diff mbox

ide save/restore current transfer fields

Message ID 1264006909-22040-1-git-send-email-quintela@redhat.com
State New
Headers show

Commit Message

Juan Quintela Jan. 20, 2010, 5:01 p.m. UTC 
From: Marcelo Tosatti <mtosatti@redhat.com>

If migration takes place between write of the bmdma address register and
write of the command register (to initiate DMA), the destination will
not properly start the DMA op, hanging the guest:

ata1.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x6 frozen
ata1.00: cmd c8/00:16:41:00:00/00:00:00:00:00/e0 tag 0 dma 11264 in
         res 40/00:00:00:00:00/00:00:00:00:00/00 Emask 0x4 (timeout)
ata1.00: status: { DRDY }

Fix by sending current transfer information in the migration data.

We need to update ide version to 4 for this to work.  As we don't
have subsectios, we need to chain the update increase until
vmstate_ide_pci (quintela)

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
---
 hw/ide/pci.c |    8 ++++++--
 1 files changed, 6 insertions(+), 2 deletions(-)

Comments

Anthony Liguori Jan. 26, 2010, 10:02 p.m. UTC | #1 
On 01/20/2010 11:01 AM, Juan Quintela wrote:
> From: Marcelo Tosatti<mtosatti@redhat.com>
>
> If migration takes place between write of the bmdma address register and
> write of the command register (to initiate DMA), the destination will
> not properly start the DMA op, hanging the guest:
>
> ata1.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x6 frozen
> ata1.00: cmd c8/00:16:41:00:00/00:00:00:00:00/e0 tag 0 dma 11264 in
>           res 40/00:00:00:00:00/00:00:00:00:00/00 Emask 0x4 (timeout)
> ata1.00: status: { DRDY }
>
> Fix by sending current transfer information in the migration data.
>
> We need to update ide version to 4 for this to work.  As we don't
> have subsectios, we need to chain the update increase until
> vmstate_ide_pci (quintela)
>
> Signed-off-by: Marcelo Tosatti<mtosatti@redhat.com>
> Signed-off-by: Juan Quintela<quintela@redhat.com>
>    

Applied.  Thanks.

This is one of those nasty stable situations.  If we backport to stable, 
we fix a potentially bug but we max migration within stable impossible.  
We really ought to blacklist the old version too without an explicit 
override.

Any thoughts on how to handle this?  We at least have to record this as 
an issue.

Regards,

Anthony Liguori
diff mbox

Patch

diff --git a/hw/ide/pci.c b/hw/ide/pci.c
index 780fc5f..4d95cc5 100644
--- a/hw/ide/pci.c
+++ b/hw/ide/pci.c
@@ -123,7 +123,7 @@  void bmdma_addr_writel(void *opaque, uint32_t addr, uint32_t val)

 static const VMStateDescription vmstate_bmdma = {
     .name = "ide bmdma",
-    .version_id = 3,
+    .version_id = 4,
     .minimum_version_id = 0,
     .minimum_version_id_old = 0,
     .fields      = (VMStateField []) {
@@ -133,6 +133,10 @@  static const VMStateDescription vmstate_bmdma = {
         VMSTATE_INT64(sector_num, BMDMAState),
         VMSTATE_UINT32(nsector, BMDMAState),
         VMSTATE_UINT8(unit, BMDMAState),
+        VMSTATE_UINT32_V(cur_addr, BMDMAState, 4),
+        VMSTATE_UINT32_V(cur_prd_last, BMDMAState, 4),
+        VMSTATE_UINT32_V(cur_prd_addr, BMDMAState, 4),
+        VMSTATE_UINT32_V(cur_prd_len, BMDMAState, 4),
         VMSTATE_END_OF_LIST()
     }
 };
@@ -152,7 +156,7 @@  static int ide_pci_post_load(void *opaque, int version_id)

 const VMStateDescription vmstate_ide_pci = {
     .name = "ide",
-    .version_id = 3,
+    .version_id = 4,
     .minimum_version_id = 0,
     .minimum_version_id_old = 0,
     .post_load = ide_pci_post_load,