summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorMiklos Szeredi <mszeredi@suse.cz>2014-11-20 16:08:59 +0100
committerAl Viro <viro@zeniv.linux.org.uk>2014-12-17 08:27:14 -0500
commit10975933da3d65f8833d4ce98dcc2ecc63a695d6 (patch)
treed68accb6a448af5031c2e4fb524e2557cdfb50d2
parent7d65cf10e3d7747033b83fa18c5f3d2a498f66bc (diff)
downloadlwn-10975933da3d65f8833d4ce98dcc2ecc63a695d6.tar.gz
lwn-10975933da3d65f8833d4ce98dcc2ecc63a695d6.zip
init: fix read-write root mount
If mount flags don't have MS_RDONLY, iso9660 returns EACCES without actually checking if it's an iso image. This tricks mount_block_root() into retrying with MS_RDONLY. This results in a read-only root despite the "rw" boot parameter if the actual filesystem was checked after iso9660. I believe the behavior of iso9660 is okay, while that of mount_block_root() is not. It should rather try all types without MS_RDONLY and only then retry with MS_RDONLY. This change also makes the code more robust against the case when EACCES is returned despite MS_RDONLY, which would've resulted in a lockup. Signed-off-by: Miklos Szeredi <mszeredi@suse.cz> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
-rw-r--r--init/do_mounts.c6
1 files changed, 4 insertions, 2 deletions
diff --git a/init/do_mounts.c b/init/do_mounts.c
index 9b3565c41502..eb410083e8e0 100644
--- a/init/do_mounts.c
+++ b/init/do_mounts.c
@@ -395,8 +395,6 @@ retry:
case 0:
goto out;
case -EACCES:
- flags |= MS_RDONLY;
- goto retry;
case -EINVAL:
continue;
}
@@ -419,6 +417,10 @@ retry:
#endif
panic("VFS: Unable to mount root fs on %s", b);
}
+ if (!(flags & MS_RDONLY)) {
+ flags |= MS_RDONLY;
+ goto retry;
+ }
printk("List of all partitions:\n");
printk_all_partitions();