Skip to contentSkip to conversation

ethtool: rss: driver tweaks and netlink context dumps

PATCH Superseded Jakub Kicinski 4 participants 19 messages Aug 6, 2024 – Aug 8, 2024

Subsystems (13)

v3 of 5 Superseded Latest is v5 →

lore
Share & tools
lore

Jakub Kicinski

author maintainer
#0 lore raw

v3 cover ethtool: rss: driver tweaks and netlink context dumps

This series is a semi-related collection of RSS patches. Main point is supporting dumping RSS contexts via ethtool netlink. At present additional RSS contexts can be queried one by one, and assuming user know the right IDs. This series uses the XArray added by Ed to provide netlink dump support for ETHTOOL_GET_RSS.

Patch 1 is a trivial selftest debug patch.
Patch 2 coverts mvpp2 for no real reason other than that I had
a grand plan of converting all drivers at some stage.
Patch 3 removes a now moot check from mlx5 so that all tests
can pass.
Patch 4 and 5 make a bit used for context support optional,
for easier grepping of drivers which need converting if nothing else.
Patch 6 OTOH adds a new cap bit; some devices don't support
using a different key per context and currently act in surprising ways.
Patch 7 and 8 update the RSS netlink code to use XArray. Patch 9 and 10 add support for dumping contexts. Patch 11 and 12 are small adjustments to spec and a new test.

I'm getting distracted with other work, so probably won't have the time soon to complete next steps, but things which are missing are (and some of these may be bad ideas):

- better discovery

Some sort of API to tell the user who many contexts the device can create. Upper bound, devices often share contexts between ports etc. so it's hard to tell exactly and upfront number of contexts for a netdev. But order of magnitude (4 vs 10s) may be enough for container management system to know whether to bother.

- create/modify/delete via netlink

The only question here is how to handle all the tricky IOCTL legacy. "No change" maps trivially to attribute not present. "reset" (indir_size = 0) probably needs to be a new NLA_FLAG?

- better table size handling

The current API assumes the LUT has fixed size, which isn't true for modern devices. We should have better APIs for the drivers to resize the tables, and in user facing API - the ability to specify pattern and min size rather than exact table expected (sort of like ethtool CLI already does).

- recounted / socket-bound contexts

Support for contexts which get "cleaned up" when their parent netlink socket gets closed. The major catch is that ntuple filters (which we don't currently track) depend on the context, so we need auto-removal for both.

v3:
- quite a few code comments and commit message changes
- mvpp2: fix interpretation of max_context_id (I'll take care of the net -> net-next merge as needed)
- filter by ifindex in the selftest
v2: https://lore.kernel.org/[email protected]
- fix bugs and build in mvpp2
v1: https://lore.kernel.org/[email protected]

Jakub Kicinski (12):
selftests: drv-net: rss_ctx: add identifier to traffic comments
eth: mvpp2: implement new RSS context API
eth: mlx5: allow disabling queues when RSS contexts exist
ethtool: make ethtool_ops::cap_rss_ctx_supported optional
eth: remove .cap_rss_ctx_supported from updated drivers
ethtool: rss: don't report key if device doesn't support it
ethtool: rss: move the device op invocation out of rss_prepare_data()
ethtool: rss: report info about additional contexts from XArray
ethtool: rss: support dumping RSS contexts
ethtool: rss: support skipping contexts during dump
netlink: specs: decode indirection table as u32 array
selftests: drv-net: rss_ctx: test dumping RSS contexts

 Documentation/netlink/specs/ethtool.yaml      |  14 +-
 Documentation/networking/ethtool-netlink.rst  |  12 +-
 .../net/ethernet/broadcom/bnxt/bnxt_ethtool.c |   2 +-
 drivers/net/ethernet/intel/ice/ice_ethtool.c  |   1 +
 .../net/ethernet/marvell/mvpp2/mvpp2_cls.c    |  18 +-
 .../net/ethernet/marvell/mvpp2/mvpp2_cls.h    |   2 +-
 .../net/ethernet/marvell/mvpp2/mvpp2_main.c   | 103 +++++---
 .../ethernet/mellanox/mlx5/core/en_ethtool.c  |  13 +-
 drivers/net/ethernet/sfc/ef100_ethtool.c      |   2 +-
 drivers/net/ethernet/sfc/ethtool.c            |   2 +-
 include/linux/ethtool.h                       |   7 +-
 include/uapi/linux/ethtool_netlink.h          |   1 +
 net/ethtool/ioctl.c                           |  31 ++-
 net/ethtool/netlink.c                         |   2 +
 net/ethtool/netlink.h                         |   4 +-
 net/ethtool/rss.c                             | 233 ++++++++++++++++--
 .../selftests/drivers/net/hw/rss_ctx.py       |  80 +++++-
 tools/testing/selftests/net/lib/py/ksft.py    |   6 +
 18 files changed, 442 insertions(+), 91 deletions(-)
Signature · 1 line
2.45.2
author maintainer
#1 lore raw

PATCH v3 1/12 selftests: drv-net: rss_ctx: add identifier to traffic comments

Include the "name" of the context in the comment for traffic checks. Makes it easier to reason about which context failed when we loop over 32 contexts (it may matter if we failed in first vs last, for example).

---
 tools/testing/selftests/drivers/net/hw/rss_ctx.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
tools/testing/selftests/drivers/net/hw/rss_ctx.py +2 -2
index 011508ca604b..1da6b214f4fe 100755
@@ -90,10 +90,10 @@ from lib.py import ethtool, ip, defer, GenerateTraffic, CmdExitFailure
     ksft_ge(directed, 20000, f"traffic on {name}: " + str(cnts))
     if params.get('noise'):
         ksft_lt(sum(cnts[i] for i in params['noise']), directed / 2,
-                "traffic on other queues:" + str(cnts))
+                f"traffic on other queues ({name})':" + str(cnts))
     if params.get('empty'):
         ksft_eq(sum(cnts[i] for i in params['empty']), 0,
-                "traffic on inactive queues: " + str(cnts))
+                f"traffic on inactive queues ({name}): " + str(cnts))
 
 
 def test_rss_key_indir(cfg):
Signature · 1 line
2.45.2
author maintainer
#2 lore raw

PATCH v3 2/12 eth: mvpp2: implement new RSS context API

In reply to Jakub Kicinski

Implement the separate create/modify/delete ops for RSS.

No problems with IDs - even tho RSS tables are per device the driver already seems to allocate IDs linearly per port. There's a translation table from per-port context ID to device context ID.

mvpp2 doesn't have a key for the hash, it defaults to an empty/previous indir table.

Note that there is no key at all, so we don't have to be concerned with reporting the wrong one (which is addressed by a patch later in the series).

Compile-tested only.

---
v3:
- fix max_context_id to be inclusive
- extend commit message with why this is different than patch 6
v2:
- copy the default indir table
- use ID from the core
- fix build issues

---
 .../net/ethernet/marvell/mvpp2/mvpp2_cls.c    |  18 +---
 .../net/ethernet/marvell/mvpp2/mvpp2_cls.h    |   2 +-
 .../net/ethernet/marvell/mvpp2/mvpp2_main.c   | 102 +++++++++++++-----
 3 files changed, 79 insertions(+), 43 deletions(-)
drivers/net/ethernet/marvell/mvpp2/mvpp2_cls.c +4 -14
index 40aeaa7bd739..1641791a2d5b 100644
@@ -1522,29 +1522,19 @@ static int mvpp22_rss_context_create(struct mvpp2_port *port, u32 *rss_ctx)
 	return 0;
 }
 
-int mvpp22_port_rss_ctx_create(struct mvpp2_port *port, u32 *port_ctx)
+int mvpp22_port_rss_ctx_create(struct mvpp2_port *port, u32 port_ctx)
 {
 	u32 rss_ctx;
-	int ret, i;
+	int ret;
 
 	ret = mvpp22_rss_context_create(port, &rss_ctx);
 	if (ret)
 		return ret;
 
-	/* Find the first available context number in the port, starting from 1.
-	 * Context 0 on each port is reserved for the default context.
-	 */
-	for (i = 1; i < MVPP22_N_RSS_TABLES; i++) {
-		if (port->rss_ctx[i] < 0)
-			break;
-	}
-
-	if (i == MVPP22_N_RSS_TABLES)
+	if (WARN_ON_ONCE(port->rss_ctx[port_ctx] >= 0))
 		return -EINVAL;
 
-	port->rss_ctx[i] = rss_ctx;
-	*port_ctx = i;
-
+	port->rss_ctx[port_ctx] = rss_ctx;
 	return 0;
 }
 
drivers/net/ethernet/marvell/mvpp2/mvpp2_cls.h +1 -1
index 663157dc8062..85c9c6e80678 100644
@@ -264,7 +264,7 @@ int mvpp22_port_rss_init(struct mvpp2_port *port);
 int mvpp22_port_rss_enable(struct mvpp2_port *port);
 int mvpp22_port_rss_disable(struct mvpp2_port *port);
 
-int mvpp22_port_rss_ctx_create(struct mvpp2_port *port, u32 *rss_ctx);
+int mvpp22_port_rss_ctx_create(struct mvpp2_port *port, u32 rss_ctx);
 int mvpp22_port_rss_ctx_delete(struct mvpp2_port *port, u32 rss_ctx);
 
 int mvpp22_port_rss_ctx_indir_set(struct mvpp2_port *port, u32 rss_ctx,
drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c +74 -28

124 lines not on this page · raw message

Signature · 1 line
2.45.2
maintainer Reviewed-by on v3 2/12
#3 lore raw

Re: [PATCH net-next v3 02/12] eth: mvpp2: implement new RSS context API

Quoted 10 lines · Jakub Kicinski

Implement the separate create/modify/delete ops for RSS.

No problems with IDs - even tho RSS tables are per device the driver already seems to allocate IDs linearly per port. There's a translation table from per-port context ID to device context ID.

mvpp2 doesn't have a key for the hash, it defaults to an empty/previous indir table.

Quoted 7 lines

Note that there is no key at all, so we don't have to be concerned with reporting the wrong one (which is addressed by a patch later in the series).

Compile-tested only.

Fwiw,

author maintainer
#4 lore raw

PATCH v3 3/12 eth: mlx5: allow disabling queues when RSS contexts exist

In reply to Jakub Kicinski

Since commit 24ac7e544081 ("ethtool: use the rss context XArray in ring deactivation safety-check") core will prevent queues from being disabled while being used by additional RSS contexts. The safety check is no longer necessary, and core will do a more accurate job of only rejecting changes which can actually break things.

---
CC: [email protected]
CC: [email protected]
CC: [email protected]
CC: [email protected]

---
 drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c | 12 ------------
 1 file changed, 12 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c +0 -12
index 36845872ae94..0b941482db30 100644
@@ -445,7 +445,6 @@ int mlx5e_ethtool_set_channels(struct mlx5e_priv *priv,
 	unsigned int count = ch->combined_count;
 	struct mlx5e_params new_params;
 	bool arfs_enabled;
-	int rss_cnt;
 	bool opened;
 	int err = 0;
 
@@ -499,17 +498,6 @@ int mlx5e_ethtool_set_channels(struct mlx5e_priv *priv,
 		goto out;
 	}
 
-	/* Don't allow changing the number of channels if non-default RSS contexts exist,
-	 * the kernel doesn't protect against set_channels operations that break them.
-	 */
-	rss_cnt = mlx5e_rx_res_rss_cnt(priv->rx_res) - 1;
-	if (rss_cnt) {
-		err = -EINVAL;
-		netdev_err(priv->netdev, "%s: Non-default RSS contexts exist (%d), cannot change the number of channels\n",
-			   __func__, rss_cnt);
-		goto out;
-	}
-
 	/* Don't allow changing the number of channels if MQPRIO mode channel offload is active,
 	 * because it defines a partition over the channels queues.
 	 */
Signature · 1 line
2.45.2
author maintainer
#5 lore raw

PATCH v3 4/12 ethtool: make ethtool_ops::cap_rss_ctx_supported optional

In reply to Jakub Kicinski

cap_rss_ctx_supported was created because the API for creating and configuring additional contexts is mux'ed with the normal RSS API. Presence of ops does not imply driver can actually support rss_context != 0 (in fact drivers mostly ignore that field). cap_rss_ctx_supported lets core check that the driver is context-aware before calling it.

Now that we have .create_rxfh_context, there is no such ambiguity. We can depend on presence of the op. Make setting the bit optional.

---
 include/linux/ethtool.h | 3 ++-
 net/ethtool/ioctl.c     | 6 ++++--
 net/ethtool/rss.c       | 3 ++-
 3 files changed, 8 insertions(+), 4 deletions(-)
include/linux/ethtool.h +2 -1
index 303fda54ef17..55c9f613ab64 100644
@@ -727,7 +727,8 @@ struct kernel_ethtool_ts_info {
  * @cap_link_lanes_supported: indicates if the driver supports lanes
  *	parameter.
  * @cap_rss_ctx_supported: indicates if the driver supports RSS
- *	contexts.
+ *	contexts via legacy API, drivers implementing @create_rxfh_context
+ *	do not have to set this bit.
  * @cap_rss_sym_xor_supported: indicates if the driver supports symmetric-xor
  *	RSS.
  * @rxfh_indir_space: max size of RSS indirection tables, if indirection table
net/ethtool/ioctl.c +4 -2
index 8ca13208d240..52dfb07393a6 100644
@@ -1227,7 +1227,8 @@ static noinline_for_stack int ethtool_get_rxfh(struct net_device *dev,
 	if (rxfh.rsvd8[0] || rxfh.rsvd8[1] || rxfh.rsvd32)
 		return -EINVAL;
 	/* Most drivers don't handle rss_context, check it's 0 as well */
-	if (rxfh.rss_context && !ops->cap_rss_ctx_supported)
+	if (rxfh.rss_context && !(ops->cap_rss_ctx_supported ||
+				  ops->create_rxfh_context))
 		return -EOPNOTSUPP;
 
 	rxfh.indir_size = rxfh_dev.indir_size;
@@ -1357,7 +1358,8 @@ static noinline_for_stack int ethtool_set_rxfh(struct net_device *dev,
 	if (rxfh.rsvd8[0] || rxfh.rsvd8[1] || rxfh.rsvd32)
 		return -EINVAL;
 	/* Most drivers don't handle rss_context, check it's 0 as well */
-	if (rxfh.rss_context && !ops->cap_rss_ctx_supported)
+	if (rxfh.rss_context && !(ops->cap_rss_ctx_supported ||
+				  ops->create_rxfh_context))
 		return -EOPNOTSUPP;
 	/* Check input data transformation capabilities */
 	if (rxfh.input_xfrm && rxfh.input_xfrm != RXH_XFRM_SYM_XOR &&
net/ethtool/rss.c +2 -1
index 5c4c4505ab9a..a06bdac8b8a2 100644
@@ -60,7 +60,8 @@ rss_prepare_data(const struct ethnl_req_info *req_base,
 		return -EOPNOTSUPP;
 
 	/* Some drivers don't handle rss_context */
-	if (request->rss_context && !ops->cap_rss_ctx_supported)
+	if (request->rss_context && !(ops->cap_rss_ctx_supported ||
+				      ops->create_rxfh_context))
 		return -EOPNOTSUPP;
 
 	ret = ethnl_ops_begin(dev);
Signature · 1 line
2.45.2
author maintainer
#6 lore raw

PATCH v3 5/12 eth: remove .cap_rss_ctx_supported from updated drivers

In reply to Jakub Kicinski

Remove .cap_rss_ctx_supported from drivers which moved to the new API. This makes it easy to grep for drivers which still need to be converted.

---
 drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c | 1 -
 drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c   | 1 -
 drivers/net/ethernet/sfc/ef100_ethtool.c          | 1 -
 drivers/net/ethernet/sfc/ethtool.c                | 1 -
 4 files changed, 4 deletions(-)
drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c +0 -1
index ab8e3f197e7b..33e8cf0a3764 100644
@@ -5289,7 +5289,6 @@ void bnxt_ethtool_free(struct bnxt *bp)
 
 const struct ethtool_ops bnxt_ethtool_ops = {
 	.cap_link_lanes_supported	= 1,
-	.cap_rss_ctx_supported		= 1,
 	.rxfh_max_context_id		= BNXT_MAX_ETH_RSS_CTX,
 	.rxfh_indir_space		= BNXT_MAX_RSS_TABLE_ENTRIES_P5,
 	.rxfh_priv_size			= sizeof(struct bnxt_rss_ctx),
drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c +0 -1
index b70aca33b928..98500b70f994 100644
@@ -5791,7 +5791,6 @@ static const struct net_device_ops mvpp2_netdev_ops = {
 };
 
 static const struct ethtool_ops mvpp2_eth_tool_ops = {
-	.cap_rss_ctx_supported	= true,
 	.rxfh_max_context_id	= MVPP22_N_RSS_TABLES - 1,
 	.supported_coalesce_params = ETHTOOL_COALESCE_USECS |
 				     ETHTOOL_COALESCE_MAX_FRAMES,
drivers/net/ethernet/sfc/ef100_ethtool.c +0 -1
index 896ffca4aee2..746b5314acb5 100644
@@ -37,7 +37,6 @@ ef100_ethtool_get_ringparam(struct net_device *net_dev,
 /*	Ethtool options available
  */
 const struct ethtool_ops ef100_ethtool_ops = {
-	.cap_rss_ctx_supported	= true,
 	.get_drvinfo		= efx_ethtool_get_drvinfo,
 	.get_msglevel		= efx_ethtool_get_msglevel,
 	.set_msglevel		= efx_ethtool_set_msglevel,
drivers/net/ethernet/sfc/ethtool.c +0 -1
index 7c887160e2ef..15245720c949 100644
@@ -240,7 +240,6 @@ static int efx_ethtool_get_ts_info(struct net_device *net_dev,
 }
 
 const struct ethtool_ops efx_ethtool_ops = {
-	.cap_rss_ctx_supported	= true,
 	.supported_coalesce_params = ETHTOOL_COALESCE_USECS |
 				     ETHTOOL_COALESCE_USECS_IRQ |
 				     ETHTOOL_COALESCE_USE_ADAPTIVE_RX,
Signature · 1 line
2.45.2
author maintainer
#7 lore raw

PATCH v3 6/12 ethtool: rss: don't report key if device doesn't support it

In reply to Jakub Kicinski

marvell/otx2 and mvpp2 do not support setting different keys for different RSS contexts. Contexts have separate indirection tables but key is shared with all other contexts. This is likely fine, indirection table is the most important piece.

Don't report the key-related parameters from such drivers. This prevents driver-errors, e.g. otx2 always writes the main key, even when user asks to change per-context key. The second reason is that without this change tracking the keys by the core gets complicated. Even if the driver correctly reject setting key with rss_context != 0, change of the main key would have to be reflected in the XArray for all additional contexts.

Since the additional contexts don't have their own keys not including the attributes (in Netlink speak) seems intuitive. ethtool CLI seems to deal with it just fine.

Having to set the flag in majority of the drivers is a bit tedious but not reporting the key is a safer default.

---
v3:
- add safety note to the commit message
- use true in sfc, drop sienna
- extend kdoc of @rxfh_per_ctx_key

---
 .../net/ethernet/broadcom/bnxt/bnxt_ethtool.c |  1 +
 drivers/net/ethernet/intel/ice/ice_ethtool.c  |  1 +
 .../ethernet/mellanox/mlx5/core/en_ethtool.c  |  1 +
 drivers/net/ethernet/sfc/ef100_ethtool.c      |  1 +
 drivers/net/ethernet/sfc/ethtool.c            |  1 +
 include/linux/ethtool.h                       |  4 +++
 net/ethtool/ioctl.c                           | 25 ++++++++++++++++---
 net/ethtool/rss.c                             | 21 +++++++++++-----
 8 files changed, 45 insertions(+), 10 deletions(-)
drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c +1 -0
index 33e8cf0a3764..77621ccfff5e 100644
@@ -5289,6 +5289,7 @@ void bnxt_ethtool_free(struct bnxt *bp)
 
 const struct ethtool_ops bnxt_ethtool_ops = {
 	.cap_link_lanes_supported	= 1,
+	.rxfh_per_ctx_key		= 1,
 	.rxfh_max_context_id		= BNXT_MAX_ETH_RSS_CTX,
 	.rxfh_indir_space		= BNXT_MAX_RSS_TABLE_ENTRIES_P5,
 	.rxfh_priv_size			= sizeof(struct bnxt_rss_ctx),
drivers/net/ethernet/intel/ice/ice_ethtool.c +1 -0
index 8c990c976132..b5b57926cdc0 100644
@@ -4725,6 +4725,7 @@ static const struct ethtool_ops ice_ethtool_ops = {
 				     ETHTOOL_COALESCE_USE_ADAPTIVE |
 				     ETHTOOL_COALESCE_RX_USECS_HIGH,
 	.cap_rss_sym_xor_supported = true,
+	.rxfh_per_ctx_key	= true,
 	.get_link_ksettings	= ice_get_link_ksettings,
 	.set_link_ksettings	= ice_set_link_ksettings,
 	.get_fec_stats		= ice_get_fec_stats,
drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c +1 -0
index 0b941482db30..2d514210aaec 100644
@@ -2593,6 +2593,7 @@ static void mlx5e_get_ts_stats(struct net_device *netdev,
 
 const struct ethtool_ops mlx5e_ethtool_ops = {
 	.cap_rss_ctx_supported	= true,
+	.rxfh_per_ctx_key	= true,
 	.supported_coalesce_params = ETHTOOL_COALESCE_USECS |
 				     ETHTOOL_COALESCE_MAX_FRAMES |
 				     ETHTOOL_COALESCE_USE_ADAPTIVE |
drivers/net/ethernet/sfc/ef100_ethtool.c +1 -0
index 746b5314acb5..5c2551369812 100644
@@ -58,6 +58,7 @@ const struct ethtool_ops ef100_ethtool_ops = {
 
 	.get_rxfh_indir_size	= efx_ethtool_get_rxfh_indir_size,
 	.get_rxfh_key_size	= efx_ethtool_get_rxfh_key_size,
+	.rxfh_per_ctx_key	= true,
 	.rxfh_priv_size		= sizeof(struct efx_rss_context_priv),
 	.get_rxfh		= efx_ethtool_get_rxfh,
 	.set_rxfh		= efx_ethtool_set_rxfh,
drivers/net/ethernet/sfc/ethtool.c +1 -0
index 15245720c949..848b1923133a 100644
@@ -267,6 +267,7 @@ const struct ethtool_ops efx_ethtool_ops = {
 	.set_rxnfc		= efx_ethtool_set_rxnfc,
 	.get_rxfh_indir_size	= efx_ethtool_get_rxfh_indir_size,
 	.get_rxfh_key_size	= efx_ethtool_get_rxfh_key_size,
+	.rxfh_per_ctx_key	= true,
 	.rxfh_priv_size		= sizeof(struct efx_rss_context_priv),
 	.get_rxfh		= efx_ethtool_get_rxfh,
 	.set_rxfh		= efx_ethtool_set_rxfh,
include/linux/ethtool.h +4 -0
index 55c9f613ab64..8c89dc33d51c 100644
@@ -731,6 +731,9 @@ struct kernel_ethtool_ts_info {
  *	do not have to set this bit.
  * @cap_rss_sym_xor_supported: indicates if the driver supports symmetric-xor
  *	RSS.
+ * @rxfh_per_ctx_key: device supports setting different RSS key for each
+ *	additional context. Netlink API should report hfunc, key, and input_xfrm
+ *	for every context, not just context 0.
  * @rxfh_indir_space: max size of RSS indirection tables, if indirection table
  *	size as returned by @get_rxfh_indir_size may change during lifetime
  *	of the device. Leave as 0 if the table size is constant.
@@ -952,6 +955,7 @@ struct ethtool_ops {
 	u32     cap_link_lanes_supported:1;
 	u32     cap_rss_ctx_supported:1;
 	u32	cap_rss_sym_xor_supported:1;
+	u32	rxfh_per_ctx_key:1;
 	u32	rxfh_indir_space;
 	u16	rxfh_key_space;
 	u16	rxfh_priv_size;
net/ethtool/ioctl.c +21 -4
index 52dfb07393a6..e32b791f8d1c 100644
@@ -1261,10 +1261,15 @@ static noinline_for_stack int ethtool_get_rxfh(struct net_device *dev,
 		if (rxfh_dev.indir)
 			memcpy(rxfh_dev.indir, ethtool_rxfh_context_indir(ctx),
 			       indir_bytes);
-		if (rxfh_dev.key)
-			memcpy(rxfh_dev.key, ethtool_rxfh_context_key(ctx),
-			       user_key_size);
-		rxfh_dev.hfunc = ctx->hfunc;
+		if (!ops->rxfh_per_ctx_key) {
+			rxfh_dev.key_size = 0;
+		} else {
+			if (rxfh_dev.key)
+				memcpy(rxfh_dev.key,
+				       ethtool_rxfh_context_key(ctx),
+				       user_key_size);
+			rxfh_dev.hfunc = ctx->hfunc;
+		}
 		rxfh_dev.input_xfrm = ctx->input_xfrm;
 		ret = 0;
 	} else {
@@ -1281,6 +1286,11 @@ static noinline_for_stack int ethtool_get_rxfh(struct net_device *dev,
 				&rxfh_dev.input_xfrm,
 				sizeof(rxfh.input_xfrm))) {
 		ret = -EFAULT;
+	} else if (copy_to_user(useraddr +
+				offsetof(struct ethtool_rxfh, key_size),
+				&rxfh_dev.key_size,
+				sizeof(rxfh.key_size))) {
+		ret = -EFAULT;
 	} else if (copy_to_user(useraddr +
 			      offsetof(struct ethtool_rxfh, rss_config[0]),
 			      rss_config, total_size)) {
@@ -1386,6 +1396,13 @@ static noinline_for_stack int ethtool_set_rxfh(struct net_device *dev,
 
 	indir_bytes = dev_indir_size * sizeof(rxfh_dev.indir[0]);
 
+	/* Check settings which may be global rather than per RSS-context */
+	if (rxfh.rss_context && !ops->rxfh_per_ctx_key)
+		if (rxfh.key_size ||
+		    (rxfh.hfunc && rxfh.hfunc != ETH_RSS_HASH_NO_CHANGE) ||
+		    (rxfh.input_xfrm && rxfh.input_xfrm != RXH_XFRM_NO_CHANGE))
+			return -EOPNOTSUPP;
+
 	rss_config = kzalloc(indir_bytes + dev_key_size, GFP_USER);
 	if (!rss_config)
 		return -ENOMEM;
net/ethtool/rss.c +15 -6
index a06bdac8b8a2..cd8100d81919 100644
@@ -10,6 +10,7 @@ struct rss_req_info {
 
 struct rss_reply_data {
 	struct ethnl_reply_data		base;
+	bool				no_key_fields;
 	u32				indir_size;
 	u32				hkey_size;
 	u32				hfunc;
@@ -60,9 +61,12 @@ rss_prepare_data(const struct ethnl_req_info *req_base,
 		return -EOPNOTSUPP;
 
 	/* Some drivers don't handle rss_context */
-	if (request->rss_context && !(ops->cap_rss_ctx_supported ||
-				      ops->create_rxfh_context))
-		return -EOPNOTSUPP;
+	if (request->rss_context) {
+		if (!ops->cap_rss_ctx_supported && !ops->create_rxfh_context)
+			return -EOPNOTSUPP;
+
+		data->no_key_fields = !ops->rxfh_per_ctx_key;
+	}
 
 	ret = ethnl_ops_begin(dev);
 	if (ret < 0)
@@ -132,13 +136,18 @@ rss_fill_reply(struct sk_buff *skb, const struct ethnl_req_info *req_base,
 	    nla_put_u32(skb, ETHTOOL_A_RSS_CONTEXT, request->rss_context))
 		return -EMSGSIZE;
 
+	if ((data->indir_size &&
+	     nla_put(skb, ETHTOOL_A_RSS_INDIR,
+		     sizeof(u32) * data->indir_size, data->indir_table)))
+		return -EMSGSIZE;
+
+	if (data->no_key_fields)
+		return 0;
+
 	if ((data->hfunc &&
 	     nla_put_u32(skb, ETHTOOL_A_RSS_HFUNC, data->hfunc)) ||
 	    (data->input_xfrm &&
 	     nla_put_u32(skb, ETHTOOL_A_RSS_INPUT_XFRM, data->input_xfrm)) ||
-	    (data->indir_size &&
-	     nla_put(skb, ETHTOOL_A_RSS_INDIR,
-		     sizeof(u32) * data->indir_size, data->indir_table)) ||
 	    (data->hkey_size &&
 	     nla_put(skb, ETHTOOL_A_RSS_HKEY, data->hkey_size, data->hkey)))
 		return -EMSGSIZE;
Signature · 1 line
2.45.2
maintainer Reviewed-by on v3 6/12
#8 lore raw

Re: [PATCH net-next v3 06/12] ethtool: rss: don't report key if device doesn't support it

Quoted 19 lines · Jakub Kicinski

marvell/otx2 and mvpp2 do not support setting different keys for different RSS contexts. Contexts have separate indirection tables but key is shared with all other contexts. This is likely fine, indirection table is the most important piece.

Don't report the key-related parameters from such drivers. This prevents driver-errors, e.g. otx2 always writes the main key, even when user asks to change per-context key. The second reason is that without this change tracking the keys by the core gets complicated. Even if the driver correctly reject setting key with rss_context != 0, change of the main key would have to be reflected in the XArray for all additional contexts.

Since the additional contexts don't have their own keys not including the attributes (in Netlink speak) seems intuitive. ethtool CLI seems to deal with it just fine.

Quoted 6 lines

Having to set the flag in majority of the drivers is a bit tedious but not reporting the key is a safer default.

author maintainer
#9 lore raw

PATCH v3 7/12 ethtool: rss: move the device op invocation out of rss_prepare_data()

In reply to Jakub Kicinski

Factor calling device ops out of rss_prepare_data(). Next patch will add alternative path using xarray. No functional changes.

---
 net/ethtool/rss.c | 43 +++++++++++++++++++++++++++----------------
 1 file changed, 27 insertions(+), 16 deletions(-)
net/ethtool/rss.c +27 -16
index cd8100d81919..5c477cc36251 100644
@@ -43,13 +43,9 @@ rss_parse_request(struct ethnl_req_info *req_info, struct nlattr **tb,
 }
 
 static int
-rss_prepare_data(const struct ethnl_req_info *req_base,
-		 struct ethnl_reply_data *reply_base,
-		 const struct genl_info *info)
+rss_prepare_get(const struct rss_req_info *request, struct net_device *dev,
+		struct rss_reply_data *data, const struct genl_info *info)
 {
-	struct rss_reply_data *data = RSS_REPDATA(reply_base);
-	struct rss_req_info *request = RSS_REQINFO(req_base);
-	struct net_device *dev = reply_base->dev;
 	struct ethtool_rxfh_param rxfh = {};
 	const struct ethtool_ops *ops;
 	u32 total_size, indir_bytes;
@@ -57,16 +53,6 @@ rss_prepare_data(const struct ethnl_req_info *req_base,
 	int ret;
 
 	ops = dev->ethtool_ops;
-	if (!ops->get_rxfh)
-		return -EOPNOTSUPP;
-
-	/* Some drivers don't handle rss_context */
-	if (request->rss_context) {
-		if (!ops->cap_rss_ctx_supported && !ops->create_rxfh_context)
-			return -EOPNOTSUPP;
-
-		data->no_key_fields = !ops->rxfh_per_ctx_key;
-	}
 
 	ret = ethnl_ops_begin(dev);
 	if (ret < 0)
@@ -109,6 +95,31 @@ rss_prepare_data(const struct ethnl_req_info *req_base,
 	return ret;
 }
 
+static int
+rss_prepare_data(const struct ethnl_req_info *req_base,
+		 struct ethnl_reply_data *reply_base,
+		 const struct genl_info *info)
+{
+	struct rss_reply_data *data = RSS_REPDATA(reply_base);
+	struct rss_req_info *request = RSS_REQINFO(req_base);
+	struct net_device *dev = reply_base->dev;
+	const struct ethtool_ops *ops;
+
+	ops = dev->ethtool_ops;
+	if (!ops->get_rxfh)
+		return -EOPNOTSUPP;
+
+	/* Some drivers don't handle rss_context */
+	if (request->rss_context) {
+		if (!ops->cap_rss_ctx_supported && !ops->create_rxfh_context)
+			return -EOPNOTSUPP;
+
+		data->no_key_fields = !ops->rxfh_per_ctx_key;
+	}
+
+	return rss_prepare_get(request, dev, data, info);
+}
+
 static int
 rss_reply_size(const struct ethnl_req_info *req_base,
 	       const struct ethnl_reply_data *reply_base)
Signature · 1 line
2.45.2
author maintainer
#10 lore raw

PATCH v3 8/12 ethtool: rss: report info about additional contexts from XArray

In reply to Jakub Kicinski

IOCTL already uses the XArray when reporting info about additional contexts. Do the same thing in netlink code.

---
 net/ethtool/rss.c | 37 ++++++++++++++++++++++++++++++++++++-
 1 file changed, 36 insertions(+), 1 deletion(-)
net/ethtool/rss.c +36 -1
index 5c477cc36251..023782ca1230 100644
@@ -82,7 +82,6 @@ rss_prepare_get(const struct rss_req_info *request, struct net_device *dev,
 	rxfh.indir = data->indir_table;
 	rxfh.key_size = data->hkey_size;
 	rxfh.key = data->hkey;
-	rxfh.rss_context = request->rss_context;
 
 	ret = ops->get_rxfh(dev, &rxfh);
 	if (ret)
@@ -95,6 +94,41 @@ rss_prepare_get(const struct rss_req_info *request, struct net_device *dev,
 	return ret;
 }
 
+static int
+rss_prepare_ctx(const struct rss_req_info *request, struct net_device *dev,
+		struct rss_reply_data *data, const struct genl_info *info)
+{
+	struct ethtool_rxfh_context *ctx;
+	u32 total_size, indir_bytes;
+	u8 *rss_config;
+
+	ctx = xa_load(&dev->ethtool->rss_ctx, request->rss_context);
+	if (!ctx)
+		return -ENOENT;
+
+	data->indir_size = ctx->indir_size;
+	data->hkey_size = ctx->key_size;
+	data->hfunc = ctx->hfunc;
+	data->input_xfrm = ctx->input_xfrm;
+
+	indir_bytes = data->indir_size * sizeof(u32);
+	total_size = indir_bytes + data->hkey_size;
+	rss_config = kzalloc(total_size, GFP_KERNEL);
+	if (!rss_config)
+		return -ENOMEM;
+
+	data->indir_table = (u32 *)rss_config;
+	memcpy(data->indir_table, ethtool_rxfh_context_indir(ctx), indir_bytes);
+
+	if (data->hkey_size) {
+		data->hkey = rss_config + indir_bytes;
+		memcpy(data->hkey, ethtool_rxfh_context_key(ctx),
+		       data->hkey_size);
+	}
+
+	return 0;
+}
+
 static int
 rss_prepare_data(const struct ethnl_req_info *req_base,
 		 struct ethnl_reply_data *reply_base,
@@ -115,6 +149,7 @@ rss_prepare_data(const struct ethnl_req_info *req_base,
 			return -EOPNOTSUPP;
 
 		data->no_key_fields = !ops->rxfh_per_ctx_key;
+		return rss_prepare_ctx(request, dev, data, info);
 	}
 
 	return rss_prepare_get(request, dev, data, info);
Signature · 1 line
2.45.2
author maintainer
#11 lore raw

PATCH v3 9/12 ethtool: rss: support dumping RSS contexts

In reply to Jakub Kicinski

Now that we track RSS contexts in the core we can easily dump them. This is a major introspection improvement, as previously the only way to find all contexts would be to try all ids (of which there may be 2^32 - 1).

Don't use the XArray iterators (like xa_for_each_start()) as they do not move the index past the end of the array once done, which caused multiple bugs in Netlink dumps in the past.

---
v3:
- one_ifindex -> match_ifindex, add a comment
- add comment about context 0 not being in XArray
- extend commit message (XArray iterators)

---
 Documentation/netlink/specs/ethtool.yaml |   9 +-
 net/ethtool/netlink.c                    |   2 +
 net/ethtool/netlink.h                    |   2 +
 net/ethtool/rss.c                        | 135 +++++++++++++++++++++++
 4 files changed, 146 insertions(+), 2 deletions(-)
Documentation/netlink/specs/ethtool.yaml +7 -2
index ea21fe135b97..cf69eedae51d 100644
@@ -1749,12 +1749,12 @@ doc: Partial family for Ethtool Netlink.
 
       attribute-set: rss
 
-      do: &rss-get-op
+      do:
         request:
           attributes:
             - header
             - context
-        reply:
+        reply: &rss-reply
           attributes:
             - header
             - context
@@ -1762,6 +1762,11 @@ doc: Partial family for Ethtool Netlink.
             - indir
             - hkey
             - input_xfrm
+      dump:
+        request:
+          attributes:
+            - header
+        reply: *rss-reply
     -
       name: plca-get-cfg
       doc: Get PLCA params.
net/ethtool/netlink.c +2 -0
index cb1eea00e349..041548e5f5e6 100644
@@ -1128,6 +1128,8 @@ static const struct genl_ops ethtool_genl_ops[] = {
 	{
 		.cmd	= ETHTOOL_MSG_RSS_GET,
 		.doit	= ethnl_default_doit,
+		.start	= ethnl_rss_dump_start,
+		.dumpit	= ethnl_rss_dumpit,
 		.policy = ethnl_rss_get_policy,
 		.maxattr = ARRAY_SIZE(ethnl_rss_get_policy) - 1,
 	},
net/ethtool/netlink.h +2 -0
index 46ec273a87c5..919371383b23 100644
@@ -464,6 +464,8 @@ int ethnl_tunnel_info_doit(struct sk_buff *skb, struct genl_info *info);
 int ethnl_tunnel_info_start(struct netlink_callback *cb);
 int ethnl_tunnel_info_dumpit(struct sk_buff *skb, struct netlink_callback *cb);
 int ethnl_act_module_fw_flash(struct sk_buff *skb, struct genl_info *info);
+int ethnl_rss_dump_start(struct netlink_callback *cb);
+int ethnl_rss_dumpit(struct sk_buff *skb, struct netlink_callback *cb);
 
 extern const char stats_std_names[__ETHTOOL_STATS_CNT][ETH_GSTRING_LEN];
 extern const char stats_eth_phy_names[__ETHTOOL_A_STATS_ETH_PHY_CNT][ETH_GSTRING_LEN];
net/ethtool/rss.c +135 -0

141 lines not on this page · raw message

Signature · 1 line
2.45.2
maintainer Reviewed-by on v3 9/12
#12 lore raw

Re: [PATCH net-next v3 09/12] ethtool: rss: support dumping RSS contexts

Quoted 11 lines · Jakub Kicinski

Now that we track RSS contexts in the core we can easily dump them. This is a major introspection improvement, as previously the only way to find all contexts would be to try all ids (of which there may be 2^32 - 1).

Don't use the XArray iterators (like xa_for_each_start()) as they do not move the index past the end of the array once done, which caused multiple bugs in Netlink dumps in the past.

author maintainer
#13 lore raw

PATCH v3 10/12 ethtool: rss: support skipping contexts during dump

In reply to Jakub Kicinski

Applications may want to deal with dynamic RSS contexts only. So dumping context 0 will be counter-productive for them. Support starting the dump from a given context ID.

Alternative would be to implement a dump flag to skip just context 0, not sure which is better...

---
 Documentation/netlink/specs/ethtool.yaml     |  4 ++++
 Documentation/networking/ethtool-netlink.rst | 12 ++++++++++--
 include/uapi/linux/ethtool_netlink.h         |  1 +
 net/ethtool/netlink.h                        |  2 +-
 net/ethtool/rss.c                            | 12 +++++++++++-
 5 files changed, 27 insertions(+), 4 deletions(-)
Documentation/netlink/specs/ethtool.yaml +4 -0
index cf69eedae51d..4c2334c213b0 100644
@@ -1028,6 +1028,9 @@ doc: Partial family for Ethtool Netlink.
       -
         name: input_xfrm
         type: u32
+      -
+        name: start-context
+        type: u32
   -
     name: plca
     attributes:
@@ -1766,6 +1769,7 @@ doc: Partial family for Ethtool Netlink.
         request:
           attributes:
             - header
+            - start-context
         reply: *rss-reply
     -
       name: plca-get-cfg
Documentation/networking/ethtool-netlink.rst +10 -2
index d5f246aceb9f..82c5542c80ce 100644
@@ -1866,10 +1866,18 @@ RSS context of an interface similar to ``ETHTOOL_GRSSH`` ioctl request.
 
 Request contents:
 
-=====================================  ======  ==========================
+=====================================  ======  ============================
   ``ETHTOOL_A_RSS_HEADER``             nested  request header
   ``ETHTOOL_A_RSS_CONTEXT``            u32     context number
-=====================================  ======  ==========================
+  ``ETHTOOL_A_RSS_START_CONTEXT``      u32     start context number (dumps)
+=====================================  ======  ============================
+
+``ETHTOOL_A_RSS_CONTEXT`` specifies which RSS context number to query,
+if not set context 0 (the main context) is queried. Dumps can be filtered
+by device (only listing contexts of a given netdev). Filtering single
+context number is not supported but ``ETHTOOL_A_RSS_START_CONTEXT``
+can be used to start dumping context from the given number (primarily
+used to ignore context 0s and only dump additional contexts).
 
 Kernel response contents:
 
include/uapi/linux/ethtool_netlink.h +1 -0
index 6d5bdcc67631..93c57525a975 100644
@@ -965,6 +965,7 @@ enum {
 	ETHTOOL_A_RSS_INDIR,		/* binary */
 	ETHTOOL_A_RSS_HKEY,		/* binary */
 	ETHTOOL_A_RSS_INPUT_XFRM,	/* u32 */
+	ETHTOOL_A_RSS_START_CONTEXT,	/* u32 */
 
 	__ETHTOOL_A_RSS_CNT,
 	ETHTOOL_A_RSS_MAX = (__ETHTOOL_A_RSS_CNT - 1),
net/ethtool/netlink.h +1 -1
index 919371383b23..236c189fc968 100644
@@ -449,7 +449,7 @@ extern const struct nla_policy ethnl_module_get_policy[ETHTOOL_A_MODULE_HEADER +
 extern const struct nla_policy ethnl_module_set_policy[ETHTOOL_A_MODULE_POWER_MODE_POLICY + 1];
 extern const struct nla_policy ethnl_pse_get_policy[ETHTOOL_A_PSE_HEADER + 1];
 extern const struct nla_policy ethnl_pse_set_policy[ETHTOOL_A_PSE_MAX + 1];
-extern const struct nla_policy ethnl_rss_get_policy[ETHTOOL_A_RSS_CONTEXT + 1];
+extern const struct nla_policy ethnl_rss_get_policy[ETHTOOL_A_RSS_START_CONTEXT + 1];
 extern const struct nla_policy ethnl_plca_get_cfg_policy[ETHTOOL_A_PLCA_HEADER + 1];
 extern const struct nla_policy ethnl_plca_set_cfg_policy[ETHTOOL_A_PLCA_MAX + 1];
 extern const struct nla_policy ethnl_plca_get_status_policy[ETHTOOL_A_PLCA_HEADER + 1];
net/ethtool/rss.c +11 -1
index 2865720ff583..e07386275e14 100644
@@ -28,6 +28,7 @@ struct rss_reply_data {
 const struct nla_policy ethnl_rss_get_policy[] = {
 	[ETHTOOL_A_RSS_HEADER] = NLA_POLICY_NESTED(ethnl_header_policy),
 	[ETHTOOL_A_RSS_CONTEXT] = { .type = NLA_U32 },
+	[ETHTOOL_A_RSS_START_CONTEXT] = { .type = NLA_U32 },
 };
 
 static int
@@ -38,6 +39,10 @@ rss_parse_request(struct ethnl_req_info *req_info, struct nlattr **tb,
 
 	if (tb[ETHTOOL_A_RSS_CONTEXT])
 		request->rss_context = nla_get_u32(tb[ETHTOOL_A_RSS_CONTEXT]);
+	if (tb[ETHTOOL_A_RSS_START_CONTEXT]) {
+		NL_SET_BAD_ATTR(extack, tb[ETHTOOL_A_RSS_START_CONTEXT]);
+		return -EINVAL;
+	}
 
 	return 0;
 }
@@ -214,6 +219,7 @@ struct rss_nl_dump_ctx {
 
 	/* User wants to only dump contexts from given ifindex */
 	unsigned int		match_ifindex;
+	unsigned int		start_ctx;
 };
 
 static struct rss_nl_dump_ctx *rss_dump_ctx(struct netlink_callback *cb)
@@ -236,6 +242,10 @@ int ethnl_rss_dump_start(struct netlink_callback *cb)
 		NL_SET_BAD_ATTR(info->extack, tb[ETHTOOL_A_RSS_CONTEXT]);
 		return -EINVAL;
 	}
+	if (tb[ETHTOOL_A_RSS_START_CONTEXT]) {
+		ctx->start_ctx = nla_get_u32(tb[ETHTOOL_A_RSS_START_CONTEXT]);
+		ctx->ctx_idx = ctx->start_ctx;
+	}
 
 	ret = ethnl_parse_header_dev_get(&req_info,
 					 tb[ETHTOOL_A_RSS_HEADER],
@@ -317,7 +327,7 @@ rss_dump_one_dev(struct sk_buff *skb, struct netlink_callback *cb,
 		if (ret)
 			return ret;
 	}
-	ctx->ctx_idx = 0;
+	ctx->ctx_idx = ctx->start_ctx;
 
 	return 0;
 }
Signature · 1 line
2.45.2
author maintainer
#14 lore raw

PATCH v3 11/12 netlink: specs: decode indirection table as u32 array

In reply to Jakub Kicinski

Indirection table is dumped as a raw u32 array, decode it. It's tempting to decode hash key, too, but it is an actual bitstream, so leave it be for now.

---
 Documentation/netlink/specs/ethtool.yaml | 1 +
 1 file changed, 1 insertion(+)
Documentation/netlink/specs/ethtool.yaml +1 -0
index 4c2334c213b0..1bbeaba5c644 100644
@@ -1022,6 +1022,7 @@ doc: Partial family for Ethtool Netlink.
       -
         name: indir
         type: binary
+        sub-type: u32
       -
         name: hkey
         type: binary
Signature · 1 line
2.45.2
Reviewed-by on v3 11/12
#15 lore raw

Re: [PATCH net-next v3 11/12] netlink: specs: decode indirection table as u32 array

Quoted 24 lines · Jakub Kicinski

Indirection table is dumped as a raw u32 array, decode it. It's tempting to decode hash key, too, but it is an actual bitstream, so leave it be for now.

---
 Documentation/netlink/specs/ethtool.yaml | 1 +
 1 file changed, 1 insertion(+)
Documentation/netlink/specs/ethtool.yaml +1 -0
index 4c2334c213b0..1bbeaba5c644 100644
@@ -1022,6 +1022,7 @@ doc: Partial family for Ethtool Netlink.
       -
         name: indir
         type: binary
+        sub-type: u32
       -
         name: hkey
         type: binary
Signature · 1 line
2.45.2
author maintainer
#16 lore raw

PATCH v3 12/12 selftests: drv-net: rss_ctx: test dumping RSS contexts

In reply to Jakub Kicinski

Add a test for dumping RSS contexts. Make sure indir table and key are sane when contexts are created with various combination of inputs. Test the dump filtering by ifname and start-context.

---
v3:
- filter results by ifindex
- check for duplicates

---
 .../selftests/drivers/net/hw/rss_ctx.py       | 76 ++++++++++++++++++-
 tools/testing/selftests/net/lib/py/ksft.py    |  6 ++
 2 files changed, 80 insertions(+), 2 deletions(-)
tools/testing/selftests/drivers/net/hw/rss_ctx.py +74 -2
index 1da6b214f4fe..9d7adb3cf33b 100755
@@ -3,7 +3,7 @@
 
 import datetime
 import random
-from lib.py import ksft_run, ksft_pr, ksft_exit, ksft_eq, ksft_ge, ksft_lt
+from lib.py import ksft_run, ksft_pr, ksft_exit, ksft_eq, ksft_ne, ksft_ge, ksft_lt
 from lib.py import NetDrvEpEnv
 from lib.py import EthtoolFamily, NetdevFamily
 from lib.py import KsftSkipEx
@@ -302,6 +302,78 @@ from lib.py import ethtool, ip, defer, GenerateTraffic, CmdExitFailure
     ksft_eq(carrier1 - carrier0, 0)
 
 
+def test_rss_context_dump(cfg):
+    """
+    Test dumping RSS contexts. This tests mostly exercises the kernel APIs.
+    """
+
+    # Get a random key of the right size
+    data = get_rss(cfg)
+    if 'rss-hash-key' in data:
+        key_data = _rss_key_rand(len(data['rss-hash-key']))
+        key = _rss_key_str(key_data)
+    else:
+        key_data = []
+        key = "ba:ad"
+
+    ids = []
+    try:
+        ids.append(ethtool_create(cfg, "-X", f"context new"))
+        defer(ethtool, f"-X {cfg.ifname} context {ids[-1]} delete")
+
+        ids.append(ethtool_create(cfg, "-X", f"context new weight 1 1"))
+        defer(ethtool, f"-X {cfg.ifname} context {ids[-1]} delete")
+
+        ids.append(ethtool_create(cfg, "-X", f"context new hkey {key}"))
+        defer(ethtool, f"-X {cfg.ifname} context {ids[-1]} delete")
+    except CmdExitFailure:
+        if not ids:
+            raise KsftSkipEx("Unable to add any contexts")
+        ksft_pr(f"Added only {len(ids)} out of 3 contexts")
+
+    expect_tuples = set([(cfg.ifname, -1)] + [(cfg.ifname, ctx_id) for ctx_id in ids])
+
+    # Dump all
+    ctxs = cfg.ethnl.rss_get({}, dump=True)
+    tuples = [(c['header']['dev-name'], c.get('context', -1)) for c in ctxs]
+    ksft_eq(len(tuples), len(set(tuples)), "duplicates in context dump")
+    ctx_tuples = set([ctx for ctx in tuples if ctx[0] == cfg.ifname])
+    ksft_eq(expect_tuples, ctx_tuples)
+
+    # Sanity-check the results
+    for data in ctxs:
+        ksft_ne(set(data['indir']), {0}, "indir table is all zero")
+        ksft_ne(set(data.get('hkey', [1])), {0}, "key is all zero")
+
+        # More specific checks
+        if len(ids) > 1 and data.get('context') == ids[1]:
+            ksft_eq(set(data['indir']), {0, 1},
+                    "ctx1 - indir table mismatch")
+        if len(ids) > 2 and data.get('context') == ids[2]:
+            ksft_eq(data['hkey'], bytes(key_data), "ctx2 - key mismatch")
+
+    # Ifindex filter
+    ctxs = cfg.ethnl.rss_get({'header': {'dev-name': cfg.ifname}}, dump=True)
+    tuples = [(c['header']['dev-name'], c.get('context', -1)) for c in ctxs]
+    ctx_tuples = set(tuples)
+    ksft_eq(len(tuples), len(ctx_tuples), "duplicates in context dump")
+    ksft_eq(expect_tuples, ctx_tuples)
+
+    # Skip ctx 0
+    expect_tuples.remove((cfg.ifname, -1))
+
+    ctxs = cfg.ethnl.rss_get({'start-context': 1}, dump=True)
+    tuples = [(c['header']['dev-name'], c.get('context', -1)) for c in ctxs]
+    ksft_eq(len(tuples), len(set(tuples)), "duplicates in context dump")
+    ctx_tuples = set([ctx for ctx in tuples if ctx[0] == cfg.ifname])
+    ksft_eq(expect_tuples, ctx_tuples)
+
+    # And finally both with ifindex and skip main
+    ctxs = cfg.ethnl.rss_get({'header': {'dev-name': cfg.ifname}, 'start-context': 1}, dump=True)
+    ctx_tuples = set([(c['header']['dev-name'], c.get('context', -1)) for c in ctxs])
+    ksft_eq(expect_tuples, ctx_tuples)
+
+
 def test_rss_context(cfg, ctx_cnt=1, create_with_cfg=None):
     """
     Test separating traffic into RSS contexts.
@@ -542,7 +614,7 @@ from lib.py import ethtool, ip, defer, GenerateTraffic, CmdExitFailure
         ksft_run([test_rss_key_indir, test_rss_queue_reconfigure,
                   test_rss_resize, test_hitless_key_update,
                   test_rss_context, test_rss_context4, test_rss_context32,
-                  test_rss_context_queue_reconfigure,
+                  test_rss_context_dump, test_rss_context_queue_reconfigure,
                   test_rss_context_overlap, test_rss_context_overlap2,
                   test_rss_context_out_of_order, test_rss_context4_create_with_cfg],
                  args=(cfg, ))
tools/testing/selftests/net/lib/py/ksft.py +6 -0
index f663e0daec0d..477ae76de93d 100644
@@ -55,6 +55,12 @@ KSFT_DISRUPTIVE = True
         _fail("Check failed", a, "!=", b, comment)
 
 
+def ksft_ne(a, b, comment=""):
+    global KSFT_RESULT
+    if a == b:
+        _fail("Check failed", a, "==", b, comment)
+
+
 def ksft_true(a, comment=""):
     if not a:
         _fail("Check failed", a, "does not eval to True", comment)
Signature · 1 line
2.45.2
Reviewed-by on v3 12/12
#17 lore raw

Re: [PATCH net-next v3 12/12] selftests: drv-net: rss_ctx: test dumping RSS contexts

Quoted 6 lines · Jakub Kicinski

Add a test for dumping RSS contexts. Make sure indir table and key are sane when contexts are created with various combination of inputs. Test the dump filtering by ifname and start-context.

maintainer Reviewed-by on v3 12/12
#18 lore raw

Re: [PATCH net-next v3 12/12] selftests: drv-net: rss_ctx: test dumping RSS contexts

In reply to Jakub Kicinski
Quoted 6 lines · Jakub Kicinski

Add a test for dumping RSS contexts. Make sure indir table and key are sane when contexts are created with various combination of inputs. Test the dump filtering by ifname and start-context.