linux-kernel - Re: [PATCH 2/2] ftrace: fast path for do_ftrace_mod

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]

Message-ID: <49C09C77.3050300@cn.fujitsu.com>
Date:	Wed, 18 Mar 2009 15:02:15 +0800
From:	Lai Jiangshan <laijs@...fujitsu.com>
To:	Steven Rostedt <srostedt@...hat.com>
CC:	Ingo Molnar <mingo@...e.hu>, LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 2/2] ftrace: fast path for do_ftrace_mod_code()

Steven Rostedt wrote:
> 
> I'm a bit nervous about this code. We do not get much benefit from it,
> because the NMI case is an anomaly, and is not a fast path anyway. This
> code only happens when we are running the stop_machine, and this adds
> added complexity for little benefit.
> 
> The original patch was to prevent an actual live lock I got in one of my
> tests. The problem was that the failure of the write caused a printk
> stack dump. But the time it took the print to go out over the serial was
> long enough that the next NMI triggered when it finished. The new NMI
> hit the same error and did another print. Thus, all I got was a lot of
> prints out over the serial, but the system was dead.
> 

Thank you. I understand.


> I like the first patch. but you remove the protection there. It should
> have been in this patch. But it should have still added the
> functionality of the previous method.

I separated it into two parts, I thought it will good for review.
But I wrote two bad patches.

>> @@ -161,6 +167,7 @@ do_ftrace_mod_code(unsigned long ip, void *new_code)
>>  {
>>  	mod_code_ip = (void *)ip;
>>  	mod_code_newcode = new_code;
>> +	mod_code_no_write = 0;
> 
> Here's another issue, if mod_code_status failed, we do not want to have
> mod_code_no_write become zero again. The logic may indeed prevent this,
> but I rather have the logic be straight forward, and just set this to
> one when we have a failure and forget about it. Yes, it is a bit more
> expensive, but it makes the code clearer.

It confused me.

do_ftrace_mod_code() is called sequently, mod_code_no_write should become zero
in new calls.

Not like old code, when the first patch is applied, there is no NMI
is attempt to call probe_kernel_write() when we just enter do_ftrace_mod_code(),
so setting mod_code_no_write to 0 is safe.(Because the flag is not set)

Lai.


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/