[PATCH 2.6] POWER5 Virtual Ethernet Checkum offload

From: Santiago Leon (santil_at_us.ibm.com)
Date: 08/27/04

  • Next message: Martin J. Bligh: "Re: Termination of the Philips Webcam Driver (pwc)" 
    Date:	Fri, 27 Aug 2004 14:04:23 -0500
To: Andrew Morton <akpm@osdl.org>

    Andrew,

    The following patch adds checksum offloading for the POWER5 Virtual
    Ethernet driver. In the case where the OS in the partitions
    communicating support this feature (i.e. both partition have this patch
    applied), no checksum will be created because the link is reliable.
    However, in the case where one of the OS in a partition does support
    this feature and the other doesn't (i.e. linux with patch applied
    talking to AIX), then the hypervisor will generate the checksum.

    Some levels of firmware will not support this feature but the code will
    figure it out and not enable it.

    Applies against the latest mainline and -mm trees. Please apply.

    Signed-off-by: Santiago Leon <santil@us.ibm.com> 

    -- 
Santiago A. Leon
Power Linux Development
IBM Linux Technology Center

    ===== drivers/net/ibmveth.c 1.17 vs edited =====
    --- 1.17/drivers/net/ibmveth.c Mon Aug 23 03:14:44 2004
    +++ edited/drivers/net/ibmveth.c Fri Aug 27 14:23:56 2004
    @@ -49,6 +49,7 @@
     #include <linux/mm.h>
     #include <linux/ethtool.h>
     #include <linux/proc_fs.h>
    +#include <linux/tcp.h>
     #include <asm/semaphore.h>
     #include <asm/hvcall.h>
     #include <asm/atomic.h>
    @@ -104,7 +105,7 @@
     
     static const char ibmveth_driver_name[] = "ibmveth";
     static const char ibmveth_driver_string[] = "IBM i/pSeries Virtual Ethernet Driver";
    -#define ibmveth_driver_version "1.02"
    +#define ibmveth_driver_version "1.03"
     
     MODULE_AUTHOR("Santiago Leon <santil@us.ibm.com>");
     MODULE_DESCRIPTION("IBM i/pSeries Virtual Ethernet Driver");
    @@ -132,6 +133,11 @@
             return (adapter->rx_queue.queue_addr[adapter->rx_queue.index].length);
     }
     
    +static inline int ibmveth_rxq_frame_checksum_good(struct ibmveth_adapter *adapter)
    +{
    + return (adapter->rx_queue.queue_addr[adapter->rx_queue.index].checksum_good);
    +}
    +
     /* setup the initial settings for a buffer pool */
     static void ibmveth_init_buffer_pool(struct ibmveth_buff_pool *pool, u32 pool_index, u32 pool_size, u32 buff_size)
     {
    @@ -403,6 +409,11 @@
     
     static void ibmveth_cleanup(struct ibmveth_adapter *adapter)
     {
    + unsigned long chkoff_result, dummy;
    +
    + h_illan_attributes(adapter->vdev->unit_address, CHECKSUM_OFFLOAD, 0,
    + chkoff_result, dummy);
    +
             if(adapter->buffer_list_addr != NULL) {
                     if(!dma_mapping_error(adapter->buffer_list_dma)) {
                             vio_unmap_single(adapter->vdev, adapter->buffer_list_dma, 4096, DMA_BIDIRECTIONAL);
    @@ -645,6 +656,12 @@
                                             desc[0].fields.length, DMA_TO_DEVICE);
             desc[0].fields.valid = 1;
     
    + if (adapter->can_checksum_offload && skb->ip_summed == CHECKSUM_HW) {
    + desc[0].fields.no_checksum = 1;
    + desc[0].fields.checksum_good = 1;
    + skb->h.th->check = 0;
    + }
    +
             if(dma_mapping_error(desc[0].fields.address)) {
                     ibmveth_error_printk("tx: unable to map initial fragment\n");
                     adapter->tx_map_failed++;
    @@ -743,6 +760,7 @@
                             } else {
                                     int length = ibmveth_rxq_frame_length(adapter);
                                     int offset = ibmveth_rxq_frame_offset(adapter);
    + int checksum_good = ibmveth_rxq_frame_checksum_good(adapter);
                                     skb = ibmveth_rxq_get_buffer(adapter);
     
                                     ibmveth_rxq_harvest_buffer(adapter);
    @@ -751,6 +769,11 @@
                                     skb_put(skb, length);
                                     skb->dev = netdev;
                                     skb->protocol = eth_type_trans(skb, netdev);
    +
    + if (checksum_good)
    + skb->ip_summed = CHECKSUM_UNNECESSARY;
    + else
    + skb->ip_summed = CHECKSUM_NONE;
     
                                     netif_receive_skb(skb); /* send it up */
     
    @@ -878,7 +901,7 @@
     
             unsigned char *mac_addr_p;
             unsigned int *mcastFilterSize_p;
    -
    + unsigned long chkoff_result, dummy;
     
             ibmveth_debug_printk_no_adapter("entering ibmveth_probe for UA 0x%x\n",
                                             dev->unit_address);
    @@ -926,6 +949,18 @@
     
             adapter->liobn = dev->iommu_table->it_index;
             
    + rc = h_illan_attributes(dev->unit_address, 0, CHECKSUM_OFFLOAD,
    + chkoff_result, dummy);
    +
    + if (rc == H_Success)
    + adapter->can_checksum_offload = 1;
    + else
    + adapter->can_checksum_offload = 0;
    +
    + ibmveth_printk("Firmware %s checksum offload, result mask=%lu\n",
    + adapter->can_checksum_offload? "supports" :
    + "does not support", chkoff_result);
    +
             netdev->irq = dev->irq;
             netdev->open = ibmveth_open;
             netdev->poll = ibmveth_poll;
    @@ -937,6 +972,8 @@
             netdev->do_ioctl = ibmveth_ioctl;
             netdev->ethtool_ops = &netdev_ethtool_ops;
             netdev->change_mtu = ibmveth_change_mtu;
    + if (adapter->can_checksum_offload)
    + netdev->features = NETIF_F_HW_CSUM;
             SET_NETDEV_DEV(netdev, &dev->dev);
     
             memcpy(&netdev->dev_addr, &adapter->mac_addr, netdev->addr_len);
    ===== drivers/net/ibmveth.h 1.2 vs edited =====
    --- 1.2/drivers/net/ibmveth.h Mon Aug 23 03:14:44 2004
    +++ edited/drivers/net/ibmveth.h Fri Aug 27 14:21:15 2004
    @@ -49,6 +49,7 @@
     #define H_SEND_LOGICAL_LAN 0x120
     #define H_MULTICAST_CTRL 0x130
     #define H_CHANGE_LOGICAL_LAN_MAC 0x14C
    +#define H_ILLAN_ATTRIBUTES 0x244
     
     /* hcall macros */
     #define h_register_logical_lan(ua, buflst, rxq, fltlst, mac) \
    @@ -69,6 +70,9 @@
     #define h_change_logical_lan_mac(ua, mac) \
       plpar_hcall_norets(H_CHANGE_LOGICAL_LAN_MAC, ua, mac)
     
    +#define h_illan_attributes(ua, reset_mask, set_mask, result, dummy) \
    + plpar_hcall(H_ILLAN_ATTRIBUTES, ua, reset_mask, set_mask, 0, &result, &dummy, &dummy)
    +
     #define IbmVethNumBufferPools 3
     #define IbmVethPool0DftSize (1024 * 2)
     #define IbmVethPool1DftSize (1024 * 4)
    @@ -79,6 +83,8 @@
     
     #define IBM_VETH_INVALID_MAP ((u16)0xffff)
     
    +#define CHECKSUM_OFFLOAD 0x6
    +
     struct ibmveth_buff_pool {
         u32 size;
         u32 index;
    @@ -115,6 +121,7 @@
         struct ibmveth_buff_pool rx_buff_pool[IbmVethNumBufferPools];
         struct ibmveth_rx_q rx_queue;
         atomic_t not_replenishing;
    + int can_checksum_offload;
     
         /* helper tasks */
         struct work_struct replenish_task;
    @@ -136,7 +143,9 @@
     struct ibmveth_buf_desc_fields {
         u32 valid : 1;
         u32 toggle : 1;
    - u32 reserved : 6;
    + u32 reserved : 4;
    + u32 no_checksum : 1;
    + u32 checksum_good :1;
         u32 length : 24;
         u32 address;
     };
    @@ -149,7 +158,10 @@
     struct ibmveth_rx_q_entry {
         u16 toggle : 1;
         u16 valid : 1;
    - u16 reserved : 14;
    + u16 reserved : 4;
    + u16 no_checksum : 1;
    + u16 checksum_good : 1;
    + u16 reserved2 : 8;
         u16 offset;
         u32 length;
         u64 correlator;

    -
    To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
    the body of a message to majordomo@vger.kernel.org
    More majordomo info at http://vger.kernel.org/majordomo-info.html
    Please read the FAQ at http://www.tux.org/lkml/

  • Next message: Martin J. Bligh: "Re: Termination of the Philips Webcam Driver (pwc)"

    Relevant Pages

    • Re: B-Tree Index Usage
      ... I asked for feedback on the need for a re-index feature that would make ... support an enhancement request that allows this sort of thing to be ... Support translates directly to something that should be ... and no one else driving this popular model truck ...
      (comp.databases.pick)
    • Re: Proposed enhancements to MD
      ... The attached patch is for the partition ... >, Adaptec, and many others must continue to support new ... >> the partitions on an array and properly boot them as it would boot a ... root fs exists on array, but md didn't see them show up, ...
      (Linux-Kernel)
    • Re: she might rip close, unless Rifaat upsets demonstrations in Maifys flow
      ... My convincing pool won't make before I feature it. ... When doesn't Vance exceed downstairs? ... Almost no respectable mls with regard to the crucial guerrilla were ... it indicates a discharge too red in support of her ...
      (sci.crypt)
    • Re: For Each
      ... the counter is a counter for the iteration - not the group ... I think the feature would be useful. ... Framework and having a counter in a for each loop, ... I meant the necessary changes to the framework to support: ...
      (microsoft.public.dotnet.languages.vb)
    • landing alongside chosen ulcer
      ... launch commonly if Mhammed's feature isn't tricky. ... in support of the gigantic harbour. ... searchs the sure wish under the television? ...
      (rec.music.christian)