search for: skb_header_clon

...bug("%s: xmit %p %pM\n", vi->dev->name, skb, dest); + if (vi->mergeable_rx_bufs) + hdr_len = sizeof hdr->mhdr; + else + hdr_len = sizeof hdr->hdr; + + can_push = vi->any_header_sg && + !((unsigned long)skb->data & (__alignof__(*hdr) - 1)) && + !skb_header_cloned(skb) && skb_headroom(skb) >= hdr_len; + /* Even if we can, don't push here yet as this would skew + * csum_start offset below. */ + if (can_push) + hdr = (struct skb_vnet_hdr *)(skb->data - hdr_len); + else + hdr = skb_vnet_hdr(skb); if (skb->ip_summed == CHECKSUM_PART...

...bug("%s: xmit %p %pM\n", vi->dev->name, skb, dest); + if (vi->mergeable_rx_bufs) + hdr_len = sizeof hdr->mhdr; + else + hdr_len = sizeof hdr->hdr; + + can_push = vi->any_header_sg && + !((unsigned long)skb->data & (__alignof__(*hdr) - 1)) && + !skb_header_cloned(skb) && skb_headroom(skb) >= hdr_len; + /* Even if we can, don't push here yet as this would skew + * csum_start offset below. */ + if (can_push) + hdr = (struct skb_vnet_hdr *)(skb->data - hdr_len); + else + hdr = skb_vnet_hdr(skb); if (skb->ip_summed == CHECKSUM_PART...

...y keeping track of an offset or similar > internally to your driver, rather than changing the SKB data pointers. What virtio does is this: can_push = vi->any_header_sg && !((unsigned long)skb->data & (__alignof__(*hdr) - 1)) && !skb_header_cloned(skb) && skb_headroom(skb) >= hdr_len; /* Even if we can, don't push here yet as this would skew * csum_start offset below. */ if (can_push) hdr = (struct virtio_net_hdr_mrg_rxbuf *)(skb->data - hdr_len); else hdr =...

...y keeping track of an offset or similar > internally to your driver, rather than changing the SKB data pointers. What virtio does is this: can_push = vi->any_header_sg && !((unsigned long)skb->data & (__alignof__(*hdr) - 1)) && !skb_header_cloned(skb) && skb_headroom(skb) >= hdr_len; /* Even if we can, don't push here yet as this would skew * csum_start offset below. */ if (can_push) hdr = (struct virtio_net_hdr_mrg_rxbuf *)(skb->data - hdr_len); else hdr =...

...t;dev->name, skb, dest); > + if (vi->mergeable_rx_bufs) > + hdr_len = sizeof hdr->mhdr; > + else > + hdr_len = sizeof hdr->hdr; > + > + can_push = vi->any_header_sg && > + !((unsigned long)skb->data & (__alignof__(*hdr) - 1)) && > + !skb_header_cloned(skb) && skb_headroom(skb) >= hdr_len; Idle thought: how often does this fail? Would it suck if we copied headers which didn't let us prepend data? Or could we bump dev->hard_header_len appropriately? Thanks, Rusty.

Looks like the GSO is involved? I got this while running Dom0 only (no guests), with a BOINC/Rosetta@home application running on all 4 cores. changeset: 10649:8e55c5c11475 Build: x86_32p (pae). ------------[ cut here ]------------ kernel BUG at net/core/dev.c:1133! invalid opcode: 0000 [#1] SMP CPU: 0 EIP: 0061:[<c04dceb0>] Not tainted VLI EFLAGS: 00210297 (2.6.16.13-xen

...bug("%s: xmit %p %pM\n", vi->dev->name, skb, dest); + if (vi->mergeable_rx_bufs) + hdr_len = sizeof hdr->mhdr; + else + hdr_len = sizeof hdr->hdr; + + can_push = vi->any_header_sg && + !((unsigned long)skb->data & (__alignof__(*hdr) - 1)) && + !skb_header_cloned(skb) && skb_headroom(skb) >= hdr_len; + /* Even if we can, don't push here yet as this would skew + * csum_start offset below. */ + if (can_push) + hdr = (struct skb_vnet_hdr *)(skb->data - hdr_len); + else + hdr = skb_vnet_hdr(skb); if (skb->ip_summed == CHECKSUM_PART...

From: Rusty Russell <rusty at rustcorp.com.au> Date: Tue, 09 Jul 2013 17:38:51 +0930 > If you convince DaveM, I won't object :) Simplifications are great, but not when the merge window opens up. Sorry, this isn't appropriate now.

From: Rusty Russell <rusty at rustcorp.com.au> Date: Tue, 09 Jul 2013 17:38:51 +0930 > If you convince DaveM, I won't object :) Simplifications are great, but not when the merge window opens up. Sorry, this isn't appropriate now.

...t;dev->name, skb, dest); > + if (vi->mergeable_rx_bufs) > + hdr_len = sizeof hdr->mhdr; > + else > + hdr_len = sizeof hdr->hdr; > + > + can_push = vi->any_header_sg && > + !((unsigned long)skb->data & (__alignof__(*hdr) - 1)) && > + !skb_header_cloned(skb) && skb_headroom(skb) >= hdr_len; > + /* Even if we can, don't push here yet as this would skew > + * csum_start offset below. */ > + if (can_push) > + hdr = (struct skb_vnet_hdr *)(skb->data - hdr_len); > + else > + hdr = skb_vnet_hdr(skb); > >...

...bug("%s: xmit %p %pM\n", vi->dev->name, skb, dest); + if (vi->mergeable_rx_bufs) + hdr_len = sizeof hdr->mhdr; + else + hdr_len = sizeof hdr->hdr; + + can_push = vi->any_header_sg && + !((unsigned long)skb->data & (__alignof__(*hdr) - 1)) && + !skb_header_cloned(skb) && skb_headroom(skb) >= hdr_len; + /* Even if we can, don't push here yet as this would skew + * csum_start offset below. */ + if (can_push) + hdr = (struct skb_vnet_hdr *)(skb->data - hdr_len); + else + hdr = skb_vnet_hdr(skb); if (skb->ip_summed == CHECKSUM_PART...

...bug("%s: xmit %p %pM\n", vi->dev->name, skb, dest); + if (vi->mergeable_rx_bufs) + hdr_len = sizeof hdr->mhdr; + else + hdr_len = sizeof hdr->hdr; + + can_push = vi->any_header_sg && + !((unsigned long)skb->data & (__alignof__(*hdr) - 1)) && + !skb_header_cloned(skb) && skb_headroom(skb) >= hdr_len; + /* Even if we can, don't push here yet as this would skew + * csum_start offset below. */ + if (can_push) + hdr = (struct skb_vnet_hdr *)(skb->data - hdr_len); + else + hdr = skb_vnet_hdr(skb); if (skb->ip_summed == CHECKSUM_PART...

Thinking about Sasha's patches, we can reduce ring usage for virtio net small packets dramatically if we put virtio net header inline with the data. This can be done for free in case guest net stack allocated extra head room for the packet, and I don't see why would this have any downsides. Even though with my recent patches qemu no longer requires header to be the first s/g element, we

Thinking about Sasha's patches, we can reduce ring usage for virtio net small packets dramatically if we put virtio net header inline with the data. This can be done for free in case guest net stack allocated extra head room for the packet, and I don't see why would this have any downsides. Even though with my recent patches qemu no longer requires header to be the first s/g element, we

...t3_GenericDesc *gdesc; + + /* conservatively estimate # of descriptors to use */ + count = VMXNET3_TXD_NEEDED(skb_headlen(skb)) + + skb_shinfo(skb)->nr_frags + 1; + + ctx.ipv4 = (skb->protocol == __constant_ntohs(ETH_P_IP)); + + ctx.mss = skb_shinfo(skb)->gso_size; + if (ctx.mss) { + if (skb_header_cloned(skb)) { + if (unlikely(pskb_expand_head(skb, 0, 0, + GFP_ATOMIC) != 0)) { + tq->stats.drop_tso++; + goto drop_pkt; + } + tq->stats.copy_skb_header++; + } + vmxnet3_prepare_tso(skb, &ctx); + } else { + if (unlikely(count > VMXNET3_MAX_TXD_PER_PKT)) { + + /*...

...t3_GenericDesc *gdesc; + + /* conservatively estimate # of descriptors to use */ + count = VMXNET3_TXD_NEEDED(skb_headlen(skb)) + + skb_shinfo(skb)->nr_frags + 1; + + ctx.ipv4 = (skb->protocol == __constant_ntohs(ETH_P_IP)); + + ctx.mss = skb_shinfo(skb)->gso_size; + if (ctx.mss) { + if (skb_header_cloned(skb)) { + if (unlikely(pskb_expand_head(skb, 0, 0, + GFP_ATOMIC) != 0)) { + tq->stats.drop_tso++; + goto drop_pkt; + } + tq->stats.copy_skb_header++; + } + vmxnet3_prepare_tso(skb, &ctx); + } else { + if (unlikely(count > VMXNET3_MAX_TXD_PER_PKT)) { + + /*...

...t3_GenericDesc *gdesc; + + /* conservatively estimate # of descriptors to use */ + count = VMXNET3_TXD_NEEDED(skb_headlen(skb)) + + skb_shinfo(skb)->nr_frags + 1; + + ctx.ipv4 = (skb->protocol == __constant_ntohs(ETH_P_IP)); + + ctx.mss = skb_shinfo(skb)->gso_size; + if (ctx.mss) { + if (skb_header_cloned(skb)) { + if (unlikely(pskb_expand_head(skb, 0, 0, + GFP_ATOMIC) != 0)) { + tq->stats.drop_tso++; + goto drop_pkt; + } + tq->stats.copy_skb_header++; + } + vmxnet3_prepare_tso(skb, &ctx); + } else { + if (unlikely(count > VMXNET3_MAX_TXD_PER_PKT)) { + + /*...

...t3_GenericDesc *gdesc; + + /* conservatively estimate # of descriptors to use */ + count = VMXNET3_TXD_NEEDED(skb_headlen(skb)) + + skb_shinfo(skb)->nr_frags + 1; + + ctx.ipv4 = (skb->protocol == __constant_ntohs(ETH_P_IP)); + + ctx.mss = skb_shinfo(skb)->gso_size; + if (ctx.mss) { + if (skb_header_cloned(skb)) { + if (unlikely(pskb_expand_head(skb, 0, 0, + GFP_ATOMIC) != 0)) { + tq->stats.drop_tso++; + goto drop_pkt; + } + tq->stats.copy_skb_header++; + } + vmxnet3_prepare_tso(skb, &ctx); + } else { + if (unlikely(count > VMXNET3_MAX_TXD_PER_PKT)) { + + /*...

...t3_GenericDesc *gdesc; + + /* conservatively estimate # of descriptors to use */ + count = VMXNET3_TXD_NEEDED(skb_headlen(skb)) + + skb_shinfo(skb)->nr_frags + 1; + + ctx.ipv4 = (skb->protocol == __constant_ntohs(ETH_P_IP)); + + ctx.mss = skb_shinfo(skb)->gso_size; + if (ctx.mss) { + if (skb_header_cloned(skb)) { + if (unlikely(pskb_expand_head(skb, 0, 0, + GFP_ATOMIC) != 0)) { + tq->stats.drop_tso++; + goto drop_pkt; + } + tq->stats.copy_skb_header++; + } + vmxnet3_prepare_tso(skb, &ctx); + } else { + if (unlikely(count > VMXNET3_MAX_TXD_PER_PKT)) { + + /*...

...t3_GenericDesc *gdesc; + + /* conservatively estimate # of descriptors to use */ + count = VMXNET3_TXD_NEEDED(skb_headlen(skb)) + + skb_shinfo(skb)->nr_frags + 1; + + ctx.ipv4 = (skb->protocol == __constant_ntohs(ETH_P_IP)); + + ctx.mss = skb_shinfo(skb)->gso_size; + if (ctx.mss) { + if (skb_header_cloned(skb)) { + if (unlikely(pskb_expand_head(skb, 0, 0, + GFP_ATOMIC) != 0)) { + tq->stats.drop_tso++; + goto drop_pkt; + } + tq->stats.copy_skb_header++; + } + vmxnet3_prepare_tso(skb, &ctx); + } else { + if (unlikely(count > VMXNET3_MAX_TXD_PER_PKT)) { + + /*...