Inode corrupted after CVS update

Inode corrupted after CVS update

[Steven Scholz]

Hi there,

I pulled the CVS yesterday, patched my linuxppc-2.4.20, build and 
flashed my image.

I am using JFFS2 as root fs. While booting I get loads of

...
VFS: Mounted root (jffs2 filesystem).
Freeing unused kernel memory: 56k init
jffs2_get_inode_nodes(): Inode corrupted at 0x000ff370, totlen 70, 
#ino  130, version 1, isize 0, csize 2, dsize 2
jffs2_do_read_inode(): No data nodes found for ino #130
Warning: unable to open an initial console.
jffs2_get_inode_nodes(): Inode corrupted at 0x000ff458, totlen 70, 
#ino  132, version 1, isize 0, csize 2, dsize 2
jffs2_do_read_inode(): No data nodes found for ino #132
Returned error for crccheck of ino #132. Expect badness...
jffs2_get_inode_nodes(): Inode corrupted at 0x000ff6e0, totlen 70, 
#ino  137, version 1, isize 0, csize 2, dsize 2
jffs2_do_read_inode(): No data nodes found for ino #137
Returned error for crccheck of ino #137. Expect badness...
jffs2_get_inode_nodes(): Inode corrupted at 0x000ff758, totlen 70, 
#ino  138, version 1, isize 0, csize 2, dsize 2
jffs2_do_read_inode(): No data nodes found for ino #138
...

Any idea?

Thanks,

Steven

Re: Inode corrupted after CVS update

[David Woodhouse]

On Tue, 2003-04-08 at 08:19, Steven Scholz wrote:
> I pulled the CVS yesterday, patched my linuxppc-2.4.20, build and 
> flashed my image.

> I am using JFFS2 as root fs. While booting I get loads of
> 
> ...
> jffs2_get_inode_nodes(): Inode corrupted at 0x000ff370, totlen 70, 
> #ino  130, version 1, isize 0, csize 2, dsize 2

Argh. device nodes have data in the nodes but isize is zero, triggering
an overly-zealous sanity check. Fixed in cvs...

-- 
dwmw2

Re: Inode corrupted after CVS update

[Steven Scholz]

David Woodhouse schrieb:
> On Tue, 2003-04-08 at 08:19, Steven Scholz wrote:
> 
>>I pulled the CVS yesterday, patched my linuxppc-2.4.20, build and 
>>flashed my image.
> 
> 
>>I am using JFFS2 as root fs. While booting I get loads of
>>
>>...
>>jffs2_get_inode_nodes(): Inode corrupted at 0x000ff370, totlen 70, 
>>#ino  130, version 1, isize 0, csize 2, dsize 2
> 
> 
> Argh. device nodes have data in the nodes but isize is zero, triggering
> an overly-zealous sanity check. Fixed in cvs...

:o)

Thanks David!

It works now.

So how stable is the MTD/JFFS2 code anyway?
Is the CVS the only place for me to get the source?
Or are there some offical, release patches?

Thanks again,

Steven

Re: Inode corrupted after CVS update

[David Woodhouse]

On Tue, 2003-04-08 at 09:33, Steven Scholz wrote:
> So how stable is the MTD/JFFS2 code anyway?
> Is the CVS the only place for me to get the source?
> Or are there some offical, release patches?

The jffs2-2_4-branch from CVS, which is in Marcelo's latest 2.4.21-pre,
is the one that's _really_ intended to be stable -- but that's because
we don't make any exciting changes there at all.

The CVS HEAD is normally _fairly_ stable, and when I have time I push it
into the handhelds.org tree so that I inflict it on an army of
iPAQ-using victims^Wtesters. The error in the previous version didn't
show up because I didn't test with device nodes -- we use devfs on the
iPAQ.

If you want to track -HEAD it's worth keeping an eye on the IRC channel
-- if Thomas and I are spending the day discussing how best to fix
stuff, and there's a lot of swearing and debugging going on, then it's
probably not a good time to update :)

-- 
dwmw2

Re: Inode corrupted after CVS update

[Steven Scholz]

David Woodhouse schrieb:
> On Tue, 2003-04-08 at 09:33, Steven Scholz wrote:
> 
>>So how stable is the MTD/JFFS2 code anyway?
>>Is the CVS the only place for me to get the source?
>>Or are there some offical, release patches?
> 
> 
> The jffs2-2_4-branch from CVS, which is in Marcelo's latest 2.4.21-pre,
> is the one that's _really_ intended to be stable -- but that's because
> we don't make any exciting changes there at all.

But the performance improvements lately done on the JFFS2 code is in 
there!?

So for production systems I should patch my 2.4.20 with the 
jffs2-2_4-branch !?

> If you want to track -HEAD it's worth keeping an eye on the IRC channel
> -- if Thomas and I are spending the day discussing how best to fix
> stuff, and there's a lot of swearing and debugging going on, then it's
> probably not a good time to update :)
:o))

Thanks,

Steven

Re: Inode corrupted after CVS update

[David Woodhouse]

On Tue, 2003-04-08 at 09:47, Steven Scholz wrote:
> But the performance improvements lately done on the JFFS2 code is in 
> there!?

No. Nothing new is in there -- possibly not even the fixes to get
mtime/nlink on directories right.
 
> So for production systems I should patch my 2.4.20 with the 
> jffs2-2_4-branch !?

That's the official answer, yes. In practice, you'll generally be OK on
CVS head, and it hasn't had much done to it recently that's going to
make it break.

-- 
dwmw2