qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Igor Mammedov <imammedo@redhat.com>
To: qemu-devel@nongnu.org
Cc: Paolo Bonzini <pbonzini@redhat.com>, Peter Xu <peterx@redhat.com>,
	"Michael S. Tsirkin" <mst@redhat.com>,
	mtosatti@redhat.com
Subject: [PATCH v4 5/8] hpet: make main counter read lock-less
Date: Thu, 14 Aug 2025 18:05:57 +0200	[thread overview]
Message-ID: <20250814160600.2327672-6-imammedo@redhat.com> (raw)
In-Reply-To: <20250814160600.2327672-1-imammedo@redhat.com>

Make access to main HPET counter lock-less.

In unlikely event of an update in progress, readers will busy wait
untill update is finished.

As result micro benchmark of concurrent reading of HPET counter
with large number of vCPU shows over 80% better (less) latency.

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
---
v3:
  * make reader busy wait during update and reuse existing seqlock API
       Peter Xu <peterx@redhat.com>
---
 hw/timer/hpet.c | 26 ++++++++++++++++++++------
 1 file changed, 20 insertions(+), 6 deletions(-)

diff --git a/hw/timer/hpet.c b/hw/timer/hpet.c
index c776afc0f2..789a31d0a0 100644
--- a/hw/timer/hpet.c
+++ b/hw/timer/hpet.c
@@ -39,6 +39,7 @@
 #include "system/address-spaces.h"
 #include "qom/object.h"
 #include "qemu/lockable.h"
+#include "qemu/seqlock.h"
 #include "trace.h"
 
 struct hpet_fw_config hpet_fw_cfg = {.count = UINT8_MAX};
@@ -74,6 +75,7 @@ struct HPETState {
     MemoryRegion iomem;
     uint64_t hpet_offset;
     bool hpet_offset_saved;
+    QemuSeqLock state_version;
     qemu_irq irqs[HPET_NUM_IRQ_ROUTES];
     uint32_t flags;
     uint8_t rtc_irq_level;
@@ -430,17 +432,25 @@ static uint64_t hpet_ram_read(void *opaque, hwaddr addr,
     trace_hpet_ram_read(addr);
     addr &= ~4;
 
-    QEMU_LOCK_GUARD(&s->lock);
     if (addr == HPET_COUNTER) {
-        if (hpet_enabled(s)) {
-            cur_tick = hpet_get_ticks(s);
-        } else {
-            cur_tick = s->hpet_counter;
-        }
+        unsigned version;
+
+        /*
+         * Write update is rare, so busywait here is unlikely to happen
+         */
+        do {
+            version = seqlock_read_begin(&s->state_version);
+            if (unlikely(!hpet_enabled(s))) {
+                cur_tick = s->hpet_counter;
+            } else {
+                cur_tick = hpet_get_ticks(s);
+            }
+        } while (seqlock_read_retry(&s->state_version, version));
         trace_hpet_ram_read_reading_counter(addr & 4, cur_tick);
         return cur_tick >> shift;
     }
 
+    QEMU_LOCK_GUARD(&s->lock);
     /*address range of all global regs*/
     if (addr <= 0xff) {
         switch (addr) {
@@ -500,6 +510,7 @@ static void hpet_ram_write(void *opaque, hwaddr addr,
             old_val = s->config;
             new_val = deposit64(old_val, shift, len, value);
             new_val = hpet_fixup_reg(new_val, old_val, HPET_CFG_WRITE_MASK);
+            seqlock_write_begin(&s->state_version);
             s->config = new_val;
             if (activating_bit(old_val, new_val, HPET_CFG_ENABLE)) {
                 /* Enable main counter and interrupt generation. */
@@ -518,6 +529,8 @@ static void hpet_ram_write(void *opaque, hwaddr addr,
                     hpet_del_timer(&s->timer[i]);
                 }
             }
+            seqlock_write_end(&s->state_version);
+
             /* i8254 and RTC output pins are disabled
              * when HPET is in legacy mode */
             if (activating_bit(old_val, new_val, HPET_CFG_LEGACY)) {
@@ -686,6 +699,7 @@ static void hpet_init(Object *obj)
     HPETState *s = HPET(obj);
 
     qemu_mutex_init(&s->lock);
+    seqlock_init(&s->state_version);
     /* HPET Area */
     memory_region_init_io(&s->iomem, obj, &hpet_ram_ops, s, "hpet", HPET_LEN);
     memory_region_enable_lockless_io(&s->iomem);
-- 
2.47.1



  parent reply	other threads:[~2025-08-14 16:07 UTC|newest]

Thread overview: 34+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-08-14 16:05 [PATCH v4 0/8] Reinvent BQL-free PIO/MMIO Igor Mammedov
2025-08-14 16:05 ` [PATCH v4 1/8] memory: reintroduce BQL-free fine-grained PIO/MMIO Igor Mammedov
2025-08-25 10:55   ` Philippe Mathieu-Daudé
2025-08-14 16:05 ` [PATCH v4 2/8] acpi: mark PMTIMER as unlocked Igor Mammedov
2025-08-14 16:05 ` [PATCH v4 3/8] hpet: switch to fain-grained device locking Igor Mammedov
2025-08-25 14:43   ` Zhao Liu
2025-08-14 16:05 ` [PATCH v4 4/8] hpet: move out main counter read into a separate block Igor Mammedov
2025-08-25 14:44   ` Zhao Liu
2025-08-14 16:05 ` Igor Mammedov [this message]
2025-08-25 14:55   ` [PATCH v4 5/8] hpet: make main counter read lock-less Zhao Liu
2025-08-25 15:10     ` Igor Mammedov
2025-08-14 16:05 ` [PATCH v4 6/8] add cpu_test_interrupt()/cpu_set_interrupt() helpers and use them tree wide Igor Mammedov
2025-08-14 19:05   ` Peter Xu
2025-08-20 15:01   ` Jason J. Herne
2025-08-21 15:57     ` Igor Mammedov
2025-08-21 15:56   ` [PATCH v5 " Igor Mammedov
2025-08-25  8:16     ` Harsh Prateek Bora
2025-08-25 15:06       ` Igor Mammedov
2025-08-25 10:35     ` Philippe Mathieu-Daudé
2025-08-25 15:02       ` Igor Mammedov
2025-08-25 15:28     ` Zhao Liu
2025-08-25 15:19       ` Igor Mammedov
2025-08-26  7:45         ` Zhao Liu
2025-08-26  8:47           ` Igor Mammedov
2025-08-26  9:27             ` Zhao Liu
2025-08-29  8:18             ` Paolo Bonzini
2025-08-29 12:33               ` Paolo Bonzini
2025-09-01 12:05                 ` Igor Mammedov
2025-09-01 12:06                   ` Paolo Bonzini
2025-08-14 16:05 ` [PATCH v4 7/8] kvm: i386: irqchip: take BQL only if there is an interrupt Igor Mammedov
2025-08-25 10:46   ` Philippe Mathieu-Daudé
2025-08-27  8:40     ` Igor Mammedov
2025-08-14 16:06 ` [PATCH v4 8/8] tcg: move interrupt caching and single step masking closer to user Igor Mammedov
2025-08-29  8:19 ` [PATCH v4 0/8] Reinvent BQL-free PIO/MMIO Paolo Bonzini

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=20250814160600.2327672-6-imammedo@redhat.com \
    --to=imammedo@redhat.com \
    --cc=mst@redhat.com \
    --cc=mtosatti@redhat.com \
    --cc=pbonzini@redhat.com \
    --cc=peterx@redhat.com \
    --cc=qemu-devel@nongnu.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).