* [PATCH net-next 1/7] cxgb4: save additional filter tuple field shifts in tp_params
2017-11-01 3:22 [PATCH net-next 0/7] cxgb4: add hash-filter support to tc-flower offload Rahul Lakkireddy
@ 2017-11-01 3:22 ` Rahul Lakkireddy
2017-11-01 3:23 ` [PATCH net-next 2/7] cxgb4: initialize hash-filter configuration Rahul Lakkireddy
` (6 subsequent siblings)
7 siblings, 0 replies; 9+ messages in thread
From: Rahul Lakkireddy @ 2017-11-01 3:22 UTC (permalink / raw)
To: netdev; +Cc: davem, kumaras, ganeshgr, nirranjan, indranil, Rahul Lakkireddy
From: Kumar Sanghvi <kumaras@chelsio.com>
Save additional filter tuple field shifts in tp_params based on
configured filter tuple fields.
Also, save the combined filter tuple mask based on configured
filter tuple fields.
Signed-off-by: Kumar Sanghvi <kumaras@chelsio.com>
Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
---
drivers/net/ethernet/chelsio/cxgb4/cxgb4.h | 12 ++++++++++--
drivers/net/ethernet/chelsio/cxgb4/t4_hw.c | 18 ++++++++++++++++--
drivers/net/ethernet/chelsio/cxgb4/t4_regs.h | 3 +++
3 files changed, 29 insertions(+), 4 deletions(-)
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
index a57761b28edc..e2c75b73595f 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
@@ -287,10 +287,18 @@ struct tp_params {
* places we store their offsets here, or a -1 if the field isn't
* present.
*/
- int vlan_shift;
- int vnic_shift;
+ int fcoe_shift;
int port_shift;
+ int vnic_shift;
+ int vlan_shift;
+ int tos_shift;
int protocol_shift;
+ int ethertype_shift;
+ int macmatch_shift;
+ int matchtype_shift;
+ int frag_shift;
+
+ u64 hash_filter_mask;
};
struct vpd_params {
diff --git a/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c b/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c
index c289ca1efc1b..efe9d3a20135 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c
@@ -8816,11 +8816,21 @@ int t4_init_tp_params(struct adapter *adap, bool sleep_ok)
* shift positions of several elements of the Compressed Filter Tuple
* for this adapter which we need frequently ...
*/
- adap->params.tp.vlan_shift = t4_filter_field_shift(adap, VLAN_F);
- adap->params.tp.vnic_shift = t4_filter_field_shift(adap, VNIC_ID_F);
+ adap->params.tp.fcoe_shift = t4_filter_field_shift(adap, FCOE_F);
adap->params.tp.port_shift = t4_filter_field_shift(adap, PORT_F);
+ adap->params.tp.vnic_shift = t4_filter_field_shift(adap, VNIC_ID_F);
+ adap->params.tp.vlan_shift = t4_filter_field_shift(adap, VLAN_F);
+ adap->params.tp.tos_shift = t4_filter_field_shift(adap, TOS_F);
adap->params.tp.protocol_shift = t4_filter_field_shift(adap,
PROTOCOL_F);
+ adap->params.tp.ethertype_shift = t4_filter_field_shift(adap,
+ ETHERTYPE_F);
+ adap->params.tp.macmatch_shift = t4_filter_field_shift(adap,
+ MACMATCH_F);
+ adap->params.tp.matchtype_shift = t4_filter_field_shift(adap,
+ MPSHITTYPE_F);
+ adap->params.tp.frag_shift = t4_filter_field_shift(adap,
+ FRAGMENTATION_F);
/* If TP_INGRESS_CONFIG.VNID == 0, then TP_VLAN_PRI_MAP.VNIC_ID
* represents the presence of an Outer VLAN instead of a VNIC ID.
@@ -8828,6 +8838,10 @@ int t4_init_tp_params(struct adapter *adap, bool sleep_ok)
if ((adap->params.tp.ingress_config & VNIC_F) == 0)
adap->params.tp.vnic_shift = -1;
+ v = t4_read_reg(adap, LE_3_DB_HASH_MASK_GEN_IPV4_T6_A);
+ adap->params.tp.hash_filter_mask = v;
+ v = t4_read_reg(adap, LE_4_DB_HASH_MASK_GEN_IPV4_T6_A);
+ adap->params.tp.hash_filter_mask |= ((u64)v << 32);
return 0;
}
diff --git a/drivers/net/ethernet/chelsio/cxgb4/t4_regs.h b/drivers/net/ethernet/chelsio/cxgb4/t4_regs.h
index 483fb7644355..44713bad0045 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/t4_regs.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/t4_regs.h
@@ -2933,6 +2933,9 @@
#define SSRAMINTPERR_V(x) ((x) << SSRAMINTPERR_S)
#define SSRAMINTPERR_F SSRAMINTPERR_V(1U)
+#define LE_3_DB_HASH_MASK_GEN_IPV4_T6_A 0x19eac
+#define LE_4_DB_HASH_MASK_GEN_IPV4_T6_A 0x19eb0
+
#define NCSI_INT_CAUSE_A 0x1a0d8
#define CIM_DM_PRTY_ERR_S 8
--
2.14.1
^ permalink raw reply related [flat|nested] 9+ messages in thread* [PATCH net-next 2/7] cxgb4: initialize hash-filter configuration
2017-11-01 3:22 [PATCH net-next 0/7] cxgb4: add hash-filter support to tc-flower offload Rahul Lakkireddy
2017-11-01 3:22 ` [PATCH net-next 1/7] cxgb4: save additional filter tuple field shifts in tp_params Rahul Lakkireddy
@ 2017-11-01 3:23 ` Rahul Lakkireddy
2017-11-01 3:23 ` [PATCH net-next 3/7] cxgb4: add support to create hash filters Rahul Lakkireddy
` (5 subsequent siblings)
7 siblings, 0 replies; 9+ messages in thread
From: Rahul Lakkireddy @ 2017-11-01 3:23 UTC (permalink / raw)
To: netdev; +Cc: davem, kumaras, ganeshgr, nirranjan, indranil, Rahul Lakkireddy
From: Kumar Sanghvi <kumaras@chelsio.com>
Add support for hash-filter configuration on T6. Also, do basic
checks for the related initialization.
Signed-off-by: Kumar Sanghvi <kumaras@chelsio.com>
Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
---
drivers/net/ethernet/chelsio/cxgb4/cxgb4.h | 6 ++++++
drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c | 22 ++++++++++++++++++++++
drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.h | 1 +
drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c | 14 ++++++++++----
drivers/net/ethernet/chelsio/cxgb4/t4_regs.h | 14 ++++++++++++++
drivers/net/ethernet/chelsio/cxgb4/t4fw_api.h | 1 +
6 files changed, 54 insertions(+), 4 deletions(-)
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
index e2c75b73595f..5f021e6062b0 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
@@ -366,6 +366,7 @@ struct adapter_params {
unsigned char crypto; /* HW capability for crypto */
unsigned char bypass;
+ unsigned char hash_filter;
unsigned int ofldq_wr_cred;
bool ulptx_memwrite_dsgl; /* use of T5 DSGL allowed */
@@ -1140,6 +1141,11 @@ static inline int is_offload(const struct adapter *adap)
return adap->params.offload;
}
+static inline int is_hashfilter(const struct adapter *adap)
+{
+ return adap->params.hash_filter;
+}
+
static inline int is_pci_uld(const struct adapter *adap)
{
return adap->params.crypto;
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c
index 89272f29f807..566bd2d3737c 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c
@@ -915,3 +915,25 @@ void filter_rpl(struct adapter *adap, const struct cpl_set_tcb_rpl *rpl)
complete(&ctx->completion);
}
}
+
+int init_hash_filter(struct adapter *adap)
+{
+ /* On T6, verify the necessary register configs and warn the user in
+ * case of improper config
+ */
+ if (is_t6(adap->params.chip)) {
+ if (TCAM_ACTV_HIT_G(t4_read_reg(adap, LE_DB_RSP_CODE_0_A)) != 4)
+ goto err;
+
+ if (HASH_ACTV_HIT_G(t4_read_reg(adap, LE_DB_RSP_CODE_1_A)) != 4)
+ goto err;
+ } else {
+ dev_err(adap->pdev_dev, "Hash filter supported only on T6\n");
+ return -EINVAL;
+ }
+ adap->params.hash_filter = 1;
+ return 0;
+err:
+ dev_warn(adap->pdev_dev, "Invalid hash filter config!\n");
+ return -EINVAL;
+}
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.h
index 23742cb1c69f..d3c1a8fafd32 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.h
@@ -45,4 +45,5 @@ int delete_filter(struct adapter *adapter, unsigned int fidx);
int writable_filter(struct filter_entry *f);
void clear_all_filters(struct adapter *adapter);
+int init_hash_filter(struct adapter *adap);
#endif /* __CXGB4_FILTER_H */
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
index e16078ddb39f..4b07cfe8c66c 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
@@ -3963,7 +3963,8 @@ static int adap_init0(struct adapter *adap)
if (ret < 0)
goto bye;
- if (caps_cmd.ofldcaps) {
+ if (caps_cmd.ofldcaps ||
+ (caps_cmd.niccaps & htons(FW_CAPS_CONFIG_NIC_HASHFILTER))) {
/* query offload-related parameters */
params[0] = FW_PARAM_DEV(NTID);
params[1] = FW_PARAM_PFVF(SERVER_START);
@@ -4000,8 +4001,13 @@ static int adap_init0(struct adapter *adap)
adap->vres.ddp.size = val[4] - val[3] + 1;
adap->params.ofldq_wr_cred = val[5];
- adap->params.offload = 1;
- adap->num_ofld_uld += 1;
+ if (caps_cmd.niccaps & htons(FW_CAPS_CONFIG_NIC_HASHFILTER)) {
+ if (init_hash_filter(adap) < 0)
+ goto bye;
+ } else {
+ adap->params.offload = 1;
+ adap->num_ofld_uld += 1;
+ }
}
if (caps_cmd.rdmacaps) {
params[0] = FW_PARAM_PFVF(STAG_START);
@@ -5171,7 +5177,7 @@ static int init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
cxgb4_init_tc_flower(adapter);
}
- if (is_offload(adapter)) {
+ if (is_offload(adapter) || is_hashfilter(adapter)) {
if (t4_read_reg(adapter, LE_DB_CONFIG_A) & HASHEN_F) {
u32 hash_base, hash_reg;
diff --git a/drivers/net/ethernet/chelsio/cxgb4/t4_regs.h b/drivers/net/ethernet/chelsio/cxgb4/t4_regs.h
index 44713bad0045..623f453bd327 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/t4_regs.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/t4_regs.h
@@ -2933,6 +2933,20 @@
#define SSRAMINTPERR_V(x) ((x) << SSRAMINTPERR_S)
#define SSRAMINTPERR_F SSRAMINTPERR_V(1U)
+#define LE_DB_RSP_CODE_0_A 0x19c74
+
+#define TCAM_ACTV_HIT_S 0
+#define TCAM_ACTV_HIT_M 0x1fU
+#define TCAM_ACTV_HIT_V(x) ((x) << TCAM_ACTV_HIT_S)
+#define TCAM_ACTV_HIT_G(x) (((x) >> TCAM_ACTV_HIT_S) & TCAM_ACTV_HIT_M)
+
+#define LE_DB_RSP_CODE_1_A 0x19c78
+
+#define HASH_ACTV_HIT_S 25
+#define HASH_ACTV_HIT_M 0x1fU
+#define HASH_ACTV_HIT_V(x) ((x) << HASH_ACTV_HIT_S)
+#define HASH_ACTV_HIT_G(x) (((x) >> HASH_ACTV_HIT_S) & HASH_ACTV_HIT_M)
+
#define LE_3_DB_HASH_MASK_GEN_IPV4_T6_A 0x19eac
#define LE_4_DB_HASH_MASK_GEN_IPV4_T6_A 0x19eb0
diff --git a/drivers/net/ethernet/chelsio/cxgb4/t4fw_api.h b/drivers/net/ethernet/chelsio/cxgb4/t4fw_api.h
index 2ba890926c73..57eb4ad3485d 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/t4fw_api.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/t4fw_api.h
@@ -1092,6 +1092,7 @@ enum fw_caps_config_switch {
enum fw_caps_config_nic {
FW_CAPS_CONFIG_NIC = 0x00000001,
FW_CAPS_CONFIG_NIC_VM = 0x00000002,
+ FW_CAPS_CONFIG_NIC_HASHFILTER = 0x00000020,
};
enum fw_caps_config_ofld {
--
2.14.1
^ permalink raw reply related [flat|nested] 9+ messages in thread* [PATCH net-next 3/7] cxgb4: add support to create hash filters
2017-11-01 3:22 [PATCH net-next 0/7] cxgb4: add hash-filter support to tc-flower offload Rahul Lakkireddy
2017-11-01 3:22 ` [PATCH net-next 1/7] cxgb4: save additional filter tuple field shifts in tp_params Rahul Lakkireddy
2017-11-01 3:23 ` [PATCH net-next 2/7] cxgb4: initialize hash-filter configuration Rahul Lakkireddy
@ 2017-11-01 3:23 ` Rahul Lakkireddy
2017-11-01 3:23 ` [PATCH net-next 4/7] cxgb4: add support to delete hash filter Rahul Lakkireddy
` (4 subsequent siblings)
7 siblings, 0 replies; 9+ messages in thread
From: Rahul Lakkireddy @ 2017-11-01 3:23 UTC (permalink / raw)
To: netdev; +Cc: davem, kumaras, ganeshgr, nirranjan, indranil, Rahul Lakkireddy
From: Kumar Sanghvi <kumaras@chelsio.com>
Add support to create hash (exact-match) filters based on the value
of 'hash' field in ch_filter_specification.
Allocate SMT/L2T entries if DMAC-rewrite/SMAC-rewrite is requested.
Allocate CLIP entry in case of IPv6 filter.
Use cpl_act_open_req[6] to send hash filter create request to hw.
Also, the filter tuple is calculated as part of sending this request.
Hash-filter reply is processed on getting cpl_act_open_rpl.
In case of success, various bits/fields in filter-tcb are set per
filter requirement, such as enabling filter hitcnts, and/or various
header rewrite operations, such as VLAN-rewrite, NAT or
(L3/L4)-rewrite, and SMAC/DMAC-rewrite. In case of failure, clear the
filter entry and release any hw resources occupied by it.
The patch also moves the functions set_tcb_field, set_tcb_tflag and
configure_filter_smac towards beginning of file.
Signed-off-by: Kumar Sanghvi <kumaras@chelsio.com>
Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
---
drivers/net/ethernet/chelsio/cxgb4/cxgb4.h | 10 +-
drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c | 573 ++++++++++++++++++++--
drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.h | 3 +
drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c | 4 +
drivers/net/ethernet/chelsio/cxgb4/t4_msg.h | 5 +
drivers/net/ethernet/chelsio/cxgb4/t4_tcb.h | 17 +
6 files changed, 562 insertions(+), 50 deletions(-)
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
index 5f021e6062b0..bb7f0e4c9a81 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
@@ -1050,6 +1050,7 @@ struct ch_filter_specification {
* matching that doesn't exist as a (value, mask) tuple.
*/
uint32_t type:1; /* 0 => IPv4, 1 => IPv6 */
+ u32 hash:1; /* 0 => wild-card, 1 => exact-match */
/* Packet dispatch information. Ingress packets which match the
* filter rules will be dropped, passed to the host or switched back
@@ -1107,7 +1108,14 @@ enum {
};
enum {
- NAT_MODE_ALL = 7, /* NAT on entire 4-tuple */
+ NAT_MODE_NONE = 0, /* No NAT performed */
+ NAT_MODE_DIP, /* NAT on Dst IP */
+ NAT_MODE_DIP_DP, /* NAT on Dst IP, Dst Port */
+ NAT_MODE_DIP_DP_SIP, /* NAT on Dst IP, Dst Port and Src IP */
+ NAT_MODE_DIP_DP_SP, /* NAT on Dst IP, Dst Port and Src Port */
+ NAT_MODE_SIP_SP, /* NAT on Src IP and Src Port */
+ NAT_MODE_DIP_SIP_SP, /* NAT on Dst IP, Src IP and Src Port */
+ NAT_MODE_ALL /* NAT on entire 4-tuple */
};
/* Host shadow copy of ingress filter entry. This is in host native format
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c
index 566bd2d3737c..cf8ca695f27d 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c
@@ -35,6 +35,8 @@
#include "cxgb4.h"
#include "t4_regs.h"
#include "t4_tcb.h"
+#include "t4_values.h"
+#include "clip_tbl.h"
#include "l2t.h"
#include "smt.h"
#include "t4fw_api.h"
@@ -50,6 +52,141 @@ static inline bool unsupported(u32 conf, u32 conf_mask, u32 val, u32 mask)
return !(conf & conf_mask) && is_field_set(val, mask);
}
+static int set_tcb_field(struct adapter *adap, struct filter_entry *f,
+ unsigned int ftid, u16 word, u64 mask, u64 val,
+ int no_reply)
+{
+ struct cpl_set_tcb_field *req;
+ struct sk_buff *skb;
+
+ skb = alloc_skb(sizeof(struct cpl_set_tcb_field), GFP_ATOMIC);
+ if (!skb)
+ return -ENOMEM;
+
+ req = (struct cpl_set_tcb_field *)__skb_put(skb, sizeof(*req));
+ memset(req, 0, sizeof(*req));
+ INIT_TP_WR_CPL(req, CPL_SET_TCB_FIELD, ftid);
+ req->reply_ctrl = htons(REPLY_CHAN_V(0) |
+ QUEUENO_V(adap->sge.fw_evtq.abs_id) |
+ NO_REPLY_V(no_reply));
+ req->word_cookie = htons(TCB_WORD_V(word) | TCB_COOKIE_V(ftid));
+ req->mask = cpu_to_be64(mask);
+ req->val = cpu_to_be64(val);
+ set_wr_txq(skb, CPL_PRIORITY_CONTROL, f->fs.val.iport & 0x3);
+ t4_ofld_send(adap, skb);
+ return 0;
+}
+
+/* Set one of the t_flags bits in the TCB.
+ */
+static int set_tcb_tflag(struct adapter *adap, struct filter_entry *f,
+ unsigned int ftid, unsigned int bit_pos,
+ unsigned int val, int no_reply)
+{
+ return set_tcb_field(adap, f, ftid, TCB_T_FLAGS_W, 1ULL << bit_pos,
+ (unsigned long long)val << bit_pos, no_reply);
+}
+
+static int configure_filter_smac(struct adapter *adap, struct filter_entry *f)
+{
+ int err;
+
+ /* do a set-tcb for smac-sel and CWR bit.. */
+ err = set_tcb_tflag(adap, f, f->tid, TF_CCTRL_CWR_S, 1, 1);
+ if (err)
+ goto smac_err;
+
+ err = set_tcb_field(adap, f, f->tid, TCB_SMAC_SEL_W,
+ TCB_SMAC_SEL_V(TCB_SMAC_SEL_M),
+ TCB_SMAC_SEL_V(f->smt->idx), 1);
+ if (!err)
+ return 0;
+
+smac_err:
+ dev_err(adap->pdev_dev, "filter %u smac config failed with error %u\n",
+ f->tid, err);
+ return err;
+}
+
+static void set_nat_params(struct adapter *adap, struct filter_entry *f,
+ unsigned int tid, bool dip, bool sip, bool dp,
+ bool sp)
+{
+ if (dip) {
+ if (f->fs.type) {
+ set_tcb_field(adap, f, tid, TCB_SND_UNA_RAW_W,
+ WORD_MASK, f->fs.nat_lip[15] |
+ f->fs.nat_lip[14] << 8 |
+ f->fs.nat_lip[13] << 16 |
+ f->fs.nat_lip[12] << 24, 1);
+
+ set_tcb_field(adap, f, tid, TCB_SND_UNA_RAW_W + 1,
+ WORD_MASK, f->fs.nat_lip[11] |
+ f->fs.nat_lip[10] << 8 |
+ f->fs.nat_lip[9] << 16 |
+ f->fs.nat_lip[8] << 24, 1);
+
+ set_tcb_field(adap, f, tid, TCB_SND_UNA_RAW_W + 2,
+ WORD_MASK, f->fs.nat_lip[7] |
+ f->fs.nat_lip[6] << 8 |
+ f->fs.nat_lip[5] << 16 |
+ f->fs.nat_lip[4] << 24, 1);
+
+ set_tcb_field(adap, f, tid, TCB_SND_UNA_RAW_W + 3,
+ WORD_MASK, f->fs.nat_lip[3] |
+ f->fs.nat_lip[2] << 8 |
+ f->fs.nat_lip[1] << 16 |
+ f->fs.nat_lip[0] << 24, 1);
+ } else {
+ set_tcb_field(adap, f, tid, TCB_RX_FRAG3_LEN_RAW_W,
+ WORD_MASK, f->fs.nat_lip[3] |
+ f->fs.nat_lip[2] << 8 |
+ f->fs.nat_lip[1] << 16 |
+ f->fs.nat_lip[0] << 24, 1);
+ }
+ }
+
+ if (sip) {
+ if (f->fs.type) {
+ set_tcb_field(adap, f, tid, TCB_RX_FRAG2_PTR_RAW_W,
+ WORD_MASK, f->fs.nat_fip[15] |
+ f->fs.nat_fip[14] << 8 |
+ f->fs.nat_fip[13] << 16 |
+ f->fs.nat_fip[12] << 24, 1);
+
+ set_tcb_field(adap, f, tid, TCB_RX_FRAG2_PTR_RAW_W + 1,
+ WORD_MASK, f->fs.nat_fip[11] |
+ f->fs.nat_fip[10] << 8 |
+ f->fs.nat_fip[9] << 16 |
+ f->fs.nat_fip[8] << 24, 1);
+
+ set_tcb_field(adap, f, tid, TCB_RX_FRAG2_PTR_RAW_W + 2,
+ WORD_MASK, f->fs.nat_fip[7] |
+ f->fs.nat_fip[6] << 8 |
+ f->fs.nat_fip[5] << 16 |
+ f->fs.nat_fip[4] << 24, 1);
+
+ set_tcb_field(adap, f, tid, TCB_RX_FRAG2_PTR_RAW_W + 3,
+ WORD_MASK, f->fs.nat_fip[3] |
+ f->fs.nat_fip[2] << 8 |
+ f->fs.nat_fip[1] << 16 |
+ f->fs.nat_fip[0] << 24, 1);
+
+ } else {
+ set_tcb_field(adap, f, tid,
+ TCB_RX_FRAG3_START_IDX_OFFSET_RAW_W,
+ WORD_MASK, f->fs.nat_fip[3] |
+ f->fs.nat_fip[2] << 8 |
+ f->fs.nat_fip[1] << 16 |
+ f->fs.nat_fip[0] << 24, 1);
+ }
+ }
+
+ set_tcb_field(adap, f, tid, TCB_PDU_HDR_LEN_W, WORD_MASK,
+ (dp ? f->fs.nat_lport : 0) |
+ (sp ? f->fs.nat_fport << 16 : 0), 1);
+}
+
/* Validate filter spec against configuration done on the card. */
static int validate_filter(struct net_device *dev,
struct ch_filter_specification *fs)
@@ -484,10 +621,8 @@ int delete_filter(struct adapter *adapter, unsigned int fidx)
void clear_filter(struct adapter *adap, struct filter_entry *f)
{
/* If the new or old filter have loopback rewriteing rules then we'll
- * need to free any existing Layer Two Table (L2T) entries of the old
- * filter rule. The firmware will handle freeing up any Source MAC
- * Table (SMT) entries used for rewriting Source MAC Addresses in
- * loopback rules.
+ * need to free any existing L2T, SMT, CLIP entries of filter
+ * rule.
*/
if (f->l2t)
cxgb4_l2t_release(f->l2t);
@@ -495,6 +630,9 @@ void clear_filter(struct adapter *adap, struct filter_entry *f)
if (f->smt)
cxgb4_smt_release(f->smt);
+ if (f->fs.hash && f->fs.type)
+ cxgb4_clip_release(f->dev, (const u32 *)&f->fs.val.lip, 1);
+
/* The zeroing of the filter rule below clears the filter valid,
* pending, locked flags, l2t pointer, etc. so it's all we need for
* this operation.
@@ -564,6 +702,269 @@ static void fill_default_mask(struct ch_filter_specification *fs)
fs->mask.fport = ~0;
}
+static u64 hash_filter_ntuple(struct ch_filter_specification *fs,
+ struct net_device *dev)
+{
+ struct adapter *adap = netdev2adap(dev);
+ struct tp_params *tp = &adap->params.tp;
+ u64 ntuple = 0;
+
+ /* Initialize each of the fields which we care about which are present
+ * in the Compressed Filter Tuple.
+ */
+ if (tp->vlan_shift >= 0 && fs->mask.ivlan)
+ ntuple |= (FT_VLAN_VLD_F | fs->val.ivlan) << tp->vlan_shift;
+
+ if (tp->port_shift >= 0 && fs->mask.iport)
+ ntuple |= (u64)fs->val.iport << tp->port_shift;
+
+ if (tp->protocol_shift >= 0) {
+ if (!fs->val.proto)
+ ntuple |= (u64)IPPROTO_TCP << tp->protocol_shift;
+ else
+ ntuple |= (u64)fs->val.proto << tp->protocol_shift;
+ }
+
+ if (tp->tos_shift >= 0 && fs->mask.tos)
+ ntuple |= (u64)(fs->val.tos) << tp->tos_shift;
+
+ if (tp->vnic_shift >= 0) {
+ if ((adap->params.tp.ingress_config & VNIC_F) &&
+ fs->mask.pfvf_vld)
+ ntuple |= (u64)((fs->val.pfvf_vld << 16) |
+ (fs->val.pf << 13) |
+ (fs->val.vf)) << tp->vnic_shift;
+ else
+ ntuple |= (u64)((fs->val.ovlan_vld << 16) |
+ (fs->val.ovlan)) << tp->vnic_shift;
+ }
+
+ if (tp->macmatch_shift >= 0 && fs->mask.macidx)
+ ntuple |= (u64)(fs->val.macidx) << tp->macmatch_shift;
+
+ if (tp->ethertype_shift >= 0 && fs->mask.ethtype)
+ ntuple |= (u64)(fs->val.ethtype) << tp->ethertype_shift;
+
+ if (tp->matchtype_shift >= 0 && fs->mask.matchtype)
+ ntuple |= (u64)(fs->val.matchtype) << tp->matchtype_shift;
+
+ if (tp->frag_shift >= 0 && fs->mask.frag)
+ ntuple |= (u64)(fs->val.frag) << tp->frag_shift;
+
+ if (tp->fcoe_shift >= 0 && fs->mask.fcoe)
+ ntuple |= (u64)(fs->val.fcoe) << tp->fcoe_shift;
+ return ntuple;
+}
+
+static void mk_act_open_req6(struct filter_entry *f, struct sk_buff *skb,
+ unsigned int qid_filterid, struct adapter *adap)
+{
+ struct cpl_t6_act_open_req6 *t6req = NULL;
+ struct cpl_act_open_req6 *req = NULL;
+
+ t6req = (struct cpl_t6_act_open_req6 *)__skb_put(skb, sizeof(*t6req));
+ INIT_TP_WR(t6req, 0);
+ req = (struct cpl_act_open_req6 *)t6req;
+ OPCODE_TID(req) = htonl(MK_OPCODE_TID(CPL_ACT_OPEN_REQ6, qid_filterid));
+ req->local_port = cpu_to_be16(f->fs.val.lport);
+ req->peer_port = cpu_to_be16(f->fs.val.fport);
+ req->local_ip_hi = *(__be64 *)(&f->fs.val.lip);
+ req->local_ip_lo = *(((__be64 *)&f->fs.val.lip) + 1);
+ req->peer_ip_hi = *(__be64 *)(&f->fs.val.fip);
+ req->peer_ip_lo = *(((__be64 *)&f->fs.val.fip) + 1);
+ req->opt0 = cpu_to_be64(NAGLE_V(f->fs.newvlan == VLAN_REMOVE ||
+ f->fs.newvlan == VLAN_REWRITE) |
+ DELACK_V(f->fs.hitcnts) |
+ L2T_IDX_V(f->l2t ? f->l2t->idx : 0) |
+ SMAC_SEL_V((cxgb4_port_viid(f->dev) &
+ 0x7F) << 1) |
+ TX_CHAN_V(f->fs.eport) |
+ NO_CONG_V(f->fs.rpttid) |
+ ULP_MODE_V(f->fs.nat_mode ?
+ ULP_MODE_TCPDDP : ULP_MODE_NONE) |
+ TCAM_BYPASS_F | NON_OFFLOAD_F);
+ t6req->params = cpu_to_be64(FILTER_TUPLE_V(hash_filter_ntuple(&f->fs,
+ f->dev)));
+ t6req->opt2 = htonl(RSS_QUEUE_VALID_F |
+ RSS_QUEUE_V(f->fs.iq) |
+ TX_QUEUE_V(f->fs.nat_mode) |
+ T5_OPT_2_VALID_F |
+ RX_CHANNEL_F |
+ CONG_CNTRL_V((f->fs.action == FILTER_DROP) |
+ (f->fs.dirsteer << 1)) |
+ PACE_V((f->fs.maskhash) |
+ ((f->fs.dirsteerhash) << 1)) |
+ CCTRL_ECN_V(f->fs.action == FILTER_SWITCH));
+}
+
+static void mk_act_open_req(struct filter_entry *f, struct sk_buff *skb,
+ unsigned int qid_filterid, struct adapter *adap)
+{
+ struct cpl_t6_act_open_req *t6req = NULL;
+ struct cpl_act_open_req *req = NULL;
+
+ t6req = (struct cpl_t6_act_open_req *)__skb_put(skb, sizeof(*t6req));
+ INIT_TP_WR(t6req, 0);
+ req = (struct cpl_act_open_req *)t6req;
+ OPCODE_TID(req) = htonl(MK_OPCODE_TID(CPL_ACT_OPEN_REQ, qid_filterid));
+ req->local_port = cpu_to_be16(f->fs.val.lport);
+ req->peer_port = cpu_to_be16(f->fs.val.fport);
+ req->local_ip = f->fs.val.lip[0] | f->fs.val.lip[1] << 8 |
+ f->fs.val.lip[2] << 16 | f->fs.val.lip[3] << 24;
+ req->peer_ip = f->fs.val.fip[0] | f->fs.val.fip[1] << 8 |
+ f->fs.val.fip[2] << 16 | f->fs.val.fip[3] << 24;
+ req->opt0 = cpu_to_be64(NAGLE_V(f->fs.newvlan == VLAN_REMOVE ||
+ f->fs.newvlan == VLAN_REWRITE) |
+ DELACK_V(f->fs.hitcnts) |
+ L2T_IDX_V(f->l2t ? f->l2t->idx : 0) |
+ SMAC_SEL_V((cxgb4_port_viid(f->dev) &
+ 0x7F) << 1) |
+ TX_CHAN_V(f->fs.eport) |
+ NO_CONG_V(f->fs.rpttid) |
+ ULP_MODE_V(f->fs.nat_mode ?
+ ULP_MODE_TCPDDP : ULP_MODE_NONE) |
+ TCAM_BYPASS_F | NON_OFFLOAD_F);
+
+ t6req->params = cpu_to_be64(FILTER_TUPLE_V(hash_filter_ntuple(&f->fs,
+ f->dev)));
+ t6req->opt2 = htonl(RSS_QUEUE_VALID_F |
+ RSS_QUEUE_V(f->fs.iq) |
+ TX_QUEUE_V(f->fs.nat_mode) |
+ T5_OPT_2_VALID_F |
+ RX_CHANNEL_F |
+ CONG_CNTRL_V((f->fs.action == FILTER_DROP) |
+ (f->fs.dirsteer << 1)) |
+ PACE_V((f->fs.maskhash) |
+ ((f->fs.dirsteerhash) << 1)) |
+ CCTRL_ECN_V(f->fs.action == FILTER_SWITCH));
+}
+
+static int cxgb4_set_hash_filter(struct net_device *dev,
+ struct ch_filter_specification *fs,
+ struct filter_ctx *ctx)
+{
+ struct adapter *adapter = netdev2adap(dev);
+ struct tid_info *t = &adapter->tids;
+ struct filter_entry *f;
+ struct sk_buff *skb;
+ int iq, atid, size;
+ int ret = 0;
+ u32 iconf;
+
+ fill_default_mask(fs);
+ ret = validate_filter(dev, fs);
+ if (ret)
+ return ret;
+
+ iq = get_filter_steerq(dev, fs);
+ if (iq < 0)
+ return iq;
+
+ f = kzalloc(sizeof(*f), GFP_KERNEL);
+ if (!f)
+ return -ENOMEM;
+
+ f->fs = *fs;
+ f->ctx = ctx;
+ f->dev = dev;
+ f->fs.iq = iq;
+
+ /* If the new filter requires loopback Destination MAC and/or VLAN
+ * rewriting then we need to allocate a Layer 2 Table (L2T) entry for
+ * the filter.
+ */
+ if (f->fs.newdmac || f->fs.newvlan) {
+ /* allocate L2T entry for new filter */
+ f->l2t = t4_l2t_alloc_switching(adapter, f->fs.vlan,
+ f->fs.eport, f->fs.dmac);
+ if (!f->l2t) {
+ ret = -ENOMEM;
+ goto out_err;
+ }
+ }
+
+ /* If the new filter requires loopback Source MAC rewriting then
+ * we need to allocate a SMT entry for the filter.
+ */
+ if (f->fs.newsmac) {
+ f->smt = cxgb4_smt_alloc_switching(f->dev, f->fs.smac);
+ if (!f->smt) {
+ if (f->l2t) {
+ cxgb4_l2t_release(f->l2t);
+ f->l2t = NULL;
+ }
+ ret = -ENOMEM;
+ goto free_l2t;
+ }
+ }
+
+ atid = cxgb4_alloc_atid(t, f);
+ if (atid < 0)
+ goto free_smt;
+
+ iconf = adapter->params.tp.ingress_config;
+ if (iconf & VNIC_F) {
+ f->fs.val.ovlan = (fs->val.pf << 13) | fs->val.vf;
+ f->fs.mask.ovlan = (fs->mask.pf << 13) | fs->mask.vf;
+ f->fs.val.ovlan_vld = fs->val.pfvf_vld;
+ f->fs.mask.ovlan_vld = fs->mask.pfvf_vld;
+ }
+
+ size = sizeof(struct cpl_t6_act_open_req);
+ if (f->fs.type) {
+ ret = cxgb4_clip_get(f->dev, (const u32 *)&f->fs.val.lip, 1);
+ if (ret)
+ goto free_atid;
+
+ skb = alloc_skb(size, GFP_KERNEL);
+ if (!skb) {
+ ret = -ENOMEM;
+ goto free_clip;
+ }
+
+ mk_act_open_req6(f, skb,
+ ((adapter->sge.fw_evtq.abs_id << 14) | atid),
+ adapter);
+ } else {
+ skb = alloc_skb(size, GFP_KERNEL);
+ if (!skb) {
+ ret = -ENOMEM;
+ goto free_atid;
+ }
+
+ mk_act_open_req(f, skb,
+ ((adapter->sge.fw_evtq.abs_id << 14) | atid),
+ adapter);
+ }
+
+ f->pending = 1;
+ set_wr_txq(skb, CPL_PRIORITY_SETUP, f->fs.val.iport & 0x3);
+ t4_ofld_send(adapter, skb);
+ return 0;
+
+free_clip:
+ cxgb4_clip_release(f->dev, (const u32 *)&f->fs.val.lip, 1);
+
+free_atid:
+ cxgb4_free_atid(t, atid);
+
+free_smt:
+ if (f->smt) {
+ cxgb4_smt_release(f->smt);
+ f->smt = NULL;
+ }
+
+free_l2t:
+ if (f->l2t) {
+ cxgb4_l2t_release(f->l2t);
+ f->l2t = NULL;
+ }
+
+out_err:
+ kfree(f);
+ return ret;
+}
+
/* Check a Chelsio Filter Request for validity, convert it into our internal
* format and send it to the hardware. Return 0 on success, an error number
* otherwise. We attach any provided filter operation context to the internal
@@ -580,6 +981,14 @@ int __cxgb4_set_filter(struct net_device *dev, int filter_id,
u32 iconf;
int iq, ret;
+ if (fs->hash) {
+ if (is_hashfilter(adapter))
+ return cxgb4_set_hash_filter(dev, fs, ctx);
+ netdev_err(dev, "%s: Exact-match filters only supported with Hash Filter configuration\n",
+ __func__);
+ return -EINVAL;
+ }
+
max_fidx = adapter->tids.nftids;
if (filter_id != (max_fidx + adapter->tids.nsftids - 1) &&
filter_id >= max_fidx)
@@ -789,60 +1198,126 @@ int cxgb4_del_filter(struct net_device *dev, int filter_id)
return ret;
}
-static int set_tcb_field(struct adapter *adap, struct filter_entry *f,
- unsigned int ftid, u16 word, u64 mask, u64 val,
- int no_reply)
+static int configure_filter_tcb(struct adapter *adap, unsigned int tid,
+ struct filter_entry *f)
{
- struct cpl_set_tcb_field *req;
- struct sk_buff *skb;
-
- skb = alloc_skb(sizeof(struct cpl_set_tcb_field), GFP_ATOMIC);
- if (!skb)
- return -ENOMEM;
-
- req = (struct cpl_set_tcb_field *)__skb_put(skb, sizeof(*req));
- memset(req, 0, sizeof(*req));
- INIT_TP_WR_CPL(req, CPL_SET_TCB_FIELD, ftid);
- req->reply_ctrl = htons(REPLY_CHAN_V(0) |
- QUEUENO_V(adap->sge.fw_evtq.abs_id) |
- NO_REPLY_V(no_reply));
- req->word_cookie = htons(TCB_WORD_V(word) | TCB_COOKIE_V(ftid));
- req->mask = cpu_to_be64(mask);
- req->val = cpu_to_be64(val);
- set_wr_txq(skb, CPL_PRIORITY_CONTROL, f->fs.val.iport & 0x3);
- t4_ofld_send(adap, skb);
+ if (f->fs.hitcnts)
+ set_tcb_field(adap, f, tid, TCB_TIMESTAMP_W,
+ TCB_TIMESTAMP_V(TCB_TIMESTAMP_M) |
+ TCB_RTT_TS_RECENT_AGE_V(TCB_RTT_TS_RECENT_AGE_M),
+ TCB_TIMESTAMP_V(0ULL) |
+ TCB_RTT_TS_RECENT_AGE_V(0ULL),
+ 1);
+
+ if (f->fs.newdmac)
+ set_tcb_tflag(adap, f, tid, TF_CCTRL_ECE_S, 1,
+ 1);
+
+ if (f->fs.newvlan == VLAN_INSERT ||
+ f->fs.newvlan == VLAN_REWRITE)
+ set_tcb_tflag(adap, f, tid, TF_CCTRL_RFR_S, 1,
+ 1);
+ if (f->fs.newsmac)
+ configure_filter_smac(adap, f);
+
+ if (f->fs.nat_mode) {
+ switch (f->fs.nat_mode) {
+ case NAT_MODE_DIP:
+ set_nat_params(adap, f, tid, true, false, false, false);
+ break;
+
+ case NAT_MODE_DIP_DP:
+ set_nat_params(adap, f, tid, true, false, true, false);
+ break;
+
+ case NAT_MODE_DIP_DP_SIP:
+ set_nat_params(adap, f, tid, true, true, true, false);
+ break;
+ case NAT_MODE_DIP_DP_SP:
+ set_nat_params(adap, f, tid, true, false, true, true);
+ break;
+
+ case NAT_MODE_SIP_SP:
+ set_nat_params(adap, f, tid, false, true, false, true);
+ break;
+
+ case NAT_MODE_DIP_SIP_SP:
+ set_nat_params(adap, f, tid, true, true, false, true);
+ break;
+
+ case NAT_MODE_ALL:
+ set_nat_params(adap, f, tid, true, true, true, true);
+ break;
+
+ default:
+ pr_err("%s: Invalid NAT mode: %d\n",
+ __func__, f->fs.nat_mode);
+ return -EINVAL;
+ }
+ }
return 0;
}
-/* Set one of the t_flags bits in the TCB.
- */
-static int set_tcb_tflag(struct adapter *adap, struct filter_entry *f,
- unsigned int ftid, unsigned int bit_pos,
- unsigned int val, int no_reply)
+void hash_filter_rpl(struct adapter *adap, const struct cpl_act_open_rpl *rpl)
{
- return set_tcb_field(adap, f, ftid, TCB_T_FLAGS_W, 1ULL << bit_pos,
- (unsigned long long)val << bit_pos, no_reply);
-}
+ unsigned int ftid = TID_TID_G(AOPEN_ATID_G(ntohl(rpl->atid_status)));
+ unsigned int status = AOPEN_STATUS_G(ntohl(rpl->atid_status));
+ struct tid_info *t = &adap->tids;
+ unsigned int tid = GET_TID(rpl);
+ struct filter_ctx *ctx = NULL;
+ struct filter_entry *f;
-static int configure_filter_smac(struct adapter *adap, struct filter_entry *f)
-{
- int err;
+ dev_dbg(adap->pdev_dev, "%s: tid = %u; atid = %u; status = %u\n",
+ __func__, tid, ftid, status);
- /* do a set-tcb for smac-sel and CWR bit.. */
- err = set_tcb_tflag(adap, f, f->tid, TF_CCTRL_CWR_S, 1, 1);
- if (err)
- goto smac_err;
+ f = lookup_atid(t, ftid);
+ if (!f) {
+ dev_err(adap->pdev_dev, "%s:could not find filter entry",
+ __func__);
+ return;
+ }
+ ctx = f->ctx;
+ f->ctx = NULL;
+
+ switch (status) {
+ case CPL_ERR_NONE:
+ f->tid = tid;
+ f->pending = 0;
+ f->valid = 1;
+ cxgb4_insert_tid(t, f, f->tid, 0);
+ cxgb4_free_atid(t, ftid);
+ if (ctx) {
+ ctx->tid = f->tid;
+ ctx->result = 0;
+ }
+ if (configure_filter_tcb(adap, tid, f)) {
+ clear_filter(adap, f);
+ cxgb4_remove_tid(t, 0, tid, 0);
+ kfree(f);
+ if (ctx) {
+ ctx->result = -EINVAL;
+ complete(&ctx->completion);
+ }
+ return;
+ }
+ break;
- err = set_tcb_field(adap, f, f->tid, TCB_SMAC_SEL_W,
- TCB_SMAC_SEL_V(TCB_SMAC_SEL_M),
- TCB_SMAC_SEL_V(f->smt->idx), 1);
- if (!err)
- return 0;
+ default:
+ dev_err(adap->pdev_dev, "%s: filter creation PROBLEM; status = %u\n",
+ __func__, status);
-smac_err:
- dev_err(adap->pdev_dev, "filter %u smac config failed with error %u\n",
- f->tid, err);
- return err;
+ if (ctx) {
+ if (status == CPL_ERR_TCAM_FULL)
+ ctx->result = -EAGAIN;
+ else
+ ctx->result = -EINVAL;
+ }
+ clear_filter(adap, f);
+ cxgb4_free_atid(t, ftid);
+ kfree(f);
+ }
+ if (ctx)
+ complete(&ctx->completion);
}
/* Handle a filter write/deletion reply. */
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.h
index d3c1a8fafd32..7480d65550a8 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.h
@@ -37,7 +37,10 @@
#include "t4_msg.h"
+#define WORD_MASK 0xffffffff
+
void filter_rpl(struct adapter *adap, const struct cpl_set_tcb_rpl *rpl);
+void hash_filter_rpl(struct adapter *adap, const struct cpl_act_open_rpl *rpl);
void clear_filter(struct adapter *adap, struct filter_entry *f);
int set_filter_wr(struct adapter *adapter, int fidx);
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
index 4b07cfe8c66c..77b4bd958748 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
@@ -572,6 +572,10 @@ static int fwevtq_handler(struct sge_rspq *q, const __be64 *rsp,
const struct cpl_set_tcb_rpl *p = (void *)rsp;
filter_rpl(q->adap, p);
+ } else if (opcode == CPL_ACT_OPEN_RPL) {
+ const struct cpl_act_open_rpl *p = (void *)rsp;
+
+ hash_filter_rpl(q->adap, p);
} else
dev_err(q->adap->pdev_dev,
"unexpected CPL %#x on FW event queue\n", opcode);
diff --git a/drivers/net/ethernet/chelsio/cxgb4/t4_msg.h b/drivers/net/ethernet/chelsio/cxgb4/t4_msg.h
index ce4838d907da..7e12f241145b 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/t4_msg.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/t4_msg.h
@@ -286,6 +286,7 @@ struct work_request_hdr {
#define RX_CHANNEL_S 26
#define RX_CHANNEL_V(x) ((x) << RX_CHANNEL_S)
+#define RX_CHANNEL_F RX_CHANNEL_V(1U)
#define WND_SCALE_EN_S 28
#define WND_SCALE_EN_V(x) ((x) << WND_SCALE_EN_S)
@@ -315,6 +316,10 @@ struct cpl_pass_open_req {
#define DELACK_V(x) ((x) << DELACK_S)
#define DELACK_F DELACK_V(1U)
+#define NON_OFFLOAD_S 7
+#define NON_OFFLOAD_V(x) ((x) << NON_OFFLOAD_S)
+#define NON_OFFLOAD_F NON_OFFLOAD_V(1U)
+
#define DSCP_S 22
#define DSCP_M 0x3F
#define DSCP_V(x) ((x) << DSCP_S)
diff --git a/drivers/net/ethernet/chelsio/cxgb4/t4_tcb.h b/drivers/net/ethernet/chelsio/cxgb4/t4_tcb.h
index c1c76663034d..c7201eb7b14c 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/t4_tcb.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/t4_tcb.h
@@ -42,6 +42,23 @@
#define TCB_T_FLAGS_W 1
+#define TF_CCTRL_ECE_S 60
#define TF_CCTRL_CWR_S 61
+#define TF_CCTRL_RFR_S 62
+#define TCB_TIMESTAMP_W 5
+#define TCB_TIMESTAMP_S 0
+#define TCB_TIMESTAMP_M 0xffffffffULL
+#define TCB_TIMESTAMP_V(x) ((x) << TCB_TIMESTAMP_S)
+
+#define TCB_RTT_TS_RECENT_AGE_W 6
+#define TCB_RTT_TS_RECENT_AGE_S 0
+#define TCB_RTT_TS_RECENT_AGE_M 0xffffffffULL
+#define TCB_RTT_TS_RECENT_AGE_V(x) ((x) << TCB_RTT_TS_RECENT_AGE_S)
+
+#define TCB_SND_UNA_RAW_W 10
+#define TCB_RX_FRAG2_PTR_RAW_W 27
+#define TCB_RX_FRAG3_LEN_RAW_W 29
+#define TCB_RX_FRAG3_START_IDX_OFFSET_RAW_W 30
+#define TCB_PDU_HDR_LEN_W 31
#endif /* __T4_TCB_H */
--
2.14.1
^ permalink raw reply related [flat|nested] 9+ messages in thread* [PATCH net-next 4/7] cxgb4: add support to delete hash filter
2017-11-01 3:22 [PATCH net-next 0/7] cxgb4: add hash-filter support to tc-flower offload Rahul Lakkireddy
` (2 preceding siblings ...)
2017-11-01 3:23 ` [PATCH net-next 3/7] cxgb4: add support to create hash filters Rahul Lakkireddy
@ 2017-11-01 3:23 ` Rahul Lakkireddy
2017-11-01 3:23 ` [PATCH net-next 5/7] cxgb4: add support to retrieve stats for hash filters Rahul Lakkireddy
` (3 subsequent siblings)
7 siblings, 0 replies; 9+ messages in thread
From: Rahul Lakkireddy @ 2017-11-01 3:23 UTC (permalink / raw)
To: netdev; +Cc: davem, kumaras, ganeshgr, nirranjan, indranil, Rahul Lakkireddy
From: Kumar Sanghvi <kumaras@chelsio.com>
Use a combined ulptx work-request to send hash filter deletion
request to hw. Hash filter deletion reply is processed on
getting cpl_abort_rpl_rss.
Release any L2T/SMT/CLIP entries on filter deletion.
Also, free up the corresponding filter entry.
Signed-off-by: Kumar Sanghvi <kumaras@chelsio.com>
Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
---
drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c | 159 ++++++++++++++++++++-
drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.h | 2 +
drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c | 4 +
.../net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c | 2 +-
drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_u32.c | 4 +-
drivers/net/ethernet/chelsio/cxgb4/cxgb4_uld.h | 4 +-
drivers/net/ethernet/chelsio/cxgb4/t4_tcb.h | 5 +
7 files changed, 173 insertions(+), 7 deletions(-)
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c
index cf8ca695f27d..eb6ba9824501 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c
@@ -87,6 +87,59 @@ static int set_tcb_tflag(struct adapter *adap, struct filter_entry *f,
(unsigned long long)val << bit_pos, no_reply);
}
+static void mk_abort_req_ulp(struct cpl_abort_req *abort_req, unsigned int tid)
+{
+ struct ulp_txpkt *txpkt = (struct ulp_txpkt *)abort_req;
+ struct ulptx_idata *sc = (struct ulptx_idata *)(txpkt + 1);
+
+ txpkt->cmd_dest = htonl(ULPTX_CMD_V(ULP_TX_PKT) | ULP_TXPKT_DEST_V(0));
+ txpkt->len = htonl(DIV_ROUND_UP(sizeof(*abort_req), 16));
+ sc->cmd_more = htonl(ULPTX_CMD_V(ULP_TX_SC_IMM));
+ sc->len = htonl(sizeof(*abort_req) - sizeof(struct work_request_hdr));
+ OPCODE_TID(abort_req) = htonl(MK_OPCODE_TID(CPL_ABORT_REQ, tid));
+ abort_req->rsvd0 = htonl(0);
+ abort_req->rsvd1 = 0;
+ abort_req->cmd = CPL_ABORT_NO_RST;
+}
+
+static void mk_abort_rpl_ulp(struct cpl_abort_rpl *abort_rpl, unsigned int tid)
+{
+ struct ulp_txpkt *txpkt = (struct ulp_txpkt *)abort_rpl;
+ struct ulptx_idata *sc = (struct ulptx_idata *)(txpkt + 1);
+
+ txpkt->cmd_dest = htonl(ULPTX_CMD_V(ULP_TX_PKT) | ULP_TXPKT_DEST_V(0));
+ txpkt->len = htonl(DIV_ROUND_UP(sizeof(*abort_rpl), 16));
+ sc->cmd_more = htonl(ULPTX_CMD_V(ULP_TX_SC_IMM));
+ sc->len = htonl(sizeof(*abort_rpl) - sizeof(struct work_request_hdr));
+ OPCODE_TID(abort_rpl) = htonl(MK_OPCODE_TID(CPL_ABORT_RPL, tid));
+ abort_rpl->rsvd0 = htonl(0);
+ abort_rpl->rsvd1 = 0;
+ abort_rpl->cmd = CPL_ABORT_NO_RST;
+}
+
+static void mk_set_tcb_ulp(struct filter_entry *f,
+ struct cpl_set_tcb_field *req,
+ unsigned int word, u64 mask, u64 val,
+ u8 cookie, int no_reply)
+{
+ struct ulp_txpkt *txpkt = (struct ulp_txpkt *)req;
+ struct ulptx_idata *sc = (struct ulptx_idata *)(txpkt + 1);
+
+ txpkt->cmd_dest = htonl(ULPTX_CMD_V(ULP_TX_PKT) | ULP_TXPKT_DEST_V(0));
+ txpkt->len = htonl(DIV_ROUND_UP(sizeof(*req), 16));
+ sc->cmd_more = htonl(ULPTX_CMD_V(ULP_TX_SC_IMM));
+ sc->len = htonl(sizeof(*req) - sizeof(struct work_request_hdr));
+ OPCODE_TID(req) = htonl(MK_OPCODE_TID(CPL_SET_TCB_FIELD, f->tid));
+ req->reply_ctrl = htons(NO_REPLY_V(no_reply) | REPLY_CHAN_V(0) |
+ QUEUENO_V(0));
+ req->word_cookie = htons(TCB_WORD_V(word) | TCB_COOKIE_V(cookie));
+ req->mask = cpu_to_be64(mask);
+ req->val = cpu_to_be64(val);
+ sc = (struct ulptx_idata *)(req + 1);
+ sc->cmd_more = htonl(ULPTX_CMD_V(ULP_TX_SC_NOOP));
+ sc->len = htonl(0);
+}
+
static int configure_filter_smac(struct adapter *adap, struct filter_entry *f)
{
int err;
@@ -1110,12 +1163,74 @@ int __cxgb4_set_filter(struct net_device *dev, int filter_id,
return ret;
}
+static int cxgb4_del_hash_filter(struct net_device *dev, int filter_id,
+ struct filter_ctx *ctx)
+{
+ struct adapter *adapter = netdev2adap(dev);
+ struct tid_info *t = &adapter->tids;
+ struct cpl_abort_req *abort_req;
+ struct cpl_abort_rpl *abort_rpl;
+ struct cpl_set_tcb_field *req;
+ struct ulptx_idata *aligner;
+ struct work_request_hdr *wr;
+ struct filter_entry *f;
+ struct sk_buff *skb;
+ unsigned int wrlen;
+ int ret;
+
+ netdev_dbg(dev, "%s: filter_id = %d ; nftids = %d\n",
+ __func__, filter_id, adapter->tids.nftids);
+
+ if (filter_id > adapter->tids.ntids)
+ return -E2BIG;
+
+ f = lookup_tid(t, filter_id);
+ if (!f) {
+ netdev_err(dev, "%s: no filter entry for filter_id = %d",
+ __func__, filter_id);
+ return -EINVAL;
+ }
+
+ ret = writable_filter(f);
+ if (ret)
+ return ret;
+
+ if (!f->valid)
+ return -EINVAL;
+
+ f->ctx = ctx;
+ f->pending = 1;
+ wrlen = roundup(sizeof(*wr) + (sizeof(*req) + sizeof(*aligner))
+ + sizeof(*abort_req) + sizeof(*abort_rpl), 16);
+ skb = alloc_skb(wrlen, GFP_KERNEL);
+ if (!skb) {
+ netdev_err(dev, "%s: could not allocate skb ..\n", __func__);
+ return -ENOMEM;
+ }
+ set_wr_txq(skb, CPL_PRIORITY_CONTROL, f->fs.val.iport & 0x3);
+ req = (struct cpl_set_tcb_field *)__skb_put(skb, wrlen);
+ INIT_ULPTX_WR(req, wrlen, 0, 0);
+ wr = (struct work_request_hdr *)req;
+ wr++;
+ req = (struct cpl_set_tcb_field *)wr;
+ mk_set_tcb_ulp(f, req, TCB_RSS_INFO_W, TCB_RSS_INFO_V(TCB_RSS_INFO_M),
+ TCB_RSS_INFO_V(adapter->sge.fw_evtq.abs_id), 0, 1);
+ aligner = (struct ulptx_idata *)(req + 1);
+ abort_req = (struct cpl_abort_req *)(aligner + 1);
+ mk_abort_req_ulp(abort_req, f->tid);
+ abort_rpl = (struct cpl_abort_rpl *)(abort_req + 1);
+ mk_abort_rpl_ulp(abort_rpl, f->tid);
+ t4_ofld_send(adapter, skb);
+ return 0;
+}
+
/* Check a delete filter request for validity and send it to the hardware.
* Return 0 on success, an error number otherwise. We attach any provided
* filter operation context to the internal filter specification in order to
* facilitate signaling completion of the operation.
*/
int __cxgb4_del_filter(struct net_device *dev, int filter_id,
+ struct ch_filter_specification *fs,
struct filter_ctx *ctx)
{
struct adapter *adapter = netdev2adap(dev);
@@ -1123,6 +1238,14 @@ int __cxgb4_del_filter(struct net_device *dev, int filter_id,
unsigned int max_fidx;
int ret;
+ if (fs && fs->hash) {
+ if (is_hashfilter(adapter))
+ return cxgb4_del_hash_filter(dev, filter_id, ctx);
+ netdev_err(dev, "%s: Exact-match filters only supported with Hash Filter configuration\n",
+ __func__);
+ return -EINVAL;
+ }
+
max_fidx = adapter->tids.nftids;
if (filter_id != (max_fidx + adapter->tids.nsftids - 1) &&
filter_id >= max_fidx)
@@ -1173,18 +1296,19 @@ int cxgb4_set_filter(struct net_device *dev, int filter_id,
return ret;
}
-int cxgb4_del_filter(struct net_device *dev, int filter_id)
+int cxgb4_del_filter(struct net_device *dev, int filter_id,
+ struct ch_filter_specification *fs)
{
struct filter_ctx ctx;
int ret;
/* If we are shutting down the adapter do not wait for completion */
if (netdev2adap(dev)->flags & SHUTTING_DOWN)
- return __cxgb4_del_filter(dev, filter_id, NULL);
+ return __cxgb4_del_filter(dev, filter_id, fs, NULL);
init_completion(&ctx.completion);
- ret = __cxgb4_del_filter(dev, filter_id, &ctx);
+ ret = __cxgb4_del_filter(dev, filter_id, fs, &ctx);
if (ret)
goto out;
@@ -1258,6 +1382,35 @@ static int configure_filter_tcb(struct adapter *adap, unsigned int tid,
return 0;
}
+void hash_del_filter_rpl(struct adapter *adap,
+ const struct cpl_abort_rpl_rss *rpl)
+{
+ unsigned int status = rpl->status;
+ struct tid_info *t = &adap->tids;
+ unsigned int tid = GET_TID(rpl);
+ struct filter_ctx *ctx = NULL;
+ struct filter_entry *f;
+
+ dev_dbg(adap->pdev_dev, "%s: status = %u; tid = %u\n",
+ __func__, status, tid);
+
+ f = lookup_tid(t, tid);
+ if (!f) {
+ dev_err(adap->pdev_dev, "%s:could not find filter entry",
+ __func__);
+ return;
+ }
+ ctx = f->ctx;
+ f->ctx = NULL;
+ clear_filter(adap, f);
+ cxgb4_remove_tid(t, 0, tid, 0);
+ kfree(f);
+ if (ctx) {
+ ctx->result = 0;
+ complete(&ctx->completion);
+ }
+}
+
void hash_filter_rpl(struct adapter *adap, const struct cpl_act_open_rpl *rpl)
{
unsigned int ftid = TID_TID_G(AOPEN_ATID_G(ntohl(rpl->atid_status)));
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.h
index 7480d65550a8..9475abd3384e 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.h
@@ -41,6 +41,8 @@
void filter_rpl(struct adapter *adap, const struct cpl_set_tcb_rpl *rpl);
void hash_filter_rpl(struct adapter *adap, const struct cpl_act_open_rpl *rpl);
+void hash_del_filter_rpl(struct adapter *adap,
+ const struct cpl_abort_rpl_rss *rpl);
void clear_filter(struct adapter *adap, struct filter_entry *f);
int set_filter_wr(struct adapter *adapter, int fidx);
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
index 77b4bd958748..35709c7f7c5b 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
@@ -576,6 +576,10 @@ static int fwevtq_handler(struct sge_rspq *q, const __be64 *rsp,
const struct cpl_act_open_rpl *p = (void *)rsp;
hash_filter_rpl(q->adap, p);
+ } else if (opcode == CPL_ABORT_RPL_RSS) {
+ const struct cpl_abort_rpl_rss *p = (void *)rsp;
+
+ hash_del_filter_rpl(q->adap, p);
} else
dev_err(q->adap->pdev_dev,
"unexpected CPL %#x on FW event queue\n", opcode);
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c
index 614db014ef18..ed377e2e9f8a 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c
@@ -730,7 +730,7 @@ int cxgb4_tc_flower_destroy(struct net_device *dev,
if (!ch_flower)
return -ENOENT;
- ret = cxgb4_del_filter(dev, ch_flower->filter_id);
+ ret = cxgb4_del_filter(dev, ch_flower->filter_id, &ch_flower->fs);
if (ret)
goto err;
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_u32.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_u32.c
index 48970ba08bdc..cd0cd13a964d 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_u32.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_u32.c
@@ -380,7 +380,7 @@ int cxgb4_delete_knode(struct net_device *dev, struct tc_cls_u32_offload *cls)
return -EINVAL;
}
- ret = cxgb4_del_filter(dev, filter_id);
+ ret = cxgb4_del_filter(dev, filter_id, NULL);
if (ret)
goto out;
@@ -399,7 +399,7 @@ int cxgb4_delete_knode(struct net_device *dev, struct tc_cls_u32_offload *cls)
if (!test_bit(j, link->tid_map))
continue;
- ret = __cxgb4_del_filter(dev, j, NULL);
+ ret = __cxgb4_del_filter(dev, j, NULL, NULL);
if (ret)
goto out;
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_uld.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_uld.h
index 52324c77a4fe..a1c850861cbf 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_uld.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_uld.h
@@ -217,10 +217,12 @@ int __cxgb4_set_filter(struct net_device *dev, int filter_id,
struct ch_filter_specification *fs,
struct filter_ctx *ctx);
int __cxgb4_del_filter(struct net_device *dev, int filter_id,
+ struct ch_filter_specification *fs,
struct filter_ctx *ctx);
int cxgb4_set_filter(struct net_device *dev, int filter_id,
struct ch_filter_specification *fs);
-int cxgb4_del_filter(struct net_device *dev, int filter_id);
+int cxgb4_del_filter(struct net_device *dev, int filter_id,
+ struct ch_filter_specification *fs);
int cxgb4_get_filter_counters(struct net_device *dev, unsigned int fidx,
u64 *hitcnt, u64 *bytecnt);
diff --git a/drivers/net/ethernet/chelsio/cxgb4/t4_tcb.h b/drivers/net/ethernet/chelsio/cxgb4/t4_tcb.h
index c7201eb7b14c..3297ce025e8b 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/t4_tcb.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/t4_tcb.h
@@ -46,6 +46,11 @@
#define TF_CCTRL_CWR_S 61
#define TF_CCTRL_RFR_S 62
+#define TCB_RSS_INFO_W 3
+#define TCB_RSS_INFO_S 0
+#define TCB_RSS_INFO_M 0x3ffULL
+#define TCB_RSS_INFO_V(x) ((x) << TCB_RSS_INFO_S)
+
#define TCB_TIMESTAMP_W 5
#define TCB_TIMESTAMP_S 0
#define TCB_TIMESTAMP_M 0xffffffffULL
--
2.14.1
^ permalink raw reply related [flat|nested] 9+ messages in thread* [PATCH net-next 5/7] cxgb4: add support to retrieve stats for hash filters
2017-11-01 3:22 [PATCH net-next 0/7] cxgb4: add hash-filter support to tc-flower offload Rahul Lakkireddy
` (3 preceding siblings ...)
2017-11-01 3:23 ` [PATCH net-next 4/7] cxgb4: add support to delete hash filter Rahul Lakkireddy
@ 2017-11-01 3:23 ` Rahul Lakkireddy
2017-11-01 3:23 ` [PATCH net-next 6/7] cxgb4: convert flower table to use rhashtable Rahul Lakkireddy
` (2 subsequent siblings)
7 siblings, 0 replies; 9+ messages in thread
From: Rahul Lakkireddy @ 2017-11-01 3:23 UTC (permalink / raw)
To: netdev; +Cc: davem, kumaras, ganeshgr, nirranjan, indranil, Rahul Lakkireddy
From: Kumar Sanghvi <kumaras@chelsio.com>
Add support to retrieve packet-count and byte-count for hash-filters
by retrieving filter-entry appropriately based on whether the
request is for hash-filter or not.
Signed-off-by: Kumar Sanghvi <kumaras@chelsio.com>
Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
---
drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c | 30 ++++++++++++++--------
.../net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c | 6 +++--
drivers/net/ethernet/chelsio/cxgb4/cxgb4_uld.h | 2 +-
3 files changed, 25 insertions(+), 13 deletions(-)
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c
index eb6ba9824501..9b3ff6209eb5 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c
@@ -341,7 +341,7 @@ static int get_filter_steerq(struct net_device *dev,
}
static int get_filter_count(struct adapter *adapter, unsigned int fidx,
- u64 *pkts, u64 *bytes)
+ u64 *pkts, u64 *bytes, bool hash)
{
unsigned int tcb_base, tcbaddr;
unsigned int word_offset;
@@ -350,14 +350,24 @@ static int get_filter_count(struct adapter *adapter, unsigned int fidx,
int ret;
tcb_base = t4_read_reg(adapter, TP_CMM_TCB_BASE_A);
- if ((fidx != (adapter->tids.nftids + adapter->tids.nsftids - 1)) &&
- fidx >= adapter->tids.nftids)
- return -E2BIG;
-
- f = &adapter->tids.ftid_tab[fidx];
- if (!f->valid)
- return -EINVAL;
+ if (is_hashfilter(adapter) && hash) {
+ if (fidx < adapter->tids.ntids) {
+ f = adapter->tids.tid_tab[fidx];
+ if (!f)
+ return -EINVAL;
+ } else {
+ return -E2BIG;
+ }
+ } else {
+ if ((fidx != (adapter->tids.nftids +
+ adapter->tids.nsftids - 1)) &&
+ fidx >= adapter->tids.nftids)
+ return -E2BIG;
+ f = &adapter->tids.ftid_tab[fidx];
+ if (!f->valid)
+ return -EINVAL;
+ }
tcbaddr = tcb_base + f->tid * TCB_SIZE;
spin_lock(&adapter->win0_lock);
@@ -409,11 +419,11 @@ static int get_filter_count(struct adapter *adapter, unsigned int fidx,
}
int cxgb4_get_filter_counters(struct net_device *dev, unsigned int fidx,
- u64 *hitcnt, u64 *bytecnt)
+ u64 *hitcnt, u64 *bytecnt, bool hash)
{
struct adapter *adapter = netdev2adap(dev);
- return get_filter_count(adapter, fidx, hitcnt, bytecnt);
+ return get_filter_count(adapter, fidx, hitcnt, bytecnt, hash);
}
int cxgb4_get_free_ftid(struct net_device *dev, int family)
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c
index ed377e2e9f8a..a26acd183eef 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c
@@ -755,7 +755,8 @@ static void ch_flower_stats_cb(struct timer_list *t)
hash_for_each_rcu(adap->flower_anymatch_tbl, i, flower_entry, link) {
ret = cxgb4_get_filter_counters(adap->port[0],
flower_entry->filter_id,
- &packets, &bytes);
+ &packets, &bytes,
+ flower_entry->fs.hash);
if (!ret) {
spin_lock(&flower_entry->lock);
ofld_stats = &flower_entry->stats;
@@ -788,7 +789,8 @@ int cxgb4_tc_flower_stats(struct net_device *dev,
}
ret = cxgb4_get_filter_counters(dev, ch_flower->filter_id,
- &packets, &bytes);
+ &packets, &bytes,
+ ch_flower->fs.hash);
if (ret < 0)
goto err;
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_uld.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_uld.h
index a1c850861cbf..08e709ab6dd4 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_uld.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_uld.h
@@ -224,7 +224,7 @@ int cxgb4_set_filter(struct net_device *dev, int filter_id,
int cxgb4_del_filter(struct net_device *dev, int filter_id,
struct ch_filter_specification *fs);
int cxgb4_get_filter_counters(struct net_device *dev, unsigned int fidx,
- u64 *hitcnt, u64 *bytecnt);
+ u64 *hitcnt, u64 *bytecnt, bool hash);
static inline void set_wr_txq(struct sk_buff *skb, int prio, int queue)
{
--
2.14.1
^ permalink raw reply related [flat|nested] 9+ messages in thread* [PATCH net-next 6/7] cxgb4: convert flower table to use rhashtable
2017-11-01 3:22 [PATCH net-next 0/7] cxgb4: add hash-filter support to tc-flower offload Rahul Lakkireddy
` (4 preceding siblings ...)
2017-11-01 3:23 ` [PATCH net-next 5/7] cxgb4: add support to retrieve stats for hash filters Rahul Lakkireddy
@ 2017-11-01 3:23 ` Rahul Lakkireddy
2017-11-01 3:23 ` [PATCH net-next 7/7] cxgb4: add support to create hash-filters via tc-flower offload Rahul Lakkireddy
2017-11-01 13:06 ` [PATCH net-next 0/7] cxgb4: add hash-filter support to " David Miller
7 siblings, 0 replies; 9+ messages in thread
From: Rahul Lakkireddy @ 2017-11-01 3:23 UTC (permalink / raw)
To: netdev; +Cc: davem, kumaras, ganeshgr, nirranjan, indranil, Rahul Lakkireddy
From: Kumar Sanghvi <kumaras@chelsio.com>
T6 supports ~500K hash filters and can theoretically climb up to
~1 million hash filters. Preallocated hash table is not efficient
in terms of memory usage. So, use rhashtable instead which gives
the flexibility to grow based on usage.
Signed-off-by: Kumar Sanghvi <kumaras@chelsio.com>
Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
---
drivers/net/ethernet/chelsio/cxgb4/cxgb4.h | 4 +-
drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c | 4 +-
.../net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c | 106 ++++++++++++++-------
.../net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.h | 4 +-
4 files changed, 82 insertions(+), 36 deletions(-)
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
index bb7f0e4c9a81..0c83ceb5a1a6 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
@@ -918,8 +918,10 @@ struct adapter {
struct chcr_stats_debug chcr_stats;
/* TC flower offload */
- DECLARE_HASHTABLE(flower_anymatch_tbl, 9);
+ struct rhashtable flower_tbl;
+ struct rhashtable_params flower_ht_params;
struct timer_list flower_stats_timer;
+ struct work_struct flower_stats_work;
/* Ethtool Dump */
struct ethtool_dump eth_dump;
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
index 35709c7f7c5b..8fd41917c07a 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
@@ -5182,7 +5182,9 @@ static int init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
dev_warn(&pdev->dev,
"could not offload tc u32, continuing\n");
- cxgb4_init_tc_flower(adapter);
+ if (cxgb4_init_tc_flower(adapter))
+ dev_warn(&pdev->dev,
+ "could not offload tc flower, continuing\n");
}
if (is_offload(adapter) || is_hashfilter(adapter)) {
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c
index a26acd183eef..3953bc1fdc20 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c
@@ -74,13 +74,8 @@ static struct ch_tc_flower_entry *allocate_flower_entry(void)
static struct ch_tc_flower_entry *ch_flower_lookup(struct adapter *adap,
unsigned long flower_cookie)
{
- struct ch_tc_flower_entry *flower_entry;
-
- hash_for_each_possible_rcu(adap->flower_anymatch_tbl, flower_entry,
- link, flower_cookie)
- if (flower_entry->tc_flower_cookie == flower_cookie)
- return flower_entry;
- return NULL;
+ return rhashtable_lookup_fast(&adap->flower_tbl, &flower_cookie,
+ adap->flower_ht_params);
}
static void cxgb4_process_flow_match(struct net_device *dev,
@@ -707,12 +702,17 @@ int cxgb4_tc_flower_replace(struct net_device *dev,
goto free_entry;
}
- INIT_HLIST_NODE(&ch_flower->link);
ch_flower->tc_flower_cookie = cls->cookie;
ch_flower->filter_id = ctx.tid;
- hash_add_rcu(adap->flower_anymatch_tbl, &ch_flower->link, cls->cookie);
+ ret = rhashtable_insert_fast(&adap->flower_tbl, &ch_flower->node,
+ adap->flower_ht_params);
+ if (ret)
+ goto del_filter;
- return ret;
+ return 0;
+
+del_filter:
+ cxgb4_del_filter(dev, ch_flower->filter_id, &ch_flower->fs);
free_entry:
kfree(ch_flower);
@@ -734,44 +734,66 @@ int cxgb4_tc_flower_destroy(struct net_device *dev,
if (ret)
goto err;
- hash_del_rcu(&ch_flower->link);
+ ret = rhashtable_remove_fast(&adap->flower_tbl, &ch_flower->node,
+ adap->flower_ht_params);
+ if (ret) {
+ netdev_err(dev, "Flow remove from rhashtable failed");
+ goto err;
+ }
kfree_rcu(ch_flower, rcu);
err:
return ret;
}
-static void ch_flower_stats_cb(struct timer_list *t)
+static void ch_flower_stats_handler(struct work_struct *work)
{
- struct adapter *adap = from_timer(adap, t, flower_stats_timer);
+ struct adapter *adap = container_of(work, struct adapter,
+ flower_stats_work);
struct ch_tc_flower_entry *flower_entry;
struct ch_tc_flower_stats *ofld_stats;
- unsigned int i;
+ struct rhashtable_iter iter;
u64 packets;
u64 bytes;
int ret;
- rcu_read_lock();
- hash_for_each_rcu(adap->flower_anymatch_tbl, i, flower_entry, link) {
- ret = cxgb4_get_filter_counters(adap->port[0],
- flower_entry->filter_id,
- &packets, &bytes,
- flower_entry->fs.hash);
- if (!ret) {
- spin_lock(&flower_entry->lock);
- ofld_stats = &flower_entry->stats;
-
- if (ofld_stats->prev_packet_count != packets) {
- ofld_stats->prev_packet_count = packets;
- ofld_stats->last_used = jiffies;
+ rhashtable_walk_enter(&adap->flower_tbl, &iter);
+ do {
+ flower_entry = ERR_PTR(rhashtable_walk_start(&iter));
+ if (IS_ERR(flower_entry))
+ goto walk_stop;
+
+ while ((flower_entry = rhashtable_walk_next(&iter)) &&
+ !IS_ERR(flower_entry)) {
+ ret = cxgb4_get_filter_counters(adap->port[0],
+ flower_entry->filter_id,
+ &packets, &bytes,
+ flower_entry->fs.hash);
+ if (!ret) {
+ spin_lock(&flower_entry->lock);
+ ofld_stats = &flower_entry->stats;
+
+ if (ofld_stats->prev_packet_count != packets) {
+ ofld_stats->prev_packet_count = packets;
+ ofld_stats->last_used = jiffies;
+ }
+ spin_unlock(&flower_entry->lock);
}
- spin_unlock(&flower_entry->lock);
}
- }
- rcu_read_unlock();
+walk_stop:
+ rhashtable_walk_stop(&iter);
+ } while (flower_entry == ERR_PTR(-EAGAIN));
+ rhashtable_walk_exit(&iter);
mod_timer(&adap->flower_stats_timer, jiffies + STATS_CHECK_PERIOD);
}
+static void ch_flower_stats_cb(struct timer_list *t)
+{
+ struct adapter *adap = from_timer(adap, t, flower_stats_timer);
+
+ schedule_work(&adap->flower_stats_work);
+}
+
int cxgb4_tc_flower_stats(struct net_device *dev,
struct tc_cls_flower_offload *cls)
{
@@ -814,15 +836,35 @@ int cxgb4_tc_flower_stats(struct net_device *dev,
return ret;
}
-void cxgb4_init_tc_flower(struct adapter *adap)
+static const struct rhashtable_params cxgb4_tc_flower_ht_params = {
+ .nelem_hint = 384,
+ .head_offset = offsetof(struct ch_tc_flower_entry, node),
+ .key_offset = offsetof(struct ch_tc_flower_entry, tc_flower_cookie),
+ .key_len = sizeof(((struct ch_tc_flower_entry *)0)->tc_flower_cookie),
+ .max_size = 524288,
+ .min_size = 512,
+ .automatic_shrinking = true
+};
+
+int cxgb4_init_tc_flower(struct adapter *adap)
{
- hash_init(adap->flower_anymatch_tbl);
+ int ret;
+
+ adap->flower_ht_params = cxgb4_tc_flower_ht_params;
+ ret = rhashtable_init(&adap->flower_tbl, &adap->flower_ht_params);
+ if (ret)
+ return ret;
+
+ INIT_WORK(&adap->flower_stats_work, ch_flower_stats_handler);
timer_setup(&adap->flower_stats_timer, ch_flower_stats_cb, 0);
mod_timer(&adap->flower_stats_timer, jiffies + STATS_CHECK_PERIOD);
+ return 0;
}
void cxgb4_cleanup_tc_flower(struct adapter *adap)
{
if (adap->flower_stats_timer.function)
del_timer_sync(&adap->flower_stats_timer);
+ cancel_work_sync(&adap->flower_stats_work);
+ rhashtable_destroy(&adap->flower_tbl);
}
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.h
index 202d5c9ec303..050c8a50ae41 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.h
@@ -48,7 +48,7 @@ struct ch_tc_flower_entry {
struct ch_filter_specification fs;
struct ch_tc_flower_stats stats;
unsigned long tc_flower_cookie;
- struct hlist_node link;
+ struct rhash_head node;
struct rcu_head rcu;
spinlock_t lock; /* lock for stats */
u32 filter_id;
@@ -115,6 +115,6 @@ int cxgb4_tc_flower_destroy(struct net_device *dev,
int cxgb4_tc_flower_stats(struct net_device *dev,
struct tc_cls_flower_offload *cls);
-void cxgb4_init_tc_flower(struct adapter *adap);
+int cxgb4_init_tc_flower(struct adapter *adap);
void cxgb4_cleanup_tc_flower(struct adapter *adap);
#endif /* __CXGB4_TC_FLOWER_H */
--
2.14.1
^ permalink raw reply related [flat|nested] 9+ messages in thread* [PATCH net-next 7/7] cxgb4: add support to create hash-filters via tc-flower offload
2017-11-01 3:22 [PATCH net-next 0/7] cxgb4: add hash-filter support to tc-flower offload Rahul Lakkireddy
` (5 preceding siblings ...)
2017-11-01 3:23 ` [PATCH net-next 6/7] cxgb4: convert flower table to use rhashtable Rahul Lakkireddy
@ 2017-11-01 3:23 ` Rahul Lakkireddy
2017-11-01 13:06 ` [PATCH net-next 0/7] cxgb4: add hash-filter support to " David Miller
7 siblings, 0 replies; 9+ messages in thread
From: Rahul Lakkireddy @ 2017-11-01 3:23 UTC (permalink / raw)
To: netdev; +Cc: davem, kumaras, ganeshgr, nirranjan, indranil, Rahul Lakkireddy
From: Kumar Sanghvi <kumaras@chelsio.com>
Determine whether the flow classifies as exact-match with respect to
4-tuple and configured tuple mask in hw. If successfully classified
as exact-match, offload the flow as hash-filter in hw.
Signed-off-by: Kumar Sanghvi <kumaras@chelsio.com>
Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
---
drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c | 148 +++++++++++++++++++++
drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.h | 2 +
.../net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c | 16 ++-
3 files changed, 161 insertions(+), 5 deletions(-)
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c
index 9b3ff6209eb5..abab67d52edb 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c
@@ -31,6 +31,7 @@
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
+#include <net/ipv6.h>
#include "cxgb4.h"
#include "t4_regs.h"
@@ -765,6 +766,153 @@ static void fill_default_mask(struct ch_filter_specification *fs)
fs->mask.fport = ~0;
}
+static bool is_addr_all_mask(u8 *ipmask, int family)
+{
+ if (family == AF_INET) {
+ struct in_addr *addr;
+
+ addr = (struct in_addr *)ipmask;
+ if (addr->s_addr == 0xffffffff)
+ return true;
+ } else if (family == AF_INET6) {
+ struct in6_addr *addr6;
+
+ addr6 = (struct in6_addr *)ipmask;
+ if (addr6->s6_addr32[0] == 0xffffffff &&
+ addr6->s6_addr32[1] == 0xffffffff &&
+ addr6->s6_addr32[2] == 0xffffffff &&
+ addr6->s6_addr32[3] == 0xffffffff)
+ return true;
+ }
+ return false;
+}
+
+static bool is_inaddr_any(u8 *ip, int family)
+{
+ int addr_type;
+
+ if (family == AF_INET) {
+ struct in_addr *addr;
+
+ addr = (struct in_addr *)ip;
+ if (addr->s_addr == htonl(INADDR_ANY))
+ return true;
+ } else if (family == AF_INET6) {
+ struct in6_addr *addr6;
+
+ addr6 = (struct in6_addr *)ip;
+ addr_type = ipv6_addr_type((const struct in6_addr *)
+ &addr6);
+ if (addr_type == IPV6_ADDR_ANY)
+ return true;
+ }
+ return false;
+}
+
+bool is_filter_exact_match(struct adapter *adap,
+ struct ch_filter_specification *fs)
+{
+ struct tp_params *tp = &adap->params.tp;
+ u64 hash_filter_mask = tp->hash_filter_mask;
+ u32 mask;
+
+ if (!is_hashfilter(adap))
+ return false;
+
+ if (fs->type) {
+ if (is_inaddr_any(fs->val.fip, AF_INET6) ||
+ !is_addr_all_mask(fs->mask.fip, AF_INET6))
+ return false;
+
+ if (is_inaddr_any(fs->val.lip, AF_INET6) ||
+ !is_addr_all_mask(fs->mask.lip, AF_INET6))
+ return false;
+ } else {
+ if (is_inaddr_any(fs->val.fip, AF_INET) ||
+ !is_addr_all_mask(fs->mask.fip, AF_INET))
+ return false;
+
+ if (is_inaddr_any(fs->val.lip, AF_INET) ||
+ !is_addr_all_mask(fs->mask.lip, AF_INET))
+ return false;
+ }
+
+ if (!fs->val.lport || fs->mask.lport != 0xffff)
+ return false;
+
+ if (!fs->val.fport || fs->mask.fport != 0xffff)
+ return false;
+
+ if (tp->fcoe_shift >= 0) {
+ mask = (hash_filter_mask >> tp->fcoe_shift) & FT_FCOE_W;
+ if (mask && !fs->mask.fcoe)
+ return false;
+ }
+
+ if (tp->port_shift >= 0) {
+ mask = (hash_filter_mask >> tp->port_shift) & FT_PORT_W;
+ if (mask && !fs->mask.iport)
+ return false;
+ }
+
+ if (tp->vnic_shift >= 0) {
+ mask = (hash_filter_mask >> tp->vnic_shift) & FT_VNIC_ID_W;
+
+ if ((adap->params.tp.ingress_config & VNIC_F)) {
+ if (mask && !fs->mask.pfvf_vld)
+ return false;
+ } else {
+ if (mask && !fs->mask.ovlan_vld)
+ return false;
+ }
+ }
+
+ if (tp->vlan_shift >= 0) {
+ mask = (hash_filter_mask >> tp->vlan_shift) & FT_VLAN_W;
+ if (mask && !fs->mask.ivlan)
+ return false;
+ }
+
+ if (tp->tos_shift >= 0) {
+ mask = (hash_filter_mask >> tp->tos_shift) & FT_TOS_W;
+ if (mask && !fs->mask.tos)
+ return false;
+ }
+
+ if (tp->protocol_shift >= 0) {
+ mask = (hash_filter_mask >> tp->protocol_shift) & FT_PROTOCOL_W;
+ if (mask && !fs->mask.proto)
+ return false;
+ }
+
+ if (tp->ethertype_shift >= 0) {
+ mask = (hash_filter_mask >> tp->ethertype_shift) &
+ FT_ETHERTYPE_W;
+ if (mask && !fs->mask.ethtype)
+ return false;
+ }
+
+ if (tp->macmatch_shift >= 0) {
+ mask = (hash_filter_mask >> tp->macmatch_shift) & FT_MACMATCH_W;
+ if (mask && !fs->mask.macidx)
+ return false;
+ }
+
+ if (tp->matchtype_shift >= 0) {
+ mask = (hash_filter_mask >> tp->matchtype_shift) &
+ FT_MPSHITTYPE_W;
+ if (mask && !fs->mask.matchtype)
+ return false;
+ }
+ if (tp->frag_shift >= 0) {
+ mask = (hash_filter_mask >> tp->frag_shift) &
+ FT_FRAGMENTATION_W;
+ if (mask && !fs->mask.frag)
+ return false;
+ }
+ return true;
+}
+
static u64 hash_filter_ntuple(struct ch_filter_specification *fs,
struct net_device *dev)
{
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.h
index 9475abd3384e..8db5fca6dcc9 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.h
@@ -51,4 +51,6 @@ int delete_filter(struct adapter *adapter, unsigned int fidx);
int writable_filter(struct filter_entry *f);
void clear_all_filters(struct adapter *adapter);
int init_hash_filter(struct adapter *adap);
+bool is_filter_exact_match(struct adapter *adap,
+ struct ch_filter_specification *fs);
#endif /* __CXGB4_FILTER_H */
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c
index 3953bc1fdc20..d4a548a6a55c 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c
@@ -38,6 +38,7 @@
#include <net/tc_act/tc_vlan.h>
#include "cxgb4.h"
+#include "cxgb4_filter.h"
#include "cxgb4_tc_flower.h"
#define STATS_CHECK_PERIOD (HZ / 2)
@@ -672,11 +673,16 @@ int cxgb4_tc_flower_replace(struct net_device *dev,
cxgb4_process_flow_match(dev, cls, fs);
cxgb4_process_flow_actions(dev, cls, fs);
- fidx = cxgb4_get_free_ftid(dev, fs->type ? PF_INET6 : PF_INET);
- if (fidx < 0) {
- netdev_err(dev, "%s: No fidx for offload.\n", __func__);
- ret = -ENOMEM;
- goto free_entry;
+ fs->hash = is_filter_exact_match(adap, fs);
+ if (fs->hash) {
+ fidx = 0;
+ } else {
+ fidx = cxgb4_get_free_ftid(dev, fs->type ? PF_INET6 : PF_INET);
+ if (fidx < 0) {
+ netdev_err(dev, "%s: No fidx for offload.\n", __func__);
+ ret = -ENOMEM;
+ goto free_entry;
+ }
}
init_completion(&ctx.completion);
--
2.14.1
^ permalink raw reply related [flat|nested] 9+ messages in thread* Re: [PATCH net-next 0/7] cxgb4: add hash-filter support to tc-flower offload
2017-11-01 3:22 [PATCH net-next 0/7] cxgb4: add hash-filter support to tc-flower offload Rahul Lakkireddy
` (6 preceding siblings ...)
2017-11-01 3:23 ` [PATCH net-next 7/7] cxgb4: add support to create hash-filters via tc-flower offload Rahul Lakkireddy
@ 2017-11-01 13:06 ` David Miller
7 siblings, 0 replies; 9+ messages in thread
From: David Miller @ 2017-11-01 13:06 UTC (permalink / raw)
To: rahul.lakkireddy; +Cc: netdev, kumaras, ganeshgr, nirranjan, indranil
From: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Date: Wed, 1 Nov 2017 08:52:58 +0530
> This series of patches add support to create hash-filters; a.k.a
> exact-match filters, to tc-flower offload. T6 supports creating
> ~500K hash-filters in hw and can theoretically be expanded up to
> ~1 million.
>
> Patch 1 fetches and saves the configured hw filter tuple field shifts
> and filter mask.
>
> Patch 2 initializes the driver to use hash-filter configuration.
>
> Patch 3 adds support to create hash filters in hw.
>
> Patch 4 adds support to delete hash filters in hw.
>
> Patch 5 adds support to retrieve filter stats for hash filters.
>
> Patch 6 converts the flower table to use rhashtable instead of
> static hlist.
>
> Patch 7 finally adds support to create hash filters via tc-flower
> offload.
Series applied, thanks.
^ permalink raw reply [flat|nested] 9+ messages in thread