PATCH: af_packet / TX_RING not fully non-blocking (w/ MSG_DONTWAIT)

PATCH: af_packet / TX_RING not fully non-blocking (w/ MSG_DONTWAIT)

[Mathias Kretschmer]

[-- Attachment #1: Type: text/plain, Size: 1467 bytes --]

Hi,

following up on my initial post and the feedback I have received,
please find attached/in-line an updated patch.

Changes:
*) keep the unlikely() in the case sock_alloc_send_skb() fails, as here 
we enter the slow path, anyway.
*) pass on the return code from sock_alloc_send_skb(), i.e. EGAIN

Please consider this for inclusion.

Cheers,

Mathias

---
diff -uNpr linux-3.16.7.orig/net/packet/af_packet.c 
linux-3.16.7/net/packet/af_packet.c
--- linux-3.16.7.orig/net/packet/af_packet.c    2014-10-30 
16:41:01.000000000 +0000
+++ linux-3.16.7/net/packet/af_packet.c 2015-04-05 08:41:14.577535454 +0000
@@ -2291,11 +2291,14 @@ static int tpacket_snd(struct packet_soc
                 tlen = dev->needed_tailroom;
                 skb = sock_alloc_send_skb(&po->sk,
                                 hlen + tlen + sizeof(struct sockaddr_ll),
-                               0, &err);
+                               !need_wait, &err);

-               if (unlikely(skb == NULL))
+               if (unlikely(skb == NULL)) {
+                       /* we assume the socket was initially writeable 
... */
+                       if (likely(len_sum > 0))
+                               err = len_sum;
                         goto out_status;
-
+               }
                 tp_len = tpacket_fill_skb(po, skb, ph, dev, size_max, 
proto,
                                           addr, hlen);
                 if (tp_len > dev->mtu + dev->hard_header_len) {

[-- Attachment #2: 005-af_packet_no_block_tx.patch --]
[-- Type: text/x-patch, Size: 802 bytes --]

diff -uNpr linux-3.16.7.orig/net/packet/af_packet.c linux-3.16.7/net/packet/af_packet.c
--- linux-3.16.7.orig/net/packet/af_packet.c	2014-10-30 16:41:01.000000000 +0000
+++ linux-3.16.7/net/packet/af_packet.c	2015-04-05 08:41:14.577535454 +0000
@@ -2291,11 +2291,14 @@ static int tpacket_snd(struct packet_soc
 		tlen = dev->needed_tailroom;
 		skb = sock_alloc_send_skb(&po->sk,
 				hlen + tlen + sizeof(struct sockaddr_ll),
-				0, &err);
+				!need_wait, &err);
 
-		if (unlikely(skb == NULL))
+		if (unlikely(skb == NULL)) {
+			/* we assume the socket was initially writeable ... */
+			if (likely(len_sum > 0))
+				err = len_sum;
 			goto out_status;
-
+		}
 		tp_len = tpacket_fill_skb(po, skb, ph, dev, size_max, proto,
 					  addr, hlen);
 		if (tp_len > dev->mtu + dev->hard_header_len) {

Re: PATCH: af_packet / TX_RING not fully non-blocking (w/ MSG_DONTWAIT)

[Daniel Borkmann]

Hi Mathias,

On 05/03/2015 01:28 PM, Mathias Kretschmer wrote:
> Hi,
>
> following up on my initial post and the feedback I have received,
> please find attached/in-line an updated patch.

Yep, thanks for following up on this.

However, your submission (https://patchwork.ozlabs.org/patch/467377/)
format is not correct. Your email client corrupted the whitespace,
your Signed-off-by is missing, a proper commit description is missing
and which tree your patch is targeted at. I presume -net tree as it's
a fix.

You're best off sending this patch via git-send-email(1), see also
Documentation/SubmittingPatches from the kernel tree.

Taking a random example submission from patchwork, this is how the end
result could look like:

   https://patchwork.ozlabs.org/patch/466743/

Thus, if you just write a proper commit message in git and use
git-format-patch(1) and git-send-email(1), you should be all set. Hope
that helps.

Cheers,
Daniel

> Changes:
> *) keep the unlikely() in the case sock_alloc_send_skb() fails, as here we enter the slow path, anyway.
> *) pass on the return code from sock_alloc_send_skb(), i.e. EGAIN
>
> Please consider this for inclusion.
>
> Cheers,
>
> Mathias