Re: [PATCH 07/10] uprobes/x86: Introduce arch_uretprobe_is_alive()

From: Oleg Nesterov <oleg@redhat.com>
To: Pratyush Anand <panand@redhat.com>
Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com>,
	Anton Arapov <arapov@gmail.com>,
	David Long <dave.long@linaro.org>,
	Denys Vlasenko <dvlasenk@redhat.com>,
	"Frank Ch. Eigler" <fche@redhat.com>,
	Ingo Molnar <mingo@kernel.org>, Jan Willeke <willeke@de.ibm.com>,
	Jim Keniston <jkenisto@us.ibm.com>,
	Mark Wielaard <mjw@redhat.com>,
	Srikar Dronamraju <srikar@linux.vnet.ibm.com>,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH 07/10] uprobes/x86: Introduce arch_uretprobe_is_alive()
Date: Wed, 20 May 2015 17:51:56 +0200	[thread overview]
Message-ID: <20150520155156.GA1343@redhat.com> (raw)
In-Reply-To: <5559D627.7080201@redhat.com>

Hi Pratyush,

sorry for delay, vacation.

On 05/18, Pratyush Anand wrote:
> Hi Oleg,
>
> On Monday 04 May 2015 06:19 PM, Oleg Nesterov wrote:
>> +bool __weak arch_uretprobe_is_alive(struct arch_uretprobe *auret, struct pt_regs *regs)
>> +{
>> +	return true;
>> +}
>
> IIUC, then this function should return false when both auret and regs
> are corresponding to same retprobe, else we need to return true, right?

Not sure I understand what you mean...

This function should return false when we know that this return_instance
was invalidated by longjmp().

> If that is the case, then should n't following work for all the cases:
>
> return sp != auret->sp;

No, this can't work.

On x86 "sp == auret->sp" only right after the "call" insn, the stack
can grow after that but this does not mean that this instance is no
longer valid.

Oleg.