[Qemu-devel] NetBSD qemu block device support

[Qemu-devel] NetBSD qemu block device support

[haad]

Hi folks,

This patch at [1] add support for NetBSD block ioctl calls to qemu
block-raw.c file. It was written for xen version of qemu but basically
it will work with vanilla qemu, too. Would anyone like to commit this
patch so it can be included into the base xen distribution ?


[1] www.netbsd.org/~haad/block-raw.diff
-- 


Regards.

Adam

Re: [Qemu-devel] NetBSD qemu block device support

[Blue Swirl]

On 3/17/10, haad <haad@netbsd.org> wrote:
> Hi folks,
>
>  This patch at [1] add support for NetBSD block ioctl calls to qemu
>  block-raw.c file. It was written for xen version of qemu but basically
>  it will work with vanilla qemu, too. Would anyone like to commit this
>  patch so it can be included into the base xen distribution ?

The patch does not apply to current development repository.

There is no description of the change suitable for changelog without
any editing.

Signed-off-by: line is missing.

The patch combines formatting (whitespace) changes with functional
changes. The formatting changes seem useless.

The essence of the patch is twofold, It adds NetBSD specific includes
and NetBSD specific code to raw_getlength(). These look OK, except
formatting may be a bit off (no space after if) and currently the code
probably should be surrounded by
#ifdef DIOCGWEDGEINFO
#endif
or just #ifdef __NetBSD__.

Re: [Qemu-devel] NetBSD qemu block device support

[Christoph Hellwig]

On Thu, Mar 18, 2010 at 10:32:36PM +0200, Blue Swirl wrote:
> The patch does not apply to current development repository.
> 
> There is no description of the change suitable for changelog without
> any editing.
> 
> Signed-off-by: line is missing.
> 
> The patch combines formatting (whitespace) changes with functional
> changes. The formatting changes seem useless.
> 
> The essence of the patch is twofold, It adds NetBSD specific includes
> and NetBSD specific code to raw_getlength(). These look OK, except
> formatting may be a bit off (no space after if) and currently the code
> probably should be surrounded by
> #ifdef DIOCGWEDGEINFO
> #endif
> or just #ifdef __NetBSD__.

Also given the extreme mess that raw_getlength is I would much prefer
to give every new OS it's own copy of it, similar to how we already
handle OpenBSD today.

[Qemu-devel] NetBSD qemu block device support

[haad]

[-- Attachment #1: Type: text/plain, Size: 1247 bytes --]

Hi,
On Mar,Thursday 18 2010, at 9:32 PM, Blue Swirl wrote:

> On 3/17/10, haad <haad@netbsd.org> wrote:
>> Hi folks,
>>
>> This patch at [1] add support for NetBSD block ioctl calls to qemu
>> block-raw.c file. It was written for xen version of qemu but basically
>> it will work with vanilla qemu, too. Would anyone like to commit this
>> patch so it can be included into the base xen distribution ?
>
> The patch does not apply to current development repository.
>
> There is no description of the change suitable for changelog without
> any editing.
>
> Signed-off-by: line is missing.
>
> The patch combines formatting (whitespace) changes with functional
> changes. The formatting changes seem useless.
>
> The essence of the patch is twofold, It adds NetBSD specific includes
> and NetBSD specific code to raw_getlength(). These look OK, except
> formatting may be a bit off (no space after if) and currently the code
> probably should be surrounded by
> #ifdef DIOCGWEDGEINFO
> #endif
> or just #ifdef __NetBSD__.

I have written this patch some time ago and then I totally forgot on
it (I'm very sorry about that)

Today I looked at it again and changed it to be clean and to properly
apply against current sources.


-- 


Regards.

Adam

[-- Attachment #2: netbsd-raw-posix.diff --]
[-- Type: text/x-diff, Size: 1200 bytes --]

diff --git a/block/raw-posix.c b/block/raw-posix.c
index 72fb8ce..444fb2d 100644
--- a/block/raw-posix.c
+++ b/block/raw-posix.c
@@ -62,6 +62,12 @@
 #include <sys/disklabel.h>
 #include <sys/dkio.h>
 #endif
+#if defined(__NetBSD__)
+#include <sys/ioctl.h>
+#include <sys/disklabel.h>
+#include <sys/dkio.h>
+#include <sys/disk.h>
+#endif
 
 #ifdef __DragonFly__
 #include <sys/ioctl.h>
@@ -595,6 +601,28 @@ static int64_t raw_getlength(BlockDriverState *bs)
     } else
         return st.st_size;
 }
+#elif defined(__NetBSD__)
+static int64_t  raw_getlength(BlockDriverState *bs)
+{
+	int fd = ((BDRVRawState*)bs->opaque)->fd;
+	struct stat st;
+	if (fstat(fd, &st))
+		return -1;
+	if (S_ISCHR(st.st_mode) || S_ISBLK(st.st_mode)) {
+		struct dkwedge_info dkw;
+		if (ioctl(fd, DIOCGWEDGEINFO, &dkw) != -1) {
+			/* NetBSD still supports only 512 bytes for disk sector */
+			return dkw.dkw_size * 512;
+		} else {
+			struct disklabel dl;
+			if(ioctl(fd, DIOCGDINFO, &dl))
+				return -1;
+			return (uint64_t)dl.d_secsize *
+			    dl.d_partitions[DISKPART(st.st_rdev)].p_size;
+		}
+	} else
+		return st.st_size;
+}
 #elif defined(__sun__)
 static int64_t raw_getlength(BlockDriverState *bs)
 {