Commit bdf2ba15 authored by David S. Miller's avatar David S. Miller

Merge branch 'am65-cpsw-rx-mq'

Roger Quadros says:

====================
net: ethernet: ti: am65-cpsw: Add multi queue RX support

am65-cpsw can support up to 8 queues at Rx. So far we have
been using only one queue (i.e. default flow) for all RX traffic.

This series adds multi-queue support. The driver starts with
1 RX queue by default. User can increase the RX queues via ethtool,
e.g. 'ethtool -L ethx rx <N>'

The series also adds regmap and regfield support to some of the
ALE registers. It adds Policer/Classifier registers and fields.

Converting the existing ALE control APIs to regfields can be a separate
exercise.

Some helper functions are added to read/write to the Policer/Classifier
registers and a default Classifier setup function is added that
routes packets based on their PCP/DSCP priority to different RX queues.
Signed-off-by: default avatarRoger Quadros <rogerq@kernel.org>
---
Changes in v4:
- Use single macro AM65_CPSW_MAX_QUEUES for both TX and RX queues
  to simplify code
- reuse am65_cpsw_get/set_per_queue_coalesce for am65_cpsw_get/set_coalesce.
- return -EINVAL if unsupported tx/rx_coalesce_usecs in
  am65_cpsw_set_coalesce.
- reverse Xmas tree declaration order fixes in cpsw_ale
- Link to v3: https://lore.kernel.org/r/20240703-am65-cpsw-multi-rx-v3-0-f11cd860fd72@kernel.org

Changes in v3:
- code style fixes
- squashed patches 5 and 6
- added comment about priority to thread mapping table.
- Added Reviewed-by Simon Horman.
- Link to v2: https://lore.kernel.org/r/20240628-am65-cpsw-multi-rx-v2-0-c399cb77db56@kernel.org

Changes in v2:
- rebase to net/next
- fixed RX stall issue during iperf
- Link to v1: https://lore.kernel.org/r/20240606-am65-cpsw-multi-rx-v1-0-0704b0cb6fdc@kernel.org
====================
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parents 52fa3b65 b7468c0f
......@@ -427,9 +427,9 @@ static void am65_cpsw_get_channels(struct net_device *ndev,
{
struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
ch->max_rx = AM65_CPSW_MAX_RX_QUEUES;
ch->max_tx = AM65_CPSW_MAX_TX_QUEUES;
ch->rx_count = AM65_CPSW_MAX_RX_QUEUES;
ch->max_rx = AM65_CPSW_MAX_QUEUES;
ch->max_tx = AM65_CPSW_MAX_QUEUES;
ch->rx_count = common->rx_ch_num_flows;
ch->tx_count = common->tx_ch_num;
}
......@@ -447,9 +447,8 @@ static int am65_cpsw_set_channels(struct net_device *ndev,
if (common->usage_count)
return -EBUSY;
am65_cpsw_nuss_remove_tx_chns(common);
return am65_cpsw_nuss_update_tx_chns(common, chs->tx_count);
return am65_cpsw_nuss_update_tx_rx_chns(common, chs->tx_count,
chs->rx_count);
}
static void
......@@ -913,80 +912,64 @@ static void am65_cpsw_get_mm_stats(struct net_device *ndev,
s->MACMergeHoldCount = readl(base + AM65_CPSW_STATN_IET_TX_HOLD);
}
static int am65_cpsw_get_coalesce(struct net_device *ndev, struct ethtool_coalesce *coal,
struct kernel_ethtool_coalesce *kernel_coal,
struct netlink_ext_ack *extack)
{
struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
struct am65_cpsw_tx_chn *tx_chn;
tx_chn = &common->tx_chns[0];
coal->rx_coalesce_usecs = common->rx_pace_timeout / 1000;
coal->tx_coalesce_usecs = tx_chn->tx_pace_timeout / 1000;
return 0;
}
static int am65_cpsw_get_per_queue_coalesce(struct net_device *ndev, u32 queue,
struct ethtool_coalesce *coal)
{
struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
struct am65_cpsw_rx_flow *rx_flow;
struct am65_cpsw_tx_chn *tx_chn;
if (queue >= AM65_CPSW_MAX_TX_QUEUES)
if (queue >= AM65_CPSW_MAX_QUEUES)
return -EINVAL;
tx_chn = &common->tx_chns[queue];
coal->tx_coalesce_usecs = tx_chn->tx_pace_timeout / 1000;
rx_flow = &common->rx_chns.flows[queue];
coal->rx_coalesce_usecs = rx_flow->rx_pace_timeout / 1000;
return 0;
}
static int am65_cpsw_set_coalesce(struct net_device *ndev, struct ethtool_coalesce *coal,
static int am65_cpsw_get_coalesce(struct net_device *ndev, struct ethtool_coalesce *coal,
struct kernel_ethtool_coalesce *kernel_coal,
struct netlink_ext_ack *extack)
{
struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
struct am65_cpsw_tx_chn *tx_chn;
tx_chn = &common->tx_chns[0];
if (coal->rx_coalesce_usecs && coal->rx_coalesce_usecs < 20)
return -EINVAL;
if (coal->tx_coalesce_usecs && coal->tx_coalesce_usecs < 20)
return -EINVAL;
common->rx_pace_timeout = coal->rx_coalesce_usecs * 1000;
tx_chn->tx_pace_timeout = coal->tx_coalesce_usecs * 1000;
return 0;
return am65_cpsw_get_per_queue_coalesce(ndev, 0, coal);
}
static int am65_cpsw_set_per_queue_coalesce(struct net_device *ndev, u32 queue,
struct ethtool_coalesce *coal)
{
struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
struct am65_cpsw_rx_flow *rx_flow;
struct am65_cpsw_tx_chn *tx_chn;
if (queue >= AM65_CPSW_MAX_TX_QUEUES)
if (queue >= AM65_CPSW_MAX_QUEUES)
return -EINVAL;
tx_chn = &common->tx_chns[queue];
if (coal->tx_coalesce_usecs && coal->tx_coalesce_usecs < 20) {
dev_info(common->dev, "defaulting to min value of 20us for tx-usecs for tx-%u\n",
queue);
coal->tx_coalesce_usecs = 20;
}
if (coal->tx_coalesce_usecs && coal->tx_coalesce_usecs < 20)
return -EINVAL;
tx_chn->tx_pace_timeout = coal->tx_coalesce_usecs * 1000;
rx_flow = &common->rx_chns.flows[queue];
if (coal->rx_coalesce_usecs && coal->rx_coalesce_usecs < 20)
return -EINVAL;
rx_flow->rx_pace_timeout = coal->rx_coalesce_usecs * 1000;
return 0;
}
static int am65_cpsw_set_coalesce(struct net_device *ndev, struct ethtool_coalesce *coal,
struct kernel_ethtool_coalesce *kernel_coal,
struct netlink_ext_ack *extack)
{
return am65_cpsw_set_per_queue_coalesce(ndev, 0, coal);
}
const struct ethtool_ops am65_cpsw_ethtool_ops_slave = {
.begin = am65_cpsw_ethtool_op_begin,
.complete = am65_cpsw_ethtool_op_complete,
......
This diff is collapsed.
......@@ -21,9 +21,7 @@ struct am65_cpts;
#define HOST_PORT_NUM 0
#define AM65_CPSW_MAX_TX_QUEUES 8
#define AM65_CPSW_MAX_RX_QUEUES 1
#define AM65_CPSW_MAX_RX_FLOWS 1
#define AM65_CPSW_MAX_QUEUES 8 /* both TX & RX */
#define AM65_CPSW_PORT_VLAN_REG_OFFSET 0x014
......@@ -58,7 +56,7 @@ struct am65_cpsw_port {
struct am65_cpsw_qos qos;
struct devlink_port devlink_port;
struct bpf_prog *xdp_prog;
struct xdp_rxq_info xdp_rxq;
struct xdp_rxq_info xdp_rxq[AM65_CPSW_MAX_QUEUES];
/* Only for suspend resume context */
u32 vid_context;
};
......@@ -94,16 +92,27 @@ struct am65_cpsw_tx_chn {
u32 rate_mbps;
};
struct am65_cpsw_rx_flow {
u32 id;
struct napi_struct napi_rx;
struct am65_cpsw_common *common;
int irq;
bool irq_disabled;
struct hrtimer rx_hrtimer;
unsigned long rx_pace_timeout;
struct page_pool *page_pool;
struct page **pages;
char name[32];
};
struct am65_cpsw_rx_chn {
struct device *dev;
struct device *dma_dev;
struct k3_cppi_desc_pool *desc_pool;
struct k3_udma_glue_rx_channel *rx_chn;
struct page_pool *page_pool;
struct page **pages;
u32 descs_num;
unsigned char dsize_log2;
int irq;
struct am65_cpsw_rx_flow flows[AM65_CPSW_MAX_QUEUES];
};
#define AM65_CPSW_QUIRK_I2027_NO_TX_CSUM BIT(0)
......@@ -145,16 +154,12 @@ struct am65_cpsw_common {
u32 tx_ch_rate_msk;
u32 rx_flow_id_base;
struct am65_cpsw_tx_chn tx_chns[AM65_CPSW_MAX_TX_QUEUES];
struct am65_cpsw_tx_chn tx_chns[AM65_CPSW_MAX_QUEUES];
struct completion tdown_complete;
atomic_t tdown_cnt;
int rx_ch_num_flows;
struct am65_cpsw_rx_chn rx_chns;
struct napi_struct napi_rx;
bool rx_irq_disabled;
struct hrtimer rx_hrtimer;
unsigned long rx_pace_timeout;
u32 nuss_ver;
u32 cpsw_ver;
......@@ -203,8 +208,8 @@ struct am65_cpsw_ndev_priv {
#define am65_common_get_host(common) (&(common)->host)
#define am65_common_get_port(common, id) (&(common)->ports[(id) - 1])
#define am65_cpsw_napi_to_common(pnapi) \
container_of(pnapi, struct am65_cpsw_common, napi_rx)
#define am65_cpsw_napi_to_rx_flow(pnapi) \
container_of(pnapi, struct am65_cpsw_rx_flow, napi_rx)
#define am65_cpsw_napi_to_tx_chn(pnapi) \
container_of(pnapi, struct am65_cpsw_tx_chn, napi_tx)
......@@ -215,8 +220,8 @@ struct am65_cpsw_ndev_priv {
extern const struct ethtool_ops am65_cpsw_ethtool_ops_slave;
void am65_cpsw_nuss_set_p0_ptype(struct am65_cpsw_common *common);
void am65_cpsw_nuss_remove_tx_chns(struct am65_cpsw_common *common);
int am65_cpsw_nuss_update_tx_chns(struct am65_cpsw_common *common, int num_tx);
int am65_cpsw_nuss_update_tx_rx_chns(struct am65_cpsw_common *common,
int num_tx, int num_rx);
bool am65_cpsw_port_dev_check(const struct net_device *dev);
......
This diff is collapsed.
......@@ -8,11 +8,14 @@
#ifndef __TI_CPSW_ALE_H__
#define __TI_CPSW_ALE_H__
struct reg_fields;
struct cpsw_ale_params {
struct device *dev;
void __iomem *ale_regs;
unsigned long ale_ageout; /* in secs */
unsigned long ale_entries;
unsigned long num_policers;
unsigned long ale_ports;
/* NU Switch has specific handling as number of bits in ALE entries
* are different than other versions of ALE. Also there are specific
......@@ -20,19 +23,69 @@ struct cpsw_ale_params {
* to identify this hardware.
*/
bool nu_switch_ale;
/* mask bit used in NU Switch ALE is 3 bits instead of 8 bits. So
* pass it from caller.
*/
u32 major_ver_mask;
const struct reg_field *reg_fields;
const char *dev_id;
unsigned long bus_freq;
};
struct ale_entry_fld;
struct regmap;
enum ale_fields {
MINOR_VER,
MAJOR_VER,
ALE_ENTRIES,
ALE_POLICERS,
POL_PORT_MEN,
POL_TRUNK_ID,
POL_PORT_NUM,
POL_PRI_MEN,
POL_PRI_VAL,
POL_OUI_MEN,
POL_OUI_INDEX,
POL_DST_MEN,
POL_DST_INDEX,
POL_SRC_MEN,
POL_SRC_INDEX,
POL_OVLAN_MEN,
POL_OVLAN_INDEX,
POL_IVLAN_MEN,
POL_IVLAN_INDEX,
POL_ETHERTYPE_MEN,
POL_ETHERTYPE_INDEX,
POL_IPSRC_MEN,
POL_IPSRC_INDEX,
POL_IPDST_MEN,
POL_IPDST_INDEX,
POL_EN,
POL_RED_DROP_EN,
POL_YELLOW_DROP_EN,
POL_YELLOW_THRESH,
POL_POL_MATCH_MODE,
POL_PRIORITY_THREAD_EN,
POL_MAC_ONLY_DEF_DIS,
POL_TEST_CLR,
POL_TEST_CLR_RED,
POL_TEST_CLR_YELLOW,
POL_TEST_CLR_SELECTED,
POL_TEST_ENTRY,
POL_STATUS_HIT,
POL_STATUS_HIT_RED,
POL_STATUS_HIT_YELLOW,
ALE_DEFAULT_THREAD_EN,
ALE_DEFAULT_THREAD_VAL,
ALE_THREAD_CLASS_INDEX,
ALE_THREAD_ENABLE,
ALE_THREAD_VALUE,
/* terminator */
ALE_FIELDS_MAX,
};
struct cpsw_ale {
struct cpsw_ale_params params;
struct timer_list timer;
struct regmap *regmap;
struct regmap_field *fields[ALE_FIELDS_MAX];
unsigned long ageout;
u32 version;
u32 features;
......@@ -140,5 +193,6 @@ int cpsw_ale_vlan_add_modify(struct cpsw_ale *ale, u16 vid, int port_mask,
int cpsw_ale_vlan_del_modify(struct cpsw_ale *ale, u16 vid, int port_mask);
void cpsw_ale_set_unreg_mcast(struct cpsw_ale *ale, int unreg_mcast_mask,
bool add);
void cpsw_ale_classifier_setup_default(struct cpsw_ale *ale, int num_rx_ch);
#endif
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment