Message ID | 20180529062838.20556-1-clg@kaod.org |
---|---|
Headers | show |
Series | ftgmac100 and NC-SI enhancements for the Aspeed SoC | expand |
Hi Cédric, On 05/29/2018 03:28 AM, Cédric Le Goater wrote: > Hello, > > Here is a couple of enhancements and fixes for the ftgmac100 NIC used > on the Aspeed SoC. It includes VLAN and multicast support. > > Following is an assorted set of changes for the NC-SI backend also > used on the Aspeed SoC when soldered on OpenPOWER boards. These fix a > few problems found when running OpenBMC on aspeed QEMU machines. > > Thanks, > > Cédric. > > Cédric Le Goater (6): > ftgmac100: compute maximum frame size depending on the protocol > ftgmac100: add IEEE 802.1Q VLAN support > net/ftgmac100: fix multicast hash routine > slirp/ncsi: fix "Get Version ID" payload length > slirp/ncsi: add a "Get Parameter" response > slirp/ncsi: add checksum support Doesn't it make more sens to have slirp patches applied previous to the ftgmac100 model? > > include/hw/net/ftgmac100.h | 7 +++++- > hw/net/ftgmac100.c | 59 +++++++++++++++++++++++++++++++++++----------- > slirp/ncsi.c | 54 +++++++++++++++++++++++++++++++++++++----- > 3 files changed, 99 insertions(+), 21 deletions(-) >
On 05/29/2018 02:34 PM, Philippe Mathieu-Daudé wrote: > Hi Cédric, > > On 05/29/2018 03:28 AM, Cédric Le Goater wrote: >> Hello, >> >> Here is a couple of enhancements and fixes for the ftgmac100 NIC used >> on the Aspeed SoC. It includes VLAN and multicast support. >> >> Following is an assorted set of changes for the NC-SI backend also >> used on the Aspeed SoC when soldered on OpenPOWER boards. These fix a >> few problems found when running OpenBMC on aspeed QEMU machines. >> >> Thanks, >> >> Cédric. >> >> Cédric Le Goater (6): >> ftgmac100: compute maximum frame size depending on the protocol >> ftgmac100: add IEEE 802.1Q VLAN support >> net/ftgmac100: fix multicast hash routine >> slirp/ncsi: fix "Get Version ID" payload length >> slirp/ncsi: add a "Get Parameter" response >> slirp/ncsi: add checksum support > > Doesn't it make more sens to have slirp patches applied previous to the > ftgmac100 model? why ? They are not necessarily dependent. C. > >> >> include/hw/net/ftgmac100.h | 7 +++++- >> hw/net/ftgmac100.c | 59 +++++++++++++++++++++++++++++++++++----------- >> slirp/ncsi.c | 54 +++++++++++++++++++++++++++++++++++++----- >> 3 files changed, 99 insertions(+), 21 deletions(-) >>
On 05/29/2018 01:57 PM, Cédric Le Goater wrote: > On 05/29/2018 02:34 PM, Philippe Mathieu-Daudé wrote: >> Hi Cédric, >> >> On 05/29/2018 03:28 AM, Cédric Le Goater wrote: >>> Hello, >>> >>> Here is a couple of enhancements and fixes for the ftgmac100 NIC used >>> on the Aspeed SoC. It includes VLAN and multicast support. >>> >>> Following is an assorted set of changes for the NC-SI backend also >>> used on the Aspeed SoC when soldered on OpenPOWER boards. These fix a >>> few problems found when running OpenBMC on aspeed QEMU machines. >>> >>> Thanks, >>> >>> Cédric. >>> >>> Cédric Le Goater (6): >>> ftgmac100: compute maximum frame size depending on the protocol >>> ftgmac100: add IEEE 802.1Q VLAN support >>> net/ftgmac100: fix multicast hash routine >>> slirp/ncsi: fix "Get Version ID" payload length >>> slirp/ncsi: add a "Get Parameter" response >>> slirp/ncsi: add checksum support >> >> Doesn't it make more sens to have slirp patches applied previous to the >> ftgmac100 model? > > why ? They are not necessarily dependent. OK, no worries.
On 05/30/2018 05:04 AM, Philippe Mathieu-Daudé wrote: > On 05/29/2018 01:57 PM, Cédric Le Goater wrote: >> On 05/29/2018 02:34 PM, Philippe Mathieu-Daudé wrote: >>> Hi Cédric, >>> >>> On 05/29/2018 03:28 AM, Cédric Le Goater wrote: >>>> Hello, >>>> >>>> Here is a couple of enhancements and fixes for the ftgmac100 NIC used >>>> on the Aspeed SoC. It includes VLAN and multicast support. >>>> >>>> Following is an assorted set of changes for the NC-SI backend also >>>> used on the Aspeed SoC when soldered on OpenPOWER boards. These fix a >>>> few problems found when running OpenBMC on aspeed QEMU machines. >>>> >>>> Thanks, >>>> >>>> Cédric. >>>> >>>> Cédric Le Goater (6): >>>> ftgmac100: compute maximum frame size depending on the protocol >>>> ftgmac100: add IEEE 802.1Q VLAN support >>>> net/ftgmac100: fix multicast hash routine >>>> slirp/ncsi: fix "Get Version ID" payload length >>>> slirp/ncsi: add a "Get Parameter" response >>>> slirp/ncsi: add checksum support >>> >>> Doesn't it make more sens to have slirp patches applied previous to the >>> ftgmac100 model? >> >> why ? They are not necessarily dependent. > > OK, no worries. > bah, I will just resend two distinct series one for 'slirp' and one for 'net'. C.
On 29 May 2018 at 15:58, Cédric Le Goater <clg@kaod.org> wrote: > Hello, > > Here is a couple of enhancements and fixes for the ftgmac100 NIC used > on the Aspeed SoC. It includes VLAN and multicast support. > > Following is an assorted set of changes for the NC-SI backend also > used on the Aspeed SoC when soldered on OpenPOWER boards. These fix a > few problems found when running OpenBMC on aspeed QEMU machines. > > Thanks, > > Cédric. > > Cédric Le Goater (6): > ftgmac100: compute maximum frame size depending on the protocol > ftgmac100: add IEEE 802.1Q VLAN support > net/ftgmac100: fix multicast hash routine > slirp/ncsi: fix "Get Version ID" payload length > slirp/ncsi: add a "Get Parameter" response > slirp/ncsi: add checksum support I tested these with -M palmetto-bmc and linux-next as the guest, and networking appeared to work fine. For the series: Tested-by: Joel Stanley <joel@jms.id.au> One question I had. I get these messages on boot: [ 1.184472] ftgmac100 1e660000.ethernet: Read MAC address 52:54:00:12:34:56 from chip [ 1.184751] ftgmac100 1e660000.ethernet: Using NCSI interface [ 1.197107] ftgmac100 1e660000.ethernet eth0: irq 19, mapped at b30b8905 [ 1.904237] ftgmac100 1e660000.ethernet eth0: NCSI: Handler for packet type 0x82 returned -19 [ 1.905682] ftgmac100 1e660000.ethernet eth0: NCSI: Handler for packet type 0x82 returned -19 [ 1.905943] ftgmac100 1e660000.ethernet eth0: NCSI: Handler for packet type 0x82 returned -19 [ 1.906182] ftgmac100 1e660000.ethernet eth0: NCSI: Handler for packet type 0x82 returned -19 [ 1.906415] ftgmac100 1e660000.ethernet eth0: NCSI: Handler for packet type 0x82 returned -19 [ 1.906643] ftgmac100 1e660000.ethernet eth0: NCSI: Handler for packet type 0x82 returned -19 [ 1.906873] ftgmac100 1e660000.ethernet eth0: NCSI: Handler for packet type 0x82 returned -19 [ 1.907103] ftgmac100 1e660000.ethernet eth0: NCSI: Handler for packet type 0x82 returned -19 [ 2.024094] ftgmac100 1e660000.ethernet eth0: NCSI: configuring channel 0 [ 2.025428] ftgmac100 1e660000.ethernet eth0: NCSI: channel 0 config done [ 2.025544] ftgmac100 1e660000.ethernet eth0: NCSI: No more channels to process [ 2.025608] ftgmac100 1e660000.ethernet eth0: NCSI interface up Comparing to hardware, I see: [ 4.071159] ftgmac100 1e660000.ethernet: Read MAC address 98:be:94:83:00:ad from chip [ 4.079024] ftgmac100 1e660000.ethernet: Using NCSI interface [ 4.085976] ftgmac100 1e660000.ethernet eth0: irq 19, mapped at a0963000 [ 40.670099] ftgmac100 1e660000.ethernet eth0: NCSI: Handler for packet type 0x82 returned -19 [ 43.852036] ftgmac100 1e660000.ethernet eth0: NCSI: configuring channel 1 [ 43.877060] ftgmac100 1e660000.ethernet eth0: NCSI: channel 1 config done [ 43.877093] ftgmac100 1e660000.ethernet eth0: NCSI: No more channels to process [ 43.877112] ftgmac100 1e660000.ethernet eth0: NCSI interface up Is it expected that we get multiple ENODEVs for packet type 0x82 in qemu? Cheers, Joel
On 05/30/2018 07:35 AM, Joel Stanley wrote: > On 29 May 2018 at 15:58, Cédric Le Goater <clg@kaod.org> wrote: >> Hello, >> >> Here is a couple of enhancements and fixes for the ftgmac100 NIC used >> on the Aspeed SoC. It includes VLAN and multicast support. >> >> Following is an assorted set of changes for the NC-SI backend also >> used on the Aspeed SoC when soldered on OpenPOWER boards. These fix a >> few problems found when running OpenBMC on aspeed QEMU machines. >> >> Thanks, >> >> Cédric. >> >> Cédric Le Goater (6): >> ftgmac100: compute maximum frame size depending on the protocol >> ftgmac100: add IEEE 802.1Q VLAN support >> net/ftgmac100: fix multicast hash routine >> slirp/ncsi: fix "Get Version ID" payload length >> slirp/ncsi: add a "Get Parameter" response >> slirp/ncsi: add checksum support > > I tested these with -M palmetto-bmc and linux-next as the guest, and > networking appeared to work fine. For the series: > > Tested-by: Joel Stanley <joel@jms.id.au> > > One question I had. I get these messages on boot: > > [ 1.184472] ftgmac100 1e660000.ethernet: Read MAC address > 52:54:00:12:34:56 from chip > [ 1.184751] ftgmac100 1e660000.ethernet: Using NCSI interface > [ 1.197107] ftgmac100 1e660000.ethernet eth0: irq 19, mapped at b30b8905 > [ 1.904237] ftgmac100 1e660000.ethernet eth0: NCSI: Handler for > packet type 0x82 returned -19 > [ 1.905682] ftgmac100 1e660000.ethernet eth0: NCSI: Handler for > packet type 0x82 returned -19 > [ 1.905943] ftgmac100 1e660000.ethernet eth0: NCSI: Handler for > packet type 0x82 returned -19 > [ 1.906182] ftgmac100 1e660000.ethernet eth0: NCSI: Handler for > packet type 0x82 returned -19 > [ 1.906415] ftgmac100 1e660000.ethernet eth0: NCSI: Handler for > packet type 0x82 returned -19 > [ 1.906643] ftgmac100 1e660000.ethernet eth0: NCSI: Handler for > packet type 0x82 returned -19 > [ 1.906873] ftgmac100 1e660000.ethernet eth0: NCSI: Handler for > packet type 0x82 returned -19 > [ 1.907103] ftgmac100 1e660000.ethernet eth0: NCSI: Handler for > packet type 0x82 returned -19 > [ 2.024094] ftgmac100 1e660000.ethernet eth0: NCSI: configuring channel 0 > [ 2.025428] ftgmac100 1e660000.ethernet eth0: NCSI: channel 0 config done > [ 2.025544] ftgmac100 1e660000.ethernet eth0: NCSI: No more > channels to process > [ 2.025608] ftgmac100 1e660000.ethernet eth0: NCSI interface up > > > Comparing to hardware, I see: > > [ 4.071159] ftgmac100 1e660000.ethernet: Read MAC address > 98:be:94:83:00:ad from chip > [ 4.079024] ftgmac100 1e660000.ethernet: Using NCSI interface > [ 4.085976] ftgmac100 1e660000.ethernet eth0: irq 19, mapped at a0963000 > [ 40.670099] ftgmac100 1e660000.ethernet eth0: NCSI: Handler for > packet type 0x82 returned -19 > [ 43.852036] ftgmac100 1e660000.ethernet eth0: NCSI: configuring channel 1 > [ 43.877060] ftgmac100 1e660000.ethernet eth0: NCSI: channel 1 config done > [ 43.877093] ftgmac100 1e660000.ethernet eth0: NCSI: No more > channels to process > [ 43.877112] ftgmac100 1e660000.ethernet eth0: NCSI interface up > > Is it expected that we get multiple ENODEVs for packet type 0x82 in qemu? When the NSCI probing is done in Linux, 8 deselect packages requests are sent and QEMU answers them all. I wonder what the real HW does. Something to check. Cheers, C.