[openib-general] Re: [PATCH][RFC][0/4] InfiniBand userspace verbs implementation

Roland Dreier roland at topspin.com
Mon Apr 25 17:04:02 PDT 2005


    Caitlin> Every RDMA related interface specification that I know of
    Caitlin> specifically excludes support of RDMA resources being
    Caitlin> inherited by child processes, with the warning that
    Caitlin> excellent implementations will give the child process an
    Caitlin> error for attempting to use the parent's RDMA resources.
    Caitlin> More streamlined implementations will simply be
    Caitlin> unpredictable.

    Caitlin> As for forking while the parent has a pending read: since
    Caitlin> the parent has not reaped the completion at the time of
    Caitlin> the fork the buffers in question are undefined. The
    Caitlin> child's buffers will be consistent, that is they are
    Caitlin> undefined.

I think you've missed the point: unless a process sets VM_DONTCOPY on
its RDMA memory regions, then incorrect memory mappings may be used if
the app does something as simple as calling system("ls").

 - R.



More information about the general mailing list