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

Hoang-Nam Nguyen HNGUYEN at de.ibm.com
Tue Mar 18 08:52:27 PDT 2008


Hi Roland!
>  > The task_struct->tgid field is about to become deprecated, due to
>  > pid namespaces make tasks have many pids, not one. The infiniband
>  > driver is one of the code, that still uses it in some places.
>
> Looks fine in terms of the changes it makes, but actually it seems
> that the ehca use of this is completely bogus and the ownership
> checking should be removed.
>
> The core ib_uverbs module has checks that make sure that objects can
> only be accessed through the file that they were created through; of
> course there are tricky ways a file can be passed from one process to
> another, but I don't think we want to disallow userspace processes
> from trying to do interesting stuff as long as it doesn't hurt anything.
>
> In other words-- ehca shouldn't be looking at tgids or anything like
> that at all.  If there are missing checks then they should be in the
> core userspace verbs stuff; but I think what we have is actually OK.
>
> ehca guys, what do you think?
Reason for above checking is to prevent a child process releasing
a resource that the parent process has created and still wants to use.
Do you think that's something we can generalize into ib_core?
Regards
Nam




More information about the general mailing list