* [PATCH net-next v2 0/3] cxgb4: add mirror action support for TC-MATCHALL
@ 2020-06-29 22:49 Rahul Lakkireddy
2020-06-29 22:49 ` [PATCH net-next v2 1/3] cxgb4: add mirror action to TC-MATCHALL offload Rahul Lakkireddy
` (2 more replies)
0 siblings, 3 replies; 5+ messages in thread
From: Rahul Lakkireddy @ 2020-06-29 22:49 UTC (permalink / raw
To: netdev; +Cc: davem, kuba, nirranjan, vishal, dt
This series of patches add support to mirror all ingress traffic
for TC-MATCHALL ingress offload.
Patch 1 adds support to dynamically create a mirror Virtual Interface
(VI) that accepts all mirror ingress traffic when mirror action is
set in TC-MATCHALL offload.
Patch 2 adds support to allocate mirror Rxqs and setup RSS for the
mirror VI.
Patch 3 adds support to replicate all the main VI configuration to
mirror VI. This includes replicating MTU, promiscuous mode,
all-multicast mode, and enabled netdev Rx feature offloads.
Thanks,
Rahul
v2:
- Add mutex to protect all mirror VI data, instead of just
mirror Rxqs, in patch 1 and 2.
- Remove the un-needed mirror Rxq mutex in patch 2.
- Simplify the replication code by refactoring t4_set_rxmode()
to handle mirror VI, instead of duplicating the t4_set_rxmode()
calls in multiple places in patch 3.
Rahul Lakkireddy (3):
cxgb4: add mirror action to TC-MATCHALL offload
cxgb4: add support for mirror Rxqs
cxgb4: add main VI to mirror VI config replication
drivers/net/ethernet/chelsio/cxgb4/cxgb4.h | 24 +-
.../ethernet/chelsio/cxgb4/cxgb4_debugfs.c | 66 ++-
.../net/ethernet/chelsio/cxgb4/cxgb4_main.c | 438 ++++++++++++++++--
.../ethernet/chelsio/cxgb4/cxgb4_tc_flower.c | 16 +-
.../ethernet/chelsio/cxgb4/cxgb4_tc_flower.h | 3 +-
.../chelsio/cxgb4/cxgb4_tc_matchall.c | 57 ++-
.../chelsio/cxgb4/cxgb4_tc_matchall.h | 1 +
drivers/net/ethernet/chelsio/cxgb4/t4_hw.c | 43 +-
8 files changed, 599 insertions(+), 49 deletions(-)
--
2.24.0
^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH net-next v2 1/3] cxgb4: add mirror action to TC-MATCHALL offload
2020-06-29 22:49 [PATCH net-next v2 0/3] cxgb4: add mirror action support for TC-MATCHALL Rahul Lakkireddy
@ 2020-06-29 22:49 ` Rahul Lakkireddy
2020-06-30 0:34 ` Jakub Kicinski
2020-06-29 22:49 ` [PATCH net-next v2 2/3] cxgb4: add support for mirror Rxqs Rahul Lakkireddy
2020-06-29 22:49 ` [PATCH net-next v2 3/3] cxgb4: add main VI to mirror VI config replication Rahul Lakkireddy
2 siblings, 1 reply; 5+ messages in thread
From: Rahul Lakkireddy @ 2020-06-29 22:49 UTC (permalink / raw
To: netdev; +Cc: davem, kuba, nirranjan, vishal, dt
Add mirror Virtual Interface (VI) support to receive all ingress
mirror traffic from the underlying device. The mirror VI is
created dynamically, if the TC-MATCHALL rule has a corresponding
mirror action. Also request MSI-X vectors needed for the mirror VI
Rxqs. If no vectors are available, then disable mirror VI support.
v2:
- Add mutex to protect all mirror VI data, instead of just
mirror Rxqs.
Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
---
drivers/net/ethernet/chelsio/cxgb4/cxgb4.h | 11 ++
.../net/ethernet/chelsio/cxgb4/cxgb4_main.c | 103 +++++++++++++++++-
.../ethernet/chelsio/cxgb4/cxgb4_tc_flower.c | 16 ++-
.../ethernet/chelsio/cxgb4/cxgb4_tc_flower.h | 3 +-
.../chelsio/cxgb4/cxgb4_tc_matchall.c | 57 +++++++++-
.../chelsio/cxgb4/cxgb4_tc_matchall.h | 1 +
drivers/net/ethernet/chelsio/cxgb4/t4_hw.c | 16 +++
7 files changed, 197 insertions(+), 10 deletions(-)
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
index d811df1b93d9..434cc2caaf5b 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
@@ -679,6 +679,12 @@ struct port_info {
u8 rx_cchan;
bool tc_block_shared;
+
+ /* Mirror VI information */
+ u16 viid_mirror;
+ refcount_t vi_mirror_refcnt;
+ u16 nmirrorqsets;
+ struct mutex vi_mirror_mutex; /* Sync access to Mirror VI info */
};
struct dentry;
@@ -960,6 +966,7 @@ struct sge {
u16 ofldqsets; /* # of active ofld queue sets */
u16 nqs_per_uld; /* # of Rx queues per ULD */
u16 eoqsets; /* # of ETHOFLD queues */
+ u16 mirrorqsets; /* # of Mirror queues */
u16 timer_val[SGE_NTIMERS];
u8 counter_val[SGE_NCOUNTERS];
@@ -1857,6 +1864,8 @@ int t4_init_rss_mode(struct adapter *adap, int mbox);
int t4_init_portinfo(struct port_info *pi, int mbox,
int port, int pf, int vf, u8 mac[]);
int t4_port_init(struct adapter *adap, int mbox, int pf, int vf);
+int t4_init_port_mirror(struct port_info *pi, u8 mbox, u8 port, u8 pf, u8 vf,
+ u16 *mirror_viid);
void t4_fatal_err(struct adapter *adapter);
unsigned int t4_chip_rss_size(struct adapter *adapter);
int t4_config_rss_range(struct adapter *adapter, int mbox, unsigned int viid,
@@ -2141,6 +2150,8 @@ int cxgb_open(struct net_device *dev);
int cxgb_close(struct net_device *dev);
void cxgb4_enable_rx(struct adapter *adap, struct sge_rspq *q);
void cxgb4_quiesce_rx(struct sge_rspq *q);
+int cxgb4_port_mirror_alloc(struct net_device *dev);
+void cxgb4_port_mirror_free(struct net_device *dev);
#ifdef CONFIG_CHELSIO_TLS_DEVICE
int cxgb4_set_ktls_feature(struct adapter *adap, bool enable);
#endif
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
index 234f01389cb1..a8b137837746 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
@@ -1285,6 +1285,55 @@ static int setup_debugfs(struct adapter *adap)
return 0;
}
+int cxgb4_port_mirror_alloc(struct net_device *dev)
+{
+ struct port_info *pi = netdev2pinfo(dev);
+ struct adapter *adap = netdev2adap(dev);
+ int ret = 0;
+
+ if (!pi->nmirrorqsets)
+ return -EOPNOTSUPP;
+
+ mutex_lock(&pi->vi_mirror_mutex);
+ if (pi->viid_mirror) {
+ refcount_inc(&pi->vi_mirror_refcnt);
+ goto out_unlock;
+ }
+
+ ret = t4_init_port_mirror(pi, adap->mbox, pi->port_id, adap->pf, 0,
+ &pi->viid_mirror);
+ if (ret)
+ goto out_unlock;
+
+ refcount_set(&pi->vi_mirror_refcnt, 1);
+
+out_unlock:
+ mutex_unlock(&pi->vi_mirror_mutex);
+ return ret;
+}
+
+void cxgb4_port_mirror_free(struct net_device *dev)
+{
+ struct port_info *pi = netdev2pinfo(dev);
+ struct adapter *adap = netdev2adap(dev);
+
+ mutex_lock(&pi->vi_mirror_mutex);
+ if (!pi->viid_mirror)
+ goto out_unlock;
+
+ if (refcount_read(&pi->vi_mirror_refcnt) > 1) {
+ refcount_dec(&pi->vi_mirror_refcnt);
+ goto out_unlock;
+ }
+
+ refcount_set(&pi->vi_mirror_refcnt, 0);
+ t4_free_vi(adap, adap->mbox, adap->pf, 0, pi->viid_mirror);
+ pi->viid_mirror = 0;
+
+out_unlock:
+ mutex_unlock(&pi->vi_mirror_mutex);
+}
+
/*
* upper-layer driver support
*/
@@ -5504,6 +5553,19 @@ static int cfg_queues(struct adapter *adap)
avail_qsets -= s->eoqsets;
}
+ /* Mirror queues must follow same scheme as normal Ethernet
+ * Queues, when there are enough queues available. Otherwise,
+ * allocate at least 1 queue per port. If even 1 queue is not
+ * available, then disable mirror queues support.
+ */
+ if (avail_qsets >= s->max_ethqsets)
+ s->mirrorqsets = s->max_ethqsets;
+ else if (avail_qsets >= adap->params.nports)
+ s->mirrorqsets = adap->params.nports;
+ else
+ s->mirrorqsets = 0;
+ avail_qsets -= s->mirrorqsets;
+
for (i = 0; i < ARRAY_SIZE(s->ethrxq); i++) {
struct sge_eth_rxq *r = &s->ethrxq[i];
@@ -5617,8 +5679,8 @@ void cxgb4_free_msix_idx_in_bmap(struct adapter *adap,
static int enable_msix(struct adapter *adap)
{
- u32 eth_need, uld_need = 0, ethofld_need = 0;
- u32 ethqsets = 0, ofldqsets = 0, eoqsets = 0;
+ u32 eth_need, uld_need = 0, ethofld_need = 0, mirror_need = 0;
+ u32 ethqsets = 0, ofldqsets = 0, eoqsets = 0, mirrorqsets = 0;
u8 num_uld = 0, nchan = adap->params.nports;
u32 i, want, need, num_vec;
struct sge *s = &adap->sge;
@@ -5649,6 +5711,12 @@ static int enable_msix(struct adapter *adap)
need += ethofld_need;
}
+ if (s->mirrorqsets) {
+ want += s->mirrorqsets;
+ mirror_need = nchan;
+ need += mirror_need;
+ }
+
want += EXTRA_VECS;
need += EXTRA_VECS;
@@ -5682,8 +5750,10 @@ static int enable_msix(struct adapter *adap)
adap->params.ethofld = 0;
s->ofldqsets = 0;
s->eoqsets = 0;
+ s->mirrorqsets = 0;
uld_need = 0;
ethofld_need = 0;
+ mirror_need = 0;
}
num_vec = allocated;
@@ -5697,6 +5767,8 @@ static int enable_msix(struct adapter *adap)
ofldqsets = nchan;
if (is_ethofld(adap))
eoqsets = ethofld_need;
+ if (s->mirrorqsets)
+ mirrorqsets = mirror_need;
num_vec -= need;
while (num_vec) {
@@ -5728,12 +5800,25 @@ static int enable_msix(struct adapter *adap)
num_vec -= uld_need;
}
}
+
+ if (s->mirrorqsets) {
+ while (num_vec) {
+ if (num_vec < mirror_need ||
+ mirrorqsets > s->mirrorqsets)
+ break;
+
+ mirrorqsets++;
+ num_vec -= mirror_need;
+ }
+ }
} else {
ethqsets = s->max_ethqsets;
if (is_uld(adap))
ofldqsets = s->ofldqsets;
if (is_ethofld(adap))
eoqsets = s->eoqsets;
+ if (s->mirrorqsets)
+ mirrorqsets = s->mirrorqsets;
}
if (ethqsets < s->max_ethqsets) {
@@ -5749,6 +5834,15 @@ static int enable_msix(struct adapter *adap)
if (is_ethofld(adap))
s->eoqsets = eoqsets;
+ if (s->mirrorqsets) {
+ s->mirrorqsets = mirrorqsets;
+ for_each_port(adap, i) {
+ pi = adap2pinfo(adap, i);
+ pi->nmirrorqsets = s->mirrorqsets / nchan;
+ mutex_init(&pi->vi_mirror_mutex);
+ }
+ }
+
/* map for msix */
ret = alloc_msix_info(adap, allocated);
if (ret)
@@ -5760,8 +5854,9 @@ static int enable_msix(struct adapter *adap)
}
dev_info(adap->pdev_dev,
- "%d MSI-X vectors allocated, nic %d eoqsets %d per uld %d\n",
- allocated, s->max_ethqsets, s->eoqsets, s->nqs_per_uld);
+ "%d MSI-X vectors allocated, nic %d eoqsets %d per uld %d mirrorqsets %d\n",
+ allocated, s->max_ethqsets, s->eoqsets, s->nqs_per_uld,
+ s->mirrorqsets);
kfree(entries);
return 0;
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c
index ae0e998d9338..6251444ffa58 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c
@@ -372,6 +372,7 @@ void cxgb4_process_flow_actions(struct net_device *in,
case FLOW_ACTION_DROP:
fs->action = FILTER_DROP;
break;
+ case FLOW_ACTION_MIRRED:
case FLOW_ACTION_REDIRECT: {
struct net_device *out = act->dev;
struct port_info *pi = netdev_priv(out);
@@ -529,7 +530,8 @@ static bool valid_pedit_action(struct net_device *dev,
int cxgb4_validate_flow_actions(struct net_device *dev,
struct flow_action *actions,
- struct netlink_ext_ack *extack)
+ struct netlink_ext_ack *extack,
+ u8 matchall_filter)
{
struct flow_action_entry *act;
bool act_redir = false;
@@ -546,11 +548,19 @@ int cxgb4_validate_flow_actions(struct net_device *dev,
case FLOW_ACTION_DROP:
/* Do nothing */
break;
+ case FLOW_ACTION_MIRRED:
case FLOW_ACTION_REDIRECT: {
struct adapter *adap = netdev2adap(dev);
struct net_device *n_dev, *target_dev;
- unsigned int i;
bool found = false;
+ unsigned int i;
+
+ if (act->id == FLOW_ACTION_MIRRED &&
+ !matchall_filter) {
+ NL_SET_ERR_MSG_MOD(extack,
+ "Egress mirror action is only supported for tc-matchall");
+ return -EOPNOTSUPP;
+ }
target_dev = act->dev;
for_each_port(adap, i) {
@@ -689,7 +699,7 @@ int cxgb4_flow_rule_replace(struct net_device *dev, struct flow_rule *rule,
u8 inet_family;
int fidx, ret;
- if (cxgb4_validate_flow_actions(dev, &rule->action, extack))
+ if (cxgb4_validate_flow_actions(dev, &rule->action, extack, 0))
return -EOPNOTSUPP;
if (cxgb4_validate_flow_match(dev, rule))
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.h
index befa459324fb..6296e1d5a12b 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.h
@@ -113,7 +113,8 @@ void cxgb4_process_flow_actions(struct net_device *in,
struct ch_filter_specification *fs);
int cxgb4_validate_flow_actions(struct net_device *dev,
struct flow_action *actions,
- struct netlink_ext_ack *extack);
+ struct netlink_ext_ack *extack,
+ u8 matchall_filter);
int cxgb4_tc_flower_replace(struct net_device *dev,
struct flow_cls_offload *cls);
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c
index c439b5bce9c9..e377e50c2492 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c
@@ -188,6 +188,49 @@ static void cxgb4_matchall_free_tc(struct net_device *dev)
tc_port_matchall->egress.state = CXGB4_MATCHALL_STATE_DISABLED;
}
+static int cxgb4_matchall_mirror_alloc(struct net_device *dev,
+ struct tc_cls_matchall_offload *cls)
+{
+ struct netlink_ext_ack *extack = cls->common.extack;
+ struct cxgb4_tc_port_matchall *tc_port_matchall;
+ struct port_info *pi = netdev2pinfo(dev);
+ struct adapter *adap = netdev2adap(dev);
+ struct flow_action_entry *act;
+ int ret;
+ u32 i;
+
+ tc_port_matchall = &adap->tc_matchall->port_matchall[pi->port_id];
+ flow_action_for_each(i, act, &cls->rule->action) {
+ if (act->id == FLOW_ACTION_MIRRED) {
+ ret = cxgb4_port_mirror_alloc(dev);
+ if (ret) {
+ NL_SET_ERR_MSG_MOD(extack,
+ "Couldn't allocate mirror");
+ return ret;
+ }
+
+ tc_port_matchall->ingress.viid_mirror = pi->viid_mirror;
+ break;
+ }
+ }
+
+ return 0;
+}
+
+static void cxgb4_matchall_mirror_free(struct net_device *dev)
+{
+ struct cxgb4_tc_port_matchall *tc_port_matchall;
+ struct port_info *pi = netdev2pinfo(dev);
+ struct adapter *adap = netdev2adap(dev);
+
+ tc_port_matchall = &adap->tc_matchall->port_matchall[pi->port_id];
+ if (!tc_port_matchall->ingress.viid_mirror)
+ return;
+
+ cxgb4_port_mirror_free(dev);
+ tc_port_matchall->ingress.viid_mirror = 0;
+}
+
static int cxgb4_matchall_alloc_filter(struct net_device *dev,
struct tc_cls_matchall_offload *cls)
{
@@ -211,6 +254,10 @@ static int cxgb4_matchall_alloc_filter(struct net_device *dev,
return -ENOMEM;
}
+ ret = cxgb4_matchall_mirror_alloc(dev, cls);
+ if (ret)
+ return ret;
+
tc_port_matchall = &adap->tc_matchall->port_matchall[pi->port_id];
fs = &tc_port_matchall->ingress.fs;
memset(fs, 0, sizeof(*fs));
@@ -229,11 +276,15 @@ static int cxgb4_matchall_alloc_filter(struct net_device *dev,
ret = cxgb4_set_filter(dev, fidx, fs);
if (ret)
- return ret;
+ goto out_free;
tc_port_matchall->ingress.tid = fidx;
tc_port_matchall->ingress.state = CXGB4_MATCHALL_STATE_ENABLED;
return 0;
+
+out_free:
+ cxgb4_matchall_mirror_free(dev);
+ return ret;
}
static int cxgb4_matchall_free_filter(struct net_device *dev)
@@ -250,6 +301,8 @@ static int cxgb4_matchall_free_filter(struct net_device *dev)
if (ret)
return ret;
+ cxgb4_matchall_mirror_free(dev);
+
tc_port_matchall->ingress.packets = 0;
tc_port_matchall->ingress.bytes = 0;
tc_port_matchall->ingress.last_used = 0;
@@ -279,7 +332,7 @@ int cxgb4_tc_matchall_replace(struct net_device *dev,
ret = cxgb4_validate_flow_actions(dev,
&cls_matchall->rule->action,
- extack);
+ extack, 1);
if (ret)
return ret;
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.h
index ab6b5683dfd3..e264b6e606c4 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.h
@@ -21,6 +21,7 @@ struct cxgb4_matchall_ingress_entry {
enum cxgb4_matchall_state state; /* Current MATCHALL offload state */
u32 tid; /* Index to hardware filter entry */
struct ch_filter_specification fs; /* Filter entry */
+ u16 viid_mirror; /* Identifier for allocated Mirror VI */
u64 bytes; /* # of bytes hitting the filter */
u64 packets; /* # of packets hitting the filter */
u64 last_used; /* Last updated jiffies time */
diff --git a/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c b/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c
index 70fe189202be..7876aa392aae 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c
@@ -9711,6 +9711,22 @@ int t4_port_init(struct adapter *adap, int mbox, int pf, int vf)
return 0;
}
+int t4_init_port_mirror(struct port_info *pi, u8 mbox, u8 port, u8 pf, u8 vf,
+ u16 *mirror_viid)
+{
+ int ret;
+
+ ret = t4_alloc_vi(pi->adapter, mbox, port, pf, vf, 1, NULL, NULL,
+ NULL, NULL);
+ if (ret < 0)
+ return ret;
+
+ if (mirror_viid)
+ *mirror_viid = ret;
+
+ return 0;
+}
+
/**
* t4_read_cimq_cfg - read CIM queue configuration
* @adap: the adapter
--
2.24.0
^ permalink raw reply related [flat|nested] 5+ messages in thread
* [PATCH net-next v2 2/3] cxgb4: add support for mirror Rxqs
2020-06-29 22:49 [PATCH net-next v2 0/3] cxgb4: add mirror action support for TC-MATCHALL Rahul Lakkireddy
2020-06-29 22:49 ` [PATCH net-next v2 1/3] cxgb4: add mirror action to TC-MATCHALL offload Rahul Lakkireddy
@ 2020-06-29 22:49 ` Rahul Lakkireddy
2020-06-29 22:49 ` [PATCH net-next v2 3/3] cxgb4: add main VI to mirror VI config replication Rahul Lakkireddy
2 siblings, 0 replies; 5+ messages in thread
From: Rahul Lakkireddy @ 2020-06-29 22:49 UTC (permalink / raw
To: netdev; +Cc: davem, kuba, nirranjan, vishal, dt
When mirror VI is enabled, allocate the mirror Rxqs and setup the
mirror VI RSS table. The mirror Rxqs are allocated/freed when
the mirror VI is created/destroyed or when underlying port is
brought up/down, respectively.
v2:
- Use mutex to protect all mirror VI data, instead of just
mirror Rxqs.
- Remove the un-needed mirror Rxq mutex.
Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
---
drivers/net/ethernet/chelsio/cxgb4/cxgb4.h | 9 +
.../ethernet/chelsio/cxgb4/cxgb4_debugfs.c | 66 ++++-
.../net/ethernet/chelsio/cxgb4/cxgb4_main.c | 235 ++++++++++++++++--
3 files changed, 288 insertions(+), 22 deletions(-)
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
index 434cc2caaf5b..8985d85a1530 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
@@ -711,6 +711,13 @@ enum {
ULP_CRYPTO_KTLS_INLINE = 1 << 3,
};
+#define CXGB4_MIRROR_RXQ_DEFAULT_DESC_NUM 1024
+#define CXGB4_MIRROR_RXQ_DEFAULT_DESC_SIZE 64
+#define CXGB4_MIRROR_RXQ_DEFAULT_INTR_USEC 5
+#define CXGB4_MIRROR_RXQ_DEFAULT_PKT_CNT 8
+
+#define CXGB4_MIRROR_FLQ_DEFAULT_DESC_NUM 72
+
struct rx_sw_desc;
struct sge_fl { /* SGE free-buffer queue state */
@@ -960,6 +967,8 @@ struct sge {
struct sge_eohw_txq *eohw_txq;
struct sge_ofld_rxq *eohw_rxq;
+ struct sge_eth_rxq *mirror_rxq[NCHAN];
+
u16 max_ethqsets; /* # of available Ethernet queue sets */
u16 ethqsets; /* # of active Ethernet queue sets */
u16 ethtxq_rover; /* Tx queue to clean up next */
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_debugfs.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_debugfs.c
index b477b8842905..d578150c7091 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_debugfs.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_debugfs.c
@@ -2742,6 +2742,58 @@ do { \
}
r -= eth_entries;
+ for_each_port(adap, j) {
+ struct port_info *pi = adap2pinfo(adap, j);
+ const struct sge_eth_rxq *rx;
+
+ mutex_lock(&pi->vi_mirror_mutex);
+ if (!refcount_read(&pi->vi_mirror_refcnt)) {
+ mutex_unlock(&pi->vi_mirror_mutex);
+ continue;
+ }
+
+ if (r >= DIV_ROUND_UP(pi->nmirrorqsets, 4)) {
+ r -= DIV_ROUND_UP(pi->nmirrorqsets, 4);
+ mutex_unlock(&pi->vi_mirror_mutex);
+ continue;
+ }
+
+ rx = &s->mirror_rxq[j][r * 4];
+ n = min(4, pi->nmirrorqsets - 4 * r);
+
+ S("QType:", "Mirror-Rxq");
+ S("Interface:",
+ rx[i].rspq.netdev ? rx[i].rspq.netdev->name : "N/A");
+ R("RspQ ID:", rspq.abs_id);
+ R("RspQ size:", rspq.size);
+ R("RspQE size:", rspq.iqe_len);
+ R("RspQ CIDX:", rspq.cidx);
+ R("RspQ Gen:", rspq.gen);
+ S3("u", "Intr delay:", qtimer_val(adap, &rx[i].rspq));
+ S3("u", "Intr pktcnt:", s->counter_val[rx[i].rspq.pktcnt_idx]);
+ R("FL ID:", fl.cntxt_id);
+ R("FL size:", fl.size - 8);
+ R("FL pend:", fl.pend_cred);
+ R("FL avail:", fl.avail);
+ R("FL PIDX:", fl.pidx);
+ R("FL CIDX:", fl.cidx);
+ RL("RxPackets:", stats.pkts);
+ RL("RxCSO:", stats.rx_cso);
+ RL("VLANxtract:", stats.vlan_ex);
+ RL("LROmerged:", stats.lro_merged);
+ RL("LROpackets:", stats.lro_pkts);
+ RL("RxDrops:", stats.rx_drops);
+ RL("RxBadPkts:", stats.bad_rx_pkts);
+ RL("FLAllocErr:", fl.alloc_failed);
+ RL("FLLrgAlcErr:", fl.large_alloc_failed);
+ RL("FLMapErr:", fl.mapping_err);
+ RL("FLLow:", fl.low);
+ RL("FLStarving:", fl.starving);
+
+ mutex_unlock(&pi->vi_mirror_mutex);
+ goto out;
+ }
+
if (!adap->tc_mqprio)
goto skip_mqprio;
@@ -3098,9 +3150,10 @@ do { \
return 0;
}
-static int sge_queue_entries(const struct adapter *adap)
+static int sge_queue_entries(struct adapter *adap)
{
int i, tot_uld_entries = 0, eohw_entries = 0, eosw_entries = 0;
+ int mirror_rxq_entries = 0;
if (adap->tc_mqprio) {
struct cxgb4_tc_port_mqprio *port_mqprio;
@@ -3123,6 +3176,15 @@ static int sge_queue_entries(const struct adapter *adap)
mutex_unlock(&adap->tc_mqprio->mqprio_mutex);
}
+ for_each_port(adap, i) {
+ struct port_info *pi = adap2pinfo(adap, i);
+
+ mutex_lock(&pi->vi_mirror_mutex);
+ if (refcount_read(&pi->vi_mirror_refcnt))
+ mirror_rxq_entries += DIV_ROUND_UP(pi->nmirrorqsets, 4);
+ mutex_unlock(&pi->vi_mirror_mutex);
+ }
+
if (!is_uld(adap))
goto lld_only;
@@ -3137,7 +3199,7 @@ static int sge_queue_entries(const struct adapter *adap)
mutex_unlock(&uld_mutex);
lld_only:
- return DIV_ROUND_UP(adap->sge.ethqsets, 4) +
+ return DIV_ROUND_UP(adap->sge.ethqsets, 4) + mirror_rxq_entries +
eohw_entries + eosw_entries + tot_uld_entries +
DIV_ROUND_UP(MAX_CTRL_QUEUES, 4) + 1;
}
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
index a8b137837746..00d8badc666c 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
@@ -822,6 +822,31 @@ static void adap_config_hpfilter(struct adapter *adapter)
"HP filter region isn't supported by FW\n");
}
+static int cxgb4_config_rss(const struct port_info *pi, u16 *rss,
+ u16 rss_size, u16 viid)
+{
+ struct adapter *adap = pi->adapter;
+ int ret;
+
+ ret = t4_config_rss_range(adap, adap->mbox, viid, 0, rss_size, rss,
+ rss_size);
+ if (ret)
+ return ret;
+
+ /* If Tunnel All Lookup isn't specified in the global RSS
+ * Configuration, then we need to specify a default Ingress
+ * Queue for any ingress packets which aren't hashed. We'll
+ * use our first ingress queue ...
+ */
+ return t4_config_vi_rss(adap, adap->mbox, viid,
+ FW_RSS_VI_CONFIG_CMD_IP6FOURTUPEN_F |
+ FW_RSS_VI_CONFIG_CMD_IP6TWOTUPEN_F |
+ FW_RSS_VI_CONFIG_CMD_IP4FOURTUPEN_F |
+ FW_RSS_VI_CONFIG_CMD_IP4TWOTUPEN_F |
+ FW_RSS_VI_CONFIG_CMD_UDPEN_F,
+ rss[0]);
+}
+
/**
* cxgb4_write_rss - write the RSS table for a given port
* @pi: the port
@@ -833,10 +858,10 @@ static void adap_config_hpfilter(struct adapter *adapter)
*/
int cxgb4_write_rss(const struct port_info *pi, const u16 *queues)
{
- u16 *rss;
- int i, err;
struct adapter *adapter = pi->adapter;
const struct sge_eth_rxq *rxq;
+ int i, err;
+ u16 *rss;
rxq = &adapter->sge.ethrxq[pi->first_qset];
rss = kmalloc_array(pi->rss_size, sizeof(u16), GFP_KERNEL);
@@ -847,21 +872,7 @@ int cxgb4_write_rss(const struct port_info *pi, const u16 *queues)
for (i = 0; i < pi->rss_size; i++, queues++)
rss[i] = rxq[*queues].rspq.abs_id;
- err = t4_config_rss_range(adapter, adapter->pf, pi->viid, 0,
- pi->rss_size, rss, pi->rss_size);
- /* If Tunnel All Lookup isn't specified in the global RSS
- * Configuration, then we need to specify a default Ingress
- * Queue for any ingress packets which aren't hashed. We'll
- * use our first ingress queue ...
- */
- if (!err)
- err = t4_config_vi_rss(adapter, adapter->mbox, pi->viid,
- FW_RSS_VI_CONFIG_CMD_IP6FOURTUPEN_F |
- FW_RSS_VI_CONFIG_CMD_IP6TWOTUPEN_F |
- FW_RSS_VI_CONFIG_CMD_IP4FOURTUPEN_F |
- FW_RSS_VI_CONFIG_CMD_IP4TWOTUPEN_F |
- FW_RSS_VI_CONFIG_CMD_UDPEN_F,
- rss[0]);
+ err = cxgb4_config_rss(pi, rss, pi->rss_size, pi->viid);
kfree(rss);
return err;
}
@@ -1285,6 +1296,151 @@ static int setup_debugfs(struct adapter *adap)
return 0;
}
+static void cxgb4_port_mirror_free_rxq(struct adapter *adap,
+ struct sge_eth_rxq *mirror_rxq)
+{
+ if ((adap->flags & CXGB4_FULL_INIT_DONE) &&
+ !(adap->flags & CXGB4_SHUTTING_DOWN))
+ cxgb4_quiesce_rx(&mirror_rxq->rspq);
+
+ if (adap->flags & CXGB4_USING_MSIX) {
+ cxgb4_clear_msix_aff(mirror_rxq->msix->vec,
+ mirror_rxq->msix->aff_mask);
+ free_irq(mirror_rxq->msix->vec, &mirror_rxq->rspq);
+ cxgb4_free_msix_idx_in_bmap(adap, mirror_rxq->msix->idx);
+ }
+
+ free_rspq_fl(adap, &mirror_rxq->rspq, &mirror_rxq->fl);
+}
+
+static int cxgb4_port_mirror_alloc_queues(struct net_device *dev)
+{
+ struct port_info *pi = netdev2pinfo(dev);
+ struct adapter *adap = netdev2adap(dev);
+ struct sge_eth_rxq *mirror_rxq;
+ struct sge *s = &adap->sge;
+ int ret = 0, msix = 0;
+ u16 i, rxqid;
+ u16 *rss;
+
+ if (!refcount_read(&pi->vi_mirror_refcnt))
+ return 0;
+
+ if (s->mirror_rxq[pi->port_id])
+ return 0;
+
+ mirror_rxq = kcalloc(pi->nmirrorqsets, sizeof(*mirror_rxq), GFP_KERNEL);
+ if (!mirror_rxq)
+ return -ENOMEM;
+
+ s->mirror_rxq[pi->port_id] = mirror_rxq;
+
+ if (!(adap->flags & CXGB4_USING_MSIX))
+ msix = -((int)adap->sge.intrq.abs_id + 1);
+
+ for (i = 0, rxqid = 0; i < pi->nmirrorqsets; i++, rxqid++) {
+ mirror_rxq = &s->mirror_rxq[pi->port_id][i];
+
+ /* Allocate Mirror Rxqs */
+ if (msix >= 0) {
+ msix = cxgb4_get_msix_idx_from_bmap(adap);
+ if (msix < 0) {
+ ret = msix;
+ goto out_free_queues;
+ }
+
+ mirror_rxq->msix = &adap->msix_info[msix];
+ snprintf(mirror_rxq->msix->desc,
+ sizeof(mirror_rxq->msix->desc),
+ "%s-mirrorrxq%d", dev->name, i);
+ }
+
+ init_rspq(adap, &mirror_rxq->rspq,
+ CXGB4_MIRROR_RXQ_DEFAULT_INTR_USEC,
+ CXGB4_MIRROR_RXQ_DEFAULT_PKT_CNT,
+ CXGB4_MIRROR_RXQ_DEFAULT_DESC_NUM,
+ CXGB4_MIRROR_RXQ_DEFAULT_DESC_SIZE);
+
+ mirror_rxq->fl.size = CXGB4_MIRROR_FLQ_DEFAULT_DESC_NUM;
+
+ ret = t4_sge_alloc_rxq(adap, &mirror_rxq->rspq, false,
+ dev, msix, &mirror_rxq->fl,
+ t4_ethrx_handler, NULL, 0);
+ if (ret)
+ goto out_free_msix_idx;
+
+ /* Setup MSI-X vectors for Mirror Rxqs */
+ if (adap->flags & CXGB4_USING_MSIX) {
+ ret = request_irq(mirror_rxq->msix->vec,
+ t4_sge_intr_msix, 0,
+ mirror_rxq->msix->desc,
+ &mirror_rxq->rspq);
+ if (ret)
+ goto out_free_rxq;
+
+ cxgb4_set_msix_aff(adap, mirror_rxq->msix->vec,
+ &mirror_rxq->msix->aff_mask, i);
+ }
+
+ /* Start NAPI for Mirror Rxqs */
+ cxgb4_enable_rx(adap, &mirror_rxq->rspq);
+ }
+
+ /* Setup RSS for Mirror Rxqs */
+ rss = kcalloc(pi->rss_size, sizeof(u16), GFP_KERNEL);
+ if (!rss) {
+ ret = -ENOMEM;
+ goto out_free_queues;
+ }
+
+ mirror_rxq = &s->mirror_rxq[pi->port_id][0];
+ for (i = 0; i < pi->rss_size; i++)
+ rss[i] = mirror_rxq[i % pi->nmirrorqsets].rspq.abs_id;
+
+ ret = cxgb4_config_rss(pi, rss, pi->rss_size, pi->viid_mirror);
+ kfree(rss);
+ if (ret)
+ goto out_free_queues;
+
+ return 0;
+
+out_free_rxq:
+ free_rspq_fl(adap, &mirror_rxq->rspq, &mirror_rxq->fl);
+
+out_free_msix_idx:
+ cxgb4_free_msix_idx_in_bmap(adap, mirror_rxq->msix->idx);
+
+out_free_queues:
+ while (rxqid-- > 0)
+ cxgb4_port_mirror_free_rxq(adap,
+ &s->mirror_rxq[pi->port_id][rxqid]);
+
+ kfree(s->mirror_rxq[pi->port_id]);
+ s->mirror_rxq[pi->port_id] = NULL;
+ return ret;
+}
+
+static void cxgb4_port_mirror_free_queues(struct net_device *dev)
+{
+ struct port_info *pi = netdev2pinfo(dev);
+ struct adapter *adap = netdev2adap(dev);
+ struct sge *s = &adap->sge;
+ u16 i;
+
+ if (!refcount_read(&pi->vi_mirror_refcnt))
+ return;
+
+ if (!s->mirror_rxq[pi->port_id])
+ return;
+
+ for (i = 0; i < pi->nmirrorqsets; i++)
+ cxgb4_port_mirror_free_rxq(adap,
+ &s->mirror_rxq[pi->port_id][i]);
+
+ kfree(s->mirror_rxq[pi->port_id]);
+ s->mirror_rxq[pi->port_id] = NULL;
+}
+
int cxgb4_port_mirror_alloc(struct net_device *dev)
{
struct port_info *pi = netdev2pinfo(dev);
@@ -1307,6 +1463,20 @@ int cxgb4_port_mirror_alloc(struct net_device *dev)
refcount_set(&pi->vi_mirror_refcnt, 1);
+ if (adap->flags & CXGB4_FULL_INIT_DONE) {
+ ret = cxgb4_port_mirror_alloc_queues(dev);
+ if (ret)
+ goto out_free_vi;
+ }
+
+ mutex_unlock(&pi->vi_mirror_mutex);
+ return 0;
+
+out_free_vi:
+ refcount_set(&pi->vi_mirror_refcnt, 0);
+ t4_free_vi(adap, adap->mbox, adap->pf, 0, pi->viid_mirror);
+ pi->viid_mirror = 0;
+
out_unlock:
mutex_unlock(&pi->vi_mirror_mutex);
return ret;
@@ -1326,6 +1496,8 @@ void cxgb4_port_mirror_free(struct net_device *dev)
goto out_unlock;
}
+ cxgb4_port_mirror_free_queues(dev);
+
refcount_set(&pi->vi_mirror_refcnt, 0);
t4_free_vi(adap, adap->mbox, adap->pf, 0, pi->viid_mirror);
pi->viid_mirror = 0;
@@ -2606,8 +2778,22 @@ int cxgb_open(struct net_device *dev)
return err;
err = link_start(dev);
- if (!err)
- netif_tx_start_all_queues(dev);
+ if (err)
+ return err;
+
+ if (pi->nmirrorqsets) {
+ mutex_lock(&pi->vi_mirror_mutex);
+ err = cxgb4_port_mirror_alloc_queues(dev);
+ if (err)
+ goto out_unlock;
+ mutex_unlock(&pi->vi_mirror_mutex);
+ }
+
+ netif_tx_start_all_queues(dev);
+ return 0;
+
+out_unlock:
+ mutex_unlock(&pi->vi_mirror_mutex);
return err;
}
@@ -2625,7 +2811,16 @@ int cxgb_close(struct net_device *dev)
cxgb4_dcb_reset(dev);
dcb_tx_queue_prio_enable(dev, false);
#endif
- return ret;
+ if (ret)
+ return ret;
+
+ if (pi->nmirrorqsets) {
+ mutex_lock(&pi->vi_mirror_mutex);
+ cxgb4_port_mirror_free_queues(dev);
+ mutex_unlock(&pi->vi_mirror_mutex);
+ }
+
+ return 0;
}
int cxgb4_create_server_filter(const struct net_device *dev, unsigned int stid,
--
2.24.0
^ permalink raw reply related [flat|nested] 5+ messages in thread
* [PATCH net-next v2 3/3] cxgb4: add main VI to mirror VI config replication
2020-06-29 22:49 [PATCH net-next v2 0/3] cxgb4: add mirror action support for TC-MATCHALL Rahul Lakkireddy
2020-06-29 22:49 ` [PATCH net-next v2 1/3] cxgb4: add mirror action to TC-MATCHALL offload Rahul Lakkireddy
2020-06-29 22:49 ` [PATCH net-next v2 2/3] cxgb4: add support for mirror Rxqs Rahul Lakkireddy
@ 2020-06-29 22:49 ` Rahul Lakkireddy
2 siblings, 0 replies; 5+ messages in thread
From: Rahul Lakkireddy @ 2020-06-29 22:49 UTC (permalink / raw
To: netdev; +Cc: davem, kuba, nirranjan, vishal, dt
When mirror VI is enabled, replicate various VI config params
enabled on main VI to mirror VI. These include replicating MTU,
promiscuous mode, all-multicast mode, and enabled netdev Rx
feature offloads.
v2:
- Simplify the replication code by refactoring t4_set_rxmode()
to handle mirror VI, instead of duplicating the t4_set_rxmode()
calls in multiple places.
Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
---
drivers/net/ethernet/chelsio/cxgb4/cxgb4.h | 4 +-
.../net/ethernet/chelsio/cxgb4/cxgb4_main.c | 100 ++++++++++++++++--
drivers/net/ethernet/chelsio/cxgb4/t4_hw.c | 27 ++++-
3 files changed, 114 insertions(+), 17 deletions(-)
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
index 8985d85a1530..cc009f7fa530 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
@@ -1984,8 +1984,8 @@ int t4_free_vi(struct adapter *adap, unsigned int mbox,
unsigned int pf, unsigned int vf,
unsigned int viid);
int t4_set_rxmode(struct adapter *adap, unsigned int mbox, unsigned int viid,
- int mtu, int promisc, int all_multi, int bcast, int vlanex,
- bool sleep_ok);
+ unsigned int viid_mirror, int mtu, int promisc, int all_multi,
+ int bcast, int vlanex, bool sleep_ok);
int t4_free_raw_mac_filt(struct adapter *adap, unsigned int viid,
const u8 *addr, const u8 *mask, unsigned int idx,
u8 lookup_type, u8 port_id, bool sleep_ok);
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
index 00d8badc666c..3742e849547b 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
@@ -435,8 +435,8 @@ static int set_rxmode(struct net_device *dev, int mtu, bool sleep_ok)
__dev_uc_sync(dev, cxgb4_mac_sync, cxgb4_mac_unsync);
__dev_mc_sync(dev, cxgb4_mac_sync, cxgb4_mac_unsync);
- return t4_set_rxmode(adapter, adapter->mbox, pi->viid, mtu,
- (dev->flags & IFF_PROMISC) ? 1 : 0,
+ return t4_set_rxmode(adapter, adapter->mbox, pi->viid, pi->viid_mirror,
+ mtu, (dev->flags & IFF_PROMISC) ? 1 : 0,
(dev->flags & IFF_ALLMULTI) ? 1 : 0, 1, -1,
sleep_ok);
}
@@ -503,15 +503,16 @@ int cxgb4_change_mac(struct port_info *pi, unsigned int viid,
*/
static int link_start(struct net_device *dev)
{
- int ret;
struct port_info *pi = netdev_priv(dev);
- unsigned int mb = pi->adapter->pf;
+ unsigned int mb = pi->adapter->mbox;
+ int ret;
/*
* We do not set address filters and promiscuity here, the stack does
* that step explicitly.
*/
- ret = t4_set_rxmode(pi->adapter, mb, pi->viid, dev->mtu, -1, -1, -1,
+ ret = t4_set_rxmode(pi->adapter, mb, pi->viid, pi->viid_mirror,
+ dev->mtu, -1, -1, -1,
!!(dev->features & NETIF_F_HW_VLAN_CTAG_RX), true);
if (ret == 0)
ret = cxgb4_update_mac_filt(pi, pi->viid, &pi->xact_addr_filt,
@@ -1270,15 +1271,15 @@ int cxgb4_set_rspq_intr_params(struct sge_rspq *q,
static int cxgb_set_features(struct net_device *dev, netdev_features_t features)
{
- const struct port_info *pi = netdev_priv(dev);
netdev_features_t changed = dev->features ^ features;
+ const struct port_info *pi = netdev_priv(dev);
int err;
if (!(changed & NETIF_F_HW_VLAN_CTAG_RX))
return 0;
- err = t4_set_rxmode(pi->adapter, pi->adapter->pf, pi->viid, -1,
- -1, -1, -1,
+ err = t4_set_rxmode(pi->adapter, pi->adapter->mbox, pi->viid,
+ pi->viid_mirror, -1, -1, -1, -1,
!!(features & NETIF_F_HW_VLAN_CTAG_RX), true);
if (unlikely(err))
dev->features = features ^ NETIF_F_HW_VLAN_CTAG_RX;
@@ -1441,6 +1442,74 @@ static void cxgb4_port_mirror_free_queues(struct net_device *dev)
s->mirror_rxq[pi->port_id] = NULL;
}
+static int cxgb4_port_mirror_start(struct net_device *dev)
+{
+ struct port_info *pi = netdev2pinfo(dev);
+ struct adapter *adap = netdev2adap(dev);
+ int ret, idx = -1;
+
+ if (!refcount_read(&pi->vi_mirror_refcnt))
+ return 0;
+
+ /* Mirror VIs can be created dynamically after stack had
+ * already setup Rx modes like MTU, promisc, allmulti, etc.
+ * on main VI. So, parse what the stack had setup on the
+ * main VI and update the same on the mirror VI.
+ */
+ ret = t4_set_rxmode(adap, adap->mbox, pi->viid, pi->viid_mirror,
+ dev->mtu, (dev->flags & IFF_PROMISC) ? 1 : 0,
+ (dev->flags & IFF_ALLMULTI) ? 1 : 0, 1,
+ !!(dev->features & NETIF_F_HW_VLAN_CTAG_RX), true);
+ if (ret) {
+ dev_err(adap->pdev_dev,
+ "Failed start up Rx mode for Mirror VI 0x%x, ret: %d\n",
+ pi->viid_mirror, ret);
+ return ret;
+ }
+
+ /* Enable replication bit for the device's MAC address
+ * in MPS TCAM, so that the packets for the main VI are
+ * replicated to mirror VI.
+ */
+ ret = cxgb4_update_mac_filt(pi, pi->viid_mirror, &idx,
+ dev->dev_addr, true, NULL);
+ if (ret) {
+ dev_err(adap->pdev_dev,
+ "Failed updating MAC filter for Mirror VI 0x%x, ret: %d\n",
+ pi->viid_mirror, ret);
+ return ret;
+ }
+
+ /* Enabling a Virtual Interface can result in an interrupt
+ * during the processing of the VI Enable command and, in some
+ * paths, result in an attempt to issue another command in the
+ * interrupt context. Thus, we disable interrupts during the
+ * course of the VI Enable command ...
+ */
+ local_bh_disable();
+ ret = t4_enable_vi_params(adap, adap->mbox, pi->viid_mirror, true, true,
+ false);
+ local_bh_enable();
+ if (ret)
+ dev_err(adap->pdev_dev,
+ "Failed starting Mirror VI 0x%x, ret: %d\n",
+ pi->viid_mirror, ret);
+
+ return ret;
+}
+
+static void cxgb4_port_mirror_stop(struct net_device *dev)
+{
+ struct port_info *pi = netdev2pinfo(dev);
+ struct adapter *adap = netdev2adap(dev);
+
+ if (!refcount_read(&pi->vi_mirror_refcnt))
+ return;
+
+ t4_enable_vi_params(adap, adap->mbox, pi->viid_mirror, false, false,
+ false);
+}
+
int cxgb4_port_mirror_alloc(struct net_device *dev)
{
struct port_info *pi = netdev2pinfo(dev);
@@ -1467,11 +1536,18 @@ int cxgb4_port_mirror_alloc(struct net_device *dev)
ret = cxgb4_port_mirror_alloc_queues(dev);
if (ret)
goto out_free_vi;
+
+ ret = cxgb4_port_mirror_start(dev);
+ if (ret)
+ goto out_free_queues;
}
mutex_unlock(&pi->vi_mirror_mutex);
return 0;
+out_free_queues:
+ cxgb4_port_mirror_free_queues(dev);
+
out_free_vi:
refcount_set(&pi->vi_mirror_refcnt, 0);
t4_free_vi(adap, adap->mbox, adap->pf, 0, pi->viid_mirror);
@@ -1496,6 +1572,7 @@ void cxgb4_port_mirror_free(struct net_device *dev)
goto out_unlock;
}
+ cxgb4_port_mirror_stop(dev);
cxgb4_port_mirror_free_queues(dev);
refcount_set(&pi->vi_mirror_refcnt, 0);
@@ -2816,6 +2893,7 @@ int cxgb_close(struct net_device *dev)
if (pi->nmirrorqsets) {
mutex_lock(&pi->vi_mirror_mutex);
+ cxgb4_port_mirror_stop(dev);
cxgb4_port_mirror_free_queues(dev);
mutex_unlock(&pi->vi_mirror_mutex);
}
@@ -3086,11 +3164,11 @@ static void cxgb_set_rxmode(struct net_device *dev)
static int cxgb_change_mtu(struct net_device *dev, int new_mtu)
{
- int ret;
struct port_info *pi = netdev_priv(dev);
+ int ret;
- ret = t4_set_rxmode(pi->adapter, pi->adapter->pf, pi->viid, new_mtu, -1,
- -1, -1, -1, true);
+ ret = t4_set_rxmode(pi->adapter, pi->adapter->mbox, pi->viid,
+ pi->viid_mirror, new_mtu, -1, -1, -1, -1, true);
if (!ret)
dev->mtu = new_mtu;
return ret;
diff --git a/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c b/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c
index 7876aa392aae..0af5ee9975df 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c
@@ -7711,6 +7711,7 @@ int t4_free_vi(struct adapter *adap, unsigned int mbox, unsigned int pf,
* @adap: the adapter
* @mbox: mailbox to use for the FW command
* @viid: the VI id
+ * @viid_mirror: the mirror VI id
* @mtu: the new MTU or -1
* @promisc: 1 to enable promiscuous mode, 0 to disable it, -1 no change
* @all_multi: 1 to enable all-multi mode, 0 to disable it, -1 no change
@@ -7721,10 +7722,11 @@ int t4_free_vi(struct adapter *adap, unsigned int mbox, unsigned int pf,
* Sets Rx properties of a virtual interface.
*/
int t4_set_rxmode(struct adapter *adap, unsigned int mbox, unsigned int viid,
- int mtu, int promisc, int all_multi, int bcast, int vlanex,
- bool sleep_ok)
+ unsigned int viid_mirror, int mtu, int promisc, int all_multi,
+ int bcast, int vlanex, bool sleep_ok)
{
- struct fw_vi_rxmode_cmd c;
+ struct fw_vi_rxmode_cmd c, c_mirror;
+ int ret;
/* convert to FW values */
if (mtu < 0)
@@ -7749,7 +7751,24 @@ int t4_set_rxmode(struct adapter *adap, unsigned int mbox, unsigned int viid,
FW_VI_RXMODE_CMD_ALLMULTIEN_V(all_multi) |
FW_VI_RXMODE_CMD_BROADCASTEN_V(bcast) |
FW_VI_RXMODE_CMD_VLANEXEN_V(vlanex));
- return t4_wr_mbox_meat(adap, mbox, &c, sizeof(c), NULL, sleep_ok);
+
+ if (viid_mirror) {
+ memcpy(&c_mirror, &c, sizeof(c_mirror));
+ c_mirror.op_to_viid =
+ cpu_to_be32(FW_CMD_OP_V(FW_VI_RXMODE_CMD) |
+ FW_CMD_REQUEST_F | FW_CMD_WRITE_F |
+ FW_VI_RXMODE_CMD_VIID_V(viid_mirror));
+ }
+
+ ret = t4_wr_mbox_meat(adap, mbox, &c, sizeof(c), NULL, sleep_ok);
+ if (ret)
+ return ret;
+
+ if (viid_mirror)
+ ret = t4_wr_mbox_meat(adap, mbox, &c_mirror, sizeof(c_mirror),
+ NULL, sleep_ok);
+
+ return ret;
}
/**
--
2.24.0
^ permalink raw reply related [flat|nested] 5+ messages in thread
* Re: [PATCH net-next v2 1/3] cxgb4: add mirror action to TC-MATCHALL offload
2020-06-29 22:49 ` [PATCH net-next v2 1/3] cxgb4: add mirror action to TC-MATCHALL offload Rahul Lakkireddy
@ 2020-06-30 0:34 ` Jakub Kicinski
0 siblings, 0 replies; 5+ messages in thread
From: Jakub Kicinski @ 2020-06-30 0:34 UTC (permalink / raw
To: Rahul Lakkireddy; +Cc: netdev, davem, nirranjan, vishal, dt
On Tue, 30 Jun 2020 04:19:51 +0530 Rahul Lakkireddy wrote:
> + if (refcount_read(&pi->vi_mirror_refcnt) > 1) {
> + refcount_dec(&pi->vi_mirror_refcnt);
> + goto out_unlock;
> + }
Please remove the use of refcount_t if you're doing this, otherwise a
person looking at this code will waste time trying to validate if this
clearly bogus use of refcount_t is harmless or not.
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2020-06-30 0:34 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2020-06-29 22:49 [PATCH net-next v2 0/3] cxgb4: add mirror action support for TC-MATCHALL Rahul Lakkireddy
2020-06-29 22:49 ` [PATCH net-next v2 1/3] cxgb4: add mirror action to TC-MATCHALL offload Rahul Lakkireddy
2020-06-30 0:34 ` Jakub Kicinski
2020-06-29 22:49 ` [PATCH net-next v2 2/3] cxgb4: add support for mirror Rxqs Rahul Lakkireddy
2020-06-29 22:49 ` [PATCH net-next v2 3/3] cxgb4: add main VI to mirror VI config replication Rahul Lakkireddy
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.