[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <1191421736.5599.18.camel@lappy>
Date: Wed, 03 Oct 2007 16:28:56 +0200
From: Peter Zijlstra <a.p.zijlstra@...llo.nl>
To: Tetsuo Handa <penguin-kernel@...ove.SAKURA.ne.jp>
Cc: jmorris@...ei.org, linux-kernel@...r.kernel.org,
linux-security-module@...r.kernel.org, chrisw@...s-sol.org
Subject: Re: [TOMOYO 05/15](repost) Domain transition handler functions.
On Wed, 2007-10-03 at 23:19 +0900, Tetsuo Handa wrote:
> Hello.
>
> Thank you for pointing out.
>
> Peter Zijlstra wrote:
> > > Currently, TOMOYO Linux avoids read_lock, on the assumption that
> > > (1) First, ptr->next is initialized with NULL.
> > > (2) Later, ptr->next is assigned non-NULL address.
> > > (3) Assigning to ptr->next is done atomically.
> > (4) wmb after asigning ptr->next
> > (5) rmb before reading ptr->next
> Excuse me, but I didn't understand why (4) and (5) are needed.
>
> append_function() {
>
> down(semaphore_for_write_protect);
> ...
> ptr = head;
> while (ptr->next) ptr = ptr->next;
> ptr->next = new_entry;
> ...
> up(semaphore_for_write_protect);
>
> }
If at all possible, use struct mutex.
> read_function() {
>
> for (ptr = head; ptr; ptr = ptr->next) {
> ...
> }
>
> }
>
> Are (4) and (5) needed even when (3) is exclusively protected by down() and up() ?
the up() would do 4. 5 ensures another cpu will actually see it. Althoug
in practise the various cache invalidations driven by the workload will
ensure it will become visible eventually anyway.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists