Hello!
> > Do you mean that this will be enough, and similar thing in tty_io.c?
> Only not here! synchronize_irq() must follow fasync list modification.
> You modify list and then wait for readers. See?
Nope.
As I see it will be:
CPU0 CPU1
kill_fasync unattaching this list node
start to work on next (*)
list node kfree_s it.
-----------------------------------------------
Access fasync->next >(OOPS)
Or do you mean that synchronize_irq must be inserted in (*) (see patch)?
If this is not right, too, then show your version, pls.
--- socket.c.orig Mon Feb 21 16:38:03 2000
+++ socket.c Mon Feb 21 16:39:34 2000
@@ -521,6 +521,7 @@
if(fa!=NULL)
{
fa->fa_fd=fd;
+ synchronize_irq();
kfree_s(fna,sizeof(struct fasync_struct));
release_sock(sock->sk);
return 0;
@@ -536,6 +537,7 @@
if (fa!=NULL)
{
*prev=fa->fa_next;
+ synchronize_irq();
kfree_s(fa,sizeof(struct fasync_struct));
}
}
> Correct reader will see either list before modification or after modification,
> both of them are valid. kill_fasync() is correct reader,
> I do not know about tty_io.c.
In tty_io.c we have function fasync_helper, which is much the same as
socket.c::sock_fasync (besides that it works not with sockets),
and the only difference is that in fasync_helper we *do* have sti/cli
(Alan quoted that part from fasync_helper), and sock_fasync does not have
sti/cli.
Bye,
Oleg
-- [RAVE *SUCKS* TEAM] [Team $Ui(iDE] [Trans-M Must Die!] [Sepultura Fans Team] [Tolkien forever team]- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Wed Feb 23 2000 - 21:00:27 EST