Merge branch 'tso-gso-limit-split'

Jakub Kicinski says:

====================
net: disambiguate the TSO and GSO limits

This series separates the device-reported TSO limitations
from the user space-controlled GSO limits. It used to be that
we only had the former (HW limits) but they were named GSO.
This probably lead to confusion and letting user override them.

The problem came up in the BIG TCP discussion between Eric and
Alex, and seems like something we should address.

Targeting net-next because (a) nobody is reporting problems;
and (b) there is a tiny but non-zero chance that some actually
wants to lift the HW limitations.
====================

Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

3 files changed, 70 insertions, 2 deletions

diff --git a/net/core/dev.c b/net/core/dev.c
index c2d73595a7c3..f036ccb61da4 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -2993,6 +2993,51 @@ undo_rx:
 EXPORT_SYMBOL(netif_set_real_num_queues);
 
 /**
+ * netif_set_tso_max_size() - set the max size of TSO frames supported
+ * @dev:	netdev to update
+ * @size:	max skb->len of a TSO frame
+ *
+ * Set the limit on the size of TSO super-frames the device can handle.
+ * Unless explicitly set the stack will assume the value of %GSO_MAX_SIZE.
+ */
+void netif_set_tso_max_size(struct net_device *dev, unsigned int size)
+{
+	dev->tso_max_size = size;
+	if (size < READ_ONCE(dev->gso_max_size))
+		netif_set_gso_max_size(dev, size);
+}
+EXPORT_SYMBOL(netif_set_tso_max_size);
+
+/**
+ * netif_set_tso_max_segs() - set the max number of segs supported for TSO
+ * @dev:	netdev to update
+ * @segs:	max number of TCP segments
+ *
+ * Set the limit on the number of TCP segments the device can generate from
+ * a single TSO super-frame.
+ * Unless explicitly set the stack will assume the value of %GSO_MAX_SEGS.
+ */
+void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs)
+{
+	dev->tso_max_segs = segs;
+	if (segs < READ_ONCE(dev->gso_max_segs))
+		netif_set_gso_max_segs(dev, segs);
+}
+EXPORT_SYMBOL(netif_set_tso_max_segs);
+
+/**
+ * netif_inherit_tso_max() - copy all TSO limits from a lower device to an upper
+ * @to:		netdev to update
+ * @from:	netdev from which to copy the limits
+ */
+void netif_inherit_tso_max(struct net_device *to, const struct net_device *from)
+{
+	netif_set_tso_max_size(to, from->tso_max_size);
+	netif_set_tso_max_segs(to, from->tso_max_segs);
+}
+EXPORT_SYMBOL(netif_inherit_tso_max);
+
+/**
  * netif_get_num_default_rss_queues - default number of RSS queues
  *
  * Default value is the number of physical cores if there are only 1 or 2, or
@@ -10560,6 +10605,8 @@ struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
 	dev->gso_max_size = GSO_MAX_SIZE;
 	dev->gso_max_segs = GSO_MAX_SEGS;
 	dev->gro_max_size = GRO_MAX_SIZE;
+	dev->tso_max_size = TSO_LEGACY_MAX_SIZE;
+	dev->tso_max_segs = TSO_MAX_SEGS;
 	dev->upper_level = 1;
 	dev->lower_level = 1;
 #ifdef CONFIG_LOCKDEP
diff --git a/net/core/dev.h b/net/core/dev.h
index 27923df00637..328b37af90ba 100644
--- a/net/core/dev.h
+++ b/net/core/dev.h
@@ -88,4 +88,25 @@ int dev_change_carrier(struct net_device *dev, bool new_carrier);
 
 void __dev_set_rx_mode(struct net_device *dev);
 
+static inline void netif_set_gso_max_size(struct net_device *dev,
+					  unsigned int size)
+{
+	/* dev->gso_max_size is read locklessly from sk_setup_caps() */
+	WRITE_ONCE(dev->gso_max_size, size);
+}
+
+static inline void netif_set_gso_max_segs(struct net_device *dev,
+					  unsigned int segs)
+{
+	/* dev->gso_max_segs is read locklessly from sk_setup_caps() */
+	WRITE_ONCE(dev->gso_max_segs, segs);
+}
+
+static inline void netif_set_gro_max_size(struct net_device *dev,
+					  unsigned int size)
+{
+	/* This pairs with the READ_ONCE() in skb_gro_receive() */
+	WRITE_ONCE(dev->gro_max_size, size);
+}
+
 #endif
diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
index eea5ed09e1bb..e6d4b9272995 100644
--- a/net/core/rtnetlink.c
+++ b/net/core/rtnetlink.c
@@ -2803,7 +2803,7 @@ static int do_setlink(const struct sk_buff *skb,
 	if (tb[IFLA_GSO_MAX_SIZE]) {
 		u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
 
-		if (max_size > GSO_MAX_SIZE) {
+		if (max_size > GSO_MAX_SIZE || max_size > dev->tso_max_size) {
 			err = -EINVAL;
 			goto errout;
 		}
@@ -2817,7 +2817,7 @@ static int do_setlink(const struct sk_buff *skb,
 	if (tb[IFLA_GSO_MAX_SEGS]) {
 		u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
 
-		if (max_segs > GSO_MAX_SEGS) {
+		if (max_segs > GSO_MAX_SEGS || max_segs > dev->tso_max_segs) {
 			err = -EINVAL;
 			goto errout;
 		}