[openib-general] Re: [PATCH 10 of 20] ipath - support for userspace apps using core driver

[Roland Dreier]

    Bryan> We're mapping some memory that the chip DMAs to into
    Bryan> userspace, so that user processes can spin on memory
    Bryan> locations without going through the kernel.  The
    Bryan> SetPageReserved hack is an attempt to stop the VM from
    Bryan> reclaiming those pages from us once a user process exits.

    Bryan> I realise that it's surely bogus, and I'd be thrilled to do
    Bryan> something correct instead.  We've tried doing both
    Bryan> SetPageReserved and get_page, but it hasn't work out too
    Bryan> well so far.

What's wrong with doing get_page()?  Surely the VM won't take pages
that you hold a reference to.

 - R.