We synchronize all threads each RAM_SAVE_FLAG_EOS. Bitmap synchronizations don't happen inside a ram section, so we are safe about two channels trying to overwrite the same memory.
Signed-off-by: Juan Quintela <quint...@redhat.com> --- migration/ram.c | 37 ++++++++++++++++++++++++++++++++++++- migration/trace-events | 1 + 2 files changed, 37 insertions(+), 1 deletion(-) diff --git a/migration/ram.c b/migration/ram.c index 153c7560cb..0266bd200c 100644 --- a/migration/ram.c +++ b/migration/ram.c @@ -551,6 +551,7 @@ struct MultiFDRecvParams { QemuMutex mutex; bool running; bool quit; + bool sync; }; typedef struct MultiFDRecvParams MultiFDRecvParams; @@ -558,6 +559,8 @@ struct { MultiFDRecvParams *params; /* number of created threads */ int count; + /* syncs main thread and channels */ + QemuSemaphore sem_main; } *multifd_recv_state; static void terminate_multifd_recv_threads(Error *errp) @@ -605,6 +608,7 @@ int multifd_load_cleanup(Error **errp) g_free(p->name); p->name = NULL; } + qemu_sem_destroy(&multifd_recv_state->sem_main); g_free(multifd_recv_state->params); multifd_recv_state->params = NULL; g_free(multifd_recv_state); @@ -613,18 +617,45 @@ int multifd_load_cleanup(Error **errp) return ret; } +static void multifd_recv_sync_main(void) +{ + int i; + + if (!migrate_use_multifd()) { + return; + } + for (i = 0; i < migrate_multifd_channels(); i++) { + MultiFDRecvParams *p = &multifd_recv_state->params[i]; + + qemu_mutex_lock(&p->mutex); + p->sync = true; + qemu_mutex_unlock(&p->mutex); + qemu_sem_post(&p->sem); + } + for (i = 0; i < migrate_multifd_channels(); i++) { + qemu_sem_wait(&multifd_recv_state->sem_main); + } + trace_multifd_recv_sync_main(); +} + static void *multifd_recv_thread(void *opaque) { MultiFDRecvParams *p = opaque; while (true) { + qemu_sem_wait(&p->sem); qemu_mutex_lock(&p->mutex); + if (p->sync) { + p->sync = false; + qemu_mutex_unlock(&p->mutex); + qemu_sem_post(&multifd_recv_state->sem_main); + continue; + } if (p->quit) { qemu_mutex_unlock(&p->mutex); break; } qemu_mutex_unlock(&p->mutex); - qemu_sem_wait(&p->sem); } return NULL; @@ -642,6 +673,7 @@ int multifd_load_setup(void) multifd_recv_state = g_malloc0(sizeof(*multifd_recv_state)); multifd_recv_state->params = g_new0(MultiFDRecvParams, thread_count); atomic_set(&multifd_recv_state->count, 0); + qemu_sem_init(&multifd_recv_state->sem_main, 0); for (i = 0; i < thread_count; i++) { MultiFDRecvParams *p = &multifd_recv_state->params[i]; @@ -655,6 +687,7 @@ int multifd_load_setup(void) QEMU_THREAD_JOINABLE); atomic_inc(&multifd_recv_state->count); } + return 0; } @@ -2868,6 +2901,7 @@ static int ram_load_postcopy(QEMUFile *f) break; case RAM_SAVE_FLAG_EOS: /* normal exit */ + multifd_recv_sync_main(); break; default: error_report("Unknown combination of migration flags: %#x" @@ -3053,6 +3087,7 @@ static int ram_load(QEMUFile *f, void *opaque, int version_id) break; case RAM_SAVE_FLAG_EOS: /* normal exit */ + multifd_recv_sync_main(); break; default: if (flags & RAM_SAVE_FLAG_HOOK) { diff --git a/migration/trace-events b/migration/trace-events index 97b5ac564f..76075c26bc 100644 --- a/migration/trace-events +++ b/migration/trace-events @@ -78,6 +78,7 @@ ram_postcopy_send_discard_bitmap(void) "" ram_save_page(const char *rbname, uint64_t offset, void *host) "%s: offset: 0x%" PRIx64 " host: %p" ram_save_queue_pages(const char *rbname, size_t start, size_t len) "%s: start: 0x%zx len: 0x%zx" multifd_send_sync_main(void) "" +multifd_recv_sync_main(void) "" # migration/migration.c await_return_path_close_on_source_close(void) "" -- 2.14.3