From: Ben Hutchings <bhutchings@solarflare.com>
To: Eric Dumazet <eric.dumazet@gmail.com>
Cc: David Miller <davem@davemloft.net>, <netdev@vger.kernel.org>,
<linux-net-drivers@solarflare.com>
Subject: Re: [PATCH net-next 3/9] sfc: Increase size of RX SKB header area
Date: Tue, 16 Jul 2013 21:57:05 +0100 [thread overview]
Message-ID: <1374008225.2120.26.camel@bwh-desktop.uk.level5networks.com> (raw)
In-Reply-To: <1374006120.6097.6.camel@edumazet-glaptop>
On Tue, 2013-07-16 at 13:22 -0700, Eric Dumazet wrote:
> On Tue, 2013-07-16 at 18:53 +0100, Ben Hutchings wrote:
>
> > Perhaps, yes. I also thought of using some of the other parser status
> > from RX completions to estimate the length of headers.
>
> That would be ideal indeed, I can provide a patch.
This is what I was intending to do, but it's untested. It doesn't allow
for a VLAN tag, but that can be done by reading the RX_EV_PKT_TYPE field
in completions.
Ben.
Subject: WIP: sfc: Pull less data into header area in efx_rx_mk_skb()
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
drivers/net/ethernet/sfc/net_driver.h | 2 ++
drivers/net/ethernet/sfc/nic.c | 3 +++
drivers/net/ethernet/sfc/rx.c | 34 ++++++++++++++++++++++++----------
3 files changed, 29 insertions(+), 10 deletions(-)
diff --git a/drivers/net/ethernet/sfc/net_driver.h b/drivers/net/ethernet/sfc/net_driver.h
index f4c7e6b..cbcf709 100644
--- a/drivers/net/ethernet/sfc/net_driver.h
+++ b/drivers/net/ethernet/sfc/net_driver.h
@@ -243,6 +243,8 @@ struct efx_rx_buffer {
#define EFX_RX_BUF_LAST_IN_PAGE 0x0001
#define EFX_RX_PKT_CSUMMED 0x0002
#define EFX_RX_PKT_DISCARD 0x0004
+#define EFX_RX_PKT_IPV4 0x0010
+#define EFX_RX_PKT_IPV6 0x0020
#define EFX_RX_PKT_TCP 0x0040
/**
diff --git a/drivers/net/ethernet/sfc/nic.c b/drivers/net/ethernet/sfc/nic.c
index 56ed3bc..b6ae31b 100644
--- a/drivers/net/ethernet/sfc/nic.c
+++ b/drivers/net/ethernet/sfc/nic.c
@@ -1093,6 +1093,9 @@ efx_handle_rx_event(struct efx_channel *channel, const efx_qword_t *event)
flags |= EFX_RX_PKT_CSUMMED;
/* fall through */
case FSE_CZ_RX_EV_HDR_TYPE_IPV4V6_OTHER:
+ flags |= EFX_QWORD_FIELD(*event, FSF_CZ_RX_EV_IPV6_PKT)
+ ? EFX_RX_PKT_IPV6 : EFX_RX_PKT_IPV4;
+ break;
case FSE_AZ_RX_EV_HDR_TYPE_OTHER:
break;
}
diff --git a/drivers/net/ethernet/sfc/rx.c b/drivers/net/ethernet/sfc/rx.c
index 65646cd..80bb7a5 100644
--- a/drivers/net/ethernet/sfc/rx.c
+++ b/drivers/net/ethernet/sfc/rx.c
@@ -451,25 +451,27 @@ efx_rx_packet_gro(struct efx_channel *channel, struct efx_rx_buffer *rx_buf,
static struct sk_buff *efx_rx_mk_skb(struct efx_channel *channel,
struct efx_rx_buffer *rx_buf,
unsigned int n_frags,
- u8 *eh, int hdr_len)
+ u8 *eh, int min_hdr_len,
+ int max_hdr_len)
{
struct efx_nic *efx = channel->efx;
struct sk_buff *skb;
/* Allocate an SKB to store the headers */
- skb = netdev_alloc_skb(efx->net_dev, hdr_len + EFX_PAGE_SKB_ALIGN);
+ skb = netdev_alloc_skb(efx->net_dev, max_hdr_len + EFX_PAGE_SKB_ALIGN);
if (unlikely(skb == NULL))
return NULL;
- EFX_BUG_ON_PARANOID(rx_buf->len < hdr_len);
+ EFX_BUG_ON_PARANOID(rx_buf->len < min_hdr_len);
+ EFX_BUG_ON_PARANOID(max_hdr_len < min_hdr_len);
skb_reserve(skb, EFX_PAGE_SKB_ALIGN);
- memcpy(__skb_put(skb, hdr_len), eh, hdr_len);
+ memcpy(__skb_put(skb, min_hdr_len), eh, min_hdr_len);
/* Append the remaining page(s) onto the frag list */
- if (rx_buf->len > hdr_len) {
- rx_buf->page_offset += hdr_len;
- rx_buf->len -= hdr_len;
+ if (rx_buf->len > min_hdr_len) {
+ rx_buf->page_offset += min_hdr_len;
+ rx_buf->len -= min_hdr_len;
for (;;) {
skb_fill_page_desc(skb, skb_shinfo(skb)->nr_frags,
@@ -587,9 +589,21 @@ static void efx_rx_deliver(struct efx_channel *channel, u8 *eh,
unsigned int n_frags)
{
struct sk_buff *skb;
- u16 hdr_len = min_t(u16, rx_buf->len, EFX_SKB_HEADERS);
-
- skb = efx_rx_mk_skb(channel, rx_buf, n_frags, eh, hdr_len);
+ int min_hdr_len = ETH_HLEN;
+
+ /* If the RX parser found an IPv4 or IPv6 header, assume we'll
+ * need to pull that and at least another 8 bytes of transport
+ * header. TODO: should handle VLAN tag too
+ */
+ if (rx_buf->flags & EFX_RX_PKT_IPV4)
+ min_hdr_len += sizeof(struct iphdr) + 8;
+ else if (rx_buf->flags & EFX_RX_PKT_IPV6)
+ min_hdr_len += sizeof(struct ipv6hdr) + 8;
+ if (unlikely(min_hdr_len) > rx_buf->len)
+ min_hdr_len = rx_buf->len;
+
+ skb = efx_rx_mk_skb(channel, rx_buf, n_frags, eh,
+ min_hdr_len, EFX_SKB_HEADERS);
if (unlikely(skb == NULL)) {
efx_free_rx_buffer(rx_buf);
return;
--
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
next prev parent reply other threads:[~2013-07-16 20:57 UTC|newest]
Thread overview: 15+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-06-24 20:11 Pull request: sfc-next 2013-06-24 Ben Hutchings
2013-06-24 20:12 ` [PATCH net-next 1/9] sfc: Fix EEH with legacy interrupts Ben Hutchings
2013-06-24 20:13 ` [PATCH net-next 2/9] sfc: Enable RX checksum offload for packets not handled by GRO Ben Hutchings
2013-06-24 20:13 ` [PATCH net-next 3/9] sfc: Increase size of RX SKB header area Ben Hutchings
2013-07-16 17:33 ` Eric Dumazet
2013-07-16 17:53 ` Ben Hutchings
2013-07-16 20:22 ` Eric Dumazet
2013-07-16 20:57 ` Ben Hutchings [this message]
2013-06-24 20:13 ` [PATCH net-next 4/9] sfc: Report software timestamping capabilities Ben Hutchings
2013-06-24 20:14 ` [PATCH net-next 5/9] sfc: Enable accelerated RFS on vlans Ben Hutchings
2013-06-24 20:14 ` [PATCH net-next 6/9] sfc: Define and set RX buffer flag for packets parsed as TCP Ben Hutchings
2013-06-24 20:14 ` [PATCH net-next 7/9] sfc: Do not pass non-TCP packets into GRO code Ben Hutchings
2013-06-24 20:15 ` [PATCH net-next 8/9] sfc: Fix IRQ cleanup in case of a probe failure Ben Hutchings
2013-06-24 20:15 ` [PATCH net-next 9/9] sfc: Improve test for IOMMU in use Ben Hutchings
2013-06-25 23:12 ` Pull request: sfc-next 2013-06-24 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=1374008225.2120.26.camel@bwh-desktop.uk.level5networks.com \
--to=bhutchings@solarflare.com \
--cc=davem@davemloft.net \
--cc=eric.dumazet@gmail.com \
--cc=linux-net-drivers@solarflare.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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).