Re: [TEST] NTFS patch

In message <41BC87EE.80007_at_mac.com>, Pete Carss writes:
>Toxa wrote:

>> Btw, if you try to mount linux ext2 partition (if exist), and then 
>> reboot (without umounting it by hands), both linux and bsd partitions
>> becomes unclean, e.g. you need to fsck.ext3 linux partition to mount it
>> again.
>> Tested on -stable and -current.

-stable being releng_5, right ?

It's strange that you would be able to unmount it cleanly by hand but
not by shutdown.  It could indicate that dirty buffers are still
stuck in the buffer cache when the unmount returns.

I've looked over the ext2fs code, but not found anything off the top
of my head that could explain it.

On -current you can try to insert a call to vinvalbuf() right before
DROP_GIANT() in ext2_unmount(), something like:

	vinvalbuf(ump->un_devvp, 0, td->td_ucred, td, 0, 0)

You can also try to insert some strategic printfs and see if the
handling of the (various) read-only flag(s) work as expected.  In particular
look after the updates of the superblock which should happen when unmounting.


-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk_at_FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.