sfdisk: failure to write partition layout to loopback devices

sfdisk: failure to write partition layout to loopback devices

[Unknown]

In util-linux 2.26 and onward a check is done to see if there are any
open filehandles on the device that is being updated. Specifically,
this check is being done in 'is_device_used': disk-utils/sfdisk.c:1055

For loopback devices the ioctl(fd, BLKRRPART) will always fail with
return code -1. Checking errno will yield error '22'. This is
"expected" and parted has considered making exceptions for this in the
past [1]. I would propose that instead of having to iss ue a '--force'
flag for basic loopback device operation, an exception should be
included in the is_device_used function.

It could be as simple as:

+    // If we are dealing with a loopback device this will fail, but
shouldn't
+    if (!strncmp("/dev/loop", fdisk_get_devname(sf->cxt), 9)) {
+        return 0;
+    }


[1] http://comments.gmane.org/gmane.comp.gnu.parted.devel/2254

Re: sfdisk: failure to write partition layout to loopback devices

[Karel Zak]

On Thu, Apr 16, 2015 at 04:06:31AM -0700, Unknown wrote:
> In util-linux 2.26 and onward a check is done to see if there are any
> open filehandles on the device that is being updated. Specifically,
> this check is being done in 'is_device_used': disk-utils/sfdisk.c:1055
> 
> For loopback devices the ioctl(fd, BLKRRPART) will always fail with
> return code -1. Checking errno will yield error '22'. This is
> "expected" and parted has considered making exceptions for this in the
> past [1]. I would propose that instead of having to iss ue a '--force'
> flag for basic loopback device operation, an exception should be
> included in the is_device_used function.
> 
> It could be as simple as:
> 
> +    // If we are dealing with a loopback device this will fail, but
> shouldn't
> +    if (!strncmp("/dev/loop", fdisk_get_devname(sf->cxt), 9)) {
> +        return 0;
> +    }

 Fixed (I have use major(st.st_rdev) == LOOPDEV_MAJOR).

    Karel

-- 
 Karel Zak  <kzak@redhat.com>
 http://karelzak.blogspot.com