On 15.09.2015 20:01, Brian C. Lane wrote:
[...]
Thanks for reviewing the patch!
Remove the necessity for DASD-specific ioctls for partition handling.
This allows to correctly handle DASD-backed virtio block devices.
I'm still not sure what the benefit of this is for the VM. Why not use
the normal virtio devices?
I'll enhance the log message with an explanation. In a nutshell it's
to support passing ECKD DASDs via virtio, which is perfectly valid on
DASD-only systems
As virtio is just a transport the intrinsic geometry specifics of
DASDs are retained even if the transport-specific ioctls are not
available. This is why upstreams s390-tools fdasd has a fallback
that I'm backporting here.
+/* kl: key length, dl: data length */
+static unsigned int recs_per_track(unsigned short dev_type, unsigned int kl,
+ unsigned int dl)
+{
+ int dn, kn;
These should be unsigned int since that's what ceil_quot returns.
} else {
+ if (ioctl(f, BLKGETSIZE64, &size_in_bytes) != 0) {
+ fdasd_error(anc, unable_to_ioctl,
+ "Could not retrieve disk size.");
+ }
This should be a translated string. Also, does it need more error
handling? I see that most of the fdasd_error calls currently fall
through without doing anything special and I'm not sure how correct that
is.
+ if (ioctl(f, BIODASDINFO, &dasd_info) != 0) {
+ /* verify that the geometry matches a 3390 DASD */
+ if (!fdasd_verify_geometry(DASD_3390_TYPE, blksize,
+ &anc->geo)) {
+ fdasd_error(anc, wrong_disk_type,
+ "Disk geometry does not match a "
+ "DASD device of type 3390.");
+ goto error;
This should also be a translated string.
Sorry about those. The backport was (too) verbatim. Will fix that in
a V2.
--
Mit freundlichen Grüßen/Kind Regards
Viktor Mihajlovski
IBM Deutschland Research & Development GmbH
Vorsitzender des Aufsichtsrats: Martina Köderitz
Geschäftsführung: Dirk Wittkopp
Sitz der Gesellschaft: Böblingen
Registergericht: Amtsgericht Stuttgart, HRB 243294
