qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Paolo Bonzini <pbonzini@redhat.com>
To: Akihiko Odaki <akihiko.odaki@daynix.com>,
	Stefan Weil <sw@weilnetz.de>, Peter Xu <peterx@redhat.com>,
	Fabiano Rosas <farosas@suse.de>,
	Hailiang Zhang <zhanghailiang@xfusion.com>
Cc: "Phil Dennis-Jordan" <phil@philjordan.eu>,
	qemu-devel@nongnu.org, devel@daynix.com,
	"Marc-André Lureau" <marcandre.lureau@redhat.com>,
	"Daniel P. Berrangé" <berrange@redhat.com>,
	"Philippe Mathieu-Daudé" <philmd@linaro.org>
Subject: Re: [PATCH v2 3/9] qemu-thread: Avoid futex abstraction for non-Linux
Date: Sat, 10 May 2025 17:02:42 +0200	[thread overview]
Message-ID: <54bb98a7-110d-41d2-a683-1428cb7d94a2@redhat.com> (raw)
In-Reply-To: <20250510-event-v2-3-7953177ce1b8@daynix.com>

On 5/10/25 10:51, Akihiko Odaki wrote:
> Add special implementations of qemu_event_set() and qemu_event_wait()
> using pthread primitives. qemu_event_wait() will ensure qemu_event_set()
> finishes, and these functions will avoid complex barrier and atomic
> operations.

Unfortunately not...

> Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
> Tested-by: Phil Dennis-Jordan <phil@philjordan.eu>
> Reviewed-by: Phil Dennis-Jordan <phil@philjordan.eu>
> ---
>   util/qemu-thread-posix.c | 45 ++++++++++++++++++---------------------------
>   1 file changed, 18 insertions(+), 27 deletions(-)
> 
> diff --git a/util/qemu-thread-posix.c b/util/qemu-thread-posix.c
> index 13459e44c768..805cac444f15 100644
> --- a/util/qemu-thread-posix.c
> +++ b/util/qemu-thread-posix.c
> @@ -319,28 +319,6 @@ void qemu_sem_wait(QemuSemaphore *sem)
>   
>   #ifdef __linux__
>   #include "qemu/futex.h"
> -#else
> -static inline void qemu_futex_wake(QemuEvent *ev, int n)
> -{
> -    assert(ev->initialized);
> -    pthread_mutex_lock(&ev->lock);
> -    if (n == 1) {
> -        pthread_cond_signal(&ev->cond);
> -    } else {
> -        pthread_cond_broadcast(&ev->cond);
> -    }
> -    pthread_mutex_unlock(&ev->lock);
> -}
> -
> -static inline void qemu_futex_wait(QemuEvent *ev, unsigned val)
> -{
> -    assert(ev->initialized);
> -    pthread_mutex_lock(&ev->lock);
> -    if (ev->value == val) {
> -        pthread_cond_wait(&ev->cond, &ev->lock);
> -    }
> -    pthread_mutex_unlock(&ev->lock);
> -}
>   #endif
>   
>   /* Valid transitions:
> @@ -363,7 +341,7 @@ static inline void qemu_futex_wait(QemuEvent *ev, unsigned val)
>   
>   void qemu_event_init(QemuEvent *ev, bool init)
>   {
> -#ifndef __linux__
> +#ifndef CONFIG_LINUX
>       pthread_mutex_init(&ev->lock, NULL);
>       pthread_cond_init(&ev->cond, NULL);
>   #endif
> @@ -376,7 +354,7 @@ void qemu_event_destroy(QemuEvent *ev)
>   {
>       assert(ev->initialized);
>       ev->initialized = false;
> -#ifndef __linux__
> +#ifndef CONFIG_LINUX
>       pthread_mutex_destroy(&ev->lock);
>       pthread_cond_destroy(&ev->cond);
>   #endif
> @@ -386,6 +364,7 @@ void qemu_event_set(QemuEvent *ev)
>   {
>       assert(ev->initialized);
>   
> +#ifdef CONFIG_LINUX
>       /*
>        * Pairs with both qemu_event_reset() and qemu_event_wait().
>        *

The user of this code will not have mutexes so some care is needed.  You 
have:

	if (!qatomic_load_acquire(&done)) {
		qemu_event_reset(&ev);
		if (!qatomic_load_acquire(&done))
			qemu_event_wait(&ev);
	}

and on the other side

	qatomic_store_release(&done, 1);
	qemu_event_set(&ev);
	--> qatomic_set(&ev.value, EV_SET);

I don't think this is correct without the memory barrier in 
qemu_event_set(), though I am not actually sure how you'd
add it.

The problem is, I don't see anything that prevents this:

				set done
				qemu_event_set()
				  pthread_mutex_lock()
				  ev.value = SET
	qemu_event_reset()
	  ev.value |= FREE
	  smp_mb__after_rmw()
				  // store buffer not flushed yet,
				  // so other thread doesn't see "done"
	read done
	pthread_mutex_lock()
				  pthread_cond_broadcast()
				  pthread_mutex_unlock()
	while ev.value != SET
		// hang
		pthread_cond_wait()

The barrier in qemu_event_reset() is not enough, you need one on each side.

> +#else
> +    pthread_mutex_lock(&ev->lock);
> +    if (qatomic_read(&ev->value) != EV_SET) {

Apart from the above this needs to be a while(), because condition 
variables also have spurious wakeups.

Paolo

> +        pthread_cond_wait(&ev->cond, &ev->lock);
> +    }
> +    pthread_mutex_unlock(&ev->lock);
> +#endif
>   }
>   
>   static __thread NotifierList thread_exit;
> 



  reply	other threads:[~2025-05-10 15:03 UTC|newest]

Thread overview: 12+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-05-10  8:51 [PATCH v2 0/9] Improve futex usage Akihiko Odaki
2025-05-10  8:51 ` [PATCH v2 1/9] futex: Check value after qemu_futex_wait() Akihiko Odaki
2025-05-10  8:51 ` [PATCH v2 2/9] futex: Support Windows Akihiko Odaki
2025-05-10  8:51 ` [PATCH v2 3/9] qemu-thread: Avoid futex abstraction for non-Linux Akihiko Odaki
2025-05-10 15:02   ` Paolo Bonzini [this message]
2025-05-11  4:41     ` Akihiko Odaki
2025-05-10  8:51 ` [PATCH v2 4/9] qemu-thread: Use futex for QemuEvent on Windows Akihiko Odaki
2025-05-10  8:51 ` [PATCH v2 5/9] qemu-thread: Use futex if available for QemuLockCnt Akihiko Odaki
2025-05-10  8:51 ` [PATCH v2 6/9] migration: Replace QemuSemaphore with QemuEvent Akihiko Odaki
2025-05-10  8:51 ` [PATCH v2 7/9] migration/colo: " Akihiko Odaki
2025-05-10  8:51 ` [PATCH v2 8/9] migration/postcopy: " Akihiko Odaki
2025-05-10  8:51 ` [PATCH v2 9/9] hw/display/apple-gfx: " Akihiko Odaki

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=54bb98a7-110d-41d2-a683-1428cb7d94a2@redhat.com \
    --to=pbonzini@redhat.com \
    --cc=akihiko.odaki@daynix.com \
    --cc=berrange@redhat.com \
    --cc=devel@daynix.com \
    --cc=farosas@suse.de \
    --cc=marcandre.lureau@redhat.com \
    --cc=peterx@redhat.com \
    --cc=phil@philjordan.eu \
    --cc=philmd@linaro.org \
    --cc=qemu-devel@nongnu.org \
    --cc=sw@weilnetz.de \
    --cc=zhanghailiang@xfusion.com \
    /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).