Add support to FCoE offload through net_device (v2)
Subsystems (3)
cover Add support to FCoE offload through net_device (v2)
This is set of revised patches for adding offload support of Fiber Channel over Ethernet (FCoE) offload support through net_device for the Open-FCoE project maintained at http://www.open-fcoe.org.
The original submission can be found at:
http://marc.info/?l=linux-scsi&m=123507707821705&w=2
The included patches would add FCoE offload support through the net_device struct as shown in the series of patches followed here. This is achieved by adding nedev feature flags for FCoE FC CRC offload and large send offload through gso. The large receive offload by direct data placement (ddp) is achieved by the two function pointers to net_device_ops, to setup and release ddp context, ndo_fcoe_ddp_setup and ndo_fcoe_ddp_done. One data member is added to net_device strut, namely fcoe_ddp_xid, which inidates the maximum exchange id allowed by LLD to perform ddp on the read I/O request.
As the following patches touch both netdev and libfc/fcoe, and they are applicable to the current tip at www.open-fcoe.org w/ patches available at:
http://marc.info/?l=linux-scsi&m=123576089125544&w=2 If they look alright, I will rework them later to be directly applicable to net-next, where the aforementioned fcoe/libfc fix patches are expected to be in kernel already through scsi-fixes.
Thanks.
yi
---
Chris Leech (2):
net: define feature flags for FCoE offloads
net: reclaim 8 upper bits of the netdev->features from GSO
Yi Zou (9):
fcoe: fcoe fc crc offload indication by skb->ip_summed
fcoe: add support to FCoE offload support in fcoe_sw through net_device
libfc: add support of large receive offload by ddp in fc_fcp
fcoe: add support to large send by gso through net_device for fcoe_sw
libfc: use lso_max for sequence offload
fcoe,libfc: check offload features from LLD through netdev
net: add FCoE offload support through net_device
net: add NETIF_F_FCOE_CRC to can_checksum_protocol net,fcoe: add ETH_P_FCOE for Fibre Channel over Ethernet (FCoE)
drivers/net/xen-netfront.c | 2 + drivers/scsi/fcoe/fcoe_sw.c | 67 ++++++++++++++++++++++++++++++++++++++++- drivers/scsi/fcoe/libfcoe.c | 16 ++++++++-- drivers/scsi/libfc/fc_exch.c | 4 ++ drivers/scsi/libfc/fc_fcp.c | 69 +++++++++++++++++++++++++++++++++++++++++- include/linux/if_ether.h | 1 + include/linux/netdevice.h | 18 ++++++++++- include/linux/skbuff.h | 2 + include/scsi/fc/fc_fcoe.h | 7 ---- include/scsi/fc_frame.h | 19 +----------- include/scsi/libfc.h | 31 +++++++++++++++++++ include/scsi/libfcoe.h | 18 ----------- net/core/dev.c | 4 ++ 13 files changed, 206 insertions(+), 52 deletions(-)
Signature · 1 line
Signature: Yi Zou <[email protected]>
David Miller
maintainerFrom: Yi Zou <[email protected]>
Date: Fri, 27 Feb 2009 14:06:32 -0800
Quoted 16 lines
This is set of revised patches for adding offload support of Fiber Channel over Ethernet (FCoE) offload support through
net_devicefor the Open-FCoE project maintained at http://www.open-fcoe.org.The original submission can be found at:
http://marc.info/?l=linux-scsi&m=123507707821705&w=2The included patches would add FCoE offload support through the
net_devicestruct as shown in the series of patches followed here. This is achieved by adding nedev feature flags for FCoE FC CRC offload and large send offload through gso. The large receive offload by direct data placement (ddp) is achieved by the two function pointers tonet_device_ops, to setup and release ddp context,ndo_fcoe_ddp_setupandndo_fcoe_ddp_done. One data member is added tonet_devicestrut, namelyfcoe_ddp_xid, which inidates the maximum exchange id allowed by LLD to perform ddp on the read I/O request.
Quoted 6 lines
As the following patches touch both netdev and libfc/fcoe, and they are applicable to the current tip at www.open-fcoe.org w/ patches available at:
http://marc.info/?l=linux-scsi&m=123576089125544&w=2 If they look alright, I will rework them later to be directly applicable to net-next, where the aforementioned fcoe/libfc fix patches are expected to be in kernel already through scsi-fixes.
I am fine with these changes, they look good.
James Bottomley
maintainerQuoted 19 lines · David Miller
From: Yi Zou <[email protected]>
Date: Fri, 27 Feb 2009 14:06:32 -0800Quoted 16 lines
This is set of revised patches for adding offload support of Fiber Channel over Ethernet (FCoE) offload support through
net_devicefor the Open-FCoE project maintained at http://www.open-fcoe.org.The original submission can be found at:
http://marc.info/?l=linux-scsi&m=123507707821705&w=2The included patches would add FCoE offload support through the
net_devicestruct as shown in the series of patches followed here. This is achieved by adding nedev feature flags for FCoE FC CRC offload and large send offload through gso. The large receive offload by direct data placement (ddp) is achieved by the two function pointers tonet_device_ops, to setup and release ddp context,ndo_fcoe_ddp_setupandndo_fcoe_ddp_done. One data member is added tonet_devicestrut, namelyfcoe_ddp_xid, which inidates the maximum exchange id allowed by LLD to perform ddp on the read I/O request.
Quoted 8 lines
Quoted 6 lines
As the following patches touch both netdev and libfc/fcoe, and they are applicable to the current tip at www.open-fcoe.org w/ patches available at:
http://marc.info/?l=linux-scsi&m=123576089125544&w=2 If they look alright, I will rework them later to be directly applicable to net-next, where the aforementioned fcoe/libfc fix patches are expected to be in kernel already through scsi-fixes.I am fine with these changes, they look good.
Did you want to take the four pure net pieces through your tree, and I'll run a postmerge tree based on your net tree, or just have me take them through the SCSI tree?
James
David Miller
maintainerFrom: James Bottomley <[email protected]>
Date: Thu, 05 Mar 2009 21:59:20 +0000
Quoted 6 lines · David Miller
Quoted 1 line
I am fine with these changes, they look good.
Did you want to take the four pure net pieces through your tree, and I'll run a postmerge tree based on your net tree, or just have me take them through the SCSI tree?
You can take them all via your SCSI tree. I think it works best that way.