Re: [Qemu-devel] [sheepdog] [PATCH 3/3] sheepdog: resend write requests when SD_RES_READONLY is received

2013-04-25 Thread MORITA Kazutaka
At Thu, 25 Apr 2013 21:08:01 +0800,
Liu Yuan wrote:
> 
> On 04/25/2013 06:37 PM, MORITA Kazutaka wrote:
> > +/* update inode with the latest state */
> > +static int coroutine_fn reload_vdi_object(BDRVSheepdogState *s)
> 
> I'd suggest function name as
> 'reload_inode(BDRVSheepdogState *s, tag, snapid)', then sd_create_branch
> and sd_snapshot_goto can make use of this function. With this change, it
> would be conflicted to my patch series applied to Stefan's block tree +
> Fix loadvm patch. So it would be better you can develop this patch set
> against Stefan's block tree + my fix loadvm patch.

sd_create_branch() doesn't use find_vdi_name() to get vid, so
reload_inode() is not suitable for it.  After all, I used
reload_inode() only for sd_snapshot_goto() and resend_aioreq() in v2,
so our patches didn't conflict.

Thanks,

Kazutaka



Re: [Qemu-devel] [sheepdog] [PATCH 3/3] sheepdog: resend write requests when SD_RES_READONLY is received

2013-04-25 Thread Liu Yuan
On 04/25/2013 06:37 PM, MORITA Kazutaka wrote:
> +/* update inode with the latest state */
> +static int coroutine_fn reload_vdi_object(BDRVSheepdogState *s)

I'd suggest function name as
'reload_inode(BDRVSheepdogState *s, tag, snapid)', then sd_create_branch
and sd_snapshot_goto can make use of this function. With this change, it
would be conflicted to my patch series applied to Stefan's block tree +
Fix loadvm patch. So it would be better you can develop this patch set
against Stefan's block tree + my fix loadvm patch.

Others looks good to me.

Thanks,
Yuan