Patchwork xen_disk: remove syncwrite option

login
register
mail settings
Submitter Stefano Stabellini
Date April 24, 2012, 11:22 a.m.
Message ID <1335266521-20875-1-git-send-email-stefano.stabellini@eu.citrix.com>
Download mbox | patch
Permalink /patch/154656/
State New
Headers show

Comments

Stefano Stabellini - April 24, 2012, 11:22 a.m.
Use the BDRV_O_CACHE_* flags instead.

Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
---
 hw/xen_disk.c |    8 +-------
 1 files changed, 1 insertions(+), 7 deletions(-)
Kevin Wolf - April 24, 2012, 11:37 a.m.
Am 24.04.2012 13:22, schrieb Stefano Stabellini:
> Use the BDRV_O_CACHE_* flags instead.
> 
> Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>

Doesn't apply to qemu.git because...

> ---
>  hw/xen_disk.c |    8 +-------
>  1 files changed, 1 insertions(+), 7 deletions(-)
> 
> diff --git a/hw/xen_disk.c b/hw/xen_disk.c
> index 4a6d89c..3e4a47b 100644
> --- a/hw/xen_disk.c
> +++ b/hw/xen_disk.c
> @@ -48,7 +48,6 @@
>  
>  /* ------------------------------------------------------------- */
>  
> -static int syncwrite    = 0;
>  static int batch_maps   = 0;
>  
>  static int max_requests = 32;
> @@ -189,15 +188,10 @@ static int ioreq_parse(struct ioreq *ioreq)
>              ioreq->presync = 1;
>              return 0;
>          }
> -        if (!syncwrite) {
> -            ioreq->postsync = 1;

...this is ioreq->presync = ioreq->postsync = 1;

And while we're at it, the commit message could mention that there was
no way to set this flag anyway, so we're just removing dead code.

Kevin
Stefano Stabellini - April 24, 2012, 11:55 a.m.
On Tue, 24 Apr 2012, Kevin Wolf wrote:
> Am 24.04.2012 13:22, schrieb Stefano Stabellini:
> > Use the BDRV_O_CACHE_* flags instead.
> > 
> > Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
> 
> Doesn't apply to qemu.git because...
> 
> > ---
> >  hw/xen_disk.c |    8 +-------
> >  1 files changed, 1 insertions(+), 7 deletions(-)
> > 
> > diff --git a/hw/xen_disk.c b/hw/xen_disk.c
> > index 4a6d89c..3e4a47b 100644
> > --- a/hw/xen_disk.c
> > +++ b/hw/xen_disk.c
> > @@ -48,7 +48,6 @@
> >  
> >  /* ------------------------------------------------------------- */
> >  
> > -static int syncwrite    = 0;
> >  static int batch_maps   = 0;
> >  
> >  static int max_requests = 32;
> > @@ -189,15 +188,10 @@ static int ioreq_parse(struct ioreq *ioreq)
> >              ioreq->presync = 1;
> >              return 0;
> >          }
> > -        if (!syncwrite) {
> > -            ioreq->postsync = 1;
> 
> ...this is ioreq->presync = ioreq->postsync = 1;

That's because it is based on top of the following two patches:

xen_disk: use bdrv_aio_flush instead of bdrv_flush
http://marc.info/?l=qemu-devel&m=133434258008959&w=2

xen_disk: implement BLKIF_OP_FLUSH_DISKCACHE, remove BLKIF_OP_WRITE_BARRIER
http://marc.info/?l=qemu-devel&m=133459581417878&w=2

I have just sent a pull request for them
(http://marc.info/?l=qemu-devel&m=133526669019403) after leaving lying
around for a while.
Next time do you want me to send xen_disk specific pull request to you
instead? Sorry if I shouldn't have sent the pull request directly...


> And while we're at it, the commit message could mention that there was
> no way to set this flag anyway, so we're just removing dead code.

good point
Kevin Wolf - April 24, 2012, noon
Am 24.04.2012 13:55, schrieb Stefano Stabellini:
> On Tue, 24 Apr 2012, Kevin Wolf wrote:
>> Am 24.04.2012 13:22, schrieb Stefano Stabellini:
>>> Use the BDRV_O_CACHE_* flags instead.
>>>
>>> Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
>>
>> Doesn't apply to qemu.git because...
>>
>>> ---
>>>  hw/xen_disk.c |    8 +-------
>>>  1 files changed, 1 insertions(+), 7 deletions(-)
>>>
>>> diff --git a/hw/xen_disk.c b/hw/xen_disk.c
>>> index 4a6d89c..3e4a47b 100644
>>> --- a/hw/xen_disk.c
>>> +++ b/hw/xen_disk.c
>>> @@ -48,7 +48,6 @@
>>>  
>>>  /* ------------------------------------------------------------- */
>>>  
>>> -static int syncwrite    = 0;
>>>  static int batch_maps   = 0;
>>>  
>>>  static int max_requests = 32;
>>> @@ -189,15 +188,10 @@ static int ioreq_parse(struct ioreq *ioreq)
>>>              ioreq->presync = 1;
>>>              return 0;
>>>          }
>>> -        if (!syncwrite) {
>>> -            ioreq->postsync = 1;
>>
>> ...this is ioreq->presync = ioreq->postsync = 1;
> 
> That's because it is based on top of the following two patches:
> 
> xen_disk: use bdrv_aio_flush instead of bdrv_flush
> http://marc.info/?l=qemu-devel&m=133434258008959&w=2
> 
> xen_disk: implement BLKIF_OP_FLUSH_DISKCACHE, remove BLKIF_OP_WRITE_BARRIER
> http://marc.info/?l=qemu-devel&m=133459581417878&w=2
> 
> I have just sent a pull request for them
> (http://marc.info/?l=qemu-devel&m=133526669019403) after leaving lying
> around for a while.
> Next time do you want me to send xen_disk specific pull request to you
> instead? Sorry if I shouldn't have sent the pull request directly...

I see. I wasn't aware that you're doing pull requests, but it isn't a
problem. It just means that you should probably do a pull request for
this one as well instead of expecting that I pick it up. But you can
have my Acked-by, if you like.

Should I ignore xen_disk patches from now on for the block branch and
assume that you pick them up?

Kevin
Stefano Stabellini - April 24, 2012, 12:43 p.m.
On Tue, 24 Apr 2012, Kevin Wolf wrote:
> > That's because it is based on top of the following two patches:
> > 
> > xen_disk: use bdrv_aio_flush instead of bdrv_flush
> > http://marc.info/?l=qemu-devel&m=133434258008959&w=2
> > 
> > xen_disk: implement BLKIF_OP_FLUSH_DISKCACHE, remove BLKIF_OP_WRITE_BARRIER
> > http://marc.info/?l=qemu-devel&m=133459581417878&w=2
> > 
> > I have just sent a pull request for them
> > (http://marc.info/?l=qemu-devel&m=133526669019403) after leaving lying
> > around for a while.
> > Next time do you want me to send xen_disk specific pull request to you
> > instead? Sorry if I shouldn't have sent the pull request directly...
> 
> I see. I wasn't aware that you're doing pull requests, but it isn't a
> problem. It just means that you should probably do a pull request for
> this one as well instead of expecting that I pick it up. But you can
> have my Acked-by, if you like.
> 
> Should I ignore xen_disk patches from now on for the block branch and
> assume that you pick them up?

I am happy to issue pull requests for xen_disk patches, but I still
welcome your reviews :-)

Patch

diff --git a/hw/xen_disk.c b/hw/xen_disk.c
index 4a6d89c..3e4a47b 100644
--- a/hw/xen_disk.c
+++ b/hw/xen_disk.c
@@ -48,7 +48,6 @@ 
 
 /* ------------------------------------------------------------- */
 
-static int syncwrite    = 0;
 static int batch_maps   = 0;
 
 static int max_requests = 32;
@@ -189,15 +188,10 @@  static int ioreq_parse(struct ioreq *ioreq)
             ioreq->presync = 1;
             return 0;
         }
-        if (!syncwrite) {
-            ioreq->postsync = 1;
-        }
+        ioreq->postsync = 1;
         /* fall through */
     case BLKIF_OP_WRITE:
         ioreq->prot = PROT_READ; /* from memory */
-        if (syncwrite) {
-            ioreq->postsync = 1;
-        }
         break;
     default:
         xen_be_printf(&blkdev->xendev, 0, "error: unknown operation (%d)\n",