* [net-next 0/8][pull request] 100GbE Intel Wired LAN Driver Updates 2020-01-25
@ 2020-01-26 6:07 Jeff Kirsher
2020-01-26 6:07 ` [net-next 1/8] ice: Enable writing hardware filtering tables Jeff Kirsher
` (8 more replies)
0 siblings, 9 replies; 10+ messages in thread
From: Jeff Kirsher @ 2020-01-26 6:07 UTC (permalink / raw)
To: davem; +Cc: Jeff Kirsher, netdev, nhorman, sassmann
This series contains updates to the ice driver to add support for RSS.
Henry and Tony enable the driver to write the filtering hardware tables
to allow for changing of RSS rules, also introduced and initialized the
structures for storing the configuration. Then followed it up by
creating an extraction sequence based on the packet header protocols to
be programmed. Next was storing the TCAM entry with the profile data
and VSI group in the respective software structures.
Md Fahad sets up the configuration to support RSS tables for the virtual
function (VF) driver (iavf). Add support for flow types TCP4, TCP6,
UDP4, UDP6, SCTP4 and SCTP6 for RSS via ethtool.
The following are changes since commit 3333e50b64fe30b7e53cf02456a2f567f689ae4f:
Merge branch 'mlxsw-Offload-TBF'
and are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next-queue 100GbE
Md Fahad Iqbal Polash (2):
ice: Initilialize VF RSS tables
ice: Implement ethtool get/set rx-flow-hash
Tony Nguyen (6):
ice: Enable writing hardware filtering tables
ice: Allocate flow profile
ice: Populate TCAM filter software structures
ice: Enable writing filtering tables
ice: Optimize table usage
ice: Bump version
drivers/net/ethernet/intel/ice/Makefile | 3 +-
.../net/ethernet/intel/ice/ice_adminq_cmd.h | 8 +
drivers/net/ethernet/intel/ice/ice_common.c | 114 +-
drivers/net/ethernet/intel/ice/ice_common.h | 8 +
drivers/net/ethernet/intel/ice/ice_ethtool.c | 243 ++
.../net/ethernet/intel/ice/ice_flex_pipe.c | 2575 ++++++++++++++++-
.../net/ethernet/intel/ice/ice_flex_pipe.h | 9 +
.../net/ethernet/intel/ice/ice_flex_type.h | 112 +
drivers/net/ethernet/intel/ice/ice_flow.c | 1275 ++++++++
drivers/net/ethernet/intel/ice/ice_flow.h | 207 ++
.../net/ethernet/intel/ice/ice_lan_tx_rx.h | 8 +
drivers/net/ethernet/intel/ice/ice_lib.c | 146 +-
drivers/net/ethernet/intel/ice/ice_main.c | 2 +-
.../ethernet/intel/ice/ice_protocol_type.h | 25 +
drivers/net/ethernet/intel/ice/ice_status.h | 1 +
drivers/net/ethernet/intel/ice/ice_switch.c | 36 -
drivers/net/ethernet/intel/ice/ice_type.h | 6 +
17 files changed, 4729 insertions(+), 49 deletions(-)
create mode 100644 drivers/net/ethernet/intel/ice/ice_flow.c
create mode 100644 drivers/net/ethernet/intel/ice/ice_flow.h
create mode 100644 drivers/net/ethernet/intel/ice/ice_protocol_type.h
--
2.24.1
^ permalink raw reply [flat|nested] 10+ messages in thread
* [net-next 1/8] ice: Enable writing hardware filtering tables
2020-01-26 6:07 [net-next 0/8][pull request] 100GbE Intel Wired LAN Driver Updates 2020-01-25 Jeff Kirsher
@ 2020-01-26 6:07 ` Jeff Kirsher
2020-01-26 6:07 ` [net-next 2/8] ice: Allocate flow profile Jeff Kirsher
` (7 subsequent siblings)
8 siblings, 0 replies; 10+ messages in thread
From: Jeff Kirsher @ 2020-01-26 6:07 UTC (permalink / raw)
To: davem
Cc: Tony Nguyen, netdev, nhorman, sassmann, Henry Tieman,
Andrew Bowers, Jeff Kirsher
From: Tony Nguyen <anthony.l.nguyen@intel.com>
Enable the driver to write the filtering hardware tables to allow for
changing of RSS rules. Upon loading of DDP package, a minimal configuration
should be written to hardware.
Introduce and initialize structures for storing configuration and make
the top level calls to configure the RSS tables to initial values. A packet
segment will be created but nothing is written to hardware yet.
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Henry Tieman <henry.w.tieman@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
drivers/net/ethernet/intel/ice/Makefile | 3 +-
drivers/net/ethernet/intel/ice/ice_common.c | 6 +-
.../net/ethernet/intel/ice/ice_flex_pipe.c | 31 ++-
drivers/net/ethernet/intel/ice/ice_flow.c | 250 ++++++++++++++++++
drivers/net/ethernet/intel/ice/ice_flow.h | 114 ++++++++
drivers/net/ethernet/intel/ice/ice_lib.c | 87 +++++-
drivers/net/ethernet/intel/ice/ice_type.h | 4 +
7 files changed, 491 insertions(+), 4 deletions(-)
create mode 100644 drivers/net/ethernet/intel/ice/ice_flow.c
create mode 100644 drivers/net/ethernet/intel/ice/ice_flow.h
diff --git a/drivers/net/ethernet/intel/ice/Makefile b/drivers/net/ethernet/intel/ice/Makefile
index 7cb829132d28..59544b0fc086 100644
--- a/drivers/net/ethernet/intel/ice/Makefile
+++ b/drivers/net/ethernet/intel/ice/Makefile
@@ -17,7 +17,8 @@ ice-y := ice_main.o \
ice_lib.o \
ice_txrx_lib.o \
ice_txrx.o \
- ice_flex_pipe.o \
+ ice_flex_pipe.o \
+ ice_flow.o \
ice_ethtool.o
ice-$(CONFIG_PCI_IOV) += ice_virtchnl_pf.o ice_sriov.o
ice-$(CONFIG_DCB) += ice_dcb.o ice_dcb_nl.o ice_dcb_lib.o
diff --git a/drivers/net/ethernet/intel/ice/ice_common.c b/drivers/net/ethernet/intel/ice/ice_common.c
index a03b4fdc01e6..dd9af9c63755 100644
--- a/drivers/net/ethernet/intel/ice/ice_common.c
+++ b/drivers/net/ethernet/intel/ice/ice_common.c
@@ -4,6 +4,7 @@
#include "ice_common.h"
#include "ice_sched.h"
#include "ice_adminq_cmd.h"
+#include "ice_flow.h"
#define ICE_PF_RESET_WAIT_COUNT 200
@@ -3406,7 +3407,10 @@ enum ice_status ice_replay_vsi(struct ice_hw *hw, u16 vsi_handle)
if (status)
return status;
}
-
+ /* Replay per VSI all RSS configurations */
+ status = ice_replay_rss_cfg(hw, vsi_handle);
+ if (status)
+ return status;
/* Replay per VSI all filters */
status = ice_replay_vsi_all_fltr(hw, vsi_handle);
return status;
diff --git a/drivers/net/ethernet/intel/ice/ice_flex_pipe.c b/drivers/net/ethernet/intel/ice/ice_flex_pipe.c
index cbd53b586c36..f37eb536e7b7 100644
--- a/drivers/net/ethernet/intel/ice/ice_flex_pipe.c
+++ b/drivers/net/ethernet/intel/ice/ice_flex_pipe.c
@@ -3,6 +3,7 @@
#include "ice_common.h"
#include "ice_flex_pipe.h"
+#include "ice_flow.h"
/**
* ice_pkg_val_buf
@@ -1379,11 +1380,18 @@ void ice_fill_blk_tbls(struct ice_hw *hw)
*/
void ice_free_hw_tbls(struct ice_hw *hw)
{
+ struct ice_rss_cfg *r, *rt;
u8 i;
for (i = 0; i < ICE_BLK_COUNT; i++) {
- hw->blk[i].is_list_init = false;
+ if (hw->blk[i].is_list_init) {
+ struct ice_es *es = &hw->blk[i].es;
+ mutex_destroy(&es->prof_map_lock);
+ mutex_destroy(&hw->fl_profs_locks[i]);
+
+ hw->blk[i].is_list_init = false;
+ }
devm_kfree(ice_hw_to_dev(hw), hw->blk[i].xlt1.ptypes);
devm_kfree(ice_hw_to_dev(hw), hw->blk[i].xlt1.ptg_tbl);
devm_kfree(ice_hw_to_dev(hw), hw->blk[i].xlt1.t);
@@ -1397,9 +1405,25 @@ void ice_free_hw_tbls(struct ice_hw *hw)
devm_kfree(ice_hw_to_dev(hw), hw->blk[i].es.written);
}
+ list_for_each_entry_safe(r, rt, &hw->rss_list_head, l_entry) {
+ list_del(&r->l_entry);
+ devm_kfree(ice_hw_to_dev(hw), r);
+ }
+ mutex_destroy(&hw->rss_locks);
memset(hw->blk, 0, sizeof(hw->blk));
}
+/**
+ * ice_init_flow_profs - init flow profile locks and list heads
+ * @hw: pointer to the hardware structure
+ * @blk_idx: HW block index
+ */
+static void ice_init_flow_profs(struct ice_hw *hw, u8 blk_idx)
+{
+ mutex_init(&hw->fl_profs_locks[blk_idx]);
+ INIT_LIST_HEAD(&hw->fl_profs[blk_idx]);
+}
+
/**
* ice_clear_hw_tbls - clear HW tables and flow profiles
* @hw: pointer to the hardware structure
@@ -1443,6 +1467,8 @@ enum ice_status ice_init_hw_tbls(struct ice_hw *hw)
{
u8 i;
+ mutex_init(&hw->rss_locks);
+ INIT_LIST_HEAD(&hw->rss_list_head);
for (i = 0; i < ICE_BLK_COUNT; i++) {
struct ice_prof_redir *prof_redir = &hw->blk[i].prof_redir;
struct ice_prof_tcam *prof = &hw->blk[i].prof;
@@ -1454,6 +1480,9 @@ enum ice_status ice_init_hw_tbls(struct ice_hw *hw)
if (hw->blk[i].is_list_init)
continue;
+ ice_init_flow_profs(hw, i);
+ mutex_init(&es->prof_map_lock);
+ INIT_LIST_HEAD(&es->prof_map);
hw->blk[i].is_list_init = true;
hw->blk[i].overwrite = blk_sizes[i].overwrite;
diff --git a/drivers/net/ethernet/intel/ice/ice_flow.c b/drivers/net/ethernet/intel/ice/ice_flow.c
new file mode 100644
index 000000000000..4828c64abe98
--- /dev/null
+++ b/drivers/net/ethernet/intel/ice/ice_flow.c
@@ -0,0 +1,250 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2019, Intel Corporation. */
+
+#include "ice_common.h"
+#include "ice_flow.h"
+
+/* Describe properties of a protocol header field */
+struct ice_flow_field_info {
+ enum ice_flow_seg_hdr hdr;
+ s16 off; /* Offset from start of a protocol header, in bits */
+ u16 size; /* Size of fields in bits */
+};
+
+#define ICE_FLOW_FLD_INFO(_hdr, _offset_bytes, _size_bytes) { \
+ .hdr = _hdr, \
+ .off = (_offset_bytes) * BITS_PER_BYTE, \
+ .size = (_size_bytes) * BITS_PER_BYTE, \
+}
+
+/* Table containing properties of supported protocol header fields */
+static const
+struct ice_flow_field_info ice_flds_info[ICE_FLOW_FIELD_IDX_MAX] = {
+ /* IPv4 / IPv6 */
+ /* ICE_FLOW_FIELD_IDX_IPV4_SA */
+ ICE_FLOW_FLD_INFO(ICE_FLOW_SEG_HDR_IPV4, 12, sizeof(struct in_addr)),
+ /* ICE_FLOW_FIELD_IDX_IPV4_DA */
+ ICE_FLOW_FLD_INFO(ICE_FLOW_SEG_HDR_IPV4, 16, sizeof(struct in_addr)),
+ /* ICE_FLOW_FIELD_IDX_IPV6_SA */
+ ICE_FLOW_FLD_INFO(ICE_FLOW_SEG_HDR_IPV6, 8, sizeof(struct in6_addr)),
+ /* ICE_FLOW_FIELD_IDX_IPV6_DA */
+ ICE_FLOW_FLD_INFO(ICE_FLOW_SEG_HDR_IPV6, 24, sizeof(struct in6_addr)),
+ /* Transport */
+ /* ICE_FLOW_FIELD_IDX_TCP_SRC_PORT */
+ ICE_FLOW_FLD_INFO(ICE_FLOW_SEG_HDR_TCP, 0, sizeof(__be16)),
+ /* ICE_FLOW_FIELD_IDX_TCP_DST_PORT */
+ ICE_FLOW_FLD_INFO(ICE_FLOW_SEG_HDR_TCP, 2, sizeof(__be16)),
+ /* ICE_FLOW_FIELD_IDX_UDP_SRC_PORT */
+ ICE_FLOW_FLD_INFO(ICE_FLOW_SEG_HDR_UDP, 0, sizeof(__be16)),
+ /* ICE_FLOW_FIELD_IDX_UDP_DST_PORT */
+ ICE_FLOW_FLD_INFO(ICE_FLOW_SEG_HDR_UDP, 2, sizeof(__be16)),
+};
+
+/**
+ * ice_flow_set_fld_ext - specifies locations of field from entry's input buffer
+ * @seg: packet segment the field being set belongs to
+ * @fld: field to be set
+ * @type: type of the field
+ * @val_loc: if not ICE_FLOW_FLD_OFF_INVAL, location of the value to match from
+ * entry's input buffer
+ * @mask_loc: if not ICE_FLOW_FLD_OFF_INVAL, location of mask value from entry's
+ * input buffer
+ * @last_loc: if not ICE_FLOW_FLD_OFF_INVAL, location of last/upper value from
+ * entry's input buffer
+ *
+ * This helper function stores information of a field being matched, including
+ * the type of the field and the locations of the value to match, the mask, and
+ * and the upper-bound value in the start of the input buffer for a flow entry.
+ * This function should only be used for fixed-size data structures.
+ *
+ * This function also opportunistically determines the protocol headers to be
+ * present based on the fields being set. Some fields cannot be used alone to
+ * determine the protocol headers present. Sometimes, fields for particular
+ * protocol headers are not matched. In those cases, the protocol headers
+ * must be explicitly set.
+ */
+static void
+ice_flow_set_fld_ext(struct ice_flow_seg_info *seg, enum ice_flow_field fld,
+ enum ice_flow_fld_match_type type, u16 val_loc,
+ u16 mask_loc, u16 last_loc)
+{
+ u64 bit = BIT_ULL(fld);
+
+ seg->match |= bit;
+ if (type == ICE_FLOW_FLD_TYPE_RANGE)
+ seg->range |= bit;
+
+ seg->fields[fld].type = type;
+ seg->fields[fld].src.val = val_loc;
+ seg->fields[fld].src.mask = mask_loc;
+ seg->fields[fld].src.last = last_loc;
+
+ ICE_FLOW_SET_HDRS(seg, ice_flds_info[fld].hdr);
+}
+
+/**
+ * ice_flow_set_fld - specifies locations of field from entry's input buffer
+ * @seg: packet segment the field being set belongs to
+ * @fld: field to be set
+ * @val_loc: if not ICE_FLOW_FLD_OFF_INVAL, location of the value to match from
+ * entry's input buffer
+ * @mask_loc: if not ICE_FLOW_FLD_OFF_INVAL, location of mask value from entry's
+ * input buffer
+ * @last_loc: if not ICE_FLOW_FLD_OFF_INVAL, location of last/upper value from
+ * entry's input buffer
+ * @range: indicate if field being matched is to be in a range
+ *
+ * This function specifies the locations, in the form of byte offsets from the
+ * start of the input buffer for a flow entry, from where the value to match,
+ * the mask value, and upper value can be extracted. These locations are then
+ * stored in the flow profile. When adding a flow entry associated with the
+ * flow profile, these locations will be used to quickly extract the values and
+ * create the content of a match entry. This function should only be used for
+ * fixed-size data structures.
+ */
+static void
+ice_flow_set_fld(struct ice_flow_seg_info *seg, enum ice_flow_field fld,
+ u16 val_loc, u16 mask_loc, u16 last_loc, bool range)
+{
+ enum ice_flow_fld_match_type t = range ?
+ ICE_FLOW_FLD_TYPE_RANGE : ICE_FLOW_FLD_TYPE_REG;
+
+ ice_flow_set_fld_ext(seg, fld, t, val_loc, mask_loc, last_loc);
+}
+
+#define ICE_FLOW_RSS_SEG_HDR_L3_MASKS \
+ (ICE_FLOW_SEG_HDR_IPV4 | ICE_FLOW_SEG_HDR_IPV6)
+
+#define ICE_FLOW_RSS_SEG_HDR_L4_MASKS \
+ (ICE_FLOW_SEG_HDR_TCP | ICE_FLOW_SEG_HDR_UDP | ICE_FLOW_SEG_HDR_SCTP)
+
+#define ICE_FLOW_RSS_SEG_HDR_VAL_MASKS \
+ (ICE_FLOW_RSS_SEG_HDR_L3_MASKS | \
+ ICE_FLOW_RSS_SEG_HDR_L4_MASKS)
+
+/**
+ * ice_flow_set_rss_seg_info - setup packet segments for RSS
+ * @segs: pointer to the flow field segment(s)
+ * @hash_fields: fields to be hashed on for the segment(s)
+ * @flow_hdr: protocol header fields within a packet segment
+ *
+ * Helper function to extract fields from hash bitmap and use flow
+ * header value to set flow field segment for further use in flow
+ * profile entry or removal.
+ */
+static enum ice_status
+ice_flow_set_rss_seg_info(struct ice_flow_seg_info *segs, u64 hash_fields,
+ u32 flow_hdr)
+{
+ u64 val;
+ u8 i;
+
+ for_each_set_bit(i, (unsigned long *)&hash_fields,
+ ICE_FLOW_FIELD_IDX_MAX)
+ ice_flow_set_fld(segs, (enum ice_flow_field)i,
+ ICE_FLOW_FLD_OFF_INVAL, ICE_FLOW_FLD_OFF_INVAL,
+ ICE_FLOW_FLD_OFF_INVAL, false);
+
+ ICE_FLOW_SET_HDRS(segs, flow_hdr);
+
+ if (segs->hdrs & ~ICE_FLOW_RSS_SEG_HDR_VAL_MASKS)
+ return ICE_ERR_PARAM;
+
+ val = (u64)(segs->hdrs & ICE_FLOW_RSS_SEG_HDR_L3_MASKS);
+ if (val && !is_power_of_2(val))
+ return ICE_ERR_CFG;
+
+ val = (u64)(segs->hdrs & ICE_FLOW_RSS_SEG_HDR_L4_MASKS);
+ if (val && !is_power_of_2(val))
+ return ICE_ERR_CFG;
+
+ return 0;
+}
+
+#define ICE_RSS_OUTER_HEADERS 1
+
+/**
+ * ice_add_rss_cfg_sync - add an RSS configuration
+ * @hashed_flds: hash bit fields (ICE_FLOW_HASH_*) to configure
+ * @addl_hdrs: protocol header fields
+ * @segs_cnt: packet segment count
+ *
+ * Assumption: lock has already been acquired for RSS list
+ */
+static enum ice_status
+ice_add_rss_cfg_sync(u64 hashed_flds, u32 addl_hdrs, u8 segs_cnt)
+{
+ struct ice_flow_seg_info *segs;
+ enum ice_status status;
+
+ if (!segs_cnt || segs_cnt > ICE_FLOW_SEG_MAX)
+ return ICE_ERR_PARAM;
+
+ segs = kcalloc(segs_cnt, sizeof(*segs), GFP_KERNEL);
+ if (!segs)
+ return ICE_ERR_NO_MEMORY;
+
+ /* Construct the packet segment info from the hashed fields */
+ status = ice_flow_set_rss_seg_info(&segs[segs_cnt - 1], hashed_flds,
+ addl_hdrs);
+
+ kfree(segs);
+ return status;
+}
+
+/**
+ * ice_add_rss_cfg - add an RSS configuration with specified hashed fields
+ * @hw: pointer to the hardware structure
+ * @vsi_handle: software VSI handle
+ * @hashed_flds: hash bit fields (ICE_FLOW_HASH_*) to configure
+ * @addl_hdrs: protocol header fields
+ *
+ * This function will generate a flow profile based on fields associated with
+ * the input fields to hash on, the flow type and use the VSI number to add
+ * a flow entry to the profile.
+ */
+enum ice_status
+ice_add_rss_cfg(struct ice_hw *hw, u16 vsi_handle, u64 hashed_flds,
+ u32 addl_hdrs)
+{
+ enum ice_status status;
+
+ if (hashed_flds == ICE_HASH_INVALID ||
+ !ice_is_vsi_valid(hw, vsi_handle))
+ return ICE_ERR_PARAM;
+
+ mutex_lock(&hw->rss_locks);
+ status = ice_add_rss_cfg_sync(hashed_flds, addl_hdrs,
+ ICE_RSS_OUTER_HEADERS);
+ mutex_unlock(&hw->rss_locks);
+
+ return status;
+}
+
+/**
+ * ice_replay_rss_cfg - replay RSS configurations associated with VSI
+ * @hw: pointer to the hardware structure
+ * @vsi_handle: software VSI handle
+ */
+enum ice_status ice_replay_rss_cfg(struct ice_hw *hw, u16 vsi_handle)
+{
+ enum ice_status status = 0;
+ struct ice_rss_cfg *r;
+
+ if (!ice_is_vsi_valid(hw, vsi_handle))
+ return ICE_ERR_PARAM;
+
+ mutex_lock(&hw->rss_locks);
+ list_for_each_entry(r, &hw->rss_list_head, l_entry) {
+ if (test_bit(vsi_handle, r->vsis)) {
+ status = ice_add_rss_cfg_sync(r->hashed_flds,
+ r->packet_hdr,
+ ICE_RSS_OUTER_HEADERS);
+ if (status)
+ break;
+ }
+ }
+ mutex_unlock(&hw->rss_locks);
+
+ return status;
+}
diff --git a/drivers/net/ethernet/intel/ice/ice_flow.h b/drivers/net/ethernet/intel/ice/ice_flow.h
new file mode 100644
index 000000000000..48c0fc09d5ff
--- /dev/null
+++ b/drivers/net/ethernet/intel/ice/ice_flow.h
@@ -0,0 +1,114 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/* Copyright (c) 2019, Intel Corporation. */
+
+#ifndef _ICE_FLOW_H_
+#define _ICE_FLOW_H_
+
+#define ICE_FLOW_FLD_OFF_INVAL 0xffff
+
+/* Generate flow hash field from flow field type(s) */
+#define ICE_FLOW_HASH_IPV4 \
+ (BIT_ULL(ICE_FLOW_FIELD_IDX_IPV4_SA) | \
+ BIT_ULL(ICE_FLOW_FIELD_IDX_IPV4_DA))
+#define ICE_FLOW_HASH_IPV6 \
+ (BIT_ULL(ICE_FLOW_FIELD_IDX_IPV6_SA) | \
+ BIT_ULL(ICE_FLOW_FIELD_IDX_IPV6_DA))
+#define ICE_FLOW_HASH_TCP_PORT \
+ (BIT_ULL(ICE_FLOW_FIELD_IDX_TCP_SRC_PORT) | \
+ BIT_ULL(ICE_FLOW_FIELD_IDX_TCP_DST_PORT))
+#define ICE_FLOW_HASH_UDP_PORT \
+ (BIT_ULL(ICE_FLOW_FIELD_IDX_UDP_SRC_PORT) | \
+ BIT_ULL(ICE_FLOW_FIELD_IDX_UDP_DST_PORT))
+
+#define ICE_HASH_INVALID 0
+#define ICE_HASH_TCP_IPV4 (ICE_FLOW_HASH_IPV4 | ICE_FLOW_HASH_TCP_PORT)
+#define ICE_HASH_TCP_IPV6 (ICE_FLOW_HASH_IPV6 | ICE_FLOW_HASH_TCP_PORT)
+#define ICE_HASH_UDP_IPV4 (ICE_FLOW_HASH_IPV4 | ICE_FLOW_HASH_UDP_PORT)
+#define ICE_HASH_UDP_IPV6 (ICE_FLOW_HASH_IPV6 | ICE_FLOW_HASH_UDP_PORT)
+
+/* Protocol header fields within a packet segment. A segment consists of one or
+ * more protocol headers that make up a logical group of protocol headers. Each
+ * logical group of protocol headers encapsulates or is encapsulated using/by
+ * tunneling or encapsulation protocols for network virtualization such as GRE,
+ * VxLAN, etc.
+ */
+enum ice_flow_seg_hdr {
+ ICE_FLOW_SEG_HDR_NONE = 0x00000000,
+ ICE_FLOW_SEG_HDR_IPV4 = 0x00000004,
+ ICE_FLOW_SEG_HDR_IPV6 = 0x00000008,
+ ICE_FLOW_SEG_HDR_TCP = 0x00000040,
+ ICE_FLOW_SEG_HDR_UDP = 0x00000080,
+ ICE_FLOW_SEG_HDR_SCTP = 0x00000100,
+};
+
+enum ice_flow_field {
+ /* L3 */
+ ICE_FLOW_FIELD_IDX_IPV4_SA,
+ ICE_FLOW_FIELD_IDX_IPV4_DA,
+ ICE_FLOW_FIELD_IDX_IPV6_SA,
+ ICE_FLOW_FIELD_IDX_IPV6_DA,
+ /* L4 */
+ ICE_FLOW_FIELD_IDX_TCP_SRC_PORT,
+ ICE_FLOW_FIELD_IDX_TCP_DST_PORT,
+ ICE_FLOW_FIELD_IDX_UDP_SRC_PORT,
+ ICE_FLOW_FIELD_IDX_UDP_DST_PORT,
+ /* The total number of enums must not exceed 64 */
+ ICE_FLOW_FIELD_IDX_MAX
+};
+
+#define ICE_FLOW_SEG_MAX 2
+#define ICE_FLOW_SET_HDRS(seg, val) ((seg)->hdrs |= (u32)(val))
+
+struct ice_flow_seg_xtrct {
+ u8 prot_id; /* Protocol ID of extracted header field */
+ u16 off; /* Starting offset of the field in header in bytes */
+ u8 idx; /* Index of FV entry used */
+ u8 disp; /* Displacement of field in bits fr. FV entry's start */
+};
+
+enum ice_flow_fld_match_type {
+ ICE_FLOW_FLD_TYPE_REG, /* Value, mask */
+ ICE_FLOW_FLD_TYPE_RANGE, /* Value, mask, last (upper bound) */
+ ICE_FLOW_FLD_TYPE_PREFIX, /* IP address, prefix, size of prefix */
+ ICE_FLOW_FLD_TYPE_SIZE, /* Value, mask, size of match */
+};
+
+struct ice_flow_fld_loc {
+ /* Describe offsets of field information relative to the beginning of
+ * input buffer provided when adding flow entries.
+ */
+ u16 val; /* Offset where the value is located */
+ u16 mask; /* Offset where the mask/prefix value is located */
+ u16 last; /* Length or offset where the upper value is located */
+};
+
+struct ice_flow_fld_info {
+ enum ice_flow_fld_match_type type;
+ /* Location where to retrieve data from an input buffer */
+ struct ice_flow_fld_loc src;
+ /* Location where to put the data into the final entry buffer */
+ struct ice_flow_fld_loc entry;
+ struct ice_flow_seg_xtrct xtrct;
+};
+
+struct ice_flow_seg_info {
+ u32 hdrs; /* Bitmask indicating protocol headers present */
+ u64 match; /* Bitmask indicating header fields to be matched */
+ u64 range; /* Bitmask indicating header fields matched as ranges */
+
+ struct ice_flow_fld_info fields[ICE_FLOW_FIELD_IDX_MAX];
+};
+
+struct ice_rss_cfg {
+ struct list_head l_entry;
+ /* bitmap of VSIs added to the RSS entry */
+ DECLARE_BITMAP(vsis, ICE_MAX_VSI);
+ u64 hashed_flds;
+ u32 packet_hdr;
+};
+
+enum ice_status ice_replay_rss_cfg(struct ice_hw *hw, u16 vsi_handle);
+enum ice_status
+ice_add_rss_cfg(struct ice_hw *hw, u16 vsi_handle, u64 hashed_flds,
+ u32 addl_hdrs);
+#endif /* _ICE_FLOW_H_ */
diff --git a/drivers/net/ethernet/intel/ice/ice_lib.c b/drivers/net/ethernet/intel/ice/ice_lib.c
index 4cfad81ba496..dd230c6664ee 100644
--- a/drivers/net/ethernet/intel/ice/ice_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_lib.c
@@ -3,6 +3,7 @@
#include "ice.h"
#include "ice_base.h"
+#include "ice_flow.h"
#include "ice_lib.h"
#include "ice_dcb_lib.h"
@@ -1086,6 +1087,88 @@ static int ice_vsi_cfg_rss_lut_key(struct ice_vsi *vsi)
return err;
}
+/**
+ * ice_vsi_set_rss_flow_fld - Sets RSS input set for different flows
+ * @vsi: VSI to be configured
+ *
+ * This function will only be called after successful download package call
+ * during initialization of PF. Since the downloaded package will erase the
+ * RSS section, this function will configure RSS input sets for different
+ * flow types. The last profile added has the highest priority, therefore 2
+ * tuple profiles (i.e. IPv4 src/dst) are added before 4 tuple profiles
+ * (i.e. IPv4 src/dst TCP src/dst port).
+ */
+static void ice_vsi_set_rss_flow_fld(struct ice_vsi *vsi)
+{
+ u16 vsi_handle = vsi->idx, vsi_num = vsi->vsi_num;
+ struct ice_pf *pf = vsi->back;
+ struct ice_hw *hw = &pf->hw;
+ enum ice_status status;
+ struct device *dev;
+
+ dev = ice_pf_to_dev(pf);
+ if (ice_is_safe_mode(pf)) {
+ dev_dbg(dev, "Advanced RSS disabled. Package download failed, vsi num = %d\n",
+ vsi_num);
+ return;
+ }
+ /* configure RSS for IPv4 with input set IP src/dst */
+ status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV4,
+ ICE_FLOW_SEG_HDR_IPV4);
+ if (status)
+ dev_dbg(dev, "ice_add_rss_cfg failed for ipv4 flow, vsi = %d, error = %d\n",
+ vsi_num, status);
+
+ /* configure RSS for IPv6 with input set IPv6 src/dst */
+ status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV6,
+ ICE_FLOW_SEG_HDR_IPV6);
+ if (status)
+ dev_dbg(dev, "ice_add_rss_cfg failed for ipv6 flow, vsi = %d, error = %d\n",
+ vsi_num, status);
+
+ /* configure RSS for tcp4 with input set IP src/dst, TCP src/dst */
+ status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_TCP_IPV4,
+ ICE_FLOW_SEG_HDR_TCP | ICE_FLOW_SEG_HDR_IPV4);
+ if (status)
+ dev_dbg(dev, "ice_add_rss_cfg failed for tcp4 flow, vsi = %d, error = %d\n",
+ vsi_num, status);
+
+ /* configure RSS for udp4 with input set IP src/dst, UDP src/dst */
+ status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_UDP_IPV4,
+ ICE_FLOW_SEG_HDR_UDP | ICE_FLOW_SEG_HDR_IPV4);
+ if (status)
+ dev_dbg(dev, "ice_add_rss_cfg failed for udp4 flow, vsi = %d, error = %d\n",
+ vsi_num, status);
+
+ /* configure RSS for sctp4 with input set IP src/dst */
+ status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV4,
+ ICE_FLOW_SEG_HDR_SCTP | ICE_FLOW_SEG_HDR_IPV4);
+ if (status)
+ dev_dbg(dev, "ice_add_rss_cfg failed for sctp4 flow, vsi = %d, error = %d\n",
+ vsi_num, status);
+
+ /* configure RSS for tcp6 with input set IPv6 src/dst, TCP src/dst */
+ status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_TCP_IPV6,
+ ICE_FLOW_SEG_HDR_TCP | ICE_FLOW_SEG_HDR_IPV6);
+ if (status)
+ dev_dbg(dev, "ice_add_rss_cfg failed for tcp6 flow, vsi = %d, error = %d\n",
+ vsi_num, status);
+
+ /* configure RSS for udp6 with input set IPv6 src/dst, UDP src/dst */
+ status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_UDP_IPV6,
+ ICE_FLOW_SEG_HDR_UDP | ICE_FLOW_SEG_HDR_IPV6);
+ if (status)
+ dev_dbg(dev, "ice_add_rss_cfg failed for udp6 flow, vsi = %d, error = %d\n",
+ vsi_num, status);
+
+ /* configure RSS for sctp6 with input set IPv6 src/dst */
+ status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV6,
+ ICE_FLOW_SEG_HDR_SCTP | ICE_FLOW_SEG_HDR_IPV6);
+ if (status)
+ dev_dbg(dev, "ice_add_rss_cfg failed for sctp6 flow, vsi = %d, error = %d\n",
+ vsi_num, status);
+}
+
/**
* ice_add_mac_to_list - Add a MAC address filter entry to the list
* @vsi: the VSI to be forwarded to
@@ -1901,8 +1984,10 @@ ice_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi,
* receive traffic on first queue. Hence no need to capture
* return value
*/
- if (test_bit(ICE_FLAG_RSS_ENA, pf->flags))
+ if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
ice_vsi_cfg_rss_lut_key(vsi);
+ ice_vsi_set_rss_flow_fld(vsi);
+ }
break;
case ICE_VSI_VF:
/* VF driver will take care of creating netdev for this type and
diff --git a/drivers/net/ethernet/intel/ice/ice_type.h b/drivers/net/ethernet/intel/ice/ice_type.h
index c4854a987130..ac38af105f90 100644
--- a/drivers/net/ethernet/intel/ice/ice_type.h
+++ b/drivers/net/ethernet/intel/ice/ice_type.h
@@ -559,6 +559,10 @@ struct ice_hw {
/* HW block tables */
struct ice_blk_info blk[ICE_BLK_COUNT];
+ struct mutex fl_profs_locks[ICE_BLK_COUNT]; /* lock fltr profiles */
+ struct list_head fl_profs[ICE_BLK_COUNT];
+ struct mutex rss_locks; /* protect RSS configuration */
+ struct list_head rss_list_head;
};
/* Statistics collected by each port, VSI, VEB, and S-channel */
--
2.24.1
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [net-next 2/8] ice: Allocate flow profile
2020-01-26 6:07 [net-next 0/8][pull request] 100GbE Intel Wired LAN Driver Updates 2020-01-25 Jeff Kirsher
2020-01-26 6:07 ` [net-next 1/8] ice: Enable writing hardware filtering tables Jeff Kirsher
@ 2020-01-26 6:07 ` Jeff Kirsher
2020-01-26 6:07 ` [net-next 3/8] ice: Populate TCAM filter software structures Jeff Kirsher
` (6 subsequent siblings)
8 siblings, 0 replies; 10+ messages in thread
From: Jeff Kirsher @ 2020-01-26 6:07 UTC (permalink / raw)
To: davem
Cc: Tony Nguyen, netdev, nhorman, sassmann, Henry Tieman,
Andrew Bowers, Jeff Kirsher
From: Tony Nguyen <anthony.l.nguyen@intel.com>
Create an extraction sequence based on the packet header protocols to be
programmed and allocate a flow profile for the extraction sequence.
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Henry Tieman <henry.w.tieman@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
.../net/ethernet/intel/ice/ice_adminq_cmd.h | 6 +
drivers/net/ethernet/intel/ice/ice_common.c | 75 +++
drivers/net/ethernet/intel/ice/ice_common.h | 6 +
.../net/ethernet/intel/ice/ice_flex_pipe.c | 216 ++++++++
.../net/ethernet/intel/ice/ice_flex_pipe.h | 3 +
.../net/ethernet/intel/ice/ice_flex_type.h | 15 +
drivers/net/ethernet/intel/ice/ice_flow.c | 490 +++++++++++++++++-
drivers/net/ethernet/intel/ice/ice_flow.h | 23 +
.../net/ethernet/intel/ice/ice_lan_tx_rx.h | 8 +
.../ethernet/intel/ice/ice_protocol_type.h | 24 +
drivers/net/ethernet/intel/ice/ice_switch.c | 36 --
drivers/net/ethernet/intel/ice/ice_type.h | 2 +
12 files changed, 864 insertions(+), 40 deletions(-)
create mode 100644 drivers/net/ethernet/intel/ice/ice_protocol_type.h
diff --git a/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h b/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h
index 5421fc413f94..ad15a772b609 100644
--- a/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h
+++ b/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h
@@ -232,6 +232,12 @@ struct ice_aqc_get_sw_cfg_resp {
*/
#define ICE_AQC_RES_TYPE_VSI_LIST_REP 0x03
#define ICE_AQC_RES_TYPE_VSI_LIST_PRUNE 0x04
+#define ICE_AQC_RES_TYPE_HASH_PROF_BLDR_PROFID 0x60
+
+#define ICE_AQC_RES_TYPE_FLAG_SCAN_BOTTOM BIT(12)
+#define ICE_AQC_RES_TYPE_FLAG_IGNORE_INDEX BIT(13)
+
+#define ICE_AQC_RES_TYPE_FLAG_DEDICATED 0x00
/* Allocate Resources command (indirect 0x0208)
* Free Resources command (indirect 0x0209)
diff --git a/drivers/net/ethernet/intel/ice/ice_common.c b/drivers/net/ethernet/intel/ice/ice_common.c
index dd9af9c63755..8f86962fd052 100644
--- a/drivers/net/ethernet/intel/ice/ice_common.c
+++ b/drivers/net/ethernet/intel/ice/ice_common.c
@@ -1497,6 +1497,81 @@ void ice_release_res(struct ice_hw *hw, enum ice_aq_res_ids res)
}
}
+/**
+ * ice_aq_alloc_free_res - command to allocate/free resources
+ * @hw: pointer to the HW struct
+ * @num_entries: number of resource entries in buffer
+ * @buf: Indirect buffer to hold data parameters and response
+ * @buf_size: size of buffer for indirect commands
+ * @opc: pass in the command opcode
+ * @cd: pointer to command details structure or NULL
+ *
+ * Helper function to allocate/free resources using the admin queue commands
+ */
+enum ice_status
+ice_aq_alloc_free_res(struct ice_hw *hw, u16 num_entries,
+ struct ice_aqc_alloc_free_res_elem *buf, u16 buf_size,
+ enum ice_adminq_opc opc, struct ice_sq_cd *cd)
+{
+ struct ice_aqc_alloc_free_res_cmd *cmd;
+ struct ice_aq_desc desc;
+
+ cmd = &desc.params.sw_res_ctrl;
+
+ if (!buf)
+ return ICE_ERR_PARAM;
+
+ if (buf_size < (num_entries * sizeof(buf->elem[0])))
+ return ICE_ERR_PARAM;
+
+ ice_fill_dflt_direct_cmd_desc(&desc, opc);
+
+ desc.flags |= cpu_to_le16(ICE_AQ_FLAG_RD);
+
+ cmd->num_entries = cpu_to_le16(num_entries);
+
+ return ice_aq_send_cmd(hw, &desc, buf, buf_size, cd);
+}
+
+/**
+ * ice_alloc_hw_res - allocate resource
+ * @hw: pointer to the HW struct
+ * @type: type of resource
+ * @num: number of resources to allocate
+ * @btm: allocate from bottom
+ * @res: pointer to array that will receive the resources
+ */
+enum ice_status
+ice_alloc_hw_res(struct ice_hw *hw, u16 type, u16 num, bool btm, u16 *res)
+{
+ struct ice_aqc_alloc_free_res_elem *buf;
+ enum ice_status status;
+ u16 buf_len;
+
+ buf_len = struct_size(buf, elem, num - 1);
+ buf = kzalloc(buf_len, GFP_KERNEL);
+ if (!buf)
+ return ICE_ERR_NO_MEMORY;
+
+ /* Prepare buffer to allocate resource. */
+ buf->num_elems = cpu_to_le16(num);
+ buf->res_type = cpu_to_le16(type | ICE_AQC_RES_TYPE_FLAG_DEDICATED |
+ ICE_AQC_RES_TYPE_FLAG_IGNORE_INDEX);
+ if (btm)
+ buf->res_type |= cpu_to_le16(ICE_AQC_RES_TYPE_FLAG_SCAN_BOTTOM);
+
+ status = ice_aq_alloc_free_res(hw, 1, buf, buf_len,
+ ice_aqc_opc_alloc_res, NULL);
+ if (status)
+ goto ice_alloc_res_exit;
+
+ memcpy(res, buf->elem, sizeof(buf->elem) * num);
+
+ice_alloc_res_exit:
+ kfree(buf);
+ return status;
+}
+
/**
* ice_get_num_per_func - determine number of resources per PF
* @hw: pointer to the HW structure
diff --git a/drivers/net/ethernet/intel/ice/ice_common.h b/drivers/net/ethernet/intel/ice/ice_common.h
index b22aa561e253..81e8ac925b9c 100644
--- a/drivers/net/ethernet/intel/ice/ice_common.h
+++ b/drivers/net/ethernet/intel/ice/ice_common.h
@@ -34,10 +34,16 @@ enum ice_status
ice_acquire_res(struct ice_hw *hw, enum ice_aq_res_ids res,
enum ice_aq_res_access_type access, u32 timeout);
void ice_release_res(struct ice_hw *hw, enum ice_aq_res_ids res);
+enum ice_status
+ice_alloc_hw_res(struct ice_hw *hw, u16 type, u16 num, bool btm, u16 *res);
enum ice_status ice_init_nvm(struct ice_hw *hw);
enum ice_status
ice_read_sr_buf(struct ice_hw *hw, u16 offset, u16 *words, u16 *data);
enum ice_status
+ice_aq_alloc_free_res(struct ice_hw *hw, u16 num_entries,
+ struct ice_aqc_alloc_free_res_elem *buf, u16 buf_size,
+ enum ice_adminq_opc opc, struct ice_sq_cd *cd);
+enum ice_status
ice_sq_send_cmd(struct ice_hw *hw, struct ice_ctl_q_info *cq,
struct ice_aq_desc *desc, void *buf, u16 buf_size,
struct ice_sq_cd *cd);
diff --git a/drivers/net/ethernet/intel/ice/ice_flex_pipe.c b/drivers/net/ethernet/intel/ice/ice_flex_pipe.c
index f37eb536e7b7..5c09d3105b8b 100644
--- a/drivers/net/ethernet/intel/ice/ice_flex_pipe.c
+++ b/drivers/net/ethernet/intel/ice/ice_flex_pipe.c
@@ -1118,6 +1118,117 @@ ice_vsig_add_mv_vsi(struct ice_hw *hw, enum ice_block blk, u16 vsi, u16 vsig)
return 0;
}
+/**
+ * ice_find_prof_id - find profile ID for a given field vector
+ * @hw: pointer to the hardware structure
+ * @blk: HW block
+ * @fv: field vector to search for
+ * @prof_id: receives the profile ID
+ */
+static enum ice_status
+ice_find_prof_id(struct ice_hw *hw, enum ice_block blk,
+ struct ice_fv_word *fv, u8 *prof_id)
+{
+ struct ice_es *es = &hw->blk[blk].es;
+ u16 off, i;
+
+ for (i = 0; i < es->count; i++) {
+ off = i * es->fvw;
+
+ if (memcmp(&es->t[off], fv, es->fvw * sizeof(*fv)))
+ continue;
+
+ *prof_id = i;
+ return 0;
+ }
+
+ return ICE_ERR_DOES_NOT_EXIST;
+}
+
+/**
+ * ice_prof_id_rsrc_type - get profile ID resource type for a block type
+ * @blk: the block type
+ * @rsrc_type: pointer to variable to receive the resource type
+ */
+static bool ice_prof_id_rsrc_type(enum ice_block blk, u16 *rsrc_type)
+{
+ switch (blk) {
+ case ICE_BLK_RSS:
+ *rsrc_type = ICE_AQC_RES_TYPE_HASH_PROF_BLDR_PROFID;
+ break;
+ default:
+ return false;
+ }
+ return true;
+}
+
+/**
+ * ice_alloc_prof_id - allocate profile ID
+ * @hw: pointer to the HW struct
+ * @blk: the block to allocate the profile ID for
+ * @prof_id: pointer to variable to receive the profile ID
+ *
+ * This function allocates a new profile ID, which also corresponds to a Field
+ * Vector (Extraction Sequence) entry.
+ */
+static enum ice_status
+ice_alloc_prof_id(struct ice_hw *hw, enum ice_block blk, u8 *prof_id)
+{
+ enum ice_status status;
+ u16 res_type;
+ u16 get_prof;
+
+ if (!ice_prof_id_rsrc_type(blk, &res_type))
+ return ICE_ERR_PARAM;
+
+ status = ice_alloc_hw_res(hw, res_type, 1, false, &get_prof);
+ if (!status)
+ *prof_id = (u8)get_prof;
+
+ return status;
+}
+
+/**
+ * ice_prof_inc_ref - increment reference count for profile
+ * @hw: pointer to the HW struct
+ * @blk: the block from which to free the profile ID
+ * @prof_id: the profile ID for which to increment the reference count
+ */
+static enum ice_status
+ice_prof_inc_ref(struct ice_hw *hw, enum ice_block blk, u8 prof_id)
+{
+ if (prof_id > hw->blk[blk].es.count)
+ return ICE_ERR_PARAM;
+
+ hw->blk[blk].es.ref_count[prof_id]++;
+
+ return 0;
+}
+
+/**
+ * ice_write_es - write an extraction sequence to hardware
+ * @hw: pointer to the HW struct
+ * @blk: the block in which to write the extraction sequence
+ * @prof_id: the profile ID to write
+ * @fv: pointer to the extraction sequence to write - NULL to clear extraction
+ */
+static void
+ice_write_es(struct ice_hw *hw, enum ice_block blk, u8 prof_id,
+ struct ice_fv_word *fv)
+{
+ u16 off;
+
+ off = prof_id * hw->blk[blk].es.fvw;
+ if (!fv) {
+ memset(&hw->blk[blk].es.t[off], 0,
+ hw->blk[blk].es.fvw * sizeof(*fv));
+ hw->blk[blk].es.written[prof_id] = false;
+ } else {
+ memcpy(&hw->blk[blk].es.t[off], fv,
+ hw->blk[blk].es.fvw * sizeof(*fv));
+ }
+}
+
/* Block / table section IDs */
static const u32 ice_blk_sids[ICE_BLK_COUNT][ICE_SID_OFF_COUNT] = {
/* SWITCH */
@@ -1576,3 +1687,108 @@ enum ice_status ice_init_hw_tbls(struct ice_hw *hw)
ice_free_hw_tbls(hw);
return ICE_ERR_NO_MEMORY;
}
+
+/**
+ * ice_add_prof - add profile
+ * @hw: pointer to the HW struct
+ * @blk: hardware block
+ * @id: profile tracking ID
+ * @ptypes: array of bitmaps indicating ptypes (ICE_FLOW_PTYPE_MAX bits)
+ * @es: extraction sequence (length of array is determined by the block)
+ *
+ * This function registers a profile, which matches a set of PTGs with a
+ * particular extraction sequence. While the hardware profile is allocated
+ * it will not be written until the first call to ice_add_flow that specifies
+ * the ID value used here.
+ */
+enum ice_status
+ice_add_prof(struct ice_hw *hw, enum ice_block blk, u64 id, u8 ptypes[],
+ struct ice_fv_word *es)
+{
+ u32 bytes = DIV_ROUND_UP(ICE_FLOW_PTYPE_MAX, BITS_PER_BYTE);
+ DECLARE_BITMAP(ptgs_used, ICE_XLT1_CNT);
+ struct ice_prof_map *prof;
+ enum ice_status status;
+ u32 byte = 0;
+ u8 prof_id;
+
+ bitmap_zero(ptgs_used, ICE_XLT1_CNT);
+
+ mutex_lock(&hw->blk[blk].es.prof_map_lock);
+
+ /* search for existing profile */
+ status = ice_find_prof_id(hw, blk, es, &prof_id);
+ if (status) {
+ /* allocate profile ID */
+ status = ice_alloc_prof_id(hw, blk, &prof_id);
+ if (status)
+ goto err_ice_add_prof;
+
+ /* and write new es */
+ ice_write_es(hw, blk, prof_id, es);
+ }
+
+ ice_prof_inc_ref(hw, blk, prof_id);
+
+ /* add profile info */
+ prof = devm_kzalloc(ice_hw_to_dev(hw), sizeof(*prof), GFP_KERNEL);
+ if (!prof)
+ goto err_ice_add_prof;
+
+ prof->profile_cookie = id;
+ prof->prof_id = prof_id;
+ prof->ptg_cnt = 0;
+ prof->context = 0;
+
+ /* build list of ptgs */
+ while (bytes && prof->ptg_cnt < ICE_MAX_PTG_PER_PROFILE) {
+ u32 bit;
+
+ if (!ptypes[byte]) {
+ bytes--;
+ byte++;
+ continue;
+ }
+
+ /* Examine 8 bits per byte */
+ for_each_set_bit(bit, (unsigned long *)&ptypes[byte],
+ BITS_PER_BYTE) {
+ u16 ptype;
+ u8 ptg;
+ u8 m;
+
+ ptype = byte * BITS_PER_BYTE + bit;
+
+ /* The package should place all ptypes in a non-zero
+ * PTG, so the following call should never fail.
+ */
+ if (ice_ptg_find_ptype(hw, blk, ptype, &ptg))
+ continue;
+
+ /* If PTG is already added, skip and continue */
+ if (test_bit(ptg, ptgs_used))
+ continue;
+
+ set_bit(ptg, ptgs_used);
+ prof->ptg[prof->ptg_cnt] = ptg;
+
+ if (++prof->ptg_cnt >= ICE_MAX_PTG_PER_PROFILE)
+ break;
+
+ /* nothing left in byte, then exit */
+ m = ~((1 << (bit + 1)) - 1);
+ if (!(ptypes[byte] & m))
+ break;
+ }
+
+ bytes--;
+ byte++;
+ }
+
+ list_add(&prof->list, &hw->blk[blk].es.prof_map);
+ status = 0;
+
+err_ice_add_prof:
+ mutex_unlock(&hw->blk[blk].es.prof_map_lock);
+ return status;
+}
diff --git a/drivers/net/ethernet/intel/ice/ice_flex_pipe.h b/drivers/net/ethernet/intel/ice/ice_flex_pipe.h
index 37eb282742d1..8cb7d7f09e0b 100644
--- a/drivers/net/ethernet/intel/ice/ice_flex_pipe.h
+++ b/drivers/net/ethernet/intel/ice/ice_flex_pipe.h
@@ -18,6 +18,9 @@
#define ICE_PKG_CNT 4
+enum ice_status
+ice_add_prof(struct ice_hw *hw, enum ice_block blk, u64 id, u8 ptypes[],
+ struct ice_fv_word *es);
enum ice_status ice_init_pkg(struct ice_hw *hw, u8 *buff, u32 len);
enum ice_status
ice_copy_and_init_pkg(struct ice_hw *hw, const u8 *buf, u32 len);
diff --git a/drivers/net/ethernet/intel/ice/ice_flex_type.h b/drivers/net/ethernet/intel/ice/ice_flex_type.h
index 5d5a7eaffa30..3005f111fb3b 100644
--- a/drivers/net/ethernet/intel/ice/ice_flex_type.h
+++ b/drivers/net/ethernet/intel/ice/ice_flex_type.h
@@ -3,6 +3,9 @@
#ifndef _ICE_FLEX_TYPE_H_
#define _ICE_FLEX_TYPE_H_
+
+#define ICE_FV_OFFSET_INVAL 0x1FF
+
/* Extraction Sequence (Field Vector) Table */
struct ice_fv_word {
u8 prot_id;
@@ -280,6 +283,17 @@ struct ice_ptg_ptype {
u8 ptg;
};
+#define ICE_MAX_PTG_PER_PROFILE 32
+
+struct ice_prof_map {
+ struct list_head list;
+ u64 profile_cookie;
+ u64 context;
+ u8 prof_id;
+ u8 ptg_cnt;
+ u8 ptg[ICE_MAX_PTG_PER_PROFILE];
+};
+
struct ice_vsig_entry {
struct list_head prop_lst;
struct ice_vsig_vsi *first_vsi;
@@ -371,4 +385,5 @@ struct ice_blk_info {
u8 is_list_init;
};
+#define ICE_FLOW_PTYPE_MAX ICE_XLT1_CNT
#endif /* _ICE_FLEX_TYPE_H_ */
diff --git a/drivers/net/ethernet/intel/ice/ice_flow.c b/drivers/net/ethernet/intel/ice/ice_flow.c
index 4828c64abe98..24fe04f8baa2 100644
--- a/drivers/net/ethernet/intel/ice/ice_flow.c
+++ b/drivers/net/ethernet/intel/ice/ice_flow.c
@@ -40,6 +40,457 @@ struct ice_flow_field_info ice_flds_info[ICE_FLOW_FIELD_IDX_MAX] = {
ICE_FLOW_FLD_INFO(ICE_FLOW_SEG_HDR_UDP, 2, sizeof(__be16)),
};
+/* Bitmaps indicating relevant packet types for a particular protocol header
+ *
+ * Packet types for packets with an Outer/First/Single IPv4 header
+ */
+static const u32 ice_ptypes_ipv4_ofos[] = {
+ 0x1DC00000, 0x04000800, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+};
+
+/* Packet types for packets with an Innermost/Last IPv4 header */
+static const u32 ice_ptypes_ipv4_il[] = {
+ 0xE0000000, 0xB807700E, 0x80000003, 0xE01DC03B,
+ 0x0000000E, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+};
+
+/* Packet types for packets with an Outer/First/Single IPv6 header */
+static const u32 ice_ptypes_ipv6_ofos[] = {
+ 0x00000000, 0x00000000, 0x77000000, 0x10002000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+};
+
+/* Packet types for packets with an Innermost/Last IPv6 header */
+static const u32 ice_ptypes_ipv6_il[] = {
+ 0x00000000, 0x03B80770, 0x000001DC, 0x0EE00000,
+ 0x00000770, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+};
+
+/* UDP Packet types for non-tunneled packets or tunneled
+ * packets with inner UDP.
+ */
+static const u32 ice_ptypes_udp_il[] = {
+ 0x81000000, 0x20204040, 0x04000010, 0x80810102,
+ 0x00000040, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+};
+
+/* Packet types for packets with an Innermost/Last TCP header */
+static const u32 ice_ptypes_tcp_il[] = {
+ 0x04000000, 0x80810102, 0x10000040, 0x02040408,
+ 0x00000102, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+};
+
+/* Packet types for packets with an Innermost/Last SCTP header */
+static const u32 ice_ptypes_sctp_il[] = {
+ 0x08000000, 0x01020204, 0x20000081, 0x04080810,
+ 0x00000204, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+};
+
+/* Manage parameters and info. used during the creation of a flow profile */
+struct ice_flow_prof_params {
+ enum ice_block blk;
+ u16 entry_length; /* # of bytes formatted entry will require */
+ u8 es_cnt;
+ struct ice_flow_prof *prof;
+
+ /* For ACL, the es[0] will have the data of ICE_RX_MDID_PKT_FLAGS_15_0
+ * This will give us the direction flags.
+ */
+ struct ice_fv_word es[ICE_MAX_FV_WORDS];
+ DECLARE_BITMAP(ptypes, ICE_FLOW_PTYPE_MAX);
+};
+
+#define ICE_FLOW_SEG_HDRS_L3_MASK \
+ (ICE_FLOW_SEG_HDR_IPV4 | ICE_FLOW_SEG_HDR_IPV6)
+#define ICE_FLOW_SEG_HDRS_L4_MASK \
+ (ICE_FLOW_SEG_HDR_TCP | ICE_FLOW_SEG_HDR_UDP | ICE_FLOW_SEG_HDR_SCTP)
+
+/**
+ * ice_flow_val_hdrs - validates packet segments for valid protocol headers
+ * @segs: array of one or more packet segments that describe the flow
+ * @segs_cnt: number of packet segments provided
+ */
+static enum ice_status
+ice_flow_val_hdrs(struct ice_flow_seg_info *segs, u8 segs_cnt)
+{
+ u8 i;
+
+ for (i = 0; i < segs_cnt; i++) {
+ /* Multiple L3 headers */
+ if (segs[i].hdrs & ICE_FLOW_SEG_HDRS_L3_MASK &&
+ !is_power_of_2(segs[i].hdrs & ICE_FLOW_SEG_HDRS_L3_MASK))
+ return ICE_ERR_PARAM;
+
+ /* Multiple L4 headers */
+ if (segs[i].hdrs & ICE_FLOW_SEG_HDRS_L4_MASK &&
+ !is_power_of_2(segs[i].hdrs & ICE_FLOW_SEG_HDRS_L4_MASK))
+ return ICE_ERR_PARAM;
+ }
+
+ return 0;
+}
+
+/**
+ * ice_flow_proc_seg_hdrs - process protocol headers present in pkt segments
+ * @params: information about the flow to be processed
+ *
+ * This function identifies the packet types associated with the protocol
+ * headers being present in packet segments of the specified flow profile.
+ */
+static enum ice_status
+ice_flow_proc_seg_hdrs(struct ice_flow_prof_params *params)
+{
+ struct ice_flow_prof *prof;
+ u8 i;
+
+ memset(params->ptypes, 0xff, sizeof(params->ptypes));
+
+ prof = params->prof;
+
+ for (i = 0; i < params->prof->segs_cnt; i++) {
+ const unsigned long *src;
+ u32 hdrs;
+
+ hdrs = prof->segs[i].hdrs;
+
+ if (hdrs & ICE_FLOW_SEG_HDR_IPV4) {
+ src = !i ? (const unsigned long *)ice_ptypes_ipv4_ofos :
+ (const unsigned long *)ice_ptypes_ipv4_il;
+ bitmap_and(params->ptypes, params->ptypes, src,
+ ICE_FLOW_PTYPE_MAX);
+ } else if (hdrs & ICE_FLOW_SEG_HDR_IPV6) {
+ src = !i ? (const unsigned long *)ice_ptypes_ipv6_ofos :
+ (const unsigned long *)ice_ptypes_ipv6_il;
+ bitmap_and(params->ptypes, params->ptypes, src,
+ ICE_FLOW_PTYPE_MAX);
+ }
+
+ if (hdrs & ICE_FLOW_SEG_HDR_UDP) {
+ src = (const unsigned long *)ice_ptypes_udp_il;
+ bitmap_and(params->ptypes, params->ptypes, src,
+ ICE_FLOW_PTYPE_MAX);
+ } else if (hdrs & ICE_FLOW_SEG_HDR_TCP) {
+ bitmap_and(params->ptypes, params->ptypes,
+ (const unsigned long *)ice_ptypes_tcp_il,
+ ICE_FLOW_PTYPE_MAX);
+ } else if (hdrs & ICE_FLOW_SEG_HDR_SCTP) {
+ src = (const unsigned long *)ice_ptypes_sctp_il;
+ bitmap_and(params->ptypes, params->ptypes, src,
+ ICE_FLOW_PTYPE_MAX);
+ }
+ }
+
+ return 0;
+}
+
+/**
+ * ice_flow_xtract_fld - Create an extraction sequence entry for the given field
+ * @hw: pointer to the HW struct
+ * @params: information about the flow to be processed
+ * @seg: packet segment index of the field to be extracted
+ * @fld: ID of field to be extracted
+ *
+ * This function determines the protocol ID, offset, and size of the given
+ * field. It then allocates one or more extraction sequence entries for the
+ * given field, and fill the entries with protocol ID and offset information.
+ */
+static enum ice_status
+ice_flow_xtract_fld(struct ice_hw *hw, struct ice_flow_prof_params *params,
+ u8 seg, enum ice_flow_field fld)
+{
+ enum ice_prot_id prot_id = ICE_PROT_ID_INVAL;
+ u8 fv_words = hw->blk[params->blk].es.fvw;
+ struct ice_flow_fld_info *flds;
+ u16 cnt, ese_bits, i;
+ u16 off;
+
+ flds = params->prof->segs[seg].fields;
+
+ switch (fld) {
+ case ICE_FLOW_FIELD_IDX_IPV4_SA:
+ case ICE_FLOW_FIELD_IDX_IPV4_DA:
+ prot_id = seg == 0 ? ICE_PROT_IPV4_OF_OR_S : ICE_PROT_IPV4_IL;
+ break;
+ case ICE_FLOW_FIELD_IDX_IPV6_SA:
+ case ICE_FLOW_FIELD_IDX_IPV6_DA:
+ prot_id = seg == 0 ? ICE_PROT_IPV6_OF_OR_S : ICE_PROT_IPV6_IL;
+ break;
+ case ICE_FLOW_FIELD_IDX_TCP_SRC_PORT:
+ case ICE_FLOW_FIELD_IDX_TCP_DST_PORT:
+ prot_id = ICE_PROT_TCP_IL;
+ break;
+ case ICE_FLOW_FIELD_IDX_UDP_SRC_PORT:
+ case ICE_FLOW_FIELD_IDX_UDP_DST_PORT:
+ prot_id = ICE_PROT_UDP_IL_OR_S;
+ break;
+ default:
+ return ICE_ERR_NOT_IMPL;
+ }
+
+ /* Each extraction sequence entry is a word in size, and extracts a
+ * word-aligned offset from a protocol header.
+ */
+ ese_bits = ICE_FLOW_FV_EXTRACT_SZ * BITS_PER_BYTE;
+
+ flds[fld].xtrct.prot_id = prot_id;
+ flds[fld].xtrct.off = (ice_flds_info[fld].off / ese_bits) *
+ ICE_FLOW_FV_EXTRACT_SZ;
+ flds[fld].xtrct.disp = (u8)(ice_flds_info[fld].off % ese_bits);
+ flds[fld].xtrct.idx = params->es_cnt;
+
+ /* Adjust the next field-entry index after accommodating the number of
+ * entries this field consumes
+ */
+ cnt = DIV_ROUND_UP(flds[fld].xtrct.disp + ice_flds_info[fld].size,
+ ese_bits);
+
+ /* Fill in the extraction sequence entries needed for this field */
+ off = flds[fld].xtrct.off;
+ for (i = 0; i < cnt; i++) {
+ u8 idx;
+
+ /* Make sure the number of extraction sequence required
+ * does not exceed the block's capability
+ */
+ if (params->es_cnt >= fv_words)
+ return ICE_ERR_MAX_LIMIT;
+
+ /* some blocks require a reversed field vector layout */
+ if (hw->blk[params->blk].es.reverse)
+ idx = fv_words - params->es_cnt - 1;
+ else
+ idx = params->es_cnt;
+
+ params->es[idx].prot_id = prot_id;
+ params->es[idx].off = off;
+ params->es_cnt++;
+
+ off += ICE_FLOW_FV_EXTRACT_SZ;
+ }
+
+ return 0;
+}
+
+/**
+ * ice_flow_create_xtrct_seq - Create an extraction sequence for given segments
+ * @hw: pointer to the HW struct
+ * @params: information about the flow to be processed
+ *
+ * This function iterates through all matched fields in the given segments, and
+ * creates an extraction sequence for the fields.
+ */
+static enum ice_status
+ice_flow_create_xtrct_seq(struct ice_hw *hw,
+ struct ice_flow_prof_params *params)
+{
+ struct ice_flow_prof *prof = params->prof;
+ enum ice_status status = 0;
+ u8 i;
+
+ for (i = 0; i < prof->segs_cnt; i++) {
+ u8 j;
+
+ for_each_set_bit(j, (unsigned long *)&prof->segs[i].match,
+ ICE_FLOW_FIELD_IDX_MAX) {
+ status = ice_flow_xtract_fld(hw, params, i,
+ (enum ice_flow_field)j);
+ if (status)
+ return status;
+ }
+ }
+
+ return status;
+}
+
+/**
+ * ice_flow_proc_segs - process all packet segments associated with a profile
+ * @hw: pointer to the HW struct
+ * @params: information about the flow to be processed
+ */
+static enum ice_status
+ice_flow_proc_segs(struct ice_hw *hw, struct ice_flow_prof_params *params)
+{
+ enum ice_status status;
+
+ status = ice_flow_proc_seg_hdrs(params);
+ if (status)
+ return status;
+
+ status = ice_flow_create_xtrct_seq(hw, params);
+ if (status)
+ return status;
+
+ switch (params->blk) {
+ case ICE_BLK_RSS:
+ /* Only header information is provided for RSS configuration.
+ * No further processing is needed.
+ */
+ status = 0;
+ break;
+ default:
+ return ICE_ERR_NOT_IMPL;
+ }
+
+ return status;
+}
+
+/**
+ * ice_flow_add_prof_sync - Add a flow profile for packet segments and fields
+ * @hw: pointer to the HW struct
+ * @blk: classification stage
+ * @dir: flow direction
+ * @prof_id: unique ID to identify this flow profile
+ * @segs: array of one or more packet segments that describe the flow
+ * @segs_cnt: number of packet segments provided
+ * @prof: stores the returned flow profile added
+ *
+ * Assumption: the caller has acquired the lock to the profile list
+ */
+static enum ice_status
+ice_flow_add_prof_sync(struct ice_hw *hw, enum ice_block blk,
+ enum ice_flow_dir dir, u64 prof_id,
+ struct ice_flow_seg_info *segs, u8 segs_cnt,
+ struct ice_flow_prof **prof)
+{
+ struct ice_flow_prof_params params;
+ enum ice_status status;
+ u8 i;
+
+ if (!prof)
+ return ICE_ERR_BAD_PTR;
+
+ memset(¶ms, 0, sizeof(params));
+ params.prof = devm_kzalloc(ice_hw_to_dev(hw), sizeof(*params.prof),
+ GFP_KERNEL);
+ if (!params.prof)
+ return ICE_ERR_NO_MEMORY;
+
+ /* initialize extraction sequence to all invalid (0xff) */
+ for (i = 0; i < ICE_MAX_FV_WORDS; i++) {
+ params.es[i].prot_id = ICE_PROT_INVALID;
+ params.es[i].off = ICE_FV_OFFSET_INVAL;
+ }
+
+ params.blk = blk;
+ params.prof->id = prof_id;
+ params.prof->dir = dir;
+ params.prof->segs_cnt = segs_cnt;
+
+ /* Make a copy of the segments that need to be persistent in the flow
+ * profile instance
+ */
+ for (i = 0; i < segs_cnt; i++)
+ memcpy(¶ms.prof->segs[i], &segs[i], sizeof(*segs));
+
+ status = ice_flow_proc_segs(hw, ¶ms);
+ if (status) {
+ ice_debug(hw, ICE_DBG_FLOW,
+ "Error processing a flow's packet segments\n");
+ goto out;
+ }
+
+ /* Add a HW profile for this flow profile */
+ status = ice_add_prof(hw, blk, prof_id, (u8 *)params.ptypes, params.es);
+ if (status) {
+ ice_debug(hw, ICE_DBG_FLOW, "Error adding a HW flow profile\n");
+ goto out;
+ }
+
+ INIT_LIST_HEAD(¶ms.prof->entries);
+ mutex_init(¶ms.prof->entries_lock);
+ *prof = params.prof;
+
+out:
+ if (status)
+ devm_kfree(ice_hw_to_dev(hw), params.prof);
+
+ return status;
+}
+
+/**
+ * ice_flow_add_prof - Add a flow profile for packet segments and matched fields
+ * @hw: pointer to the HW struct
+ * @blk: classification stage
+ * @dir: flow direction
+ * @prof_id: unique ID to identify this flow profile
+ * @segs: array of one or more packet segments that describe the flow
+ * @segs_cnt: number of packet segments provided
+ */
+static enum ice_status
+ice_flow_add_prof(struct ice_hw *hw, enum ice_block blk, enum ice_flow_dir dir,
+ u64 prof_id, struct ice_flow_seg_info *segs, u8 segs_cnt)
+{
+ struct ice_flow_prof *prof = NULL;
+ enum ice_status status;
+
+ if (segs_cnt > ICE_FLOW_SEG_MAX)
+ return ICE_ERR_MAX_LIMIT;
+
+ if (!segs_cnt)
+ return ICE_ERR_PARAM;
+
+ if (!segs)
+ return ICE_ERR_BAD_PTR;
+
+ status = ice_flow_val_hdrs(segs, segs_cnt);
+ if (status)
+ return status;
+
+ mutex_lock(&hw->fl_profs_locks[blk]);
+
+ status = ice_flow_add_prof_sync(hw, blk, dir, prof_id, segs, segs_cnt,
+ &prof);
+ if (!status)
+ list_add(&prof->l_entry, &hw->fl_profs[blk]);
+
+ mutex_unlock(&hw->fl_profs_locks[blk]);
+
+ return status;
+}
+
/**
* ice_flow_set_fld_ext - specifies locations of field from entry's input buffer
* @seg: packet segment the field being set belongs to
@@ -161,10 +612,28 @@ ice_flow_set_rss_seg_info(struct ice_flow_seg_info *segs, u64 hash_fields,
return 0;
}
+#define ICE_FLOW_PROF_HASH_S 0
+#define ICE_FLOW_PROF_HASH_M (0xFFFFFFFFULL << ICE_FLOW_PROF_HASH_S)
+#define ICE_FLOW_PROF_HDR_S 32
+#define ICE_FLOW_PROF_HDR_M (0x3FFFFFFFULL << ICE_FLOW_PROF_HDR_S)
+#define ICE_FLOW_PROF_ENCAP_S 63
+#define ICE_FLOW_PROF_ENCAP_M (BIT_ULL(ICE_FLOW_PROF_ENCAP_S))
+
#define ICE_RSS_OUTER_HEADERS 1
+/* Flow profile ID format:
+ * [0:31] - Packet match fields
+ * [32:62] - Protocol header
+ * [63] - Encapsulation flag, 0 if non-tunneled, 1 if tunneled
+ */
+#define ICE_FLOW_GEN_PROFID(hash, hdr, segs_cnt) \
+ (u64)(((u64)(hash) & ICE_FLOW_PROF_HASH_M) | \
+ (((u64)(hdr) << ICE_FLOW_PROF_HDR_S) & ICE_FLOW_PROF_HDR_M) | \
+ ((u8)((segs_cnt) - 1) ? ICE_FLOW_PROF_ENCAP_M : 0))
+
/**
* ice_add_rss_cfg_sync - add an RSS configuration
+ * @hw: pointer to the hardware structure
* @hashed_flds: hash bit fields (ICE_FLOW_HASH_*) to configure
* @addl_hdrs: protocol header fields
* @segs_cnt: packet segment count
@@ -172,7 +641,8 @@ ice_flow_set_rss_seg_info(struct ice_flow_seg_info *segs, u64 hash_fields,
* Assumption: lock has already been acquired for RSS list
*/
static enum ice_status
-ice_add_rss_cfg_sync(u64 hashed_flds, u32 addl_hdrs, u8 segs_cnt)
+ice_add_rss_cfg_sync(struct ice_hw *hw, u64 hashed_flds, u32 addl_hdrs,
+ u8 segs_cnt)
{
struct ice_flow_seg_info *segs;
enum ice_status status;
@@ -187,7 +657,19 @@ ice_add_rss_cfg_sync(u64 hashed_flds, u32 addl_hdrs, u8 segs_cnt)
/* Construct the packet segment info from the hashed fields */
status = ice_flow_set_rss_seg_info(&segs[segs_cnt - 1], hashed_flds,
addl_hdrs);
-
+ if (status)
+ goto exit;
+
+ /* Create a new flow profile with generated profile and packet
+ * segment information.
+ */
+ status = ice_flow_add_prof(hw, ICE_BLK_RSS, ICE_FLOW_RX,
+ ICE_FLOW_GEN_PROFID(hashed_flds,
+ segs[segs_cnt - 1].hdrs,
+ segs_cnt),
+ segs, segs_cnt);
+
+exit:
kfree(segs);
return status;
}
@@ -214,7 +696,7 @@ ice_add_rss_cfg(struct ice_hw *hw, u16 vsi_handle, u64 hashed_flds,
return ICE_ERR_PARAM;
mutex_lock(&hw->rss_locks);
- status = ice_add_rss_cfg_sync(hashed_flds, addl_hdrs,
+ status = ice_add_rss_cfg_sync(hw, hashed_flds, addl_hdrs,
ICE_RSS_OUTER_HEADERS);
mutex_unlock(&hw->rss_locks);
@@ -237,7 +719,7 @@ enum ice_status ice_replay_rss_cfg(struct ice_hw *hw, u16 vsi_handle)
mutex_lock(&hw->rss_locks);
list_for_each_entry(r, &hw->rss_list_head, l_entry) {
if (test_bit(vsi_handle, r->vsis)) {
- status = ice_add_rss_cfg_sync(r->hashed_flds,
+ status = ice_add_rss_cfg_sync(hw, r->hashed_flds,
r->packet_hdr,
ICE_RSS_OUTER_HEADERS);
if (status)
diff --git a/drivers/net/ethernet/intel/ice/ice_flow.h b/drivers/net/ethernet/intel/ice/ice_flow.h
index 48c0fc09d5ff..05b0dab4793c 100644
--- a/drivers/net/ethernet/intel/ice/ice_flow.h
+++ b/drivers/net/ethernet/intel/ice/ice_flow.h
@@ -56,7 +56,13 @@ enum ice_flow_field {
ICE_FLOW_FIELD_IDX_MAX
};
+enum ice_flow_dir {
+ ICE_FLOW_RX = 0x02,
+};
+
#define ICE_FLOW_SEG_MAX 2
+#define ICE_FLOW_FV_EXTRACT_SZ 2
+
#define ICE_FLOW_SET_HDRS(seg, val) ((seg)->hdrs |= (u32)(val))
struct ice_flow_seg_xtrct {
@@ -99,6 +105,23 @@ struct ice_flow_seg_info {
struct ice_flow_fld_info fields[ICE_FLOW_FIELD_IDX_MAX];
};
+struct ice_flow_prof {
+ struct list_head l_entry;
+
+ u64 id;
+ enum ice_flow_dir dir;
+ u8 segs_cnt;
+
+ /* Keep track of flow entries associated with this flow profile */
+ struct mutex entries_lock;
+ struct list_head entries;
+
+ struct ice_flow_seg_info segs[ICE_FLOW_SEG_MAX];
+
+ /* software VSI handles referenced by this flow profile */
+ DECLARE_BITMAP(vsis, ICE_MAX_VSI);
+};
+
struct ice_rss_cfg {
struct list_head l_entry;
/* bitmap of VSIs added to the RSS entry */
diff --git a/drivers/net/ethernet/intel/ice/ice_lan_tx_rx.h b/drivers/net/ethernet/intel/ice/ice_lan_tx_rx.h
index 0997d352709b..878e125d8b42 100644
--- a/drivers/net/ethernet/intel/ice/ice_lan_tx_rx.h
+++ b/drivers/net/ethernet/intel/ice/ice_lan_tx_rx.h
@@ -199,6 +199,14 @@ enum ice_rxdid {
/* Receive Flex Descriptor Rx opcode values */
#define ICE_RX_OPC_MDID 0x01
+/* Receive Descriptor MDID values that access packet flags */
+enum ice_flex_mdid_pkt_flags {
+ ICE_RX_MDID_PKT_FLAGS_15_0 = 20,
+ ICE_RX_MDID_PKT_FLAGS_31_16,
+ ICE_RX_MDID_PKT_FLAGS_47_32,
+ ICE_RX_MDID_PKT_FLAGS_63_48,
+};
+
/* Receive Descriptor MDID values */
enum ice_flex_rx_mdid {
ICE_RX_MDID_FLOW_ID_LOWER = 5,
diff --git a/drivers/net/ethernet/intel/ice/ice_protocol_type.h b/drivers/net/ethernet/intel/ice/ice_protocol_type.h
new file mode 100644
index 000000000000..97db46ce7a21
--- /dev/null
+++ b/drivers/net/ethernet/intel/ice/ice_protocol_type.h
@@ -0,0 +1,24 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/* Copyright (c) 2019, Intel Corporation. */
+
+#ifndef _ICE_PROTOCOL_TYPE_H_
+#define _ICE_PROTOCOL_TYPE_H_
+/* Decoders for ice_prot_id:
+ * - F: First
+ * - I: Inner
+ * - L: Last
+ * - O: Outer
+ * - S: Single
+ */
+enum ice_prot_id {
+ ICE_PROT_ID_INVAL = 0,
+ ICE_PROT_IPV4_OF_OR_S = 32,
+ ICE_PROT_IPV4_IL = 33,
+ ICE_PROT_IPV6_OF_OR_S = 40,
+ ICE_PROT_IPV6_IL = 41,
+ ICE_PROT_TCP_IL = 49,
+ ICE_PROT_UDP_IL_OR_S = 53,
+ ICE_PROT_META_ID = 255, /* when offset == metadata */
+ ICE_PROT_INVALID = 255 /* when offset == ICE_FV_OFFSET_INVAL */
+};
+#endif /* _ICE_PROTOCOL_TYPE_H_ */
diff --git a/drivers/net/ethernet/intel/ice/ice_switch.c b/drivers/net/ethernet/intel/ice/ice_switch.c
index b5a53f862a83..431266081a80 100644
--- a/drivers/net/ethernet/intel/ice/ice_switch.c
+++ b/drivers/net/ethernet/intel/ice/ice_switch.c
@@ -49,42 +49,6 @@ static const u8 dummy_eth_header[DUMMY_ETH_HDR_LEN] = { 0x2, 0, 0, 0, 0, 0,
sizeof(((struct ice_sw_rule_vsi_list *)0)->vsi) + \
((n) * sizeof(((struct ice_sw_rule_vsi_list *)0)->vsi)))
-/**
- * ice_aq_alloc_free_res - command to allocate/free resources
- * @hw: pointer to the HW struct
- * @num_entries: number of resource entries in buffer
- * @buf: Indirect buffer to hold data parameters and response
- * @buf_size: size of buffer for indirect commands
- * @opc: pass in the command opcode
- * @cd: pointer to command details structure or NULL
- *
- * Helper function to allocate/free resources using the admin queue commands
- */
-static enum ice_status
-ice_aq_alloc_free_res(struct ice_hw *hw, u16 num_entries,
- struct ice_aqc_alloc_free_res_elem *buf, u16 buf_size,
- enum ice_adminq_opc opc, struct ice_sq_cd *cd)
-{
- struct ice_aqc_alloc_free_res_cmd *cmd;
- struct ice_aq_desc desc;
-
- cmd = &desc.params.sw_res_ctrl;
-
- if (!buf)
- return ICE_ERR_PARAM;
-
- if (buf_size < (num_entries * sizeof(buf->elem[0])))
- return ICE_ERR_PARAM;
-
- ice_fill_dflt_direct_cmd_desc(&desc, opc);
-
- desc.flags |= cpu_to_le16(ICE_AQ_FLAG_RD);
-
- cmd->num_entries = cpu_to_le16(num_entries);
-
- return ice_aq_send_cmd(hw, &desc, buf, buf_size, cd);
-}
-
/**
* ice_init_def_sw_recp - initialize the recipe book keeping tables
* @hw: pointer to the HW struct
diff --git a/drivers/net/ethernet/intel/ice/ice_type.h b/drivers/net/ethernet/intel/ice/ice_type.h
index ac38af105f90..b361ffabb0ca 100644
--- a/drivers/net/ethernet/intel/ice/ice_type.h
+++ b/drivers/net/ethernet/intel/ice/ice_type.h
@@ -13,6 +13,7 @@
#include "ice_controlq.h"
#include "ice_lan_tx_rx.h"
#include "ice_flex_type.h"
+#include "ice_protocol_type.h"
static inline bool ice_is_tc_ena(unsigned long bitmap, u8 tc)
{
@@ -41,6 +42,7 @@ static inline u32 ice_round_to_num(u32 N, u32 R)
#define ICE_DBG_QCTX BIT_ULL(6)
#define ICE_DBG_NVM BIT_ULL(7)
#define ICE_DBG_LAN BIT_ULL(8)
+#define ICE_DBG_FLOW BIT_ULL(9)
#define ICE_DBG_SW BIT_ULL(13)
#define ICE_DBG_SCHED BIT_ULL(14)
#define ICE_DBG_PKG BIT_ULL(16)
--
2.24.1
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [net-next 3/8] ice: Populate TCAM filter software structures
2020-01-26 6:07 [net-next 0/8][pull request] 100GbE Intel Wired LAN Driver Updates 2020-01-25 Jeff Kirsher
2020-01-26 6:07 ` [net-next 1/8] ice: Enable writing hardware filtering tables Jeff Kirsher
2020-01-26 6:07 ` [net-next 2/8] ice: Allocate flow profile Jeff Kirsher
@ 2020-01-26 6:07 ` Jeff Kirsher
2020-01-26 6:07 ` [net-next 4/8] ice: Enable writing filtering tables Jeff Kirsher
` (5 subsequent siblings)
8 siblings, 0 replies; 10+ messages in thread
From: Jeff Kirsher @ 2020-01-26 6:07 UTC (permalink / raw)
To: davem
Cc: Tony Nguyen, netdev, nhorman, sassmann, Henry Tieman,
Andrew Bowers, Jeff Kirsher
From: Tony Nguyen <anthony.l.nguyen@intel.com>
Store the TCAM entry with the profile data and the VSI group in the
respective SW structures. This will be subsequently used to write out
the tables to hardware.
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Henry Tieman <henry.w.tieman@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
.../net/ethernet/intel/ice/ice_adminq_cmd.h | 1 +
drivers/net/ethernet/intel/ice/ice_common.c | 33 +
drivers/net/ethernet/intel/ice/ice_common.h | 2 +
.../net/ethernet/intel/ice/ice_flex_pipe.c | 1318 +++++++++++++++++
.../net/ethernet/intel/ice/ice_flex_pipe.h | 2 +
.../net/ethernet/intel/ice/ice_flex_type.h | 51 +
drivers/net/ethernet/intel/ice/ice_flow.c | 61 +-
drivers/net/ethernet/intel/ice/ice_status.h | 1 +
8 files changed, 1459 insertions(+), 10 deletions(-)
diff --git a/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h b/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h
index ad15a772b609..9967973a7c42 100644
--- a/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h
+++ b/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h
@@ -233,6 +233,7 @@ struct ice_aqc_get_sw_cfg_resp {
#define ICE_AQC_RES_TYPE_VSI_LIST_REP 0x03
#define ICE_AQC_RES_TYPE_VSI_LIST_PRUNE 0x04
#define ICE_AQC_RES_TYPE_HASH_PROF_BLDR_PROFID 0x60
+#define ICE_AQC_RES_TYPE_HASH_PROF_BLDR_TCAM 0x61
#define ICE_AQC_RES_TYPE_FLAG_SCAN_BOTTOM BIT(12)
#define ICE_AQC_RES_TYPE_FLAG_IGNORE_INDEX BIT(13)
diff --git a/drivers/net/ethernet/intel/ice/ice_common.c b/drivers/net/ethernet/intel/ice/ice_common.c
index 8f86962fd052..0207e28c2682 100644
--- a/drivers/net/ethernet/intel/ice/ice_common.c
+++ b/drivers/net/ethernet/intel/ice/ice_common.c
@@ -1572,6 +1572,39 @@ ice_alloc_hw_res(struct ice_hw *hw, u16 type, u16 num, bool btm, u16 *res)
return status;
}
+/**
+ * ice_free_hw_res - free allocated HW resource
+ * @hw: pointer to the HW struct
+ * @type: type of resource to free
+ * @num: number of resources
+ * @res: pointer to array that contains the resources to free
+ */
+enum ice_status
+ice_free_hw_res(struct ice_hw *hw, u16 type, u16 num, u16 *res)
+{
+ struct ice_aqc_alloc_free_res_elem *buf;
+ enum ice_status status;
+ u16 buf_len;
+
+ buf_len = struct_size(buf, elem, num - 1);
+ buf = kzalloc(buf_len, GFP_KERNEL);
+ if (!buf)
+ return ICE_ERR_NO_MEMORY;
+
+ /* Prepare buffer to free resource. */
+ buf->num_elems = cpu_to_le16(num);
+ buf->res_type = cpu_to_le16(type);
+ memcpy(buf->elem, res, sizeof(buf->elem) * num);
+
+ status = ice_aq_alloc_free_res(hw, num, buf, buf_len,
+ ice_aqc_opc_free_res, NULL);
+ if (status)
+ ice_debug(hw, ICE_DBG_SW, "CQ CMD Buffer:\n");
+
+ kfree(buf);
+ return status;
+}
+
/**
* ice_get_num_per_func - determine number of resources per PF
* @hw: pointer to the HW structure
diff --git a/drivers/net/ethernet/intel/ice/ice_common.h b/drivers/net/ethernet/intel/ice/ice_common.h
index 81e8ac925b9c..b5c013fdaaf9 100644
--- a/drivers/net/ethernet/intel/ice/ice_common.h
+++ b/drivers/net/ethernet/intel/ice/ice_common.h
@@ -36,6 +36,8 @@ ice_acquire_res(struct ice_hw *hw, enum ice_aq_res_ids res,
void ice_release_res(struct ice_hw *hw, enum ice_aq_res_ids res);
enum ice_status
ice_alloc_hw_res(struct ice_hw *hw, u16 type, u16 num, bool btm, u16 *res);
+enum ice_status
+ice_free_hw_res(struct ice_hw *hw, u16 type, u16 num, u16 *res);
enum ice_status ice_init_nvm(struct ice_hw *hw);
enum ice_status
ice_read_sr_buf(struct ice_hw *hw, u16 offset, u16 *words, u16 *data);
diff --git a/drivers/net/ethernet/intel/ice/ice_flex_pipe.c b/drivers/net/ethernet/intel/ice/ice_flex_pipe.c
index 5c09d3105b8b..8c93c303a4a5 100644
--- a/drivers/net/ethernet/intel/ice/ice_flex_pipe.c
+++ b/drivers/net/ethernet/intel/ice/ice_flex_pipe.c
@@ -159,6 +159,176 @@ ice_pkg_enum_section(struct ice_seg *ice_seg, struct ice_pkg_enum *state,
return state->sect;
}
+/* Key creation */
+
+#define ICE_DC_KEY 0x1 /* don't care */
+#define ICE_DC_KEYINV 0x1
+#define ICE_NM_KEY 0x0 /* never match */
+#define ICE_NM_KEYINV 0x0
+#define ICE_0_KEY 0x1 /* match 0 */
+#define ICE_0_KEYINV 0x0
+#define ICE_1_KEY 0x0 /* match 1 */
+#define ICE_1_KEYINV 0x1
+
+/**
+ * ice_gen_key_word - generate 16-bits of a key/mask word
+ * @val: the value
+ * @valid: valid bits mask (change only the valid bits)
+ * @dont_care: don't care mask
+ * @nvr_mtch: never match mask
+ * @key: pointer to an array of where the resulting key portion
+ * @key_inv: pointer to an array of where the resulting key invert portion
+ *
+ * This function generates 16-bits from a 8-bit value, an 8-bit don't care mask
+ * and an 8-bit never match mask. The 16-bits of output are divided into 8 bits
+ * of key and 8 bits of key invert.
+ *
+ * '0' = b01, always match a 0 bit
+ * '1' = b10, always match a 1 bit
+ * '?' = b11, don't care bit (always matches)
+ * '~' = b00, never match bit
+ *
+ * Input:
+ * val: b0 1 0 1 0 1
+ * dont_care: b0 0 1 1 0 0
+ * never_mtch: b0 0 0 0 1 1
+ * ------------------------------
+ * Result: key: b01 10 11 11 00 00
+ */
+static enum ice_status
+ice_gen_key_word(u8 val, u8 valid, u8 dont_care, u8 nvr_mtch, u8 *key,
+ u8 *key_inv)
+{
+ u8 in_key = *key, in_key_inv = *key_inv;
+ u8 i;
+
+ /* 'dont_care' and 'nvr_mtch' masks cannot overlap */
+ if ((dont_care ^ nvr_mtch) != (dont_care | nvr_mtch))
+ return ICE_ERR_CFG;
+
+ *key = 0;
+ *key_inv = 0;
+
+ /* encode the 8 bits into 8-bit key and 8-bit key invert */
+ for (i = 0; i < 8; i++) {
+ *key >>= 1;
+ *key_inv >>= 1;
+
+ if (!(valid & 0x1)) { /* change only valid bits */
+ *key |= (in_key & 0x1) << 7;
+ *key_inv |= (in_key_inv & 0x1) << 7;
+ } else if (dont_care & 0x1) { /* don't care bit */
+ *key |= ICE_DC_KEY << 7;
+ *key_inv |= ICE_DC_KEYINV << 7;
+ } else if (nvr_mtch & 0x1) { /* never match bit */
+ *key |= ICE_NM_KEY << 7;
+ *key_inv |= ICE_NM_KEYINV << 7;
+ } else if (val & 0x01) { /* exact 1 match */
+ *key |= ICE_1_KEY << 7;
+ *key_inv |= ICE_1_KEYINV << 7;
+ } else { /* exact 0 match */
+ *key |= ICE_0_KEY << 7;
+ *key_inv |= ICE_0_KEYINV << 7;
+ }
+
+ dont_care >>= 1;
+ nvr_mtch >>= 1;
+ valid >>= 1;
+ val >>= 1;
+ in_key >>= 1;
+ in_key_inv >>= 1;
+ }
+
+ return 0;
+}
+
+/**
+ * ice_bits_max_set - determine if the number of bits set is within a maximum
+ * @mask: pointer to the byte array which is the mask
+ * @size: the number of bytes in the mask
+ * @max: the max number of set bits
+ *
+ * This function determines if there are at most 'max' number of bits set in an
+ * array. Returns true if the number for bits set is <= max or will return false
+ * otherwise.
+ */
+static bool ice_bits_max_set(const u8 *mask, u16 size, u16 max)
+{
+ u16 count = 0;
+ u16 i;
+
+ /* check each byte */
+ for (i = 0; i < size; i++) {
+ /* if 0, go to next byte */
+ if (!mask[i])
+ continue;
+
+ /* We know there is at least one set bit in this byte because of
+ * the above check; if we already have found 'max' number of
+ * bits set, then we can return failure now.
+ */
+ if (count == max)
+ return false;
+
+ /* count the bits in this byte, checking threshold */
+ count += hweight8(mask[i]);
+ if (count > max)
+ return false;
+ }
+
+ return true;
+}
+
+/**
+ * ice_set_key - generate a variable sized key with multiples of 16-bits
+ * @key: pointer to where the key will be stored
+ * @size: the size of the complete key in bytes (must be even)
+ * @val: array of 8-bit values that makes up the value portion of the key
+ * @upd: array of 8-bit masks that determine what key portion to update
+ * @dc: array of 8-bit masks that make up the don't care mask
+ * @nm: array of 8-bit masks that make up the never match mask
+ * @off: the offset of the first byte in the key to update
+ * @len: the number of bytes in the key update
+ *
+ * This function generates a key from a value, a don't care mask and a never
+ * match mask.
+ * upd, dc, and nm are optional parameters, and can be NULL:
+ * upd == NULL --> udp mask is all 1's (update all bits)
+ * dc == NULL --> dc mask is all 0's (no don't care bits)
+ * nm == NULL --> nm mask is all 0's (no never match bits)
+ */
+static enum ice_status
+ice_set_key(u8 *key, u16 size, u8 *val, u8 *upd, u8 *dc, u8 *nm, u16 off,
+ u16 len)
+{
+ u16 half_size;
+ u16 i;
+
+ /* size must be a multiple of 2 bytes. */
+ if (size % 2)
+ return ICE_ERR_CFG;
+
+ half_size = size / 2;
+ if (off + len > half_size)
+ return ICE_ERR_CFG;
+
+ /* Make sure at most one bit is set in the never match mask. Having more
+ * than one never match mask bit set will cause HW to consume excessive
+ * power otherwise; this is a power management efficiency check.
+ */
+#define ICE_NVR_MTCH_BITS_MAX 1
+ if (nm && !ice_bits_max_set(nm, len, ICE_NVR_MTCH_BITS_MAX))
+ return ICE_ERR_CFG;
+
+ for (i = 0; i < len; i++)
+ if (ice_gen_key_word(val[i], upd ? upd[i] : 0xff,
+ dc ? dc[i] : 0, nm ? nm[i] : 0,
+ key + off + i, key + half_size + off + i))
+ return ICE_ERR_CFG;
+
+ return 0;
+}
+
/**
* ice_acquire_global_cfg_lock
* @hw: pointer to the HW structure
@@ -952,6 +1122,48 @@ enum ice_sid_all {
ICE_SID_OFF_COUNT,
};
+/* Characteristic handling */
+
+/**
+ * ice_match_prop_lst - determine if properties of two lists match
+ * @list1: first properties list
+ * @list2: second properties list
+ *
+ * Count, cookies and the order must match in order to be considered equivalent.
+ */
+static bool
+ice_match_prop_lst(struct list_head *list1, struct list_head *list2)
+{
+ struct ice_vsig_prof *tmp1;
+ struct ice_vsig_prof *tmp2;
+ u16 chk_count = 0;
+ u16 count = 0;
+
+ /* compare counts */
+ list_for_each_entry(tmp1, list1, list)
+ count++;
+ list_for_each_entry(tmp2, list2, list)
+ chk_count++;
+ if (!count || count != chk_count)
+ return false;
+
+ tmp1 = list_first_entry(list1, struct ice_vsig_prof, list);
+ tmp2 = list_first_entry(list2, struct ice_vsig_prof, list);
+
+ /* profile cookies must compare, and in the exact same order to take
+ * into account priority
+ */
+ while (count--) {
+ if (tmp2->profile_cookie != tmp1->profile_cookie)
+ return false;
+
+ tmp1 = list_next_entry(tmp1, list);
+ tmp2 = list_next_entry(tmp2, list);
+ }
+
+ return true;
+}
+
/* VSIG Management */
/**
@@ -999,6 +1211,117 @@ static u16 ice_vsig_alloc_val(struct ice_hw *hw, enum ice_block blk, u16 vsig)
return ICE_VSIG_VALUE(idx, hw->pf_id);
}
+/**
+ * ice_vsig_alloc - Finds a free entry and allocates a new VSIG
+ * @hw: pointer to the hardware structure
+ * @blk: HW block
+ *
+ * This function will iterate through the VSIG list and mark the first
+ * unused entry for the new VSIG entry as used and return that value.
+ */
+static u16 ice_vsig_alloc(struct ice_hw *hw, enum ice_block blk)
+{
+ u16 i;
+
+ for (i = 1; i < ICE_MAX_VSIGS; i++)
+ if (!hw->blk[blk].xlt2.vsig_tbl[i].in_use)
+ return ice_vsig_alloc_val(hw, blk, i);
+
+ return ICE_DEFAULT_VSIG;
+}
+
+/**
+ * ice_find_dup_props_vsig - find VSI group with a specified set of properties
+ * @hw: pointer to the hardware structure
+ * @blk: HW block
+ * @chs: characteristic list
+ * @vsig: returns the VSIG with the matching profiles, if found
+ *
+ * Each VSIG is associated with a characteristic set; i.e. all VSIs under
+ * a group have the same characteristic set. To check if there exists a VSIG
+ * which has the same characteristics as the input characteristics; this
+ * function will iterate through the XLT2 list and return the VSIG that has a
+ * matching configuration. In order to make sure that priorities are accounted
+ * for, the list must match exactly, including the order in which the
+ * characteristics are listed.
+ */
+static enum ice_status
+ice_find_dup_props_vsig(struct ice_hw *hw, enum ice_block blk,
+ struct list_head *chs, u16 *vsig)
+{
+ struct ice_xlt2 *xlt2 = &hw->blk[blk].xlt2;
+ u16 i;
+
+ for (i = 0; i < xlt2->count; i++)
+ if (xlt2->vsig_tbl[i].in_use &&
+ ice_match_prop_lst(chs, &xlt2->vsig_tbl[i].prop_lst)) {
+ *vsig = ICE_VSIG_VALUE(i, hw->pf_id);
+ return 0;
+ }
+
+ return ICE_ERR_DOES_NOT_EXIST;
+}
+
+/**
+ * ice_vsig_free - free VSI group
+ * @hw: pointer to the hardware structure
+ * @blk: HW block
+ * @vsig: VSIG to remove
+ *
+ * The function will remove all VSIs associated with the input VSIG and move
+ * them to the DEFAULT_VSIG and mark the VSIG available.
+ */
+static enum ice_status
+ice_vsig_free(struct ice_hw *hw, enum ice_block blk, u16 vsig)
+{
+ struct ice_vsig_prof *dtmp, *del;
+ struct ice_vsig_vsi *vsi_cur;
+ u16 idx;
+
+ idx = vsig & ICE_VSIG_IDX_M;
+ if (idx >= ICE_MAX_VSIGS)
+ return ICE_ERR_PARAM;
+
+ if (!hw->blk[blk].xlt2.vsig_tbl[idx].in_use)
+ return ICE_ERR_DOES_NOT_EXIST;
+
+ hw->blk[blk].xlt2.vsig_tbl[idx].in_use = false;
+
+ vsi_cur = hw->blk[blk].xlt2.vsig_tbl[idx].first_vsi;
+ /* If the VSIG has at least 1 VSI then iterate through the
+ * list and remove the VSIs before deleting the group.
+ */
+ if (vsi_cur) {
+ /* remove all vsis associated with this VSIG XLT2 entry */
+ do {
+ struct ice_vsig_vsi *tmp = vsi_cur->next_vsi;
+
+ vsi_cur->vsig = ICE_DEFAULT_VSIG;
+ vsi_cur->changed = 1;
+ vsi_cur->next_vsi = NULL;
+ vsi_cur = tmp;
+ } while (vsi_cur);
+
+ /* NULL terminate head of VSI list */
+ hw->blk[blk].xlt2.vsig_tbl[idx].first_vsi = NULL;
+ }
+
+ /* free characteristic list */
+ list_for_each_entry_safe(del, dtmp,
+ &hw->blk[blk].xlt2.vsig_tbl[idx].prop_lst,
+ list) {
+ list_del(&del->list);
+ devm_kfree(ice_hw_to_dev(hw), del);
+ }
+
+ /* if VSIG characteristic list was cleared for reset
+ * re-initialize the list head
+ */
+ INIT_LIST_HEAD(&hw->blk[blk].xlt2.vsig_tbl[idx].prop_lst);
+
+ return 0;
+}
+
/**
* ice_vsig_remove_vsi - remove VSI from VSIG
* @hw: pointer to the hardware structure
@@ -1162,6 +1485,62 @@ static bool ice_prof_id_rsrc_type(enum ice_block blk, u16 *rsrc_type)
return true;
}
+/**
+ * ice_tcam_ent_rsrc_type - get TCAM entry resource type for a block type
+ * @blk: the block type
+ * @rsrc_type: pointer to variable to receive the resource type
+ */
+static bool ice_tcam_ent_rsrc_type(enum ice_block blk, u16 *rsrc_type)
+{
+ switch (blk) {
+ case ICE_BLK_RSS:
+ *rsrc_type = ICE_AQC_RES_TYPE_HASH_PROF_BLDR_TCAM;
+ break;
+ default:
+ return false;
+ }
+ return true;
+}
+
+/**
+ * ice_alloc_tcam_ent - allocate hardware TCAM entry
+ * @hw: pointer to the HW struct
+ * @blk: the block to allocate the TCAM for
+ * @tcam_idx: pointer to variable to receive the TCAM entry
+ *
+ * This function allocates a new entry in a Profile ID TCAM for a specific
+ * block.
+ */
+static enum ice_status
+ice_alloc_tcam_ent(struct ice_hw *hw, enum ice_block blk, u16 *tcam_idx)
+{
+ u16 res_type;
+
+ if (!ice_tcam_ent_rsrc_type(blk, &res_type))
+ return ICE_ERR_PARAM;
+
+ return ice_alloc_hw_res(hw, res_type, 1, true, tcam_idx);
+}
+
+/**
+ * ice_free_tcam_ent - free hardware TCAM entry
+ * @hw: pointer to the HW struct
+ * @blk: the block from which to free the TCAM entry
+ * @tcam_idx: the TCAM entry to free
+ *
+ * This function frees an entry in a Profile ID TCAM for a specific block.
+ */
+static enum ice_status
+ice_free_tcam_ent(struct ice_hw *hw, enum ice_block blk, u16 tcam_idx)
+{
+ u16 res_type;
+
+ if (!ice_tcam_ent_rsrc_type(blk, &res_type))
+ return ICE_ERR_PARAM;
+
+ return ice_free_hw_res(hw, res_type, 1, &tcam_idx);
+}
+
/**
* ice_alloc_prof_id - allocate profile ID
* @hw: pointer to the HW struct
@@ -1688,6 +2067,146 @@ enum ice_status ice_init_hw_tbls(struct ice_hw *hw)
return ICE_ERR_NO_MEMORY;
}
+/**
+ * ice_prof_gen_key - generate profile ID key
+ * @hw: pointer to the HW struct
+ * @blk: the block in which to write profile ID to
+ * @ptg: packet type group (PTG) portion of key
+ * @vsig: VSIG portion of key
+ * @cdid: CDID portion of key
+ * @flags: flag portion of key
+ * @vl_msk: valid mask
+ * @dc_msk: don't care mask
+ * @nm_msk: never match mask
+ * @key: output of profile ID key
+ */
+static enum ice_status
+ice_prof_gen_key(struct ice_hw *hw, enum ice_block blk, u8 ptg, u16 vsig,
+ u8 cdid, u16 flags, u8 vl_msk[ICE_TCAM_KEY_VAL_SZ],
+ u8 dc_msk[ICE_TCAM_KEY_VAL_SZ], u8 nm_msk[ICE_TCAM_KEY_VAL_SZ],
+ u8 key[ICE_TCAM_KEY_SZ])
+{
+ struct ice_prof_id_key inkey;
+
+ inkey.xlt1 = ptg;
+ inkey.xlt2_cdid = cpu_to_le16(vsig);
+ inkey.flags = cpu_to_le16(flags);
+
+ switch (hw->blk[blk].prof.cdid_bits) {
+ case 0:
+ break;
+ case 2:
+#define ICE_CD_2_M 0xC000U
+#define ICE_CD_2_S 14
+ inkey.xlt2_cdid &= ~cpu_to_le16(ICE_CD_2_M);
+ inkey.xlt2_cdid |= cpu_to_le16(BIT(cdid) << ICE_CD_2_S);
+ break;
+ case 4:
+#define ICE_CD_4_M 0xF000U
+#define ICE_CD_4_S 12
+ inkey.xlt2_cdid &= ~cpu_to_le16(ICE_CD_4_M);
+ inkey.xlt2_cdid |= cpu_to_le16(BIT(cdid) << ICE_CD_4_S);
+ break;
+ case 8:
+#define ICE_CD_8_M 0xFF00U
+#define ICE_CD_8_S 16
+ inkey.xlt2_cdid &= ~cpu_to_le16(ICE_CD_8_M);
+ inkey.xlt2_cdid |= cpu_to_le16(BIT(cdid) << ICE_CD_8_S);
+ break;
+ default:
+ ice_debug(hw, ICE_DBG_PKG, "Error in profile config\n");
+ break;
+ }
+
+ return ice_set_key(key, ICE_TCAM_KEY_SZ, (u8 *)&inkey, vl_msk, dc_msk,
+ nm_msk, 0, ICE_TCAM_KEY_SZ / 2);
+}
+
+/**
+ * ice_tcam_write_entry - write TCAM entry
+ * @hw: pointer to the HW struct
+ * @blk: the block in which to write profile ID to
+ * @idx: the entry index to write to
+ * @prof_id: profile ID
+ * @ptg: packet type group (PTG) portion of key
+ * @vsig: VSIG portion of key
+ * @cdid: CDID portion of key
+ * @flags: flag portion of key
+ * @vl_msk: valid mask
+ * @dc_msk: don't care mask
+ * @nm_msk: never match mask
+ */
+static enum ice_status
+ice_tcam_write_entry(struct ice_hw *hw, enum ice_block blk, u16 idx,
+ u8 prof_id, u8 ptg, u16 vsig, u8 cdid, u16 flags,
+ u8 vl_msk[ICE_TCAM_KEY_VAL_SZ],
+ u8 dc_msk[ICE_TCAM_KEY_VAL_SZ],
+ u8 nm_msk[ICE_TCAM_KEY_VAL_SZ])
+{
+ struct ice_prof_tcam_entry;
+ enum ice_status status;
+
+ status = ice_prof_gen_key(hw, blk, ptg, vsig, cdid, flags, vl_msk,
+ dc_msk, nm_msk, hw->blk[blk].prof.t[idx].key);
+ if (!status) {
+ hw->blk[blk].prof.t[idx].addr = cpu_to_le16(idx);
+ hw->blk[blk].prof.t[idx].prof_id = prof_id;
+ }
+
+ return status;
+}
+
+/**
+ * ice_vsig_get_ref - returns number of VSIs belong to a VSIG
+ * @hw: pointer to the hardware structure
+ * @blk: HW block
+ * @vsig: VSIG to query
+ * @refs: pointer to variable to receive the reference count
+ */
+static enum ice_status
+ice_vsig_get_ref(struct ice_hw *hw, enum ice_block blk, u16 vsig, u16 *refs)
+{
+ u16 idx = vsig & ICE_VSIG_IDX_M;
+ struct ice_vsig_vsi *ptr;
+
+ *refs = 0;
+
+ if (!hw->blk[blk].xlt2.vsig_tbl[idx].in_use)
+ return ICE_ERR_DOES_NOT_EXIST;
+
+ ptr = hw->blk[blk].xlt2.vsig_tbl[idx].first_vsi;
+ while (ptr) {
+ (*refs)++;
+ ptr = ptr->next_vsi;
+ }
+
+ return 0;
+}
+
+/**
+ * ice_has_prof_vsig - check to see if VSIG has a specific profile
+ * @hw: pointer to the hardware structure
+ * @blk: HW block
+ * @vsig: VSIG to check against
+ * @hdl: profile handle
+ */
+static bool
+ice_has_prof_vsig(struct ice_hw *hw, enum ice_block blk, u16 vsig, u64 hdl)
+{
+ u16 idx = vsig & ICE_VSIG_IDX_M;
+ struct ice_vsig_prof *ent;
+
+ list_for_each_entry(ent, &hw->blk[blk].xlt2.vsig_tbl[idx].prop_lst,
+ list)
+ if (ent->profile_cookie == hdl)
+ return true;
+
+ ice_debug(hw, ICE_DBG_INIT,
+ "Characteristic list for VSI group %d not found.\n",
+ vsig);
+ return false;
+}
+
/**
* ice_add_prof - add profile
* @hw: pointer to the HW struct
@@ -1792,3 +2311,802 @@ ice_add_prof(struct ice_hw *hw, enum ice_block blk, u64 id, u8 ptypes[],
mutex_unlock(&hw->blk[blk].es.prof_map_lock);
return status;
}
+
+/**
+ * ice_search_prof_id_low - Search for a profile tracking ID low level
+ * @hw: pointer to the HW struct
+ * @blk: hardware block
+ * @id: profile tracking ID
+ *
+ * This will search for a profile tracking ID which was previously added. This
+ * version assumes that the caller has already acquired the prof map lock.
+ */
+static struct ice_prof_map *
+ice_search_prof_id_low(struct ice_hw *hw, enum ice_block blk, u64 id)
+{
+ struct ice_prof_map *entry = NULL;
+ struct ice_prof_map *map;
+
+ list_for_each_entry(map, &hw->blk[blk].es.prof_map, list)
+ if (map->profile_cookie == id) {
+ entry = map;
+ break;
+ }
+
+ return entry;
+}
+
+/**
+ * ice_search_prof_id - Search for a profile tracking ID
+ * @hw: pointer to the HW struct
+ * @blk: hardware block
+ * @id: profile tracking ID
+ *
+ * This will search for a profile tracking ID which was previously added.
+ */
+static struct ice_prof_map *
+ice_search_prof_id(struct ice_hw *hw, enum ice_block blk, u64 id)
+{
+ struct ice_prof_map *entry;
+
+ mutex_lock(&hw->blk[blk].es.prof_map_lock);
+ entry = ice_search_prof_id_low(hw, blk, id);
+ mutex_unlock(&hw->blk[blk].es.prof_map_lock);
+
+ return entry;
+}
+
+/**
+ * ice_rel_tcam_idx - release a TCAM index
+ * @hw: pointer to the HW struct
+ * @blk: hardware block
+ * @idx: the index to release
+ */
+static enum ice_status
+ice_rel_tcam_idx(struct ice_hw *hw, enum ice_block blk, u16 idx)
+{
+ /* Masks to invoke a never match entry */
+ u8 vl_msk[ICE_TCAM_KEY_VAL_SZ] = { 0xFF, 0xFF, 0xFF, 0xFF, 0xFF };
+ u8 dc_msk[ICE_TCAM_KEY_VAL_SZ] = { 0xFE, 0xFF, 0xFF, 0xFF, 0xFF };
+ u8 nm_msk[ICE_TCAM_KEY_VAL_SZ] = { 0x01, 0x00, 0x00, 0x00, 0x00 };
+ enum ice_status status;
+
+ /* write the TCAM entry */
+ status = ice_tcam_write_entry(hw, blk, idx, 0, 0, 0, 0, 0, vl_msk,
+ dc_msk, nm_msk);
+ if (status)
+ return status;
+
+ /* release the TCAM entry */
+ status = ice_free_tcam_ent(hw, blk, idx);
+
+ return status;
+}
+
+/**
+ * ice_rem_prof_id - remove one profile from a VSIG
+ * @hw: pointer to the HW struct
+ * @blk: hardware block
+ * @prof: pointer to profile structure to remove
+ */
+static enum ice_status
+ice_rem_prof_id(struct ice_hw *hw, enum ice_block blk,
+ struct ice_vsig_prof *prof)
+{
+ enum ice_status status;
+ u16 i;
+
+ for (i = 0; i < prof->tcam_count; i++)
+ if (prof->tcam[i].in_use) {
+ prof->tcam[i].in_use = false;
+ status = ice_rel_tcam_idx(hw, blk,
+ prof->tcam[i].tcam_idx);
+ if (status)
+ return ICE_ERR_HW_TABLE;
+ }
+
+ return 0;
+}
+
+/**
+ * ice_rem_vsig - remove VSIG
+ * @hw: pointer to the HW struct
+ * @blk: hardware block
+ * @vsig: the VSIG to remove
+ * @chg: the change list
+ */
+static enum ice_status
+ice_rem_vsig(struct ice_hw *hw, enum ice_block blk, u16 vsig,
+ struct list_head *chg)
+{
+ u16 idx = vsig & ICE_VSIG_IDX_M;
+ struct ice_vsig_vsi *vsi_cur;
+ struct ice_vsig_prof *d, *t;
+ enum ice_status status;
+
+ /* remove TCAM entries */
+ list_for_each_entry_safe(d, t,
+ &hw->blk[blk].xlt2.vsig_tbl[idx].prop_lst,
+ list) {
+ status = ice_rem_prof_id(hw, blk, d);
+ if (status)
+ return status;
+
+ list_del(&d->list);
+ devm_kfree(ice_hw_to_dev(hw), d);
+ }
+
+ /* Move all VSIS associated with this VSIG to the default VSIG */
+ vsi_cur = hw->blk[blk].xlt2.vsig_tbl[idx].first_vsi;
+ /* If the VSIG has at least 1 VSI then iterate through the list
+ * and remove the VSIs before deleting the group.
+ */
+ if (vsi_cur)
+ do {
+ struct ice_vsig_vsi *tmp = vsi_cur->next_vsi;
+ struct ice_chs_chg *p;
+
+ p = devm_kzalloc(ice_hw_to_dev(hw), sizeof(*p),
+ GFP_KERNEL);
+ if (!p)
+ return ICE_ERR_NO_MEMORY;
+
+ p->type = ICE_VSIG_REM;
+ p->orig_vsig = vsig;
+ p->vsig = ICE_DEFAULT_VSIG;
+ p->vsi = vsi_cur - hw->blk[blk].xlt2.vsis;
+
+ list_add(&p->list_entry, chg);
+
+ vsi_cur = tmp;
+ } while (vsi_cur);
+
+ return ice_vsig_free(hw, blk, vsig);
+}
+
+/**
+ * ice_get_prof - get profile
+ * @hw: pointer to the HW struct
+ * @blk: hardware block
+ * @hdl: profile handle
+ * @chg: change list
+ */
+static enum ice_status
+ice_get_prof(struct ice_hw *hw, enum ice_block blk, u64 hdl,
+ struct list_head *chg)
+{
+ struct ice_prof_map *map;
+ struct ice_chs_chg *p;
+ u16 i;
+
+ /* Get the details on the profile specified by the handle ID */
+ map = ice_search_prof_id(hw, blk, hdl);
+ if (!map)
+ return ICE_ERR_DOES_NOT_EXIST;
+
+ for (i = 0; i < map->ptg_cnt; i++)
+ if (!hw->blk[blk].es.written[map->prof_id]) {
+ /* add ES to change list */
+ p = devm_kzalloc(ice_hw_to_dev(hw), sizeof(*p),
+ GFP_KERNEL);
+ if (!p)
+ goto err_ice_get_prof;
+
+ p->type = ICE_PTG_ES_ADD;
+ p->ptype = 0;
+ p->ptg = map->ptg[i];
+ p->add_ptg = 0;
+
+ p->add_prof = 1;
+ p->prof_id = map->prof_id;
+
+ hw->blk[blk].es.written[map->prof_id] = true;
+
+ list_add(&p->list_entry, chg);
+ }
+
+ return 0;
+
+err_ice_get_prof:
+ /* let caller clean up the change list */
+ return ICE_ERR_NO_MEMORY;
+}
+
+/**
+ * ice_get_profs_vsig - get a copy of the list of profiles from a VSIG
+ * @hw: pointer to the HW struct
+ * @blk: hardware block
+ * @vsig: VSIG from which to copy the list
+ * @lst: output list
+ *
+ * This routine makes a copy of the list of profiles in the specified VSIG.
+ */
+static enum ice_status
+ice_get_profs_vsig(struct ice_hw *hw, enum ice_block blk, u16 vsig,
+ struct list_head *lst)
+{
+ struct ice_vsig_prof *ent1, *ent2;
+ u16 idx = vsig & ICE_VSIG_IDX_M;
+
+ list_for_each_entry(ent1, &hw->blk[blk].xlt2.vsig_tbl[idx].prop_lst,
+ list) {
+ struct ice_vsig_prof *p;
+
+ /* copy to the input list */
+ p = devm_kmemdup(ice_hw_to_dev(hw), ent1, sizeof(*p),
+ GFP_KERNEL);
+ if (!p)
+ goto err_ice_get_profs_vsig;
+
+ list_add_tail(&p->list, lst);
+ }
+
+ return 0;
+
+err_ice_get_profs_vsig:
+ list_for_each_entry_safe(ent1, ent2, lst, list) {
+ list_del(&ent1->list);
+ devm_kfree(ice_hw_to_dev(hw), ent1);
+ }
+
+ return ICE_ERR_NO_MEMORY;
+}
+
+/**
+ * ice_add_prof_to_lst - add profile entry to a list
+ * @hw: pointer to the HW struct
+ * @blk: hardware block
+ * @lst: the list to be added to
+ * @hdl: profile handle of entry to add
+ */
+static enum ice_status
+ice_add_prof_to_lst(struct ice_hw *hw, enum ice_block blk,
+ struct list_head *lst, u64 hdl)
+{
+ struct ice_prof_map *map;
+ struct ice_vsig_prof *p;
+ u16 i;
+
+ map = ice_search_prof_id(hw, blk, hdl);
+ if (!map)
+ return ICE_ERR_DOES_NOT_EXIST;
+
+ p = devm_kzalloc(ice_hw_to_dev(hw), sizeof(*p), GFP_KERNEL);
+ if (!p)
+ return ICE_ERR_NO_MEMORY;
+
+ p->profile_cookie = map->profile_cookie;
+ p->prof_id = map->prof_id;
+ p->tcam_count = map->ptg_cnt;
+
+ for (i = 0; i < map->ptg_cnt; i++) {
+ p->tcam[i].prof_id = map->prof_id;
+ p->tcam[i].tcam_idx = ICE_INVALID_TCAM;
+ p->tcam[i].ptg = map->ptg[i];
+ }
+
+ list_add(&p->list, lst);
+
+ return 0;
+}
+
+/**
+ * ice_move_vsi - move VSI to another VSIG
+ * @hw: pointer to the HW struct
+ * @blk: hardware block
+ * @vsi: the VSI to move
+ * @vsig: the VSIG to move the VSI to
+ * @chg: the change list
+ */
+static enum ice_status
+ice_move_vsi(struct ice_hw *hw, enum ice_block blk, u16 vsi, u16 vsig,
+ struct list_head *chg)
+{
+ enum ice_status status;
+ struct ice_chs_chg *p;
+ u16 orig_vsig;
+
+ p = devm_kzalloc(ice_hw_to_dev(hw), sizeof(*p), GFP_KERNEL);
+ if (!p)
+ return ICE_ERR_NO_MEMORY;
+
+ status = ice_vsig_find_vsi(hw, blk, vsi, &orig_vsig);
+ if (!status)
+ status = ice_vsig_add_mv_vsi(hw, blk, vsi, vsig);
+
+ if (status) {
+ devm_kfree(ice_hw_to_dev(hw), p);
+ return status;
+ }
+
+ p->type = ICE_VSI_MOVE;
+ p->vsi = vsi;
+ p->orig_vsig = orig_vsig;
+ p->vsig = vsig;
+
+ list_add(&p->list_entry, chg);
+
+ return 0;
+}
+
+/**
+ * ice_prof_tcam_ena_dis - add enable or disable TCAM change
+ * @hw: pointer to the HW struct
+ * @blk: hardware block
+ * @enable: true to enable, false to disable
+ * @vsig: the VSIG of the TCAM entry
+ * @tcam: pointer the TCAM info structure of the TCAM to disable
+ * @chg: the change list
+ *
+ * This function appends an enable or disable TCAM entry in the change log
+ */
+static enum ice_status
+ice_prof_tcam_ena_dis(struct ice_hw *hw, enum ice_block blk, bool enable,
+ u16 vsig, struct ice_tcam_inf *tcam,
+ struct list_head *chg)
+{
+ enum ice_status status;
+ struct ice_chs_chg *p;
+
+ /* Default: enable means change the low flag bit to don't care */
+ u8 dc_msk[ICE_TCAM_KEY_VAL_SZ] = { 0x01, 0x00, 0x00, 0x00, 0x00 };
+ u8 nm_msk[ICE_TCAM_KEY_VAL_SZ] = { 0x00, 0x00, 0x00, 0x00, 0x00 };
+ u8 vl_msk[ICE_TCAM_KEY_VAL_SZ] = { 0x01, 0x00, 0x00, 0x00, 0x00 };
+
+ /* if disabling, free the TCAM */
+ if (!enable) {
+ status = ice_free_tcam_ent(hw, blk, tcam->tcam_idx);
+ tcam->tcam_idx = 0;
+ tcam->in_use = 0;
+ return status;
+ }
+
+ /* for re-enabling, reallocate a TCAM */
+ status = ice_alloc_tcam_ent(hw, blk, &tcam->tcam_idx);
+ if (status)
+ return status;
+
+ /* add TCAM to change list */
+ p = devm_kzalloc(ice_hw_to_dev(hw), sizeof(*p), GFP_KERNEL);
+ if (!p)
+ return ICE_ERR_NO_MEMORY;
+
+ status = ice_tcam_write_entry(hw, blk, tcam->tcam_idx, tcam->prof_id,
+ tcam->ptg, vsig, 0, 0, vl_msk, dc_msk,
+ nm_msk);
+ if (status)
+ goto err_ice_prof_tcam_ena_dis;
+
+ tcam->in_use = 1;
+
+ p->type = ICE_TCAM_ADD;
+ p->add_tcam_idx = true;
+ p->prof_id = tcam->prof_id;
+ p->ptg = tcam->ptg;
+ p->vsig = 0;
+ p->tcam_idx = tcam->tcam_idx;
+
+ /* log change */
+ list_add(&p->list_entry, chg);
+
+ return 0;
+
+err_ice_prof_tcam_ena_dis:
+ devm_kfree(ice_hw_to_dev(hw), p);
+ return status;
+}
+
+/**
+ * ice_adj_prof_priorities - adjust profile based on priorities
+ * @hw: pointer to the HW struct
+ * @blk: hardware block
+ * @vsig: the VSIG for which to adjust profile priorities
+ * @chg: the change list
+ */
+static enum ice_status
+ice_adj_prof_priorities(struct ice_hw *hw, enum ice_block blk, u16 vsig,
+ struct list_head *chg)
+{
+ DECLARE_BITMAP(ptgs_used, ICE_XLT1_CNT);
+ struct ice_vsig_prof *t;
+ enum ice_status status;
+ u16 idx;
+
+ bitmap_zero(ptgs_used, ICE_XLT1_CNT);
+ idx = vsig & ICE_VSIG_IDX_M;
+
+ /* Priority is based on the order in which the profiles are added. The
+ * newest added profile has highest priority and the oldest added
+ * profile has the lowest priority. Since the profile property list for
+ * a VSIG is sorted from newest to oldest, this code traverses the list
+ * in order and enables the first of each PTG that it finds (that is not
+ * already enabled); it also disables any duplicate PTGs that it finds
+ * in the older profiles (that are currently enabled).
+ */
+
+ list_for_each_entry(t, &hw->blk[blk].xlt2.vsig_tbl[idx].prop_lst,
+ list) {
+ u16 i;
+
+ for (i = 0; i < t->tcam_count; i++) {
+ /* Scan the priorities from newest to oldest.
+ * Make sure that the newest profiles take priority.
+ */
+ if (test_bit(t->tcam[i].ptg, ptgs_used) &&
+ t->tcam[i].in_use) {
+ /* need to mark this PTG as never match, as it
+ * was already in use and therefore duplicate
+ * (and lower priority)
+ */
+ status = ice_prof_tcam_ena_dis(hw, blk, false,
+ vsig,
+ &t->tcam[i],
+ chg);
+ if (status)
+ return status;
+ } else if (!test_bit(t->tcam[i].ptg, ptgs_used) &&
+ !t->tcam[i].in_use) {
+ /* need to enable this PTG, as it in not in use
+ * and not enabled (highest priority)
+ */
+ status = ice_prof_tcam_ena_dis(hw, blk, true,
+ vsig,
+ &t->tcam[i],
+ chg);
+ if (status)
+ return status;
+ }
+
+ /* keep track of used ptgs */
+ set_bit(t->tcam[i].ptg, ptgs_used);
+ }
+ }
+
+ return 0;
+}
+
+/**
+ * ice_add_prof_id_vsig - add profile to VSIG
+ * @hw: pointer to the HW struct
+ * @blk: hardware block
+ * @vsig: the VSIG to which this profile is to be added
+ * @hdl: the profile handle indicating the profile to add
+ * @chg: the change list
+ */
+static enum ice_status
+ice_add_prof_id_vsig(struct ice_hw *hw, enum ice_block blk, u16 vsig, u64 hdl,
+ struct list_head *chg)
+{
+ /* Masks that ignore flags */
+ u8 vl_msk[ICE_TCAM_KEY_VAL_SZ] = { 0xFF, 0xFF, 0xFF, 0xFF, 0xFF };
+ u8 dc_msk[ICE_TCAM_KEY_VAL_SZ] = { 0xFF, 0xFF, 0x00, 0x00, 0x00 };
+ u8 nm_msk[ICE_TCAM_KEY_VAL_SZ] = { 0x00, 0x00, 0x00, 0x00, 0x00 };
+ struct ice_prof_map *map;
+ struct ice_vsig_prof *t;
+ struct ice_chs_chg *p;
+ u16 i;
+
+ /* Get the details on the profile specified by the handle ID */
+ map = ice_search_prof_id(hw, blk, hdl);
+ if (!map)
+ return ICE_ERR_DOES_NOT_EXIST;
+
+ /* Error, if this VSIG already has this profile */
+ if (ice_has_prof_vsig(hw, blk, vsig, hdl))
+ return ICE_ERR_ALREADY_EXISTS;
+
+ /* new VSIG profile structure */
+ t = devm_kzalloc(ice_hw_to_dev(hw), sizeof(*t), GFP_KERNEL);
+ if (!t)
+ return ICE_ERR_NO_MEMORY;
+
+ t->profile_cookie = map->profile_cookie;
+ t->prof_id = map->prof_id;
+ t->tcam_count = map->ptg_cnt;
+
+ /* create TCAM entries */
+ for (i = 0; i < map->ptg_cnt; i++) {
+ enum ice_status status;
+ u16 tcam_idx;
+
+ /* add TCAM to change list */
+ p = devm_kzalloc(ice_hw_to_dev(hw), sizeof(*p), GFP_KERNEL);
+ if (!p)
+ goto err_ice_add_prof_id_vsig;
+
+ /* allocate the TCAM entry index */
+ status = ice_alloc_tcam_ent(hw, blk, &tcam_idx);
+ if (status) {
+ devm_kfree(ice_hw_to_dev(hw), p);
+ goto err_ice_add_prof_id_vsig;
+ }
+
+ t->tcam[i].ptg = map->ptg[i];
+ t->tcam[i].prof_id = map->prof_id;
+ t->tcam[i].tcam_idx = tcam_idx;
+ t->tcam[i].in_use = true;
+
+ p->type = ICE_TCAM_ADD;
+ p->add_tcam_idx = true;
+ p->prof_id = t->tcam[i].prof_id;
+ p->ptg = t->tcam[i].ptg;
+ p->vsig = vsig;
+ p->tcam_idx = t->tcam[i].tcam_idx;
+
+ /* write the TCAM entry */
+ status = ice_tcam_write_entry(hw, blk, t->tcam[i].tcam_idx,
+ t->tcam[i].prof_id,
+ t->tcam[i].ptg, vsig, 0, 0,
+ vl_msk, dc_msk, nm_msk);
+ if (status)
+ goto err_ice_add_prof_id_vsig;
+
+ /* log change */
+ list_add(&p->list_entry, chg);
+ }
+
+ /* add profile to VSIG */
+ list_add(&t->list,
+ &hw->blk[blk].xlt2.vsig_tbl[(vsig & ICE_VSIG_IDX_M)].prop_lst);
+
+ return 0;
+
+err_ice_add_prof_id_vsig:
+ /* let caller clean up the change list */
+ devm_kfree(ice_hw_to_dev(hw), t);
+ return ICE_ERR_NO_MEMORY;
+}
+
+/**
+ * ice_create_prof_id_vsig - add a new VSIG with a single profile
+ * @hw: pointer to the HW struct
+ * @blk: hardware block
+ * @vsi: the initial VSI that will be in VSIG
+ * @hdl: the profile handle of the profile that will be added to the VSIG
+ * @chg: the change list
+ */
+static enum ice_status
+ice_create_prof_id_vsig(struct ice_hw *hw, enum ice_block blk, u16 vsi, u64 hdl,
+ struct list_head *chg)
+{
+ enum ice_status status;
+ struct ice_chs_chg *p;
+ u16 new_vsig;
+
+ p = devm_kzalloc(ice_hw_to_dev(hw), sizeof(*p), GFP_KERNEL);
+ if (!p)
+ return ICE_ERR_NO_MEMORY;
+
+ new_vsig = ice_vsig_alloc(hw, blk);
+ if (!new_vsig) {
+ status = ICE_ERR_HW_TABLE;
+ goto err_ice_create_prof_id_vsig;
+ }
+
+ status = ice_move_vsi(hw, blk, vsi, new_vsig, chg);
+ if (status)
+ goto err_ice_create_prof_id_vsig;
+
+ status = ice_add_prof_id_vsig(hw, blk, new_vsig, hdl, chg);
+ if (status)
+ goto err_ice_create_prof_id_vsig;
+
+ p->type = ICE_VSIG_ADD;
+ p->vsi = vsi;
+ p->orig_vsig = ICE_DEFAULT_VSIG;
+ p->vsig = new_vsig;
+
+ list_add(&p->list_entry, chg);
+
+ return 0;
+
+err_ice_create_prof_id_vsig:
+ /* let caller clean up the change list */
+ devm_kfree(ice_hw_to_dev(hw), p);
+ return status;
+}
+
+/**
+ * ice_create_vsig_from_lst - create a new VSIG with a list of profiles
+ * @hw: pointer to the HW struct
+ * @blk: hardware block
+ * @vsi: the initial VSI that will be in VSIG
+ * @lst: the list of profile that will be added to the VSIG
+ * @chg: the change list
+ */
+static enum ice_status
+ice_create_vsig_from_lst(struct ice_hw *hw, enum ice_block blk, u16 vsi,
+ struct list_head *lst, struct list_head *chg)
+{
+ struct ice_vsig_prof *t;
+ enum ice_status status;
+ u16 vsig;
+
+ vsig = ice_vsig_alloc(hw, blk);
+ if (!vsig)
+ return ICE_ERR_HW_TABLE;
+
+ status = ice_move_vsi(hw, blk, vsi, vsig, chg);
+ if (status)
+ return status;
+
+ list_for_each_entry(t, lst, list) {
+ status = ice_add_prof_id_vsig(hw, blk, vsig, t->profile_cookie,
+ chg);
+ if (status)
+ return status;
+ }
+
+ return 0;
+}
+
+/**
+ * ice_find_prof_vsig - find a VSIG with a specific profile handle
+ * @hw: pointer to the HW struct
+ * @blk: hardware block
+ * @hdl: the profile handle of the profile to search for
+ * @vsig: returns the VSIG with the matching profile
+ */
+static bool
+ice_find_prof_vsig(struct ice_hw *hw, enum ice_block blk, u64 hdl, u16 *vsig)
+{
+ struct ice_vsig_prof *t;
+ enum ice_status status;
+ struct list_head lst;
+
+ INIT_LIST_HEAD(&lst);
+
+ t = kzalloc(sizeof(*t), GFP_KERNEL);
+ if (!t)
+ return false;
+
+ t->profile_cookie = hdl;
+ list_add(&t->list, &lst);
+
+ status = ice_find_dup_props_vsig(hw, blk, &lst, vsig);
+
+ list_del(&t->list);
+ kfree(t);
+
+ return !status;
+}
+
+/**
+ * ice_add_prof_id_flow - add profile flow
+ * @hw: pointer to the HW struct
+ * @blk: hardware block
+ * @vsi: the VSI to enable with the profile specified by ID
+ * @hdl: profile handle
+ *
+ * Calling this function will update the hardware tables to enable the
+ * profile indicated by the ID parameter for the VSIs specified in the VSI
+ * array. Once successfully called, the flow will be enabled.
+ */
+enum ice_status
+ice_add_prof_id_flow(struct ice_hw *hw, enum ice_block blk, u16 vsi, u64 hdl)
+{
+ struct ice_vsig_prof *tmp1, *del1;
+ struct ice_chs_chg *tmp, *del;
+ struct list_head union_lst;
+ enum ice_status status;
+ struct list_head chg;
+ u16 vsig;
+
+ INIT_LIST_HEAD(&union_lst);
+ INIT_LIST_HEAD(&chg);
+
+ /* Get profile */
+ status = ice_get_prof(hw, blk, hdl, &chg);
+ if (status)
+ return status;
+
+ /* determine if VSI is already part of a VSIG */
+ status = ice_vsig_find_vsi(hw, blk, vsi, &vsig);
+ if (!status && vsig) {
+ bool only_vsi;
+ u16 or_vsig;
+ u16 ref;
+
+ /* found in VSIG */
+ or_vsig = vsig;
+
+ /* make sure that there is no overlap/conflict between the new
+ * characteristics and the existing ones; we don't support that
+ * scenario
+ */
+ if (ice_has_prof_vsig(hw, blk, vsig, hdl)) {
+ status = ICE_ERR_ALREADY_EXISTS;
+ goto err_ice_add_prof_id_flow;
+ }
+
+ /* last VSI in the VSIG? */
+ status = ice_vsig_get_ref(hw, blk, vsig, &ref);
+ if (status)
+ goto err_ice_add_prof_id_flow;
+ only_vsi = (ref == 1);
+
+ /* create a union of the current profiles and the one being
+ * added
+ */
+ status = ice_get_profs_vsig(hw, blk, vsig, &union_lst);
+ if (status)
+ goto err_ice_add_prof_id_flow;
+
+ status = ice_add_prof_to_lst(hw, blk, &union_lst, hdl);
+ if (status)
+ goto err_ice_add_prof_id_flow;
+
+ /* search for an existing VSIG with an exact charc match */
+ status = ice_find_dup_props_vsig(hw, blk, &union_lst, &vsig);
+ if (!status) {
+ /* move VSI to the VSIG that matches */
+ status = ice_move_vsi(hw, blk, vsi, vsig, &chg);
+ if (status)
+ goto err_ice_add_prof_id_flow;
+
+ /* VSI has been moved out of or_vsig. If the or_vsig had
+ * only that VSI it is now empty and can be removed.
+ */
+ if (only_vsi) {
+ status = ice_rem_vsig(hw, blk, or_vsig, &chg);
+ if (status)
+ goto err_ice_add_prof_id_flow;
+ }
+ } else if (only_vsi) {
+ /* If the original VSIG only contains one VSI, then it
+ * will be the requesting VSI. In this case the VSI is
+ * not sharing entries and we can simply add the new
+ * profile to the VSIG.
+ */
+ status = ice_add_prof_id_vsig(hw, blk, vsig, hdl, &chg);
+ if (status)
+ goto err_ice_add_prof_id_flow;
+
+ /* Adjust priorities */
+ status = ice_adj_prof_priorities(hw, blk, vsig, &chg);
+ if (status)
+ goto err_ice_add_prof_id_flow;
+ } else {
+ /* No match, so we need a new VSIG */
+ status = ice_create_vsig_from_lst(hw, blk, vsi,
+ &union_lst, &chg);
+ if (status)
+ goto err_ice_add_prof_id_flow;
+
+ /* Adjust priorities */
+ status = ice_adj_prof_priorities(hw, blk, vsig, &chg);
+ if (status)
+ goto err_ice_add_prof_id_flow;
+ }
+ } else {
+ /* need to find or add a VSIG */
+ /* search for an existing VSIG with an exact charc match */
+ if (ice_find_prof_vsig(hw, blk, hdl, &vsig)) {
+ /* found an exact match */
+ /* add or move VSI to the VSIG that matches */
+ status = ice_move_vsi(hw, blk, vsi, vsig, &chg);
+ if (status)
+ goto err_ice_add_prof_id_flow;
+ } else {
+ /* we did not find an exact match */
+ /* we need to add a VSIG */
+ status = ice_create_prof_id_vsig(hw, blk, vsi, hdl,
+ &chg);
+ if (status)
+ goto err_ice_add_prof_id_flow;
+ }
+ }
+
+err_ice_add_prof_id_flow:
+ list_for_each_entry_safe(del, tmp, &chg, list_entry) {
+ list_del(&del->list_entry);
+ devm_kfree(ice_hw_to_dev(hw), del);
+ }
+
+ list_for_each_entry_safe(del1, tmp1, &union_lst, list) {
+ list_del(&del1->list);
+ devm_kfree(ice_hw_to_dev(hw), del1);
+ }
+
+ return status;
+}
diff --git a/drivers/net/ethernet/intel/ice/ice_flex_pipe.h b/drivers/net/ethernet/intel/ice/ice_flex_pipe.h
index 8cb7d7f09e0b..33e4510da24c 100644
--- a/drivers/net/ethernet/intel/ice/ice_flex_pipe.h
+++ b/drivers/net/ethernet/intel/ice/ice_flex_pipe.h
@@ -21,6 +21,8 @@
enum ice_status
ice_add_prof(struct ice_hw *hw, enum ice_block blk, u64 id, u8 ptypes[],
struct ice_fv_word *es);
+enum ice_status
+ice_add_prof_id_flow(struct ice_hw *hw, enum ice_block blk, u16 vsi, u64 hdl);
enum ice_status ice_init_pkg(struct ice_hw *hw, u8 *buff, u32 len);
enum ice_status
ice_copy_and_init_pkg(struct ice_hw *hw, const u8 *buf, u32 len);
diff --git a/drivers/net/ethernet/intel/ice/ice_flex_type.h b/drivers/net/ethernet/intel/ice/ice_flex_type.h
index 3005f111fb3b..9d95d51bc760 100644
--- a/drivers/net/ethernet/intel/ice/ice_flex_type.h
+++ b/drivers/net/ethernet/intel/ice/ice_flex_type.h
@@ -283,6 +283,7 @@ struct ice_ptg_ptype {
u8 ptg;
};
+#define ICE_MAX_TCAM_PER_PROFILE 32
#define ICE_MAX_PTG_PER_PROFILE 32
struct ice_prof_map {
@@ -294,6 +295,23 @@ struct ice_prof_map {
u8 ptg[ICE_MAX_PTG_PER_PROFILE];
};
+#define ICE_INVALID_TCAM 0xFFFF
+
+struct ice_tcam_inf {
+ u16 tcam_idx;
+ u8 ptg;
+ u8 prof_id;
+ u8 in_use;
+};
+
+struct ice_vsig_prof {
+ struct list_head list;
+ u64 profile_cookie;
+ u8 prof_id;
+ u8 tcam_count;
+ struct ice_tcam_inf tcam[ICE_MAX_TCAM_PER_PROFILE];
+};
+
struct ice_vsig_entry {
struct list_head prop_lst;
struct ice_vsig_vsi *first_vsi;
@@ -343,6 +361,13 @@ struct ice_xlt2 {
u16 count;
};
+/* Profile ID Management */
+struct ice_prof_id_key {
+ __le16 flags;
+ u8 xlt1;
+ __le16 xlt2_cdid;
+} __packed;
+
/* Keys are made up of two values, each one-half the size of the key.
* For TCAM, the entire key is 80 bits wide (or 2, 40-bit wide values)
*/
@@ -385,5 +410,31 @@ struct ice_blk_info {
u8 is_list_init;
};
+enum ice_chg_type {
+ ICE_TCAM_NONE = 0,
+ ICE_PTG_ES_ADD,
+ ICE_TCAM_ADD,
+ ICE_VSIG_ADD,
+ ICE_VSIG_REM,
+ ICE_VSI_MOVE,
+};
+
+struct ice_chs_chg {
+ struct list_head list_entry;
+ enum ice_chg_type type;
+
+ u8 add_ptg;
+ u8 add_vsig;
+ u8 add_tcam_idx;
+ u8 add_prof;
+ u16 ptype;
+ u8 ptg;
+ u8 prof_id;
+ u16 vsi;
+ u16 vsig;
+ u16 orig_vsig;
+ u16 tcam_idx;
+};
+
#define ICE_FLOW_PTYPE_MAX ICE_XLT1_CNT
#endif /* _ICE_FLEX_TYPE_H_ */
diff --git a/drivers/net/ethernet/intel/ice/ice_flow.c b/drivers/net/ethernet/intel/ice/ice_flow.c
index 24fe04f8baa2..1bbf4b6ed7d2 100644
--- a/drivers/net/ethernet/intel/ice/ice_flow.c
+++ b/drivers/net/ethernet/intel/ice/ice_flow.c
@@ -450,6 +450,38 @@ ice_flow_add_prof_sync(struct ice_hw *hw, enum ice_block blk,
return status;
}
+/**
+ * ice_flow_assoc_prof - associate a VSI with a flow profile
+ * @hw: pointer to the hardware structure
+ * @blk: classification stage
+ * @prof: pointer to flow profile
+ * @vsi_handle: software VSI handle
+ *
+ * Assumption: the caller has acquired the lock to the profile list
+ * and the software VSI handle has been validated
+ */
+static enum ice_status
+ice_flow_assoc_prof(struct ice_hw *hw, enum ice_block blk,
+ struct ice_flow_prof *prof, u16 vsi_handle)
+{
+ enum ice_status status = 0;
+
+ if (!test_bit(vsi_handle, prof->vsis)) {
+ status = ice_add_prof_id_flow(hw, blk,
+ ice_get_hw_vsi_num(hw,
+ vsi_handle),
+ prof->id);
+ if (!status)
+ set_bit(vsi_handle, prof->vsis);
+ else
+ ice_debug(hw, ICE_DBG_FLOW,
+ "HW profile add failed, %d\n",
+ status);
+ }
+
+ return status;
+}
+
/**
* ice_flow_add_prof - Add a flow profile for packet segments and matched fields
* @hw: pointer to the HW struct
@@ -458,12 +490,13 @@ ice_flow_add_prof_sync(struct ice_hw *hw, enum ice_block blk,
* @prof_id: unique ID to identify this flow profile
* @segs: array of one or more packet segments that describe the flow
* @segs_cnt: number of packet segments provided
+ * @prof: stores the returned flow profile added
*/
static enum ice_status
ice_flow_add_prof(struct ice_hw *hw, enum ice_block blk, enum ice_flow_dir dir,
- u64 prof_id, struct ice_flow_seg_info *segs, u8 segs_cnt)
+ u64 prof_id, struct ice_flow_seg_info *segs, u8 segs_cnt,
+ struct ice_flow_prof **prof)
{
- struct ice_flow_prof *prof = NULL;
enum ice_status status;
if (segs_cnt > ICE_FLOW_SEG_MAX)
@@ -482,9 +515,9 @@ ice_flow_add_prof(struct ice_hw *hw, enum ice_block blk, enum ice_flow_dir dir,
mutex_lock(&hw->fl_profs_locks[blk]);
status = ice_flow_add_prof_sync(hw, blk, dir, prof_id, segs, segs_cnt,
- &prof);
+ prof);
if (!status)
- list_add(&prof->l_entry, &hw->fl_profs[blk]);
+ list_add(&(*prof)->l_entry, &hw->fl_profs[blk]);
mutex_unlock(&hw->fl_profs_locks[blk]);
@@ -634,6 +667,7 @@ ice_flow_set_rss_seg_info(struct ice_flow_seg_info *segs, u64 hash_fields,
/**
* ice_add_rss_cfg_sync - add an RSS configuration
* @hw: pointer to the hardware structure
+ * @vsi_handle: software VSI handle
* @hashed_flds: hash bit fields (ICE_FLOW_HASH_*) to configure
* @addl_hdrs: protocol header fields
* @segs_cnt: packet segment count
@@ -641,9 +675,11 @@ ice_flow_set_rss_seg_info(struct ice_flow_seg_info *segs, u64 hash_fields,
* Assumption: lock has already been acquired for RSS list
*/
static enum ice_status
-ice_add_rss_cfg_sync(struct ice_hw *hw, u64 hashed_flds, u32 addl_hdrs,
- u8 segs_cnt)
+ice_add_rss_cfg_sync(struct ice_hw *hw, u16 vsi_handle, u64 hashed_flds,
+ u32 addl_hdrs, u8 segs_cnt)
{
+ const enum ice_block blk = ICE_BLK_RSS;
+ struct ice_flow_prof *prof = NULL;
struct ice_flow_seg_info *segs;
enum ice_status status;
@@ -663,11 +699,15 @@ ice_add_rss_cfg_sync(struct ice_hw *hw, u64 hashed_flds, u32 addl_hdrs,
/* Create a new flow profile with generated profile and packet
* segment information.
*/
- status = ice_flow_add_prof(hw, ICE_BLK_RSS, ICE_FLOW_RX,
+ status = ice_flow_add_prof(hw, blk, ICE_FLOW_RX,
ICE_FLOW_GEN_PROFID(hashed_flds,
segs[segs_cnt - 1].hdrs,
segs_cnt),
- segs, segs_cnt);
+ segs, segs_cnt, &prof);
+ if (status)
+ goto exit;
+
+ status = ice_flow_assoc_prof(hw, blk, prof, vsi_handle);
exit:
kfree(segs);
@@ -696,7 +736,7 @@ ice_add_rss_cfg(struct ice_hw *hw, u16 vsi_handle, u64 hashed_flds,
return ICE_ERR_PARAM;
mutex_lock(&hw->rss_locks);
- status = ice_add_rss_cfg_sync(hw, hashed_flds, addl_hdrs,
+ status = ice_add_rss_cfg_sync(hw, vsi_handle, hashed_flds, addl_hdrs,
ICE_RSS_OUTER_HEADERS);
mutex_unlock(&hw->rss_locks);
@@ -719,7 +759,8 @@ enum ice_status ice_replay_rss_cfg(struct ice_hw *hw, u16 vsi_handle)
mutex_lock(&hw->rss_locks);
list_for_each_entry(r, &hw->rss_list_head, l_entry) {
if (test_bit(vsi_handle, r->vsis)) {
- status = ice_add_rss_cfg_sync(hw, r->hashed_flds,
+ status = ice_add_rss_cfg_sync(hw, vsi_handle,
+ r->hashed_flds,
r->packet_hdr,
ICE_RSS_OUTER_HEADERS);
if (status)
diff --git a/drivers/net/ethernet/intel/ice/ice_status.h b/drivers/net/ethernet/intel/ice/ice_status.h
index c01597885629..a9a8bc3aca42 100644
--- a/drivers/net/ethernet/intel/ice/ice_status.h
+++ b/drivers/net/ethernet/intel/ice/ice_status.h
@@ -26,6 +26,7 @@ enum ice_status {
ICE_ERR_IN_USE = -16,
ICE_ERR_MAX_LIMIT = -17,
ICE_ERR_RESET_ONGOING = -18,
+ ICE_ERR_HW_TABLE = -19,
ICE_ERR_NVM_CHECKSUM = -51,
ICE_ERR_BUF_TOO_SHORT = -52,
ICE_ERR_NVM_BLANK_MODE = -53,
--
2.24.1
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [net-next 4/8] ice: Enable writing filtering tables
2020-01-26 6:07 [net-next 0/8][pull request] 100GbE Intel Wired LAN Driver Updates 2020-01-25 Jeff Kirsher
` (2 preceding siblings ...)
2020-01-26 6:07 ` [net-next 3/8] ice: Populate TCAM filter software structures Jeff Kirsher
@ 2020-01-26 6:07 ` Jeff Kirsher
2020-01-26 6:07 ` [net-next 5/8] ice: Optimize table usage Jeff Kirsher
` (4 subsequent siblings)
8 siblings, 0 replies; 10+ messages in thread
From: Jeff Kirsher @ 2020-01-26 6:07 UTC (permalink / raw)
To: davem
Cc: Tony Nguyen, netdev, nhorman, sassmann, Henry Tieman,
Andrew Bowers, Jeff Kirsher
From: Tony Nguyen <anthony.l.nguyen@intel.com>
Write the hardware tables based on the populated software structures.
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Henry Tieman <henry.w.tieman@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
.../net/ethernet/intel/ice/ice_adminq_cmd.h | 1 +
.../net/ethernet/intel/ice/ice_flex_pipe.c | 604 ++++++++++++++++++
.../net/ethernet/intel/ice/ice_flex_type.h | 46 ++
3 files changed, 651 insertions(+)
diff --git a/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h b/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h
index 9967973a7c42..4459bc564b11 100644
--- a/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h
+++ b/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h
@@ -1856,6 +1856,7 @@ enum ice_adminq_opc {
/* package commands */
ice_aqc_opc_download_pkg = 0x0C40,
+ ice_aqc_opc_update_pkg = 0x0C42,
ice_aqc_opc_get_pkg_info_list = 0x0C43,
/* debug commands */
diff --git a/drivers/net/ethernet/intel/ice/ice_flex_pipe.c b/drivers/net/ethernet/intel/ice/ice_flex_pipe.c
index 8c93c303a4a5..6dca611b408a 100644
--- a/drivers/net/ethernet/intel/ice/ice_flex_pipe.c
+++ b/drivers/net/ethernet/intel/ice/ice_flex_pipe.c
@@ -5,6 +5,86 @@
#include "ice_flex_pipe.h"
#include "ice_flow.h"
+static const u32 ice_sect_lkup[ICE_BLK_COUNT][ICE_SECT_COUNT] = {
+ /* SWITCH */
+ {
+ ICE_SID_XLT0_SW,
+ ICE_SID_XLT_KEY_BUILDER_SW,
+ ICE_SID_XLT1_SW,
+ ICE_SID_XLT2_SW,
+ ICE_SID_PROFID_TCAM_SW,
+ ICE_SID_PROFID_REDIR_SW,
+ ICE_SID_FLD_VEC_SW,
+ ICE_SID_CDID_KEY_BUILDER_SW,
+ ICE_SID_CDID_REDIR_SW
+ },
+
+ /* ACL */
+ {
+ ICE_SID_XLT0_ACL,
+ ICE_SID_XLT_KEY_BUILDER_ACL,
+ ICE_SID_XLT1_ACL,
+ ICE_SID_XLT2_ACL,
+ ICE_SID_PROFID_TCAM_ACL,
+ ICE_SID_PROFID_REDIR_ACL,
+ ICE_SID_FLD_VEC_ACL,
+ ICE_SID_CDID_KEY_BUILDER_ACL,
+ ICE_SID_CDID_REDIR_ACL
+ },
+
+ /* FD */
+ {
+ ICE_SID_XLT0_FD,
+ ICE_SID_XLT_KEY_BUILDER_FD,
+ ICE_SID_XLT1_FD,
+ ICE_SID_XLT2_FD,
+ ICE_SID_PROFID_TCAM_FD,
+ ICE_SID_PROFID_REDIR_FD,
+ ICE_SID_FLD_VEC_FD,
+ ICE_SID_CDID_KEY_BUILDER_FD,
+ ICE_SID_CDID_REDIR_FD
+ },
+
+ /* RSS */
+ {
+ ICE_SID_XLT0_RSS,
+ ICE_SID_XLT_KEY_BUILDER_RSS,
+ ICE_SID_XLT1_RSS,
+ ICE_SID_XLT2_RSS,
+ ICE_SID_PROFID_TCAM_RSS,
+ ICE_SID_PROFID_REDIR_RSS,
+ ICE_SID_FLD_VEC_RSS,
+ ICE_SID_CDID_KEY_BUILDER_RSS,
+ ICE_SID_CDID_REDIR_RSS
+ },
+
+ /* PE */
+ {
+ ICE_SID_XLT0_PE,
+ ICE_SID_XLT_KEY_BUILDER_PE,
+ ICE_SID_XLT1_PE,
+ ICE_SID_XLT2_PE,
+ ICE_SID_PROFID_TCAM_PE,
+ ICE_SID_PROFID_REDIR_PE,
+ ICE_SID_FLD_VEC_PE,
+ ICE_SID_CDID_KEY_BUILDER_PE,
+ ICE_SID_CDID_REDIR_PE
+ }
+};
+
+/**
+ * ice_sect_id - returns section ID
+ * @blk: block type
+ * @sect: section type
+ *
+ * This helper function returns the proper section ID given a block type and a
+ * section type.
+ */
+static u32 ice_sect_id(enum ice_block blk, enum ice_sect sect)
+{
+ return ice_sect_lkup[blk][sect];
+}
+
/**
* ice_pkg_val_buf
* @buf: pointer to the ice buffer
@@ -375,6 +455,31 @@ static void ice_release_global_cfg_lock(struct ice_hw *hw)
ice_release_res(hw, ICE_GLOBAL_CFG_LOCK_RES_ID);
}
+/**
+ * ice_acquire_change_lock
+ * @hw: pointer to the HW structure
+ * @access: access type (read or write)
+ *
+ * This function will request ownership of the change lock.
+ */
+static enum ice_status
+ice_acquire_change_lock(struct ice_hw *hw, enum ice_aq_res_access_type access)
+{
+ return ice_acquire_res(hw, ICE_CHANGE_LOCK_RES_ID, access,
+ ICE_CHANGE_LOCK_TIMEOUT);
+}
+
+/**
+ * ice_release_change_lock
+ * @hw: pointer to the HW structure
+ *
+ * This function will release the change lock using the proper Admin Command.
+ */
+static void ice_release_change_lock(struct ice_hw *hw)
+{
+ ice_release_res(hw, ICE_CHANGE_LOCK_RES_ID);
+}
+
/**
* ice_aq_download_pkg
* @hw: pointer to the hardware structure
@@ -423,6 +528,54 @@ ice_aq_download_pkg(struct ice_hw *hw, struct ice_buf_hdr *pkg_buf,
return status;
}
+/**
+ * ice_aq_update_pkg
+ * @hw: pointer to the hardware structure
+ * @pkg_buf: the package cmd buffer
+ * @buf_size: the size of the package cmd buffer
+ * @last_buf: last buffer indicator
+ * @error_offset: returns error offset
+ * @error_info: returns error information
+ * @cd: pointer to command details structure or NULL
+ *
+ * Update Package (0x0C42)
+ */
+static enum ice_status
+ice_aq_update_pkg(struct ice_hw *hw, struct ice_buf_hdr *pkg_buf, u16 buf_size,
+ bool last_buf, u32 *error_offset, u32 *error_info,
+ struct ice_sq_cd *cd)
+{
+ struct ice_aqc_download_pkg *cmd;
+ struct ice_aq_desc desc;
+ enum ice_status status;
+
+ if (error_offset)
+ *error_offset = 0;
+ if (error_info)
+ *error_info = 0;
+
+ cmd = &desc.params.download_pkg;
+ ice_fill_dflt_direct_cmd_desc(&desc, ice_aqc_opc_update_pkg);
+ desc.flags |= cpu_to_le16(ICE_AQ_FLAG_RD);
+
+ if (last_buf)
+ cmd->flags |= ICE_AQC_DOWNLOAD_PKG_LAST_BUF;
+
+ status = ice_aq_send_cmd(hw, &desc, pkg_buf, buf_size, cd);
+ if (status == ICE_ERR_AQ_ERROR) {
+ /* Read error from buffer only when the FW returned an error */
+ struct ice_aqc_download_pkg_resp *resp;
+
+ resp = (struct ice_aqc_download_pkg_resp *)pkg_buf;
+ if (error_offset)
+ *error_offset = le32_to_cpu(resp->error_offset);
+ if (error_info)
+ *error_info = le32_to_cpu(resp->error_info);
+ }
+
+ return status;
+}
+
/**
* ice_find_seg_in_pkg
* @hw: pointer to the hardware structure
@@ -457,6 +610,44 @@ ice_find_seg_in_pkg(struct ice_hw *hw, u32 seg_type,
return NULL;
}
+/**
+ * ice_update_pkg
+ * @hw: pointer to the hardware structure
+ * @bufs: pointer to an array of buffers
+ * @count: the number of buffers in the array
+ *
+ * Obtains change lock and updates package.
+ */
+static enum ice_status
+ice_update_pkg(struct ice_hw *hw, struct ice_buf *bufs, u32 count)
+{
+ enum ice_status status;
+ u32 offset, info, i;
+
+ status = ice_acquire_change_lock(hw, ICE_RES_WRITE);
+ if (status)
+ return status;
+
+ for (i = 0; i < count; i++) {
+ struct ice_buf_hdr *bh = (struct ice_buf_hdr *)(bufs + i);
+ bool last = ((i + 1) == count);
+
+ status = ice_aq_update_pkg(hw, bh, le16_to_cpu(bh->data_end),
+ last, &offset, &info, NULL);
+
+ if (status) {
+ ice_debug(hw, ICE_DBG_PKG,
+ "Update pkg failed: err %d off %d inf %d\n",
+ status, offset, info);
+ break;
+ }
+ }
+
+ ice_release_change_lock(hw);
+
+ return status;
+}
+
/**
* ice_dwnld_cfg_bufs
* @hw: pointer to the hardware structure
@@ -938,6 +1129,169 @@ enum ice_status ice_copy_and_init_pkg(struct ice_hw *hw, const u8 *buf, u32 len)
return status;
}
+/**
+ * ice_pkg_buf_alloc
+ * @hw: pointer to the HW structure
+ *
+ * Allocates a package buffer and returns a pointer to the buffer header.
+ * Note: all package contents must be in Little Endian form.
+ */
+static struct ice_buf_build *ice_pkg_buf_alloc(struct ice_hw *hw)
+{
+ struct ice_buf_build *bld;
+ struct ice_buf_hdr *buf;
+
+ bld = devm_kzalloc(ice_hw_to_dev(hw), sizeof(*bld), GFP_KERNEL);
+ if (!bld)
+ return NULL;
+
+ buf = (struct ice_buf_hdr *)bld;
+ buf->data_end = cpu_to_le16(offsetof(struct ice_buf_hdr,
+ section_entry));
+ return bld;
+}
+
+/**
+ * ice_pkg_buf_free
+ * @hw: pointer to the HW structure
+ * @bld: pointer to pkg build (allocated by ice_pkg_buf_alloc())
+ *
+ * Frees a package buffer
+ */
+static void ice_pkg_buf_free(struct ice_hw *hw, struct ice_buf_build *bld)
+{
+ devm_kfree(ice_hw_to_dev(hw), bld);
+}
+
+/**
+ * ice_pkg_buf_reserve_section
+ * @bld: pointer to pkg build (allocated by ice_pkg_buf_alloc())
+ * @count: the number of sections to reserve
+ *
+ * Reserves one or more section table entries in a package buffer. This routine
+ * can be called multiple times as long as they are made before calling
+ * ice_pkg_buf_alloc_section(). Once ice_pkg_buf_alloc_section()
+ * is called once, the number of sections that can be allocated will not be able
+ * to be increased; not using all reserved sections is fine, but this will
+ * result in some wasted space in the buffer.
+ * Note: all package contents must be in Little Endian form.
+ */
+static enum ice_status
+ice_pkg_buf_reserve_section(struct ice_buf_build *bld, u16 count)
+{
+ struct ice_buf_hdr *buf;
+ u16 section_count;
+ u16 data_end;
+
+ if (!bld)
+ return ICE_ERR_PARAM;
+
+ buf = (struct ice_buf_hdr *)&bld->buf;
+
+ /* already an active section, can't increase table size */
+ section_count = le16_to_cpu(buf->section_count);
+ if (section_count > 0)
+ return ICE_ERR_CFG;
+
+ if (bld->reserved_section_table_entries + count > ICE_MAX_S_COUNT)
+ return ICE_ERR_CFG;
+ bld->reserved_section_table_entries += count;
+
+ data_end = le16_to_cpu(buf->data_end) +
+ (count * sizeof(buf->section_entry[0]));
+ buf->data_end = cpu_to_le16(data_end);
+
+ return 0;
+}
+
+/**
+ * ice_pkg_buf_alloc_section
+ * @bld: pointer to pkg build (allocated by ice_pkg_buf_alloc())
+ * @type: the section type value
+ * @size: the size of the section to reserve (in bytes)
+ *
+ * Reserves memory in the buffer for a section's content and updates the
+ * buffers' status accordingly. This routine returns a pointer to the first
+ * byte of the section start within the buffer, which is used to fill in the
+ * section contents.
+ * Note: all package contents must be in Little Endian form.
+ */
+static void *
+ice_pkg_buf_alloc_section(struct ice_buf_build *bld, u32 type, u16 size)
+{
+ struct ice_buf_hdr *buf;
+ u16 sect_count;
+ u16 data_end;
+
+ if (!bld || !type || !size)
+ return NULL;
+
+ buf = (struct ice_buf_hdr *)&bld->buf;
+
+ /* check for enough space left in buffer */
+ data_end = le16_to_cpu(buf->data_end);
+
+ /* section start must align on 4 byte boundary */
+ data_end = ALIGN(data_end, 4);
+
+ if ((data_end + size) > ICE_MAX_S_DATA_END)
+ return NULL;
+
+ /* check for more available section table entries */
+ sect_count = le16_to_cpu(buf->section_count);
+ if (sect_count < bld->reserved_section_table_entries) {
+ void *section_ptr = ((u8 *)buf) + data_end;
+
+ buf->section_entry[sect_count].offset = cpu_to_le16(data_end);
+ buf->section_entry[sect_count].size = cpu_to_le16(size);
+ buf->section_entry[sect_count].type = cpu_to_le32(type);
+
+ data_end += size;
+ buf->data_end = cpu_to_le16(data_end);
+
+ buf->section_count = cpu_to_le16(sect_count + 1);
+ return section_ptr;
+ }
+
+ /* no free section table entries */
+ return NULL;
+}
+
+/**
+ * ice_pkg_buf_get_active_sections
+ * @bld: pointer to pkg build (allocated by ice_pkg_buf_alloc())
+ *
+ * Returns the number of active sections. Before using the package buffer
+ * in an update package command, the caller should make sure that there is at
+ * least one active section - otherwise, the buffer is not legal and should
+ * not be used.
+ * Note: all package contents must be in Little Endian form.
+ */
+static u16 ice_pkg_buf_get_active_sections(struct ice_buf_build *bld)
+{
+ struct ice_buf_hdr *buf;
+
+ if (!bld)
+ return 0;
+
+ buf = (struct ice_buf_hdr *)&bld->buf;
+ return le16_to_cpu(buf->section_count);
+}
+
+/**
+ * ice_pkg_buf
+ * @bld: pointer to pkg build (allocated by ice_pkg_buf_alloc())
+ *
+ * Return a pointer to the buffer's header
+ */
+static struct ice_buf *ice_pkg_buf(struct ice_buf_build *bld)
+{
+ if (!bld)
+ return NULL;
+
+ return &bld->buf;
+}
+
/* PTG Management */
/**
@@ -2207,6 +2561,252 @@ ice_has_prof_vsig(struct ice_hw *hw, enum ice_block blk, u16 vsig, u64 hdl)
return false;
}
+/**
+ * ice_prof_bld_es - build profile ID extraction sequence changes
+ * @hw: pointer to the HW struct
+ * @blk: hardware block
+ * @bld: the update package buffer build to add to
+ * @chgs: the list of changes to make in hardware
+ */
+static enum ice_status
+ice_prof_bld_es(struct ice_hw *hw, enum ice_block blk,
+ struct ice_buf_build *bld, struct list_head *chgs)
+{
+ u16 vec_size = hw->blk[blk].es.fvw * sizeof(struct ice_fv_word);
+ struct ice_chs_chg *tmp;
+
+ list_for_each_entry(tmp, chgs, list_entry)
+ if (tmp->type == ICE_PTG_ES_ADD && tmp->add_prof) {
+ u16 off = tmp->prof_id * hw->blk[blk].es.fvw;
+ struct ice_pkg_es *p;
+ u32 id;
+
+ id = ice_sect_id(blk, ICE_VEC_TBL);
+ p = (struct ice_pkg_es *)
+ ice_pkg_buf_alloc_section(bld, id, sizeof(*p) +
+ vec_size -
+ sizeof(p->es[0]));
+
+ if (!p)
+ return ICE_ERR_MAX_LIMIT;
+
+ p->count = cpu_to_le16(1);
+ p->offset = cpu_to_le16(tmp->prof_id);
+
+ memcpy(p->es, &hw->blk[blk].es.t[off], vec_size);
+ }
+
+ return 0;
+}
+
+/**
+ * ice_prof_bld_tcam - build profile ID TCAM changes
+ * @hw: pointer to the HW struct
+ * @blk: hardware block
+ * @bld: the update package buffer build to add to
+ * @chgs: the list of changes to make in hardware
+ */
+static enum ice_status
+ice_prof_bld_tcam(struct ice_hw *hw, enum ice_block blk,
+ struct ice_buf_build *bld, struct list_head *chgs)
+{
+ struct ice_chs_chg *tmp;
+
+ list_for_each_entry(tmp, chgs, list_entry)
+ if (tmp->type == ICE_TCAM_ADD && tmp->add_tcam_idx) {
+ struct ice_prof_id_section *p;
+ u32 id;
+
+ id = ice_sect_id(blk, ICE_PROF_TCAM);
+ p = (struct ice_prof_id_section *)
+ ice_pkg_buf_alloc_section(bld, id, sizeof(*p));
+
+ if (!p)
+ return ICE_ERR_MAX_LIMIT;
+
+ p->count = cpu_to_le16(1);
+ p->entry[0].addr = cpu_to_le16(tmp->tcam_idx);
+ p->entry[0].prof_id = tmp->prof_id;
+
+ memcpy(p->entry[0].key,
+ &hw->blk[blk].prof.t[tmp->tcam_idx].key,
+ sizeof(hw->blk[blk].prof.t->key));
+ }
+
+ return 0;
+}
+
+/**
+ * ice_prof_bld_xlt1 - build XLT1 changes
+ * @blk: hardware block
+ * @bld: the update package buffer build to add to
+ * @chgs: the list of changes to make in hardware
+ */
+static enum ice_status
+ice_prof_bld_xlt1(enum ice_block blk, struct ice_buf_build *bld,
+ struct list_head *chgs)
+{
+ struct ice_chs_chg *tmp;
+
+ list_for_each_entry(tmp, chgs, list_entry)
+ if (tmp->type == ICE_PTG_ES_ADD && tmp->add_ptg) {
+ struct ice_xlt1_section *p;
+ u32 id;
+
+ id = ice_sect_id(blk, ICE_XLT1);
+ p = (struct ice_xlt1_section *)
+ ice_pkg_buf_alloc_section(bld, id, sizeof(*p));
+
+ if (!p)
+ return ICE_ERR_MAX_LIMIT;
+
+ p->count = cpu_to_le16(1);
+ p->offset = cpu_to_le16(tmp->ptype);
+ p->value[0] = tmp->ptg;
+ }
+
+ return 0;
+}
+
+/**
+ * ice_prof_bld_xlt2 - build XLT2 changes
+ * @blk: hardware block
+ * @bld: the update package buffer build to add to
+ * @chgs: the list of changes to make in hardware
+ */
+static enum ice_status
+ice_prof_bld_xlt2(enum ice_block blk, struct ice_buf_build *bld,
+ struct list_head *chgs)
+{
+ struct ice_chs_chg *tmp;
+
+ list_for_each_entry(tmp, chgs, list_entry) {
+ struct ice_xlt2_section *p;
+ u32 id;
+
+ switch (tmp->type) {
+ case ICE_VSIG_ADD:
+ case ICE_VSI_MOVE:
+ case ICE_VSIG_REM:
+ id = ice_sect_id(blk, ICE_XLT2);
+ p = (struct ice_xlt2_section *)
+ ice_pkg_buf_alloc_section(bld, id, sizeof(*p));
+
+ if (!p)
+ return ICE_ERR_MAX_LIMIT;
+
+ p->count = cpu_to_le16(1);
+ p->offset = cpu_to_le16(tmp->vsi);
+ p->value[0] = cpu_to_le16(tmp->vsig);
+ break;
+ default:
+ break;
+ }
+ }
+
+ return 0;
+}
+
+/**
+ * ice_upd_prof_hw - update hardware using the change list
+ * @hw: pointer to the HW struct
+ * @blk: hardware block
+ * @chgs: the list of changes to make in hardware
+ */
+static enum ice_status
+ice_upd_prof_hw(struct ice_hw *hw, enum ice_block blk,
+ struct list_head *chgs)
+{
+ struct ice_buf_build *b;
+ struct ice_chs_chg *tmp;
+ enum ice_status status;
+ u16 pkg_sects;
+ u16 xlt1 = 0;
+ u16 xlt2 = 0;
+ u16 tcam = 0;
+ u16 es = 0;
+ u16 sects;
+
+ /* count number of sections we need */
+ list_for_each_entry(tmp, chgs, list_entry) {
+ switch (tmp->type) {
+ case ICE_PTG_ES_ADD:
+ if (tmp->add_ptg)
+ xlt1++;
+ if (tmp->add_prof)
+ es++;
+ break;
+ case ICE_TCAM_ADD:
+ tcam++;
+ break;
+ case ICE_VSIG_ADD:
+ case ICE_VSI_MOVE:
+ case ICE_VSIG_REM:
+ xlt2++;
+ break;
+ default:
+ break;
+ }
+ }
+ sects = xlt1 + xlt2 + tcam + es;
+
+ if (!sects)
+ return 0;
+
+ /* Build update package buffer */
+ b = ice_pkg_buf_alloc(hw);
+ if (!b)
+ return ICE_ERR_NO_MEMORY;
+
+ status = ice_pkg_buf_reserve_section(b, sects);
+ if (status)
+ goto error_tmp;
+
+ /* Preserve order of table update: ES, TCAM, PTG, VSIG */
+ if (es) {
+ status = ice_prof_bld_es(hw, blk, b, chgs);
+ if (status)
+ goto error_tmp;
+ }
+
+ if (tcam) {
+ status = ice_prof_bld_tcam(hw, blk, b, chgs);
+ if (status)
+ goto error_tmp;
+ }
+
+ if (xlt1) {
+ status = ice_prof_bld_xlt1(blk, b, chgs);
+ if (status)
+ goto error_tmp;
+ }
+
+ if (xlt2) {
+ status = ice_prof_bld_xlt2(blk, b, chgs);
+ if (status)
+ goto error_tmp;
+ }
+
+ /* After package buffer build check if the section count in buffer is
+ * non-zero and matches the number of sections detected for package
+ * update.
+ */
+ pkg_sects = ice_pkg_buf_get_active_sections(b);
+ if (!pkg_sects || pkg_sects != sects) {
+ status = ICE_ERR_INVAL_SIZE;
+ goto error_tmp;
+ }
+
+ /* update package */
+ status = ice_update_pkg(hw, ice_pkg_buf(b), 1);
+ if (status == ICE_ERR_AQ_ERROR)
+ ice_debug(hw, ICE_DBG_INIT, "Unable to update HW profile\n");
+
+error_tmp:
+ ice_pkg_buf_free(hw, b);
+ return status;
+}
+
/**
* ice_add_prof - add profile
* @hw: pointer to the HW struct
@@ -3097,6 +3697,10 @@ ice_add_prof_id_flow(struct ice_hw *hw, enum ice_block blk, u16 vsi, u64 hdl)
}
}
+ /* update hardware */
+ if (!status)
+ status = ice_upd_prof_hw(hw, blk, &chg);
+
err_ice_add_prof_id_flow:
list_for_each_entry_safe(del, tmp, &chg, list_entry) {
list_del(&del->list_entry);
diff --git a/drivers/net/ethernet/intel/ice/ice_flex_type.h b/drivers/net/ethernet/intel/ice/ice_flex_type.h
index 9d95d51bc760..0fb3fe3ff3ea 100644
--- a/drivers/net/ethernet/intel/ice/ice_flex_type.h
+++ b/drivers/net/ethernet/intel/ice/ice_flex_type.h
@@ -108,37 +108,57 @@ struct ice_buf_hdr {
sizeof(struct ice_buf_hdr) - (hd_sz)) / (ent_sz))
/* ice package section IDs */
+#define ICE_SID_XLT0_SW 10
+#define ICE_SID_XLT_KEY_BUILDER_SW 11
#define ICE_SID_XLT1_SW 12
#define ICE_SID_XLT2_SW 13
#define ICE_SID_PROFID_TCAM_SW 14
#define ICE_SID_PROFID_REDIR_SW 15
#define ICE_SID_FLD_VEC_SW 16
+#define ICE_SID_CDID_KEY_BUILDER_SW 17
+#define ICE_SID_CDID_REDIR_SW 18
+#define ICE_SID_XLT0_ACL 20
+#define ICE_SID_XLT_KEY_BUILDER_ACL 21
#define ICE_SID_XLT1_ACL 22
#define ICE_SID_XLT2_ACL 23
#define ICE_SID_PROFID_TCAM_ACL 24
#define ICE_SID_PROFID_REDIR_ACL 25
#define ICE_SID_FLD_VEC_ACL 26
+#define ICE_SID_CDID_KEY_BUILDER_ACL 27
+#define ICE_SID_CDID_REDIR_ACL 28
+#define ICE_SID_XLT0_FD 30
+#define ICE_SID_XLT_KEY_BUILDER_FD 31
#define ICE_SID_XLT1_FD 32
#define ICE_SID_XLT2_FD 33
#define ICE_SID_PROFID_TCAM_FD 34
#define ICE_SID_PROFID_REDIR_FD 35
#define ICE_SID_FLD_VEC_FD 36
+#define ICE_SID_CDID_KEY_BUILDER_FD 37
+#define ICE_SID_CDID_REDIR_FD 38
+#define ICE_SID_XLT0_RSS 40
+#define ICE_SID_XLT_KEY_BUILDER_RSS 41
#define ICE_SID_XLT1_RSS 42
#define ICE_SID_XLT2_RSS 43
#define ICE_SID_PROFID_TCAM_RSS 44
#define ICE_SID_PROFID_REDIR_RSS 45
#define ICE_SID_FLD_VEC_RSS 46
+#define ICE_SID_CDID_KEY_BUILDER_RSS 47
+#define ICE_SID_CDID_REDIR_RSS 48
#define ICE_SID_RXPARSER_BOOST_TCAM 56
+#define ICE_SID_XLT0_PE 80
+#define ICE_SID_XLT_KEY_BUILDER_PE 81
#define ICE_SID_XLT1_PE 82
#define ICE_SID_XLT2_PE 83
#define ICE_SID_PROFID_TCAM_PE 84
#define ICE_SID_PROFID_REDIR_PE 85
#define ICE_SID_FLD_VEC_PE 86
+#define ICE_SID_CDID_KEY_BUILDER_PE 87
+#define ICE_SID_CDID_REDIR_PE 88
/* Label Metadata section IDs */
#define ICE_SID_LBL_FIRST 0x80000010
@@ -155,6 +175,19 @@ enum ice_block {
ICE_BLK_COUNT
};
+enum ice_sect {
+ ICE_XLT0 = 0,
+ ICE_XLT_KB,
+ ICE_XLT1,
+ ICE_XLT2,
+ ICE_PROF_TCAM,
+ ICE_PROF_REDIR,
+ ICE_VEC_TBL,
+ ICE_CDID_KB,
+ ICE_CDID_REDIR,
+ ICE_SECT_COUNT
+};
+
/* package labels */
struct ice_label {
__le16 value;
@@ -237,6 +270,13 @@ struct ice_prof_redir_section {
u8 redir_value[1];
};
+/* package buffer building */
+
+struct ice_buf_build {
+ struct ice_buf buf;
+ u16 reserved_section_table_entries;
+};
+
struct ice_pkg_enum {
struct ice_buf_table *buf_table;
u32 buf_idx;
@@ -251,6 +291,12 @@ struct ice_pkg_enum {
void *(*handler)(u32 sect_type, void *section, u32 index, u32 *offset);
};
+struct ice_pkg_es {
+ __le16 count;
+ __le16 offset;
+ struct ice_fv_word es[1];
+};
+
struct ice_es {
u32 sid;
u16 count;
--
2.24.1
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [net-next 5/8] ice: Optimize table usage
2020-01-26 6:07 [net-next 0/8][pull request] 100GbE Intel Wired LAN Driver Updates 2020-01-25 Jeff Kirsher
` (3 preceding siblings ...)
2020-01-26 6:07 ` [net-next 4/8] ice: Enable writing filtering tables Jeff Kirsher
@ 2020-01-26 6:07 ` Jeff Kirsher
2020-01-26 6:07 ` [net-next 6/8] ice: Initilialize VF RSS tables Jeff Kirsher
` (3 subsequent siblings)
8 siblings, 0 replies; 10+ messages in thread
From: Jeff Kirsher @ 2020-01-26 6:07 UTC (permalink / raw)
To: davem
Cc: Tony Nguyen, netdev, nhorman, sassmann, Henry Tieman,
Andrew Bowers, Jeff Kirsher
From: Tony Nguyen <anthony.l.nguyen@intel.com>
Attempt to optimize TCAM entries and reduce table resource usage by
searching for profiles that can be reused. Provide resource cleanup
of both hardware and software structures.
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Henry Tieman <henry.w.tieman@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
.../net/ethernet/intel/ice/ice_flex_pipe.c | 394 ++++++++++++++++++
.../net/ethernet/intel/ice/ice_flex_pipe.h | 4 +
drivers/net/ethernet/intel/ice/ice_flow.c | 338 +++++++++++++++
drivers/net/ethernet/intel/ice/ice_flow.h | 10 +
drivers/net/ethernet/intel/ice/ice_lib.c | 28 +-
5 files changed, 773 insertions(+), 1 deletion(-)
diff --git a/drivers/net/ethernet/intel/ice/ice_flex_pipe.c b/drivers/net/ethernet/intel/ice/ice_flex_pipe.c
index 6dca611b408a..99208946224c 100644
--- a/drivers/net/ethernet/intel/ice/ice_flex_pipe.c
+++ b/drivers/net/ethernet/intel/ice/ice_flex_pipe.c
@@ -1921,6 +1921,26 @@ ice_alloc_prof_id(struct ice_hw *hw, enum ice_block blk, u8 *prof_id)
return status;
}
+/**
+ * ice_free_prof_id - free profile ID
+ * @hw: pointer to the HW struct
+ * @blk: the block from which to free the profile ID
+ * @prof_id: the profile ID to free
+ *
+ * This function frees a profile ID, which also corresponds to a Field Vector.
+ */
+static enum ice_status
+ice_free_prof_id(struct ice_hw *hw, enum ice_block blk, u8 prof_id)
+{
+ u16 tmp_prof_id = (u16)prof_id;
+ u16 res_type;
+
+ if (!ice_prof_id_rsrc_type(blk, &res_type))
+ return ICE_ERR_PARAM;
+
+ return ice_free_hw_res(hw, res_type, 1, &tmp_prof_id);
+}
+
/**
* ice_prof_inc_ref - increment reference count for profile
* @hw: pointer to the HW struct
@@ -1962,6 +1982,28 @@ ice_write_es(struct ice_hw *hw, enum ice_block blk, u8 prof_id,
}
}
+/**
+ * ice_prof_dec_ref - decrement reference count for profile
+ * @hw: pointer to the HW struct
+ * @blk: the block from which to free the profile ID
+ * @prof_id: the profile ID for which to decrement the reference count
+ */
+static enum ice_status
+ice_prof_dec_ref(struct ice_hw *hw, enum ice_block blk, u8 prof_id)
+{
+ if (prof_id > hw->blk[blk].es.count)
+ return ICE_ERR_PARAM;
+
+ if (hw->blk[blk].es.ref_count[prof_id] > 0) {
+ if (!--hw->blk[blk].es.ref_count[prof_id]) {
+ ice_write_es(hw, blk, prof_id, NULL);
+ return ice_free_prof_id(hw, blk, prof_id);
+ }
+ }
+
+ return 0;
+}
+
/* Block / table section IDs */
static const u32 ice_blk_sids[ICE_BLK_COUNT][ICE_SID_OFF_COUNT] = {
/* SWITCH */
@@ -2218,6 +2260,64 @@ void ice_fill_blk_tbls(struct ice_hw *hw)
ice_init_sw_db(hw);
}
+/**
+ * ice_free_prof_map - free profile map
+ * @hw: pointer to the hardware structure
+ * @blk_idx: HW block index
+ */
+static void ice_free_prof_map(struct ice_hw *hw, u8 blk_idx)
+{
+ struct ice_es *es = &hw->blk[blk_idx].es;
+ struct ice_prof_map *del, *tmp;
+
+ mutex_lock(&es->prof_map_lock);
+ list_for_each_entry_safe(del, tmp, &es->prof_map, list) {
+ list_del(&del->list);
+ devm_kfree(ice_hw_to_dev(hw), del);
+ }
+ INIT_LIST_HEAD(&es->prof_map);
+ mutex_unlock(&es->prof_map_lock);
+}
+
+/**
+ * ice_free_flow_profs - free flow profile entries
+ * @hw: pointer to the hardware structure
+ * @blk_idx: HW block index
+ */
+static void ice_free_flow_profs(struct ice_hw *hw, u8 blk_idx)
+{
+ struct ice_flow_prof *p, *tmp;
+
+ mutex_lock(&hw->fl_profs_locks[blk_idx]);
+ list_for_each_entry_safe(p, tmp, &hw->fl_profs[blk_idx], l_entry) {
+ list_del(&p->l_entry);
+ devm_kfree(ice_hw_to_dev(hw), p);
+ }
+ mutex_unlock(&hw->fl_profs_locks[blk_idx]);
+
+ /* if driver is in reset and tables are being cleared
+ * re-initialize the flow profile list heads
+ */
+ INIT_LIST_HEAD(&hw->fl_profs[blk_idx]);
+}
+
+/**
+ * ice_free_vsig_tbl - free complete VSIG table entries
+ * @hw: pointer to the hardware structure
+ * @blk: the HW block on which to free the VSIG table entries
+ */
+static void ice_free_vsig_tbl(struct ice_hw *hw, enum ice_block blk)
+{
+ u16 i;
+
+ if (!hw->blk[blk].xlt2.vsig_tbl)
+ return;
+
+ for (i = 1; i < ICE_MAX_VSIGS; i++)
+ if (hw->blk[blk].xlt2.vsig_tbl[i].in_use)
+ ice_vsig_free(hw, blk, i);
+}
+
/**
* ice_free_hw_tbls - free hardware table memory
* @hw: pointer to the hardware structure
@@ -2231,11 +2331,15 @@ void ice_free_hw_tbls(struct ice_hw *hw)
if (hw->blk[i].is_list_init) {
struct ice_es *es = &hw->blk[i].es;
+ ice_free_prof_map(hw, i);
mutex_destroy(&es->prof_map_lock);
+
+ ice_free_flow_profs(hw, i);
mutex_destroy(&hw->fl_profs_locks[i]);
hw->blk[i].is_list_init = false;
}
+ ice_free_vsig_tbl(hw, (enum ice_block)i);
devm_kfree(ice_hw_to_dev(hw), hw->blk[i].xlt1.ptypes);
devm_kfree(ice_hw_to_dev(hw), hw->blk[i].xlt1.ptg_tbl);
devm_kfree(ice_hw_to_dev(hw), hw->blk[i].xlt1.t);
@@ -2283,6 +2387,13 @@ void ice_clear_hw_tbls(struct ice_hw *hw)
struct ice_xlt2 *xlt2 = &hw->blk[i].xlt2;
struct ice_es *es = &hw->blk[i].es;
+ if (hw->blk[i].is_list_init) {
+ ice_free_prof_map(hw, i);
+ ice_free_flow_profs(hw, i);
+ }
+
+ ice_free_vsig_tbl(hw, (enum ice_block)i);
+
memset(xlt1->ptypes, 0, xlt1->count * sizeof(*xlt1->ptypes));
memset(xlt1->ptg_tbl, 0,
ICE_MAX_PTGS * sizeof(*xlt1->ptg_tbl));
@@ -2956,6 +3067,25 @@ ice_search_prof_id(struct ice_hw *hw, enum ice_block blk, u64 id)
return entry;
}
+/**
+ * ice_vsig_prof_id_count - count profiles in a VSIG
+ * @hw: pointer to the HW struct
+ * @blk: hardware block
+ * @vsig: VSIG to remove the profile from
+ */
+static u16
+ice_vsig_prof_id_count(struct ice_hw *hw, enum ice_block blk, u16 vsig)
+{
+ u16 idx = vsig & ICE_VSIG_IDX_M, count = 0;
+ struct ice_vsig_prof *p;
+
+ list_for_each_entry(p, &hw->blk[blk].xlt2.vsig_tbl[idx].prop_lst,
+ list)
+ count++;
+
+ return count;
+}
+
/**
* ice_rel_tcam_idx - release a TCAM index
* @hw: pointer to the HW struct
@@ -3064,6 +3194,117 @@ ice_rem_vsig(struct ice_hw *hw, enum ice_block blk, u16 vsig,
return ice_vsig_free(hw, blk, vsig);
}
+/**
+ * ice_rem_prof_id_vsig - remove a specific profile from a VSIG
+ * @hw: pointer to the HW struct
+ * @blk: hardware block
+ * @vsig: VSIG to remove the profile from
+ * @hdl: profile handle indicating which profile to remove
+ * @chg: list to receive a record of changes
+ */
+static enum ice_status
+ice_rem_prof_id_vsig(struct ice_hw *hw, enum ice_block blk, u16 vsig, u64 hdl,
+ struct list_head *chg)
+{
+ u16 idx = vsig & ICE_VSIG_IDX_M;
+ struct ice_vsig_prof *p, *t;
+ enum ice_status status;
+
+ list_for_each_entry_safe(p, t,
+ &hw->blk[blk].xlt2.vsig_tbl[idx].prop_lst,
+ list)
+ if (p->profile_cookie == hdl) {
+ if (ice_vsig_prof_id_count(hw, blk, vsig) == 1)
+ /* this is the last profile, remove the VSIG */
+ return ice_rem_vsig(hw, blk, vsig, chg);
+
+ status = ice_rem_prof_id(hw, blk, p);
+ if (!status) {
+ list_del(&p->list);
+ devm_kfree(ice_hw_to_dev(hw), p);
+ }
+ return status;
+ }
+
+ return ICE_ERR_DOES_NOT_EXIST;
+}
+
+/**
+ * ice_rem_flow_all - remove all flows with a particular profile
+ * @hw: pointer to the HW struct
+ * @blk: hardware block
+ * @id: profile tracking ID
+ */
+static enum ice_status
+ice_rem_flow_all(struct ice_hw *hw, enum ice_block blk, u64 id)
+{
+ struct ice_chs_chg *del, *tmp;
+ enum ice_status status;
+ struct list_head chg;
+ u16 i;
+
+ INIT_LIST_HEAD(&chg);
+
+ for (i = 1; i < ICE_MAX_VSIGS; i++)
+ if (hw->blk[blk].xlt2.vsig_tbl[i].in_use) {
+ if (ice_has_prof_vsig(hw, blk, i, id)) {
+ status = ice_rem_prof_id_vsig(hw, blk, i, id,
+ &chg);
+ if (status)
+ goto err_ice_rem_flow_all;
+ }
+ }
+
+ status = ice_upd_prof_hw(hw, blk, &chg);
+
+err_ice_rem_flow_all:
+ list_for_each_entry_safe(del, tmp, &chg, list_entry) {
+ list_del(&del->list_entry);
+ devm_kfree(ice_hw_to_dev(hw), del);
+ }
+
+ return status;
+}
+
+/**
+ * ice_rem_prof - remove profile
+ * @hw: pointer to the HW struct
+ * @blk: hardware block
+ * @id: profile tracking ID
+ *
+ * This will remove the profile specified by the ID parameter, which was
+ * previously created through ice_add_prof. If any existing entries
+ * are associated with this profile, they will be removed as well.
+ */
+enum ice_status ice_rem_prof(struct ice_hw *hw, enum ice_block blk, u64 id)
+{
+ struct ice_prof_map *pmap;
+ enum ice_status status;
+
+ mutex_lock(&hw->blk[blk].es.prof_map_lock);
+
+ pmap = ice_search_prof_id_low(hw, blk, id);
+ if (!pmap) {
+ status = ICE_ERR_DOES_NOT_EXIST;
+ goto err_ice_rem_prof;
+ }
+
+ /* remove all flows with this profile */
+ status = ice_rem_flow_all(hw, blk, pmap->profile_cookie);
+ if (status)
+ goto err_ice_rem_prof;
+
+ /* dereference profile, and possibly remove */
+ ice_prof_dec_ref(hw, blk, pmap->prof_id);
+
+ list_del(&pmap->list);
+ devm_kfree(ice_hw_to_dev(hw), pmap);
+
+err_ice_rem_prof:
+ mutex_unlock(&hw->blk[blk].es.prof_map_lock);
+ return status;
+}
+
/**
* ice_get_prof - get profile
* @hw: pointer to the HW struct
@@ -3714,3 +3955,156 @@ ice_add_prof_id_flow(struct ice_hw *hw, enum ice_block blk, u16 vsi, u64 hdl)
return status;
}
+
+/**
+ * ice_rem_prof_from_list - remove a profile from list
+ * @hw: pointer to the HW struct
+ * @lst: list to remove the profile from
+ * @hdl: the profile handle indicating the profile to remove
+ */
+static enum ice_status
+ice_rem_prof_from_list(struct ice_hw *hw, struct list_head *lst, u64 hdl)
+{
+ struct ice_vsig_prof *ent, *tmp;
+
+ list_for_each_entry_safe(ent, tmp, lst, list)
+ if (ent->profile_cookie == hdl) {
+ list_del(&ent->list);
+ devm_kfree(ice_hw_to_dev(hw), ent);
+ return 0;
+ }
+
+ return ICE_ERR_DOES_NOT_EXIST;
+}
+
+/**
+ * ice_rem_prof_id_flow - remove flow
+ * @hw: pointer to the HW struct
+ * @blk: hardware block
+ * @vsi: the VSI from which to remove the profile specified by ID
+ * @hdl: profile tracking handle
+ *
+ * Calling this function will update the hardware tables to remove the
+ * profile indicated by the ID parameter for the VSIs specified in the VSI
+ * array. Once successfully called, the flow will be disabled.
+ */
+enum ice_status
+ice_rem_prof_id_flow(struct ice_hw *hw, enum ice_block blk, u16 vsi, u64 hdl)
+{
+ struct ice_vsig_prof *tmp1, *del1;
+ struct ice_chs_chg *tmp, *del;
+ struct list_head chg, copy;
+ enum ice_status status;
+ u16 vsig;
+
+ INIT_LIST_HEAD(©);
+ INIT_LIST_HEAD(&chg);
+
+ /* determine if VSI is already part of a VSIG */
+ status = ice_vsig_find_vsi(hw, blk, vsi, &vsig);
+ if (!status && vsig) {
+ bool last_profile;
+ bool only_vsi;
+ u16 ref;
+
+ /* found in VSIG */
+ last_profile = ice_vsig_prof_id_count(hw, blk, vsig) == 1;
+ status = ice_vsig_get_ref(hw, blk, vsig, &ref);
+ if (status)
+ goto err_ice_rem_prof_id_flow;
+ only_vsi = (ref == 1);
+
+ if (only_vsi) {
+ /* If the original VSIG only contains one reference,
+ * which will be the requesting VSI, then the VSI is not
+ * sharing entries and we can simply remove the specific
+ * characteristics from the VSIG.
+ */
+
+ if (last_profile) {
+ /* If there are no profiles left for this VSIG,
+ * then simply remove the the VSIG.
+ */
+ status = ice_rem_vsig(hw, blk, vsig, &chg);
+ if (status)
+ goto err_ice_rem_prof_id_flow;
+ } else {
+ status = ice_rem_prof_id_vsig(hw, blk, vsig,
+ hdl, &chg);
+ if (status)
+ goto err_ice_rem_prof_id_flow;
+
+ /* Adjust priorities */
+ status = ice_adj_prof_priorities(hw, blk, vsig,
+ &chg);
+ if (status)
+ goto err_ice_rem_prof_id_flow;
+ }
+
+ } else {
+ /* Make a copy of the VSIG's list of Profiles */
+ status = ice_get_profs_vsig(hw, blk, vsig, ©);
+ if (status)
+ goto err_ice_rem_prof_id_flow;
+
+ /* Remove specified profile entry from the list */
+ status = ice_rem_prof_from_list(hw, ©, hdl);
+ if (status)
+ goto err_ice_rem_prof_id_flow;
+
+ if (list_empty(©)) {
+ status = ice_move_vsi(hw, blk, vsi,
+ ICE_DEFAULT_VSIG, &chg);
+ if (status)
+ goto err_ice_rem_prof_id_flow;
+
+ } else if (!ice_find_dup_props_vsig(hw, blk, ©,
+ &vsig)) {
+ /* found an exact match */
+ /* add or move VSI to the VSIG that matches */
+ /* Search for a VSIG with a matching profile
+ * list
+ */
+
+ /* Found match, move VSI to the matching VSIG */
+ status = ice_move_vsi(hw, blk, vsi, vsig, &chg);
+ if (status)
+ goto err_ice_rem_prof_id_flow;
+ } else {
+ /* since no existing VSIG supports this
+ * characteristic pattern, we need to create a
+ * new VSIG and TCAM entries
+ */
+ status = ice_create_vsig_from_lst(hw, blk, vsi,
+ ©, &chg);
+ if (status)
+ goto err_ice_rem_prof_id_flow;
+
+ /* Adjust priorities */
+ status = ice_adj_prof_priorities(hw, blk, vsig,
+ &chg);
+ if (status)
+ goto err_ice_rem_prof_id_flow;
+ }
+ }
+ } else {
+ status = ICE_ERR_DOES_NOT_EXIST;
+ }
+
+ /* update hardware tables */
+ if (!status)
+ status = ice_upd_prof_hw(hw, blk, &chg);
+
+err_ice_rem_prof_id_flow:
+ list_for_each_entry_safe(del, tmp, &chg, list_entry) {
+ list_del(&del->list_entry);
+ devm_kfree(ice_hw_to_dev(hw), del);
+ }
+
+ list_for_each_entry_safe(del1, tmp1, ©, list) {
+ list_del(&del1->list);
+ devm_kfree(ice_hw_to_dev(hw), del1);
+ }
+
+ return status;
+}
diff --git a/drivers/net/ethernet/intel/ice/ice_flex_pipe.h b/drivers/net/ethernet/intel/ice/ice_flex_pipe.h
index 33e4510da24c..c7b5e1a6ea2b 100644
--- a/drivers/net/ethernet/intel/ice/ice_flex_pipe.h
+++ b/drivers/net/ethernet/intel/ice/ice_flex_pipe.h
@@ -23,6 +23,8 @@ ice_add_prof(struct ice_hw *hw, enum ice_block blk, u64 id, u8 ptypes[],
struct ice_fv_word *es);
enum ice_status
ice_add_prof_id_flow(struct ice_hw *hw, enum ice_block blk, u16 vsi, u64 hdl);
+enum ice_status
+ice_rem_prof_id_flow(struct ice_hw *hw, enum ice_block blk, u16 vsi, u64 hdl);
enum ice_status ice_init_pkg(struct ice_hw *hw, u8 *buff, u32 len);
enum ice_status
ice_copy_and_init_pkg(struct ice_hw *hw, const u8 *buf, u32 len);
@@ -31,4 +33,6 @@ void ice_free_seg(struct ice_hw *hw);
void ice_fill_blk_tbls(struct ice_hw *hw);
void ice_clear_hw_tbls(struct ice_hw *hw);
void ice_free_hw_tbls(struct ice_hw *hw);
+enum ice_status
+ice_rem_prof(struct ice_hw *hw, enum ice_block blk, u64 id);
#endif /* _ICE_FLEX_PIPE_H_ */
diff --git a/drivers/net/ethernet/intel/ice/ice_flow.c b/drivers/net/ethernet/intel/ice/ice_flow.c
index 1bbf4b6ed7d2..ffce2284d8fa 100644
--- a/drivers/net/ethernet/intel/ice/ice_flow.c
+++ b/drivers/net/ethernet/intel/ice/ice_flow.c
@@ -377,6 +377,77 @@ ice_flow_proc_segs(struct ice_hw *hw, struct ice_flow_prof_params *params)
return status;
}
+#define ICE_FLOW_FIND_PROF_CHK_FLDS 0x00000001
+#define ICE_FLOW_FIND_PROF_CHK_VSI 0x00000002
+#define ICE_FLOW_FIND_PROF_NOT_CHK_DIR 0x00000004
+
+/**
+ * ice_flow_find_prof_conds - Find a profile matching headers and conditions
+ * @hw: pointer to the HW struct
+ * @blk: classification stage
+ * @dir: flow direction
+ * @segs: array of one or more packet segments that describe the flow
+ * @segs_cnt: number of packet segments provided
+ * @vsi_handle: software VSI handle to check VSI (ICE_FLOW_FIND_PROF_CHK_VSI)
+ * @conds: additional conditions to be checked (ICE_FLOW_FIND_PROF_CHK_*)
+ */
+static struct ice_flow_prof *
+ice_flow_find_prof_conds(struct ice_hw *hw, enum ice_block blk,
+ enum ice_flow_dir dir, struct ice_flow_seg_info *segs,
+ u8 segs_cnt, u16 vsi_handle, u32 conds)
+{
+ struct ice_flow_prof *p, *prof = NULL;
+
+ mutex_lock(&hw->fl_profs_locks[blk]);
+ list_for_each_entry(p, &hw->fl_profs[blk], l_entry)
+ if ((p->dir == dir || conds & ICE_FLOW_FIND_PROF_NOT_CHK_DIR) &&
+ segs_cnt && segs_cnt == p->segs_cnt) {
+ u8 i;
+
+ /* Check for profile-VSI association if specified */
+ if ((conds & ICE_FLOW_FIND_PROF_CHK_VSI) &&
+ ice_is_vsi_valid(hw, vsi_handle) &&
+ !test_bit(vsi_handle, p->vsis))
+ continue;
+
+ /* Protocol headers must be checked. Matched fields are
+ * checked if specified.
+ */
+ for (i = 0; i < segs_cnt; i++)
+ if (segs[i].hdrs != p->segs[i].hdrs ||
+ ((conds & ICE_FLOW_FIND_PROF_CHK_FLDS) &&
+ segs[i].match != p->segs[i].match))
+ break;
+
+ /* A match is found if all segments are matched */
+ if (i == segs_cnt) {
+ prof = p;
+ break;
+ }
+ }
+ mutex_unlock(&hw->fl_profs_locks[blk]);
+
+ return prof;
+}
+
+/**
+ * ice_flow_find_prof_id - Look up a profile with given profile ID
+ * @hw: pointer to the HW struct
+ * @blk: classification stage
+ * @prof_id: unique ID to identify this flow profile
+ */
+static struct ice_flow_prof *
+ice_flow_find_prof_id(struct ice_hw *hw, enum ice_block blk, u64 prof_id)
+{
+ struct ice_flow_prof *p;
+
+ list_for_each_entry(p, &hw->fl_profs[blk], l_entry)
+ if (p->id == prof_id)
+ return p;
+
+ return NULL;
+}
+
/**
* ice_flow_add_prof_sync - Add a flow profile for packet segments and fields
* @hw: pointer to the HW struct
@@ -450,6 +521,31 @@ ice_flow_add_prof_sync(struct ice_hw *hw, enum ice_block blk,
return status;
}
+/**
+ * ice_flow_rem_prof_sync - remove a flow profile
+ * @hw: pointer to the hardware structure
+ * @blk: classification stage
+ * @prof: pointer to flow profile to remove
+ *
+ * Assumption: the caller has acquired the lock to the profile list
+ */
+static enum ice_status
+ice_flow_rem_prof_sync(struct ice_hw *hw, enum ice_block blk,
+ struct ice_flow_prof *prof)
+{
+ enum ice_status status;
+
+ /* Remove all hardware profiles associated with this flow profile */
+ status = ice_rem_prof(hw, blk, prof->id);
+ if (!status) {
+ list_del(&prof->l_entry);
+ mutex_destroy(&prof->entries_lock);
+ devm_kfree(ice_hw_to_dev(hw), prof);
+ }
+
+ return status;
+}
+
/**
* ice_flow_assoc_prof - associate a VSI with a flow profile
* @hw: pointer to the hardware structure
@@ -482,6 +578,38 @@ ice_flow_assoc_prof(struct ice_hw *hw, enum ice_block blk,
return status;
}
+/**
+ * ice_flow_disassoc_prof - disassociate a VSI from a flow profile
+ * @hw: pointer to the hardware structure
+ * @blk: classification stage
+ * @prof: pointer to flow profile
+ * @vsi_handle: software VSI handle
+ *
+ * Assumption: the caller has acquired the lock to the profile list
+ * and the software VSI handle has been validated
+ */
+static enum ice_status
+ice_flow_disassoc_prof(struct ice_hw *hw, enum ice_block blk,
+ struct ice_flow_prof *prof, u16 vsi_handle)
+{
+ enum ice_status status = 0;
+
+ if (test_bit(vsi_handle, prof->vsis)) {
+ status = ice_rem_prof_id_flow(hw, blk,
+ ice_get_hw_vsi_num(hw,
+ vsi_handle),
+ prof->id);
+ if (!status)
+ clear_bit(vsi_handle, prof->vsis);
+ else
+ ice_debug(hw, ICE_DBG_FLOW,
+ "HW profile remove failed, %d\n",
+ status);
+ }
+
+ return status;
+}
+
/**
* ice_flow_add_prof - Add a flow profile for packet segments and matched fields
* @hw: pointer to the HW struct
@@ -524,6 +652,35 @@ ice_flow_add_prof(struct ice_hw *hw, enum ice_block blk, enum ice_flow_dir dir,
return status;
}
+/**
+ * ice_flow_rem_prof - Remove a flow profile and all entries associated with it
+ * @hw: pointer to the HW struct
+ * @blk: the block for which the flow profile is to be removed
+ * @prof_id: unique ID of the flow profile to be removed
+ */
+static enum ice_status
+ice_flow_rem_prof(struct ice_hw *hw, enum ice_block blk, u64 prof_id)
+{
+ struct ice_flow_prof *prof;
+ enum ice_status status;
+
+ mutex_lock(&hw->fl_profs_locks[blk]);
+
+ prof = ice_flow_find_prof_id(hw, blk, prof_id);
+ if (!prof) {
+ status = ICE_ERR_DOES_NOT_EXIST;
+ goto out;
+ }
+
+ /* prof becomes invalid after the call */
+ status = ice_flow_rem_prof_sync(hw, blk, prof);
+
+out:
+ mutex_unlock(&hw->fl_profs_locks[blk]);
+
+ return status;
+}
+
/**
* ice_flow_set_fld_ext - specifies locations of field from entry's input buffer
* @seg: packet segment the field being set belongs to
@@ -645,6 +802,132 @@ ice_flow_set_rss_seg_info(struct ice_flow_seg_info *segs, u64 hash_fields,
return 0;
}
+/**
+ * ice_rem_vsi_rss_list - remove VSI from RSS list
+ * @hw: pointer to the hardware structure
+ * @vsi_handle: software VSI handle
+ *
+ * Remove the VSI from all RSS configurations in the list.
+ */
+void ice_rem_vsi_rss_list(struct ice_hw *hw, u16 vsi_handle)
+{
+ struct ice_rss_cfg *r, *tmp;
+
+ if (list_empty(&hw->rss_list_head))
+ return;
+
+ mutex_lock(&hw->rss_locks);
+ list_for_each_entry_safe(r, tmp, &hw->rss_list_head, l_entry)
+ if (test_and_clear_bit(vsi_handle, r->vsis))
+ if (bitmap_empty(r->vsis, ICE_MAX_VSI)) {
+ list_del(&r->l_entry);
+ devm_kfree(ice_hw_to_dev(hw), r);
+ }
+ mutex_unlock(&hw->rss_locks);
+}
+
+/**
+ * ice_rem_vsi_rss_cfg - remove RSS configurations associated with VSI
+ * @hw: pointer to the hardware structure
+ * @vsi_handle: software VSI handle
+ *
+ * This function will iterate through all flow profiles and disassociate
+ * the VSI from that profile. If the flow profile has no VSIs it will
+ * be removed.
+ */
+enum ice_status ice_rem_vsi_rss_cfg(struct ice_hw *hw, u16 vsi_handle)
+{
+ const enum ice_block blk = ICE_BLK_RSS;
+ struct ice_flow_prof *p, *t;
+ enum ice_status status = 0;
+
+ if (!ice_is_vsi_valid(hw, vsi_handle))
+ return ICE_ERR_PARAM;
+
+ if (list_empty(&hw->fl_profs[blk]))
+ return 0;
+
+ mutex_lock(&hw->fl_profs_locks[blk]);
+ list_for_each_entry_safe(p, t, &hw->fl_profs[blk], l_entry)
+ if (test_bit(vsi_handle, p->vsis)) {
+ status = ice_flow_disassoc_prof(hw, blk, p, vsi_handle);
+ if (status)
+ break;
+
+ if (bitmap_empty(p->vsis, ICE_MAX_VSI)) {
+ status = ice_flow_rem_prof_sync(hw, blk, p);
+ if (status)
+ break;
+ }
+ }
+ mutex_unlock(&hw->fl_profs_locks[blk]);
+
+ return status;
+}
+
+/**
+ * ice_rem_rss_list - remove RSS configuration from list
+ * @hw: pointer to the hardware structure
+ * @vsi_handle: software VSI handle
+ * @prof: pointer to flow profile
+ *
+ * Assumption: lock has already been acquired for RSS list
+ */
+static void
+ice_rem_rss_list(struct ice_hw *hw, u16 vsi_handle, struct ice_flow_prof *prof)
+{
+ struct ice_rss_cfg *r, *tmp;
+
+ /* Search for RSS hash fields associated to the VSI that match the
+ * hash configurations associated to the flow profile. If found
+ * remove from the RSS entry list of the VSI context and delete entry.
+ */
+ list_for_each_entry_safe(r, tmp, &hw->rss_list_head, l_entry)
+ if (r->hashed_flds == prof->segs[prof->segs_cnt - 1].match &&
+ r->packet_hdr == prof->segs[prof->segs_cnt - 1].hdrs) {
+ clear_bit(vsi_handle, r->vsis);
+ if (bitmap_empty(r->vsis, ICE_MAX_VSI)) {
+ list_del(&r->l_entry);
+ devm_kfree(ice_hw_to_dev(hw), r);
+ }
+ return;
+ }
+}
+
+/**
+ * ice_add_rss_list - add RSS configuration to list
+ * @hw: pointer to the hardware structure
+ * @vsi_handle: software VSI handle
+ * @prof: pointer to flow profile
+ *
+ * Assumption: lock has already been acquired for RSS list
+ */
+static enum ice_status
+ice_add_rss_list(struct ice_hw *hw, u16 vsi_handle, struct ice_flow_prof *prof)
+{
+ struct ice_rss_cfg *r, *rss_cfg;
+
+ list_for_each_entry(r, &hw->rss_list_head, l_entry)
+ if (r->hashed_flds == prof->segs[prof->segs_cnt - 1].match &&
+ r->packet_hdr == prof->segs[prof->segs_cnt - 1].hdrs) {
+ set_bit(vsi_handle, r->vsis);
+ return 0;
+ }
+
+ rss_cfg = devm_kzalloc(ice_hw_to_dev(hw), sizeof(*rss_cfg),
+ GFP_KERNEL);
+ if (!rss_cfg)
+ return ICE_ERR_NO_MEMORY;
+
+ rss_cfg->hashed_flds = prof->segs[prof->segs_cnt - 1].match;
+ rss_cfg->packet_hdr = prof->segs[prof->segs_cnt - 1].hdrs;
+ set_bit(vsi_handle, rss_cfg->vsis);
+
+ list_add_tail(&rss_cfg->l_entry, &hw->rss_list_head);
+
+ return 0;
+}
+
#define ICE_FLOW_PROF_HASH_S 0
#define ICE_FLOW_PROF_HASH_M (0xFFFFFFFFULL << ICE_FLOW_PROF_HASH_S)
#define ICE_FLOW_PROF_HDR_S 32
@@ -696,6 +979,52 @@ ice_add_rss_cfg_sync(struct ice_hw *hw, u16 vsi_handle, u64 hashed_flds,
if (status)
goto exit;
+ /* Search for a flow profile that has matching headers, hash fields
+ * and has the input VSI associated to it. If found, no further
+ * operations required and exit.
+ */
+ prof = ice_flow_find_prof_conds(hw, blk, ICE_FLOW_RX, segs, segs_cnt,
+ vsi_handle,
+ ICE_FLOW_FIND_PROF_CHK_FLDS |
+ ICE_FLOW_FIND_PROF_CHK_VSI);
+ if (prof)
+ goto exit;
+
+ /* Check if a flow profile exists with the same protocol headers and
+ * associated with the input VSI. If so disassociate the VSI from
+ * this profile. The VSI will be added to a new profile created with
+ * the protocol header and new hash field configuration.
+ */
+ prof = ice_flow_find_prof_conds(hw, blk, ICE_FLOW_RX, segs, segs_cnt,
+ vsi_handle, ICE_FLOW_FIND_PROF_CHK_VSI);
+ if (prof) {
+ status = ice_flow_disassoc_prof(hw, blk, prof, vsi_handle);
+ if (!status)
+ ice_rem_rss_list(hw, vsi_handle, prof);
+ else
+ goto exit;
+
+ /* Remove profile if it has no VSIs associated */
+ if (bitmap_empty(prof->vsis, ICE_MAX_VSI)) {
+ status = ice_flow_rem_prof(hw, blk, prof->id);
+ if (status)
+ goto exit;
+ }
+ }
+
+ /* Search for a profile that has same match fields only. If this
+ * exists then associate the VSI to this profile.
+ */
+ prof = ice_flow_find_prof_conds(hw, blk, ICE_FLOW_RX, segs, segs_cnt,
+ vsi_handle,
+ ICE_FLOW_FIND_PROF_CHK_FLDS);
+ if (prof) {
+ status = ice_flow_assoc_prof(hw, blk, prof, vsi_handle);
+ if (!status)
+ status = ice_add_rss_list(hw, vsi_handle, prof);
+ goto exit;
+ }
+
/* Create a new flow profile with generated profile and packet
* segment information.
*/
@@ -708,6 +1037,15 @@ ice_add_rss_cfg_sync(struct ice_hw *hw, u16 vsi_handle, u64 hashed_flds,
goto exit;
status = ice_flow_assoc_prof(hw, blk, prof, vsi_handle);
+ /* If association to a new flow profile failed then this profile can
+ * be removed.
+ */
+ if (status) {
+ ice_flow_rem_prof(hw, blk, prof->id);
+ goto exit;
+ }
+
+ status = ice_add_rss_list(hw, vsi_handle, prof);
exit:
kfree(segs);
diff --git a/drivers/net/ethernet/intel/ice/ice_flow.h b/drivers/net/ethernet/intel/ice/ice_flow.h
index 05b0dab4793c..38669b077209 100644
--- a/drivers/net/ethernet/intel/ice/ice_flow.h
+++ b/drivers/net/ethernet/intel/ice/ice_flow.h
@@ -4,6 +4,7 @@
#ifndef _ICE_FLOW_H_
#define _ICE_FLOW_H_
+#define ICE_FLOW_ENTRY_HANDLE_INVAL 0
#define ICE_FLOW_FLD_OFF_INVAL 0xffff
/* Generate flow hash field from flow field type(s) */
@@ -60,6 +61,12 @@ enum ice_flow_dir {
ICE_FLOW_RX = 0x02,
};
+enum ice_flow_priority {
+ ICE_FLOW_PRIO_LOW,
+ ICE_FLOW_PRIO_NORMAL,
+ ICE_FLOW_PRIO_HIGH
+};
+
#define ICE_FLOW_SEG_MAX 2
#define ICE_FLOW_FV_EXTRACT_SZ 2
@@ -130,7 +137,10 @@ struct ice_rss_cfg {
u32 packet_hdr;
};
+enum ice_status ice_flow_rem_entry(struct ice_hw *hw, u64 entry_h);
+void ice_rem_vsi_rss_list(struct ice_hw *hw, u16 vsi_handle);
enum ice_status ice_replay_rss_cfg(struct ice_hw *hw, u16 vsi_handle);
+enum ice_status ice_rem_vsi_rss_cfg(struct ice_hw *hw, u16 vsi_handle);
enum ice_status
ice_add_rss_cfg(struct ice_hw *hw, u16 vsi_handle, u64 hashed_flds,
u32 addl_hdrs);
diff --git a/drivers/net/ethernet/intel/ice/ice_lib.c b/drivers/net/ethernet/intel/ice/ice_lib.c
index dd230c6664ee..caa3edd5d3d6 100644
--- a/drivers/net/ethernet/intel/ice/ice_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_lib.c
@@ -494,7 +494,28 @@ bool ice_is_safe_mode(struct ice_pf *pf)
}
/**
- * ice_rss_clean - Delete RSS related VSI structures that hold user inputs
+ * ice_vsi_clean_rss_flow_fld - Delete RSS configuration
+ * @vsi: the VSI being cleaned up
+ *
+ * This function deletes RSS input set for all flows that were configured
+ * for this VSI
+ */
+static void ice_vsi_clean_rss_flow_fld(struct ice_vsi *vsi)
+{
+ struct ice_pf *pf = vsi->back;
+ enum ice_status status;
+
+ if (ice_is_safe_mode(pf))
+ return;
+
+ status = ice_rem_vsi_rss_cfg(&pf->hw, vsi->idx);
+ if (status)
+ dev_dbg(ice_pf_to_dev(pf), "ice_rem_vsi_rss_cfg failed for vsi = %d, error = %d\n",
+ vsi->vsi_num, status);
+}
+
+/**
+ * ice_rss_clean - Delete RSS related VSI structures and configuration
* @vsi: the VSI being removed
*/
static void ice_rss_clean(struct ice_vsi *vsi)
@@ -508,6 +529,11 @@ static void ice_rss_clean(struct ice_vsi *vsi)
devm_kfree(dev, vsi->rss_hkey_user);
if (vsi->rss_lut_user)
devm_kfree(dev, vsi->rss_lut_user);
+
+ ice_vsi_clean_rss_flow_fld(vsi);
+ /* remove RSS replay list */
+ if (!ice_is_safe_mode(pf))
+ ice_rem_vsi_rss_list(&pf->hw, vsi->idx);
}
/**
--
2.24.1
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [net-next 6/8] ice: Initilialize VF RSS tables
2020-01-26 6:07 [net-next 0/8][pull request] 100GbE Intel Wired LAN Driver Updates 2020-01-25 Jeff Kirsher
` (4 preceding siblings ...)
2020-01-26 6:07 ` [net-next 5/8] ice: Optimize table usage Jeff Kirsher
@ 2020-01-26 6:07 ` Jeff Kirsher
2020-01-26 6:07 ` [net-next 7/8] ice: Implement ethtool get/set rx-flow-hash Jeff Kirsher
` (2 subsequent siblings)
8 siblings, 0 replies; 10+ messages in thread
From: Jeff Kirsher @ 2020-01-26 6:07 UTC (permalink / raw)
To: davem
Cc: Md Fahad Iqbal Polash, netdev, nhorman, sassmann, Tony Nguyen,
Andrew Bowers, Jeff Kirsher
From: Md Fahad Iqbal Polash <md.fahad.iqbal.polash@intel.com>
Set configuration for hardware RSS tables for VFs.
Signed-off-by: Md Fahad Iqbal Polash <md.fahad.iqbal.polash@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
drivers/net/ethernet/intel/ice/ice_flow.c | 135 ++++++++++++++++++
drivers/net/ethernet/intel/ice/ice_flow.h | 59 ++++++++
drivers/net/ethernet/intel/ice/ice_lib.c | 31 +++-
.../ethernet/intel/ice/ice_protocol_type.h | 1 +
4 files changed, 225 insertions(+), 1 deletion(-)
diff --git a/drivers/net/ethernet/intel/ice/ice_flow.c b/drivers/net/ethernet/intel/ice/ice_flow.c
index ffce2284d8fa..4ecac0ae565a 100644
--- a/drivers/net/ethernet/intel/ice/ice_flow.c
+++ b/drivers/net/ethernet/intel/ice/ice_flow.c
@@ -38,6 +38,11 @@ struct ice_flow_field_info ice_flds_info[ICE_FLOW_FIELD_IDX_MAX] = {
ICE_FLOW_FLD_INFO(ICE_FLOW_SEG_HDR_UDP, 0, sizeof(__be16)),
/* ICE_FLOW_FIELD_IDX_UDP_DST_PORT */
ICE_FLOW_FLD_INFO(ICE_FLOW_SEG_HDR_UDP, 2, sizeof(__be16)),
+ /* ICE_FLOW_FIELD_IDX_SCTP_SRC_PORT */
+ ICE_FLOW_FLD_INFO(ICE_FLOW_SEG_HDR_SCTP, 0, sizeof(__be16)),
+ /* ICE_FLOW_FIELD_IDX_SCTP_DST_PORT */
+ ICE_FLOW_FLD_INFO(ICE_FLOW_SEG_HDR_SCTP, 2, sizeof(__be16)),
+
};
/* Bitmaps indicating relevant packet types for a particular protocol header
@@ -266,6 +271,10 @@ ice_flow_xtract_fld(struct ice_hw *hw, struct ice_flow_prof_params *params,
case ICE_FLOW_FIELD_IDX_UDP_DST_PORT:
prot_id = ICE_PROT_UDP_IL_OR_S;
break;
+ case ICE_FLOW_FIELD_IDX_SCTP_SRC_PORT:
+ case ICE_FLOW_FIELD_IDX_SCTP_DST_PORT:
+ prot_id = ICE_PROT_SCTP_IL;
+ break;
default:
return ICE_ERR_NOT_IMPL;
}
@@ -1081,6 +1090,132 @@ ice_add_rss_cfg(struct ice_hw *hw, u16 vsi_handle, u64 hashed_flds,
return status;
}
+/* Mapping of AVF hash bit fields to an L3-L4 hash combination.
+ * As the ice_flow_avf_hdr_field represent individual bit shifts in a hash,
+ * convert its values to their appropriate flow L3, L4 values.
+ */
+#define ICE_FLOW_AVF_RSS_IPV4_MASKS \
+ (BIT_ULL(ICE_AVF_FLOW_FIELD_IPV4_OTHER) | \
+ BIT_ULL(ICE_AVF_FLOW_FIELD_FRAG_IPV4))
+#define ICE_FLOW_AVF_RSS_TCP_IPV4_MASKS \
+ (BIT_ULL(ICE_AVF_FLOW_FIELD_IPV4_TCP_SYN_NO_ACK) | \
+ BIT_ULL(ICE_AVF_FLOW_FIELD_IPV4_TCP))
+#define ICE_FLOW_AVF_RSS_UDP_IPV4_MASKS \
+ (BIT_ULL(ICE_AVF_FLOW_FIELD_UNICAST_IPV4_UDP) | \
+ BIT_ULL(ICE_AVF_FLOW_FIELD_MULTICAST_IPV4_UDP) | \
+ BIT_ULL(ICE_AVF_FLOW_FIELD_IPV4_UDP))
+#define ICE_FLOW_AVF_RSS_ALL_IPV4_MASKS \
+ (ICE_FLOW_AVF_RSS_TCP_IPV4_MASKS | ICE_FLOW_AVF_RSS_UDP_IPV4_MASKS | \
+ ICE_FLOW_AVF_RSS_IPV4_MASKS | BIT_ULL(ICE_AVF_FLOW_FIELD_IPV4_SCTP))
+
+#define ICE_FLOW_AVF_RSS_IPV6_MASKS \
+ (BIT_ULL(ICE_AVF_FLOW_FIELD_IPV6_OTHER) | \
+ BIT_ULL(ICE_AVF_FLOW_FIELD_FRAG_IPV6))
+#define ICE_FLOW_AVF_RSS_UDP_IPV6_MASKS \
+ (BIT_ULL(ICE_AVF_FLOW_FIELD_UNICAST_IPV6_UDP) | \
+ BIT_ULL(ICE_AVF_FLOW_FIELD_MULTICAST_IPV6_UDP) | \
+ BIT_ULL(ICE_AVF_FLOW_FIELD_IPV6_UDP))
+#define ICE_FLOW_AVF_RSS_TCP_IPV6_MASKS \
+ (BIT_ULL(ICE_AVF_FLOW_FIELD_IPV6_TCP_SYN_NO_ACK) | \
+ BIT_ULL(ICE_AVF_FLOW_FIELD_IPV6_TCP))
+#define ICE_FLOW_AVF_RSS_ALL_IPV6_MASKS \
+ (ICE_FLOW_AVF_RSS_TCP_IPV6_MASKS | ICE_FLOW_AVF_RSS_UDP_IPV6_MASKS | \
+ ICE_FLOW_AVF_RSS_IPV6_MASKS | BIT_ULL(ICE_AVF_FLOW_FIELD_IPV6_SCTP))
+
+/**
+ * ice_add_avf_rss_cfg - add an RSS configuration for AVF driver
+ * @hw: pointer to the hardware structure
+ * @vsi_handle: software VSI handle
+ * @avf_hash: hash bit fields (ICE_AVF_FLOW_FIELD_*) to configure
+ *
+ * This function will take the hash bitmap provided by the AVF driver via a
+ * message, convert it to ICE-compatible values, and configure RSS flow
+ * profiles.
+ */
+enum ice_status
+ice_add_avf_rss_cfg(struct ice_hw *hw, u16 vsi_handle, u64 avf_hash)
+{
+ enum ice_status status = 0;
+ u64 hash_flds;
+
+ if (avf_hash == ICE_AVF_FLOW_FIELD_INVALID ||
+ !ice_is_vsi_valid(hw, vsi_handle))
+ return ICE_ERR_PARAM;
+
+ /* Make sure no unsupported bits are specified */
+ if (avf_hash & ~(ICE_FLOW_AVF_RSS_ALL_IPV4_MASKS |
+ ICE_FLOW_AVF_RSS_ALL_IPV6_MASKS))
+ return ICE_ERR_CFG;
+
+ hash_flds = avf_hash;
+
+ /* Always create an L3 RSS configuration for any L4 RSS configuration */
+ if (hash_flds & ICE_FLOW_AVF_RSS_ALL_IPV4_MASKS)
+ hash_flds |= ICE_FLOW_AVF_RSS_IPV4_MASKS;
+
+ if (hash_flds & ICE_FLOW_AVF_RSS_ALL_IPV6_MASKS)
+ hash_flds |= ICE_FLOW_AVF_RSS_IPV6_MASKS;
+
+ /* Create the corresponding RSS configuration for each valid hash bit */
+ while (hash_flds) {
+ u64 rss_hash = ICE_HASH_INVALID;
+
+ if (hash_flds & ICE_FLOW_AVF_RSS_ALL_IPV4_MASKS) {
+ if (hash_flds & ICE_FLOW_AVF_RSS_IPV4_MASKS) {
+ rss_hash = ICE_FLOW_HASH_IPV4;
+ hash_flds &= ~ICE_FLOW_AVF_RSS_IPV4_MASKS;
+ } else if (hash_flds &
+ ICE_FLOW_AVF_RSS_TCP_IPV4_MASKS) {
+ rss_hash = ICE_FLOW_HASH_IPV4 |
+ ICE_FLOW_HASH_TCP_PORT;
+ hash_flds &= ~ICE_FLOW_AVF_RSS_TCP_IPV4_MASKS;
+ } else if (hash_flds &
+ ICE_FLOW_AVF_RSS_UDP_IPV4_MASKS) {
+ rss_hash = ICE_FLOW_HASH_IPV4 |
+ ICE_FLOW_HASH_UDP_PORT;
+ hash_flds &= ~ICE_FLOW_AVF_RSS_UDP_IPV4_MASKS;
+ } else if (hash_flds &
+ BIT_ULL(ICE_AVF_FLOW_FIELD_IPV4_SCTP)) {
+ rss_hash = ICE_FLOW_HASH_IPV4 |
+ ICE_FLOW_HASH_SCTP_PORT;
+ hash_flds &=
+ ~BIT_ULL(ICE_AVF_FLOW_FIELD_IPV4_SCTP);
+ }
+ } else if (hash_flds & ICE_FLOW_AVF_RSS_ALL_IPV6_MASKS) {
+ if (hash_flds & ICE_FLOW_AVF_RSS_IPV6_MASKS) {
+ rss_hash = ICE_FLOW_HASH_IPV6;
+ hash_flds &= ~ICE_FLOW_AVF_RSS_IPV6_MASKS;
+ } else if (hash_flds &
+ ICE_FLOW_AVF_RSS_TCP_IPV6_MASKS) {
+ rss_hash = ICE_FLOW_HASH_IPV6 |
+ ICE_FLOW_HASH_TCP_PORT;
+ hash_flds &= ~ICE_FLOW_AVF_RSS_TCP_IPV6_MASKS;
+ } else if (hash_flds &
+ ICE_FLOW_AVF_RSS_UDP_IPV6_MASKS) {
+ rss_hash = ICE_FLOW_HASH_IPV6 |
+ ICE_FLOW_HASH_UDP_PORT;
+ hash_flds &= ~ICE_FLOW_AVF_RSS_UDP_IPV6_MASKS;
+ } else if (hash_flds &
+ BIT_ULL(ICE_AVF_FLOW_FIELD_IPV6_SCTP)) {
+ rss_hash = ICE_FLOW_HASH_IPV6 |
+ ICE_FLOW_HASH_SCTP_PORT;
+ hash_flds &=
+ ~BIT_ULL(ICE_AVF_FLOW_FIELD_IPV6_SCTP);
+ }
+ }
+
+ if (rss_hash == ICE_HASH_INVALID)
+ return ICE_ERR_OUT_OF_RANGE;
+
+ status = ice_add_rss_cfg(hw, vsi_handle, rss_hash,
+ ICE_FLOW_SEG_HDR_NONE);
+ if (status)
+ break;
+ }
+
+ return status;
+}
+
/**
* ice_replay_rss_cfg - replay RSS configurations associated with VSI
* @hw: pointer to the hardware structure
diff --git a/drivers/net/ethernet/intel/ice/ice_flow.h b/drivers/net/ethernet/intel/ice/ice_flow.h
index 38669b077209..475a025d3a2d 100644
--- a/drivers/net/ethernet/intel/ice/ice_flow.h
+++ b/drivers/net/ethernet/intel/ice/ice_flow.h
@@ -20,6 +20,9 @@
#define ICE_FLOW_HASH_UDP_PORT \
(BIT_ULL(ICE_FLOW_FIELD_IDX_UDP_SRC_PORT) | \
BIT_ULL(ICE_FLOW_FIELD_IDX_UDP_DST_PORT))
+#define ICE_FLOW_HASH_SCTP_PORT \
+ (BIT_ULL(ICE_FLOW_FIELD_IDX_SCTP_SRC_PORT) | \
+ BIT_ULL(ICE_FLOW_FIELD_IDX_SCTP_DST_PORT))
#define ICE_HASH_INVALID 0
#define ICE_HASH_TCP_IPV4 (ICE_FLOW_HASH_IPV4 | ICE_FLOW_HASH_TCP_PORT)
@@ -53,10 +56,64 @@ enum ice_flow_field {
ICE_FLOW_FIELD_IDX_TCP_DST_PORT,
ICE_FLOW_FIELD_IDX_UDP_SRC_PORT,
ICE_FLOW_FIELD_IDX_UDP_DST_PORT,
+ ICE_FLOW_FIELD_IDX_SCTP_SRC_PORT,
+ ICE_FLOW_FIELD_IDX_SCTP_DST_PORT,
/* The total number of enums must not exceed 64 */
ICE_FLOW_FIELD_IDX_MAX
};
+/* Flow headers and fields for AVF support */
+enum ice_flow_avf_hdr_field {
+ /* Values 0 - 28 are reserved for future use */
+ ICE_AVF_FLOW_FIELD_INVALID = 0,
+ ICE_AVF_FLOW_FIELD_UNICAST_IPV4_UDP = 29,
+ ICE_AVF_FLOW_FIELD_MULTICAST_IPV4_UDP,
+ ICE_AVF_FLOW_FIELD_IPV4_UDP,
+ ICE_AVF_FLOW_FIELD_IPV4_TCP_SYN_NO_ACK,
+ ICE_AVF_FLOW_FIELD_IPV4_TCP,
+ ICE_AVF_FLOW_FIELD_IPV4_SCTP,
+ ICE_AVF_FLOW_FIELD_IPV4_OTHER,
+ ICE_AVF_FLOW_FIELD_FRAG_IPV4,
+ /* Values 37-38 are reserved */
+ ICE_AVF_FLOW_FIELD_UNICAST_IPV6_UDP = 39,
+ ICE_AVF_FLOW_FIELD_MULTICAST_IPV6_UDP,
+ ICE_AVF_FLOW_FIELD_IPV6_UDP,
+ ICE_AVF_FLOW_FIELD_IPV6_TCP_SYN_NO_ACK,
+ ICE_AVF_FLOW_FIELD_IPV6_TCP,
+ ICE_AVF_FLOW_FIELD_IPV6_SCTP,
+ ICE_AVF_FLOW_FIELD_IPV6_OTHER,
+ ICE_AVF_FLOW_FIELD_FRAG_IPV6,
+ ICE_AVF_FLOW_FIELD_RSVD47,
+ ICE_AVF_FLOW_FIELD_FCOE_OX,
+ ICE_AVF_FLOW_FIELD_FCOE_RX,
+ ICE_AVF_FLOW_FIELD_FCOE_OTHER,
+ /* Values 51-62 are reserved */
+ ICE_AVF_FLOW_FIELD_L2_PAYLOAD = 63,
+ ICE_AVF_FLOW_FIELD_MAX
+};
+
+/* Supported RSS offloads This macro is defined to support
+ * VIRTCHNL_OP_GET_RSS_HENA_CAPS ops. PF driver sends the RSS hardware
+ * capabilities to the caller of this ops.
+ */
+#define ICE_DEFAULT_RSS_HENA ( \
+ BIT_ULL(ICE_AVF_FLOW_FIELD_IPV4_UDP) | \
+ BIT_ULL(ICE_AVF_FLOW_FIELD_IPV4_SCTP) | \
+ BIT_ULL(ICE_AVF_FLOW_FIELD_IPV4_TCP) | \
+ BIT_ULL(ICE_AVF_FLOW_FIELD_IPV4_OTHER) | \
+ BIT_ULL(ICE_AVF_FLOW_FIELD_FRAG_IPV4) | \
+ BIT_ULL(ICE_AVF_FLOW_FIELD_IPV6_UDP) | \
+ BIT_ULL(ICE_AVF_FLOW_FIELD_IPV6_TCP) | \
+ BIT_ULL(ICE_AVF_FLOW_FIELD_IPV6_SCTP) | \
+ BIT_ULL(ICE_AVF_FLOW_FIELD_IPV6_OTHER) | \
+ BIT_ULL(ICE_AVF_FLOW_FIELD_FRAG_IPV6) | \
+ BIT_ULL(ICE_AVF_FLOW_FIELD_IPV4_TCP_SYN_NO_ACK) | \
+ BIT_ULL(ICE_AVF_FLOW_FIELD_UNICAST_IPV4_UDP) | \
+ BIT_ULL(ICE_AVF_FLOW_FIELD_MULTICAST_IPV4_UDP) | \
+ BIT_ULL(ICE_AVF_FLOW_FIELD_IPV6_TCP_SYN_NO_ACK) | \
+ BIT_ULL(ICE_AVF_FLOW_FIELD_UNICAST_IPV6_UDP) | \
+ BIT_ULL(ICE_AVF_FLOW_FIELD_MULTICAST_IPV6_UDP))
+
enum ice_flow_dir {
ICE_FLOW_RX = 0x02,
};
@@ -140,6 +197,8 @@ struct ice_rss_cfg {
enum ice_status ice_flow_rem_entry(struct ice_hw *hw, u64 entry_h);
void ice_rem_vsi_rss_list(struct ice_hw *hw, u16 vsi_handle);
enum ice_status ice_replay_rss_cfg(struct ice_hw *hw, u16 vsi_handle);
+enum ice_status
+ice_add_avf_rss_cfg(struct ice_hw *hw, u16 vsi_handle, u64 hashed_flds);
enum ice_status ice_rem_vsi_rss_cfg(struct ice_hw *hw, u16 vsi_handle);
enum ice_status
ice_add_rss_cfg(struct ice_hw *hw, u16 vsi_handle, u64 hashed_flds,
diff --git a/drivers/net/ethernet/intel/ice/ice_lib.c b/drivers/net/ethernet/intel/ice/ice_lib.c
index caa3edd5d3d6..1874c9f51a32 100644
--- a/drivers/net/ethernet/intel/ice/ice_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_lib.c
@@ -1113,6 +1113,33 @@ static int ice_vsi_cfg_rss_lut_key(struct ice_vsi *vsi)
return err;
}
+/**
+ * ice_vsi_set_vf_rss_flow_fld - Sets VF VSI RSS input set for different flows
+ * @vsi: VSI to be configured
+ *
+ * This function will only be called during the VF VSI setup. Upon successful
+ * completion of package download, this function will configure default RSS
+ * input sets for VF VSI.
+ */
+static void ice_vsi_set_vf_rss_flow_fld(struct ice_vsi *vsi)
+{
+ struct ice_pf *pf = vsi->back;
+ enum ice_status status;
+ struct device *dev;
+
+ dev = ice_pf_to_dev(pf);
+ if (ice_is_safe_mode(pf)) {
+ dev_dbg(dev, "Advanced RSS disabled. Package download failed, vsi num = %d\n",
+ vsi->vsi_num);
+ return;
+ }
+
+ status = ice_add_avf_rss_cfg(&pf->hw, vsi->idx, ICE_DEFAULT_RSS_HENA);
+ if (status)
+ dev_dbg(dev, "ice_add_avf_rss_cfg failed for vsi = %d, error = %d\n",
+ vsi->vsi_num, status);
+}
+
/**
* ice_vsi_set_rss_flow_fld - Sets RSS input set for different flows
* @vsi: VSI to be configured
@@ -2037,8 +2064,10 @@ ice_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi,
* receive traffic on first queue. Hence no need to capture
* return value
*/
- if (test_bit(ICE_FLAG_RSS_ENA, pf->flags))
+ if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
ice_vsi_cfg_rss_lut_key(vsi);
+ ice_vsi_set_vf_rss_flow_fld(vsi);
+ }
break;
case ICE_VSI_LB:
ret = ice_vsi_alloc_rings(vsi);
diff --git a/drivers/net/ethernet/intel/ice/ice_protocol_type.h b/drivers/net/ethernet/intel/ice/ice_protocol_type.h
index 97db46ce7a21..71647566964e 100644
--- a/drivers/net/ethernet/intel/ice/ice_protocol_type.h
+++ b/drivers/net/ethernet/intel/ice/ice_protocol_type.h
@@ -18,6 +18,7 @@ enum ice_prot_id {
ICE_PROT_IPV6_IL = 41,
ICE_PROT_TCP_IL = 49,
ICE_PROT_UDP_IL_OR_S = 53,
+ ICE_PROT_SCTP_IL = 96,
ICE_PROT_META_ID = 255, /* when offset == metadata */
ICE_PROT_INVALID = 255 /* when offset == ICE_FV_OFFSET_INVAL */
};
--
2.24.1
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [net-next 7/8] ice: Implement ethtool get/set rx-flow-hash
2020-01-26 6:07 [net-next 0/8][pull request] 100GbE Intel Wired LAN Driver Updates 2020-01-25 Jeff Kirsher
` (5 preceding siblings ...)
2020-01-26 6:07 ` [net-next 6/8] ice: Initilialize VF RSS tables Jeff Kirsher
@ 2020-01-26 6:07 ` Jeff Kirsher
2020-01-26 6:07 ` [net-next 8/8] ice: Bump version Jeff Kirsher
2020-01-26 10:50 ` [net-next 0/8][pull request] 100GbE Intel Wired LAN Driver Updates 2020-01-25 David Miller
8 siblings, 0 replies; 10+ messages in thread
From: Jeff Kirsher @ 2020-01-26 6:07 UTC (permalink / raw)
To: davem
Cc: Md Fahad Iqbal Polash, netdev, nhorman, sassmann, Tony Nguyen,
Andrew Bowers, Jeff Kirsher
From: Md Fahad Iqbal Polash <md.fahad.iqbal.polash@intel.com>
Provide support to change or retrieve RSS hash options for a flow type.
The supported flow-types are: tcp4, tcp6, udp4, udp6, sctp4, sctp6.
Signed-off-by: Md Fahad Iqbal Polash <md.fahad.iqbal.polash@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
drivers/net/ethernet/intel/ice/ice_ethtool.c | 243 +++++++++++++++++++
drivers/net/ethernet/intel/ice/ice_flow.c | 29 +++
drivers/net/ethernet/intel/ice/ice_flow.h | 1 +
3 files changed, 273 insertions(+)
diff --git a/drivers/net/ethernet/intel/ice/ice_ethtool.c b/drivers/net/ethernet/intel/ice/ice_ethtool.c
index f395457b728f..90c6a3ca20c9 100644
--- a/drivers/net/ethernet/intel/ice/ice_ethtool.c
+++ b/drivers/net/ethernet/intel/ice/ice_ethtool.c
@@ -4,6 +4,7 @@
/* ethtool support for ice */
#include "ice.h"
+#include "ice_flow.h"
#include "ice_lib.h"
#include "ice_dcb_lib.h"
@@ -2533,6 +2534,243 @@ ice_set_link_ksettings(struct net_device *netdev,
return err;
}
+/**
+ * ice_parse_hdrs - parses headers from RSS hash input
+ * @nfc: ethtool rxnfc command
+ *
+ * This function parses the rxnfc command and returns intended
+ * header types for RSS configuration
+ */
+static u32 ice_parse_hdrs(struct ethtool_rxnfc *nfc)
+{
+ u32 hdrs = ICE_FLOW_SEG_HDR_NONE;
+
+ switch (nfc->flow_type) {
+ case TCP_V4_FLOW:
+ hdrs |= ICE_FLOW_SEG_HDR_TCP | ICE_FLOW_SEG_HDR_IPV4;
+ break;
+ case UDP_V4_FLOW:
+ hdrs |= ICE_FLOW_SEG_HDR_UDP | ICE_FLOW_SEG_HDR_IPV4;
+ break;
+ case SCTP_V4_FLOW:
+ hdrs |= ICE_FLOW_SEG_HDR_SCTP | ICE_FLOW_SEG_HDR_IPV4;
+ break;
+ case TCP_V6_FLOW:
+ hdrs |= ICE_FLOW_SEG_HDR_TCP | ICE_FLOW_SEG_HDR_IPV6;
+ break;
+ case UDP_V6_FLOW:
+ hdrs |= ICE_FLOW_SEG_HDR_UDP | ICE_FLOW_SEG_HDR_IPV6;
+ break;
+ case SCTP_V6_FLOW:
+ hdrs |= ICE_FLOW_SEG_HDR_SCTP | ICE_FLOW_SEG_HDR_IPV6;
+ break;
+ default:
+ break;
+ }
+ return hdrs;
+}
+
+#define ICE_FLOW_HASH_FLD_IPV4_SA BIT_ULL(ICE_FLOW_FIELD_IDX_IPV4_SA)
+#define ICE_FLOW_HASH_FLD_IPV6_SA BIT_ULL(ICE_FLOW_FIELD_IDX_IPV6_SA)
+#define ICE_FLOW_HASH_FLD_IPV4_DA BIT_ULL(ICE_FLOW_FIELD_IDX_IPV4_DA)
+#define ICE_FLOW_HASH_FLD_IPV6_DA BIT_ULL(ICE_FLOW_FIELD_IDX_IPV6_DA)
+#define ICE_FLOW_HASH_FLD_TCP_SRC_PORT BIT_ULL(ICE_FLOW_FIELD_IDX_TCP_SRC_PORT)
+#define ICE_FLOW_HASH_FLD_TCP_DST_PORT BIT_ULL(ICE_FLOW_FIELD_IDX_TCP_DST_PORT)
+#define ICE_FLOW_HASH_FLD_UDP_SRC_PORT BIT_ULL(ICE_FLOW_FIELD_IDX_UDP_SRC_PORT)
+#define ICE_FLOW_HASH_FLD_UDP_DST_PORT BIT_ULL(ICE_FLOW_FIELD_IDX_UDP_DST_PORT)
+#define ICE_FLOW_HASH_FLD_SCTP_SRC_PORT \
+ BIT_ULL(ICE_FLOW_FIELD_IDX_SCTP_SRC_PORT)
+#define ICE_FLOW_HASH_FLD_SCTP_DST_PORT \
+ BIT_ULL(ICE_FLOW_FIELD_IDX_SCTP_DST_PORT)
+
+/**
+ * ice_parse_hash_flds - parses hash fields from RSS hash input
+ * @nfc: ethtool rxnfc command
+ *
+ * This function parses the rxnfc command and returns intended
+ * hash fields for RSS configuration
+ */
+static u64 ice_parse_hash_flds(struct ethtool_rxnfc *nfc)
+{
+ u64 hfld = ICE_HASH_INVALID;
+
+ if (nfc->data & RXH_IP_SRC || nfc->data & RXH_IP_DST) {
+ switch (nfc->flow_type) {
+ case TCP_V4_FLOW:
+ case UDP_V4_FLOW:
+ case SCTP_V4_FLOW:
+ if (nfc->data & RXH_IP_SRC)
+ hfld |= ICE_FLOW_HASH_FLD_IPV4_SA;
+ if (nfc->data & RXH_IP_DST)
+ hfld |= ICE_FLOW_HASH_FLD_IPV4_DA;
+ break;
+ case TCP_V6_FLOW:
+ case UDP_V6_FLOW:
+ case SCTP_V6_FLOW:
+ if (nfc->data & RXH_IP_SRC)
+ hfld |= ICE_FLOW_HASH_FLD_IPV6_SA;
+ if (nfc->data & RXH_IP_DST)
+ hfld |= ICE_FLOW_HASH_FLD_IPV6_DA;
+ break;
+ default:
+ break;
+ }
+ }
+
+ if (nfc->data & RXH_L4_B_0_1 || nfc->data & RXH_L4_B_2_3) {
+ switch (nfc->flow_type) {
+ case TCP_V4_FLOW:
+ case TCP_V6_FLOW:
+ if (nfc->data & RXH_L4_B_0_1)
+ hfld |= ICE_FLOW_HASH_FLD_TCP_SRC_PORT;
+ if (nfc->data & RXH_L4_B_2_3)
+ hfld |= ICE_FLOW_HASH_FLD_TCP_DST_PORT;
+ break;
+ case UDP_V4_FLOW:
+ case UDP_V6_FLOW:
+ if (nfc->data & RXH_L4_B_0_1)
+ hfld |= ICE_FLOW_HASH_FLD_UDP_SRC_PORT;
+ if (nfc->data & RXH_L4_B_2_3)
+ hfld |= ICE_FLOW_HASH_FLD_UDP_DST_PORT;
+ break;
+ case SCTP_V4_FLOW:
+ case SCTP_V6_FLOW:
+ if (nfc->data & RXH_L4_B_0_1)
+ hfld |= ICE_FLOW_HASH_FLD_SCTP_SRC_PORT;
+ if (nfc->data & RXH_L4_B_2_3)
+ hfld |= ICE_FLOW_HASH_FLD_SCTP_DST_PORT;
+ break;
+ default:
+ break;
+ }
+ }
+
+ return hfld;
+}
+
+/**
+ * ice_set_rss_hash_opt - Enable/Disable flow types for RSS hash
+ * @vsi: the VSI being configured
+ * @nfc: ethtool rxnfc command
+ *
+ * Returns Success if the flow input set is supported.
+ */
+static int
+ice_set_rss_hash_opt(struct ice_vsi *vsi, struct ethtool_rxnfc *nfc)
+{
+ struct ice_pf *pf = vsi->back;
+ enum ice_status status;
+ struct device *dev;
+ u64 hashed_flds;
+ u32 hdrs;
+
+ dev = ice_pf_to_dev(pf);
+ if (ice_is_safe_mode(pf)) {
+ dev_dbg(dev, "Advanced RSS disabled. Package download failed, vsi num = %d\n",
+ vsi->vsi_num);
+ return -EINVAL;
+ }
+
+ hashed_flds = ice_parse_hash_flds(nfc);
+ if (hashed_flds == ICE_HASH_INVALID) {
+ dev_dbg(dev, "Invalid hash fields, vsi num = %d\n",
+ vsi->vsi_num);
+ return -EINVAL;
+ }
+
+ hdrs = ice_parse_hdrs(nfc);
+ if (hdrs == ICE_FLOW_SEG_HDR_NONE) {
+ dev_dbg(dev, "Header type is not valid, vsi num = %d\n",
+ vsi->vsi_num);
+ return -EINVAL;
+ }
+
+ status = ice_add_rss_cfg(&pf->hw, vsi->idx, hashed_flds, hdrs);
+ if (status) {
+ dev_dbg(dev, "ice_add_rss_cfg failed, vsi num = %d, error = %d\n",
+ vsi->vsi_num, status);
+ return -EINVAL;
+ }
+
+ return 0;
+}
+
+/**
+ * ice_get_rss_hash_opt - Retrieve hash fields for a given flow-type
+ * @vsi: the VSI being configured
+ * @nfc: ethtool rxnfc command
+ */
+static void
+ice_get_rss_hash_opt(struct ice_vsi *vsi, struct ethtool_rxnfc *nfc)
+{
+ struct ice_pf *pf = vsi->back;
+ struct device *dev;
+ u64 hash_flds;
+ u32 hdrs;
+
+ dev = ice_pf_to_dev(pf);
+
+ nfc->data = 0;
+ if (ice_is_safe_mode(pf)) {
+ dev_dbg(dev, "Advanced RSS disabled. Package download failed, vsi num = %d\n",
+ vsi->vsi_num);
+ return;
+ }
+
+ hdrs = ice_parse_hdrs(nfc);
+ if (hdrs == ICE_FLOW_SEG_HDR_NONE) {
+ dev_dbg(dev, "Header type is not valid, vsi num = %d\n",
+ vsi->vsi_num);
+ return;
+ }
+
+ hash_flds = ice_get_rss_cfg(&pf->hw, vsi->idx, hdrs);
+ if (hash_flds == ICE_HASH_INVALID) {
+ dev_dbg(dev, "No hash fields found for the given header type, vsi num = %d\n",
+ vsi->vsi_num);
+ return;
+ }
+
+ if (hash_flds & ICE_FLOW_HASH_FLD_IPV4_SA ||
+ hash_flds & ICE_FLOW_HASH_FLD_IPV6_SA)
+ nfc->data |= (u64)RXH_IP_SRC;
+
+ if (hash_flds & ICE_FLOW_HASH_FLD_IPV4_DA ||
+ hash_flds & ICE_FLOW_HASH_FLD_IPV6_DA)
+ nfc->data |= (u64)RXH_IP_DST;
+
+ if (hash_flds & ICE_FLOW_HASH_FLD_TCP_SRC_PORT ||
+ hash_flds & ICE_FLOW_HASH_FLD_UDP_SRC_PORT ||
+ hash_flds & ICE_FLOW_HASH_FLD_SCTP_SRC_PORT)
+ nfc->data |= (u64)RXH_L4_B_0_1;
+
+ if (hash_flds & ICE_FLOW_HASH_FLD_TCP_DST_PORT ||
+ hash_flds & ICE_FLOW_HASH_FLD_UDP_DST_PORT ||
+ hash_flds & ICE_FLOW_HASH_FLD_SCTP_DST_PORT)
+ nfc->data |= (u64)RXH_L4_B_2_3;
+}
+
+/**
+ * ice_set_rxnfc - command to set Rx flow rules.
+ * @netdev: network interface device structure
+ * @cmd: ethtool rxnfc command
+ *
+ * Returns 0 for success and negative values for errors
+ */
+static int ice_set_rxnfc(struct net_device *netdev, struct ethtool_rxnfc *cmd)
+{
+ struct ice_netdev_priv *np = netdev_priv(netdev);
+ struct ice_vsi *vsi = np->vsi;
+
+ switch (cmd->cmd) {
+ case ETHTOOL_SRXFH:
+ return ice_set_rss_hash_opt(vsi, cmd);
+ default:
+ break;
+ }
+ return -EOPNOTSUPP;
+}
+
/**
* ice_get_rxnfc - command to get Rx flow classification rules
* @netdev: network interface device structure
@@ -2554,6 +2792,10 @@ ice_get_rxnfc(struct net_device *netdev, struct ethtool_rxnfc *cmd,
cmd->data = vsi->rss_size;
ret = 0;
break;
+ case ETHTOOL_GRXFH:
+ ice_get_rss_hash_opt(vsi, cmd);
+ ret = 0;
+ break;
default:
break;
}
@@ -3857,6 +4099,7 @@ static const struct ethtool_ops ice_ethtool_ops = {
.set_priv_flags = ice_set_priv_flags,
.get_sset_count = ice_get_sset_count,
.get_rxnfc = ice_get_rxnfc,
+ .set_rxnfc = ice_set_rxnfc,
.get_ringparam = ice_get_ringparam,
.set_ringparam = ice_set_ringparam,
.nway_reset = ice_nway_reset,
diff --git a/drivers/net/ethernet/intel/ice/ice_flow.c b/drivers/net/ethernet/intel/ice/ice_flow.c
index 4ecac0ae565a..a05ceb59863b 100644
--- a/drivers/net/ethernet/intel/ice/ice_flow.c
+++ b/drivers/net/ethernet/intel/ice/ice_flow.c
@@ -1244,3 +1244,32 @@ enum ice_status ice_replay_rss_cfg(struct ice_hw *hw, u16 vsi_handle)
return status;
}
+
+/**
+ * ice_get_rss_cfg - returns hashed fields for the given header types
+ * @hw: pointer to the hardware structure
+ * @vsi_handle: software VSI handle
+ * @hdrs: protocol header type
+ *
+ * This function will return the match fields of the first instance of flow
+ * profile having the given header types and containing input VSI
+ */
+u64 ice_get_rss_cfg(struct ice_hw *hw, u16 vsi_handle, u32 hdrs)
+{
+ struct ice_rss_cfg *r, *rss_cfg = NULL;
+
+ /* verify if the protocol header is non zero and VSI is valid */
+ if (hdrs == ICE_FLOW_SEG_HDR_NONE || !ice_is_vsi_valid(hw, vsi_handle))
+ return ICE_HASH_INVALID;
+
+ mutex_lock(&hw->rss_locks);
+ list_for_each_entry(r, &hw->rss_list_head, l_entry)
+ if (test_bit(vsi_handle, r->vsis) &&
+ r->packet_hdr == hdrs) {
+ rss_cfg = r;
+ break;
+ }
+ mutex_unlock(&hw->rss_locks);
+
+ return rss_cfg ? rss_cfg->hashed_flds : ICE_HASH_INVALID;
+}
diff --git a/drivers/net/ethernet/intel/ice/ice_flow.h b/drivers/net/ethernet/intel/ice/ice_flow.h
index 475a025d3a2d..5558627bd5eb 100644
--- a/drivers/net/ethernet/intel/ice/ice_flow.h
+++ b/drivers/net/ethernet/intel/ice/ice_flow.h
@@ -203,4 +203,5 @@ enum ice_status ice_rem_vsi_rss_cfg(struct ice_hw *hw, u16 vsi_handle);
enum ice_status
ice_add_rss_cfg(struct ice_hw *hw, u16 vsi_handle, u64 hashed_flds,
u32 addl_hdrs);
+u64 ice_get_rss_cfg(struct ice_hw *hw, u16 vsi_handle, u32 hdrs);
#endif /* _ICE_FLOW_H_ */
--
2.24.1
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [net-next 8/8] ice: Bump version
2020-01-26 6:07 [net-next 0/8][pull request] 100GbE Intel Wired LAN Driver Updates 2020-01-25 Jeff Kirsher
` (6 preceding siblings ...)
2020-01-26 6:07 ` [net-next 7/8] ice: Implement ethtool get/set rx-flow-hash Jeff Kirsher
@ 2020-01-26 6:07 ` Jeff Kirsher
2020-01-26 10:50 ` [net-next 0/8][pull request] 100GbE Intel Wired LAN Driver Updates 2020-01-25 David Miller
8 siblings, 0 replies; 10+ messages in thread
From: Jeff Kirsher @ 2020-01-26 6:07 UTC (permalink / raw)
To: davem; +Cc: Tony Nguyen, netdev, nhorman, sassmann, Andrew Bowers, Jeff Kirsher
From: Tony Nguyen <anthony.l.nguyen@intel.com>
Bump version to 0.8.2-k
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
drivers/net/ethernet/intel/ice/ice_main.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c
index eb9d00608e9a..5ae671609f98 100644
--- a/drivers/net/ethernet/intel/ice/ice_main.c
+++ b/drivers/net/ethernet/intel/ice/ice_main.c
@@ -13,7 +13,7 @@
#define DRV_VERSION_MAJOR 0
#define DRV_VERSION_MINOR 8
-#define DRV_VERSION_BUILD 1
+#define DRV_VERSION_BUILD 2
#define DRV_VERSION __stringify(DRV_VERSION_MAJOR) "." \
__stringify(DRV_VERSION_MINOR) "." \
--
2.24.1
^ permalink raw reply related [flat|nested] 10+ messages in thread
* Re: [net-next 0/8][pull request] 100GbE Intel Wired LAN Driver Updates 2020-01-25
2020-01-26 6:07 [net-next 0/8][pull request] 100GbE Intel Wired LAN Driver Updates 2020-01-25 Jeff Kirsher
` (7 preceding siblings ...)
2020-01-26 6:07 ` [net-next 8/8] ice: Bump version Jeff Kirsher
@ 2020-01-26 10:50 ` David Miller
8 siblings, 0 replies; 10+ messages in thread
From: David Miller @ 2020-01-26 10:50 UTC (permalink / raw)
To: jeffrey.t.kirsher; +Cc: netdev, nhorman, sassmann
From: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Date: Sat, 25 Jan 2020 22:07:29 -0800
> This series contains updates to the ice driver to add support for RSS.
>
> Henry and Tony enable the driver to write the filtering hardware tables
> to allow for changing of RSS rules, also introduced and initialized the
> structures for storing the configuration. Then followed it up by
> creating an extraction sequence based on the packet header protocols to
> be programmed. Next was storing the TCAM entry with the profile data
> and VSI group in the respective software structures.
>
> Md Fahad sets up the configuration to support RSS tables for the virtual
> function (VF) driver (iavf). Add support for flow types TCP4, TCP6,
> UDP4, UDP6, SCTP4 and SCTP6 for RSS via ethtool.
>
> The following are changes since commit 3333e50b64fe30b7e53cf02456a2f567f689ae4f:
> Merge branch 'mlxsw-Offload-TBF'
> and are available in the git repository at:
> git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next-queue 100GbE
Pulled, thanks Jeff.
^ permalink raw reply [flat|nested] 10+ messages in thread
end of thread, other threads:[~2020-01-26 10:50 UTC | newest]
Thread overview: 10+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-01-26 6:07 [net-next 0/8][pull request] 100GbE Intel Wired LAN Driver Updates 2020-01-25 Jeff Kirsher
2020-01-26 6:07 ` [net-next 1/8] ice: Enable writing hardware filtering tables Jeff Kirsher
2020-01-26 6:07 ` [net-next 2/8] ice: Allocate flow profile Jeff Kirsher
2020-01-26 6:07 ` [net-next 3/8] ice: Populate TCAM filter software structures Jeff Kirsher
2020-01-26 6:07 ` [net-next 4/8] ice: Enable writing filtering tables Jeff Kirsher
2020-01-26 6:07 ` [net-next 5/8] ice: Optimize table usage Jeff Kirsher
2020-01-26 6:07 ` [net-next 6/8] ice: Initilialize VF RSS tables Jeff Kirsher
2020-01-26 6:07 ` [net-next 7/8] ice: Implement ethtool get/set rx-flow-hash Jeff Kirsher
2020-01-26 6:07 ` [net-next 8/8] ice: Bump version Jeff Kirsher
2020-01-26 10:50 ` [net-next 0/8][pull request] 100GbE Intel Wired LAN Driver Updates 2020-01-25 David Miller
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.