[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20160307120318.62a168ed@gandalf.local.home>
Date: Mon, 7 Mar 2016 12:03:18 -0500
From: Steven Rostedt <rostedt@...dmis.org>
To: Alan Stern <stern@...land.harvard.edu>
Cc: Sedat Dilek <sedat.dilek@...il.com>,
Jiri Kosina <jikos@...nel.org>, Tejun Heo <tj@...nel.org>,
Lai Jiangshan <jiangshanlai@...il.com>,
Benjamin Tissoires <benjamin.tissoires@...hat.com>,
Paul McKenney <paulmck@...ux.vnet.ibm.com>,
Andy Lutomirski <luto@...nel.org>,
LKML <linux-kernel@...r.kernel.org>,
USB list <linux-usb@...r.kernel.org>,
Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
Peter Zijlstra <peterz@...radead.org>,
Ingo Molnar <mingo@...hat.com>
Subject: Re: [PATCH] usbhid: Fix lockdep unannotated irqs-off warning
On Mon, 7 Mar 2016 11:41:37 -0500 (EST)
Alan Stern <stern@...land.harvard.edu> wrote:
> It's hard to call this a compiler bug, but perhaps it is -- I don't
> know how programmers are supposed to tell CLANG that a subroutine
> modifies the Interrupt Flag in a way that the compiler shouldn't mess
> up.
Really! This is what's is happening??
Clang takes this:
if (!--hid->open) {
spin_unlock_irq(X);
do_something();
} else {
spin_unlock_irq(X);
}
Thus it's basically doing:
FLAG = !--hid->open;
push flags;
spin_unlock_irq(X)
pop flags;
if (FLAG zero set) {
do_something();
}
OUCH!!! There's gotta be a way to turn that off, otherwise Clang can
not be used to compile the kernel.
Nice detective work.
-- Steve
Powered by blists - more mailing lists