Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S964936AbVJUNDh (ORCPT ); Fri, 21 Oct 2005 09:03:37 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S964937AbVJUNDg (ORCPT ); Fri, 21 Oct 2005 09:03:36 -0400 Received: from ms-smtp-04.nyroc.rr.com ([24.24.2.58]:45311 "EHLO ms-smtp-04.nyroc.rr.com") by vger.kernel.org with ESMTP id S964936AbVJUNDg (ORCPT ); Fri, 21 Oct 2005 09:03:36 -0400 Date: Fri, 21 Oct 2005 09:03:18 -0400 (EDT) From: Steven Rostedt X-X-Sender: rostedt@localhost.localdomain To: Yitzchak Eidus cc: linux-kernel@vger.kernel.org Subject: Re: question about code from the linux kernel development ( se ) book In-Reply-To: Message-ID: References: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 2593 Lines: 70 Oh God, please indent code examples, and if your email client strips white space, change your email client. On Fri, 21 Oct 2005, Yitzchak Eidus wrote: > first i am very sorry if it isnt the place to ask questions like this > but i didnt know where else to ask ( i tryed irc channels and i was > send from there to this list ) > anyway: > does this following code look buggy? : [ Indention added ] > DECLARE_WAITQUEUE ( wait , current ); > add_wait_queue ( q , &wait ); > while ( !condition ) { > set_current_stat ( TASK_INTERRUPTABLE ); i > if ( signal_pending ( current ) ) > /* handle signal */ I assume that the signal_pending is the if result and not the schedule. Since there was no indentation I couldn't tell. > schedule ( ); /* Moved brace down added */ } > set_current_state ( TASK_RUNNING ); > remove_wait_queue ( q , &wait ); Before I go to your questions, I'll first answer that this _is_ buggy code. If the condition is checked and you are woken up between the while (!condition) and the set_current_state, then you will end up sleeping forever or until someone sends you a signal. > first:doesnt in the way from checking the !condition to > set_current_state the condition can be changed no? Yes, and then put it again after schedule. > > second:why not putting the schedule ( ); right after the > set_current_state ( ) , what the point in checking the if ( > signal_pending ( ) first, if the proccess doesnt started to sleep yet? Yes, I would put the signal_pending check after the schedule (as most of the kernel does this). > third: in the cleaning in the way from putting the set_current_state ( > TASK_RUNNING ) into remove_wait_queue , cant the queue wait list ( q ) > wake up again the wait procsess? Yeah, so? There's no harm in that, except for an extra cpu cycles that are done to wake it up. That, I wouldn't change. > ( thnks for the help , please if it can be done answer quickly i am > tanker in the idf and need to come back to the army soon , ( no > internet there... ) ) BTW, this is not an IRC, we use normal capitalization and normal spelling (when we know how to spell a word ;-). So the next time you send to the list, send it as if you were writing a serious letter, or you may just be ignored. (as you might have been if I didn't respond). -- Steve - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/