Re: [patch V4 part 5 02/31] x86/entry: Provide helpers for execute on irqstack

From: Lai Jiangshan <jiangshanlai+lkml@gmail.com>
To: Thomas Gleixner <tglx@linutronix.de>
Cc: LKML <linux-kernel@vger.kernel.org>,
	x86@kernel.org, "Paul E. McKenney" <paulmck@kernel.org>,
	Andy Lutomirski <luto@kernel.org>,
	Alexandre Chartre <alexandre.chartre@oracle.com>,
	Frederic Weisbecker <frederic@kernel.org>,
	Paolo Bonzini <pbonzini@redhat.com>,
	Sean Christopherson <sean.j.christopherson@intel.com>,
	Masami Hiramatsu <mhiramat@kernel.org>,
	Petr Mladek <pmladek@suse.com>,
	Steven Rostedt <rostedt@goodmis.org>,
	Joel Fernandes <joel@joelfernandes.org>,
	Boris Ostrovsky <boris.ostrovsky@oracle.com>,
	Juergen Gross <jgross@suse.com>, Brian Gerst <brgerst@gmail.com>,
	Mathieu Desnoyers <mathieu.desnoyers@efficios.com>,
	Josh Poimboeuf <jpoimboe@redhat.com>,
	Will Deacon <will@kernel.org>
Subject: Re: [patch V4 part 5 02/31] x86/entry: Provide helpers for execute on irqstack
Date: Sun, 10 May 2020 12:33:33 +0800	[thread overview]
Message-ID: <CAJhGHyCGOfD6pQQ58CysfjyvqPLc9dc54WpqE2wSuVvyEK-9jw@mail.gmail.com> (raw)
In-Reply-To: <20200505135828.316937774@linutronix.de>

On Tue, May 5, 2020 at 10:19 PM Thomas Gleixner <tglx@linutronix.de> wrote:
>
> Device interrupt handlers and system vector handlers are executed on the
> interrupt stack. The stack switch happens in the low level assembly entry
> code. This conflicts with the efforts to consolidate the exit code in C to
> ensure correctness vs. RCU and tracing.
>
> As there is no way to move #DB away from IST due to the MOV SS issue, the
> requirements vs. #DB and NMI for switching to the interrupt stack do not
> exist anymore. The only requirement is that interrupts are disabled.

Hi, tglx and Andy Lutomirski,

Is there any information about "no way to move #DB away from IST
due to the MOV SS issue"? IST-based #DB results to ist_shift(for
nested #DB) and debug_idt(for #NMI vs. #DB) which are somewhat
ugly. If IST-less #DB should work, debug stack should be switched
in software manner like interrupt stack.

There was a "POP/MOV SS" CVE/issue about #BP which lead to
moving #BP to IST-less by d8ba61ba58c8
(x86/entry/64: Don't use IST entry for #BP stack)

#DB #BP are considered as #NMI due to their super-interrupt
ability. But the kernel has much more control over #DB and #BP
which can be disabled by putting the code snip into non-instrument
sections like __entry noinstr etc.

Is it possible to implement IST-less #DB?

Thanks,
Lai