From: Stephen Hemminger <shemminger@linux-foundation.org>
To: Patrick McHardy <kaber@trash.net>
Cc: Herbert Xu <herbert.xu@redhat.com>,
"David S. Miller" <davem@davemloft.net>,
netdev@vger.kernel.org, Michael Chan <mchan@broadcom.com>
Subject: Re: [RFC] IPV6 checksum offloading in network devices
Date: Wed, 30 May 2007 14:00:34 -0700 [thread overview]
Message-ID: <20070530140034.13bbf971@freepuppy> (raw)
In-Reply-To: <465DA2A4.2060608@trash.net>
This is better.
There is still a possiblity when a device allows IPV6 and not IPV4
checksumming, that the checksum will be done in the fixup code in
dev_queue_xmit.
The existing model for checksum offload does not correctly handle
devices that can offload IPV4 and IPV6 only. The NETIF_F_HW_CSUM flag
implies device can do any arbitrary protocol.
This patch:
* adds NETIF_F_IPV6_CSUM for those devices
* fixes bnx2 and tg3 devices that need it
* add NETIF_F_IPV6_CSUM to ipv6 output (incl GSO)
* fixes assumptions about NETIF_F_ALL_CSUM in nat
* adjusts bridge union of checksumming computation
---
drivers/net/bnx2.c | 6 +++---
drivers/net/tg3.c | 7 +++----
include/linux/netdevice.h | 8 ++++++--
net/bridge/br_if.c | 10 +++++++++-
net/core/dev.c | 24 +++++++++++++++++++++---
net/ipv4/af_inet.c | 3 +++
net/ipv4/ip_output.c | 2 +-
net/ipv4/netfilter/nf_nat_helper.c | 4 ++--
net/ipv6/ipv6_sockglue.c | 2 +-
9 files changed, 49 insertions(+), 17 deletions(-)
--- a/drivers/net/bnx2.c 2007-05-30 08:26:18.000000000 -0700
+++ b/drivers/net/bnx2.c 2007-05-30 08:26:32.000000000 -0700
@@ -6471,10 +6471,10 @@ bnx2_init_one(struct pci_dev *pdev, cons
memcpy(dev->perm_addr, bp->mac_addr, 6);
bp->name = board_info[ent->driver_data].name;
+ dev->features |= NETIF_F_IP_CSUM | NETIF_F_SG;
if (CHIP_NUM(bp) == CHIP_NUM_5709)
- dev->features |= NETIF_F_HW_CSUM | NETIF_F_SG;
- else
- dev->features |= NETIF_F_IP_CSUM | NETIF_F_SG;
+ dev->features |= NETIF_F_IPV6_CSUM;
+
#ifdef BCM_VLAN
dev->features |= NETIF_F_HW_VLAN_TX | NETIF_F_HW_VLAN_RX;
#endif
--- a/drivers/net/tg3.c 2007-05-30 08:26:18.000000000 -0700
+++ b/drivers/net/tg3.c 2007-05-30 08:26:32.000000000 -0700
@@ -11959,12 +11959,11 @@ static int __devinit tg3_init_one(struct
* checksumming.
*/
if ((tp->tg3_flags & TG3_FLAG_BROKEN_CHECKSUMS) == 0) {
+ dev->features |= NETIF_F_IP_CSUM | NETIF_F_SG;
if (GET_ASIC_REV(tp->pci_chip_rev_id) == ASIC_REV_5755 ||
GET_ASIC_REV(tp->pci_chip_rev_id) == ASIC_REV_5787)
- dev->features |= NETIF_F_HW_CSUM;
- else
- dev->features |= NETIF_F_IP_CSUM;
- dev->features |= NETIF_F_SG;
+ dev->features |= NETIF_F_IPV6_CSUM;
+
tp->tg3_flags |= TG3_FLAG_RX_CHECKSUMS;
} else
tp->tg3_flags &= ~TG3_FLAG_RX_CHECKSUMS;
--- a/include/linux/netdevice.h 2007-05-30 08:26:18.000000000 -0700
+++ b/include/linux/netdevice.h 2007-05-30 10:10:57.000000000 -0700
@@ -314,9 +314,10 @@ struct net_device
/* Net device features */
unsigned long features;
#define NETIF_F_SG 1 /* Scatter/gather IO. */
-#define NETIF_F_IP_CSUM 2 /* Can checksum only TCP/UDP over IPv4. */
+#define NETIF_F_IP_CSUM 2 /* Can checksum TCP/UDP over IPv4. */
#define NETIF_F_NO_CSUM 4 /* Does not require checksum. F.e. loopack. */
#define NETIF_F_HW_CSUM 8 /* Can checksum all the packets. */
+#define NETIF_F_IPV6_CSUM 16 /* Can checksum TCP/UDP over IPV6 */
#define NETIF_F_HIGHDMA 32 /* Can DMA to high memory. */
#define NETIF_F_FRAGLIST 64 /* Scatter/gather IO. */
#define NETIF_F_HW_VLAN_TX 128 /* Transmit VLAN hw acceleration */
@@ -338,8 +339,11 @@ struct net_device
/* List of features with software fallbacks. */
#define NETIF_F_GSO_SOFTWARE (NETIF_F_TSO | NETIF_F_TSO_ECN | NETIF_F_TSO6)
+
#define NETIF_F_GEN_CSUM (NETIF_F_NO_CSUM | NETIF_F_HW_CSUM)
-#define NETIF_F_ALL_CSUM (NETIF_F_IP_CSUM | NETIF_F_GEN_CSUM)
+#define NETIF_F_V4_CSUM (NETIF_F_GEN_CSUM | NETIF_F_IP_CSUM)
+#define NETIF_F_V6_CSUM (NETIF_F_GEN_CSUM | NETIF_F_IPV6_CSUM)
+#define NETIF_F_ALL_CSUM (NETIF_F_V4_CSUM | NETIF_F_V6_CSUM)
struct net_device *next_sched;
--- a/net/bridge/br_if.c 2007-05-30 08:26:18.000000000 -0700
+++ b/net/bridge/br_if.c 2007-05-30 10:07:22.000000000 -0700
@@ -368,10 +368,18 @@ void br_features_recompute(struct net_br
list_for_each_entry(p, &br->port_list, list) {
unsigned long feature = p->dev->features;
+ /* if device needs checksumming, downgrade to hw checksumming */
if (checksum & NETIF_F_NO_CSUM && !(feature & NETIF_F_NO_CSUM))
checksum ^= NETIF_F_NO_CSUM | NETIF_F_HW_CSUM;
+
+ /* if device can't do all checksum, downgrade to ipv4/ipv6 */
if (checksum & NETIF_F_HW_CSUM && !(feature & NETIF_F_HW_CSUM))
- checksum ^= NETIF_F_HW_CSUM | NETIF_F_IP_CSUM;
+ checksum ^= NETIF_F_HW_CSUM
+ | NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
+
+ if (checksum & NETIF_F_IPV6_CSUM && !(feature & NETIF_F_IPV6_CSUM))
+ checksum &= ~NETIF_F_IPV6_CSUM;
+
if (!(feature & NETIF_F_IP_CSUM))
checksum = 0;
--- a/net/core/dev.c 2007-05-30 08:26:18.000000000 -0700
+++ b/net/core/dev.c 2007-05-30 10:07:31.000000000 -0700
@@ -1509,9 +1509,11 @@ int dev_queue_xmit(struct sk_buff *skb)
skb_set_transport_header(skb, skb->csum_start -
skb_headroom(skb));
- if (!(dev->features & NETIF_F_GEN_CSUM) &&
- (!(dev->features & NETIF_F_IP_CSUM) ||
- skb->protocol != htons(ETH_P_IP)))
+ if (!(dev->features & NETIF_F_GEN_CSUM)
+ || ((dev->features & NETIF_F_IP_CSUM)
+ && skb->protocol == htons(ETH_P_IP))
+ || ((dev->features & NETIF_F_IPV6_CSUM)
+ && skb->protocol == htons(ETH_P_IPV6)))
if (skb_checksum_help(skb))
goto out_kfree_skb;
}
@@ -3105,6 +3107,22 @@ int register_netdevice(struct net_device
}
}
+ /* Fix illegal checksum combinations */
+ if ((dev->features & NETIF_F_HW_CSUM) &&
+ (dev->features & (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) {
+ printk(KERN_NOTICE "%s: mixed HW and IP checksum settings.\n",
+ dev->name);
+ dev->features &= ~(NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
+ }
+
+ if ((dev->features & NETIF_F_NO_CSUM) &&
+ (dev->features & (NETIF_F_HW_CSUM|NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) {
+ printk(KERN_NOTICE "%s: mixed no checksumming and other settings.\n",
+ dev->name);
+ dev->features &= ~(NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM|NETIF_F_HW_CSUM);
+ }
+
+
/* Fix illegal SG+CSUM combinations. */
if ((dev->features & NETIF_F_SG) &&
!(dev->features & NETIF_F_ALL_CSUM)) {
--- a/net/ipv6/ipv6_sockglue.c 2007-05-29 16:44:45.000000000 -0700
+++ b/net/ipv6/ipv6_sockglue.c 2007-05-30 09:54:52.000000000 -0700
@@ -123,7 +123,7 @@ static struct sk_buff *ipv6_gso_segment(
struct ipv6hdr *ipv6h;
struct inet6_protocol *ops;
- if (!(features & NETIF_F_HW_CSUM))
+ if (!(features & NETIF_F_V6_CSUM))
features &= ~NETIF_F_SG;
if (unlikely(skb_shinfo(skb)->gso_type &
--- a/net/ipv4/af_inet.c 2007-05-29 16:44:45.000000000 -0700
+++ b/net/ipv4/af_inet.c 2007-05-30 09:55:11.000000000 -0700
@@ -1170,6 +1170,9 @@ static struct sk_buff *inet_gso_segment(
int ihl;
int id;
+ if (!(features & NETIF_F_V4_CSUM))
+ features &= ~NETIF_F_SG;
+
if (unlikely(skb_shinfo(skb)->gso_type &
~(SKB_GSO_TCPV4 |
SKB_GSO_UDP |
--- a/net/ipv4/ip_output.c 2007-05-29 16:44:45.000000000 -0700
+++ b/net/ipv4/ip_output.c 2007-05-30 09:51:33.000000000 -0700
@@ -837,7 +837,7 @@ int ip_append_data(struct sock *sk,
*/
if (transhdrlen &&
length + fragheaderlen <= mtu &&
- rt->u.dst.dev->features & NETIF_F_ALL_CSUM &&
+ rt->u.dst.dev->features & NETIF_F_V4_CSUM &&
!exthdrlen)
csummode = CHECKSUM_PARTIAL;
--- a/net/ipv4/netfilter/nf_nat_helper.c 2007-05-29 16:44:45.000000000 -0700
+++ b/net/ipv4/netfilter/nf_nat_helper.c 2007-05-30 10:01:33.000000000 -0700
@@ -178,7 +178,7 @@ nf_nat_mangle_tcp_packet(struct sk_buff
datalen = (*pskb)->len - iph->ihl*4;
if ((*pskb)->ip_summed != CHECKSUM_PARTIAL) {
if (!(rt->rt_flags & RTCF_LOCAL) &&
- (*pskb)->dev->features & NETIF_F_ALL_CSUM) {
+ (*pskb)->dev->features & NETIF_F_V4_CSUM) {
(*pskb)->ip_summed = CHECKSUM_PARTIAL;
(*pskb)->csum_start = skb_headroom(*pskb) +
skb_network_offset(*pskb) +
@@ -265,7 +265,7 @@ nf_nat_mangle_udp_packet(struct sk_buff
if ((*pskb)->ip_summed != CHECKSUM_PARTIAL) {
if (!(rt->rt_flags & RTCF_LOCAL) &&
- (*pskb)->dev->features & NETIF_F_ALL_CSUM) {
+ (*pskb)->dev->features & NETIF_F_V4_CSUM) {
(*pskb)->ip_summed = CHECKSUM_PARTIAL;
(*pskb)->csum_start = skb_headroom(*pskb) +
skb_network_offset(*pskb) +
next prev parent reply other threads:[~2007-05-30 21:12 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2007-05-29 20:58 Definition and usage of NETIF_F_HW_SUM? Stephen Hemminger
2007-05-29 21:36 ` Herbert Xu
2007-05-29 21:58 ` Stephen Hemminger
2007-05-30 0:10 ` Michael Chan
2007-05-29 23:45 ` Stephen Hemminger
2007-06-04 15:35 ` Ron Mercer
2007-05-30 15:53 ` [RFC] IPV6 checksum offloading in network devices Stephen Hemminger
2007-05-30 16:13 ` Patrick McHardy
2007-05-30 21:00 ` Stephen Hemminger [this message]
2007-06-27 7:44 ` David Miller
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20070530140034.13bbf971@freepuppy \
--to=shemminger@linux-foundation.org \
--cc=davem@davemloft.net \
--cc=herbert.xu@redhat.com \
--cc=kaber@trash.net \
--cc=mchan@broadcom.com \
--cc=netdev@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.