From patchwork Sat Oct 17 06:28:06 2020 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Venkat Duvvuru X-Patchwork-Id: 81193 X-Patchwork-Delegate: ajit.khaparde@broadcom.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from dpdk.org (dpdk.org [92.243.14.124]) by inbox.dpdk.org (Postfix) with ESMTP id 20194A04DB; Sat, 17 Oct 2020 08:32:24 +0200 (CEST) Received: from [92.243.14.124] (localhost [127.0.0.1]) by dpdk.org (Postfix) with ESMTP id CE1C5FC42; Sat, 17 Oct 2020 08:28:34 +0200 (CEST) Received: from relay.smtp-ext.broadcom.com (unknown [192.19.221.30]) by dpdk.org (Postfix) with ESMTP id E7C75E2C3 for ; Sat, 17 Oct 2020 08:28:15 +0200 (CEST) Received: from S60.dhcp.broadcom.net (unknown [10.123.66.170]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-SHA256 (128/128 bits)) (No client certificate requested) by relay.smtp-ext.broadcom.com (Postfix) with ESMTPS id A1F708260; Fri, 16 Oct 2020 23:28:14 -0700 (PDT) DKIM-Filter: OpenDKIM Filter v2.11.0 relay.smtp-ext.broadcom.com A1F708260 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=broadcom.com; s=dkimrelay; t=1602916094; bh=noRJEe51jpdO/Sia077eKbZbnDmzDIElIkLzln3tRuk=; h=From:To:Cc:Subject:Date:In-Reply-To:References:From; b=FUspktDpXgAqoxoW/un0AqoYI8waDa43ECf9HDxnT+DPtl+6EUi4N+0TYLsKceLf+ lYHNua+BZKt09OjjOvb9gEbrXsHgyvF62P39NvTDCrt9cAOY312GoAE2VsTWm1Uhb2 fYZZ39zxpFuSCwlmd/VROWbz8ksrYtER+zn/jq1s= From: Venkat Duvvuru To: dev@dpdk.org Cc: Kishore Padmanabha Date: Sat, 17 Oct 2020 11:58:06 +0530 Message-Id: <1602916089-18576-12-git-send-email-venkatkumar.duvvuru@broadcom.com> X-Mailer: git-send-email 2.7.4 In-Reply-To: <1602916089-18576-1-git-send-email-venkatkumar.duvvuru@broadcom.com> References: <1602916089-18576-1-git-send-email-venkatkumar.duvvuru@broadcom.com> Subject: [dpdk-dev] [PATCH 11/14] net/bnxt: add mapper support for wildcard TCAM entry X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" From: Kishore Padmanabha Added support for the key and mask fields encoding for the wildcard tcam entry. Signed-off-by: Kishore Padmanabha Reviewed-by: Michael Baucom --- drivers/net/bnxt/tf_ulp/ulp_mapper.c | 47 +++++++++++-- drivers/net/bnxt/tf_ulp/ulp_utils.c | 125 +++++++++++++++++++++++++++++++++++ drivers/net/bnxt/tf_ulp/ulp_utils.h | 48 +++++++++++++- 3 files changed, 212 insertions(+), 8 deletions(-) diff --git a/drivers/net/bnxt/tf_ulp/ulp_mapper.c b/drivers/net/bnxt/tf_ulp/ulp_mapper.c index b74cb92..27b4780 100644 --- a/drivers/net/bnxt/tf_ulp/ulp_mapper.c +++ b/drivers/net/bnxt/tf_ulp/ulp_mapper.c @@ -1497,6 +1497,29 @@ ulp_mapper_tcam_tbl_entry_write(struct bnxt_ulp_mapper_parms *parms, return rc; } +#define BNXT_ULP_WC_TCAM_SLICE_SIZE 80 +/* internal function to post process the key/mask blobs for wildcard tcam tbl */ +static void ulp_mapper_wc_tcam_tbl_post_process(struct ulp_blob *blob, + uint32_t len) +{ + uint8_t mode[2] = {0x0, 0x0}; + uint32_t mode_len = len / BNXT_ULP_WC_TCAM_SLICE_SIZE; + uint32_t size, idx; + + /* Add the mode bits to the key and mask*/ + if (mode_len == 2) + mode[1] = 2; + else if (mode_len > 2) + mode[1] = 3; + + size = BNXT_ULP_WC_TCAM_SLICE_SIZE + ULP_BYTE_2_BITS(sizeof(mode)); + for (idx = 0; idx < mode_len; idx++) + ulp_blob_insert(blob, (size * idx), mode, + ULP_BYTE_2_BITS(sizeof(mode))); + ulp_blob_perform_64B_word_swap(blob); + ulp_blob_perform_64B_byte_swap(blob); +} + static int32_t ulp_mapper_tcam_tbl_process(struct bnxt_ulp_mapper_parms *parms, struct bnxt_ulp_mapper_tbl_info *tbl) @@ -1533,9 +1556,9 @@ ulp_mapper_tcam_tbl_process(struct bnxt_ulp_mapper_parms *parms, return -EINVAL; } - if (!ulp_blob_init(&key, tbl->key_bit_size, + if (!ulp_blob_init(&key, tbl->blob_key_bit_size, parms->device_params->byte_order) || - !ulp_blob_init(&mask, tbl->key_bit_size, + !ulp_blob_init(&mask, tbl->blob_key_bit_size, parms->device_params->byte_order) || !ulp_blob_init(&data, tbl->result_bit_size, parms->device_params->byte_order) || @@ -1545,6 +1568,11 @@ ulp_mapper_tcam_tbl_process(struct bnxt_ulp_mapper_parms *parms, return -EINVAL; } + if (tbl->resource_type == TF_TCAM_TBL_TYPE_WC_TCAM) { + key.byte_order = BNXT_ULP_BYTE_ORDER_BE; + mask.byte_order = BNXT_ULP_BYTE_ORDER_BE; + } + /* create the key/mask */ /* * NOTE: The WC table will require some kind of flag to handle the @@ -1570,6 +1598,11 @@ ulp_mapper_tcam_tbl_process(struct bnxt_ulp_mapper_parms *parms, } } + if (tbl->resource_type == TF_TCAM_TBL_TYPE_WC_TCAM) { + ulp_mapper_wc_tcam_tbl_post_process(&key, tbl->key_bit_size); + ulp_mapper_wc_tcam_tbl_post_process(&mask, tbl->key_bit_size); + } + if (tbl->srch_b4_alloc == BNXT_ULP_SEARCH_BEFORE_ALLOC_NO) { /* * No search for re-use is requested, so simply allocate the @@ -1578,18 +1611,18 @@ ulp_mapper_tcam_tbl_process(struct bnxt_ulp_mapper_parms *parms, aparms.dir = tbl->direction; aparms.tcam_tbl_type = tbl->resource_type; aparms.search_enable = tbl->srch_b4_alloc; - aparms.key_sz_in_bits = tbl->key_bit_size; aparms.key = ulp_blob_data_get(&key, &tmplen); - if (tbl->key_bit_size != tmplen) { + aparms.key_sz_in_bits = tmplen; + if (tbl->blob_key_bit_size != tmplen) { BNXT_TF_DBG(ERR, "Key len (%d) != Expected (%d)\n", - tmplen, tbl->key_bit_size); + tmplen, tbl->blob_key_bit_size); return -EINVAL; } aparms.mask = ulp_blob_data_get(&mask, &tmplen); - if (tbl->key_bit_size != tmplen) { + if (tbl->blob_key_bit_size != tmplen) { BNXT_TF_DBG(ERR, "Mask len (%d) != Expected (%d)\n", - tmplen, tbl->key_bit_size); + tmplen, tbl->blob_key_bit_size); return -EINVAL; } diff --git a/drivers/net/bnxt/tf_ulp/ulp_utils.c b/drivers/net/bnxt/tf_ulp/ulp_utils.c index 24474e2..a13a3bb 100644 --- a/drivers/net/bnxt/tf_ulp/ulp_utils.c +++ b/drivers/net/bnxt/tf_ulp/ulp_utils.c @@ -274,6 +274,69 @@ ulp_blob_push(struct ulp_blob *blob, } /* + * Insert data into the binary blob at the given offset. + * + * blob [in] The blob that data is added to. The blob must + * be initialized prior to pushing data. + * + * offset [in] The offset where the data needs to be inserted. + * + * data [in/out] A pointer to bytes to be added to the blob. + * + * datalen [in] The number of bits to be added to the blob. + * + * The offset of the data is updated after each push of data. + * NULL returned on error. + */ +uint32_t +ulp_blob_insert(struct ulp_blob *blob, uint32_t offset, + uint8_t *data, uint32_t datalen) +{ + uint32_t rc; + uint8_t local_data[BNXT_ULP_FLMP_BLOB_SIZE]; + uint16_t mov_len; + + /* validate the arguments */ + if (!blob || datalen > (uint32_t)(blob->bitlen - blob->write_idx) || + offset > blob->write_idx) { + BNXT_TF_DBG(ERR, "invalid argument\n"); + return 0; /* failure */ + } + + mov_len = blob->write_idx - offset; + /* If offset and data len are not 8 bit aligned then return error */ + if (ULP_BITS_IS_BYTE_NOT_ALIGNED(offset) || + ULP_BITS_IS_BYTE_NOT_ALIGNED(datalen)) { + BNXT_TF_DBG(ERR, "invalid argument, not aligned\n"); + return 0; /* failure */ + } + + /* copy the data so we can move the data */ + memcpy(local_data, &blob->data[ULP_BITS_2_BYTE_NR(offset)], + ULP_BITS_2_BYTE(mov_len)); + blob->write_idx = offset; + if (blob->byte_order == BNXT_ULP_BYTE_ORDER_BE) + rc = ulp_bs_push_msb(blob->data, + blob->write_idx, + datalen, + data); + else + rc = ulp_bs_push_lsb(blob->data, + blob->write_idx, + datalen, + data); + if (!rc) { + BNXT_TF_DBG(ERR, "Failed ro write blob\n"); + return 0; + } + /* copy the previously stored data */ + memcpy(&blob->data[ULP_BITS_2_BYTE_NR(offset + datalen)], local_data, + ULP_BITS_2_BYTE(mov_len)); + blob->write_idx += (mov_len + datalen); + return datalen; +} + +/* * Add data to the binary blob at the current offset. * * blob [in] The blob that data is added to. The blob must @@ -604,6 +667,68 @@ ulp_blob_perform_byte_reverse(struct ulp_blob *blob) } /* + * Perform the blob buffer 64 bit word swap. + * This api makes the first 4 bytes the last in + * a given 64 bit value and vice-versa. + * + * blob [in] The blob's data to be used for swap. + * + * returns void. + */ +void +ulp_blob_perform_64B_word_swap(struct ulp_blob *blob) +{ + uint32_t i, j, num; + uint8_t xchar; + uint32_t word_size = ULP_64B_IN_BYTES / 2; + + /* validate the arguments */ + if (!blob) { + BNXT_TF_DBG(ERR, "invalid argument\n"); + return; /* failure */ + } + num = ULP_BITS_2_BYTE(blob->write_idx); + for (i = 0; i < num; i = i + ULP_64B_IN_BYTES) { + for (j = 0; j < word_size; j++) { + xchar = blob->data[i + j]; + blob->data[i + j] = blob->data[i + j + word_size]; + blob->data[i + j + word_size] = xchar; + } + } +} + +/* + * Perform the blob buffer 64 bit byte swap. + * This api makes the first byte the last in + * a given 64 bit value and vice-versa. + * + * blob [in] The blob's data to be used for swap. + * + * returns void. + */ +void +ulp_blob_perform_64B_byte_swap(struct ulp_blob *blob) +{ + uint32_t i, j, num; + uint8_t xchar; + uint32_t offset = ULP_64B_IN_BYTES - 1; + + /* validate the arguments */ + if (!blob) { + BNXT_TF_DBG(ERR, "invalid argument\n"); + return; /* failure */ + } + num = ULP_BITS_2_BYTE(blob->write_idx); + for (i = 0; i < num; i = i + ULP_64B_IN_BYTES) { + for (j = 0; j < (ULP_64B_IN_BYTES / 2); j++) { + xchar = blob->data[i + j]; + blob->data[i + j] = blob->data[i + offset - j]; + blob->data[i + offset - j] = xchar; + } + } +} + +/* * Read data from the operand * * operand [in] A pointer to a 16 Byte operand diff --git a/drivers/net/bnxt/tf_ulp/ulp_utils.h b/drivers/net/bnxt/tf_ulp/ulp_utils.h index 898e851..749ac06 100644 --- a/drivers/net/bnxt/tf_ulp/ulp_utils.h +++ b/drivers/net/bnxt/tf_ulp/ulp_utils.h @@ -12,7 +12,7 @@ #define ULP_BUFFER_ALIGN_8_BYTE 8 #define ULP_BUFFER_ALIGN_16_BYTE 16 #define ULP_BUFFER_ALIGN_64_BYTE 64 - +#define ULP_64B_IN_BYTES 8 /* * Macros for bitmap sets and gets * These macros can be used if the val are power of 2. @@ -57,6 +57,9 @@ /* Macro to round off to next multiple of 8*/ #define ULP_BYTE_ROUND_OFF_8(x) (((x) + 7) & ~7) +/* Macro to check bits are byte aligned */ +#define ULP_BITS_IS_BYTE_NOT_ALIGNED(x) ((x) % 8) + /* Macros to read the computed fields */ #define ULP_COMP_FLD_IDX_RD(params, idx) \ rte_be_to_cpu_32((params)->comp_fld[(idx)]) @@ -168,6 +171,25 @@ ulp_blob_push(struct ulp_blob *blob, uint32_t datalen); /* + * Insert data into the binary blob at the given offset. + * + * blob [in] The blob that data is added to. The blob must + * be initialized prior to pushing data. + * + * offset [in] The offset where the data needs to be inserted. + * + * data [in/out] A pointer to bytes to be added to the blob. + * + * datalen [in] The number of bits to be added to the blob. + * + * The offset of the data is updated after each push of data. + * NULL returned on error. + */ +uint32_t +ulp_blob_insert(struct ulp_blob *blob, uint32_t offset, + uint8_t *data, uint32_t datalen); + +/* * Add data to the binary blob at the current offset. * * blob [in] The blob that data is added to. The blob must @@ -300,6 +322,30 @@ void ulp_blob_perform_byte_reverse(struct ulp_blob *blob); /* + * Perform the blob buffer 64 bit word swap. + * This api makes the first 4 bytes the last in + * a given 64 bit value and vice-versa. + * + * blob [in] The blob's data to be used for swap. + * + * returns void. + */ +void +ulp_blob_perform_64B_word_swap(struct ulp_blob *blob); + +/* + * Perform the blob buffer 64 bit byte swap. + * This api makes the first byte the last in + * a given 64 bit value and vice-versa. + * + * blob [in] The blob's data to be used for swap. + * + * returns void. + */ +void +ulp_blob_perform_64B_byte_swap(struct ulp_blob *blob); + +/* * Read data from the operand * * operand [in] A pointer to a 16 Byte operand