Re: [Qemu-devel] [PATCH] Fix pflash migration

[Laszlo Ersek <lersek@redhat.com>]

On 04/15/16 13:41, Dr. David Alan Gilbert (git) wrote:
> From: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
> 
> Pflash migration (e.g. q35 + EFI variable storage) fails
> with the assert:
> 
> bdrv_co_do_pwritev: Assertion `!(bs->open_flags & 0x0800)' failed.
> 
> This avoids the problem by delaying the pflash update until after
> the device loads complete.
> 
> Tested by:
>   Migrating Q35/EFI vm.
>   Changing efi variable content (with efiboot in the guest)
>   md5sum'ing the variable file before migration and after.
> 
> This is a fix that Paolo posted in the message
>   570244B3.4070105@redhat.com
> 
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
> ---
>  hw/block/pflash_cfi01.c | 18 ++++++++++++++++--
>  1 file changed, 16 insertions(+), 2 deletions(-)
> 
> diff --git a/hw/block/pflash_cfi01.c b/hw/block/pflash_cfi01.c
> index c475c2a..106a775 100644
> --- a/hw/block/pflash_cfi01.c
> +++ b/hw/block/pflash_cfi01.c
> @@ -46,6 +46,7 @@
>  #include "exec/address-spaces.h"
>  #include "qemu/host-utils.h"
>  #include "hw/sysbus.h"
> +#include "sysemu/sysemu.h"
>  
>  #define PFLASH_BUG(fmt, ...) \
>  do { \
> @@ -97,6 +98,7 @@ struct pflash_t {
>      MemoryRegion mem;
>      char *name;
>      void *storage;
> +    VMChangeStateEntry *vmstate;
>  };
>  
>  static int pflash_post_load(void *opaque, int version_id);
> @@ -944,13 +946,25 @@ MemoryRegion *pflash_cfi01_get_memory(pflash_t *fl)
>      return &fl->mem;
>  }
>  
> +static void postload_update_cb(void *opaque, int running, RunState state)
> +{
> +    pflash_t *pfl = opaque;
> +
> +    /* This is called after bdrv_invalidate_cache_all.  */
> +    qemu_del_vm_change_state_handler(pfl->vmstate);
> +    pfl->vmstate = NULL;
> +
> +    DPRINTF("%s: updating bdrv for %s\n", __func__, pfl->name);
> +    pflash_update(pfl, 0, pfl->sector_len * pfl->nb_blocs);
> +}
> +
>  static int pflash_post_load(void *opaque, int version_id)
>  {
>      pflash_t *pfl = opaque;
>  
>      if (!pfl->ro) {
> -        DPRINTF("%s: updating bdrv for %s\n", __func__, pfl->name);
> -        pflash_update(pfl, 0, pfl->sector_len * pfl->nb_blocs);
> +        pfl->vmstate = qemu_add_vm_change_state_handler(postload_update_cb,
> +                                                        pfl);
>      }
>      return 0;
>  }
> 

I can't offer an in-depth review (I don't know migration internals /
vmstate changes well enough, sorry), but I agree with the fix (and am
thankful for it), and both the patch and the testing look good to me.

Perhaps some assert()s could be added to postload_update_cb(), about
"running" and "state", but I'm not certain.

Acked-by: Laszlo Ersek <lersek@redhat.com>