BUG: pty.c in linux 2.6.31-rc5+

BUG: pty.c in linux 2.6.31-rc5+

[MRU.Ilya]

Synchronic's line disciplines(ppp_synctty, hdlc) can't work with pseudo 
terminals.

If frames send very fast from one side(master/slave) to other one,
frames merge in one (merge point "flip buffer") and the other side can 
read only one big frame.
But sync line discipline need one frame in one read.

For push one frame to ldisc I propose small patch, but this regression for
commit e043e42bdb66885b3ac10d27a01ccb9972e2b0a3.

I don't find any mutexes and other things forbidden in not user context,
in function flush_to_ldisc() , and I think this patch
is true and this function can called from any context. But I not sure.

Relative commit  86d23a057e718f73adc15e463c643d6e014a19cd.

BUG: pty.c in linux 2.6.31-rc5+

[MRU.Ilya]

[-- Attachment #1: Type: text/plain, Size: 688 bytes --]

Synchronic's line disciplines(ppp_synctty, hdlc) can't work with pseudo 
terminals.

If frames send very fast from one side(master/slave) to other one,
frames merge in one (merge point "flip buffer") and the other side can 
read only one big frame.
But sync line discipline need one frame in one read.

For push one frame to ldisc I propose small patch, but this regression for
commit e043e42bdb66885b3ac10d27a01ccb9972e2b0a3.

I don't find any mutexes and other things forbidden in not user context,
in function flush_to_ldisc() , and I think this patch
is true and this function can called from any context. But I not sure.

Relative commit  86d23a057e718f73adc15e463c643d6e014a19cd.



[-- Attachment #2: pty_set_low_latency.patch --]
[-- Type: text/plain, Size: 275 bytes --]

--- orig/pty.c	2011-07-13 07:29:43.000000000 +0400
+++ work/pty.c	2011-08-03 20:59:01.000000000 +0400
@@ -204,6 +204,7 @@
 	clear_bit(TTY_OTHER_CLOSED, &tty->link->flags);
 	set_bit(TTY_THROTTLED, &tty->flags);
 	retval = 0;
+	tty->low_latency = 1;
 out:
 	return retval;
 }

Re: BUG: pty.c in linux 2.6.31-rc5+

[Greg KH]

On Thu, Aug 04, 2011 at 10:43:19AM +0400, MRU.Ilya wrote:
> Synchronic's line disciplines(ppp_synctty, hdlc) can't work with
> pseudo terminals.
> 
> If frames send very fast from one side(master/slave) to other one,
> frames merge in one (merge point "flip buffer") and the other side
> can read only one big frame.
> But sync line discipline need one frame in one read.
> 
> For push one frame to ldisc I propose small patch, but this regression for
> commit e043e42bdb66885b3ac10d27a01ccb9972e2b0a3.
> 
> I don't find any mutexes and other things forbidden in not user context,
> in function flush_to_ldisc() , and I think this patch
> is true and this function can called from any context. But I not sure.
> 
> Relative commit  86d23a057e718f73adc15e463c643d6e014a19cd.

I don't see a patch here.

<formletter>

This is not the correct way to submit patches for inclusion in the
stable kernel tree.  Please read Documentation/stable_kernel_rules.txt
for how to do this properly.

</formletter>

Re: BUG: pty.c in linux 2.6.31-rc5+

[Greg KH]

On Thu, Aug 04, 2011 at 10:46:43AM +0400, MRU.Ilya wrote:
> Synchronic's line disciplines(ppp_synctty, hdlc) can't work with
> pseudo terminals.
> 
> If frames send very fast from one side(master/slave) to other one,
> frames merge in one (merge point "flip buffer") and the other side
> can read only one big frame.
> But sync line discipline need one frame in one read.
> 
> For push one frame to ldisc I propose small patch, but this regression for
> commit e043e42bdb66885b3ac10d27a01ccb9972e2b0a3.
> 
> I don't find any mutexes and other things forbidden in not user context,
> in function flush_to_ldisc() , and I think this patch
> is true and this function can called from any context. But I not sure.
> 
> Relative commit  86d23a057e718f73adc15e463c643d6e014a19cd.
> 
> 

> --- orig/pty.c	2011-07-13 07:29:43.000000000 +0400
> +++ work/pty.c	2011-08-03 20:59:01.000000000 +0400
> @@ -204,6 +204,7 @@
>  	clear_bit(TTY_OTHER_CLOSED, &tty->link->flags);
>  	set_bit(TTY_THROTTLED, &tty->flags);
>  	retval = 0;
> +	tty->low_latency = 1;

I really don't think this is correct.

Also, you forgot to read Documentation/SubmittingPatches for how to
properly format and send a patch and who to send it to...

thanks,

greg k-h

Re: BUG: pty.c in linux 2.6.31-rc5+

[Alan Cox]

On Thu, 04 Aug 2011 10:43:19 +0400
"MRU.Ilya" <utdilya@mail.ru> wrote:

> Synchronic's line disciplines(ppp_synctty, hdlc) can't work with pseudo 
> terminals.

Correct - tty devices do not support framing. This is intentional.