@ -170,7 +170,7 @@ and ``idle=nomwait``.  If any of them is present in the kernel command line, the
  ``MWAIT`` instruction is not allowed to be used, so the initialization of
  ``intel_idle`` will fail.
 
-Apart from that there are four module parameters recognized by ``intel_idle``
+Apart from that there are five module parameters recognized by ``intel_idle``
  itself that can be set via the kernel command line (they cannot be updated via
  sysfs, so that is the only way to change their values).
 
@@ -216,6 +216,21 @@ are ignored).
  The idle states disabled this way can be enabled (on a per-CPU basis) from user
  space via ``sysfs``.
 
+The ``ibrs_off`` module parameter is a boolean flag (defaults to
+false). If set, it is used to control if IBRS (Indirect Branch Restricted
+Speculation) should be turned off when the CPU enters an idle state.
+This flag does not affect CPUs that use Enhanced IBRS which can remain
+on with little performance impact.
+
+For some CPUs, IBRS will be selected as mitigation for Spectre v2 and Retbleed
+security vulnerabilities by default.  Leaving the IBRS mode on while idling may
+have a performance impact on its sibling CPU.  The IBRS mode will be turned off
+by default when the CPU enters into a deep idle state, but not in some
+shallower ones.  Setting the ``ibrs_off`` module parameter will force the IBRS
+mode to off when the CPU is in any one of the available idle states.  This may
+help performance of a sibling CPU at the expense of a slightly higher wakeup
+latency for the idle CPU.
+
 
  .. _intel-idle-core-and-package-idle-states:
 
drivers/idle/intel_idle.c +10 -1
index c9479f089037..e1b826344682 100644
@@ -69,6 +69,7 @@ static int max_cstate = CPUIDLE_STATE_MAX - 1;
  static unsigned int disabled_states_mask __read_mostly;
  static unsigned int preferred_states_mask __read_mostly;
  static bool force_irq_on __read_mostly;
+static bool ibrs_off __read_mostly;
 
  static struct cpuidle_device __percpu *intel_idle_cpuidle_devices;
 
@@ -1919,11 +1920,13 @@ static void state_update_enter_method(struct cpuidle_state *state, int cstate)
  	}
 
  	if (cpu_feature_enabled(X86_FEATURE_KERNEL_IBRS) &&
-			   state->flags & CPUIDLE_FLAG_IBRS) {
+			((state->flags & CPUIDLE_FLAG_IBRS) || ibrs_off)) {
  		/*
  		 * IBRS mitigation requires that C-states are entered
  		 * with interrupts disabled.
  		 */
+		if (ibrs_off && (state->flags & CPUIDLE_FLAG_IRQ_ENABLE))
+			state->flags &= ~CPUIDLE_FLAG_IRQ_ENABLE;
  		WARN_ON_ONCE(state->flags & CPUIDLE_FLAG_IRQ_ENABLE);
  		state->enter = intel_idle_ibrs;
  		return;
@@ -2346,3 +2349,9 @@ MODULE_PARM_DESC(preferred_cstates, "Mask of preferred idle states");
   * 'CPUIDLE_FLAG_INIT_XSTATE' and 'CPUIDLE_FLAG_IBRS' flags.
   */
  module_param(force_irq_on, bool, 0444);
+/*
+ * Force the disabling of IBRS when X86_FEATURE_KERNEL_IBRS is on and
+ * CPUIDLE_FLAG_IRQ_ENABLE isn't set.
+ */
+module_param(ibrs_off, bool, 0444);
+MODULE_PARM_DESC(ibrs_off, "Disable IBRS when idle");
maintainer
#6 lore raw

Re: [PATCH v5 4/4] intel_idle: Add ibrs_off module parameter to force disable IBRS

Quoted 36 lines · Waiman Long
Quoted 28 lines · Waiman Long

Commit bf5835bcdb96 ("intel_idle: Disable IBRS during long idle") disables IBRS when the cstate is 6 or lower. However, there are some use cases where a customer may want to use max_cstate=1 to lower latency. Such use cases will suffer from the performance degradation caused by the enabling of IBRS in the sibling idle thread. Add a "ibrs_off" module parameter to force disable IBRS and the CPUIDLE_FLAG_IRQ_ENABLE flag if set.

In the case of a Skylake server with max_cstate=1, this new ibrs_off option will likely increase the IRQ response latency as IRQ will now be disabled.

When running SPECjbb2015 with cstates set to C1 on a Skylake system.

First test when the kernel is booted with: "intel_idle.ibrs_off"
max-jOPS = 117828, critical-jOPS = 66047

Then retest when the kernel is booted without the "intel_idle.ibrs_off" added.
max-jOPS = 116408, critical-jOPS = 58958

That means booting with "intel_idle.ibrs_off" improves performance by:
max-jOPS: 1.2%, which could be considered noise range.
critical-jOPS: 12%, which is definitely a solid improvement.

The admin-guide/pm/intel_idle.rst file is updated to add a description about the new "ibrs_off" module parameter.

Quoted 6 lines
---
  Documentation/admin-guide/pm/intel_idle.rst | 17 ++++++++++++++++-
  drivers/idle/intel_idle.c                   | 11 ++++++++++-
  2 files changed, 26 insertions(+), 2 deletions(-)
Quoted 2 lines

Ping! Is there further suggested changes for this patch series or is it good enough to be merged?

I suppose I'll go stick it in x86/core unless someone objects.

Tested-by on v5 4/4
#7 lore raw

Re: [PATCH v5 4/4] intel_idle: Add ibrs_off module parameter to force disable IBRS

In reply to Waiman Long

Waiman Long, Jul 10, 2023 at 21:48:

Quoted 25 lines

Commit bf5835bcdb96 ("intel_idle: Disable IBRS during long idle") disables IBRS when the cstate is 6 or lower. However, there are some use cases where a customer may want to use max_cstate=1 to lower latency. Such use cases will suffer from the performance degradation caused by the enabling of IBRS in the sibling idle thread. Add a "ibrs_off" module parameter to force disable IBRS and the CPUIDLE_FLAG_IRQ_ENABLE flag if set.

In the case of a Skylake server with max_cstate=1, this new ibrs_off option will likely increase the IRQ response latency as IRQ will now be disabled.

When running SPECjbb2015 with cstates set to C1 on a Skylake system.

First test when the kernel is booted with: "intel_idle.ibrs_off"
max-jOPS = 117828, critical-jOPS = 66047

Then retest when the kernel is booted without the "intel_idle.ibrs_off" added.
max-jOPS = 116408, critical-jOPS = 58958

That means booting with "intel_idle.ibrs_off" improves performance by:
max-jOPS: 1.2%, which could be considered noise range.
critical-jOPS: 12%, which is definitely a solid improvement.

Quoted 5 lines

The admin-guide/pm/intel_idle.rst file is updated to add a description about the new "ibrs_off" module parameter.

Hi Longman,

Adding back my tag that I had set in v1. Thanks!