[PATCH net 0/4] Fix broken backpressure during FDB dump in DSA drivers

[PATCH net 0/4] Fix broken backpressure during FDB dump in DSA drivers

[Vladimir Oltean]

rtnl_fdb_dump() has logic to split a dump of PF_BRIDGE neighbors into
multiple netlink skbs if the buffer provided by user space is too small
(one buffer will typically handle a few hundred FDB entries).

When the current buffer becomes full, nlmsg_put() in
dsa_slave_port_fdb_do_dump() returns -EMSGSIZE and DSA saves the index
of the last dumped FDB entry, returns to rtnl_fdb_dump() up to that
point, and then the dump resumes on the same port with a new skb, and
FDB entries up to the saved index are simply skipped.

Since dsa_slave_port_fdb_do_dump() is pointed to by the "cb" passed to
drivers, then drivers must check for the -EMSGSIZE error code returned
by it. Otherwise, when a netlink skb becomes full, DSA will no longer
save newly dumped FDB entries to it, but the driver will continue
dumping. So FDB entries will be missing from the dump.

DSA is one of the few switchdev drivers that have an .ndo_fdb_dump
implementation, because of the assumption that the hardware and software
FDBs cannot be efficiently kept in sync via SWITCHDEV_FDB_ADD_TO_BRIDGE.
Other drivers with a home-cooked .ndo_fdb_dump implementation are
ocelot and dpaa2-switch. These appear to do the correct thing, as do the
other DSA drivers, so nothing else appears to need fixing.

Vladimir Oltean (4):
  net: dsa: hellcreek: fix broken backpressure in .port_fdb_dump
  net: dsa: lan9303: fix broken backpressure in .port_fdb_dump
  net: dsa: lantiq: fix broken backpressure in .port_fdb_dump
  net: dsa: sja1105: fix broken backpressure in .port_fdb_dump

 drivers/net/dsa/hirschmann/hellcreek.c |  7 ++++--
 drivers/net/dsa/lan9303-core.c         | 34 ++++++++++++++------------
 drivers/net/dsa/lantiq_gswip.c         | 14 ++++++++---
 drivers/net/dsa/sja1105/sja1105_main.c |  4 ++-
 4 files changed, 37 insertions(+), 22 deletions(-)

-- 
2.25.1

[PATCH net 1/4] net: dsa: hellcreek: fix broken backpressure in .port_fdb_dump

[Vladimir Oltean]

rtnl_fdb_dump() has logic to split a dump of PF_BRIDGE neighbors into
multiple netlink skbs if the buffer provided by user space is too small
(one buffer will typically handle a few hundred FDB entries).

When the current buffer becomes full, nlmsg_put() in
dsa_slave_port_fdb_do_dump() returns -EMSGSIZE and DSA saves the index
of the last dumped FDB entry, returns to rtnl_fdb_dump() up to that
point, and then the dump resumes on the same port with a new skb, and
FDB entries up to the saved index are simply skipped.

Since dsa_slave_port_fdb_do_dump() is pointed to by the "cb" passed to
drivers, then drivers must check for the -EMSGSIZE error code returned
by it. Otherwise, when a netlink skb becomes full, DSA will no longer
save newly dumped FDB entries to it, but the driver will continue
dumping. So FDB entries will be missing from the dump.

Fix the broken backpressure by propagating the "cb" return code and
allow rtnl_fdb_dump() to restart the FDB dump with a new skb.

Fixes: e4b27ebc780f ("net: dsa: Add DSA driver for Hirschmann Hellcreek switches")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
---
 drivers/net/dsa/hirschmann/hellcreek.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/drivers/net/dsa/hirschmann/hellcreek.c b/drivers/net/dsa/hirschmann/hellcreek.c
index 9fdcc4bde480..5c54ae1be62c 100644
--- a/drivers/net/dsa/hirschmann/hellcreek.c
+++ b/drivers/net/dsa/hirschmann/hellcreek.c
@@ -912,6 +912,7 @@ static int hellcreek_fdb_dump(struct dsa_switch *ds, int port,
 {
 	struct hellcreek *hellcreek = ds->priv;
 	u16 entries;
+	int ret = 0;
 	size_t i;
 
 	mutex_lock(&hellcreek->reg_lock);
@@ -943,12 +944,14 @@ static int hellcreek_fdb_dump(struct dsa_switch *ds, int port,
 		if (!(entry.portmask & BIT(port)))
 			continue;
 
-		cb(entry.mac, 0, entry.is_static, data);
+		ret = cb(entry.mac, 0, entry.is_static, data);
+		if (ret)
+			break;
 	}
 
 	mutex_unlock(&hellcreek->reg_lock);
 
-	return 0;
+	return ret;
 }
 
 static int hellcreek_vlan_filtering(struct dsa_switch *ds, int port,
-- 
2.25.1

[PATCH net 2/4] net: dsa: lan9303: fix broken backpressure in .port_fdb_dump

[Vladimir Oltean]

rtnl_fdb_dump() has logic to split a dump of PF_BRIDGE neighbors into
multiple netlink skbs if the buffer provided by user space is too small
(one buffer will typically handle a few hundred FDB entries).

When the current buffer becomes full, nlmsg_put() in
dsa_slave_port_fdb_do_dump() returns -EMSGSIZE and DSA saves the index
of the last dumped FDB entry, returns to rtnl_fdb_dump() up to that
point, and then the dump resumes on the same port with a new skb, and
FDB entries up to the saved index are simply skipped.

Since dsa_slave_port_fdb_do_dump() is pointed to by the "cb" passed to
drivers, then drivers must check for the -EMSGSIZE error code returned
by it. Otherwise, when a netlink skb becomes full, DSA will no longer
save newly dumped FDB entries to it, but the driver will continue
dumping. So FDB entries will be missing from the dump.

Fix the broken backpressure by propagating the "cb" return code and
allow rtnl_fdb_dump() to restart the FDB dump with a new skb.

Fixes: ab335349b852 ("net: dsa: lan9303: Add port_fast_age and port_fdb_dump methods")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
---
 drivers/net/dsa/lan9303-core.c | 34 +++++++++++++++++++---------------
 1 file changed, 19 insertions(+), 15 deletions(-)

diff --git a/drivers/net/dsa/lan9303-core.c b/drivers/net/dsa/lan9303-core.c
index 344374025426..d7ce281570b5 100644
--- a/drivers/net/dsa/lan9303-core.c
+++ b/drivers/net/dsa/lan9303-core.c
@@ -557,12 +557,12 @@ static int lan9303_alr_make_entry_raw(struct lan9303 *chip, u32 dat0, u32 dat1)
 	return 0;
 }
 
-typedef void alr_loop_cb_t(struct lan9303 *chip, u32 dat0, u32 dat1,
-			   int portmap, void *ctx);
+typedef int alr_loop_cb_t(struct lan9303 *chip, u32 dat0, u32 dat1,
+			  int portmap, void *ctx);
 
-static void lan9303_alr_loop(struct lan9303 *chip, alr_loop_cb_t *cb, void *ctx)
+static int lan9303_alr_loop(struct lan9303 *chip, alr_loop_cb_t *cb, void *ctx)
 {
-	int i;
+	int ret = 0, i;
 
 	mutex_lock(&chip->alr_mutex);
 	lan9303_write_switch_reg(chip, LAN9303_SWE_ALR_CMD,
@@ -582,13 +582,17 @@ static void lan9303_alr_loop(struct lan9303 *chip, alr_loop_cb_t *cb, void *ctx)
 						LAN9303_ALR_DAT1_PORT_BITOFFS;
 		portmap = alrport_2_portmap[alrport];
 
-		cb(chip, dat0, dat1, portmap, ctx);
+		ret = cb(chip, dat0, dat1, portmap, ctx);
+		if (ret)
+			break;
 
 		lan9303_write_switch_reg(chip, LAN9303_SWE_ALR_CMD,
 					 LAN9303_ALR_CMD_GET_NEXT);
 		lan9303_write_switch_reg(chip, LAN9303_SWE_ALR_CMD, 0);
 	}
 	mutex_unlock(&chip->alr_mutex);
+
+	return ret;
 }
 
 static void alr_reg_to_mac(u32 dat0, u32 dat1, u8 mac[6])
@@ -606,18 +610,20 @@ struct del_port_learned_ctx {
 };
 
 /* Clear learned (non-static) entry on given port */
-static void alr_loop_cb_del_port_learned(struct lan9303 *chip, u32 dat0,
-					 u32 dat1, int portmap, void *ctx)
+static int alr_loop_cb_del_port_learned(struct lan9303 *chip, u32 dat0,
+					u32 dat1, int portmap, void *ctx)
 {
 	struct del_port_learned_ctx *del_ctx = ctx;
 	int port = del_ctx->port;
 
 	if (((BIT(port) & portmap) == 0) || (dat1 & LAN9303_ALR_DAT1_STATIC))
-		return;
+		return 0;
 
 	/* learned entries has only one port, we can just delete */
 	dat1 &= ~LAN9303_ALR_DAT1_VALID; /* delete entry */
 	lan9303_alr_make_entry_raw(chip, dat0, dat1);
+
+	return 0;
 }
 
 struct port_fdb_dump_ctx {
@@ -626,19 +632,19 @@ struct port_fdb_dump_ctx {
 	dsa_fdb_dump_cb_t *cb;
 };
 
-static void alr_loop_cb_fdb_port_dump(struct lan9303 *chip, u32 dat0,
-				      u32 dat1, int portmap, void *ctx)
+static int alr_loop_cb_fdb_port_dump(struct lan9303 *chip, u32 dat0,
+				     u32 dat1, int portmap, void *ctx)
 {
 	struct port_fdb_dump_ctx *dump_ctx = ctx;
 	u8 mac[ETH_ALEN];
 	bool is_static;
 
 	if ((BIT(dump_ctx->port) & portmap) == 0)
-		return;
+		return 0;
 
 	alr_reg_to_mac(dat0, dat1, mac);
 	is_static = !!(dat1 & LAN9303_ALR_DAT1_STATIC);
-	dump_ctx->cb(mac, 0, is_static, dump_ctx->data);
+	return dump_ctx->cb(mac, 0, is_static, dump_ctx->data);
 }
 
 /* Set a static ALR entry. Delete entry if port_map is zero */
@@ -1210,9 +1216,7 @@ static int lan9303_port_fdb_dump(struct dsa_switch *ds, int port,
 	};
 
 	dev_dbg(chip->dev, "%s(%d)\n", __func__, port);
-	lan9303_alr_loop(chip, alr_loop_cb_fdb_port_dump, &dump_ctx);
-
-	return 0;
+	return lan9303_alr_loop(chip, alr_loop_cb_fdb_port_dump, &dump_ctx);
 }
 
 static int lan9303_port_mdb_prepare(struct dsa_switch *ds, int port,
-- 
2.25.1

[PATCH net 3/4] net: dsa: lantiq: fix broken backpressure in .port_fdb_dump

[Vladimir Oltean]

rtnl_fdb_dump() has logic to split a dump of PF_BRIDGE neighbors into
multiple netlink skbs if the buffer provided by user space is too small
(one buffer will typically handle a few hundred FDB entries).

When the current buffer becomes full, nlmsg_put() in
dsa_slave_port_fdb_do_dump() returns -EMSGSIZE and DSA saves the index
of the last dumped FDB entry, returns to rtnl_fdb_dump() up to that
point, and then the dump resumes on the same port with a new skb, and
FDB entries up to the saved index are simply skipped.

Since dsa_slave_port_fdb_do_dump() is pointed to by the "cb" passed to
drivers, then drivers must check for the -EMSGSIZE error code returned
by it. Otherwise, when a netlink skb becomes full, DSA will no longer
save newly dumped FDB entries to it, but the driver will continue
dumping. So FDB entries will be missing from the dump.

Fix the broken backpressure by propagating the "cb" return code and
allow rtnl_fdb_dump() to restart the FDB dump with a new skb.

Fixes: 58c59ef9e930 ("net: dsa: lantiq: Add Forwarding Database access")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
---
 drivers/net/dsa/lantiq_gswip.c | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)

diff --git a/drivers/net/dsa/lantiq_gswip.c b/drivers/net/dsa/lantiq_gswip.c
index 314ae78bbdd6..e78026ef6d8c 100644
--- a/drivers/net/dsa/lantiq_gswip.c
+++ b/drivers/net/dsa/lantiq_gswip.c
@@ -1404,11 +1404,17 @@ static int gswip_port_fdb_dump(struct dsa_switch *ds, int port,
 		addr[1] = mac_bridge.key[2] & 0xff;
 		addr[0] = (mac_bridge.key[2] >> 8) & 0xff;
 		if (mac_bridge.val[1] & GSWIP_TABLE_MAC_BRIDGE_STATIC) {
-			if (mac_bridge.val[0] & BIT(port))
-				cb(addr, 0, true, data);
+			if (mac_bridge.val[0] & BIT(port)) {
+				err = cb(addr, 0, true, data);
+				if (err)
+					return err;
+			}
 		} else {
-			if (((mac_bridge.val[0] & GENMASK(7, 4)) >> 4) == port)
-				cb(addr, 0, false, data);
+			if (((mac_bridge.val[0] & GENMASK(7, 4)) >> 4) == port) {
+				err = cb(addr, 0, false, data);
+				if (err)
+					return err;
+			}
 		}
 	}
 	return 0;
-- 
2.25.1

[PATCH net 4/4] net: dsa: sja1105: fix broken backpressure in .port_fdb_dump

[Vladimir Oltean]

rtnl_fdb_dump() has logic to split a dump of PF_BRIDGE neighbors into
multiple netlink skbs if the buffer provided by user space is too small
(one buffer will typically handle a few hundred FDB entries).

When the current buffer becomes full, nlmsg_put() in
dsa_slave_port_fdb_do_dump() returns -EMSGSIZE and DSA saves the index
of the last dumped FDB entry, returns to rtnl_fdb_dump() up to that
point, and then the dump resumes on the same port with a new skb, and
FDB entries up to the saved index are simply skipped.

Since dsa_slave_port_fdb_do_dump() is pointed to by the "cb" passed to
drivers, then drivers must check for the -EMSGSIZE error code returned
by it. Otherwise, when a netlink skb becomes full, DSA will no longer
save newly dumped FDB entries to it, but the driver will continue
dumping. So FDB entries will be missing from the dump.

Fix the broken backpressure by propagating the "cb" return code and
allow rtnl_fdb_dump() to restart the FDB dump with a new skb.

Fixes: 291d1e72b756 ("net: dsa: sja1105: Add support for FDB and MDB management")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
---
 drivers/net/dsa/sja1105/sja1105_main.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/net/dsa/sja1105/sja1105_main.c b/drivers/net/dsa/sja1105/sja1105_main.c
index 8667c9754330..b188a80eeec6 100644
--- a/drivers/net/dsa/sja1105/sja1105_main.c
+++ b/drivers/net/dsa/sja1105/sja1105_main.c
@@ -1635,7 +1635,9 @@ static int sja1105_fdb_dump(struct dsa_switch *ds, int port,
 		/* We need to hide the dsa_8021q VLANs from the user. */
 		if (priv->vlan_state == SJA1105_VLAN_UNAWARE)
 			l2_lookup.vlanid = 0;
-		cb(macaddr, l2_lookup.vlanid, l2_lookup.lockeds, data);
+		rc = cb(macaddr, l2_lookup.vlanid, l2_lookup.lockeds, data);
+		if (rc)
+			return rc;
 	}
 	return 0;
 }
-- 
2.25.1

Re: [PATCH net 1/4] net: dsa: hellcreek: fix broken backpressure in .port_fdb_dump

[Kurt Kanzenbach]

[-- Attachment #1: Type: text/plain, Size: 1256 bytes --]

On Tue Aug 10 2021, Vladimir Oltean wrote:
> rtnl_fdb_dump() has logic to split a dump of PF_BRIDGE neighbors into
> multiple netlink skbs if the buffer provided by user space is too small
> (one buffer will typically handle a few hundred FDB entries).
>
> When the current buffer becomes full, nlmsg_put() in
> dsa_slave_port_fdb_do_dump() returns -EMSGSIZE and DSA saves the index
> of the last dumped FDB entry, returns to rtnl_fdb_dump() up to that
> point, and then the dump resumes on the same port with a new skb, and
> FDB entries up to the saved index are simply skipped.
>
> Since dsa_slave_port_fdb_do_dump() is pointed to by the "cb" passed to
> drivers, then drivers must check for the -EMSGSIZE error code returned
> by it. Otherwise, when a netlink skb becomes full, DSA will no longer
> save newly dumped FDB entries to it, but the driver will continue
> dumping. So FDB entries will be missing from the dump.
>
> Fix the broken backpressure by propagating the "cb" return code and
> allow rtnl_fdb_dump() to restart the FDB dump with a new skb.
>
> Fixes: e4b27ebc780f ("net: dsa: Add DSA driver for Hirschmann Hellcreek switches")
> Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>

Acked-by: Kurt Kanzenbach <kurt@linutronix.de>

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 861 bytes --]

Re: [PATCH net 0/4] Fix broken backpressure during FDB dump in DSA drivers

[patchwork-bot+netdevbpf]

Hello:

This series was applied to netdev/net.git (refs/heads/master):

On Tue, 10 Aug 2021 14:19:52 +0300 you wrote:
> rtnl_fdb_dump() has logic to split a dump of PF_BRIDGE neighbors into
> multiple netlink skbs if the buffer provided by user space is too small
> (one buffer will typically handle a few hundred FDB entries).
> 
> When the current buffer becomes full, nlmsg_put() in
> dsa_slave_port_fdb_do_dump() returns -EMSGSIZE and DSA saves the index
> of the last dumped FDB entry, returns to rtnl_fdb_dump() up to that
> point, and then the dump resumes on the same port with a new skb, and
> FDB entries up to the saved index are simply skipped.
> 
> [...]

Here is the summary with links:
  - [net,1/4] net: dsa: hellcreek: fix broken backpressure in .port_fdb_dump
    https://git.kernel.org/netdev/net/c/cd391280bf46
  - [net,2/4] net: dsa: lan9303: fix broken backpressure in .port_fdb_dump
    https://git.kernel.org/netdev/net/c/ada2fee185d8
  - [net,3/4] net: dsa: lantiq: fix broken backpressure in .port_fdb_dump
    https://git.kernel.org/netdev/net/c/871a73a1c8f5
  - [net,4/4] net: dsa: sja1105: fix broken backpressure in .port_fdb_dump
    https://git.kernel.org/netdev/net/c/21b52fed928e

You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html