qemu: fix physical memory migration

Commit Message

Yaniv Kamay March 24, 2009, 8:51 p.m. UTC

Hi,

Attaching patch that:

    1. Fix physical memory live migration.
        - Stop dirty memory tracking after completion of memory transfer.
        - In stage 3, updating  dirty memory bits before collecting  
remaining dirty blocks
             in order to prevent missing dirty blokes in the range of 
current_addr to phys_ram_size while
             no dirty blocks exist in the range 0 - (current_addr - 1)
     
    2. Improve migration error handling

Thanks,
Yaniv

Comments

Avi Kivity March 25, 2009, 12:05 p.m. UTC | #1

Yaniv Kamay wrote:
> Hi,
>
> Attaching patch that:
>
>    1. Fix physical memory live migration.
>        - Stop dirty memory tracking after completion of memory transfer.
>        - In stage 3, updating  dirty memory bits before collecting  
> remaining dirty blocks
>             in order to prevent missing dirty blokes in the range of 
> current_addr to phys_ram_size while
>             no dirty blocks exist in the range 0 - (current_addr - 1)
>        2. Improve migration error handling

Please split into three separate patches.

>
> +void qemu_file_set_has_error(QEMUFile *f);

Better named qemu_file_set_error()?

> diff --git a/qemu/vl.c b/qemu/vl.c
> index 7ae266e..405212f 100644
> --- a/qemu/vl.c
> +++ b/qemu/vl.c
> @@ -3221,8 +3221,14 @@ static int ram_save_block(QEMUFile *f)
>      int found = 0;
>  
>      while (addr < phys_ram_size) {
> -        if (kvm_enabled() && current_addr == 0)
> -            kvm_update_dirty_pages_log(); /* FIXME: propagate errors */
> +        if (kvm_enabled() && current_addr == 0) {
> +            int r;
> +            if ((r = kvm_update_dirty_pages_log())) {

Compare against negative; it could return positive numbers in the future 
even on no error.

> +                printf("%s: update dirty pages log failed %d\n", 
> __FUNCTION__, r);

qemu_log (in qemu-log.h)

Patch

From d565b6b2246b0ac7c8ca292680f9134edb35caa7 Mon Sep 17 00:00:00 2001
From: Yaniv Kamay <[e-mail]>
Date: Tue, 24 Mar 2009 20:12:35 +0200
Subject: [PATCH] qemu: fix physical memory migration

---
 qemu/hw/hw.h     |    1 +
 qemu/migration.c |    7 ++++++-
 qemu/savevm.c    |    5 +++++
 qemu/vl.c        |   19 +++++++++++++++----
 4 files changed, 27 insertions(+), 5 deletions(-)

diff --git a/qemu/hw/hw.h b/qemu/hw/hw.h
index eab7bb4..6aad964 100644
--- a/qemu/hw/hw.h
+++ b/qemu/hw/hw.h
@@ -67,6 +67,7 @@  unsigned int qemu_get_be32(QEMUFile *f);
 uint64_t qemu_get_be64(QEMUFile *f);
 int qemu_file_rate_limit(QEMUFile *f);
 int qemu_file_has_error(QEMUFile *f);
+void qemu_file_set_has_error(QEMUFile *f);
 
 /* Try to send any outstanding data.  This function is useful when output is
  * halted due to rate limiting or EAGAIN errors occur as it can be used to
diff --git a/qemu/migration.c b/qemu/migration.c
index b3904b2..1cee6e9 100644
--- a/qemu/migration.c
+++ b/qemu/migration.c
@@ -225,7 +225,12 @@  void migrate_fd_put_ready(void *opaque)
 
         bdrv_flush_all();
         qemu_savevm_state_complete(s->file);
-        s->state = MIG_STATE_COMPLETED;
+        if (qemu_file_has_error(s->file)) {
+            vm_start();
+            s->state = MIG_STATE_ERROR;
+        } else {
+            s->state = MIG_STATE_COMPLETED;
+        }
         migrate_fd_cleanup(s);
     }
 }
diff --git a/qemu/savevm.c b/qemu/savevm.c
index 72c3709..7cd312d 100644
--- a/qemu/savevm.c
+++ b/qemu/savevm.c
@@ -370,6 +370,11 @@  int qemu_file_has_error(QEMUFile *f)
     return f->has_error;
 }
 
+void qemu_file_set_has_error(QEMUFile *f)
+{
+    f->has_error = 1;
+}
+
 void qemu_fflush(QEMUFile *f)
 {
     if (!f->put_buffer)
diff --git a/qemu/vl.c b/qemu/vl.c
index 7ae266e..405212f 100644
--- a/qemu/vl.c
+++ b/qemu/vl.c
@@ -3221,8 +3221,14 @@  static int ram_save_block(QEMUFile *f)
     int found = 0;
 
     while (addr < phys_ram_size) {
-        if (kvm_enabled() && current_addr == 0)
-            kvm_update_dirty_pages_log(); /* FIXME: propagate errors */
+        if (kvm_enabled() && current_addr == 0) {
+            int r;
+            if ((r = kvm_update_dirty_pages_log())) {
+                printf("%s: update dirty pages log failed %d\n", __FUNCTION__, r);
+                qemu_file_set_has_error(f); // for now: replacing FIXME with ugly hack
+                return 0;
+            }
+        }
         if (cpu_physical_memory_get_dirty(current_addr, MIGRATION_DIRTY_FLAG)) {
             uint8_t ch;
 
@@ -3293,10 +3299,15 @@  static int ram_save_live(QEMUFile *f, int stage, void *opaque)
     /* try transferring iterative blocks of memory */
 
     if (stage == 3) {
-        cpu_physical_memory_set_dirty_tracking(0);
-
+        int r;
+        if ((r = kvm_update_dirty_pages_log())) {
+            printf("%s: update dirty pages log failed %d\n", __FUNCTION__, r);
+            qemu_file_set_has_error(f);
+            return 0;
+        }
         /* flush all remaining blocks regardless of rate limiting */
         while (ram_save_block(f) != 0);
+        cpu_physical_memory_set_dirty_tracking(0);
     }
 
     qemu_put_be64(f, RAM_SAVE_FLAG_EOS);
-- 
1.6.0.2