Re: Fixing the tty layer was Re: any chance of 2.6.0-test*?

[Dipankar Sarma <dipankar@in.ibm.com>]

On Mon, Jan 13, 2003 at 08:25:39AM +0100, Andi Kleen wrote:
> > Oh, yes, I have spent hours and hours trying to untangle tty locking
> > and it isn't simple.
> 
> Oops. Could you quickly summarize your findings so far ?

I only found more confusions - I can't figure how tty_files list
is locked - sure files_lock is supposed to protect it but there
are deletions done without any lock. Another thing that needs
looking into is to avoid or reduce use of the tasklist_lock there.

> > What does that BKL protect ? I can't seem to ever figure our if
> > all the races are plugged or not.
> 
> Well, one has to start somewhere. Just starting by plugging most of the
> obvious races, then the more subtle ones can be attacked later.
> 
> The idea of the BKL was to protect the protect context code against
> itself (code lock) and also the few global data structures that 
> are only accessed from process context (like the tty drivers list)

In that case would it not be better to replace all BKLs by a single tty
lock ?

> 
> I attached my current patch, it isn't too well tested however and needs
> more work.
> 
> Mostly just adds lock_kernel()s to the high level code so far and a few comments.

Cool, I will start off by testing this stuff.

Thanks
Dipankar