On 06/07/2017 03:43 PM, Juan Quintela wrote:
Alexey Perevalov <a.pereva...@samsung.com> wrote:
This patch adds request to kernel space for UFFD_FEATURE_THREAD_ID,
in case when this feature is provided by kernel.
I think this function is wrong
migration_exit_cb will be called at QEMU exit, but see later
+static void migration_exit_cb(Notifier *n, void *data)
+{
+ PostcopyBlocktimeContext *ctx = container_of(n, PostcopyBlocktimeContext,
+ exit_notifier);
+ destroy_blocktime_context(ctx);
+}
+
+static struct PostcopyBlocktimeContext *blocktime_context_new(void)
+{
+ PostcopyBlocktimeContext *ctx = g_new0(PostcopyBlocktimeContext, 1);
+ ctx->page_fault_vcpu_time = g_new0(int64_t, smp_cpus);
+ ctx->vcpu_addr = g_new0(uint64_t, smp_cpus);
+ ctx->vcpu_blocktime = g_new0(int64_t, smp_cpus);
+
+ ctx->exit_notifier.notify = migration_exit_cb;
+ qemu_add_exit_notifier(&ctx->exit_notifier);
+ add_migration_state_change_notifier(&ctx->postcopy_notifier);
Or you don't want to call it this awy.
This will destroy the context at every migration state change.
Or I am missing something here? Look at ui/spice-core.c to see how to
use it only for some states (I guess you will need to do it for
error/cleanup/completion changes only).
I forgot, to remove
add_migration_state_change_notifier(&ctx->postcopy_notifier);
in previous version, here was callback with migration state check.
Later, Juan.
+ return ctx;
+}
/**
* receive_ufd_features: check userfault fd features, to request only
supported
@@ -155,6 +207,19 @@ static bool ufd_check_and_apply(int ufd,
MigrationIncomingState *mis)
}
}
+#ifdef UFFD_FEATURE_THREAD_ID
+ if (migrate_postcopy_blocktime() && mis &&
+ UFFD_FEATURE_THREAD_ID & supported_features) {
+ /* kernel supports that feature */
+ /* don't create blocktime_context if it exists */
+ if (!mis->blocktime_ctx) {
+ mis->blocktime_ctx = blocktime_context_new();
+ }
+
+ asked_features |= UFFD_FEATURE_THREAD_ID;
+ }
+#endif
+
/*
* request features, even if asked_features is 0, due to
* kernel expects UFFD_API before UFFDIO_REGISTER, per
--
Best regards,
Alexey Perevalov
