Skip to contentSkip to conversation

locking/qspinlock: Enhance pvqspinlock performance

PATCH In review Waiman Long 2 participants 9 messages Aug 8, 2015 – Aug 14, 2015

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

v5 cover locking/qspinlock: Enhance pvqspinlock performance

v4->v5:
- Rebased the patch to the latest tip tree.
- Corrected the comments and commit log for patch 1.
- Removed the v4 patch 5 as PV kick deferment is no longer needed with the new tip tree.
- Simplified the adaptive spinning patch (patch 6) & improve its performance a bit further.
- Re-ran the benchmark test with the new patch.

v3->v4:
- Patch 1: add comment about possible racing condition in PV unlock.
- Patch 2: simplified the pv_pending_lock() function as suggested by Davidlohr.
- Move PV unlock optimization patch forward to patch 4 & rerun performance test.

v2->v3:
- Moved deferred kicking enablement patch forward & move back the kick-ahead patch to make the effect of kick-ahead more visible.
- Reworked patch 6 to make it more readable.
- Reverted back to use state as a tri-state variable instead of adding an additional bistate variable.
- Added performance data for different values of PV_KICK_AHEAD_MAX.
- Add a new patch to optimize PV unlock code path performance.

v1->v2:
- Take out the queued unfair lock patches
- Add a patch to simplify the PV unlock code
- Move pending bit and statistics collection patches to the front
- Keep vCPU kicking in pv_kick_node(), but defer it to unlock time when appropriate.
- Change the wait-early patch to use adaptive spinning to better balance the difference effect on normal and over-committed guests.
- Add patch-to-patch performance changes in the patch commit logs.

This patchset tries to improve the performance of both normal and over-commmitted VM guests. The kick-ahead and adaptive spinning patches are inspired by the "Do Virtual Machines Really Scale?" blog from Sanidhya Kashyap.

Patch 1 simplifies the unlock code by removing the unnecessary state check.

Patch 2 adds pending bit support to pvqspinlock improving performance at light load.

Patch 3 allows the collection of various count data that are useful to see what is happening in the system. They do add a bit of overhead when enabled slowing performance a tiny bit.

Patch 4 optimizes the PV unlock code path performance for x86-64 architecture.

Patch 5 enables multiple vCPU kick-ahead's at unlock time, outside of the critical section which can improve performance in overcommitted guests and sometime even in normal guests.

Patch 6 enables adaptive spinning in the queue nodes. This patch can lead to pretty big performance increase in over-committed guest at the expense of a slight performance hit in normal guests.

Patches 2 & 4 improves performance of common uncontended and lightly contended cases. Patches 5-6 are for improving performance in over-committed VM guests.

Performance measurements were done on a 32-CPU Westmere-EX and Haswell-EX systems. The Westmere-EX system got the most performance gain from patch 5, whereas the Haswell-EX system got the most gain from patch 6 for over-committed guests.

The table below shows the Linux kernel build times for various values of PV_KICK_AHEAD_MAX on an over-committed 48-vCPU guest on the Westmere-EX system:

PV_KICK_AHEAD_MAX Patches 1-5 Patches 1-6
----------------- ----------- -----------

	  1		  9m46.9s	 11m10.1s
	  2		  9m40.2s	 10m08.3s
	  3		  9m36.8s	  9m49.8s
	  4		  9m35.9s	  9m38.7s
	  5		  9m35.1s	  9m33.0s
	  6		  9m35.7s	  9m28.5s

With patches 1-5, the performance wasn't very sensitive to different PV_KICK_AHEAD_MAX values. Adding patch 6 into the mix, however, changes the picture quite dramatically. There is a performance regression if PV_KICK_AHEAD_MAX is too small. Starting with a value of 4, increasing PV_KICK_AHEAD_MAX only gets us a minor benefit.

Waiman Long (6):
locking/pvqspinlock: Unconditional PV kick with _Q_SLOW_VAL locking/pvqspinlock: Add pending bit support locking/pvqspinlock: Collect slowpath lock statistics locking/pvqspinlock, x86: Optimize PV unlock code path locking/pvqspinlock: Allow vCPUs kick-ahead locking/pvqspinlock: Queue node adaptive spinning

 arch/x86/Kconfig                          |    7 +
 arch/x86/include/asm/qspinlock_paravirt.h |   59 ++++
 kernel/locking/qspinlock.c                |   32 ++-
 kernel/locking/qspinlock_paravirt.h       |  475 +++++++++++++++++++++++++++--
 4 files changed, 542 insertions(+), 31 deletions(-)
#1 lore raw

PATCH v5 1/6 locking/pvqspinlock: Unconditional PV kick with _Q_SLOW_VAL

If _Q_SLOW_VAL has been set, the vCPU state must have been vcpu_hashed. The extra check at the end of __pv_queued_spin_unlock() is unnecessary and so is removed.

---
 kernel/locking/qspinlock_paravirt.h |    3 +--
 1 files changed, 1 insertions(+), 2 deletions(-)
kernel/locking/qspinlock_paravirt.h +1 -2
index c8e6e9a..6eafb9e 100644
@@ -364,8 +364,7 @@ __visible void __pv_queued_spin_unlock(struct qspinlock *lock)
 	 * vCPU is harmless other than the additional latency in completing
 	 * the unlock.
 	 */
-	if (READ_ONCE(node->state) == vcpu_hashed)
-		pv_kick(node->cpu);
+	pv_kick(node->cpu);
 }
 /*
  * Include the architecture specific callee-save thunk of the
Signature · 1 line
1.7.1
maintainer
#2 lore raw

Re: [PATCH v5 1/6] locking/pvqspinlock: Unconditional PV kick with _Q_SLOW_VAL

Quoted 3 lines · Waiman Long

If _Q_SLOW_VAL has been set, the vCPU state must have been vcpu_hashed. The extra check at the end of __pv_queued_spin_unlock() is unnecessary and so is removed.

This is half the patch it should be.

Because if the load is not needed, then the store is not either, and then there's the comments to update.

#3 lore raw

Re: [PATCH v5 1/6] locking/pvqspinlock: Unconditional PV kick with _Q_SLOW_VAL

Quoted 8 lines · Peter Zijlstra
Quoted 3 lines · Waiman Long

If _Q_SLOW_VAL has been set, the vCPU state must have been vcpu_hashed. The extra check at the end of __pv_queued_spin_unlock() is unnecessary and so is removed.

This is half the patch it should be.

Because if the load is not needed, then the store is not either, and then there's the comments to update.

Sorry for the late reply.

That is true. Setting state to vcpu_hashed in pv_wait_head() isn't really necessary. I kept it there for consistency sake as the state may be set to vcpu_hashed in pv_kick_node(). We can certainly take that out.

BTW, could you also review the other patches when you have time? I am coming to the LinuxCon/Plumbers next week. Hopefully, I can chat with you again.

Cheers,
Longman

#4 lore raw

PATCH v5 2/6 locking/pvqspinlock: Add pending bit support

In reply to Waiman Long

Like the native qspinlock, using the pending bit when it is lightly loaded to acquire the lock is faster than going through the PV queuing process which is even slower than the native queuing process. It also avoids loading two additional cachelines (the MCS and PV nodes).

This patch adds the pending bit support for PV qspinlock. The pending bit code has a smaller spin threshold (1<<10). It will default back to the queuing method if it cannot acquired the lock within a certain time limit.

On a VM with 32 vCPUs on a 32-core Westmere-EX box, the kernel build times on 4.2-rc1 based kernels were:

Kernel Build Time Sys Time
------ ---------- --------
w/o patch 3m28.5s 28m17.5s
with patch 3m19.3s 23m55.7s

Using a locking microbenchmark on the same system, the locking rates in (kops/s) were:

Threads Rate w/o patch Rate with patch
------- -------------- ---------------
2 (same socket) 6,515,265 7,077,476
2 (diff sockets) 2,967,145 4,353,851

---
 kernel/locking/qspinlock.c          |   27 ++++++++++++-
 kernel/locking/qspinlock_paravirt.h |   73 +++++++++++++++++++++++++++++++++++
 2 files changed, 99 insertions(+), 1 deletions(-)
kernel/locking/qspinlock.c +26 -1
index 337c881..94fdd27 100644
@@ -162,6 +162,17 @@ static __always_inline void clear_pending_set_locked(struct qspinlock *lock)
 	WRITE_ONCE(l->locked_pending, _Q_LOCKED_VAL);
 }
 
+/**
+ * clear_pending - clear the pending bit.
+ * @lock: Pointer to queued spinlock structure
+ */
+static __always_inline void clear_pending(struct qspinlock *lock)
+{
+	struct __qspinlock *l = (void *)lock;
+
+	WRITE_ONCE(l->pending, 0);
+}
+
 /*
  * xchg_tail - Put in the new queue tail code word & retrieve previous one
  * @lock : Pointer to queued spinlock structure
@@ -193,6 +204,15 @@ static __always_inline void clear_pending_set_locked(struct qspinlock *lock)
 }
 
 /**
+ * clear_pending - clear the pending bit.
+ * @lock: Pointer to queued spinlock structure
+ */
+static __always_inline void clear_pending(struct qspinlock *lock)
+{
+	atomic_add(-_Q_PENDING_VAL, &lock->val);
+}
+
+/**
  * xchg_tail - Put in the new queue tail code word & retrieve previous one
  * @lock : Pointer to queued spinlock structure
  * @tail : The new queue tail code word
@@ -245,6 +265,7 @@ static __always_inline void __pv_wait_head(struct qspinlock *lock,
 					   struct mcs_spinlock *node) { }
 
 #define pv_enabled()		false
+#define pv_pending_lock(l, v)	false
 
 #define pv_init_node		__pv_init_node
 #define pv_wait_node		__pv_wait_node
@@ -286,8 +307,11 @@ void queued_spin_lock_slowpath(struct qspinlock *lock, u32 val)
 
 	BUILD_BUG_ON(CONFIG_NR_CPUS >= (1U << _Q_TAIL_CPU_BITS));
 
-	if (pv_enabled())
+	if (pv_enabled()) {
+		if (pv_pending_lock(lock, val))
+			return;	/* Got the lock via pending bit */
 		goto queue;
+	}
 
 	if (virt_queued_spin_lock(lock))
 		return;
@@ -463,6 +487,7 @@ EXPORT_SYMBOL(queued_spin_lock_slowpath);
 #undef pv_wait_node
 #undef pv_kick_node
 #undef pv_wait_head
+#undef pv_pending_lock
 
 #undef  queued_spin_lock_slowpath
 #define queued_spin_lock_slowpath	__pv_queued_spin_lock_slowpath
kernel/locking/qspinlock_paravirt.h +73 -0

85 lines not on this page · raw message

Signature · 1 line
1.7.1
#5 lore raw

PATCH v5 3/6 locking/pvqspinlock: Collect slowpath lock statistics

In reply to Waiman Long

This patch enables the accumulation of kicking and waiting related PV qspinlock statistics when the new QUEUED_LOCK_STAT configuration option is selected. It also enables the collection of kicking and wakeup latencies which have a heavy dependency on the CPUs being used.

The measured latencies for different CPUs are:

	CPU		Wakeup		Kicking
	---		------		-------
	Haswell-EX	89.8us		 7.4us
	Westmere-EX	67.6us		 9.3us

The measured latencies varied a bit from run-to-run. The wakeup latency is much higher than the kicking latency.

A sample of statistics counts after a kernel build (no CPU overcommit) was:

hash_hops_count=576912
kick_latencies=5258025484
kick_unlock_count=576911
kick_wait_count=576903
pending_fail_count=10722
pending_lock_count=6123545
spurious_wakeup=92
wait_again_count=75
wait_head_count=60
wait_node_count=576936
wake_latencies=37061460652

---
 arch/x86/Kconfig                    |    7 ++
 kernel/locking/qspinlock_paravirt.h |  178 ++++++++++++++++++++++++++++++++++-
 2 files changed, 180 insertions(+), 5 deletions(-)
arch/x86/Kconfig +7 -0
index 4c9c8b8..86bf53e 100644
@@ -719,6 +719,13 @@ config PARAVIRT_SPINLOCKS
 
 	  If you are unsure how to answer this question, answer Y.
 
+config QUEUED_LOCK_STAT
+	bool "Paravirt queued lock statistics"
+	depends on PARAVIRT && DEBUG_FS && QUEUED_SPINLOCKS
+	---help---
+	  Enable the collection of statistical data on the behavior of
+	  paravirtualized queued spinlocks and report them on debugfs.
+
 source "arch/x86/xen/Kconfig"
 
 config KVM_GUEST
kernel/locking/qspinlock_paravirt.h +173 -5

269 lines not on this page · raw message

Signature · 1 line
1.7.1
#6 lore raw

PATCH v5 4/6 locking/pvqspinlock, x86: Optimize PV unlock code path

In reply to Waiman Long

The unlock function in queued spinlocks was optimized for better performance on bare metal systems at the expense of virtualized guests.

For x86-64 systems, the unlock call needs to go through a PV_CALLEE_SAVE_REGS_THUNK() which saves and restores 8 64-bit registers before calling the real __pv_queued_spin_unlock() function. The thunk code may also be in a separate cacheline from __pv_queued_spin_unlock().

This patch optimizes the PV unlock code path by:
1) Moving the unlock slowpath code from the fastpath into a separate __pv_queued_spin_unlock_slowpath() function to make the fastpath as simple as possible..
2) For x86-64, hand-coded an assembly function to combine the register saving thunk code with the fastpath code. Only registers that are used in the fastpath will be saved and restored. If the fastpath fails, the slowpath function will be called via another PV_CALLEE_SAVE_REGS_THUNK(). For 32-bit, it falls back to the C __pv_queued_spin_unlock() code as the thunk saves and restores only one 32-bit register.

With a microbenchmark of 5M lock-unlock loop, the table below shows the execution times before and after the patch with different number of threads in a VM running on a 32-core Westmere-EX box with x86-64 4.2-rc1 based kernels:

Threads Before patch After patch % Change
------- ------------ ----------- --------

     1		   134.1 ms	  119.3 ms	  -11%
     2		   1286  ms	   953  ms	  -26%
     3		   3715  ms	  3480  ms	  -6.3%
     4		   4092  ms	  3764  ms	  -8.0%
---
 arch/x86/include/asm/qspinlock_paravirt.h |   59 +++++++++++++++++++++++++++++
 kernel/locking/qspinlock_paravirt.h       |   43 +++++++++++++--------
 2 files changed, 86 insertions(+), 16 deletions(-)
arch/x86/include/asm/qspinlock_paravirt.h +59 -0
index b002e71..3001972 100644
@@ -1,6 +1,65 @@
 #ifndef __ASM_QSPINLOCK_PARAVIRT_H
 #define __ASM_QSPINLOCK_PARAVIRT_H
 
+/*
+ * For x86-64, PV_CALLEE_SAVE_REGS_THUNK() saves and restores 8 64-bit
+ * registers. For i386, however, only 1 32-bit register needs to be saved
+ * and restored. So an optimized version of __pv_queued_spin_unlock() is
+ * hand-coded for 64-bit, but it isn't worthwhile to do it for 32-bit.
+ */
+#ifdef CONFIG_64BIT
+
+PV_CALLEE_SAVE_REGS_THUNK(__pv_queued_spin_unlock_slowpath);
+#define __pv_queued_spin_unlock	__pv_queued_spin_unlock
+#define PV_UNLOCK		"__raw_callee_save___pv_queued_spin_unlock"
+#define PV_UNLOCK_SLOWPATH	"__raw_callee_save___pv_queued_spin_unlock_slowpath"
+
+/*
+ * Optimized assembly version of __raw_callee_save___pv_queued_spin_unlock
+ * which combines the registers saving trunk and the body of the following
+ * C code:
+ *
+ * void __pv_queued_spin_unlock(struct qspinlock *lock)
+ * {
+ *	struct __qspinlock *l = (void *)lock;
+ *	u8 lockval = cmpxchg(&l->locked, _Q_LOCKED_VAL, 0);
+ *
+ *	if (likely(lockval == _Q_LOCKED_VAL))
+ *		return;
+ *	pv_queued_spin_unlock_slowpath(lock, lockval);
+ * }
+ *
+ * For x86-64,
+ *   rdi = lock 	(first argument)
+ *   rsi = lockval	(second argument)
+ *   rdx = internal variable (set to 0)
+ */
+asm(".pushsection .text;"
+    ".globl " PV_UNLOCK ";"
+    ".align 4,0x90;"
+    PV_UNLOCK ": "
+    "push  %rdx;"
+    "mov   $0x1,%eax;"
+    "xor   %edx,%edx;"
+    "lock cmpxchg %dl,(%rdi);"
+    "cmp   $0x1,%al;"
+    "jne   .slowpath;"
+    "pop   %rdx;"
+    "ret;"
+    ".slowpath: "
+    "push   %rsi;"
+    "movzbl %al,%esi;"
+    "call " PV_UNLOCK_SLOWPATH ";"
+    "pop    %rsi;"
+    "pop    %rdx;"
+    "ret;"
+    ".size " PV_UNLOCK ", .-" PV_UNLOCK ";"
+    ".popsection");
+
+#else /* CONFIG_64BIT */
+
+extern void __pv_queued_spin_unlock(struct qspinlock *lock);
 PV_CALLEE_SAVE_REGS_THUNK(__pv_queued_spin_unlock);
 
+#endif /* CONFIG_64BIT */
 #endif
kernel/locking/qspinlock_paravirt.h +27 -16

62 lines not on this page · raw message

Signature · 1 line
1.7.1
#7 lore raw

PATCH v5 5/6 locking/pvqspinlock: Allow vCPUs kick-ahead

In reply to Waiman Long

Frequent CPU halting (vmexit) and CPU kicking (vmenter) lengthens critical section and block forward progress. This patch implements a kick-ahead mechanism where the unlocker will kick the queue head vCPUs as well as up to four additional vCPUs next to the queue head if they were halted. The kickings are done after exiting the critical section to improve parallelism.

The amount of kick-ahead allowed depends on the number of vCPUs in the VM guest. Currently it allows up to 1 vCPU kick-ahead per 4 vCPUs available up to a maximum of PV_KICK_AHEAD_MAX (4). There are diminishing returns in increasing the maximum value. The current value of 4 is a compromise of getting a nice performance boost without penalizing too much on the one vCPU that is doing all the kickings.

Linux kernel builds were run in KVM guest on an 8-socket, 4 cores/socket Westmere-EX system and a 4-socket, 8 cores/socket Haswell-EX system. Both systems are configured to have 32 physical CPUs. The kernel build times before and after the patch were:

Westmere Haswell
Patch 32 vCPUs 48 vCPUs 32 vCPUs 48 vCPUs
----- -------- -------- -------- --------
Before patch 3m21.9s 11m20.6s 2m08.6s 17m12.8s
After patch 3m03.2s 9m21.1s 2m08.9s 16m14.8s

This improves performance quite substantially on Westmere, but not so much on Haswell.

---
 kernel/locking/qspinlock_paravirt.h |   71 +++++++++++++++++++++++++++++++++-
 1 files changed, 68 insertions(+), 3 deletions(-)
kernel/locking/qspinlock_paravirt.h +68 -3
index 7c9d6ed..9996609 100644
@@ -57,6 +57,7 @@ enum pv_qlock_stat {
 	pvstat_wait_again,
 	pvstat_kick_wait,
 	pvstat_kick_unlock,
+	pvstat_kick_ahead,
 	pvstat_pend_lock,
 	pvstat_pend_fail,
 	pvstat_spurious,
@@ -77,6 +78,7 @@ static const char * const stat_fsnames[pvstat_num] = {
 	[pvstat_wait_again]  = "wait_again_count",
 	[pvstat_kick_wait]   = "kick_wait_count",
 	[pvstat_kick_unlock] = "kick_unlock_count",
+	[pvstat_kick_ahead]  = "kick_ahead_count",
 	[pvstat_pend_lock]   = "pending_lock_count",
 	[pvstat_pend_fail]   = "pending_fail_count",
 	[pvstat_spurious]    = "spurious_wakeup",
@@ -89,7 +91,7 @@ static atomic_t pvstats[pvstat_num];
  * pv_kick_latencies = sum of all pv_kick latencies in ns
  * pv_wake_latencies = sum of all wakeup latencies in ns
  *
- * Avg kick latency   = pv_kick_latencies/kick_unlock_count
+ * Avg kick latency   = pv_kick_latencies/(kick_unlock_count + kick_ahead_count)
  * Avg wake latency   = pv_wake_latencies/kick_wait_count
  * Avg # of hops/hash = hash_hops_count/kick_unlock_count
  */
@@ -221,6 +223,18 @@ static struct pv_hash_entry *pv_lock_hash;
 static unsigned int pv_lock_hash_bits __read_mostly;
 
 /*
+ * Allow kick-ahead of vCPUs at unlock time
+ *
+ * The pv_kick_ahead value is set by a simple formula that 1 vCPU kick-ahead
+ * is allowed per 4 vCPUs available up to a maximum of PV_KICK_AHEAD_MAX.
+ * There are diminishing returns in increasing PV_KICK_AHEAD_MAX. The current
+ * value of 4 is a good compromise that gives a good performance boost without
+ * penalizing the vCPU that is doing the kicking by too much.
+ */
+#define PV_KICK_AHEAD_MAX	4
+static int pv_kick_ahead __read_mostly;
+
+/*
  * Allocate memory for the PV qspinlock hash buckets
  *
  * This function should be called from the paravirt spinlock initialization
@@ -228,7 +242,8 @@ static unsigned int pv_lock_hash_bits __read_mostly;
  */
 void __init __pv_init_lock_hash(void)
 {
-	int pv_hash_size = ALIGN(4 * num_possible_cpus(), PV_HE_PER_LINE);
+	int ncpus = num_possible_cpus();
+	int pv_hash_size = ALIGN(4 * ncpus, PV_HE_PER_LINE);
 
 	if (pv_hash_size < PV_HE_MIN)
 		pv_hash_size = PV_HE_MIN;
@@ -242,6 +257,13 @@ void __init __pv_init_lock_hash(void)
 					       pv_hash_size, 0, HASH_EARLY,
 					       &pv_lock_hash_bits, NULL,
 					       pv_hash_size, pv_hash_size);
+	/*
+	 * Enable the unlock kick ahead mode according to the number of
+	 * vCPUs available.
+	 */
+	pv_kick_ahead = min(ncpus/4, PV_KICK_AHEAD_MAX);
+	if (pv_kick_ahead)
+		pr_info("PV unlock kick ahead max count = %d\n", pv_kick_ahead);
 }
 
 #define for_each_hash_entry(he, offset, hash)						\
@@ -551,6 +573,26 @@ static void pv_wait_head(struct qspinlock *lock, struct mcs_spinlock *node)
 }
 
 /*
+ * Helper to get the address of the next kickable node
+ *
+ * The node has to be in the halted state. The state will then be
+ * transitioned to the running state. If no kickable node is found, NULL
+ * will be returned.
+ */
+static inline struct pv_node *pv_get_kick_node(struct pv_node *node)
+{
+	struct pv_node *next = (struct pv_node *)READ_ONCE(node->mcs.next);
+
+	if (!next || (READ_ONCE(next->state) != vcpu_halted))
+		return NULL;
+
+	if (xchg(&next->state, vcpu_running) != vcpu_halted)
+		next = NULL;	/* No kicking is needed */
+
+	return next;
+}
+
+/*
  * PV versions of the unlock fastpath and slowpath functions to be used
  * instead of queued_spin_unlock().
  */
@@ -558,7 +600,8 @@ __visible void
 __pv_queued_spin_unlock_slowpath(struct qspinlock *lock, u8 locked)
 {
 	struct __qspinlock *l = (void *)lock;
-	struct pv_node *node;
+	struct pv_node *node, *next;
+	int i, nr_kick, cpus[PV_KICK_AHEAD_MAX];
 
 	if (unlikely(locked != _Q_SLOW_VAL)) {
 		WARN(!debug_locks_silent,
@@ -583,6 +626,20 @@ __pv_queued_spin_unlock_slowpath(struct qspinlock *lock, u8 locked)
 	node = pv_unhash(lock);
 
 	/*
+	 * Implement unlock kick-ahead
+	 *
+	 * Access the next group of nodes, if available, and prepare to kick
+	 * them after releasing the lock if they are in the halted state. This
+	 * should improve performance on an overcommitted system.
+	 */
+	for (nr_kick = 0, next = node; nr_kick < pv_kick_ahead; nr_kick++) {
+		next = pv_get_kick_node(next);
+		if (!next)
+			break;
+		cpus[nr_kick] = next->cpu;
+	}
+
+	/*
 	 * Now that we have a reference to the (likely) blocked pv_node,
 	 * release the lock.
 	 */
@@ -597,6 +654,14 @@ __pv_queued_spin_unlock_slowpath(struct qspinlock *lock, u8 locked)
 	 */
 	pvstat_inc(pvstat_kick_unlock);
 	pv_kick(node->cpu);
+
+	/*
+	 * Kick the next group of vCPUs, if available.
+	 */
+	for (i = 0; i < nr_kick; i++) {
+		pvstat_inc(pvstat_kick_ahead);
+		pv_kick(cpus[i]);
+	}
 }
 
 /*
Signature · 1 line
1.7.1
#8 lore raw

PATCH v5 6/6 locking/pvqspinlock: Queue node adaptive spinning

In reply to Waiman Long

In an overcommitted guest where some vCPUs have to be halted to make forward progress in other areas, it is highly likely that a vCPU later in the spinlock queue will be spinning while the ones earlier in the queue would have been halted. The spinning in the later vCPUs is then just a waste of precious CPU cycles because they are not going to get the lock soon as the earlier ones have to be woken up and take their turn to get the lock.

Reducing the spinning threshold is found to improve performance in an overcommitted VM guest, but decrease performance when there is no overcommittment.

This patch implements an adaptive spinning mechanism where the vCPU will call pv_wait() earlier if all the following conditions are true:

1) the vCPU has not been halted before;
2) the previous vCPU is in the halted state;
3) there are a lot of pv_wait() for the current vCPU recently.

Linux kernel builds were run in KVM guest on an 8-socket, 4 cores/socket Westmere-EX system and a 4-socket, 8 cores/socket Haswell-EX system. Both systems are configured to have 32 physical CPUs. The kernel build times before and after the patch were:

Westmere Haswell
Patch 32 vCPUs 48 vCPUs 32 vCPUs 48 vCPUs
----- -------- -------- -------- --------
Before patch 3m03.2s 9m21.1s 2m08.9s 16m14.8s
After patch 3m04.1s 9m28.5s 2m09.5s 8m29.3s

This patch seemed to cause a tiny bit of performance degraduation for 32 vCPUs. For 48 vCPUs, there wasn't much change for Westmere, but a pretty big performance jump for Haswell.

---
 kernel/locking/qspinlock.c          |    5 +-
 kernel/locking/qspinlock_paravirt.h |  111 +++++++++++++++++++++++++++++++++-
 2 files changed, 110 insertions(+), 6 deletions(-)
kernel/locking/qspinlock.c +3 -2
index 94fdd27..da39d43 100644
@@ -258,7 +258,8 @@ static __always_inline void set_locked(struct qspinlock *lock)
  */
 
 static __always_inline void __pv_init_node(struct mcs_spinlock *node) { }
-static __always_inline void __pv_wait_node(struct mcs_spinlock *node) { }
+static __always_inline void __pv_wait_node(struct mcs_spinlock *node,
+					   struct mcs_spinlock *prev) { }
 static __always_inline void __pv_kick_node(struct qspinlock *lock,
 					   struct mcs_spinlock *node) { }
 static __always_inline void __pv_wait_head(struct qspinlock *lock,
@@ -415,7 +416,7 @@ queue:
 		prev = decode_tail(old);
 		WRITE_ONCE(prev->next, node);
 
-		pv_wait_node(node);
+		pv_wait_node(node, prev);
 		arch_mcs_spin_lock_contended(&node->locked);
 	}
 
kernel/locking/qspinlock_paravirt.h +107 -4

188 lines not on this page · raw message

Signature · 1 line
1.7.1