From: Karel Zak <kzak@redhat.com>
Date: Wed, 20 Jul 2016 09:24:36 +0000 (+0200)
Subject: Revert "sfdisk: exit with error if rereading partition table fails"
X-Git-Tag: v2.29-rc1~147
X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=78e7f78896f2c9cf1b2615c8c555a64dcd9d2642;p=thirdparty%2Futil-linux.git

Revert "sfdisk: exit with error if rereading partition table fails"

This reverts commit 14f644f386a1708483ed446e983c0976e3976a9d.

It seems we have mess in reread-after-write:

all old versions to v2.20  -- returns 0
from v2.20 to v2.26        -- returns 1
since v2.26                -- returns 0

I think re-read errors should not be interpreted as fatal errors,
because it's pretty common that you want to modify only one partition
(e.g. resize) and then another partitions are probably still in use
and re-read all PT does not make sense.

What we need is to improve granularity for re-read and calls it only
when really necessary (all PT modified) and otherwise call BLKPG
(add/delete/resize) ioctls.

Reported-by: Nikhil Valluru <vvnikhil@gmail.com>
Signed-off-by: Karel Zak <kzak@redhat.com>
---

diff --git a/disk-utils/sfdisk.c b/disk-utils/sfdisk.c
index 0f329ed397..d51b2613df 100644
--- a/disk-utils/sfdisk.c
+++ b/disk-utils/sfdisk.c
@@ -535,7 +535,7 @@ static int write_changes(struct sfdisk *sf)
 			rc = move_partition_data(sf, sf->partno, sf->orig_pa);
 		if (!rc) {
 			fdisk_info(sf->cxt, _("\nThe partition table has been altered."));
-			rc = fdisk_reread_partition_table(sf->cxt);
+			fdisk_reread_partition_table(sf->cxt);
 		}
 	}
 	if (!rc)