[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [PATCH RFC 6/7] migration: Split multifd pending_job into two boolea
From: |
Fabiano Rosas |
Subject: |
Re: [PATCH RFC 6/7] migration: Split multifd pending_job into two booleans |
Date: |
Mon, 23 Oct 2023 12:15:49 -0300 |
Peter Xu <peterx@redhat.com> writes:
> Signed-off-by: Peter Xu <peterx@redhat.com>
> ---
> migration/multifd.h | 16 ++++++++++------
> migration/multifd.c | 33 +++++++++++++++++++++++----------
> 2 files changed, 33 insertions(+), 16 deletions(-)
>
> diff --git a/migration/multifd.h b/migration/multifd.h
> index 2acf400085..ddee7b8d8a 100644
> --- a/migration/multifd.h
> +++ b/migration/multifd.h
> @@ -101,12 +101,16 @@ typedef struct {
> uint32_t flags;
> /* global number of generated multifd packets */
> uint64_t packet_num;
> - /* thread has work to do */
> - int pending_job;
> - /* array of pages to sent.
> - * The owner of 'pages' depends of 'pending_job' value:
> - * pending_job == 0 -> migration_thread can use it.
> - * pending_job != 0 -> multifd_channel can use it.
> + /* thread has a request to sync all data */
> + bool pending_sync;
> + /* thread has something to send */
> + bool pending_job;
> + /*
> + * Array of pages to sent. The owner of 'pages' depends of
> + * 'pending_job' value:
> + *
> + * - true -> multifd_channel owns it.
> + * - false -> migration_thread owns it.
> */
> MultiFDPages_t *pages;
>
> diff --git a/migration/multifd.c b/migration/multifd.c
> index 8140520843..fe8d746ff9 100644
> --- a/migration/multifd.c
> +++ b/migration/multifd.c
> @@ -425,7 +425,7 @@ static int multifd_send_pages(QEMUFile *f)
> p = &multifd_send_state->params[i];
> qemu_mutex_lock(&p->mutex);
> if (!p->pending_job) {
> - p->pending_job++;
> + p->pending_job = true;
> next_channel = (i + 1) % migrate_multifd_channels();
> break;
> }
> @@ -615,8 +615,7 @@ int multifd_send_sync_main(QEMUFile *f)
>
> qemu_mutex_lock(&p->mutex);
> p->packet_num = multifd_send_state->packet_num++;
> - p->flags |= MULTIFD_FLAG_SYNC;
> - p->pending_job++;
> + p->pending_sync = true;
> qemu_mutex_unlock(&p->mutex);
> qemu_sem_post(&p->sem);
> }
> @@ -747,8 +746,6 @@ static void *multifd_send_thread(void *opaque)
>
> qemu_mutex_lock(&p->mutex);
> if (p->pending_job) {
> - bool need_sync = p->flags & MULTIFD_FLAG_SYNC;
> -
> if (!multifd_send_prepare(p, &local_err)) {
> assert(local_err);
> qemu_mutex_unlock(&p->mutex);
> @@ -764,12 +761,27 @@ static void *multifd_send_thread(void *opaque)
> qemu_mutex_lock(&p->mutex);
>
> /* Send successful, mark the task completed */
> - p->pending_job--;
> + p->pending_job = false;
> +
> + } else if (p->pending_sync) {
Is your intention here to stop sending the SYNC along with the pages?
This would have to loop once more to send the sync.
- [PATCH RFC 3/7] migration: multifd_send_kick_main(), (continued)
[PATCH RFC 5/7] migration: Modulize multifd send threads with a few helpers, Peter Xu, 2023/10/22
[PATCH RFC 7/7] migration: Further unify paths for multifd normal or sync requests, Peter Xu, 2023/10/22
[PATCH RFC 6/7] migration: Split multifd pending_job into two booleans, Peter Xu, 2023/10/22
- Re: [PATCH RFC 6/7] migration: Split multifd pending_job into two booleans,
Fabiano Rosas <=