From: Alexander Bluhm Subject: Re: bnxt(4) tso To: Hrvoje Popovski Cc: Jonathan Matthew , Jan Klemkow , tech@openbsd.org Date: Fri, 12 Jan 2024 10:35:21 +0100 On Fri, Jan 12, 2024 at 10:27:30AM +0100, Hrvoje Popovski wrote: > On 12.1.2024. 9:32, Alexander Bluhm wrote: > > On Fri, Jan 12, 2024 at 12:43:44AM +0100, Alexander Bluhm wrote: > >> I have thrown your diff on my test setup. > > Sending TCP works as expected with line speed. > > > > Jumbo frames and vlan looks completely broken. > > Also socket splicing and userland relay copy have problems. > > > > On my text matrix I see a lot of failed tests. > > http://bluhm.genua.de/netlink/results/2024-01-11T10%3A37%3A05Z/netlink.html > > > > It is possible that only one test transfers the hardware in a bad > > state and all other tests after that fail. > > > > bluhm > > > Hi, > > mbuhl@ ping me few day ago to test vlan/tso and with or without vlan I'm > getting same good results. > For me vlan is working just fine ... My bnxt interfaces are 10GBase-T mbuhl@ is testing on the same setup as I do. And he has seen simmilar failures with bnxt TSO diff a few days ago. So it might be a measurement specific problem. But it works without the diff and also just sending with TSO works fine. bnxt0 at pci4 dev 0 function 0 "Broadcom BCM57412" rev 0x01: fw ver 214.4.91, msix, 8 queues, address 14:23:f2:a0:71:e0 bnxt1 at pci4 dev 0 function 1 "Broadcom BCM57412" rev 0x01: fw ver 214.4.91, msix, 8 queues, address 14:23:f2:a0:71:e1 root@ot42:.../~# ifconfig bnxt0 hwfeatures bnxt0: flags=8c43 mtu 9000 hwfeatures=31b7 hardmtu 9500 lladdr 14:23:f2:a0:71:e0 index 1 priority 0 llprio 3 media: Ethernet autoselect (10GSFP+Cu full-duplex,rxpause,txpause) status: active bluhm