diff mbox series

[net-next,02/10] eth: fbnic: support querying RSS config

Message ID 20241220025241.1522781-3-kuba@kernel.org (mailing list archive)
State New
Delegated to: Netdev Maintainers
Headers show
Series eth: fbnic: support basic RSS config and setting channel count | expand

Checks

Context Check Description
netdev/series_format success Posting correctly formatted
netdev/tree_selection success Clearly marked for net-next
netdev/ynl success Generated files up to date; no warnings/errors; no diff in generated;
netdev/fixes_present success Fixes tag not required for -next series
netdev/header_inline success No static functions without inline keyword in header files
netdev/build_32bit success Errors and warnings before: 1 this patch: 1
netdev/build_tools success No tools touched, skip
netdev/cc_maintainers warning 3 maintainers not CCed: andrew+netdev@lunn.ch kernel-team@meta.com vadim.fedorenko@linux.dev
netdev/build_clang success Errors and warnings before: 2 this patch: 2
netdev/verify_signedoff success Signed-off-by tag matches author and committer
netdev/deprecated_api success None detected
netdev/check_selftest success No net selftest shell script
netdev/verify_fixes success No Fixes tag
netdev/build_allmodconfig_warn success Errors and warnings before: 1 this patch: 1
netdev/checkpatch success total: 0 errors, 0 warnings, 0 checks, 115 lines checked
netdev/build_clang_rust success No Rust files in patch. Skipping build
netdev/kdoc success Errors and warnings before: 0 this patch: 0
netdev/source_inline success Was 0 now: 0
netdev/contest success net-next-2024-12-20--06-00 (tests: 881)

Commit Message

Jakub Kicinski Dec. 20, 2024, 2:52 a.m. UTC 
From: Alexander Duyck <alexanderduyck@fb.com>

The initial driver submission already added all the RSS state,
as part of multi-queue support. Expose the configuration via
the ethtool APIs.

Signed-off-by: Alexander Duyck <alexanderduyck@fb.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
 .../net/ethernet/meta/fbnic/fbnic_ethtool.c   | 103 ++++++++++++++++++
 1 file changed, 103 insertions(+)

Comments

Przemek Kitszel Dec. 20, 2024, 11:42 a.m. UTC | #1 
On 12/20/24 03:52, Jakub Kicinski wrote:
> From: Alexander Duyck <alexanderduyck@fb.com>
> 
> The initial driver submission already added all the RSS state,
> as part of multi-queue support. Expose the configuration via
> the ethtool APIs.
> 
> Signed-off-by: Alexander Duyck <alexanderduyck@fb.com>
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
> ---
>   .../net/ethernet/meta/fbnic/fbnic_ethtool.c   | 103 ++++++++++++++++++
>   1 file changed, 103 insertions(+)
> 


> +static int
> +fbnic_get_rxfh(struct net_device *netdev, struct ethtool_rxfh_param *rxfh)
> +{
> +	struct fbnic_net *fbn = netdev_priv(netdev);
> +	unsigned int i;

AFAIK index type should be spelled as u32
And will be best declared in the first clause of the for()

> +
> +	rxfh->hfunc = ETH_RSS_HASH_TOP;
> +
> +	if (rxfh->key) {
> +		for (i = 0; i < FBNIC_RPC_RSS_KEY_BYTE_LEN; i++) {
> +			u32 rss_key = fbn->rss_key[i / 4] << ((i % 4) * 8);

are you dropping 75% of entropy provided in fbn->rss_key?

> +
> +			rxfh->key[i] = rss_key >> 24;
> +		}
> +	}
> +
> +	if (rxfh->indir) {
> +		for (i = 0; i < FBNIC_RPC_RSS_TBL_SIZE; i++)
> +			rxfh->indir[i] = fbn->indir_tbl[0][i];
> +	}
> +
> +	return 0;
> +}
> +
>   static int
>   fbnic_get_ts_info(struct net_device *netdev,
>   		  struct kernel_ethtool_ts_info *tsinfo)
> @@ -209,6 +308,10 @@ static const struct ethtool_ops fbnic_ethtool_ops = {
>   	.get_strings		= fbnic_get_strings,
>   	.get_ethtool_stats	= fbnic_get_ethtool_stats,
>   	.get_sset_count		= fbnic_get_sset_count,
> +	.get_rxnfc		= fbnic_get_rxnfc,
> +	.get_rxfh_key_size	= fbnic_get_rxfh_key_size,
> +	.get_rxfh_indir_size	= fbnic_get_rxfh_indir_size,
> +	.get_rxfh		= fbnic_get_rxfh,
>   	.get_ts_info		= fbnic_get_ts_info,
>   	.get_ts_stats		= fbnic_get_ts_stats,
>   	.get_eth_mac_stats	= fbnic_get_eth_mac_stats,
Jakub Kicinski Dec. 20, 2024, 2:08 p.m. UTC | #2 
On Fri, 20 Dec 2024 12:42:42 +0100 Przemek Kitszel wrote:
> > +static int
> > +fbnic_get_rxfh(struct net_device *netdev, struct ethtool_rxfh_param *rxfh)
> > +{
> > +	struct fbnic_net *fbn = netdev_priv(netdev);
> > +	unsigned int i;  
> 
> AFAIK index type should be spelled as u32

Does it matter? I have a weak preference for not using explicitly sized
types unless the bit width is itself meaningful.

> And will be best declared in the first clause of the for()

I don't see woohaii.

> > +
> > +	rxfh->hfunc = ETH_RSS_HASH_TOP;
> > +
> > +	if (rxfh->key) {
> > +		for (i = 0; i < FBNIC_RPC_RSS_KEY_BYTE_LEN; i++) {
> > +			u32 rss_key = fbn->rss_key[i / 4] << ((i % 4) * 8);  
> 
> are you dropping 75% of entropy provided in fbn->rss_key?

Nope, it's shifting out the unused bits. And below we shift back
down to the lowest byte. 
We store the key as u32 (register width) while the uAPI is in u8.

> > +
> > +			rxfh->key[i] = rss_key >> 24;
> > +		}
> > +	}
> > +
> > +	if (rxfh->indir) {
> > +		for (i = 0; i < FBNIC_RPC_RSS_TBL_SIZE; i++)
> > +			rxfh->indir[i] = fbn->indir_tbl[0][i];
> > +	}
> > +
> > +	return 0;
Przemek Kitszel Dec. 20, 2024, 2:23 p.m. UTC | #3 
On 12/20/24 15:08, Jakub Kicinski wrote:
> On Fri, 20 Dec 2024 12:42:42 +0100 Przemek Kitszel wrote:

with [1] this patch looks good to me, thank you for explanation:
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>

the rest of the series is also fine for me, but I didn't spend that
much time to inflate my RB count

>>> +static int
>>> +fbnic_get_rxfh(struct net_device *netdev, struct ethtool_rxfh_param *rxfh)
>>> +{
>>> +	struct fbnic_net *fbn = netdev_priv(netdev);
>>> +	unsigned int i;
>>
>> AFAIK index type should be spelled as u32
> 
> Does it matter? I have a weak preference for not using explicitly sized
> types unless the bit width is itself meaningful.

me too, unless it is "unsiged int", but I', not gonna fight for it

I would say that plain int is fine for iterator unless you want to
explicitly get the non-UB wrap-around (or count more than 31 bit wide).
But again, does not matter

> 
>> And will be best declared in the first clause of the for()
> 
> I don't see woohaii.

with such long to type types, it would be inconvenient, but the usual
argument about the proper scope of the variable

> 
>>> +
>>> +	rxfh->hfunc = ETH_RSS_HASH_TOP;
>>> +
>>> +	if (rxfh->key) {
>>> +		for (i = 0; i < FBNIC_RPC_RSS_KEY_BYTE_LEN; i++) {
>>> +			u32 rss_key = fbn->rss_key[i / 4] << ((i % 4) * 8);
>>
>> are you dropping 75% of entropy provided in fbn->rss_key?
> 
> Nope, it's shifting out the unused bits. And below we shift back
> down to the lowest byte.
> We store the key as u32 (register width) while the uAPI is in u8.

[1]
OK, I've double checked and the fbn->rss_key array has indeed
FBNIC_RPC_RSS_KEY_BYTE_LEN/4 entries


> 
>>> +
>>> +			rxfh->key[i] = rss_key >> 24;
>>> +		}
>>> +	}
>>> +
>>> +	if (rxfh->indir) {
>>> +		for (i = 0; i < FBNIC_RPC_RSS_TBL_SIZE; i++)
>>> +			rxfh->indir[i] = fbn->indir_tbl[0][i];
>>> +	}
>>> +
>>> +	return 0;
diff mbox series

Patch

diff --git a/drivers/net/ethernet/meta/fbnic/fbnic_ethtool.c b/drivers/net/ethernet/meta/fbnic/fbnic_ethtool.c
index 777e083acae9..e71ae6abb0f5 100644
--- a/drivers/net/ethernet/meta/fbnic/fbnic_ethtool.c
+++ b/drivers/net/ethernet/meta/fbnic/fbnic_ethtool.c
@@ -102,6 +102,105 @@  static int fbnic_get_sset_count(struct net_device *dev, int sset)
 	}
 }
 
+static int fbnic_get_rss_hash_idx(u32 flow_type)
+{
+	switch (flow_type & ~(FLOW_EXT | FLOW_MAC_EXT | FLOW_RSS)) {
+	case TCP_V4_FLOW:
+		return FBNIC_TCP4_HASH_OPT;
+	case TCP_V6_FLOW:
+		return FBNIC_TCP6_HASH_OPT;
+	case UDP_V4_FLOW:
+		return FBNIC_UDP4_HASH_OPT;
+	case UDP_V6_FLOW:
+		return FBNIC_UDP6_HASH_OPT;
+	case AH_V4_FLOW:
+	case ESP_V4_FLOW:
+	case AH_ESP_V4_FLOW:
+	case SCTP_V4_FLOW:
+	case IPV4_FLOW:
+	case IPV4_USER_FLOW:
+		return FBNIC_IPV4_HASH_OPT;
+	case AH_V6_FLOW:
+	case ESP_V6_FLOW:
+	case AH_ESP_V6_FLOW:
+	case SCTP_V6_FLOW:
+	case IPV6_FLOW:
+	case IPV6_USER_FLOW:
+		return FBNIC_IPV6_HASH_OPT;
+	case ETHER_FLOW:
+		return FBNIC_ETHER_HASH_OPT;
+	}
+
+	return -1;
+}
+
+static int
+fbnic_get_rss_hash_opts(struct fbnic_net *fbn, struct ethtool_rxnfc *cmd)
+{
+	int hash_opt_idx = fbnic_get_rss_hash_idx(cmd->flow_type);
+
+	if (hash_opt_idx < 0)
+		return -EINVAL;
+
+	/* Report options from rss_en table in fbn */
+	cmd->data = fbn->rss_flow_hash[hash_opt_idx];
+
+	return 0;
+}
+
+static int fbnic_get_rxnfc(struct net_device *netdev,
+			   struct ethtool_rxnfc *cmd, u32 *rule_locs)
+{
+	struct fbnic_net *fbn = netdev_priv(netdev);
+	int ret = -EOPNOTSUPP;
+
+	switch (cmd->cmd) {
+	case ETHTOOL_GRXRINGS:
+		cmd->data = fbn->num_rx_queues;
+		ret = 0;
+		break;
+	case ETHTOOL_GRXFH:
+		ret = fbnic_get_rss_hash_opts(fbn, cmd);
+		break;
+	}
+
+	return ret;
+}
+
+static u32 fbnic_get_rxfh_key_size(struct net_device *netdev)
+{
+	return FBNIC_RPC_RSS_KEY_BYTE_LEN;
+}
+
+static u32 fbnic_get_rxfh_indir_size(struct net_device *netdev)
+{
+	return FBNIC_RPC_RSS_TBL_SIZE;
+}
+
+static int
+fbnic_get_rxfh(struct net_device *netdev, struct ethtool_rxfh_param *rxfh)
+{
+	struct fbnic_net *fbn = netdev_priv(netdev);
+	unsigned int i;
+
+	rxfh->hfunc = ETH_RSS_HASH_TOP;
+
+	if (rxfh->key) {
+		for (i = 0; i < FBNIC_RPC_RSS_KEY_BYTE_LEN; i++) {
+			u32 rss_key = fbn->rss_key[i / 4] << ((i % 4) * 8);
+
+			rxfh->key[i] = rss_key >> 24;
+		}
+	}
+
+	if (rxfh->indir) {
+		for (i = 0; i < FBNIC_RPC_RSS_TBL_SIZE; i++)
+			rxfh->indir[i] = fbn->indir_tbl[0][i];
+	}
+
+	return 0;
+}
+
 static int
 fbnic_get_ts_info(struct net_device *netdev,
 		  struct kernel_ethtool_ts_info *tsinfo)
@@ -209,6 +308,10 @@  static const struct ethtool_ops fbnic_ethtool_ops = {
 	.get_strings		= fbnic_get_strings,
 	.get_ethtool_stats	= fbnic_get_ethtool_stats,
 	.get_sset_count		= fbnic_get_sset_count,
+	.get_rxnfc		= fbnic_get_rxnfc,
+	.get_rxfh_key_size	= fbnic_get_rxfh_key_size,
+	.get_rxfh_indir_size	= fbnic_get_rxfh_indir_size,
+	.get_rxfh		= fbnic_get_rxfh,
 	.get_ts_info		= fbnic_get_ts_info,
 	.get_ts_stats		= fbnic_get_ts_stats,
 	.get_eth_mac_stats	= fbnic_get_eth_mac_stats,