tech-kern archive

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

re: Updated drm code



   On Fri, Jun 19, 2009 at 04:13:31PM +1000, matthew green wrote:
   > this change lets drm work on amd64 where it wants to have mmap
   > work on offsets that are an unsigned value and end up triggering
   > a < 0 test inside uvm_device.c.  it's kind of a hack but it seems
   > cleaner than any other change i've seen proposed.
   
   Can you please explain why this can't be dealt with more generally (i.e. the
   offset arg from mmap be interpreted as an unsigned value always and the neg
   test restricted to -1)?
   
   Last time jared tried to explain it to me some (maybe broken) int32_t ->
   off_t transition was involved "somewhere", but I may misremember this.


we could audit *all* the driver mmap routines to check again,
and then keep the audit up for newer drivers.

this is more of a stop-gap solution that is the least ugly
method i've seen for this so far.  ideally, it would go away
the same way that D_MPSAFE should go away some day :-)


.mrg.


Home | Main Index | Thread Index | Old Index