Peter Zijlstra wrote:
On Mon, 2010-01-18 at 17:40 +0100, Petr Baudis wrote:
On Fri, Jan 15, 2010 at 06:11:24PM +0100, Peter Zijlstra wrote:
On Fri, 2010-01-15 at 08:58 -0800, Darren Hart wrote:
The futex man-pages (2,7) from version 3.15-1 and at least a few earlier
versions are considerably out of date with respect to the current
implementation. They don't document newer op codes, such as:
FUTEX_WAIT_BITSET
FUTEX_WAKE_BITSET
FUTEX_WAKE_OP
FUTEX_LOCK_PI
FUTEX_UNLOCK_PI
FUTEX_WAIT_REQUEUE_PI
FUTEX_CMP_REQUEUE_PI
These new wait op codes now use absolute timeouts, while the original
FUTEX_WAIT uses a relative timeout. Lastly, and most importantly, glibc
has removed the futex() wrapper to syscall(SYS_futex, ...). With that in
mind, would people prefer that we simply remove the futex man-pages
(2,7)?
A readable text about the interaction between the futex value and the
various futex ops and how to build proper locking primitives with them
would be helpful I think, however doing that in the form of a locking
library (as has been suggested at KS) seems plenty fine to me.
Readable code is much better than rambling English at conveying this
stuff.
http://people.redhat.com/drepper/futex.pdf is pretty good description,
unfortunately also outdated by now. Perhaps it would be interesting to
convert this into some kind of wiki form with Ulrich's permission?
People keep saying that, but I found it highly unreadable.
Interesting. I found it to be a good introduction into the usage of
futexes as well as cautionary of the many non-obvious race conditions
possible with locking structures built on futexes.
--
Darren Hart
IBM Linux Technology Center
Real-Time Linux Team
--
To unsubscribe from this list: send the line "unsubscribe linux-man" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
