Re: [PATCH RESEND 3.18-rc3 v2 0/2] trace: kdb: Bug fixes for ftdump

[Daniel Thompson <daniel.thompson@linaro.org>]

On 14/11/14 02:16, Steven Rostedt wrote:
> On Mon, 10 Nov 2014 09:41:32 +0000
> Daniel Thompson <daniel.thompson@linaro.org> wrote:
> 
>> On 06/11/14 13:27, Steven Rostedt wrote:
>>> On Thu,  6 Nov 2014 12:41:54 +0000
>>> Daniel Thompson <daniel.thompson@linaro.org> wrote:
>>>
>>>> This patchset makes it possible to use kdb's ftdump command without
>>>> panicing, crashing or livelocking. The ftdump command cannot be used
>>>> at all without these changes.
>>>>
>>>> IIRC this patches are still pending Jason's ack.
>>>
>>> I haven't heard from Jason in a long time. Is he still active?
>>
>> [sorry for the delay, I wanted to give Jason a chance to answer this]
>>
>> Very occasionally.
>>
>> I can't find anything on lkml in the last three months, and I have
>> unreviewed kdb patches that stretch back well beyond that.
>>
>> That said he still helps people on kgdb-bugreport@ from time-to-time
>> (and as recently as last week). I've also had a little bit of private
>> contact although nothing very recent.
>>
>> On that basis I'd say you shouldn't feel guilty if you have to accept a
>> change here without an ack.
>>
> 
> He had more than enough time to respond. OK, I'll take it.
> 
> Looking at the first patch, I notice that there's no protection of the
> static buffer_iter array. I also noticed that there's no protection of
> the static iter itself (which was there before your patch). I take it
> that this code is not re-entrant.

No.

k(g)db halts all other processors before entering the command dispatch
loop and it forces a kernel panic if the debugger is reentered by the
same CPU.


Daniel.