RFC: No more deadlock detection for POSIX locks

From: Matthew Wilcox (willy@debian.org)
Date: Thu Oct 10 2002 - 15:42:33 EST

Next message: Paolo Ciarrocchi: "[Benchmark] Contest 0.50 against 2.5.41-mm2"
Previous message: Matt D. Robinson: "[PATCH] 2.5.41: lkcd (8/8): dump driver and build files"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

The deadlock detection for posix locks really isn't worth anything
any more. It was always slightly dubious, since a parent/child could
remove each other's locks (thanks, POSIX!). But now it's really dubious
since we store the TID, not the PID of the requesting process, and any
thread can unlock a lock set by another thread.

Here's one situation in which it can falsely return -EDEADLK:

TID 1001, PID 1002 takes lock A
TID 1003, PID 1004 takes lock B
TID 1001, PID 1005 takes lock B, blocks
TID 1003, PID 1004 takes lock A, gets -EDEADLK.
Even though (1001,1002) isn't blocking on any lock and will release lock A
in the future.

So how about we just delete the nasty deadlock detection code? I've never
been fond of the user-triggerable O(N^2) algorithm, and we're permitted
to not implement it (POSIX suggests applications set a timer to detect
deadlock themselves, so anyone writing a portable application is already
doing this).

Objections?

-- 
Revolutions do not require corporate support.
-
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/

Next message: Paolo Ciarrocchi: "[Benchmark] Contest 0.50 against 2.5.41-mm2"
Previous message: Matt D. Robinson: "[PATCH] 2.5.41: lkcd (8/8): dump driver and build files"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

This archive was generated by hypermail 2b29 : Tue Oct 15 2002 - 22:00:39 EST