From: Nicholas Piggin <npiggin@gmail.com>
To: linuxppc-dev@lists.ozlabs.org
Cc: Jordan Niethe <jniethe5@gmail.com>,
Laurent Dufour <ldufour@linux.ibm.com>,
Nicholas Piggin <npiggin@gmail.com>
Subject: [PATCH v3 05/17] powerpc/qspinlock: theft prevention to control latency
Date: Sat, 26 Nov 2022 19:59:20 +1000 [thread overview]
Message-ID: <20221126095932.1234527-6-npiggin@gmail.com> (raw)
In-Reply-To: <20221126095932.1234527-1-npiggin@gmail.com>
Give the queue head the ability to stop stealers. After a number of
spins without sucessfully acquiring the lock, the queue head sets
this, which halts stealing and will assure it is the next owner.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
---
arch/powerpc/include/asm/qspinlock_types.h | 8 +++-
arch/powerpc/lib/qspinlock.c | 53 ++++++++++++++++++++++
2 files changed, 60 insertions(+), 1 deletion(-)
diff --git a/arch/powerpc/include/asm/qspinlock_types.h b/arch/powerpc/include/asm/qspinlock_types.h
index b9a5a52fa670..1911a8a16237 100644
--- a/arch/powerpc/include/asm/qspinlock_types.h
+++ b/arch/powerpc/include/asm/qspinlock_types.h
@@ -29,7 +29,8 @@ typedef struct qspinlock {
* Bitfields in the lock word:
*
* 0: locked bit
- * 1-16: unused bits
+ * 1-15: unused bits
+ * 16: must queue bit
* 17-31: tail cpu (+1)
*/
#define _Q_SET_MASK(type) (((1U << _Q_ ## type ## _BITS) - 1)\
@@ -39,6 +40,11 @@ typedef struct qspinlock {
#define _Q_LOCKED_BITS 1
#define _Q_LOCKED_VAL (1U << _Q_LOCKED_OFFSET)
+/* 0x00010000 */
+#define _Q_MUST_Q_OFFSET 16
+#define _Q_MUST_Q_BITS 1
+#define _Q_MUST_Q_VAL (1U << _Q_MUST_Q_OFFSET)
+
/* 0xfffe0000 */
#define _Q_TAIL_CPU_OFFSET 17
#define _Q_TAIL_CPU_BITS 15
diff --git a/arch/powerpc/lib/qspinlock.c b/arch/powerpc/lib/qspinlock.c
index 6d67bc38b122..979b17ac7bd1 100644
--- a/arch/powerpc/lib/qspinlock.c
+++ b/arch/powerpc/lib/qspinlock.c
@@ -22,6 +22,7 @@ struct qnodes {
/* Tuning parameters */
static int steal_spins __read_mostly = (1<<5);
static bool maybe_stealers __read_mostly = true;
+static int head_spins __read_mostly = (1<<8);
static DEFINE_PER_CPU_ALIGNED(struct qnodes, qnodes);
@@ -30,6 +31,11 @@ static __always_inline int get_steal_spins(void)
return steal_spins;
}
+static __always_inline int get_head_spins(void)
+{
+ return head_spins;
+}
+
static inline u32 encode_tail_cpu(int cpu)
{
return (cpu + 1) << _Q_TAIL_CPU_OFFSET;
@@ -104,6 +110,22 @@ static __always_inline u32 publish_tail_cpu(struct qspinlock *lock, u32 tail)
return prev;
}
+static __always_inline u32 set_mustq(struct qspinlock *lock)
+{
+ u32 prev;
+
+ asm volatile(
+"1: lwarx %0,0,%1 # set_mustq \n"
+" or %0,%0,%2 \n"
+" stwcx. %0,0,%1 \n"
+" bne- 1b \n"
+ : "=&r" (prev)
+ : "r" (&lock->val), "r" (_Q_MUST_Q_VAL)
+ : "cr0", "memory");
+
+ return prev;
+}
+
static struct qnode *get_tail_qnode(struct qspinlock *lock, u32 val)
{
int cpu = decode_tail_cpu(val);
@@ -139,6 +161,9 @@ static inline bool try_to_steal_lock(struct qspinlock *lock)
do {
u32 val = READ_ONCE(lock->val);
+ if (val & _Q_MUST_Q_VAL)
+ break;
+
if (unlikely(!(val & _Q_LOCKED_VAL))) {
if (__queued_spin_trylock_steal(lock))
return true;
@@ -157,7 +182,9 @@ static inline void queued_spin_lock_mcs_queue(struct qspinlock *lock)
struct qnodes *qnodesp;
struct qnode *next, *node;
u32 val, old, tail;
+ bool mustq = false;
int idx;
+ int iters = 0;
BUILD_BUG_ON(CONFIG_NR_CPUS >= (1U << _Q_TAIL_CPU_BITS));
@@ -209,6 +236,15 @@ static inline void queued_spin_lock_mcs_queue(struct qspinlock *lock)
break;
cpu_relax();
+ if (!maybe_stealers)
+ continue;
+ iters++;
+
+ if (!mustq && iters >= get_head_spins()) {
+ mustq = true;
+ set_mustq(lock);
+ val |= _Q_MUST_Q_VAL;
+ }
}
/* If we're the last queued, must clean up the tail. */
@@ -293,9 +329,26 @@ static int steal_spins_get(void *data, u64 *val)
DEFINE_SIMPLE_ATTRIBUTE(fops_steal_spins, steal_spins_get, steal_spins_set, "%llu\n");
+static int head_spins_set(void *data, u64 val)
+{
+ head_spins = val;
+
+ return 0;
+}
+
+static int head_spins_get(void *data, u64 *val)
+{
+ *val = head_spins;
+
+ return 0;
+}
+
+DEFINE_SIMPLE_ATTRIBUTE(fops_head_spins, head_spins_get, head_spins_set, "%llu\n");
+
static __init int spinlock_debugfs_init(void)
{
debugfs_create_file("qspl_steal_spins", 0600, arch_debugfs_dir, NULL, &fops_steal_spins);
+ debugfs_create_file("qspl_head_spins", 0600, arch_debugfs_dir, NULL, &fops_head_spins);
return 0;
}
--
2.37.2
next prev parent reply other threads:[~2022-11-26 10:05 UTC|newest]
Thread overview: 21+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-11-26 9:59 [PATCH v3 00/17] powerpc: alternate queued spinlock implementation Nicholas Piggin
2022-11-26 9:59 ` [PATCH v3 01/17] powerpc/qspinlock: add mcs queueing for contended waiters Nicholas Piggin
2022-11-26 9:59 ` [PATCH v3 02/17] powerpc/qspinlock: use a half-word store to unlock to avoid larx/stcx Nicholas Piggin
2022-11-26 9:59 ` [PATCH v3 03/17] powerpc/qspinlock: convert atomic operations to assembly Nicholas Piggin
2022-11-26 9:59 ` [PATCH v3 04/17] powerpc/qspinlock: allow new waiters to steal the lock before queueing Nicholas Piggin
2022-11-26 9:59 ` Nicholas Piggin [this message]
2022-11-26 9:59 ` [PATCH v3 06/17] powerpc/qspinlock: store owner CPU in lock word Nicholas Piggin
2022-11-26 9:59 ` [PATCH v3 07/17] powerpc/qspinlock: paravirt yield to lock owner Nicholas Piggin
2022-11-26 9:59 ` [PATCH v3 08/17] powerpc/qspinlock: implement option to yield to previous node Nicholas Piggin
2022-11-26 9:59 ` [PATCH v3 09/17] powerpc/qspinlock: allow stealing when head of queue yields Nicholas Piggin
2022-11-26 9:59 ` [PATCH v3 10/17] powerpc/qspinlock: allow propagation of yield CPU down the queue Nicholas Piggin
2022-11-26 9:59 ` [PATCH v3 11/17] powerpc/qspinlock: add ability to prod new queue head CPU Nicholas Piggin
2022-11-26 9:59 ` [PATCH v3 12/17] powerpc/qspinlock: allow lock stealing in trylock and lock fastpath Nicholas Piggin
2022-11-26 9:59 ` [PATCH v3 13/17] powerpc/qspinlock: use spin_begin/end API Nicholas Piggin
2022-11-26 9:59 ` [PATCH v3 14/17] powerpc/qspinlock: reduce remote node steal spins Nicholas Piggin
2022-11-26 9:59 ` [PATCH v3 15/17] powerpc/qspinlock: allow indefinite spinning on a preempted owner Nicholas Piggin
2022-11-26 9:59 ` [PATCH v3 16/17] powerpc/qspinlock: provide accounting and options for sleepy locks Nicholas Piggin
2022-11-26 9:59 ` [PATCH v3 17/17] powerpc/qspinlock: add compile-time tuning adjustments Nicholas Piggin
2022-11-28 3:11 ` [PATCH v3 real 01/17] powerpc/qspinlock: powerpc qspinlock implementation Nicholas Piggin
2022-12-08 12:39 ` (subset) [PATCH v3 00/17] powerpc: alternate queued spinlock implementation Michael Ellerman
2023-04-13 10:58 ` Shrikanth Hegde
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20221126095932.1234527-6-npiggin@gmail.com \
--to=npiggin@gmail.com \
--cc=jniethe5@gmail.com \
--cc=ldufour@linux.ibm.com \
--cc=linuxppc-dev@lists.ozlabs.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).