On Sat, Sep 11, 2010 at 3:04 PM, Anthony Liguori <aligu...@us.ibm.com> wrote:
> This fixes a couple nasty problems relating to live migration.
>
> 1) When dealing with shared storage with weak coherence (i.e. NFS), even if
> we re-read, we may end up with undesired caching. By delaying any reads
> until we absolutely have to, we decrease the likelihood of any undesirable
> caching.
>
> 2) When dealing with copy-on-read, the local storage acts as a cache. We need
> to make sure to avoid any reads to avoid polluting the local cache.
>
> Signed-off-by: Anthony Liguori <aligu...@us.ibm.com>
>
> diff --git a/hw/ide/core.c b/hw/ide/core.c
> index 1e466d1..57d8db3 100644
> --- a/hw/ide/core.c
> +++ b/hw/ide/core.c
> @@ -105,6 +132,8 @@ static void ide_identify(IDEState *s)
> return;
> }
>
> + guess_geometry(s);
> +
Does the same change need to be made in ide_cfata_identify()?
I quickly checked the VMStateDescription and don't see cylinders,
heads, sectors being saved for migration. I am concerned that IDE
will break after migration if the following happens:
1. Guest resumes and does not issue ATA IDENTIFY so cylinders, heads,
sectors are not initialized.
2. Normal I/O is performed, invoking ide_get_sector() which uses
geometry information that has not been initialized.
Did I miss something?
> diff --git a/hw/virtio-blk.c b/hw/virtio-blk.c
> index bd6bbe6..0bf17ec 100644
> --- a/hw/virtio-blk.c
> +++ b/hw/virtio-blk.c
> @@ -427,6 +427,10 @@ static void virtio_blk_update_config(VirtIODevice *vdev,
> uint8_t *config)
>
> bdrv_get_geometry(s->bs, &capacity);
> bdrv_get_geometry_hint(s->bs, &cylinders, &heads, &secs);
> + if (cylinders == 0) {
> + bdrv_guess_geometry(s->bs, &cylinders, &heads, &secs);
> + }
> +
bdrv_guess_geometry() can be called unconditionally. The call to
bdrv_get_geometry_hint() can be eliminated. bdrv_guess_geometry()
updates the geometry hint and does not probe the boot sector after the
first time.
Stefan
