[ofa-general] Re: [PATCH 1/3] Infiniband: make ehca_pd use struct pid pointer rather than pid_t

Roland Dreier rdreier at cisco.com
Tue Mar 18 11:27:00 PDT 2008


 > Right, above checking is based on a very simple policy "creator of a
 > resource is also the owner in term of releasing it" and will not cover
 > other customized patterns. We had a case - believe on sles9, in which
 > a child process manipulated/released resources from parent one, and
 > it was not easy to find the bug.

Hmm, I can see how that might be ugly.  On the other hand it doesn't
seem like the unix way for a process not to be able to do something
with a file if it has a valid fd.

 > Wrt/ your actual question: we can remove the tgid stuff from ehca kernel
 > code. When do you expect me to send a patch at latest?

I don't think it's super-urgent.  If you can't do it, say, by the end
of this week, then I'll apply Pavel's patch so we don't block his
progress on namespace stuff.  But I would still like to get a patch to
move this out of ehca at some point, so please don't drop it.

If you guys think there is value in having the checks, then please
send a patch to add the ownership stuff to the uverbs core and we can
argue about it then.

Thanks,
  Roland




More information about the general mailing list