From: Ben Hutchings <ben@decadent.org.uk>
To: linux-kernel@vger.kernel.org, stable@vger.kernel.org
Cc: akpm@linux-foundation.org,
"Greg Kroah-Hartman" <gregkh@linuxfoundation.org>,
"Paul Stewart" <pstew@chromium.org>,
"Aurelien Jacobs" <aurel@gnuage.org>,
"Trond Wuellner" <trond@chromium.org>,
"Eric Dumazet" <eric.dumazet@gmail.com>,
"Emil Goode" <emilgoode@gmail.com>,
"Grant Grundler" <grundler@chromium.org>,
"David S. Miller" <davem@davemloft.net>
Subject: [PATCH 3.2 12/18] asix: asix_rx_fixup surgery to reduce skb truesizes
Date: Mon, 07 Apr 2014 00:35:48 +0100 [thread overview]
Message-ID: <lsq.1396827348.590224716@decadent.org.uk> (raw)
In-Reply-To: <lsq.1396827347.462803115@decadent.org.uk>
3.2.57-rc1 review patch. If anyone has any objections, please let me know.
------------------
From: Eric Dumazet <eric.dumazet@gmail.com>
commit a9e0aca4b37885b5599e52211f098bd7f565e749 upstream.
asix_rx_fixup() is complex, and does some unnecessary memory copies (at
least on x86 where NET_IP_ALIGN is 0)
Also, it tends to provide skbs with a big truesize (4096+256 with
MTU=1500) to upper stack, so incoming trafic consume a lot of memory and
I noticed early packet drops because we hit socket rcvbuf too fast.
Switch to a different strategy, using copybreak so that we provide nice
skbs to upper stack (including the NET_SKB_PAD to avoid future head
reallocations in some paths)
With this patch, I no longer see packets drops or tcp collapses on
various tcp workload with a AX88772 adapter.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Aurelien Jacobs <aurel@gnuage.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Trond Wuellner <trond@chromium.org>
Cc: Grant Grundler <grundler@chromium.org>
Cc: Paul Stewart <pstew@chromium.org>
Reviewed-by: Grant Grundler <grundler@chromium.org>
Reviewed-by: Grant Grundler <grundler@chromium.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
[ Emil: Backported to 3.2: fixed small conflict ]
Signed-off-by: Emil Goode <emilgoode@gmail.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
---
drivers/net/usb/asix.c | 88 +++++++++++-------------------------------------
1 file changed, 20 insertions(+), 68 deletions(-)
--- a/drivers/net/usb/asix.c
+++ b/drivers/net/usb/asix.c
@@ -306,88 +306,40 @@ asix_write_cmd_async(struct usbnet *dev,
static int asix_rx_fixup(struct usbnet *dev, struct sk_buff *skb)
{
- u8 *head;
- u32 header;
- char *packet;
- struct sk_buff *ax_skb;
- u16 size;
-
- head = (u8 *) skb->data;
- memcpy(&header, head, sizeof(header));
- le32_to_cpus(&header);
- packet = head + sizeof(header);
+ int offset = 0;
- skb_pull(skb, 4);
+ while (offset + sizeof(u32) < skb->len) {
+ struct sk_buff *ax_skb;
+ u16 size;
+ u32 header = get_unaligned_le32(skb->data + offset);
- while (skb->len > 0) {
- if ((header & 0x07ff) != ((~header >> 16) & 0x07ff))
- netdev_err(dev->net, "asix_rx_fixup() Bad Header Length\n");
+ offset += sizeof(u32);
/* get the packet length */
- size = (u16) (header & 0x000007ff);
-
- if ((skb->len) - ((size + 1) & 0xfffe) == 0) {
- u8 alignment = (unsigned long)skb->data & 0x3;
- if (alignment != 0x2) {
- /*
- * not 16bit aligned so use the room provided by
- * the 32 bit header to align the data
- *
- * note we want 16bit alignment as MAC header is
- * 14bytes thus ip header will be aligned on
- * 32bit boundary so accessing ipheader elements
- * using a cast to struct ip header wont cause
- * an unaligned accesses.
- */
- u8 realignment = (alignment + 2) & 0x3;
- memmove(skb->data - realignment,
- skb->data,
- size);
- skb->data -= realignment;
- skb_set_tail_pointer(skb, size);
- }
- return 2;
+ size = (u16) (header & 0x7ff);
+ if (size != ((~header >> 16) & 0x07ff)) {
+ netdev_err(dev->net, "asix_rx_fixup() Bad Header Length\n");
+ return 0;
}
- if (size > dev->net->mtu + ETH_HLEN + VLAN_HLEN) {
+ if ((size > dev->net->mtu + ETH_HLEN + VLAN_HLEN) ||
+ (size + offset > skb->len)) {
netdev_err(dev->net, "asix_rx_fixup() Bad RX Length %d\n",
size);
return 0;
}
- ax_skb = skb_clone(skb, GFP_ATOMIC);
- if (ax_skb) {
- u8 alignment = (unsigned long)packet & 0x3;
- ax_skb->len = size;
-
- if (alignment != 0x2) {
- /*
- * not 16bit aligned use the room provided by
- * the 32 bit header to align the data
- */
- u8 realignment = (alignment + 2) & 0x3;
- memmove(packet - realignment, packet, size);
- packet -= realignment;
- }
- ax_skb->data = packet;
- skb_set_tail_pointer(ax_skb, size);
- usbnet_skb_return(dev, ax_skb);
- } else {
+ ax_skb = netdev_alloc_skb_ip_align(dev->net, size);
+ if (!ax_skb)
return 0;
- }
-
- skb_pull(skb, (size + 1) & 0xfffe);
- if (skb->len < sizeof(header))
- break;
+ skb_put(ax_skb, size);
+ memcpy(ax_skb->data, skb->data + offset, size);
+ usbnet_skb_return(dev, ax_skb);
- head = (u8 *) skb->data;
- memcpy(&header, head, sizeof(header));
- le32_to_cpus(&header);
- packet = head + sizeof(header);
- skb_pull(skb, 4);
+ offset += (size + 1) & 0xfffe;
}
- if (skb->len < 0) {
+ if (skb->len != offset) {
netdev_err(dev->net, "asix_rx_fixup() Bad SKB Length %d\n",
skb->len);
return 0;
@@ -1538,7 +1490,7 @@ static const struct driver_info ax88772_
.status = asix_status,
.link_reset = ax88772_link_reset,
.reset = ax88772_reset,
- .flags = FLAG_ETHER | FLAG_FRAMING_AX | FLAG_LINK_INTR,
+ .flags = FLAG_ETHER | FLAG_FRAMING_AX | FLAG_LINK_INTR | FLAG_MULTI_PACKET,
.rx_fixup = asix_rx_fixup,
.tx_fixup = asix_tx_fixup,
};
next prev parent reply other threads:[~2014-04-06 23:40 UTC|newest]
Thread overview: 25+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-04-06 23:35 [PATCH 3.2 00/18] 3.2.57-rc1 review Ben Hutchings
2014-04-06 23:35 ` [PATCH 3.2 07/18] net: add and use skb_gso_transport_seglen() Ben Hutchings
2014-04-06 23:35 ` [PATCH 3.2 08/18] net: ip, ipv6: handle gso skbs in forwarding path Ben Hutchings
2014-04-06 23:35 ` [PATCH 3.2 18/18] s390: fix kernel crash due to linkage stack instructions Ben Hutchings
2014-04-06 23:35 ` [PATCH 3.2 09/18] deb-pkg: use KCONFIG_CONFIG instead of .config file directly Ben Hutchings
2014-04-06 23:35 ` [PATCH 3.2 05/18] netfilter: nf_conntrack_dccp: fix skb_header_pointer API usages Ben Hutchings
2014-04-06 23:35 ` [PATCH 3.2 03/18] staging: speakup: Prefix set_mask_bits() symbol Ben Hutchings
2014-04-06 23:35 ` [PATCH 3.2 15/18] KVM: MMU: handle invalid root_hpa at __direct_map Ben Hutchings
2014-04-06 23:35 ` [PATCH 3.2 11/18] deb-pkg: Fix cross-building linux-headers package Ben Hutchings
2014-04-06 23:35 ` [PATCH 3.2 01/18] Input: synaptics - add manual min/max quirk Ben Hutchings
2014-04-06 23:35 ` [PATCH 3.2 13/18] net: asix: handle packets crossing URB boundaries Ben Hutchings
2014-04-06 23:35 ` [PATCH 3.2 14/18] net: asix: add missing flag to struct driver_info Ben Hutchings
2014-04-06 23:35 ` [PATCH 3.2 06/18] ipc/msg: fix race around refcount Ben Hutchings
2014-04-06 23:35 ` [PATCH 3.2 02/18] Input: synaptics - add manual min/max quirk for ThinkPad X240 Ben Hutchings
2014-04-06 23:35 ` [PATCH 3.2 04/18] ext4: atomically set inode->i_flags in ext4_set_inode_flags() Ben Hutchings
2014-04-06 23:35 ` [PATCH 3.2 10/18] deb-pkg: Fix building for MIPS big-endian or ARM OABI Ben Hutchings
2014-04-06 23:35 ` [PATCH 3.2 16/18] KVM: VMX: fix use after free of vmx->loaded_vmcs Ben Hutchings
2014-04-06 23:35 ` Ben Hutchings [this message]
2014-04-06 23:35 ` [PATCH 3.2 17/18] cifs: ensure that uncached writes handle unmapped areas correctly Ben Hutchings
2014-04-07 1:41 ` Ben Hutchings
2014-04-07 13:45 ` Raphael Geissert
2014-04-07 19:14 ` Ben Hutchings
2014-04-07 1:42 ` [PATCH 3.2 00/18] 3.2.57-rc1 review Ben Hutchings
2014-04-07 3:55 ` Guenter Roeck
2014-04-07 12:30 ` Ben Hutchings
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=lsq.1396827348.590224716@decadent.org.uk \
--to=ben@decadent.org.uk \
--cc=akpm@linux-foundation.org \
--cc=aurel@gnuage.org \
--cc=davem@davemloft.net \
--cc=emilgoode@gmail.com \
--cc=eric.dumazet@gmail.com \
--cc=gregkh@linuxfoundation.org \
--cc=grundler@chromium.org \
--cc=linux-kernel@vger.kernel.org \
--cc=pstew@chromium.org \
--cc=stable@vger.kernel.org \
--cc=trond@chromium.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
all inboxes | Powered by JetHome®