Patchwork Error on O_DIRECT for physical CDROM/DVD drives

login
register
mail settings
Submitter Jes Sorensen
Date July 20, 2010, 4:09 p.m.
Message ID <1279642177-21229-1-git-send-email-Jes.Sorensen@redhat.com>
Download mbox | patch
Permalink /patch/59343/
State New
Headers show

Comments

Jes Sorensen - July 20, 2010, 4:09 p.m.
From: Jes Sorensen <Jes.Sorensen@redhat.com>

O_DIRECT (cache=none) requires sector alignment, however the physical
sector size of CDROM/DVD drives is 2048, as opposed to most disk
devices which use 512. QEMU is hard coding 512 all over the place, so
allowing O_DIRECT for CDROM/DVD devices does not work. Exit with an
error!

Signed-off-by: Jes Sorensen <Jes.Sorensen@redhat.com>
---
 block/raw-posix.c |    5 +++++
 1 files changed, 5 insertions(+), 0 deletions(-)
David S. Ahern - July 20, 2010, 4:35 p.m.
On 07/20/10 10:09, Jes.Sorensen@redhat.com wrote:
> From: Jes Sorensen <Jes.Sorensen@redhat.com>
> 
> O_DIRECT (cache=none) requires sector alignment, however the physical
> sector size of CDROM/DVD drives is 2048, as opposed to most disk
> devices which use 512. QEMU is hard coding 512 all over the place, so
> allowing O_DIRECT for CDROM/DVD devices does not work. Exit with an
> error!
> 
> Signed-off-by: Jes Sorensen <Jes.Sorensen@redhat.com>
> ---
>  block/raw-posix.c |    5 +++++
>  1 files changed, 5 insertions(+), 0 deletions(-)
> 
> diff --git a/block/raw-posix.c b/block/raw-posix.c
> index 291699f..1b840c4 100644
> --- a/block/raw-posix.c
> +++ b/block/raw-posix.c
> @@ -1139,6 +1139,11 @@ static int cdrom_open(BlockDriverState *bs, const char *filename, int flags)
>      BDRVRawState *s = bs->opaque;
>  
>      s->type = FTYPE_CD;
> +    if (flags & BDRV_O_NOCACHE) {
> +        fprintf(stderr, "O_DIRECT (cache=none) for CDROM/DVD device (%s) "
> +                "is unsupported\n", filename);
> +        exit(1);
> +    }

If a cdrom is added via the monitor this would abruptly terminate the VM
- which is not good.

David

>  
>      /* open will not fail even if no CD is inserted, so add O_NONBLOCK */
>      return raw_open_common(bs, filename, flags, O_NONBLOCK);
Jes Sorensen - July 20, 2010, 4:44 p.m.
On 07/20/10 18:35, David S. Ahern wrote:
> On 07/20/10 10:09, Jes.Sorensen@redhat.com wrote:
>> diff --git a/block/raw-posix.c b/block/raw-posix.c
>> index 291699f..1b840c4 100644
>> --- a/block/raw-posix.c
>> +++ b/block/raw-posix.c
>> @@ -1139,6 +1139,11 @@ static int cdrom_open(BlockDriverState *bs, const char *filename, int flags)
>>      BDRVRawState *s = bs->opaque;
>>  
>>      s->type = FTYPE_CD;
>> +    if (flags & BDRV_O_NOCACHE) {
>> +        fprintf(stderr, "O_DIRECT (cache=none) for CDROM/DVD device (%s) "
>> +                "is unsupported\n", filename);
>> +        exit(1);
>> +    }
> 
> If a cdrom is added via the monitor this would abruptly terminate the VM
> - which is not good.

True, but this would only happen if the user explicitly marks it
cache=none which is not the default value for a new drive, so it would
be a corner case.

Cheers,
Jes
Markus Armbruster - July 20, 2010, 5:19 p.m.
Jes Sorensen <Jes.Sorensen@redhat.com> writes:

> On 07/20/10 18:35, David S. Ahern wrote:
>> On 07/20/10 10:09, Jes.Sorensen@redhat.com wrote:
>>> diff --git a/block/raw-posix.c b/block/raw-posix.c
>>> index 291699f..1b840c4 100644
>>> --- a/block/raw-posix.c
>>> +++ b/block/raw-posix.c
>>> @@ -1139,6 +1139,11 @@ static int cdrom_open(BlockDriverState *bs, const char *filename, int flags)
>>>      BDRVRawState *s = bs->opaque;
>>>  
>>>      s->type = FTYPE_CD;
>>> +    if (flags & BDRV_O_NOCACHE) {
>>> +        fprintf(stderr, "O_DIRECT (cache=none) for CDROM/DVD device (%s) "
>>> +                "is unsupported\n", filename);
>>> +        exit(1);
>>> +    }
>> 
>> If a cdrom is added via the monitor this would abruptly terminate the VM
>> - which is not good.
>
> True, but this would only happen if the user explicitly marks it
> cache=none which is not the default value for a new drive, so it would
> be a corner case.

Why not return -ENOTSUP?
Jes Sorensen - July 21, 2010, 7:47 a.m.
On 07/20/10 19:19, Markus Armbruster wrote:
> Jes Sorensen <Jes.Sorensen@redhat.com> writes:
> 
>> On 07/20/10 18:35, David S. Ahern wrote:
>>> On 07/20/10 10:09, Jes.Sorensen@redhat.com wrote:
>>> If a cdrom is added via the monitor this would abruptly terminate the VM
>>> - which is not good.
>>
>> True, but this would only happen if the user explicitly marks it
>> cache=none which is not the default value for a new drive, so it would
>> be a corner case.
> 
> Why not return -ENOTSUP?

Good point, just posted v3 of the patch that does this instead of exit()

Jes

Patch

diff --git a/block/raw-posix.c b/block/raw-posix.c
index 291699f..1b840c4 100644
--- a/block/raw-posix.c
+++ b/block/raw-posix.c
@@ -1139,6 +1139,11 @@  static int cdrom_open(BlockDriverState *bs, const char *filename, int flags)
     BDRVRawState *s = bs->opaque;
 
     s->type = FTYPE_CD;
+    if (flags & BDRV_O_NOCACHE) {
+        fprintf(stderr, "O_DIRECT (cache=none) for CDROM/DVD device (%s) "
+                "is unsupported\n", filename);
+        exit(1);
+    }
 
     /* open will not fail even if no CD is inserted, so add O_NONBLOCK */
     return raw_open_common(bs, filename, flags, O_NONBLOCK);