[9fans] getting a pid from threadpostmountsrv

[9fans] getting a pid from threadpostmountsrv

[Skip Tavakkolian]

is there an easy way to get the pid of the proc that
threadpostmountsrv creates to run srv()?

thought about sending the current pid over a global
channel in Srv.attach(); seems wrong. maybe a
Srv.start() is warranted?

Re: [9fans] getting a pid from threadpostmountsrv

[Russ Cox]

> is there an easy way to get the pid of the proc that
> threadpostmountsrv creates to run srv()?
>
> thought about sending the current pid over a global
> channel in Srv.attach(); seems wrong. maybe a
> Srv.start() is warranted?

why do you care?

russ

Re: [9fans] getting a pid from threadpostmountsrv

[Skip Tavakkolian]

> why do you care?

i handle part of a ns that is exported, with an in-proc
server.  i need to be able to post a note to it.

Re: [9fans] getting a pid from threadpostmountsrv

[Russ Cox]

> i handle part of a ns that is exported, with an in-proc
> server.  i need to be able to post a note to it.

why?

i know this line of questioning is annoying, but
unless you are posting the note to interrupt a
system call, it's really not a good idea to have
notes flying around in threaded programs.
the problem is that notes can come in at arbitrary
times and really screw things up in unexpected ways.
there is always a better, more coordinated way.

and in the case of interrupting a system call
you'd be better off using iocall and iointerrupt.

russ

Re: [9fans] getting a pid from threadpostmountsrv

[Skip Tavakkolian]

> why?

shouldn't an in-prog fs -- notefs in cpu.c -- be possible with lib9p?

for snsa ref implementaton, a sibling proc running srv() handles
/dev/sysctl, that controls the workings of the agent.  agent needs to
be able to kill the srv proc if the exec'd exportfs dies.

Re: [9fans] getting a pid from threadpostmountsrv

[Russ Cox]

> shouldn't an in-prog fs -- notefs in cpu.c -- be possible with lib9p?
>
> for snsa ref implementaton, a sibling proc running srv() handles
> /dev/sysctl, that controls the workings of the agent.  agent needs to
> be able to kill the srv proc if the exec'd exportfs dies.

sure but you didn't answer the question of why you are
sending notes.  if you just want to clean up on exit
you can use threadexitsall and the right things happen.
if you are trying to kill off one proc but not the whole
program using a note, you are asking for trouble.

russ

Re: [9fans] getting a pid from threadpostmountsrv

[Skip Tavakkolian]

> and in the case of interrupting a system call
> you'd be better off using iocall and iointerrupt.

for now, setting Srv.leavefdsopen has the right effect. it
also leaves the proc in the same process group.