From: Ming Lei <ming.lei@canonical.com>
To: Eric Dumazet <eric.dumazet@gmail.com>
Cc: "David S. Miller" <davem@davemloft.net>,
Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
Oliver Neukum <oneukum@suse.de>, Freddy Xin <freddy@asix.com.tw>,
Ben Hutchings <bhutchings@solarflare.com>,
Grant Grundler <grundler@google.com>,
netdev@vger.kernel.org, linux-usb@vger.kernel.org
Subject: Re: [PATCH 4/4] USBNET: ax88179_178a: enable tso if host supports sg dma
Date: Thu, 1 Aug 2013 12:41:08 +0800 [thread overview]
Message-ID: <CACVXFVNSZJ+ojjBCwYDKsWWvs03aL8BugcX7u483MvBCQCH1sg@mail.gmail.com> (raw)
In-Reply-To: <1375329066.10515.125.camel@edumazet-glaptop>
Hi Eric,
Thanks for your review.
On Thu, Aug 1, 2013 at 11:51 AM, Eric Dumazet <eric.dumazet@gmail.com> wrote:
> On Wed, 2013-07-31 at 18:51 +0800, Ming Lei wrote:
>> This patch enables 'can_dma_sg' flag for ax88179_178a device
>> if the attached host controller supports building packet from
>> discontinuous buffers(DMA SG is possible), so both frame header
>> and skb data buffers can be passed to usb stack via urb->sg,
>> then skb data copy can be saved.
>>
>> With the patch, CPU utilization decreased much in iperf test at
>> client mode.
>>
>> Signed-off-by: Ming Lei <ming.lei@canonical.com>
>> ---
>> drivers/net/usb/ax88179_178a.c | 30 ++++++++++++++++++++++++++++++
>> 1 file changed, 30 insertions(+)
>>
>> diff --git a/drivers/net/usb/ax88179_178a.c b/drivers/net/usb/ax88179_178a.c
>> index 5a468f3..c75bded 100644
>> --- a/drivers/net/usb/ax88179_178a.c
>> +++ b/drivers/net/usb/ax88179_178a.c
>> @@ -1031,12 +1031,20 @@ static int ax88179_bind(struct usbnet *dev, struct usb_interface *intf)
>> dev->mii.phy_id = 0x03;
>> dev->mii.supports_gmii = 1;
>>
>> + if (dev->udev->bus->no_sg_limit)
>> + dev->can_dma_sg = 1;
>> +
>> dev->net->features |= NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
>> NETIF_F_RXCSUM;
>>
>> dev->net->hw_features |= NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
>> NETIF_F_RXCSUM;
>>
>> + if (dev->can_dma_sg) {
>> + dev->net->features |= NETIF_F_SG | NETIF_F_TSO;
>> + dev->net->hw_features |= NETIF_F_SG | NETIF_F_TSO;
>> + }
>> +
>> /* Enable checksum offload */
>> *tmp = AX_RXCOE_IP | AX_RXCOE_TCP | AX_RXCOE_UDP |
>> AX_RXCOE_TCPV6 | AX_RXCOE_UDPV6;
>> @@ -1170,12 +1178,30 @@ ax88179_tx_fixup(struct usbnet *dev, struct sk_buff *skb, gfp_t flags)
>> int mss = skb_shinfo(skb)->gso_size;
>> int headroom;
>> int tailroom;
>> + struct skb_data *entry = (struct skb_data *)skb->cb;
>>
>> tx_hdr1 = skb->len;
>> tx_hdr2 = mss;
>> if (((skb->len + 8) % frame_size) == 0)
>> tx_hdr2 |= 0x80008000; /* Enable padding */
>>
>> + if (dev->can_dma_sg) {
>
>
>
>> + if (!(entry->header = kmalloc(8, GFP_ATOMIC)))
>> + goto no_sg;
>> +
>> + entry->length = 8;
>> + cpu_to_le32s(&tx_hdr1);
>> + cpu_to_le32s(&tx_hdr2);
>
> You could do these cpu_to_le32s() calls before the if (dev->can_dma_sg)
> and remove them before the skb_copy_to_linear_data() calls.
OK.
>
>> + memcpy(entry->header, &tx_hdr1, 4);
>> + memcpy(entry->header + 4, &tx_hdr2, 4);
>> +
>> + return skb;
>> + } else {
>> +no_sg:
>> + entry->header = NULL;
>> + entry->length = 0;
>> + }
>> +
>> headroom = skb_headroom(skb);
>> tailroom = skb_tailroom(skb);
>>
>> @@ -1323,6 +1349,10 @@ static int ax88179_reset(struct usbnet *dev)
>>
>> dev->net->hw_features |= NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
>> NETIF_F_RXCSUM;
>> + if (dev->can_dma_sg) {
>> + dev->net->features |= NETIF_F_SG | NETIF_F_TSO;
>> + dev->net->hw_features |= NETIF_F_SG | NETIF_F_TSO;
>> + }
>
> Are you sure this part is needed ?
I add flags here because the part existed previously, and if you are sure
it isn't needed anymore, I can remove this part.
>
> Also it looks like that you are going through this sg allocation/setup
> even for linear skb ?
>
> For linear skb, with 8+ bytes of headroom, I guess this adds significant
> overhead (2 kmalloc() + sg setup)
>From my trace result, lots of linear SKBs are cloned or header-cloned, so
it needs skb copy too.
Is it normal in xmit path to see cloned SKBs for driver? If not, I can add check
to avoid allocation of 8 bytes header for non-cloned skb.
But for the kmalloc of urb->sg, I don't think there is one way to save that.
Thanks,
--
Ming Lei
next prev parent reply other threads:[~2013-08-01 4:41 UTC|newest]
Thread overview: 29+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-07-31 10:51 [PATCH 0/4] USB & USBNET: loose DMA SG check and support usbnet DMA SG Ming Lei
2013-07-31 10:51 ` [PATCH 1/4] USB: introduce no_sg_limit field into usb_bus Ming Lei
[not found] ` <1375267909-30373-1-git-send-email-ming.lei-Z7WLFzj8eWMS+FvcfC7Uqw@public.gmane.org>
2013-07-31 10:51 ` [PATCH 2/4] USB: XHCI: mark no_sg_limit Ming Lei
2013-07-31 16:40 ` Sarah Sharp
2013-08-01 7:30 ` Greg Kroah-Hartman
2013-08-01 8:15 ` Oliver Neukum
2013-08-01 20:47 ` Sarah Sharp
2013-07-31 15:25 ` [PATCH 0/4] USB & USBNET: loose DMA SG check and support usbnet DMA SG Eric Dumazet
2013-07-31 10:51 ` [PATCH 3/4] USBNET: support " Ming Lei
2013-07-31 10:51 ` [PATCH 4/4] USBNET: ax88179_178a: enable tso if host supports sg dma Ming Lei
2013-07-31 12:47 ` Greg Kroah-Hartman
2013-07-31 13:50 ` Ming Lei
2013-07-31 14:02 ` Oliver Neukum
[not found] ` <1375279366.14846.1.camel-B2T3B9s34ElbnMAlSieJcQ@public.gmane.org>
2013-07-31 15:15 ` Eric Dumazet
2013-07-31 15:25 ` Ming Lei
2013-08-01 2:04 ` Ming Lei
[not found] ` <1375267909-30373-5-git-send-email-ming.lei-Z7WLFzj8eWMS+FvcfC7Uqw@public.gmane.org>
2013-08-01 3:51 ` Eric Dumazet
2013-08-01 4:41 ` Ming Lei [this message]
2013-08-01 5:04 ` Eric Dumazet
2013-08-01 8:10 ` Ming Lei
[not found] ` <CACVXFVMt-TgRVU+iA3NVuUzJSeXUPGvV89OnDdHaLEB1Ua6aAw-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
2013-08-01 13:21 ` Eric Dumazet
2013-08-01 13:49 ` [PATCH net-next] ax88179_178a: avoid copy of tx tcp packets Eric Dumazet
2013-08-01 13:52 ` Eric Dumazet
2013-08-04 4:56 ` David Miller
2013-08-01 15:30 ` [PATCH 4/4] USBNET: ax88179_178a: enable tso if host supports sg dma Grant Grundler
2013-08-01 15:36 ` Eric Dumazet
2013-08-01 16:12 ` Grant Grundler
2013-07-31 11:12 ` [PATCH 0/4] USB & USBNET: loose DMA SG check and support usbnet DMA SG Oliver Neukum
2013-07-31 11:55 ` Ming Lei
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=CACVXFVNSZJ+ojjBCwYDKsWWvs03aL8BugcX7u483MvBCQCH1sg@mail.gmail.com \
--to=ming.lei@canonical.com \
--cc=bhutchings@solarflare.com \
--cc=davem@davemloft.net \
--cc=eric.dumazet@gmail.com \
--cc=freddy@asix.com.tw \
--cc=gregkh@linuxfoundation.org \
--cc=grundler@google.com \
--cc=linux-usb@vger.kernel.org \
--cc=netdev@vger.kernel.org \
--cc=oneukum@suse.de \
/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).