Re: + s390-tty-prepare-for-put_char-to-return-success-fail.patch added to -mm tree

Re: + s390-tty-prepare-for-put_char-to-return-success-fail.patch added to -mm tree

[Heiko Carstens]

>   * NOTE: include/linux/tty_driver.h specifies that a character should be
>   * ignored if there is no room in the queue. This driver implements a different
>   * semantic in that it will block when there is no more room left.
> + *
> + * FIXME: putchar can currently be called from BH and other non blocking
> + * handlers so  this semantic isn't a good idea.
>   */
> -static void
> +static int
>  sclp_vt220_put_char(struct tty_struct *tty, unsigned char ch)
>  {
>  	__sclp_vt220_write(&ch, 1, 0, 0, 1);
> +	return 1;

Uh, oh... the comment above used to mean "block if in schedulable context or
busy wait otherwise". Figuring out which context we were in was done via
in_atomic(), which was broken on !SMP anyway, so that had to go.
So the last parameter for __sclp_vt220_write is now an indicator if scheduling
is allowed or not (1 means yes). Somebody told me that the put_char routine
will only be called from schedulable context, which now doesn't seem to be
true?!
For the console functions we pass indeed 0 to __sclp_vt220_write since printk
may be called within any context. Is that also true for the tty put_char
routine?

Re: + s390-tty-prepare-for-put_char-to-return-success-fail.patch added to -mm tree

[Alan Cox]

> is allowed or not (1 means yes). Somebody told me that the put_char routine
> will only be called from schedulable context, which now doesn't seem to be
> true?!

It isnt true.

> For the console functions we pass indeed 0 to __sclp_vt220_write since printk
> may be called within any context. Is that also true for the tty put_char
> routine?

Yes, but failing a put_char is perfectly acceptable.

Alan

Re: + s390-tty-prepare-for-put_char-to-return-success-fail.patch added to -mm tree

[Heiko Carstens]

On Tue, Apr 15, 2008 at 11:11:01AM +0100, Alan Cox wrote:
> > is allowed or not (1 means yes). Somebody told me that the put_char routine
> > will only be called from schedulable context, which now doesn't seem to be
> > true?!
> 
> It isnt true.
> 
> > For the console functions we pass indeed 0 to __sclp_vt220_write since printk
> > may be called within any context. Is that also true for the tty put_char
> > routine?
> 
> Yes, but failing a put_char is perfectly acceptable.

Ok, thanks for pointing this out! We are going to fix this as soon as your
patch is merged, so we can avoid further merge conflicts.
I assume your patch is for 2.6.26?

Re: + s390-tty-prepare-for-put_char-to-return-success-fail.patch added to -mm tree

[Alan Cox]

On Tue, Apr 15, 2008 at 12:45:07PM +0200, Heiko Carstens wrote:
> Ok, thanks for pointing this out! We are going to fix this as soon as your
> patch is merged, so we can avoid further merge conflicts.
> I assume your patch is for 2.6.26?

I would hope so - assuming no other bugs fall out of this.

Alan

Re: + s390-tty-prepare-for-put_char-to-return-success-fail.patch added to -mm tree

[Andrew Morton]

On Tue, 15 Apr 2008 07:07:45 -0400 Alan Cox <alan@redhat.com> wrote:

> On Tue, Apr 15, 2008 at 12:45:07PM +0200, Heiko Carstens wrote:
> > Ok, thanks for pointing this out! We are going to fix this as soon as your
> > patch is merged, so we can avoid further merge conflicts.
> > I assume your patch is for 2.6.26?
> 
> I would hope so - assuming no other bugs fall out of this.
> 

It might be a bit late to find out - I'll _try_ to squeeze a -mm out this
week, but I have an ill-timed trip next week and then we're into the merge
window.

So please do double-check that it's not all catastrophically buggy in some
way ;)