Re: [patch 1/2] fork_connector: add a fork connector

[Jay Lan <jlan@engr.sgi.com>]

Paul Jackson wrote:
> Guillaume wrote:
> 
>>  The goal of the fork connector is to inform a user space application
>>that a fork occurs in the kernel. This information (cpu ID, parent PID
>>and child PID) can be used by several user space applications. It's not
>>only for accounting. Accounting and fork_connector are two different
>>things and thus, fork_connector doesn't do the merge of any kinds of
>>data (and it will never do). 
> 
> 
> Yes - it is clear that the fork_connector does this - inform user space
> of fork information <cpu, parent, child>.  I'm not saying that
> fork_connector should merge data; I'm observing that it doesn't, and
> that this would seem to serve the needs of accounting poorly.

Paul,

You probably can look at it this way: the accounting data being
written out by BSD are per process data and the fork connector
provides information needed to group processes into process
aggregates.

Thanks,
  - jay

> 
> Out of curiosity, what are these 'several user space applications?'  The
> only one I know of is this extension to bsd accounting to include
> capturing parent and child pid at fork.  Probably you've mentioned some
> other uses of fork_connector before here, but I missed it.
> 
> 
>>The relayfs is done, like Evgeniy said, for large amount of
>>datas. So I think that it's not suitable for what we want to achieve
>>with the fork connector.
> 
> 
> I never claimed that relayfs was appropriate for fork_connector.
> 
> I'm not trying to tape a rock to Evgeniy's screwdriver.  I'm saying that
> accounting looks like a nail to me, so let us see what rocks and hammers
> we have in our tool box.
>