MIME-Version: 1.0
In-Reply-To: <20150724162550.GG3612@1wt.eu>
References: <CALCETrUf9s-o-ETMiSxxjMGxVeH7di4O9vTi0Oe7wS-RCiVXLA@mail.gmail.com>
 <CA+55aFwR8mHw=wm+Uecy0ERgrD7WbijBn9kj_ZAd47L4GyG5Xw@mail.gmail.com>
 <CALCETrVAzhE7w3BDjqRack54BLncZALbnAOZyeXHx1cSTryy4g@mail.gmail.com>
 <CA+55aFyxs8Q5WrjN9o4Zmfd_4+muLkcoO8cXyv5Nt+Pf8c0TBQ@mail.gmail.com>
 <20150723173105.6795c0dc@gandalf.local.home> <CA+55aFy0-rj7hp3zOUAZD5y5Zp=v6Cu3TG0SHB-buj3oYTJcZg@mail.gmail.com>
 <CALCETrWMgsrgEYWpzPFapOj+-SvZfadDAZ7SH7O8bFsR2b6F1Q@mail.gmail.com>
 <CA+55aFzma9NgODkzz08zpEKSWVnwxuCvwPt_JnO8HaHwRnBPdQ@mail.gmail.com>
 <20150724081326.GO25159@twins.programming.kicks-ass.net> <CALCETrWjzU79ASDK+0RJQyCy6qTdM3FPTa4ZM0d5sVW66yhcug@mail.gmail.com>
 <20150724162550.GG3612@1wt.eu>
From: Andy Lutomirski <luto@amacapital.net>
Date: Fri, 24 Jul 2015 10:21:16 -0700
Message-ID: <CALCETrV2p9j5zQAmEL66YUQv33Z7Bk6mL4eeJUUpcXpvFFafYg@mail.gmail.com>
Subject: Re: Dealing with the NMI mess
To: Willy Tarreau <w@1wt.eu>
Cc: Peter Zijlstra <peterz@infradead.org>,
        Linus Torvalds <torvalds@linux-foundation.org>,
        Steven Rostedt <rostedt@goodmis.org>, X86 ML <x86@kernel.org>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        Borislav Petkov <bp@alien8.de>, Thomas Gleixner <tglx@linutronix.de>,
        Brian Gerst <brgerst@gmail.com>
Content-Type: text/plain; charset=UTF-8
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1282
Lines: 27

On Fri, Jul 24, 2015 at 9:25 AM, Willy Tarreau <w@1wt.eu> wrote:
> On Fri, Jul 24, 2015 at 08:48:57AM -0700, Andy Lutomirski wrote:
>> So by the time we detect that we've hit a watchpoint, the instruction
>> that tripped it is done and we don't need RF.  Furthermore, after
>> reading 17.3.1.1: I *think* that regs->flags withh have RF *clear* if
>> we hit a watchpoint.
>
> Apparently after reading 17.3.1.1, it seems like RF can still be set
> if a data breakpoint triggers in a repeated string instruction before
> the last iteration. However I don't think we care because as long as
> we return to the string instruction, since the data location was already
> visited it won't trigger again so the loss of the flag should be safe.
>

Oh, right.  So my proposal is wrong: it'll clear a watchpoint
incorrectly if we hit it in the middle of a string operation.

So we should either parse dr0..dr3 (whichever one triggered) or do
Peter's think and clear dr7 entirely.  I still prefer just clearing
the breakpoint that triggered.

--Andy
--
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/