Date: Fri, 24 Jul 2015 15:05:11 -0400
From: Steven Rostedt <rostedt@goodmis.org>
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>, Willy Tarreau <w@1wt.eu>,
        Andy Lutomirski <luto@amacapital.net>, 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>
Subject: Re: Dealing with the NMI mess
Message-ID: <20150724150511.72a09300@gandalf.local.home>
In-Reply-To: <CA+55aFzhf7uGw_Mf_WMS-FH9236-Q-A13zfiVpLZqUKrVRYY=A@mail.gmail.com>
References: <CA+55aFzma9NgODkzz08zpEKSWVnwxuCvwPt_JnO8HaHwRnBPdQ@mail.gmail.com>
	<20150724081326.GO25159@twins.programming.kicks-ass.net>
	<20150724075841.40f209f4@gandalf.local.home>
	<20150724124304.GH19282@twins.programming.kicks-ass.net>
	<20150724090342.6d11e16d@gandalf.local.home>
	<20150724132128.GA3612@1wt.eu>
	<20150724103127.3c3f4693@gandalf.local.home>
	<20150724145901.GB3612@1wt.eu>
	<20150724111621.34713023@gandalf.local.home>
	<20150724152637.GC3612@1wt.eu>
	<20150724153054.GK19282@twins.programming.kicks-ass.net>
	<CA+55aFz0=bVcRs2M3etVqU3+tgz2ZPJDsTiUieWpshVuiR_utg@mail.gmail.com>
	<CA+55aFzhf7uGw_Mf_WMS-FH9236-Q-A13zfiVpLZqUKrVRYY=A@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 2161
Lines: 65

On Fri, 24 Jul 2015 11:41:55 -0700
Linus Torvalds <torvalds@linux-foundation.org> wrote:

> On Fri, Jul 24, 2015 at 11:29 AM, Linus Torvalds
> <torvalds@linux-foundation.org> wrote:
> >
> > So in the #DB handler, we would basically only clear instruction
> > breakpoints, and only when they trigger. If we have a data breakpoint
> > that triggers (even in kernel mode, and with interrupts disabled), let
> > it trigger and return with "ret" anyway. No biggie.
> 
> So we'd not only look at "which breakpoint triggered", we'd also look
> at the actual debug register and check that "R/Wn == 0", and only
> disable it for that case.
> 
> So you'd read %dr6 and %dr7, and then iterate 0..3 and check whether
> it triggerd (bit #n in %dr6), and that R/Wn (bits 16-17+n*4 of %dr7)
> is zero, and if so, clear LGn bits (bits 0-1+n*2) in %dr7.
> 
> Something like
> 
>         unsigned long mask = 0;
>         unsigned int dr6 = debug_read(6);
>         unsigned int dr7 = debug_read(7)
>         int i;
> 
>         for (i = 0; i < 4; i++) {
>                 if ((dr6 >> i) & 1) {
>                         if (!((dr7 >> (4*i+16)) & 3))
>                                 mask |= 3 << (i*2);
>                 }
>         }
> 
>         if (mask)
>                 debug_write(dr7 & ~mask, 7);

Macros would be nice for readability.

	for (i = 0; i < 4; i++) {
		if ((dr6 >> i) & 1) {
			int shift = DR_CONTROL_SIZE * i + DR_CONTROL_SHIFT;
			if (!((dr7 >> shift) & DR_RW_READ))
				mask |= (DR_LOCAL_ENABLE|DR_GLOBAL_ENABLE) << (i * DR_ENABLE_SIZE);
		}
	}

-- Steve

> 
> (yeah, I could easily have screwed that up)
> 
> But the above should only clear bits in dr7 that are actually
> associated with the instruction breakpoint that triggered, and since
> it's a _kernel_ instruction breakpoint, not a user one, we can clear
> it and forget it. No need to re-enable at all.
> 
> Hmm?
> 
>                        Linus

--
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/