[PATCH net-next 0/2] Make [IP6]GRE[TAP] devices always NETIF_F_LLTX

Subsystems: networking [general], networking [ipv4/ipv6], the rest · Maintainers: "David S. Miller", Eric Dumazet, Jakub Kicinski, Paolo Abeni, David Ahern, Ido Schimmel, Linus Torvalds

Series: 2 patches touching 2 files, +35/-49

STALE1646d

5 messages, 2 authors, 2022-04-25 · open the first message on its own page

[PATCH net-next 0/2] Make [IP6]GRE[TAP] devices always NETIF_F_LLTX

From: Peilin Ye <hidden>
Date: 2022-04-21 22:44:02 

From: Peilin Ye <redacted>

Hi all,

This patchset depends on these fixes [1].  Since o_seqno is now atomic_t,
we can always turn on NETIF_F_LLTX for [IP6]GRE[TAP] devices, since we no
longer need the TX lock (&txq->_xmit_lock).

We could probably do the same thing to [IP6]ERSPAN devices as well, but
I'm not familiar with them yet.  For example, ERSPAN devices are
initialized as |= GRE_FEATURES in erspan_tunnel_init(), but I don't see
IP6ERSPAN devices being initialized as |= GRE6_FEATURES.  Please suggest
if I'm missing something, thanks!

[1] https://lore.kernel.org/netdev/cover.1650575919.git.peilin.ye@bytedance.com/

Thanks,
Peilin Ye (2):
  ip_gre: Make GRE and GRETAP devices always NETIF_F_LLTX
  ip6_gre: Make IP6GRE and IP6GRETAP devices always NETIF_F_LLTX

 net/ipv4/ip_gre.c  | 50 ++++++++++++++++++++--------------------------
 net/ipv6/ip6_gre.c | 34 ++++++++++++-------------------
 2 files changed, 35 insertions(+), 49 deletions(-)

-- 
2.20.1

[PATCH net-next 1/2] ip_gre: Make GRE and GRETAP devices always NETIF_F_LLTX

From: Peilin Ye <hidden>
Date: 2022-04-21 22:45:04 

From: Peilin Ye <redacted>

Recently we made o_seqno atomic_t.  Stop special-casing TUNNEL_SEQ, and
always mark GRE[TAP] devices as NETIF_F_LLTX, since we no longer need
the TX lock (&txq->_xmit_lock).

Depends on patch "ip_gre, ip6_gre: Fix race condition on o_seqno in
collect_md mode".

Signed-off-by: Peilin Ye <redacted>
---
 net/ipv4/ip_gre.c | 50 +++++++++++++++++++++--------------------------
 1 file changed, 22 insertions(+), 28 deletions(-)

diff --git a/net/ipv4/ip_gre.c b/net/ipv4/ip_gre.c
index 8cf86e42c1d1..a81c2964f70b 100644
--- a/net/ipv4/ip_gre.c
+++ b/net/ipv4/ip_gre.c
@@ -748,6 +748,7 @@ static netdev_tx_t gre_tap_xmit(struct sk_buff *skb,
 static void ipgre_link_update(struct net_device *dev, bool set_mtu)
 {
 	struct ip_tunnel *tunnel = netdev_priv(dev);
+	__be16 flags;
 	int len;
 
 	len = tunnel->tun_hlen;
@@ -763,19 +764,15 @@ static void ipgre_link_update(struct net_device *dev, bool set_mtu)
 	if (set_mtu)
 		dev->mtu = max_t(int, dev->mtu - len, 68);
 
-	if (!(tunnel->parms.o_flags & TUNNEL_SEQ)) {
-		if (!(tunnel->parms.o_flags & TUNNEL_CSUM) ||
-		    tunnel->encap.type == TUNNEL_ENCAP_NONE) {
-			dev->features |= NETIF_F_GSO_SOFTWARE;
-			dev->hw_features |= NETIF_F_GSO_SOFTWARE;
-		} else {
-			dev->features &= ~NETIF_F_GSO_SOFTWARE;
-			dev->hw_features &= ~NETIF_F_GSO_SOFTWARE;
-		}
-		dev->features |= NETIF_F_LLTX;
-	} else {
+	flags = tunnel->parms.o_flags;
+
+	if (flags & TUNNEL_SEQ ||
+	    (flags & TUNNEL_CSUM && tunnel->encap.type != TUNNEL_ENCAP_NONE)) {
+		dev->features &= ~NETIF_F_GSO_SOFTWARE;
 		dev->hw_features &= ~NETIF_F_GSO_SOFTWARE;
-		dev->features &= ~(NETIF_F_LLTX | NETIF_F_GSO_SOFTWARE);
+	} else {
+		dev->features |= NETIF_F_GSO_SOFTWARE;
+		dev->hw_features |= NETIF_F_GSO_SOFTWARE;
 	}
 }
 
@@ -949,6 +946,7 @@ static void ipgre_tunnel_setup(struct net_device *dev)
 static void __gre_tunnel_init(struct net_device *dev)
 {
 	struct ip_tunnel *tunnel;
+	__be16 flags;
 
 	tunnel = netdev_priv(dev);
 	tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
@@ -957,25 +955,21 @@ static void __gre_tunnel_init(struct net_device *dev)
 	tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
 	dev->needed_headroom = tunnel->hlen + sizeof(tunnel->parms.iph);
 
-	dev->features		|= GRE_FEATURES;
+	dev->features		|= GRE_FEATURES | NETIF_F_LLTX;
 	dev->hw_features	|= GRE_FEATURES;
 
-	if (!(tunnel->parms.o_flags & TUNNEL_SEQ)) {
-		/* TCP offload with GRE SEQ is not supported, nor
-		 * can we support 2 levels of outer headers requiring
-		 * an update.
-		 */
-		if (!(tunnel->parms.o_flags & TUNNEL_CSUM) ||
-		    (tunnel->encap.type == TUNNEL_ENCAP_NONE)) {
-			dev->features    |= NETIF_F_GSO_SOFTWARE;
-			dev->hw_features |= NETIF_F_GSO_SOFTWARE;
-		}
+	flags = tunnel->parms.o_flags;
 
-		/* Can use a lockless transmit, unless we generate
-		 * output sequences
-		 */
-		dev->features |= NETIF_F_LLTX;
-	}
+	/* TCP offload with GRE SEQ is not supported, nor can we support 2
+	 * levels of outer headers requiring an update.
+	 */
+	if (flags & TUNNEL_SEQ)
+		return;
+	if (flags & TUNNEL_CSUM && tunnel->encap.type != TUNNEL_ENCAP_NONE)
+		return;
+
+	dev->features |= NETIF_F_GSO_SOFTWARE;
+	dev->hw_features |= NETIF_F_GSO_SOFTWARE;
 }
 
 static int ipgre_tunnel_init(struct net_device *dev)
-- 
2.20.1

[PATCH net-next 2/2] ip6_gre: Make IP6GRE and IP6GRETAP devices always NETIF_F_LLTX

From: Peilin Ye <hidden>
Date: 2022-04-21 22:45:33 

From: Peilin Ye <redacted>

Recently we made o_seqno atomic_t.  Stop special-casing TUNNEL_SEQ, and
always mark IP6GRE[TAP] devices as NETIF_F_LLTX, since we no longer need
the TX lock (&txq->_xmit_lock).

Signed-off-by: Peilin Ye <redacted>
---
 net/ipv6/ip6_gre.c | 34 +++++++++++++---------------------
 1 file changed, 13 insertions(+), 21 deletions(-)

diff --git a/net/ipv6/ip6_gre.c b/net/ipv6/ip6_gre.c
index 5136959b3dc5..4e37f7c29900 100644
--- a/net/ipv6/ip6_gre.c
+++ b/net/ipv6/ip6_gre.c
@@ -382,11 +382,6 @@ static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
 		goto failed_free;
 
 	ip6gre_tnl_link_config(nt, 1);
-
-	/* Can use a lockless transmit, unless we generate output sequences */
-	if (!(nt->parms.o_flags & TUNNEL_SEQ))
-		dev->features |= NETIF_F_LLTX;
-
 	ip6gre_tunnel_link(ign, nt);
 	return nt;
 
@@ -1445,26 +1440,23 @@ static void ip6gre_tunnel_setup(struct net_device *dev)
 static void ip6gre_tnl_init_features(struct net_device *dev)
 {
 	struct ip6_tnl *nt = netdev_priv(dev);
+	__be16 flags;
 
-	dev->features		|= GRE6_FEATURES;
+	dev->features		|= GRE6_FEATURES | NETIF_F_LLTX;
 	dev->hw_features	|= GRE6_FEATURES;
 
-	if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
-		/* TCP offload with GRE SEQ is not supported, nor
-		 * can we support 2 levels of outer headers requiring
-		 * an update.
-		 */
-		if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
-		    nt->encap.type == TUNNEL_ENCAP_NONE) {
-			dev->features    |= NETIF_F_GSO_SOFTWARE;
-			dev->hw_features |= NETIF_F_GSO_SOFTWARE;
-		}
+	flags = nt->parms.o_flags;
 
-		/* Can use a lockless transmit, unless we generate
-		 * output sequences
-		 */
-		dev->features |= NETIF_F_LLTX;
-	}
+	/* TCP offload with GRE SEQ is not supported, nor can we support 2
+	 * levels of outer headers requiring an update.
+	 */
+	if (flags & TUNNEL_SEQ)
+		return;
+	if (flags & TUNNEL_CSUM && nt->encap.type != TUNNEL_ENCAP_NONE)
+		return;
+
+	dev->features |= NETIF_F_GSO_SOFTWARE;
+	dev->hw_features |= NETIF_F_GSO_SOFTWARE;
 }
 
 static int ip6gre_tunnel_init_common(struct net_device *dev)
-- 
2.20.1

Re: [PATCH net-next 0/2] Make [IP6]GRE[TAP] devices always NETIF_F_LLTX

From: Jakub Kicinski <kuba@kernel.org>
Date: 2022-04-25 18:03:08 

On Thu, 21 Apr 2022 15:43:42 -0700 Peilin Ye wrote:
This patchset depends on these fixes [1].  Since o_seqno is now atomic_t,
we can always turn on NETIF_F_LLTX for [IP6]GRE[TAP] devices, since we no
longer need the TX lock (&txq->_xmit_lock).
LGTM, thanks, but please repost on Thu / Fri. The fixes make their way
into net-next on Thu so until then we can't apply.

We could probably do the same thing to [IP6]ERSPAN devices as well, but
I'm not familiar with them yet.  For example, ERSPAN devices are
initialized as |= GRE_FEATURES in erspan_tunnel_init(), but I don't see
IP6ERSPAN devices being initialized as |= GRE6_FEATURES.  Please suggest
if I'm missing something, thanks!
Probably good to CC William when you repost.

Re: [PATCH net-next 0/2] Make [IP6]GRE[TAP] devices always NETIF_F_LLTX

From: Peilin Ye <hidden>
Date: 2022-04-25 19:05:59 

Hi Jakub,

On Mon, Apr 25, 2022 at 11:02:59AM -0700, Jakub Kicinski wrote:
On Thu, 21 Apr 2022 15:43:42 -0700 Peilin Ye wrote:
quoted
This patchset depends on these fixes [1].  Since o_seqno is now atomic_t,
we can always turn on NETIF_F_LLTX for [IP6]GRE[TAP] devices, since we no
longer need the TX lock (&txq->_xmit_lock).
LGTM, thanks, but please repost on Thu / Fri. The fixes make their way
into net-next on Thu so until then we can't apply.
Thanks for the review!
quoted
We could probably do the same thing to [IP6]ERSPAN devices as well, but
I'm not familiar with them yet.  For example, ERSPAN devices are
initialized as |= GRE_FEATURES in erspan_tunnel_init(), but I don't see
IP6ERSPAN devices being initialized as |= GRE6_FEATURES.  Please suggest
if I'm missing something, thanks!
Probably good to CC William when you repost.
Sure, I will resend then.

Thanks,
Peilin Ye
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help