Support MACsec VLAN
Subsystems (6)
v2 cover Support MACsec VLAN
Dear maintainers,
This patch series introduces support for hardware (HW) offload MACsec devices with VLAN configuration. The patches address both scenarios where the VLAN header is both the inner and outer header for MACsec.
The changes include:
1. Adding MACsec offload operation for VLAN.
2. Considering VLAN when accessing MACsec net device.
3. Currently offloading MACsec when it's configured over VLAN with current MACsec TX steering rules would wrongly insert the MACsec sec tag after inserting the VLAN header. This resulted in an ETHERNET | SECTAG | VLAN packet when ETHERNET | VLAN | SECTAG is configured. The patche handles this issue when configuring steering rules.
4. Adding MACsec rx_handler change support in case of a marked skb and a mismatch on the dst MAC address.
Please review these changes and let me know if you have any feedback or concerns.
Updates since v1:
- Consult vlan_features when adding NETIF_F_HW_MACSEC.
- Allow grep for the functions.
- Add helper function to get the macsec operation to allow the compiler to make some choice.
Thanks,
Emeel
Emeel Hakim (4):
vlan: Add MACsec offload operations for VLAN interface net/mlx5: Support MACsec over VLAN
net/mlx5: Consider VLAN interface in MACsec TX steering rules
macsec: Add MACsec rx_handler change support
.../mellanox/mlx5/core/en_accel/macsec.c | 42 +++++--- .../mellanox/mlx5/core/en_accel/macsec_fs.c | 7 ++ .../net/ethernet/mellanox/mlx5/core/en_main.c | 1 + drivers/net/macsec.c | 9 ++ net/8021q/vlan_dev.c | 101 ++++++++++++++++++ 5 files changed, 144 insertions(+), 16 deletions(-)
Signature · 1 line
2.21.3
PATCH v2 1/4 vlan: Add MACsec offload operations for VLAN interface
Add support for MACsec offload operations for VLAN driver to allow offloading MACsec when VLAN's real device supports Macsec offload by forwarding the offload request to it.
---
V1 -> V2: - Consult vlan_features when adding NETIF_F_HW_MACSEC.
- Allow grep for the functions.
- Add helper function to get the macsec operation to allow the compiler to make some choice.
.../net/ethernet/mellanox/mlx5/core/en_main.c | 1 + net/8021q/vlan_dev.c | 101 ++++++++++++++++++ 2 files changed, 102 insertions(+)
drivers/net/ethernet/mellanox/mlx5/core/en_main.c +1 -0
@@ -5076,6 +5076,7 @@ static void mlx5e_build_nic_netdev(struct net_device *netdev) netdev->vlan_features |=NETIF_F_SG ; netdev->vlan_features |=NETIF_F_HW_CSUM ; + netdev->vlan_features |=NETIF_F_HW_MACSEC ; netdev->vlan_features |=NETIF_F_GRO ; netdev->vlan_features |=NETIF_F_TSO ; netdev->vlan_features |=NETIF_F_TSO6 ;
net/8021q/vlan_dev.c +101 -0
125 lines not on this page · raw message
Signature · 1 line
2.21.3
Sabrina Dubroca
maintainerRe: [PATCH net-next v2 1/4] vlan: Add MACsec offload operations for VLAN interface
Quoted 7 lines · Emeel Hakim
Add support for MACsec offload operations for VLAN driver to allow offloading MACsec when VLAN's real device supports Macsec offload by forwarding the offload request to it.
Signed-off-by Emeel Hakim <[email protected]>---
V1 -> V2: - Consultvlan_featureswhen addingNETIF_F_HW_MACSEC.
Uh? You're not actually doing that? You also dropped the changes to vlan_dev_fix_features without explaining why.
[...]
Quoted 11 lines
net/8021q/vlan_dev.c +3
-0@@ -572,6 +573,9 @@ static int vlan_dev_init(struct net_device *dev)NETIF_F_HIGHDMA |NETIF_F_SCTP_CRC |NETIF_F_ALL_FCOE ; +if (real_dev->features &NETIF_F_HW_MACSEC ) + dev->hw_features |=NETIF_F_HW_MACSEC ; + dev->features |= dev->hw_features |NETIF_F_LLTX ;netif_inherit_tso_max (dev, real_dev);if (dev->features &NETIF_F_VLAN_FEATURES )
net/8021q/vlan_dev.c static int vlan_dev_fill_forward_path(struct net_device_path_ctx *ctx, · line 812
Quoted 6 lines
net/8021q/vlan_dev.c +3
-0@@ -803,6 +807,100 @@ static int vlan_dev_fill_forward_path(struct net_device_path_ctx *ctx,return 0 ; } +#if IS_ENABLED (CONFIG_MACSEC ) + +static const struct macsec_ops *vlan_get_macsec_ops (struct macsec_context *ctx)
^ const?
net/8021q/vlan_dev.c static int vlan_dev_fill_forward_path(struct net_device_path_ctx *ctx, · line 824
Quoted 12 lines
net/8021q/vlan_dev.c +12
-0+{ +return vlan_dev_priv (ctx->netdev)->real_dev->macsec_ops; +} + +#define _BUILD_VLAN_MACSEC_MDO (mdo) \ + const struct macsec_ops *ops; \ + ops = vlan_get_macsec_ops(ctx); \ + return ops ? ops->mdo_ ## mdo(ctx) : -EOPNOTSUPP + +static int vlan_macsec_add_txsa (struct macsec_context *ctx) +{ +_BUILD_VLAN_MACSEC_MDO (add_txsa);
Shouldn't those be indented?
Quoted 2 lines
(quoted diff) +2
-0+} +
[...]
net/8021q/vlan_dev.c static int vlan_dev_fill_forward_path(struct net_device_path_ctx *ctx, · line 887
Quoted 5 lines
net/8021q/vlan_dev.c +5
-0+ +#define VLAN_MACSEC_DECLARE_MDO (mdo) .mdo_ ## mdo = vlan_macsec_ ## mdo + +static const struct macsec_ops macsec_offload_ops = { +VLAN_MACSEC_DECLARE_MDO (add_txsa),
This completely breaks the ability to use cscope when looking for implementations of mdo_add_txsa. I'm not very fond of the c/p, but I don't think we should be using macros at all here. At least to me, being able to navigate directly from mdo_add_txsa to its implementation without expanding the macro manually is important.
So, IMHO, those should be:
.mdo_add_txsa = vlan_macsec_add_txsa,
(etc)
Signature · 1 line
Sabrina
Re: [PATCH net-next v2 1/4] vlan: Add MACsec offload operations for VLAN interface
Quoted 11 lines · Sabrina Dubroca
Quoted 7 lines · Emeel Hakim
Add support for MACsec offload operations for VLAN driver to allow offloading MACsec when VLAN's real device supports Macsec offload by forwarding the offload request to it.
Signed-off-by Emeel Hakim <[email protected]>---
V1 -> V2: - Consultvlan_featureswhen addingNETIF_F_HW_MACSEC.Uh? You're not actually doing that? You also dropped the changes to
vlan_dev_fix_featureswithout explaining why.
vlan_dev_fix_features() relies on real_dev->vlan_features which was set in mlx5 part of this patch.
643 static netdev_features_t vlan_dev_fix_features(struct net_device *dev,
644 netdev_features_t features)
645 {
...
649
650 lower_features = netdev_intersect_features((real_dev->vlan_features |
651 NETIF_F_RXCSUM),
652 real_dev->features);
This part ensure that once real_dev->vlan_features and real_dev->features have NETIF_F_HW_MACSEC, the returned features will include NETIF_F_HW_MACSEC too.
Quoted 52 lines
[...]
Quoted 11 lines
net/8021q/vlan_dev.c +3
-0@@ -572,6 +573,9 @@ static int vlan_dev_init(struct net_device *dev)NETIF_F_HIGHDMA |NETIF_F_SCTP_CRC |NETIF_F_ALL_FCOE ; +if (real_dev->features &NETIF_F_HW_MACSEC ) + dev->hw_features |=NETIF_F_HW_MACSEC ; + dev->features |= dev->hw_features |NETIF_F_LLTX ;netif_inherit_tso_max (dev, real_dev);if (dev->features &NETIF_F_VLAN_FEATURES )net/8021q/vlan_dev.c static int vlan_dev_fill_forward_path(struct net_device_path_ctx *ctx, · line 812
Quoted 6 lines
net/8021q/vlan_dev.c +3
-0@@ -803,6 +807,100 @@ static int vlan_dev_fill_forward_path(struct net_device_path_ctx *ctx,return 0 ; } +#if IS_ENABLED (CONFIG_MACSEC ) + +static const struct macsec_ops *vlan_get_macsec_ops (struct macsec_context *ctx)^ const?
net/8021q/vlan_dev.c static int vlan_dev_fill_forward_path(struct net_device_path_ctx *ctx, · line 824
Quoted 12 lines
net/8021q/vlan_dev.c +12
-0+{ +return vlan_dev_priv (ctx->netdev)->real_dev->macsec_ops; +} + +#define _BUILD_VLAN_MACSEC_MDO (mdo) \ + const struct macsec_ops *ops; \ + ops = vlan_get_macsec_ops(ctx); \ + return ops ? ops->mdo_ ## mdo(ctx) : -EOPNOTSUPP + +static int vlan_macsec_add_txsa (struct macsec_context *ctx) +{ +_BUILD_VLAN_MACSEC_MDO (add_txsa);Shouldn't those be indented?
Quoted 2 lines
(quoted diff) +2
-0+} +[...]
net/8021q/vlan_dev.c static int vlan_dev_fill_forward_path(struct net_device_path_ctx *ctx, · line 887
Quoted 5 lines
net/8021q/vlan_dev.c +5
-0+ +#define VLAN_MACSEC_DECLARE_MDO (mdo) .mdo_ ## mdo = vlan_macsec_ ## mdo + +static const struct macsec_ops macsec_offload_ops = { +VLAN_MACSEC_DECLARE_MDO (add_txsa),This completely breaks the ability to use cscope when looking for implementations of
mdo_add_txsa. I'm not very fond of the c/p, but I don't think we should be using macros at all here. At least to me, being able to navigate directly frommdo_add_txsato its implementation without expanding the macro manually is important.
Quoted 3 lines
So, IMHO, those should be:
.mdo_add_txsa =
vlan_macsec_add_txsa,
Completely agree with you.
Quoted 6 lines
(etc)
Signature · 1 line
Sabrina
Sabrina Dubroca
maintainerRe: [PATCH net-next v2 1/4] vlan: Add MACsec offload operations for VLAN interface
Quoted 25 lines · Leon Romanovsky
Quoted 11 lines · Sabrina Dubroca
Quoted 7 lines · Emeel Hakim
Add support for MACsec offload operations for VLAN driver to allow offloading MACsec when VLAN's real device supports Macsec offload by forwarding the offload request to it.
Signed-off-by Emeel Hakim <[email protected]>---
V1 -> V2: - Consultvlan_featureswhen addingNETIF_F_HW_MACSEC.Uh? You're not actually doing that? You also dropped the changes to
vlan_dev_fix_featureswithout explaining why.
vlan_dev_fix_features()relies onreal_dev->vlan_featureswhich was set in mlx5 part of this patch.643 static
netdev_features_tvlan_dev_fix_features(structnet_device*dev,
644netdev_features_tfeatures)
645 {
...
649
650lower_features= netdev_intersect_features((real_dev->vlan_features |
651NETIF_F_RXCSUM),
652real_dev->features);
Quoted 2 lines
This part ensure that once
real_dev->vlan_featuresandreal_dev->featureshaveNETIF_F_HW_MACSEC, the returned features will includeNETIF_F_HW_MACSECtoo.
Ok, thanks.
But back to the issue of vlan_features, in vlan_dev_init: I'm not convinced NETIF_F_HW_MACSEC should be added to hw_features based on ->features. That would result in a new vlan device that can't offload macsec at all if it was created at the wrong time (while the lower device's macsec offload was temporarily disabled). AFAIU, vlandev->hw_features should be based on realdev->vlan_features. I don't see a reason to advertise a feature in the vlan device if we won't ever be able to turn it on because it's not in ->vlan_features ("grmbl why can't I enable it, ethtool says it's here?!").
Emeel, I'm not a maintainer, but I don't think you should be reposting until the existing discussion has settled down.
Quoted 12 lines
Quoted 12 lines
[...]
Quoted 10 lines
net/8021q/vlan_dev.c +3
-0@@ -572,6 +573,9 @@ static int vlan_dev_init(struct net_device *dev)NETIF_F_HIGHDMA |NETIF_F_SCTP_CRC |NETIF_F_ALL_FCOE ; +if (real_dev->features &NETIF_F_HW_MACSEC ) + dev->hw_features |=NETIF_F_HW_MACSEC ; + dev->features |= dev->hw_features |NETIF_F_LLTX ;netif_inherit_tso_max (dev, real_dev);if (dev->features &NETIF_F_VLAN_FEATURES )
Signature · 1 line
Sabrina
Jakub Kicinski
maintainerRe: [PATCH net-next v2 1/4] vlan: Add MACsec offload operations for VLAN interface
Quoted 2 lines · On Thu, 30 Mar 2023 19:19:21 +0200 Sabrina Dubroca wrote:
I don't think you should be reposting
until the existing discussion has settled down.
Thanks for that note, very true, v3 discarded.
RE: [PATCH net-next v2 1/4] vlan: Add MACsec offload operations for VLAN interface
Quoted 12 lines
-----Original Message-----
From: Jakub Kicinski <[email protected]>
Sent: Thursday, 30 March 2023 20:43
To: Sabrina Dubroca <[email protected]>
Cc: Leon Romanovsky <[email protected]>; Emeel Hakim <[email protected]>; [email protected]; [email protected]; [email protected]; [email protected]
Subject: Re: [PATCH net-next v2 1/4] vlan: Add MACsec offload operations for VLAN interface
Quoted 8 lines
External email: Use caution opening links or attachments
Quoted 2 lines · On Thu, 30 Mar 2023 19:19:21 +0200 Sabrina Dubroca wrote:
I don't think you should be reposting
until the existing discussion has settled down.Thanks for that note, very true, v3 discarded.
Sure, I thought it was closed after Leon's reply, so I addressed the comments and reposted. My bad.
Re: [PATCH net-next v2 1/4] vlan: Add MACsec offload operations for VLAN interface
In reply to Sabrina DubrocaQuoted 29 lines · Sabrina Dubroca
Quoted 25 lines · Leon Romanovsky
Quoted 11 lines · Sabrina Dubroca
Quoted 7 lines · Emeel Hakim
Add support for MACsec offload operations for VLAN driver to allow offloading MACsec when VLAN's real device supports Macsec offload by forwarding the offload request to it.
Signed-off-by Emeel Hakim <[email protected]>---
V1 -> V2: - Consultvlan_featureswhen addingNETIF_F_HW_MACSEC.Uh? You're not actually doing that? You also dropped the changes to
vlan_dev_fix_featureswithout explaining why.
vlan_dev_fix_features()relies onreal_dev->vlan_featureswhich was set in mlx5 part of this patch.643 static
netdev_features_tvlan_dev_fix_features(structnet_device*dev,
644netdev_features_tfeatures)
645 {
...
649
650lower_features= netdev_intersect_features((real_dev->vlan_features |
651NETIF_F_RXCSUM),
652real_dev->features);Quoted 2 lines
This part ensure that once
real_dev->vlan_featuresandreal_dev->featureshaveNETIF_F_HW_MACSEC, the returned features will includeNETIF_F_HW_MACSECtoo.
Quoted 7 lines
Ok, thanks.
But back to the issue of
vlan_features, invlan_dev_init: I'm not convincedNETIF_F_HW_MACSECshould be added tohw_featuresbased on ->features. That would result in a new vlan device that can't offload macsec at all if it was created at the wrong time (while the lower device's macsec offload was temporarily disabled).
Sorry, I'm new to this netdev features zoo, but if I read correctly Documentation/networking/netdev-features.rst, the ->features is the list of enabled ones:
29 2. netdev->features set contains features which are currently enabled
30 for a device. This should be changed only by network core or in
31 error paths of ndo_set_features callback.
And user will have a chance to disable it for VLAN because it was added to ->hw_features:
24 1. netdev->hw_features set contains features whose state may possibly
25 be changed (enabled or disabled) for a particular device by user's 26 request. This set should be initialized in ndo_init callback and not 27 changed later.
So how can VLAN be created with NETIF_F_HW_MACSEC while real_dev mcasec offload is disabled?
Quoted 1 line
AFAIU,
vlandev->hw_featuresshould be based onrealdev->vlan_features.
Is this macsec offloaded VLAN can be called "child VLAN device"?
33 3. netdev->vlan_features set contains features whose state is inherited
34 by child VLAN devices (limits netdev->features set). This is currently 35 used for all VLAN devices whether tags are stripped or inserted in 36 hardware or software.
Quoted 24 lines
I don't see a reason to advertise a feature in the vlan device if we won't ever be able to turn it on because it's not in ->vlan_features ("grmbl why can't I enable it, ethtool says it's here?!").
Emeel, I'm not a maintainer, but I don't think you should be reposting until the existing discussion has settled down.
Quoted 12 lines
Quoted 12 lines
[...]
Quoted 10 lines
net/8021q/vlan_dev.c +3
-0@@ -572,6 +573,9 @@ static int vlan_dev_init(struct net_device *dev)NETIF_F_HIGHDMA |NETIF_F_SCTP_CRC |NETIF_F_ALL_FCOE ; +if (real_dev->features &NETIF_F_HW_MACSEC ) + dev->hw_features |=NETIF_F_HW_MACSEC ; + dev->features |= dev->hw_features |NETIF_F_LLTX ;netif_inherit_tso_max (dev, real_dev);if (dev->features &NETIF_F_VLAN_FEATURES )Signature · 1 line
Sabrina
Sabrina Dubroca
maintainerRe: [PATCH net-next v2 1/4] vlan: Add MACsec offload operations for VLAN interface
Quoted 49 lines · Leon Romanovsky
Quoted 29 lines · Sabrina Dubroca
Quoted 25 lines · Leon Romanovsky
Quoted 11 lines · Sabrina Dubroca
Quoted 7 lines · Emeel Hakim
Add support for MACsec offload operations for VLAN driver to allow offloading MACsec when VLAN's real device supports Macsec offload by forwarding the offload request to it.
Signed-off-by Emeel Hakim <[email protected]>---
V1 -> V2: - Consultvlan_featureswhen addingNETIF_F_HW_MACSEC.Uh? You're not actually doing that? You also dropped the changes to
vlan_dev_fix_featureswithout explaining why.
vlan_dev_fix_features()relies onreal_dev->vlan_featureswhich was set in mlx5 part of this patch.643 static
netdev_features_tvlan_dev_fix_features(structnet_device*dev,
644netdev_features_tfeatures)
645 {
...
649
650lower_features= netdev_intersect_features((real_dev->vlan_features |
651NETIF_F_RXCSUM),
652real_dev->features);Quoted 2 lines
This part ensure that once
real_dev->vlan_featuresandreal_dev->featureshaveNETIF_F_HW_MACSEC, the returned features will includeNETIF_F_HW_MACSECtoo.Quoted 7 lines
Ok, thanks.
But back to the issue of
vlan_features, invlan_dev_init: I'm not convincedNETIF_F_HW_MACSECshould be added tohw_featuresbased on ->features. That would result in a new vlan device that can't offload macsec at all if it was created at the wrong time (while the lower device's macsec offload was temporarily disabled).Sorry, I'm new to this netdev features zoo, but if I read correctly
Documentation/networking/netdev-features.rst, the ->features is the list of enabled ones:29 2.
netdev->featuresset contains features which are currently enabled
30 for a device. This should be changed only by network core or in
31 error paths ofndo_set_featurescallback.And user will have a chance to disable it for VLAN because it was added to ->hw_features:
Quoted 7 lines
24 1.
netdev->hw_featuresset contains features whose state may possibly25 be changed (enabled or disabled) for a particular device by user's 26 request. This set should be initialized in ndo_init callback and not 27 changed later.So how can VLAN be created with
NETIF_F_HW_MACSECwhilereal_devmcasec offload is disabled?
I'm proposing that be VLAN device be created with the capability (->hw_features contains NETIF_F_HW_MACSEC) but disabled (->features doesn't contain NETIF_F_HW_MACSEC). That way, if NETIF_F_HW_MACSEC is re-enabled on the lower device, you don't need to destroy the VLAN device to enable macsec offload on it as well. You still won't be able to enable macsec offload on the VLAN device unless it's active on the real NIC.
I think whether the lower device currently has NETIF_F_HW_MACSEC should only affect whether you can enable the feature on the vlan device right now. What feature is enabled at creation time should be irrelevant.
Quoted 8 lines
Quoted 1 line
AFAIU,
vlandev->hw_featuresshould be based onrealdev->vlan_features.Is this macsec offloaded VLAN can be called "child VLAN device"?
33 3.
netdev->vlan_featuresset contains features whose state is inherited34 by child VLAN devices (limits netdev->features set). This is currently 35 used for all VLAN devices whether tags are stripped or inserted in 36 hardware or software.
Yes. In this patch, we're talking about this situation:
eth0 --------------> vlan0 --------------> macsec0
real NIC
(capable of (can also
offloading MACsec) offload MACsec)
And vlan0 would be a "child VLAN device" of eth0.
"limits netdev->features set" is the netdev_intersect_features you quoted in your previous email.
Quoted 24 lines
Quoted 24 lines
I don't see a reason to advertise a feature in the vlan device if we won't ever be able to turn it on because it's not in ->vlan_features ("grmbl why can't I enable it, ethtool says it's here?!").
Emeel, I'm not a maintainer, but I don't think you should be reposting until the existing discussion has settled down.
Quoted 12 lines
Quoted 12 lines
[...]
Quoted 10 lines
net/8021q/vlan_dev.c +3
-0@@ -572,6 +573,9 @@ static int vlan_dev_init(struct net_device *dev)NETIF_F_HIGHDMA |NETIF_F_SCTP_CRC |NETIF_F_ALL_FCOE ; +if (real_dev->features &NETIF_F_HW_MACSEC ) + dev->hw_features |=NETIF_F_HW_MACSEC ; + dev->features |= dev->hw_features |NETIF_F_LLTX ;netif_inherit_tso_max (dev, real_dev);if (dev->features &NETIF_F_VLAN_FEATURES )Signature · 1 line
Sabrina
Signature · 1 line
Sabrina
RE: [PATCH net-next v2 1/4] vlan: Add MACsec offload operations for VLAN interface
Quoted 82 lines
-----Original Message-----
From: Sabrina Dubroca <[email protected]>
Sent: Thursday, 30 March 2023 23:33
To: Leon Romanovsky <[email protected]>
Cc: Emeel Hakim <[email protected]>; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]
Subject: Re: [PATCH net-next v2 1/4] vlan: Add MACsec offload operations for VLAN interfaceExternal email: Use caution opening links or attachments
Quoted 18 lines · Leon Romanovsky
Quoted 17 lines · Sabrina Dubroca
Quoted 16 lines · Leon Romanovsky
Quoted 11 lines · Sabrina Dubroca
Quoted 7 lines · Emeel Hakim
Add support for MACsec offload operations for VLAN driver to allow offloading MACsec when VLAN's real device supports Macsec offload by forwarding the offload request to it.
Signed-off-by Emeel Hakim <[email protected]>---
V1 -> V2: - Consultvlan_featureswhen addingNETIF_F_HW_MACSEC.Uh? You're not actually doing that? You also dropped the changes to
vlan_dev_fix_featureswithout explaining why.
vlan_dev_fix_features()relies onreal_dev->vlan_featureswhich was set in mlx5 part of this patch.Quoted 12 lines
Quoted 6 lines
Quoted 6 lines
643 static
netdev_features_tvlan_dev_fix_features(structnet_device*dev,
644netdev_features_tfeatures)
645 {
...
649
650lower_features= netdev_intersect_features((real_dev-
vlan_features|Quoted 5 lines
Quoted 5 lines
651
NETIF_F_RXCSUM),
652real_dev->features);This part ensure that once
real_dev->vlan_featuresandreal_dev->featureshaveNETIF_F_HW_MACSEC, the returned features willinclude
NETIF_F_HW_MACSECtoo.Quoted 21 lines
Quoted 9 lines
Ok, thanks.
But back to the issue of
vlan_features, invlan_dev_init: I'm not convincedNETIF_F_HW_MACSECshould be added tohw_featuresbased on ->features. That would result in a new vlan device that can't ->offload
macsec at all if it was created at the wrong time (while the lower device's macsec offload was temporarily disabled).Sorry, I'm new to this netdev features zoo, but if I read correctly
Documentation/networking/netdev-features.rst, the ->features is the list of enabled ones:29 2.
netdev->featuresset contains features which are currently enabled
30 for a device. This should be changed only by network core or in
31 error paths ofndo_set_featurescallback.And user will have a chance to disable it for VLAN because it was added to ->hw_features:
Quoted 7 lines
24 1.
netdev->hw_featuresset contains features whose state may possibly25 be changed (enabled or disabled) for a particular device by user's 26 request. This set should be initialized in ndo_init callback and not 27 changed later.So how can VLAN be created with
NETIF_F_HW_MACSECwhilereal_devmcasec offload is disabled?I'm proposing that be VLAN device be created with the capability (->hw_features contains
NETIF_F_HW_MACSEC) but disabled (->features doesn't containNETIF_F_HW_MACSEC). That way, ifNETIF_F_HW_MACSECis re-enabled on the lower device, you don't need to destroy the VLAN device to enable macsec offload on it as well. You still won't be able to enable macsec offload on the VLAN device unless it's active on the real NIC.
Quoted 3 lines
I think whether the lower device currently has
NETIF_F_HW_MACSECshould only affect whether you can enable the feature on the vlan device right now. What feature is enabled at creation time should be irrelevant.
Thanks for the proposal Sabrina, I'm also new to this netdev features zone so IIUC your'e proposing that we have NETIF_F_HW_MACSEC added to the dev->hw_features uponvlan_dev_init, but disabled (we don’t add it to dev->features) , and upon vlan_dev_fix_features we check if the real_device have NETIF_F_HW_MACSEC enabled (after the intersect with the real_dev->vlan_features) and if so we add it to the features.
So something like:
static int vlan_dev_init(struct net_device *dev)
{
...
dev->features |= dev->hw_features | NETIF_F_LLTX; dev->hw_features |= NETIF_F_HW_MACSEC;
...
}
static netdev_features_t vlan_dev_fix_features(struct net_device *dev,
netdev_features_t features)
{
...
if (lower_features & NETIF_F_HW_MACSEC)
features |= NETIF_F_HW_MACSEC;
return features;
}
if this is the case then I totally agree that this is a better approach.
Quoted 48 lines
Quoted 8 lines
Quoted 1 line
AFAIU,
vlandev->hw_featuresshould be based onrealdev->vlan_features.Is this macsec offloaded VLAN can be called "child VLAN device"?
33 3.
netdev->vlan_featuresset contains features whose state is inherited34 by child VLAN devices (limits netdev->features set). This is currently 35 used for all VLAN devices whether tags are stripped or inserted in 36 hardware or software.Yes. In this patch, we're talking about this situation:
eth0 --------------> vlan0 --------------> macsec0
real NIC
(capable of (can also
offloading MACsec) offload MACsec)And vlan0 would be a "child VLAN device" of eth0.
"limits
netdev->featuresset" is thenetdev_intersect_featuresyou quoted in your previous email.Quoted 14 lines
Quoted 14 lines
I don't see a reason to advertise a feature in the vlan device if we won't ever be able to turn it on because it's not in ->vlan_features ("grmbl why can't I enable it, ethtool says it's here?!").
Emeel, I'm not a maintainer, but I don't think you should be reposting until the existing discussion has settled down.
Quoted 6 lines
Quoted 6 lines
[...]
Quoted 4 lines
net/8021q/vlan_dev.c +0
-0@@ -572,6 +573,9 @@ static int vlan_dev_init(struct net_device *dev)NETIF_F_HIGHDMA |NETIF_F_SCTP_CRC |NETIF_F_ALL_FCOE ;Quoted 10 lines
Quoted 10 lines
Quoted 6 lines
Quoted 6 lines
net/8021q/vlan_dev.c static int vlan_dev_init(struct net_device *dev) · line 581
Quoted 6 lines
net/8021q/vlan_dev.c +3
-0+if (real_dev->features &NETIF_F_HW_MACSEC ) + dev->hw_features |=NETIF_F_HW_MACSEC ; + dev->features |= dev->hw_features |NETIF_F_LLTX ;netif_inherit_tso_max (dev, real_dev);if (dev->features &NETIF_F_VLAN_FEATURES )--
Sabrina--
Sabrina
Sabrina Dubroca
maintainerRe: [PATCH net-next v2 1/4] vlan: Add MACsec offload operations for VLAN interface
Quoted 94 lines · Emeel Hakim
Quoted 82 lines
-----Original Message-----
From: Sabrina Dubroca <[email protected]>
Sent: Thursday, 30 March 2023 23:33
To: Leon Romanovsky <[email protected]>
Cc: Emeel Hakim <[email protected]>; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]
Subject: Re: [PATCH net-next v2 1/4] vlan: Add MACsec offload operations for VLAN interfaceExternal email: Use caution opening links or attachments
Quoted 18 lines · Leon Romanovsky
Quoted 17 lines · Sabrina Dubroca
Quoted 16 lines · Leon Romanovsky
Quoted 11 lines · Sabrina Dubroca
Quoted 7 lines · Emeel Hakim
Add support for MACsec offload operations for VLAN driver to allow offloading MACsec when VLAN's real device supports Macsec offload by forwarding the offload request to it.
Signed-off-by Emeel Hakim <[email protected]>---
V1 -> V2: - Consultvlan_featureswhen addingNETIF_F_HW_MACSEC.Uh? You're not actually doing that? You also dropped the changes to
vlan_dev_fix_featureswithout explaining why.
vlan_dev_fix_features()relies onreal_dev->vlan_featureswhich was set in mlx5 part of this patch.Quoted 12 lines
Quoted 6 lines
Quoted 6 lines
643 static
netdev_features_tvlan_dev_fix_features(structnet_device*dev,
644netdev_features_tfeatures)
645 {
...
649
650lower_features= netdev_intersect_features((real_dev-
vlan_features|Quoted 5 lines
Quoted 5 lines
651
NETIF_F_RXCSUM),
652real_dev->features);This part ensure that once
real_dev->vlan_featuresandreal_dev->featureshaveNETIF_F_HW_MACSEC, the returned features willinclude
NETIF_F_HW_MACSECtoo.Quoted 21 lines
Quoted 9 lines
Ok, thanks.
But back to the issue of
vlan_features, invlan_dev_init: I'm not convincedNETIF_F_HW_MACSECshould be added tohw_featuresbased on ->features. That would result in a new vlan device that can't ->offload
macsec at all if it was created at the wrong time (while the lower device's macsec offload was temporarily disabled).Sorry, I'm new to this netdev features zoo, but if I read correctly
Documentation/networking/netdev-features.rst, the ->features is the list of enabled ones:29 2.
netdev->featuresset contains features which are currently enabled
30 for a device. This should be changed only by network core or in
31 error paths ofndo_set_featurescallback.And user will have a chance to disable it for VLAN because it was added to ->hw_features:
Quoted 7 lines
24 1.
netdev->hw_featuresset contains features whose state may possibly25 be changed (enabled or disabled) for a particular device by user's 26 request. This set should be initialized in ndo_init callback and not 27 changed later.So how can VLAN be created with
NETIF_F_HW_MACSECwhilereal_devmcasec offload is disabled?I'm proposing that be VLAN device be created with the capability (->hw_features contains
NETIF_F_HW_MACSEC) but disabled (->features doesn't containNETIF_F_HW_MACSEC). That way, ifNETIF_F_HW_MACSECis re-enabled on the lower device, you don't need to destroy the VLAN device to enable macsec offload on it as well. You still won't be able to enable macsec offload on the VLAN device unless it's active on the real NIC.Quoted 3 lines
I think whether the lower device currently has
NETIF_F_HW_MACSECshould only affect whether you can enable the feature on the vlan device right now. What feature is enabled at creation time should be irrelevant.Thanks for the proposal Sabrina, I'm also new to this netdev features zone so IIUC your'e proposing that we have
NETIF_F_HW_MACSECadded to thedev->hw_featuresuponvlan_dev_init, but disabled (we don’t add it todev->features) , and uponvlan_dev_fix_featureswe check if thereal_devicehaveNETIF_F_HW_MACSECenabled (after the intersect with thereal_dev->vlan_features) and if so we add it to the features.So something like:
Quoted 7 lines
static int
vlan_dev_init(struct net_device *dev)
{
...dev->features |= dev->hw_features | NETIF_F_LLTX; dev->hw_features |= NETIF_F_HW_MACSEC;...
}
That would be adding the NETIF_F_HW_MACSEC to all VLAN devices, whether the lower device advertises this feature or not. That's wrong.
What I had in mind was:
if (real_dev->vlan_features & NETIF_F_HW_MACSEC) dev->hw_features |= NETIF_F_HW_MACSEC;
And we should enable it by default when the lower device has it enabled, which would be the case with this:
@@ -572,6 +572,9 @@ static int vlan_dev_init(struct net_device *dev)
NETIF_F_HIGHDMA | NETIF_F_SCTP_CRC | NETIF_F_ALL_FCOE;
+ if (real_dev->vlan_features & NETIF_F_HW_MACSEC)
+ dev->hw_features |= NETIF_F_HW_MACSEC;
+
dev->features |= dev->hw_features | NETIF_F_LLTX; netif_inherit_tso_max(dev, real_dev);
if (dev->features & NETIF_F_VLAN_FEATURES)
What I meant by "but disabled" in my previous email was that if the lower device currently has NETIF_F_HW_MACSEC, the new vlan device should also have it disabled, not that it should always be disabled on creation.
Quoted 9 lines
static
netdev_features_tvlan_dev_fix_features(structnet_device*dev,
netdev_features_tfeatures)
{
...(quoted diff) +0
-0if (lower_features & NETIF_F_HW_MACSEC) features |= NETIF_F_HW_MACSEC;return features;
}
I don't think NETIF_F_HW_MACSEC is "special" enough to require hacks in vlan_dev_fix_features. IMHO modifying vlan_dev_fix_features should only happen if we have no other way to implement a consistent and useful behavior. I don't think that's the case here.
Signature · 1 line
Sabrina
RE: [PATCH net-next v2 1/4] vlan: Add MACsec offload operations for VLAN interface
Quoted 149 lines
-----Original Message-----
From: Sabrina Dubroca <[email protected]>
Sent: Tuesday, 4 April 2023 15:54
To: Emeel Hakim <[email protected]>
Cc: Leon Romanovsky <[email protected]>; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]
Subject: Re: [PATCH net-next v2 1/4] vlan: Add MACsec offload operations for VLAN interfaceExternal email: Use caution opening links or attachments
Quoted 19 lines · Emeel Hakim
Quoted 19 lines
-----Original Message-----
From: Sabrina Dubroca <[email protected]>
Sent: Thursday, 30 March 2023 23:33
To: Leon Romanovsky <[email protected]>
Cc: Emeel Hakim <[email protected]>; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]
Subject: Re: [PATCH net-next v2 1/4] vlan: Add MACsec offload operations for VLAN interfaceExternal email: Use caution opening links or attachments
Quoted 3 lines · Leon Romanovsky
Quoted 2 lines · Sabrina Dubroca
Quoted 1 line · Leon Romanovsky
On Wed, Mar 29, 2023 at 04:43:59PM +0200, Sabrina Dubroca wrote:
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 7 lines · Emeel Hakim
Add support for MACsec offload operations for VLAN driver to allow offloading MACsec when VLAN's real device supports Macsec offload by forwarding the offload request to it.
Signed-off-by Emeel Hakim <[email protected]>---
V1 -> V2: - Consultvlan_featureswhen adding
NETIF_F_HW_MACSEC.Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 3 lines
Uh? You're not actually doing that? You also dropped the changes to
vlan_dev_fix_featureswithout explaining why.
vlan_dev_fix_features()relies onreal_dev->vlan_featureswhich was set in mlx5 part of this patch.643 static
netdev_features_tvlan_dev_fix_features(structnet_device*dev,
Quoted 28 lines
Quoted 28 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
644 netdev_features_t features) 645 { ... 649 650 lower_features = netdev_intersect_features((real_dev-Quoted 7 lines
vlan_features|Quoted 6 lines
Quoted 6 lines
651
NETIF_F_RXCSUM),
652real_dev->features);This part ensure that once
real_dev->vlan_featuresandreal_dev->featureshaveNETIF_F_HW_MACSEC, the returned features willinclude
NETIF_F_HW_MACSECtoo.Quoted 15 lines
Quoted 10 lines
Ok, thanks.
But back to the issue of
vlan_features, invlan_dev_init: I'm not convincedNETIF_F_HW_MACSECshould be added tohw_featuresbased on
->features. That would result in a new vlan device that can't ->offload
macsec at all if it was created at the wrong time (while the lower device's macsec offload was temporarily disabled).Sorry, I'm new to this netdev features zoo, but if I read correctly
Documentation/networking/netdev-features.rst, the ->features is the list of enabled ones:Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
29 2.
netdev->featuresset contains features which are currently enabled
30 for a device. This should be changed only by network core or in
31 error paths ofndo_set_featurescallback.And user will have a chance to disable it for VLAN because it was added to ->hw_features:
24 1.
netdev->hw_featuresset contains features whose state maypossibly
Quoted 7 lines
Quoted 7 lines
Quoted 6 lines
25 be changed (enabled or disabled) for a particular device by user's 26 request. This set should be initialized in ndo_init callback and not 27 changed later.So how can VLAN be created with
NETIF_F_HW_MACSECwhilereal_devmcasec offload is disabled?Quoted 6 lines
Quoted 6 lines
I'm proposing that be VLAN device be created with the capability (->hw_features contains
NETIF_F_HW_MACSEC) but disabled (->features doesn't containNETIF_F_HW_MACSEC). That way, ifNETIF_F_HW_MACSECis re-enabled on the lower device, you don't need to destroy the VLAN device to enable macsec offload on it as well. You still won't be able to enable macsec offload on the VLAN device unless it's active on thereal NIC.
Quoted 4 lines
Quoted 4 lines
I think whether the lower device currently has
NETIF_F_HW_MACSECshould only affect whether you can enable the feature on the vlan device right now. What feature is enabled at creation time should beirrelevant.
Quoted 6 lines
Thanks for the proposal Sabrina, I'm also new to this netdev features zone so IIUC your'e proposing that we have
NETIF_F_HW_MACSECadded to thedev->hw_featuresuponvlan_dev_init, but disabled (we don’t add it todev->features) , and uponvlan_dev_fix_featureswe check if thereal_devicehaveNETIF_F_HW_MACSECenabled (after the intersect with the
real_dev->vlan_features) and if so we add it to the features.Quoted 7 lines
So something like:
static int
vlan_dev_init(struct net_device *dev){ ...dev->features |= dev->hw_features | NETIF_F_LLTX; dev->hw_features |= NETIF_F_HW_MACSEC; ... }That would be adding the
NETIF_F_HW_MACSECto all VLAN devices, whether the lower device advertises this feature or not. That's wrong.What I had in mind was:
if (real_dev->vlan_features & NETIF_F_HW_MACSEC) dev->hw_features |= NETIF_F_HW_MACSEC;And we should enable it by default when the lower device has it enabled, which would be the case with this:
(quoted diff) +3
-0@@ -572,6 +572,9 @@ static int vlan_dev_init(struct net_device *dev) NETIF_F_HIGHDMA | NETIF_F_SCTP_CRC | NETIF_F_ALL_FCOE; + if (real_dev->vlan_features & NETIF_F_HW_MACSEC) + dev->hw_features |= NETIF_F_HW_MACSEC; + dev->features |= dev->hw_features | NETIF_F_LLTX; netif_inherit_tso_max(dev, real_dev); if (dev->features & NETIF_F_VLAN_FEATURES)
Quoted 4 lines
What I meant by "but disabled" in my previous email was that if the lower device currently has
NETIF_F_HW_MACSEC, the new vlan device should also have it disabled, not that it should always be disabled on creation.
Thanks for the explanation its clear for me now, I tested it and its working for me. I agree with this approach.
I can prepare a new version if we are closed on everything.
Should I send a v3 (since previous v3 got discarded) or I send it as a v4 ?
Quoted 7 lines
Quoted 7 lines
static
netdev_features_tvlan_dev_fix_features(structnet_device*dev,
netdev_features_tfeatures)
{
...
if (lower_features&NETIF_F_HW_MACSEC)
features |=NETIF_F_HW_MACSEC;
Quoted 7 lines
Quoted 2 lines
return features;
}I don't think
NETIF_F_HW_MACSECis "special" enough to require hacks invlan_dev_fix_features. IMHO modifyingvlan_dev_fix_featuresshould only happen if we have no other way to implement a consistent and useful behavior. I don't think that's the case here.
I agree, no need to touch vlan_dev_fix_features with your approach.
Quoted 2 lines
--
Sabrina
Sabrina Dubroca
maintainerRe: [PATCH net-next v2 1/4] vlan: Add MACsec offload operations for VLAN interface
Quoted 149 lines · Emeel Hakim
Quoted 149 lines
-----Original Message-----
From: Sabrina Dubroca <[email protected]>
Sent: Tuesday, 4 April 2023 15:54
To: Emeel Hakim <[email protected]>
Cc: Leon Romanovsky <[email protected]>; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]
Subject: Re: [PATCH net-next v2 1/4] vlan: Add MACsec offload operations for VLAN interfaceExternal email: Use caution opening links or attachments
Quoted 19 lines · Emeel Hakim
Quoted 19 lines
-----Original Message-----
From: Sabrina Dubroca <[email protected]>
Sent: Thursday, 30 March 2023 23:33
To: Leon Romanovsky <[email protected]>
Cc: Emeel Hakim <[email protected]>; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]
Subject: Re: [PATCH net-next v2 1/4] vlan: Add MACsec offload operations for VLAN interfaceExternal email: Use caution opening links or attachments
Quoted 3 lines · Leon Romanovsky
Quoted 2 lines · Sabrina Dubroca
Quoted 1 line · Leon Romanovsky
On Wed, Mar 29, 2023 at 04:43:59PM +0200, Sabrina Dubroca wrote:
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 7 lines · Emeel Hakim
Add support for MACsec offload operations for VLAN driver to allow offloading MACsec when VLAN's real device supports Macsec offload by forwarding the offload request to it.
Signed-off-by Emeel Hakim <[email protected]>---
V1 -> V2: - Consultvlan_featureswhen adding
NETIF_F_HW_MACSEC.Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 3 lines
Uh? You're not actually doing that? You also dropped the changes to
vlan_dev_fix_featureswithout explaining why.
vlan_dev_fix_features()relies onreal_dev->vlan_featureswhich was set in mlx5 part of this patch.643 static
netdev_features_tvlan_dev_fix_features(structnet_device*dev,
Quoted 28 lines
Quoted 28 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
644 netdev_features_t features) 645 { ... 649 650 lower_features = netdev_intersect_features((real_dev-Quoted 7 lines
vlan_features|Quoted 6 lines
Quoted 6 lines
651
NETIF_F_RXCSUM),
652real_dev->features);This part ensure that once
real_dev->vlan_featuresandreal_dev->featureshaveNETIF_F_HW_MACSEC, the returned features willinclude
NETIF_F_HW_MACSECtoo.Quoted 15 lines
Quoted 10 lines
Ok, thanks.
But back to the issue of
vlan_features, invlan_dev_init: I'm not convincedNETIF_F_HW_MACSECshould be added tohw_featuresbased on
->features. That would result in a new vlan device that can't ->offload
macsec at all if it was created at the wrong time (while the lower device's macsec offload was temporarily disabled).Sorry, I'm new to this netdev features zoo, but if I read correctly
Documentation/networking/netdev-features.rst, the ->features is the list of enabled ones:Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
29 2.
netdev->featuresset contains features which are currently enabled
30 for a device. This should be changed only by network core or in
31 error paths ofndo_set_featurescallback.And user will have a chance to disable it for VLAN because it was added to ->hw_features:
24 1.
netdev->hw_featuresset contains features whose state maypossibly
Quoted 7 lines
Quoted 7 lines
Quoted 6 lines
25 be changed (enabled or disabled) for a particular device by user's 26 request. This set should be initialized in ndo_init callback and not 27 changed later.So how can VLAN be created with
NETIF_F_HW_MACSECwhilereal_devmcasec offload is disabled?Quoted 6 lines
Quoted 6 lines
I'm proposing that be VLAN device be created with the capability (->hw_features contains
NETIF_F_HW_MACSEC) but disabled (->features doesn't containNETIF_F_HW_MACSEC). That way, ifNETIF_F_HW_MACSECis re-enabled on the lower device, you don't need to destroy the VLAN device to enable macsec offload on it as well. You still won't be able to enable macsec offload on the VLAN device unless it's active on thereal NIC.
Quoted 4 lines
Quoted 4 lines
I think whether the lower device currently has
NETIF_F_HW_MACSECshould only affect whether you can enable the feature on the vlan device right now. What feature is enabled at creation time should beirrelevant.
Quoted 6 lines
Thanks for the proposal Sabrina, I'm also new to this netdev features zone so IIUC your'e proposing that we have
NETIF_F_HW_MACSECadded to thedev->hw_featuresuponvlan_dev_init, but disabled (we don’t add it todev->features) , and uponvlan_dev_fix_featureswe check if thereal_devicehaveNETIF_F_HW_MACSECenabled (after the intersect with the
real_dev->vlan_features) and if so we add it to the features.Quoted 7 lines
So something like:
static int
vlan_dev_init(struct net_device *dev){ ...dev->features |= dev->hw_features | NETIF_F_LLTX; dev->hw_features |= NETIF_F_HW_MACSEC; ... }That would be adding the
NETIF_F_HW_MACSECto all VLAN devices, whether the lower device advertises this feature or not. That's wrong.What I had in mind was:
if (real_dev->vlan_features & NETIF_F_HW_MACSEC) dev->hw_features |= NETIF_F_HW_MACSEC;And we should enable it by default when the lower device has it enabled, which would be the case with this:
(quoted diff) +3
-0@@ -572,6 +572,9 @@ static int vlan_dev_init(struct net_device *dev) NETIF_F_HIGHDMA | NETIF_F_SCTP_CRC | NETIF_F_ALL_FCOE; + if (real_dev->vlan_features & NETIF_F_HW_MACSEC) + dev->hw_features |= NETIF_F_HW_MACSEC; + dev->features |= dev->hw_features | NETIF_F_LLTX; netif_inherit_tso_max(dev, real_dev); if (dev->features & NETIF_F_VLAN_FEATURES)
Quoted 9 lines
Quoted 4 lines
What I meant by "but disabled" in my previous email was that if the lower device currently has
NETIF_F_HW_MACSEC, the new vlan device should also have it disabled, not that it should always be disabled on creation.Thanks for the explanation its clear for me now, I tested it and its working for me. I agree with this approach.
I can prepare a new version if we are closed on everything.
Should I send a v3 (since previous v3 got discarded) or I send it as a v4 ?
I'd call that v4 so that nobody gets confused. It's going to be different from the v3 you already posted. I'm going through the patches again, I have a couple of other comments.
Signature · 1 line
Sabrina
Sabrina Dubroca
maintainerRe: [PATCH net-next v2 1/4] vlan: Add MACsec offload operations for VLAN interface
In reply to Emeel HakimQuoted 18 lines · Emeel Hakim
Add support for MACsec offload operations for VLAN driver to allow offloading MACsec when VLAN's real device supports Macsec offload by forwarding the offload request to it.
Signed-off-by Emeel Hakim <[email protected]>---
V1 -> V2: - Consultvlan_featureswhen addingNETIF_F_HW_MACSEC.
- Allow grep for the functions.
- Add helper function to get the macsec operation to allow the compiler to make some choice..../net/ethernet/mellanox/mlx5/core/en_main.c | 1 + net/8021q/vlan_dev.c | 101 ++++++++++++++++++ 2 files changed, 102 insertions(+)drivers/net/ethernet/mellanox/mlx5/core/en_main.c +0
-0
drivers/net/ethernet/mellanox/mlx5/core/en_main.c static void mlx5e_build_nic_netdev(struct net_device *netdev) · line 5082
Quoted 7 lines
drivers/net/ethernet/mellanox/mlx5/core/en_main.c +1
-0@@ -5076,6 +5076,7 @@ static void mlx5e_build_nic_netdev(struct net_device *netdev) netdev->vlan_features |=NETIF_F_SG ; netdev->vlan_features |=NETIF_F_HW_CSUM ; + netdev->vlan_features |=NETIF_F_HW_MACSEC ; netdev->vlan_features |=NETIF_F_GRO ; netdev->vlan_features |=NETIF_F_TSO ; netdev->vlan_features |=NETIF_F_TSO6 ;
IMO that shouldn't be part of this patch. One patch for VLAN to add MACsec offload, then one for mlx5 to enable the feature.
Signature · 1 line
Sabrina
RE: [PATCH net-next v2 1/4] vlan: Add MACsec offload operations for VLAN interface
Quoted 37 lines
-----Original Message-----
From: Sabrina Dubroca <[email protected]>
Sent: Wednesday, 5 April 2023 12:27
To: Emeel Hakim <[email protected]>
Cc: [email protected]; [email protected]; [email protected]; [email protected]; [email protected]
Subject: Re: [PATCH net-next v2 1/4] vlan: Add MACsec offload operations for VLAN interfaceExternal email: Use caution opening links or attachments
Quoted 9 lines · Emeel Hakim
Add support for MACsec offload operations for VLAN driver to allow offloading MACsec when VLAN's real device supports Macsec offload by forwarding the offload request to it.
Signed-off-by Emeel Hakim <[email protected]>---
V1 -> V2: - Consultvlan_featureswhen addingNETIF_F_HW_MACSEC.
- Allow grep for the functions.
- Add helper function to get the macsec operation to allow the compilerto make some choice.
Quoted 12 lines
.../net/ethernet/mellanox/mlx5/core/en_main.c | 1 + net/8021q/vlan_dev.c | 101 ++++++++++++++++++ 2 files changed, 102 insertions(+)drivers/net/ethernet/mellanox/mlx5/core/en_main.c +0
-0@@ -5076,6 +5076,7 @@ static void mlx5e_build_nic_netdev(struct net_device *netdev)
Quoted 9 lines
Quoted 6 lines
drivers/net/ethernet/mellanox/mlx5/core/en_main.c +1
-0netdev->vlan_features |=NETIF_F_SG ; netdev->vlan_features |=NETIF_F_HW_CSUM ; + netdev->vlan_features |=NETIF_F_HW_MACSEC ; netdev->vlan_features |=NETIF_F_GRO ; netdev->vlan_features |=NETIF_F_TSO ;netdev->vlan_features |= NETIF_F_TSO6;IMO that shouldn't be part of this patch. One patch for VLAN to add MACsec offload, then one for mlx5 to enable the feature.
I can split the patch into two patches and send all as new version, should It be as a new V3 (since last V3 got discarded) or I send as a V4?
Quoted 2 lines
--
Sabrina
PATCH v2 2/4 net/mlx5: Support MACsec over VLAN
In reply to Emeel HakimMACsec device may have a VLAN device on top of it. Detect MACsec state correctly under this condition, and return the correct net device accordingly.
--- .../mellanox/mlx5/core/en_accel/macsec.c | 42 ++++++++++++------- 1 file changed, 26 insertions(+), 16 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/en_accel/macsec.c +26 -16
@@ -4,6 +4,7 @@#include <linux/mlx5/device.h> #include <linux/mlx5/mlx5_ifc.h> #include <linux/xarray.h> +#include <linux/if_vlan.h> #include "en.h" #include "lib/aso.h"
@@ -348,12 +349,21 @@ static void mlx5e_macsec_cleanup_sa(struct mlx5e_macsec *macsec, sa->macsec_rule =NULL ; } +static inline struct mlx5e_priv *macsec_netdev_priv (const struct net_device *dev) +{ +#if defined(CONFIG_VLAN_8021Q ) || defined(CONFIG_VLAN_8021Q_MODULE ) +if (is_vlan_dev (dev)) +return netdev_priv (vlan_dev_priv (dev)->real_dev); +#endif +return netdev_priv (dev); +} +static int mlx5e_macsec_init_sa (struct macsec_context *ctx,struct mlx5e_macsec_sa *sa,bool encrypt,bool is_tx) { -struct mlx5e_priv *priv =netdev_priv (ctx->netdev); +struct mlx5e_priv *priv =macsec_netdev_priv (ctx->netdev);struct mlx5e_macsec *macsec = priv->macsec;struct mlx5_macsec_rule_attrs rule_attrs;struct mlx5_core_dev *mdev = priv->mdev;
@@ -427,7 +437,7 @@ static int macsec_rx_sa_active_update(struct macsec_context *ctx, struct mlx5e_macsec_sa *rx_sa, bool active) { -struct mlx5e_priv *priv =netdev_priv (ctx->netdev); +struct mlx5e_priv *priv =macsec_netdev_priv (ctx->netdev);struct mlx5e_macsec *macsec = priv->macsec;int err =0 ;
@@ -510,7 +520,7 @@ static int mlx5e_macsec_add_txsa(struct macsec_context *ctx) {const struct macsec_tx_sc *tx_sc = &ctx->secy->tx_sc;const struct macsec_tx_sa *ctx_tx_sa = ctx->sa.tx_sa; -struct mlx5e_priv *priv =netdev_priv (ctx->netdev); +struct mlx5e_priv *priv =macsec_netdev_priv (ctx->netdev);const struct macsec_secy *secy = ctx->secy;struct mlx5e_macsec_device *macsec_device;struct mlx5_core_dev *mdev = priv->mdev;
@@ -585,7 +595,7 @@ static int mlx5e_macsec_upd_txsa(struct macsec_context *ctx) {const struct macsec_tx_sc *tx_sc = &ctx->secy->tx_sc;const struct macsec_tx_sa *ctx_tx_sa = ctx->sa.tx_sa; -struct mlx5e_priv *priv =netdev_priv (ctx->netdev); +struct mlx5e_priv *priv =macsec_netdev_priv (ctx->netdev);struct mlx5e_macsec_device *macsec_device;u8 assoc_num = ctx->sa.assoc_num;struct mlx5e_macsec_sa *tx_sa;
@@ -645,7 +655,7 @@ static int mlx5e_macsec_upd_txsa(struct macsec_context *ctx)static int mlx5e_macsec_del_txsa (struct macsec_context *ctx) { -struct mlx5e_priv *priv =netdev_priv (ctx->netdev); +struct mlx5e_priv *priv =macsec_netdev_priv (ctx->netdev);struct mlx5e_macsec_device *macsec_device;u8 assoc_num = ctx->sa.assoc_num;struct mlx5e_macsec_sa *tx_sa;
@@ -696,7 +706,7 @@ static u32 mlx5e_macsec_get_sa_from_hashtable(struct rhashtable *sci_hash, sci_tstatic int mlx5e_macsec_add_rxsc (struct macsec_context *ctx) {struct mlx5e_macsec_rx_sc_xarray_element *sc_xarray_element; -struct mlx5e_priv *priv =netdev_priv (ctx->netdev); +struct mlx5e_priv *priv =macsec_netdev_priv (ctx->netdev);const struct macsec_rx_sc *ctx_rx_sc = ctx->rx_sc;struct mlx5e_macsec_device *macsec_device;struct mlx5e_macsec_rx_sc *rx_sc;
@@ -776,7 +786,7 @@ static int mlx5e_macsec_add_rxsc(struct macsec_context *ctx)static int mlx5e_macsec_upd_rxsc (struct macsec_context *ctx) { -struct mlx5e_priv *priv =netdev_priv (ctx->netdev); +struct mlx5e_priv *priv =macsec_netdev_priv (ctx->netdev);const struct macsec_rx_sc *ctx_rx_sc = ctx->rx_sc;struct mlx5e_macsec_device *macsec_device;struct mlx5e_macsec_rx_sc *rx_sc;
@@ -854,7 +864,7 @@ static void macsec_del_rxsc_ctx(struct mlx5e_macsec *macsec, struct mlx5e_macsecstatic int mlx5e_macsec_del_rxsc (struct macsec_context *ctx) { -struct mlx5e_priv *priv =netdev_priv (ctx->netdev); +struct mlx5e_priv *priv =macsec_netdev_priv (ctx->netdev);struct mlx5e_macsec_device *macsec_device;struct mlx5e_macsec_rx_sc *rx_sc;struct mlx5e_macsec *macsec;
@@ -890,8 +900,8 @@ static int mlx5e_macsec_del_rxsc(struct macsec_context *ctx)static int mlx5e_macsec_add_rxsa (struct macsec_context *ctx) { +struct mlx5e_priv *priv =macsec_netdev_priv (ctx->netdev);const struct macsec_rx_sa *ctx_rx_sa = ctx->sa.rx_sa; -struct mlx5e_priv *priv =netdev_priv (ctx->netdev);struct mlx5e_macsec_device *macsec_device;struct mlx5_core_dev *mdev = priv->mdev;u8 assoc_num = ctx->sa.assoc_num;
@@ -976,8 +986,8 @@ static int mlx5e_macsec_add_rxsa(struct macsec_context *ctx)static int mlx5e_macsec_upd_rxsa (struct macsec_context *ctx) { +struct mlx5e_priv *priv =macsec_netdev_priv (ctx->netdev);const struct macsec_rx_sa *ctx_rx_sa = ctx->sa.rx_sa; -struct mlx5e_priv *priv =netdev_priv (ctx->netdev);struct mlx5e_macsec_device *macsec_device;u8 assoc_num = ctx->sa.assoc_num;struct mlx5e_macsec_rx_sc *rx_sc;
@@ -1033,7 +1043,7 @@ static int mlx5e_macsec_upd_rxsa(struct macsec_context *ctx)static int mlx5e_macsec_del_rxsa (struct macsec_context *ctx) { -struct mlx5e_priv *priv =netdev_priv (ctx->netdev); +struct mlx5e_priv *priv =macsec_netdev_priv (ctx->netdev);struct mlx5e_macsec_device *macsec_device;sci_t sci = ctx->sa.rx_sa->sc->sci;struct mlx5e_macsec_rx_sc *rx_sc;
@@ -1085,7 +1095,7 @@ static int mlx5e_macsec_del_rxsa(struct macsec_context *ctx)static int mlx5e_macsec_add_secy (struct macsec_context *ctx) { -struct mlx5e_priv *priv =netdev_priv (ctx->netdev); +struct mlx5e_priv *priv =macsec_netdev_priv (ctx->netdev);const struct net_device *dev = ctx->secy->netdev;const struct net_device *netdev = ctx->netdev;struct mlx5e_macsec_device *macsec_device;
@@ -1137,7 +1147,7 @@ static int mlx5e_macsec_add_secy(struct macsec_context *ctx)static int macsec_upd_secy_hw_address (struct macsec_context *ctx,struct mlx5e_macsec_device *macsec_device) { -struct mlx5e_priv *priv =netdev_priv (ctx->netdev); +struct mlx5e_priv *priv =macsec_netdev_priv (ctx->netdev);const struct net_device *dev = ctx->secy->netdev;struct mlx5e_macsec *macsec = priv->macsec;struct mlx5e_macsec_rx_sc *rx_sc, *tmp;
@@ -1184,8 +1194,8 @@ static int macsec_upd_secy_hw_address(struct macsec_context *ctx,*/ static int mlx5e_macsec_upd_secy (struct macsec_context *ctx) { +struct mlx5e_priv *priv =macsec_netdev_priv (ctx->netdev);const struct macsec_tx_sc *tx_sc = &ctx->secy->tx_sc; -struct mlx5e_priv *priv =netdev_priv (ctx->netdev);const struct net_device *dev = ctx->secy->netdev;struct mlx5e_macsec_device *macsec_device;struct mlx5e_macsec_sa *tx_sa;
@@ -1240,7 +1250,7 @@ static int mlx5e_macsec_upd_secy(struct macsec_context *ctx)static int mlx5e_macsec_del_secy (struct macsec_context *ctx) { -struct mlx5e_priv *priv =netdev_priv (ctx->netdev); +struct mlx5e_priv *priv =macsec_netdev_priv (ctx->netdev);struct mlx5e_macsec_device *macsec_device;struct mlx5e_macsec_rx_sc *rx_sc, *tmp;struct mlx5e_macsec_sa *tx_sa;
@@ -1741,7 +1751,7 @@ void mlx5e_macsec_offload_handle_rx_skb(struct net_device *netdev, {struct mlx5e_macsec_rx_sc_xarray_element *sc_xarray_element;u32 macsec_meta_data =be32_to_cpu (cqe->ft_metadata); -struct mlx5e_priv *priv =netdev_priv (netdev); +struct mlx5e_priv *priv =macsec_netdev_priv (netdev);struct mlx5e_macsec_rx_sc *rx_sc;struct mlx5e_macsec *macsec;u32 fs_id;
Signature · 1 line
2.21.3
PATCH v2 3/4 net/mlx5: Consider VLAN interface in MACsec TX steering rules
In reply to Emeel HakimOffloading MACsec when its configured over VLAN with current MACsec TX steering rules will wrongly insert MACsec sec tag after inserting the VLAN header leading to a ETHERNET | SECTAG | VLAN packet when ETHERNET | VLAN | SECTAG is configured.
The above issue is due to adding the SECTAG by HW which is a later stage compared to the VLAN header insertion stage.
Detect such a case and adjust TX steering rules to insert the SECTAG in the correct place by using reformat_param_0 field in the packet reformat to indicate the offset of SECTAG from end of the MAC header to account for VLANs in granularity of 4Bytes.
--- .../net/ethernet/mellanox/mlx5/core/en_accel/macsec_fs.c | 7 +++++++ 1 file changed, 7 insertions(+)
drivers/net/ethernet/mellanox/mlx5/core/en_accel/macsec_fs.c +7 -0
@@ -4,6 +4,7 @@#include <net/macsec.h> #include <linux/netdevice.h> #include <linux/mlx5/qp.h> +#include <linux/if_vlan.h> #include "fs_core.h" #include "en/fs.h" #include "en_accel/macsec_fs.h"
@@ -510,6 +511,8 @@ static void macsec_fs_tx_del_rule(struct mlx5e_macsec_fs *macsec_fs,macsec_fs_tx_ft_put (macsec_fs); } +#define MLX5_REFORMAT_PARAM_ADD_MACSEC_OFFSET_4_BYTES 1 +static union mlx5e_macsec_rule * macsec_fs_tx_add_rule(struct mlx5e_macsec_fs *macsec_fs, const struct macsec_context *macsec_ctx,
@@ -555,6 +558,10 @@ macsec_fs_tx_add_rule(struct mlx5e_macsec_fs *macsec_fs, reformat_params.type =MLX5_REFORMAT_TYPE_ADD_MACSEC ; reformat_params.size = reformat_size; reformat_params.data = reformatbf; + +if (is_vlan_dev (macsec_ctx->netdev)) + reformat_params.param_0 =MLX5_REFORMAT_PARAM_ADD_MACSEC_OFFSET_4_BYTES ; + flow_act.pkt_reformat =mlx5_packet_reformat_alloc (macsec_fs->mdev, &reformat_params,MLX5_FLOW_NAMESPACE_EGRESS_MACSEC );
Signature · 1 line
2.21.3
PATCH v2 4/4 macsec: Add MACsec rx_handler change support
In reply to Emeel HakimOffloading device drivers will mark offloaded MACsec SKBs with the corresponding SCI in the skb_metadata_dst so the macsec rx handler will know to which interface to divert those skbs, in case of a marked skb and a mismatch on the dst MAC address, divert the skb to the macsec net_device where the macsec rx_handler will be called.
Example of such a case is having a MACsec with VLAN as an inner header ETHERNET | SECTAG | VLAN packet.
--- drivers/net/macsec.c | 9 +++++++++ 1 file changed, 9 insertions(+)
drivers/net/macsec.c +9 -0
@@ -1048,6 +1048,15 @@ static enum rx_handler_result handle_not_macsec(struct sk_buff *skb)__netif_rx (nskb); } + +if (md_dst && md_dst->type ==METADATA_MACSEC && + (find_rx_sc (&macsec->secy, md_dst->u.macsec_info.sci))) { + skb->dev = ndev; + skb->pkt_type =PACKET_HOST ; + ret =RX_HANDLER_ANOTHER ; + goto out; + } +continue ; }
Signature · 1 line
2.21.3
Sabrina Dubroca
maintainerRe: [PATCH net-next 4/4] macsec: Add MACsec rx_handler change support
Quoted 5 lines · Emeel Hakim
Offloading device drivers will mark offloaded MACsec SKBs with the corresponding SCI in the
skb_metadata_dstso the macsec rx handler will know to which interface to divert those skbs, in case of a marked skb and a mismatch on the dst MAC address, divert the skb to the macsecnet_devicewhere the macsecrx_handlerwill be called.
Sorry, I don't understand what you're trying to say here and in the subject line.
To me, "Add MACsec rx_handler change support" sounds like you're changing what function is used as ->rx_handler, which is not what this patch is doing.
Quoted 13 lines
Example of such a case is having a MACsec with VLAN as an inner header ETHERNET | SECTAG | VLAN packet.
Signed-off-by Emeel Hakim <[email protected]>--- drivers/net/macsec.c | 9 +++++++++ 1 file changed, 9 insertions(+)drivers/net/macsec.c +0
-0
drivers/net/macsec.c static enum rx_handler_result handle_not_macsec(struct sk_buff *skb) · line 1053
Quoted 6 lines
drivers/net/macsec.c +3
-0@@ -1048,6 +1048,15 @@ static enum rx_handler_result handle_not_macsec(struct sk_buff *skb)__netif_rx (nskb); } + +if (md_dst && md_dst->type ==METADATA_MACSEC && + (find_rx_sc (&macsec->secy, md_dst->u.macsec_info.sci))) {
We already do that exact find_rx_sc call earlier in the same loop, can't we skip it now?
Quoted 12 lines
drivers/net/macsec.c +6
-0+ skb->dev = ndev; + skb->pkt_type =PACKET_HOST ; + ret =RX_HANDLER_ANOTHER ; + goto out; + } +continue ; }Signature · 1 line
2.21.3
Signature · 1 line
Sabrina
RE: [PATCH net-next 4/4] macsec: Add MACsec rx_handler change support
Quoted 41 lines
-----Original Message-----
From: Sabrina Dubroca <[email protected]>
Sent: Wednesday, 5 April 2023 12:36
To: Emeel Hakim <[email protected]>
Cc: [email protected]; [email protected]; [email protected]; [email protected]; [email protected]
Subject: Re: [PATCH net-next 4/4] macsec: Add MACsecrx_handlerchange supportExternal email: Use caution opening links or attachments
Quoted 5 lines · Emeel Hakim
Offloading device drivers will mark offloaded MACsec SKBs with the corresponding SCI in the
skb_metadata_dstso the macsec rx handler will know to which interface to divert those skbs, in case of a marked skb and a mismatch on the dst MAC address, divert the skb to the macsecnet_devicewhere the macsecrx_handlerwill be called.Sorry, I don't understand what you're trying to say here and in the subject line.
To me, "Add MACsec
rx_handlerchange support" sounds like you're changing what function is used as ->rx_handler, which is not what this patch is doing.Quoted 15 lines
Example of such a case is having a MACsec with VLAN as an inner header ETHERNET | SECTAG | VLAN packet.
Signed-off-by Emeel Hakim <[email protected]>--- drivers/net/macsec.c | 9 +++++++++ 1 file changed, 9 insertions(+)drivers/net/macsec.c index renamed from drivers/net/macsec.c +0
-0
25616247d7a5..88b00ea4af68100644drivers/net/macsec.c +0
-0@@ -1048,6 +1048,15 @@ static enum rx_handler_result handle_not_macsec(struct sk_buff *skb)
Quoted 9 lines
drivers/net/macsec.c static enum rx_handler_result handle_not_macsec(struct sk_buff *skb)
Quoted 6 lines
drivers/net/macsec.c +4
-0__netif_rx (nskb); } + +if (md_dst && md_dst->type ==METADATA_MACSEC && + (find_rx_sc (&macsec->secy, + md_dst->u.macsec_info.sci))) {We already do that exact
find_rx_sccall earlier in the same loop, can't we skip it now?
I can save the result into a parameter and use that. Thanks for pointing that.
Quoted 15 lines
Quoted 10 lines
drivers/net/macsec.c +6
-1+ skb->dev = ndev; + skb->pkt_type =PACKET_HOST ; + ret =RX_HANDLER_ANOTHER ; + goto out; + } +continue ; } --Quoted 2 lines
2.21.3
--
Sabrina
Quoted 18 lines · Emeel Hakim
Dear maintainers,
This patch series introduces support for hardware (HW) offload MACsec devices with VLAN configuration. The patches address both scenarios where the VLAN header is both the inner and outer header for MACsec.
The changes include:
1. Adding MACsec offload operation for VLAN.
2. Considering VLAN when accessing MACsec net device.
3. Currently offloading MACsec when it's configured over VLAN with current MACsec TX steering rules would wrongly insert the MACsec sec tag after inserting the VLAN header. This resulted in an ETHERNET | SECTAG | VLAN packet when ETHERNET | VLAN | SECTAG is configured. The patche handles this issue when configuring steering rules.
4. Adding MACsecrx_handlerchange support in case of a marked skb and a mismatch on the dst MAC address.
Quoted 8 lines
Please review these changes and let me know if you have any feedback or concerns.
Updates since v1:
- Consultvlan_featureswhen addingNETIF_F_HW_MACSEC.
- Allow grep for the functions.
- Add helper function to get the macsec operation to allow the compiler to make some choice.
Please mark all your patches as vXXX and not only one.
Thanks