tech-kern archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Idea for *at system calls



On Tue, Oct 06, 2009 at 04:22:12AM -0400, der Mouse wrote:
> > I had an idea how to implement the *at system calls in a less nasty
> > way.  We invent to new system calls:
> 
> >     int push_thread_cwd(int filedes);
> >     int pop_thread_cwd();
> 
> I don't think that's quite right.  A signal delivered between push and
> pop would see the thread's (or process's - these calls aren't
> threading-only, are they?) cwd set to the "wrong" value.

I think making the point of making these calls thread specific is
to try to avoid the inconsistent cwd you are describing.
Is there a well defined behaviour for how a per-thread cwd is supposed
 to behave with regards to signals?  If not, then the fact that it 
 might be different during this call doesn't seem too bad.
On the other hand, having to do three separate system calls does kind of suck.

eric


Home | Main Index | Thread Index | Old Index