linux-c-programming.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* linux and pthread
@ 2004-11-10 20:02 Simon Drouin
  2004-11-11  8:18 ` Jan-Benedict Glaw
  0 siblings, 1 reply; 2+ messages in thread
From: Simon Drouin @ 2004-11-10 20:02 UTC (permalink / raw)
  To: Linux c programming list

Hi all,

Not sure it is the best place to post that. Tell me if you have any 
other suggestion.

I'm running Mandrake 10 and developping a C/C++ program using multiple 
threads. My program has a main thread that manages UI and a second 
thread is in a loop that continuously grabs data from a serial port. The 
communication between the 2 threads is controled by a mutex. The 
grabbing thread holds the mutex for most of the time in its loop. My 
problem is that when the main thread tries to lock the mutex, it is 
suspended forever,  like if the grabbing thread didn't release the mutex 
lock for long enough for the main thread to lock it.

The FAQ on the LinuxThreads page says that it used to be a problem with 
versions < 0.8 of LinuxThreads. In versions 0.8 and later, a thread that 
locked a mutex while another one was holding it was sure to be scheduled 
as soon as the other thread released the mutex. And I suppose Mandrake 
10 has a version >= 0.8 since the faq was pretty old.

Now, I've been trying to figure out which version of LinuxThread I was 
using. Here's the output of find / -name 'libpthread*.so*':

/usr/lib/libpthread.so
/usr/lib/valgrind/libpthread.so
/usr/lib/valgrind/libpthread.so.0
/lib/i686/libpthread-0.10.so
/lib/i686/libpthread.so.0
/lib/tls/libpthread-0.10.so
/lib/tls/libpthread.so.0
/lib/libpthread-0.10.so
/lib/libpthread.so.0

When I do a 'ldd' on my program, I can see it is using 
/lib/tls/libpthread.so.0. Is there a documentation somewhere on what is 
what in this list. Seems like there are 2 versions of libpthread out 
there: the original LinuxThreads lib and the "Native Posix Thread 
Library"(nptl) (correct me if I'm wrong). Some forum I ran into suggests 
that what is in /lib/tls is related to nptl. Is it the case? If so, what 
are all the others? Are they all different builds of LinuxThreads?

ld seems to be falling back on /lib/i686/libpthread.so.0 if I rename the 
/lib/tls directory. If I run my program with /lib/i686/libpthread.so.0, 
my schedulling problem seems to disapeer. Is there any known bugs in 
nptl and if so, where can I get the information?

Thanks in advance.

Simon Drouin

^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: linux and pthread
  2004-11-10 20:02 linux and pthread Simon Drouin
@ 2004-11-11  8:18 ` Jan-Benedict Glaw
  0 siblings, 0 replies; 2+ messages in thread
From: Jan-Benedict Glaw @ 2004-11-11  8:18 UTC (permalink / raw)
  To: Linux c programming list

[-- Attachment #1: Type: text/plain, Size: 1845 bytes --]

On Wed, 2004-11-10 15:02:36 -0500, Simon Drouin <sdrouin@bic.mni.mcgill.ca>
wrote in message <419273DC.3050208@bic.mni.mcgill.ca>:
> Not sure it is the best place to post that. Tell me if you have any 
> other suggestion.

At least, it's appropriate to ask here.

> I'm running Mandrake 10 and developping a C/C++ program using multiple 
> threads. My program has a main thread that manages UI and a second 
> thread is in a loop that continuously grabs data from a serial port. The 
> communication between the 2 threads is controled by a mutex. The 
> grabbing thread holds the mutex for most of the time in its loop. My 

I guess here's the real problem.

> problem is that when the main thread tries to lock the mutex, it is 
> suspended forever,  like if the grabbing thread didn't release the mutex 
> lock for long enough for the main thread to lock it.

Why do you keep the lock locked? Most of the time, mutexes tend to be
unlocked nearly always and only get locked if (eg. for your example) the
serial thread got a byte and queues it in for the UI thread for
displaying/dispatching/whatnot.

Maybe the trick is to actually use a condition here: let the UI thread
(or a new one) just sleep on the condition. If you got a byte, lock a
mutex (if there are other placed that can access your global serial byte
buffer), queue it in, unlock and wake up the thread(s) sleeping in the
condition. Then, all of them (or only the UI thread) can re-display the
buffer state.

MfG, JBG

-- 
Jan-Benedict Glaw       jbglaw@lug-owl.de    . +49-172-7608481             _ O _
"Eine Freie Meinung in  einem Freien Kopf    | Gegen Zensur | Gegen Krieg  _ _ O
 fuer einen Freien Staat voll Freier Bürger" | im Internet! |   im Irak!   O O O
ret = do_actions((curr | FREE_SPEECH) & ~(NEW_COPYRIGHT_LAW | DRM | TCPA));

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2004-11-11  8:18 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2004-11-10 20:02 linux and pthread Simon Drouin
2004-11-11  8:18 ` Jan-Benedict Glaw

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).