[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#47490: base: mke2fs makes unmountable filesystems; mount doesn't honor -o nocheck

> Jim Dennis <jimd@starshine.org> writes:

>> Package: base
>> Version: 19991015
>> Severity: important

>> 	I was wrong.  Earlier I claimed that mount was
>> 	unable to handle a filesystem with a non-default
>> 	inode/block ratio ("couldn't mount RDWR").  However,
>> 	my reasoning was wrong.

>> 	In actuality the new mke2fs is making filesystems that
>> 	the latest mount command will not tolerate.  

>> 	I've heard that we need to start using the 
>> 	mount -o nocheck (or check=none) option to mount the 
>> 	newly created filesystems with the "sparse superblock" 
>> 	features.  

>> 	However, that option is not honored with my copy of 
>> 	mount (2.9x).  mke2fs version is 1.15.  

>> 	e2fsprogs version 1.15-3
> The problem goes away with a 2.2.x kernel, which can mount sparse
> superblock ext2 filesystems. Under 2.0.x you can mount the filesystem
> read-only and then remount read-write to trick linux into ignoring the 
> feature, but mke2fs should just be fixed not to create sparse
> superblocks under 2.0.x.
> May the Source be with you.
> 			Goswin

	There are several problems and possible solutions.
	1st we could backport the sparse superblock support
	to the 2.0 (and even the 1.2 and 1.0) kernels.

	We could patch mke2fs to check the current kernel version
	and default to non-sparse creation when running under
	older kernels (with a warning message, perhaps).  

	We could improve the mount(8) error messages so that it
	specified which "optional feature" is involved, and update
	the man pages and FAQs so that they have the appropriate
	error message(s) listed with recommended fixes and

	There is also a workaround involving the use of 
	tune2fs to force the bits back to the state.

	When user space utilities start *requiring* features
	of new kernels, we should be more graceful about 
	the implementation, documentation, errors and warning

	Concurrent to my finding this problem on my system 
	I was reading about it on Tom Oehser's Root/Boot
	mailing list.  He has a set of preliminary patches
	to the 2.0.35 kernel to backport the requisite fs

Jim Dennis                                             jdennis@linuxcare.com
Linuxcare: Linux Corporate Support Team:            http://www.linuxcare.com

Reply to: