Re: Linux 2.6.25-rc4

[Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>]

On Sunday 16 March 2008, Anders Eriksson wrote:
> 
> bzolnier@gmail.com said:
> > This commit is also _after_ the previously "guilty" one (commit
> > 852738f39258deafb3d89c187cb1a4050820d555)?
> 
> Yep.
> 
> > IIRC we've agreed that bisection between 2.6.24 and commit 8527* was needed?
> 
> I did that first, and it pointed to  8527* as the bad one. During this bisect 
> I took care to always start from a 2.6.24 .config, and do a make oldconfig from 
> that. A clean series of good emerged.. Unfortunately I don't have the configs 
> from the previous runs saved.
> 
> Given this, I concluded that the configs had to play a role in making a 
> version go bad, so I decided to do another one between  8527* and 2.6.25-rc1
> always based on the  8527* config (which _worked_ when tested with the 2.6.24 
> config dragged forward. So the upped end of the first bisect was badly chosen)
> 
> The attahed was the result of that latter run...

Given that the problem manifested with 8527* in the past (w/ whatever config)
then 4d97* just cannot be a guilty commit (== the one introducing the problem).

IOW the 8527* and later commits just make the issue more likely to occur
(moreover there should be no difference in configs between 8527* and 4d97*).

I guess that re-doing bisection between 2.6.24 and 8527* may be fruitless
but maybe somebody has some idea how we can proceed further with debugging
the hang instead?  Jens/Ingo?

Thanks,
Bart