Re: pthread_exit() patch to fix cancellation bug

[Nick Garnett <nickg at ecoscentric dot com>]

Dan Jakubiec <djakubiec@yahoo.com> writes:

> Attached is a patch to fix a cancellation problem with
> cleanup handlers for POSIX threads.  Your feedback is
> appreciated.
> 
> This bug affects POSIX threads that are configured
> with cleanup handlers via the pthread_cleanup_push()
> function.  The bug is that thread cancellation is not
> disabled by pthread_exit() while running thread
> cleanup handlers.
> 
> The symptoms are as follows: if a thread is cancelled
> with pthread_cancel() while cancellation is enabled in
> the thread, any (cancellable) system call issued out
> of its cleanup handler(s) will once again trip a
> cancellation point.  This causes the system call to be
> aborted AND the pthread_exit() routine to be
> re-entered.  Although pthread_exit() seems to deal
> with this okay, the net effect is that the system
> calls in the cleanup handlers do not execute and
> system resources do not get cleaned up.
> 
> According to POSIX:
> ---POSIX
> When a cancellation request is acted upon, the
> routines in the list are invoked one by one in LIFO
> sequence; that is, the last routine pushed onto the
> list (Last In) is the first to be invoked (First Out).
> The thread invokes the cancellation cleanup handler
> with cancellation disabled until the last cancellation
> cleanup handler returns. When the cancellation cleanup
> handler for a scope is invoked, the storage for that
> scope remains valid. If the last cancellation cleanup
> handler returns, thread execution is terminated and a
> status of PTHREAD_CANCELED is made available to any
> threads joining with the target. The symbolic constant
> PTHREAD_CANCELED expands to a constant expression of
> type ( void *) whose value matches no pointer to an
> object in memory nor the value NULL.
> ---POSIX
> 
> 
> So, this fix modifies pthread_exit() to disable
> cancellation for the thread prior to calling its
> cleanup handlers.  I went ahead and left cancellation
> disabled for the remainder of pthread_exit() for two
> reasons:
> 
> 1) The thread is exitting anyway.
> 
> 2) The key destructors might also call cancellable
> system calls.
> 
> 
> Your comments are appreciated.
> 

This all seems perfectly reasonable, I've now committed your patch.


-- 
Nick Garnett                    eCos Kernel Architect
http://www.ecoscentric.com      The eCos and RedBoot experts