The kernel can change spinlock behavior when running as a guest. But
this guest-friendly behavior causes performance problems on bare metal.
So there's a 'virt_spin_lock_key' static key to switch between the two
modes.
In theory, the static key is enabled by default (run in guest mode) and
should be disabled for bare metal (and in some guests that want native
behavior or paravirt spinlock).
Performance drop is reported when running encode/decode workload and
BenchSEE cache sub-workload.
Bisect points to commit ce0a1b608bfc ("x86/paravirt: Silence unused
native_pv_lock_init() function warning"). When CONFIG_PARAVIRT_SPINLOCKS
is disabled the virt_spin_lock_key is incorrectly set to true on bare
metal. The qspinlock degenerates to test-and-set spinlock, which
decrease the performance on bare metal.
Set the default value of virt_spin_lock_key to false. If booting in a VM,
enable this key. Later during the VM initialization, if other
high-efficient spinlock is preferred(paravirt-spinlock eg), or the user
wants the native qspinlock(via nopvspin boot commandline), the
virt_spin_lock_key is disabled accordingly.
The relation is described as below:
X86_FEATURE_HYPERVISOR Y Y Y N
CONFIG_PARAVIRT_SPINLOCKS Y Y N Y/N
PV spinlock Y N N Y/N
virt_spin_lock_key N Y/N Y N
Fixes: ce0a1b608bfc ("x86/paravirt: Silence unused native_pv_lock_init() function warning")
Suggested-by: Dave Hansen <dave.hansen@xxxxxxxxxxxxxxx>
Suggested-by: Qiuxu Zhuo <qiuxu.zhuo@xxxxxxxxx>
Suggested-by: Nikolay Borisov <nik.borisov@xxxxxxxx>
Reported-by: Prem Nath Dey <prem.nath.dey@xxxxxxxxx>
Reported-by: Xiaoping Zhou <xiaoping.zhou@xxxxxxxxx>
Reviewed-by: Nikolay Borisov <nik.borisov@xxxxxxxx>
Signed-off-by: Chen Yu <yu.c.chen@xxxxxxxxx>
---
v4->v5:
Fix the comments around virt_spin_lock_key. (Thomas Gleixner)
v3->v4:
Refine the commit log.
Added Reviewed-by tag from Nikolay.
v2->v3:
Change the default value of virt_spin_lock_key from true to false.
Enable this key when it is in the VM, and disable it when needed.
This makes the code more readable. (Nikolay Borisov)
Dropped Reviewed-by because the code has been changed.
v1->v2:
Refine the commit log per Dave's suggestion.
Simplify the fix by directly disabling the virt_spin_lock_key on bare metal.
Collect Reviewed-by from Juergen.
---
arch/x86/include/asm/qspinlock.h | 8 ++++----
arch/x86/kernel/paravirt.c | 7 +++----
2 files changed, 7 insertions(+), 8 deletions(-)
diff --git a/arch/x86/include/asm/qspinlock.h b/arch/x86/include/asm/qspinlock.h
index a053c1293975..dd5b0ea4955f 100644
--- a/arch/x86/include/asm/qspinlock.h
+++ b/arch/x86/include/asm/qspinlock.h
@@ -66,13 +66,13 @@ static inline bool vcpu_is_preempted(long cpu)
#ifdef CONFIG_PARAVIRT
/*
- * virt_spin_lock_key - enables (by default) the virt_spin_lock() hijack.
+ * virt_spin_lock_key - disables (by default) the virt_spin_lock() hijack.
*
* Native (and PV wanting native due to vCPU pinning) should disable this key.
- * It is done in this backwards fashion to only have a single direction change,
- * which removes ordering between native_pv_spin_init() and HV setup.
+ * native_pv_lock_init() is invoked before the HV setup. The former enable
+ * the key by default on the VM guest and the latter disable it ondemand.
*/
-DECLARE_STATIC_KEY_TRUE(virt_spin_lock_key);
+DECLARE_STATIC_KEY_FALSE(virt_spin_lock_key);
/*
* Shortcut for the queued_spin_lock_slowpath() function that allows
diff --git a/arch/x86/kernel/paravirt.c b/arch/x86/kernel/paravirt.c
index 5358d43886ad..fec381533555 100644
--- a/arch/x86/kernel/paravirt.c
+++ b/arch/x86/kernel/paravirt.c
@@ -51,13 +51,12 @@ DEFINE_ASM_FUNC(pv_native_irq_enable, "sti", .noinstr.text);
DEFINE_ASM_FUNC(pv_native_read_cr2, "mov %cr2, %rax", .noinstr.text);
#endif
-DEFINE_STATIC_KEY_TRUE(virt_spin_lock_key);
+DEFINE_STATIC_KEY_FALSE(virt_spin_lock_key);
void __init native_pv_lock_init(void)
{
- if (IS_ENABLED(CONFIG_PARAVIRT_SPINLOCKS) &&
- !boot_cpu_has(X86_FEATURE_HYPERVISOR))
- static_branch_disable(&virt_spin_lock_key);
+ if (boot_cpu_has(X86_FEATURE_HYPERVISOR))
+ static_branch_enable(&virt_spin_lock_key);
}
static void native_tlb_remove_table(struct mmu_gather *tlb, void *table)
--
2.25.1
