Re: [Nbd] NBD, DISCARD/TRIM

[Wouter Verhelst <w@...112...>]

On 29-03-13 15:30, Paul Clements wrote:
> On Thu, Mar 28, 2013 at 12:36 PM, folkert <folkert@...421...> wrote:
>>> Very good. Would it be at all possible to try the attached patch,
>>> which includes some additional debugging information, so that we can
>>> see why this is happening?
>>
>> It is unclear which process is the trigger:
>>
>> [   71.078260] got ERESTARTSYS: sig:17 code:262145 sender:2853
> 
> OK, so sig 17 is SIGCHLD???

AAAARGh!

> This is weird. I'm not that familiar with
> latest upstream nbd-client, but is there maybe some fork or thread
> that is happening?

Yes, that's for the partitioning stuff. Originally I had the parent do
that, but then the "-nofork" parameter would behave counterintuitive.

Now we fork(), but that obviously means I need to eat SIGCHLD. Sigh.

>> root      2852     1  3 17:31 ?        00:00:03 nbd-client 172.16.86.1 12342 /dev/nbd0
>> root      2854     2 46 17:31 ?        00:00:43 [nbd0]
>>
>> I would almost say that it is something in nbd-client itself?
> 
> Agreed...it certainly looks that way...
> 
> I'll look at latest nbd-client source and see if anything looks
> suspicious, but in the meantime if anyone more familiar with that code
> can chime in, that would be great.

-- 
Copyshops should do vouchers. So that next time some bureaucracy
requires you to mail a form in triplicate, you can mail it just once,
add a voucher, and save on postage.