On Fri, 21 May 2010, Linus Torvalds wrote:
>
> > Either way, of course, we need the patch back ...
>
> I'll fix it up.
Hmm. Pushed that out as appended, since that is the correct resolve.
HOWEVER - the code still doesn't actually make any sense. It does
if (sk_sleep(sock->sk)) {
and that sk_sleep() today is an inline function that just does
return &sk->sk_wq->wait;
and testing the result of an address-of operation for NULL is almost
certainly totally non-sensical. Sure, it _might_ work (maybe 'wait' is the
first element in the 'sk_wq' data structure, and sk_wq is NULL), but that
kind of code is always total and utterl crap regardless.
So I pushed it out because I had done the resolve already, and it's no
worse than it was before, but it's still a steaming buggy pile of shit.
It being iscsi, I can't bring myself to care. But somebody who does,
should really look at it. The most likely resolution is to remove the test
entirely, since I don't think it's ever valid to have a socket that
doesn't have a sk_wq (there's a _lot_ of unconditional use of sk_sleep()).
Linus
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
