Skip to contentSkip to conversation

macsec: add support for IFLA_MACSEC_OFFLOAD in macsec_changelink

PATCH Superseded [email protected] 4 participants 9 messages Dec 8, 2022 – Dec 9, 2022

Subsystems (3)

v4 of 9 Superseded Latest is v9 →

lore
Share & tools
lore
#0 lore raw

PATCH v4 1/2 macsec: add support for IFLA_MACSEC_OFFLOAD in macsec_changelink

From: Emeel Hakim <[email protected]>

Add support for changing Macsec offload selection through the netlink layer by implementing the relevant changes in macsec_changelink.

Since the handling in macsec_changelink is similar to macsec_upd_offload, update macsec_upd_offload to use a common helper function to avoid duplication.

Example for setting offload for a macsec device:
ip link set macsec0 type macsec offload mac

---
V3 -> V4: - Dont pass whole attributes data to macsec_update_offload, just pass relevant attribute.
- Fix code style.
- Remove macsec_changelink_upd_offload
V2 -> V3: - Split the original patch into 3 patches, the macsec_rtnl_policy related change (separate patch)
to be sent to "net" branch as a fix.
- Change the original patch title to make it clear that it's only adding IFLA_MACSEC_OFFLOAD to changelink
V1 -> V2: Add common helper to avoid duplicating code

 drivers/net/macsec.c | 100 +++++++++++++++++++++++--------------------
 1 file changed, 53 insertions(+), 47 deletions(-)
drivers/net/macsec.c +53 -47
index d73b9d535b7a..abfe4a612a2d 100644
@@ -2583,38 +2583,16 @@ static bool macsec_is_configured(struct macsec_dev *macsec)
 	return false;
 }
 
-static int macsec_upd_offload(struct sk_buff *skb, struct genl_info *info)
+static int macsec_update_offload(struct net_device *dev, enum macsec_offload offload)
 {
-	struct nlattr *tb_offload[MACSEC_OFFLOAD_ATTR_MAX + 1];
-	enum macsec_offload offload, prev_offload;
-	int (*func)(struct macsec_context *ctx);
-	struct nlattr **attrs = info->attrs;
-	struct net_device *dev;
+	enum macsec_offload prev_offload;
 	const struct macsec_ops *ops;
 	struct macsec_context ctx;
 	struct macsec_dev *macsec;
-	int ret;
-
-	if (!attrs[MACSEC_ATTR_IFINDEX])
-		return -EINVAL;
-
-	if (!attrs[MACSEC_ATTR_OFFLOAD])
-		return -EINVAL;
-
-	if (nla_parse_nested_deprecated(tb_offload, MACSEC_OFFLOAD_ATTR_MAX,
-					attrs[MACSEC_ATTR_OFFLOAD],
-					macsec_genl_offload_policy, NULL))
-		return -EINVAL;
+	int ret = 0;
 
-	dev = get_dev_from_nl(genl_info_net(info), attrs);
-	if (IS_ERR(dev))
-		return PTR_ERR(dev);
 	macsec = macsec_priv(dev);
 
-	if (!tb_offload[MACSEC_OFFLOAD_ATTR_TYPE])
-		return -EINVAL;
-
-	offload = nla_get_u8(tb_offload[MACSEC_OFFLOAD_ATTR_TYPE]);
 	if (macsec->offload == offload)
 		return 0;
 
@@ -2627,43 +2605,65 @@ static int macsec_upd_offload(struct sk_buff *skb, struct genl_info *info)
 	if (netif_running(dev))
 		return -EBUSY;
 
-	rtnl_lock();
-
-	prev_offload = macsec->offload;
-	macsec->offload = offload;
-
 	/* Check if the device already has rules configured: we do not support
 	 * rules migration.
 	 */
-	if (macsec_is_configured(macsec)) {
-		ret = -EBUSY;
-		goto rollback;
-	}
+	if (macsec_is_configured(macsec))
+		return -EBUSY;
+
+	prev_offload = macsec->offload;
 
 	ops = __macsec_get_ops(offload == MACSEC_OFFLOAD_OFF ? prev_offload : offload,
 			       macsec, &ctx);
-	if (!ops) {
+	if (!ops)
 		ret = -EOPNOTSUPP;
-		goto rollback;
-	}
 
-	if (prev_offload == MACSEC_OFFLOAD_OFF)
-		func = ops->mdo_add_secy;
-	else
-		func = ops->mdo_del_secy;
+	macsec->offload = offload;
 
 	ctx.secy = &macsec->secy;
-	ret = macsec_offload(func, &ctx);
+	ret = offload == MACSEC_OFFLOAD_OFF ? macsec_offload(ops->mdo_del_secy, &ctx)
+					    : macsec_offload(ops->mdo_add_secy, &ctx);
+
 	if (ret)
-		goto rollback;
+		macsec->offload = prev_offload;
 
-	rtnl_unlock();
-	return 0;
+	return ret;
+}
+
+static int macsec_upd_offload(struct sk_buff *skb, struct genl_info *info)
+{
+	struct nlattr *tb_offload[MACSEC_OFFLOAD_ATTR_MAX + 1];
+	struct nlattr **attrs = info->attrs;
+	enum macsec_offload offload;
+	struct net_device *dev;
+	int ret;
+
+	if (!attrs[MACSEC_ATTR_IFINDEX])
+		return -EINVAL;
+
+	if (!attrs[MACSEC_ATTR_OFFLOAD])
+		return -EINVAL;
+
+	if (nla_parse_nested_deprecated(tb_offload, MACSEC_OFFLOAD_ATTR_MAX,
+					attrs[MACSEC_ATTR_OFFLOAD],
+					macsec_genl_offload_policy, NULL))
+		return -EINVAL;
+
+	dev = get_dev_from_nl(genl_info_net(info), attrs);
+	if (IS_ERR(dev))
+		return PTR_ERR(dev);
 
-rollback:
-	macsec->offload = prev_offload;
+	if (!tb_offload[MACSEC_OFFLOAD_ATTR_TYPE])
+		return -EINVAL;
+
+	offload = nla_get_u8(tb_offload[MACSEC_OFFLOAD_ATTR_TYPE]);
+
+	rtnl_lock();
+
+	ret = macsec_update_offload(dev, offload);
 
 	rtnl_unlock();
+
 	return ret;
 }
 
@@ -3831,6 +3831,12 @@ static int macsec_changelink(struct net_device *dev, struct nlattr *tb[],
 	if (ret)
 		goto cleanup;
 
+	if (data[IFLA_MACSEC_OFFLOAD]) {
+		ret = macsec_update_offload(dev, nla_get_u8(data[IFLA_MACSEC_OFFLOAD]));
+		if (ret)
+			goto cleanup;
+	}
+
 	/* If h/w offloading is available, propagate to the device */
 	if (macsec_is_offloaded(macsec)) {
 		const struct macsec_ops *ops;
Signature · 1 line
2.21.3
#1 lore raw

PATCH v2 2/2 macsec: dump IFLA_MACSEC_OFFLOAD attribute as part of macsec dump

From: Emeel Hakim <[email protected]>

Support dumping offload netlink attribute in macsec's device attributes dump.
Change macsec_get_size to consider the offload attribute in the calculations of the required room for dumping the device netlink attributes.

---
V1 -> V2: Update commit message

 drivers/net/macsec.c | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)
drivers/net/macsec.c +9 -2
index abfe4a612a2d..85208a64f259 100644
@@ -4237,16 +4237,22 @@ static size_t macsec_get_size(const struct net_device *dev)
 		nla_total_size(1) + /* IFLA_MACSEC_SCB */
 		nla_total_size(1) + /* IFLA_MACSEC_REPLAY_PROTECT */
 		nla_total_size(1) + /* IFLA_MACSEC_VALIDATION */
+		nla_total_size(1) + /* IFLA_MACSEC_OFFLOAD */
 		0;
 }
 
 static int macsec_fill_info(struct sk_buff *skb,
 			    const struct net_device *dev)
 {
-	struct macsec_secy *secy = &macsec_priv(dev)->secy;
-	struct macsec_tx_sc *tx_sc = &secy->tx_sc;
+	struct macsec_tx_sc *tx_sc;
+	struct macsec_dev *macsec;
+	struct macsec_secy *secy;
 	u64 csid;
 
+	macsec = macsec_priv(dev);
+	secy = &macsec->secy;
+	tx_sc = &secy->tx_sc;
+
 	switch (secy->key_len) {
 	case MACSEC_GCM_AES_128_SAK_LEN:
 		csid = secy->xpn ? MACSEC_CIPHER_ID_GCM_AES_XPN_128 : MACSEC_DEFAULT_CIPHER_ID;
@@ -4271,6 +4277,7 @@ static int macsec_fill_info(struct sk_buff *skb,
 	    nla_put_u8(skb, IFLA_MACSEC_SCB, tx_sc->scb) ||
 	    nla_put_u8(skb, IFLA_MACSEC_REPLAY_PROTECT, secy->replay_protect) ||
 	    nla_put_u8(skb, IFLA_MACSEC_VALIDATION, secy->validate_frames) ||
+	    nla_put_u8(skb, IFLA_MACSEC_OFFLOAD, macsec->offload) ||
 	    0)
 		goto nla_put_failure;
 
Signature · 1 line
2.21.3
Reviewed-by on v2 2/2
#2 lore raw

Re: [PATCH net-next v2 2/2] macsec: dump IFLA_MACSEC_OFFLOAD attribute as part of macsec dump

Quoted 10 lines · Thu, Dec 08, 2022 at 12:55:17PM CET, [email protected] wrote:

From: Emeel Hakim <[email protected]>

Support dumping offload netlink attribute in macsec's device attributes dump.
Change macsec_get_size to consider the offload attribute in the calculations of the required room for dumping the device netlink attributes.

Reviewed-by on v4 1/2
#3 lore raw
In reply to [email protected]
Quoted 10 lines · Thu, Dec 08, 2022 at 12:55:16PM CET, [email protected] wrote:

From: Emeel Hakim <[email protected]>

Add support for changing Macsec offload selection through the netlink layer by implementing the relevant changes in macsec_changelink.

Since the handling in macsec_changelink is similar to macsec_upd_offload, update macsec_upd_offload to use a common helper function to avoid duplication.

Quoted 5 lines

Example for setting offload for a macsec device:
ip link set macsec0 type macsec offload mac

Looks fine to me.

maintainer
#4 lore raw
In reply to [email protected]
Quoted 90 lines · 2022-12-08, 13:55:16 +0200, [email protected] wrote:

From: Emeel Hakim <[email protected]>

Add support for changing Macsec offload selection through the netlink layer by implementing the relevant changes in macsec_changelink.

Since the handling in macsec_changelink is similar to macsec_upd_offload, update macsec_upd_offload to use a common helper function to avoid duplication.

Example for setting offload for a macsec device:
ip link set macsec0 type macsec offload mac

---
V3 -> V4: - Dont pass whole attributes data to macsec_update_offload, just pass relevant attribute.
- Fix code style.
- Remove macsec_changelink_upd_offload
V2 -> V3: - Split the original patch into 3 patches, the macsec_rtnl_policy related change (separate patch)
to be sent to "net" branch as a fix.
- Change the original patch title to make it clear that it's only adding IFLA_MACSEC_OFFLOAD to changelink
V1 -> V2: Add common helper to avoid duplicating code

 drivers/net/macsec.c | 100 +++++++++++++++++++++++--------------------
 1 file changed, 53 insertions(+), 47 deletions(-)
drivers/net/macsec.c +3 -34
index d73b9d535b7a..abfe4a612a2d 100644
@@ -2583,38 +2583,16 @@ static bool macsec_is_configured(struct macsec_dev *macsec)
 	return false;
 }
 
-static int macsec_upd_offload(struct sk_buff *skb, struct genl_info *info)
+static int macsec_update_offload(struct net_device *dev, enum macsec_offload offload)
 {
-	struct nlattr *tb_offload[MACSEC_OFFLOAD_ATTR_MAX + 1];
-	enum macsec_offload offload, prev_offload;
-	int (*func)(struct macsec_context *ctx);
-	struct nlattr **attrs = info->attrs;
-	struct net_device *dev;
+	enum macsec_offload prev_offload;
 	const struct macsec_ops *ops;
 	struct macsec_context ctx;
 	struct macsec_dev *macsec;
-	int ret;
-
-	if (!attrs[MACSEC_ATTR_IFINDEX])
-		return -EINVAL;
-
-	if (!attrs[MACSEC_ATTR_OFFLOAD])
-		return -EINVAL;
-
-	if (nla_parse_nested_deprecated(tb_offload, MACSEC_OFFLOAD_ATTR_MAX,
-					attrs[MACSEC_ATTR_OFFLOAD],
-					macsec_genl_offload_policy, NULL))
-		return -EINVAL;
+	int ret = 0;
 
-	dev = get_dev_from_nl(genl_info_net(info), attrs);
-	if (IS_ERR(dev))
-		return PTR_ERR(dev);
 	macsec = macsec_priv(dev);
 
-	if (!tb_offload[MACSEC_OFFLOAD_ATTR_TYPE])
-		return -EINVAL;
-
-	offload = nla_get_u8(tb_offload[MACSEC_OFFLOAD_ATTR_TYPE]);
 	if (macsec->offload == offload)
 		return 0;
 
@@ -2627,43 +2605,65 @@ static int macsec_upd_offload(struct sk_buff *skb, struct genl_info *info)
 	if (netif_running(dev))
 		return -EBUSY;
 
-	rtnl_lock();
-
-	prev_offload = macsec->offload;
-	macsec->offload = offload;
-
 	/* Check if the device already has rules configured: we do not support
 	 * rules migration.
 	 */
-	if (macsec_is_configured(macsec)) {
-		ret = -EBUSY;
-		goto rollback;
-	}

drivers/net/macsec.c static int macsec_upd_offload(struct sk_buff *skb, struct genl_info *info) · line 2619

Quoted 10 lines
drivers/net/macsec.c +5 -1
+	if (macsec_is_configured(macsec))
+		return -EBUSY;
+
+	prev_offload = macsec->offload;
 
 	ops = __macsec_get_ops(offload == MACSEC_OFFLOAD_OFF ? prev_offload : offload,
 			       macsec, &ctx);
-	if (!ops) {
+	if (!ops)
 		ret = -EOPNOTSUPP;
		return -EOPNOTSUPP;

It's probably not a problem because macsec_check_offload should catch that, but it looks wrong.

Quoted 4 lines
drivers/net/macsec.c +0 -3
-		goto rollback;
-	}
 
-	if (prev_offload == MACSEC_OFFLOAD_OFF)

drivers/net/macsec.c static int macsec_upd_offload(struct sk_buff *skb, struct genl_info *info) · line 2626

Quoted 10 lines
drivers/net/macsec.c +4 -4
-		func = ops->mdo_add_secy;
-	else
-		func = ops->mdo_del_secy;
+	macsec->offload = offload;
 
 	ctx.secy = &macsec->secy;
-	ret = macsec_offload(func, &ctx);
+	ret = offload == MACSEC_OFFLOAD_OFF ? macsec_offload(ops->mdo_del_secy, &ctx)
+					    : macsec_offload(ops->mdo_add_secy, &ctx);
+

nit: unnecessary blank line (sorry, I should have spotted that earlier)

Quoted 61 lines
drivers/net/macsec.c +41 -5
 	if (ret)
-		goto rollback;
+		macsec->offload = prev_offload;
 
-	rtnl_unlock();
-	return 0;
+	return ret;
+}
+
+static int macsec_upd_offload(struct sk_buff *skb, struct genl_info *info)
+{
+	struct nlattr *tb_offload[MACSEC_OFFLOAD_ATTR_MAX + 1];
+	struct nlattr **attrs = info->attrs;
+	enum macsec_offload offload;
+	struct net_device *dev;
+	int ret;
+
+	if (!attrs[MACSEC_ATTR_IFINDEX])
+		return -EINVAL;
+
+	if (!attrs[MACSEC_ATTR_OFFLOAD])
+		return -EINVAL;
+
+	if (nla_parse_nested_deprecated(tb_offload, MACSEC_OFFLOAD_ATTR_MAX,
+					attrs[MACSEC_ATTR_OFFLOAD],
+					macsec_genl_offload_policy, NULL))
+		return -EINVAL;
+
+	dev = get_dev_from_nl(genl_info_net(info), attrs);
+	if (IS_ERR(dev))
+		return PTR_ERR(dev);
 
-rollback:
-	macsec->offload = prev_offload;
+	if (!tb_offload[MACSEC_OFFLOAD_ATTR_TYPE])
+		return -EINVAL;
+
+	offload = nla_get_u8(tb_offload[MACSEC_OFFLOAD_ATTR_TYPE]);
+
+	rtnl_lock();
+
+	ret = macsec_update_offload(dev, offload);
 
 	rtnl_unlock();
+
 	return ret;
 }
 
@@ -3831,6 +3831,12 @@ static int macsec_changelink(struct net_device *dev, struct nlattr *tb[],
 	if (ret)
 		goto cleanup;
 
+	if (data[IFLA_MACSEC_OFFLOAD]) {
+		ret = macsec_update_offload(dev, nla_get_u8(data[IFLA_MACSEC_OFFLOAD]));
+		if (ret)
+			goto cleanup;
+	}
+
 	/* If h/w offloading is available, propagate to the device */
 	if (macsec_is_offloaded(macsec)) {
 		const struct macsec_ops *ops;
Signature · 1 line
Sabrina
maintainer
#5 lore raw
In reply to [email protected]

drivers/net/macsec.c static int macsec_upd_offload(struct sk_buff *skb, struct genl_info *info) · line 2652

Quoted 1 line · On Thu, 8 Dec 2022 13:55:16 +0200 [email protected] wrote:
drivers/net/macsec.c +1 -0
+	dev = get_dev_from_nl(genl_info_net(info), attrs);

What prevents this dev from disappearing before...

drivers/net/macsec.c static int macsec_upd_offload(struct sk_buff *skb, struct genl_info *info) · line 2661

Quoted 8 lines
drivers/net/macsec.c +8 -0
+	if (IS_ERR(dev))
+		return PTR_ERR(dev);
+	if (!tb_offload[MACSEC_OFFLOAD_ATTR_TYPE])
+		return -EINVAL;
+
+	offload = nla_get_u8(tb_offload[MACSEC_OFFLOAD_ATTR_TYPE]);
+
+	rtnl_lock();

... we finally take the lock?

I think you're just moving this code, but still.

Quoted 3 lines
drivers/net/macsec.c +1 -0
+	ret = macsec_update_offload(dev, offload);
 
 	rtnl_unlock();
maintainer
#6 lore raw
Quoted 1 line · On Thu, 8 Dec 2022 18:32:44 -0800 Jakub Kicinski wrote:

I think you're just moving this code, but still.

And by "by still" I mean - it's still a bug, so it needs to be fixed first.

#7 lore raw
Quoted 12 lines

-----Original Message-----
From: Jakub Kicinski <[email protected]>
Sent: Friday, 9 December 2022 4:33
To: Emeel Hakim <[email protected]>
Cc: [email protected]; Raed Salem <[email protected]>; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]
Subject: Re: [PATCH net-next v4 1/2] macsec: add support for IFLA_MACSEC_OFFLOAD in macsec_changelink

Quoted 7 lines

External email: Use caution opening links or attachments

Quoted 1 line · On Thu, 8 Dec 2022 18:32:44 -0800 Jakub Kicinski wrote:

I think you're just moving this code, but still.

And by "by still" I mean - it's still a bug, so it needs to be fixed first.

The code added by those patches does not use the rtnl_lock, the lock is just getting moved as part of sharing similar code, but the new code is still not using it, I don’t think those patches need to wait until a fix of an existing locking issue as long as the new code is not inserting any bugs.

#8 lore raw
Quoted 29 lines

-----Original Message-----
From: Emeel Hakim
Sent: Friday, 9 December 2022 12:19
To: Jakub Kicinski <[email protected]>
Cc: [email protected]; Raed Salem <[email protected]>; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]
Subject: RE: [PATCH net-next v4 1/2] macsec: add support for IFLA_MACSEC_OFFLOAD in macsec_changelink

Quoted 15 lines

-----Original Message-----
From: Jakub Kicinski <[email protected]>
Sent: Friday, 9 December 2022 4:33
To: Emeel Hakim <[email protected]>
Cc: [email protected]; Raed Salem <[email protected]>; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]
Subject: Re: [PATCH net-next v4 1/2] macsec: add support for IFLA_MACSEC_OFFLOAD in macsec_changelink

External email: Use caution opening links or attachments

On Thu, 8 Dec 2022 18:32:44 -0800 Jakub Kicinski wrote:

Quoted 8 lines
Quoted 3 lines
Quoted 1 line

I think you're just moving this code, but still.

And by "by still" I mean - it's still a bug, so it needs to be fixed first.

The code added by those patches does not use the rtnl_lock, the lock is just getting moved as part of sharing similar code, but the new code is still not using it, I don’t think those patches need to wait until a fix of an existing locking issue as long as the new code is not inserting any bugs.

I will send a fix patch shortly as requested.