[pve-devel] [PATCH qemu 1/2] fix #4476: savevm-async: avoid looping without progress

[Fiona Ebner]

when pend_postcopy is large. By definition, pend_postcopy won't
decrease when iterating, so a value larger than the cutoff of 400000
would lead to essentially empty iterations, filling up the state file
until only 30 MiB + pending_size remain and the second half of the
check would trigger.

Avoid this, by not considering pend_postcopy for the cutoff to enter
the final phase.

Signed-off-by: Fiona Ebner <f.ebner at proxmox.com>
---

Many thanks to Fabian for discussing the issue with me!

 ...vevm-async-for-background-state-snapshots.patch | 14 ++++++++------
 ...-PVE-add-optional-buffer-size-to-QEMUFile.patch |  6 +++---
 ...async-register-yank-before-migration_inco.patch |  4 ++--
 3 files changed, 13 insertions(+), 11 deletions(-)

diff --git a/debian/patches/pve/0017-PVE-add-savevm-async-for-background-state-snapshots.patch b/debian/patches/pve/0017-PVE-add-savevm-async-for-background-state-snapshots.patch
index fcbe4bc..91d4710 100644
--- a/debian/patches/pve/0017-PVE-add-savevm-async-for-background-state-snapshots.patch
+++ b/debian/patches/pve/0017-PVE-add-savevm-async-for-background-state-snapshots.patch
@@ -24,7 +24,8 @@ Signed-off-by: Wolfgang Bumiller <w.bumiller at proxmox.com>
 [improve aborting]
 Signed-off-by: Stefan Reiter <s.reiter at proxmox.com>
 [FE: further improve aborting
-     adapt to removal of QEMUFileOps]
+     adapt to removal of QEMUFileOps
+     improve condition for entering final stage]
 Signed-off-by: Fiona Ebner <f.ebner at proxmox.com>
 ---
  hmp-commands-info.hx         |  13 +
@@ -32,13 +33,13 @@ Signed-off-by: Fiona Ebner <f.ebner at proxmox.com>
  include/migration/snapshot.h |   2 +
  include/monitor/hmp.h        |   5 +
  migration/meson.build        |   1 +
- migration/savevm-async.c     | 531 +++++++++++++++++++++++++++++++++++
+ migration/savevm-async.c     | 532 +++++++++++++++++++++++++++++++++++
  monitor/hmp-cmds.c           |  57 ++++
  qapi/migration.json          |  34 +++
  qapi/misc.json               |  32 +++
  qemu-options.hx              |  12 +
  softmmu/vl.c                 |  10 +
- 11 files changed, 730 insertions(+)
+ 11 files changed, 731 insertions(+)
  create mode 100644 migration/savevm-async.c
 
 diff --git a/hmp-commands-info.hx b/hmp-commands-info.hx
@@ -154,10 +155,10 @@ index 8cac83c06c..0842d00cd2 100644
  ), gnutls)
 diff --git a/migration/savevm-async.c b/migration/savevm-async.c
 new file mode 100644
-index 0000000000..05d394c0e2
+index 0000000000..4a4e91a26d
 --- /dev/null
 +++ b/migration/savevm-async.c
-@@ -0,0 +1,531 @@
+@@ -0,0 +1,532 @@
 +#include "qemu/osdep.h"
 +#include "migration/channel-savevm-async.h"
 +#include "migration/migration.h"
@@ -415,7 +416,8 @@ index 0000000000..05d394c0e2
 +
 +        maxlen = blk_getlength(snap_state.target) - 30*1024*1024;
 +
-+        if (pending_size > 400000 && snap_state.bs_pos + pending_size < maxlen) {
++        /* Note that there is no progress for pend_postcopy when iterating */
++        if (pending_size - pend_postcopy > 400000 && snap_state.bs_pos + pending_size < maxlen) {
 +            ret = qemu_savevm_state_iterate(snap_state.file, false);
 +            if (ret < 0) {
 +                save_snapshot_error("qemu_savevm_state_iterate error %d", ret);
diff --git a/debian/patches/pve/0018-PVE-add-optional-buffer-size-to-QEMUFile.patch b/debian/patches/pve/0018-PVE-add-optional-buffer-size-to-QEMUFile.patch
index 5cf894b..956a09d 100644
--- a/debian/patches/pve/0018-PVE-add-optional-buffer-size-to-QEMUFile.patch
+++ b/debian/patches/pve/0018-PVE-add-optional-buffer-size-to-QEMUFile.patch
@@ -192,10 +192,10 @@ index fa13d04d78..914f1a63a8 100644
  int qemu_fclose(QEMUFile *f);
  
 diff --git a/migration/savevm-async.c b/migration/savevm-async.c
-index 05d394c0e2..bafe6ae5eb 100644
+index 4a4e91a26d..20aae335ca 100644
 --- a/migration/savevm-async.c
 +++ b/migration/savevm-async.c
-@@ -367,7 +367,7 @@ void qmp_savevm_start(bool has_statefile, const char *statefile, Error **errp)
+@@ -368,7 +368,7 @@ void qmp_savevm_start(bool has_statefile, const char *statefile, Error **errp)
  
      QIOChannel *ioc = QIO_CHANNEL(qio_channel_savevm_async_new(snap_state.target,
                                                                 &snap_state.bs_pos));
@@ -204,7 +204,7 @@ index 05d394c0e2..bafe6ae5eb 100644
  
      if (!snap_state.file) {
          error_set(errp, ERROR_CLASS_GENERIC_ERROR, "failed to open '%s'", statefile);
-@@ -500,7 +500,8 @@ int load_snapshot_from_blockdev(const char *filename, Error **errp)
+@@ -501,7 +501,8 @@ int load_snapshot_from_blockdev(const char *filename, Error **errp)
      blk_op_block_all(be, blocker);
  
      /* restore the VM state */
diff --git a/debian/patches/pve/0049-PVE-savevm-async-register-yank-before-migration_inco.patch b/debian/patches/pve/0049-PVE-savevm-async-register-yank-before-migration_inco.patch
index b25c9fc..746b735 100644
--- a/debian/patches/pve/0049-PVE-savevm-async-register-yank-before-migration_inco.patch
+++ b/debian/patches/pve/0049-PVE-savevm-async-register-yank-before-migration_inco.patch
@@ -11,7 +11,7 @@ Signed-off-by: Thomas Lamprecht <t.lamprecht at proxmox.com>
  1 file changed, 5 insertions(+)
 
 diff --git a/migration/savevm-async.c b/migration/savevm-async.c
-index bafe6ae5eb..da3634048f 100644
+index 20aae335ca..94c5ae1c81 100644
 --- a/migration/savevm-async.c
 +++ b/migration/savevm-async.c
 @@ -20,6 +20,7 @@
@@ -22,7 +22,7 @@ index bafe6ae5eb..da3634048f 100644
  
  /* #define DEBUG_SAVEVM_STATE */
  
-@@ -514,6 +515,10 @@ int load_snapshot_from_blockdev(const char *filename, Error **errp)
+@@ -515,6 +516,10 @@ int load_snapshot_from_blockdev(const char *filename, Error **errp)
      dirty_bitmap_mig_before_vm_start();
  
      qemu_fclose(f);
-- 
2.30.2