Skip to contentSkip to conversation

multifd: Implement yank for multifd send side

PATCH In review Lukas Straub 2 participants 3 messages Aug 4, 2021 – Aug 17, 2021

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 multifd: Implement yank for multifd send side

When introducing yank functionality in the migration code I forgot to cover the multifd send side.

---

@Leonardo Could you check if this fixes your issue?

 migration/multifd.c | 6 +++++-
 migration/multifd.h | 2 ++
 2 files changed, 7 insertions(+), 1 deletion(-)
migration/multifd.c +5 -1
index 377da78f5b..5a4f158f3c 100644
@@ -546,6 +546,9 @@ void multifd_save_cleanup(void)
         MultiFDSendParams *p = &multifd_send_state->params[i];
         Error *local_err = NULL;
 
+        if (p->registered_yank) {
+            migration_ioc_unregister_yank(p->c);
+        }
         socket_send_channel_destroy(p->c);
         p->c = NULL;
         qemu_mutex_destroy(&p->mutex);
@@ -813,7 +816,8 @@ static bool multifd_channel_connect(MultiFDSendParams *p,
                 return false;
             }
         } else {
-            /* update for tls qio channel */
+            migration_ioc_register_yank(ioc);
+            p->registered_yank = true;
             p->c = ioc;
             qemu_thread_create(&p->thread, p->name, multifd_send_thread, p,
                                    QEMU_THREAD_JOINABLE);
migration/multifd.h +2 -0
index 8d6751f5ed..16c4d112d1 100644
@@ -85,6 +85,8 @@ typedef struct {
     bool running;
     /* should this thread finish */
     bool quit;
+    /* is the yank function registered */
+    bool registered_yank;
     /* thread has work to do */
     int pending_job;
     /* array of pages to sent */
Signature · 1 line
2.32.0
#1 lore raw

Hello Lukas,

Quoted 8 lines · Lukas Straub

When introducing yank functionality in the migration code I forgot to cover the multifd send side.

---

@Leonardo Could you check if this fixes your issue?

In the same scenario I was testing my previous patch,
(http://patchwork.ozlabs.org/project/qemu-devel/patch/[email protected]/)
this patch also seems to fix the issue .
(https://bugzilla.redhat.com/show_bug.cgi?id=1970337).

Quoted 44 lines
 migration/multifd.c | 6 +++++-
 migration/multifd.h | 2 ++
 2 files changed, 7 insertions(+), 1 deletion(-)
migration/multifd.c +5 -1
index 377da78f5b..5a4f158f3c 100644
@@ -546,6 +546,9 @@ void multifd_save_cleanup(void)
         MultiFDSendParams *p = &multifd_send_state->params[i];
         Error *local_err = NULL;
 
+        if (p->registered_yank) {
+            migration_ioc_unregister_yank(p->c);
+        }
         socket_send_channel_destroy(p->c);
         p->c = NULL;
         qemu_mutex_destroy(&p->mutex);
@@ -813,7 +816,8 @@ static bool multifd_channel_connect(MultiFDSendParams *p,
                 return false;
             }
         } else {
-            /* update for tls qio channel */
+            migration_ioc_register_yank(ioc);
+            p->registered_yank = true;
             p->c = ioc;
             qemu_thread_create(&p->thread, p->name, multifd_send_thread, p,
                                    QEMU_THREAD_JOINABLE);
migration/multifd.h +2 -1
index 8d6751f5ed..16c4d112d1 100644
@@ -85,6 +85,8 @@ typedef struct {
     bool running;
     /* should this thread finish */
     bool quit;
+    /* is the yank function registered */
+    bool registered_yank;
     /* thread has work to do */
     int pending_job;
     /* array of pages to sent */
--

2.32.0

Tested-by Reviewed-by
#2 lore raw
Quoted 11 lines · Leonardo Bras Soares Passos

Hello Lukas,

Quoted 7 lines · Lukas Straub

When introducing yank functionality in the migration code I forgot to cover the multifd send side.

---

Quoted 6 lines
Quoted 1 line

@Leonardo Could you check if this fixes your issue?

In the same scenario I was testing my previous patch,
(http://patchwork.ozlabs.org/project/qemu-devel/patch/[email protected]/)
this patch also seems to fix the issue .
(https://bugzilla.redhat.com/show_bug.cgi?id=1970337).

Regarding this single test:

I am by no means a yank or migration expert, but the change seems to make sense based on what I could learn in the above BZ.

So, FWIW:

Although I think it would be great if a more experienced person could also review.

Best regards,
Leonardo Bras

Quoted 46 lines
Quoted 46 lines


 migration/multifd.c | 6 +++++-
 migration/multifd.h | 2 ++
 2 files changed, 7 insertions(+), 1 deletion(-)
migration/multifd.c +5 -1
index 377da78f5b..5a4f158f3c 100644
@@ -546,6 +546,9 @@ void multifd_save_cleanup(void)
         MultiFDSendParams *p = &multifd_send_state->params[i];
         Error *local_err = NULL;
 
+        if (p->registered_yank) {
+            migration_ioc_unregister_yank(p->c);
+        }
         socket_send_channel_destroy(p->c);
         p->c = NULL;
         qemu_mutex_destroy(&p->mutex);
@@ -813,7 +816,8 @@ static bool multifd_channel_connect(MultiFDSendParams *p,
                 return false;
             }
         } else {
-            /* update for tls qio channel */
+            migration_ioc_register_yank(ioc);
+            p->registered_yank = true;
             p->c = ioc;
             qemu_thread_create(&p->thread, p->name, multifd_send_thread, p,
                                    QEMU_THREAD_JOINABLE);
migration/multifd.h +2 -1
index 8d6751f5ed..16c4d112d1 100644
@@ -85,6 +85,8 @@ typedef struct {
     bool running;
     /* should this thread finish */
     bool quit;
+    /* is the yank function registered */
+    bool registered_yank;
     /* thread has work to do */
     int pending_job;
     /* array of pages to sent */
--

2.32.0