I just applied mthca_cosmetic_icm_page_size.patch with one minor
change: reviewing the patch made me realize that the cleanup code in
mthca_unmap_eq_icm() was wrong.  We only map one _HCA_ page in
mthca_map_eq_icm(), which means that we should only unmap one HCA page
in mthca_unmap_eq_icm().  The old code was unmapping enough HCA pages
to cover a kernel (PAGE_SIZE) page, which would be wrong if PAGE_SIZE
is not 4096.  This has been harmless until now because there are no
machines (that I know of) with both PCIe and PAGE_SIZE != 4096.

Thanks,
  Roland
_______________________________________________
openib-general mailing list
[email protected]
http://openib.org/mailman/listinfo/openib-general

To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general

Reply via email to