Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Tue, 2 Feb 2021 17:45:47 +0100
From:   Peter Zijlstra <peterz@infradead.org>
To:     Steven Rostedt <rostedt@goodmis.org>
Cc:     Alexei Starovoitov <alexei.starovoitov@gmail.com>,
        Masami Hiramatsu <mhiramat@kernel.org>,
        Nikolay Borisov <nborisov@suse.com>,
        LKML <linux-kernel@vger.kernel.org>,
        Alexei Starovoitov <ast@kernel.org>, bpf <bpf@vger.kernel.org>,
        Josh Poimboeuf <jpoimboe@redhat.com>
Subject: Re: kprobes broken since 0d00449c7a28 ("x86: Replace ist_enter()
 with nmi_enter()")
Message-ID: <YBmBu0c24RjNYFet@hirez.programming.kicks-ass.net>
References: <20210129105952.74dc8464@gandalf.local.home>
 <20210129162438.GC8912@worktop.programming.kicks-ass.net>
 <CAADnVQLMqHpSsZ1OdZRFmKqNWKiRq3dxRxw+y=kvMdmkN7htUw@mail.gmail.com>
 <20210129175943.GH8912@worktop.programming.kicks-ass.net>
 <20210129140103.3ce971b7@gandalf.local.home>
 <20210129162454.293523c6@gandalf.local.home>
 <YBUYsFlxjsQxuvfB@hirez.programming.kicks-ass.net>
 <20210130074410.6384c2e2@oasis.local.home>
 <YBktVT+z7sV/vEPU@hirez.programming.kicks-ass.net>
 <20210202095249.5abd6780@gandalf.local.home>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20210202095249.5abd6780@gandalf.local.home>
Precedence: bulk

On Tue, Feb 02, 2021 at 09:52:49AM -0500, Steven Rostedt wrote:

> But from a handler, you could do:
> 
> 	if (in_nmi())
> 		return;
> 	local_irq_save(flags);
> 	/* Now you are safe from being re-entrant. */

But that's an utter crap thing to do. That's like saying I don't care
about my events, at which point you might as well not bother at all.

And you can still do that, you just get less coverage today than you
used to. You used to throw things under the bus, now you throw more
under the bus. If you didn't care, I can't seem to find myself caring
either.

> Where as there's no equivalent in a NMI handler. That's what makes
> kprobe/ftrace handlers different than NMI handlers.

I don't see how.

> > Also, given how everything can nest, it had better all be lockless
> > anyway. You can get your regular function trace interrupted, which can
> > hit a #DB, which can function trace, which can #BP which can function
> > trace again which can get #NMI etc.. Many wonderfun nestings possible.
> 
> I would call #DB an #BP handlers very special.

They are, just like NMI is special, which is why they're classed
together.

> Question: Do #DB and #BP set "in_interrupt()"? Because the function tracer
> has infrastructure to prevent recursion in the same context.

Sure we _could_ do that, but then we get into the 'fun' problem of
getting a breakpoint/int3 at random places and calling random code and
having deadlocks because they take the same lock.

There was very little that stopped that from happening.

> That is, a
> ftrace handler calls something that gets traced, the recursion protection
> will detect that and prevent the handler from being called again. But the
> recursion protection is interrupt context aware and lets the handler get
> called again if the recursion happens from a different context:

> If #DB and #BP do not change the in_interrupt() context, then the above
> still will protect the ftrace handlers from recursion due to them.

But it doesn't help with:

	spin_lock_irq(&foo); // task context
	#DB
	  spin_lock_irq(&foo); // interrupt context per your above

All you need to do is put a breakpoint on a piece of code that holds a
spinlock and a handler that takes the same spinlock.

There was very little from stopping that.

> That would require refactoring all the code that's been around since 2008.

Because I couldn't tell why/if any of that was correct at all. #DB/#BP
don't play by the normal rules. They're _far_ more NMI-like than they're
IRQ-like due to ignoring IF.