2.6.0-test2: cursor started to disappear

2.6.0-test2: cursor started to disappear

[Pavel Machek]

Hi!

Under 2.6.0-test2 cursor problems got worse... Sometimes cursor just
disappears. Switching consoles does not help.

Plus I'm seeing some silent data corruption. It may be swsusp or loop
related, don't yet know how to reproduce it (ext2 on hp omnibook xe3).
								Pavel
-- 
When do you have a heart between your knees?
[Johanka's followup: and *two* hearts?]

Re: 2.6.0-test2: cursor started to disappear

[Andries Brouwer]

On Mon, Jul 28, 2003 at 08:14:08PM +0200, Pavel Machek wrote:

> Plus I'm seeing some silent data corruption. It may be
> swsusp or loop related

Loop is not stable at all. Unsuitable for daily use.

Re: 2.6.0-test2: cursor started to disappear

[Pavel Machek]

Hi!

> > Plus I'm seeing some silent data corruption. It may be
> > swsusp or loop related
> 
> Loop is not stable at all. Unsuitable for daily use.

Ouch... I have my most important filesystem on loop. Time to go back
to 2.4.X? Or do you have some patches you want me to try?

							Pavel
-- 
When do you have a heart between your knees?
[Johanka's followup: and *two* hearts?]

Re: 2.6.0-test2: cursor started to disappear

[Andries Brouwer]

On Mon, Jul 28, 2003 at 08:34:43PM +0200, Pavel Machek wrote:

> > > Plus I'm seeing some silent data corruption. It may be
> > > swsusp or loop related
> > 
> > Loop is not stable at all. Unsuitable for daily use.
> 
> Ouch... I have my most important filesystem on loop. Time to go back
> to 2.4.X? Or do you have some patches you want me to try?

[maybe an extra backup?]

Jari's version of loop.c has always been solid.

Re: 2.6.0-test2: cursor started to disappear

[Pavel Machek]

Hi!

> > > > Plus I'm seeing some silent data corruption. It may be
> > > > swsusp or loop related
> > > 
> > > Loop is not stable at all. Unsuitable for daily use.
> > 
> > Ouch... I have my most important filesystem on loop. Time to go back
> > to 2.4.X? Or do you have some patches you want me to try?
> 
> [maybe an extra backup?]

Hmm, looks like I should do that.

> Jari's version of loop.c has always been solid.

Jari Ruusu? Okay, I'll test that.

								Pavel
-- 
When do you have a heart between your knees?
[Johanka's followup: and *two* hearts?]

Re: 2.6.0-test2: cursor started to disappear

[Andrew Morton]

Andries Brouwer <aebr@win.tue.nl> wrote:
>
> On Mon, Jul 28, 2003 at 08:14:08PM +0200, Pavel Machek wrote:
> 
> > Plus I'm seeing some silent data corruption. It may be
> > swsusp or loop related
> 
> Loop is not stable at all. Unsuitable for daily use.

That's the first I've heard about it.  Do you have some details on this?  A
test case perhaps?

Re: 2.6.0-test2: cursor started to disappear

[Andrew Morton]

Andries Brouwer <aebr@win.tue.nl> wrote:
>
> On Mon, Jul 28, 2003 at 01:17:41PM -0700, Andrew Morton wrote:
> > Andries Brouwer <aebr@win.tue.nl> wrote:
> > >
> > > On Mon, Jul 28, 2003 at 08:14:08PM +0200, Pavel Machek wrote:
> > > 
> > > > Plus I'm seeing some silent data corruption. It may be
> > > > swsusp or loop related
> > > 
> > > Loop is not stable at all. Unsuitable for daily use.
> > 
> > That's the first I've heard about it.  Do you have some details on this?  A
> > test case perhaps?
> 
> Yes, there are many reports, and it is easy to confirm.
> 
> By some coincidence just a moment ago we saw the announcement of
> Bugzilla bug 1000:
> 
> [Bug 1000] New: file corruption using cryptoloop on ext2/ext3

That's cryptoloop.  I thought you were referring to vanilla loop.

Are you saying that the problems are only with cryptoloop, or does normal
old loop have some bug?

Re: 2.6.0-test2: cursor started to disappear

[Andries Brouwer]

On Mon, Jul 28, 2003 at 01:17:41PM -0700, Andrew Morton wrote:
> Andries Brouwer <aebr@win.tue.nl> wrote:
> >
> > On Mon, Jul 28, 2003 at 08:14:08PM +0200, Pavel Machek wrote:
> > 
> > > Plus I'm seeing some silent data corruption. It may be
> > > swsusp or loop related
> > 
> > Loop is not stable at all. Unsuitable for daily use.
> 
> That's the first I've heard about it.  Do you have some details on this?  A
> test case perhaps?

Yes, there are many reports, and it is easy to confirm.

By some coincidence just a moment ago we saw the announcement of
Bugzilla bug 1000:

[Bug 1000] New: file corruption using cryptoloop on ext2/ext3


Andries

Re: 2.6.0-test2: cursor started to disappear

[Andries Brouwer]

On Mon, Jul 28, 2003 at 01:34:31PM -0700, Andrew Morton wrote:

> That's cryptoloop.  I thought you were referring to vanilla loop.
> 
> Are you saying that the problems are only with cryptoloop, or does
> normal old loop have some bug?

It would be too optimistic to claim that it has none,
but the easily provoked corruption I mentioned is for cryptoloop.
Sorry for being imprecise.

Re: 2.6.0-test2: cursor started to disappear

[Pavel Machek]

Hi!

> > > Plus I'm seeing some silent data corruption. It may be
> > > swsusp or loop related
> > 
> > Loop is not stable at all. Unsuitable for daily use.
> 
> That's the first I've heard about it.  Do you have some details on this?  A
> test case perhaps?

After a while (week of use), kernel started complaining about "block
already freed in inode X" or something like that. I rebooted and fsck
found nothing serious (that is normal loop). I've now switched to
Jari's version of loop -- I'd like to keep this filesystem.

								Pavel
-- 
When do you have a heart between your knees?
[Johanka's followup: and *two* hearts?]

Re: 2.6.0-test2: cursor started to disappear

[Mikulas Patocka]

On Mon, 28 Jul 2003, Pavel Machek wrote:

> Hi!
>
> > > Plus I'm seeing some silent data corruption. It may be
> > > swsusp or loop related
> >
> > Loop is not stable at all. Unsuitable for daily use.
>
> Ouch... I have my most important filesystem on loop. Time to go back
> to 2.4.X? Or do you have some patches you want me to try?

Time to go back to 2.0.x --- it's the only kernel where loop was
correctly (almost) deadlockless implemented :)

Mikulas