On Tue, Jul 19, 2016 at 01:34:53PM -0600, Jason Gunthorpe wrote: > On Tue, Jul 19, 2016 at 03:55:57PM +0300, Yuval Shaia wrote: > > On Mon, Jul 18, 2016 at 11:19:21AM -0600, Jason Gunthorpe wrote: > > > On Mon, Jul 18, 2016 at 12:40:48AM -0700, Yuval Shaia wrote: > > > > Decouple SG support from HW ability to do UD checksum. > > > > This coupling is for historical reasons and removed with 'commit > > > > ec5f06156423 ("net: Kill link between CSUM and SG features.")' > > > > > > Shouldn't this also check that the QP's sg limit is something > > > appropriate? At least > 1 .. > > > > What is special with 1? > > One means the hardware doesn't support SG at all. > > Is there a reason to sg_linearize down to ipoib if we know the > hardware doesn't support SG at all? Presumably it is better to let the > netstack deal with that case. ok, i see the point but as Leon Romanovsky mentioned - this is not an issue with the patch itself, right? i.e. the original code relays on the fact that HW supports UD-CSUM and did not checked QP's sg limit (we even do not have any QP at this stage yet). > > Jason > -- > To unsubscribe from this list: send the line "unsubscribe linux-rdma" in > the body of a message to majordomo@xxxxxxxxxxxxxxx > More majordomo info at http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html