Skip to contentSkip to conversation

multifd: Fix the number of channels ready

PATCH In review Juan Quintela 2 participants 5 messages Apr 26, 2023

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 multifd: Fix the number of channels ready

We don't wait in the sem when we are doing a sync_main. Make it wait there. To make things clearer, we mark the channel ready at the begining of the thread loop.

---
 migration/multifd.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
migration/multifd.c +2 -1
index 903df2117b..e625e8725e 100644
@@ -635,6 +635,7 @@ int multifd_send_sync_main(QEMUFile *f)
     for (i = 0; i < migrate_multifd_channels(); i++) {
         MultiFDSendParams *p = &multifd_send_state->params[i];
 
+        qemu_sem_wait(&multifd_send_state->channels_ready);
         trace_multifd_send_sync_main_wait(p->id);
         qemu_sem_wait(&p->sem_sync);
 
@@ -668,6 +669,7 @@ static void *multifd_send_thread(void *opaque)
     p->num_packets = 1;
 
     while (true) {
+        qemu_sem_post(&multifd_send_state->channels_ready);
         qemu_sem_wait(&p->sem);
 
         if (qatomic_read(&multifd_send_state->exiting)) {
@@ -736,7 +738,6 @@ static void *multifd_send_thread(void *opaque)
             if (flags & MULTIFD_FLAG_SYNC) {
                 qemu_sem_post(&p->sem_sync);
             }
-            qemu_sem_post(&multifd_send_state->channels_ready);
         } else if (p->quit) {
             qemu_mutex_unlock(&p->mutex);
             break;
Signature · 1 line
2.40.0
#1 lore raw
Quoted 3 lines · Juan Quintela

We don't wait in the sem when we are doing a sync_main. Make it wait there. To make things clearer, we mark the channel ready at the begining of the thread loop.

So in other words we're estabilishing that "channel ready" means ready to send, regardless of having sent the sync packet. Is that it?

#2 lore raw
Quoted 8 lines · Fabiano Rosas
Quoted 3 lines · Juan Quintela

We don't wait in the sem when we are doing a sync_main. Make it wait there. To make things clearer, we mark the channel ready at the begining of the thread loop.

So in other words we're estabilishing that "channel ready" means ready to send, regardless of having sent the sync packet. Is that it?

Yeap.

There was a bug (from the beggining) that made the counter always get up and up. This fixes it.

It was always supposed to work this way.

/me puts (second time in the week) a brown paper bag on head

Later, Juan.

#3 lore raw
Quoted 10 lines · Juan Quintela
Quoted 8 lines · Fabiano Rosas
Quoted 3 lines · Juan Quintela

We don't wait in the sem when we are doing a sync_main. Make it wait there. To make things clearer, we mark the channel ready at the begining of the thread loop.

So in other words we're estabilishing that "channel ready" means ready to send, regardless of having sent the sync packet. Is that it?

Quoted 6 lines

Yeap.

There was a bug (from the beggining) that made the counter always get up and up. This fixes it.

It was always supposed to work this way.

Ah, great. I'm proposing a multifd variant without the sync packet in my fixed-ram series and moving the channels_ready to the top of the loop means I can stop issuing an extra qemu_sem_post(&p->sem) just to skip the sync packet.