]> err.no Git - linux-2.6/commitdiff
[PATCH] s390: dasd diag with block sizes > 512
authorPeter Oberparleiter <peter.oberparleiter@de.ibm.com>
Mon, 7 Nov 2005 08:59:08 +0000 (00:59 -0800)
committerLinus Torvalds <torvalds@g5.osdl.org>
Mon, 7 Nov 2005 15:53:34 +0000 (07:53 -0800)
Access to FBA disks via DIAG fails for block sizes > 512 byte.  The device
analysis code of the DIAG discipline does not properly initialize the DIAG250
device environment after completion of the analysis.  This results in VM only
serving 512 bytes per block I/O request whereas Linux expects larger block
sizes.  Add proper device environment setup to end of analysis code.

Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
drivers/s390/block/dasd_diag.c

index 9aa608f528eb0be62e049cfd9f629831005bd880..ab8754e566bcbdff183f22982dbe91d2ba431cac 100644 (file)
@@ -6,7 +6,7 @@
  * Bugreports.to..: <Linux390@de.ibm.com>
  * (C) IBM Corporation, IBM Deutschland Entwicklung GmbH, 1999,2000
  *
- * $Revision: 1.50 $
+ * $Revision: 1.51 $
  */
 
 #include <linux/config.h>
@@ -404,37 +404,47 @@ dasd_diag_check_device(struct dasd_device *device)
                private->iob.bio_list = &bio;
                private->iob.flaga = DASD_DIAG_FLAGA_DEFAULT;
                rc = dia250(&private->iob, RW_BIO);
-               if (rc == 0 || rc == 3)
-                       break;
+               if (rc == 3) {
+                       DEV_MESSAGE(KERN_WARNING, device, "%s",
+                               "DIAG call failed");
+                       rc = -EOPNOTSUPP;
+                       goto out;
+               }
                mdsk_term_io(device);
+               if (rc == 0)
+                       break;
        }
-       if (rc == 3) {
-               DEV_MESSAGE(KERN_WARNING, device, "%s", "DIAG call failed");
-               rc = -EOPNOTSUPP;
-       } else if (rc != 0) {
+       if (bsize > PAGE_SIZE) {
                DEV_MESSAGE(KERN_WARNING, device, "device access failed "
                            "(rc=%d)", rc);
                rc = -EIO;
+               goto out;
+       }
+       /* check for label block */
+       if (memcmp(label->label_id, DASD_DIAG_CMS1,
+                 sizeof(DASD_DIAG_CMS1)) == 0) {
+               /* get formatted blocksize from label block */
+               bsize = (unsigned int) label->block_size;
+               device->blocks = (unsigned long) label->block_count;
+       } else
+               device->blocks = end_block;
+       device->bp_block = bsize;
+       device->s2b_shift = 0;  /* bits to shift 512 to get a block */
+       for (sb = 512; sb < bsize; sb = sb << 1)
+               device->s2b_shift++;
+       rc = mdsk_init_io(device, device->bp_block, 0, NULL);
+       if (rc) {
+               DEV_MESSAGE(KERN_WARNING, device, "DIAG initialization "
+                       "failed (rc=%d)", rc);
+               rc = -EIO;
        } else {
-               if (memcmp(label->label_id, DASD_DIAG_CMS1,
-                         sizeof(DASD_DIAG_CMS1)) == 0) {
-                       /* get formatted blocksize from label block */
-                       bsize = (unsigned int) label->block_size;
-                       device->blocks = (unsigned long) label->block_count;
-               } else
-                       device->blocks = end_block;
-               device->bp_block = bsize;
-               device->s2b_shift = 0;  /* bits to shift 512 to get a block */
-               for (sb = 512; sb < bsize; sb = sb << 1)
-                       device->s2b_shift++;
-               
                DEV_MESSAGE(KERN_INFO, device,
                            "(%ld B/blk): %ldkB",
                            (unsigned long) device->bp_block,
                            (unsigned long) (device->blocks <<
                                device->s2b_shift) >> 1);
-               rc = 0;
        }
+out:
        free_page((long) label);
        return rc;
 }