Re: [PATCH 2.6.31-rc4] gigaset: really fix chars_in_buffer

[Tilman Schmidt <tilman@imap.cc>]

[-- Attachment #1: Type: text/plain, Size: 1452 bytes --]

Hi Alan,

thanks a lot for getting in contact with me, and for your
explanation of the rationale behind your patch.

On Sat, 1 Aug 2009 17:11:41 +0100, Alan Cox wrote:
> Is that correct. If you take a signal so the mutex_lock_interruptible
> takes the error path we'll exit any waits for characters to be flushed
> from buffers and lose the bytes in some cases ?

You are right. I'll have to change it to use non-interruptible
mutex_lock. I'll prepare a new patch.

> Thats why I didn't
> convert it - I didn't understand why it was using _interruptible at all ?

No specific reason. just the general rule to prefer sleeping
interuptibly whenever possible, because unkillable processes are
so annoying. Seeing that chars_in_buffer was declared as returning
int, I assumed (wrongly as it now turns out) that it would be
allowed to return the customary negative values for errors, so
that it could use the _interruptible variant. When I saw your
patch I read into it that zero would be the appropriate return
value for all error cases, including a signal while waiting for
the mutex. I guess there was a bit of wishful thinking in that.

Well, thanks again for your help. I'll follow up with a revised
patch.

Regards,
Tilman

-- 
Tilman Schmidt                    E-Mail: tilman@imap.cc
Bonn, Germany
Diese Nachricht besteht zu 100% aus wiederverwerteten Bits.
Ungeöffnet mindestens haltbar bis: (siehe Rückseite)


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 254 bytes --]