linux-kernel - Re: [BUGFIX PATCH tip/master V2 3/3] kprobes/arm: Fix a possible deadlock case in kretprobe

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Thu, 9 Feb 2017 16:49:00 +0000
From:   Russell King - ARM Linux <linux@...linux.org.uk>
To:     Masami Hiramatsu <mhiramat@...nel.org>
Cc:     Ingo Molnar <mingo@...hat.com>, linux-kernel@...r.kernel.org,
        Peter Zijlstra <peterz@...radead.org>,
        Ananth N Mavinakayanahalli <ananth@...ux.vnet.ibm.com>,
        Thomas Gleixner <tglx@...utronix.de>,
        "H . Peter Anvin" <hpa@...or.com>, Jon Medhurst <tixy@...aro.org>,
        Wang Nan <wangnan0@...wei.com>,
        Catalin Marinas <catalin.marinas@....com>,
        Will Deacon <will.deacon@....com>,
        "David A . Long" <dave.long@...aro.org>,
        Sandeepa Prabhu <sandeepa.s.prabhu@...il.com>
Subject: Re: [BUGFIX PATCH tip/master V2 3/3] kprobes/arm: Fix a possible
 deadlock case in kretprobe

On Fri, Feb 10, 2017 at 01:32:22AM +0900, Masami Hiramatsu wrote:
> Fix a possibility of deadlock case in kretprobe on arm
> implementation. There may be a chance that the kretprobe
> hash table lock can cause a dead lock.
> 
> The senario is that a user puts 2 kretprobes, one on normal
> function and one on a function which can be called from
> somewhare which can interrupt in irq disabled critical
> section like FIQ.

If we:
- hit a kernel tracing feature from FIQ context
- the tracing feature takes a lock
- the lock is also taken elsewhere on the same CPU with IRQs disabled

we will quite simply deadlock.

In this case, kretprobe_hash_lock() takes the hlist_lock using
raw_spin_lock_irqsave().

Now, from what I can see in the kprobes code, this lock is taken in
other contexts (eg, kprobe_flush_task()), which means even with this
fix, it's still risky if a kprobe is placed on a FIQ-called function.

> In this case, if the kernel hits the 1st kretprobe on a
> normal function return which calls trampoline_handler(),
> acquire a spinlock on the hash table in kretprobe_hash_lock()
> and disable irqs. After that, if the 2nd kretprobe is kicked
> from FIQ, it also calls trampoline_handler() and tries to
> acquire the same spinlock (since the hash is based on
> current task, same as the 1st kretprobe), it causes
> a deadlock.

So my deadlock scenario is:

- we're in the middle of kprobe_flush_task()
- FIQ happens, calls trampoline_handler()
- deadlock in kretprobe_hash_lock()

>From what I can see, kretprobes in FIQ are just unsafe.

I suspect that avoiding these deadlocks means that we have to deny
kprobes from FIQ context - making trampoline_handler() return
immediately if in_nmi() is true.

-- 
RMK's Patch system: http://www.armlinux.org.uk/developer/patches/
FTTC broadband for 0.8mile line: currently at 9.6Mbps down 400kbps up
according to speedtest.net.