xref: /linux/drivers/infiniband/hw/bnxt_re/hw_counters.c (revision bb5b94f5bbe75470912b70fb08880fc5273aa62d)
1 /*
2  * Broadcom NetXtreme-E RoCE driver.
3  *
4  * Copyright (c) 2016 - 2017, Broadcom. All rights reserved.  The term
5  * Broadcom refers to Broadcom Limited and/or its subsidiaries.
6  *
7  * This software is available to you under a choice of one of two
8  * licenses.  You may choose to be licensed under the terms of the GNU
9  * General Public License (GPL) Version 2, available from the file
10  * COPYING in the main directory of this source tree, or the
11  * BSD license below:
12  *
13  * Redistribution and use in source and binary forms, with or without
14  * modification, are permitted provided that the following conditions
15  * are met:
16  *
17  * 1. Redistributions of source code must retain the above copyright
18  *    notice, this list of conditions and the following disclaimer.
19  * 2. Redistributions in binary form must reproduce the above copyright
20  *    notice, this list of conditions and the following disclaimer in
21  *    the documentation and/or other materials provided with the
22  *    distribution.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS''
25  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
26  * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
27  * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS
28  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
31  * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
32  * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE
33  * OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN
34  * IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
35  *
36  * Description: Statistics
37  *
38  */
39 
40 #include <linux/interrupt.h>
41 #include <linux/types.h>
42 #include <linux/spinlock.h>
43 #include <linux/sched.h>
44 #include <linux/slab.h>
45 #include <linux/pci.h>
46 #include <linux/prefetch.h>
47 #include <linux/delay.h>
48 
49 #include <rdma/ib_addr.h>
50 
51 #include "bnxt_ulp.h"
52 #include "roce_hsi.h"
53 #include "qplib_res.h"
54 #include "qplib_sp.h"
55 #include "qplib_fp.h"
56 #include "qplib_rcfw.h"
57 #include "bnxt_re.h"
58 #include "hw_counters.h"
59 
60 static const char * const bnxt_re_stat_name[] = {
61 	[BNXT_RE_ACTIVE_QP]		=  "active_qps",
62 	[BNXT_RE_ACTIVE_SRQ]		=  "active_srqs",
63 	[BNXT_RE_ACTIVE_CQ]		=  "active_cqs",
64 	[BNXT_RE_ACTIVE_MR]		=  "active_mrs",
65 	[BNXT_RE_ACTIVE_MW]		=  "active_mws",
66 	[BNXT_RE_RX_PKTS]		=  "rx_pkts",
67 	[BNXT_RE_RX_BYTES]		=  "rx_bytes",
68 	[BNXT_RE_TX_PKTS]		=  "tx_pkts",
69 	[BNXT_RE_TX_BYTES]		=  "tx_bytes",
70 	[BNXT_RE_RECOVERABLE_ERRORS]	=  "recoverable_errors",
71 	[BNXT_RE_RX_DROPS]		=  "rx_roce_drops",
72 	[BNXT_RE_RX_DISCARDS]		=  "rx_roce_discards",
73 	[BNXT_RE_TO_RETRANSMITS]        = "to_retransmits",
74 	[BNXT_RE_SEQ_ERR_NAKS_RCVD]     = "seq_err_naks_rcvd",
75 	[BNXT_RE_MAX_RETRY_EXCEEDED]    = "max_retry_exceeded",
76 	[BNXT_RE_RNR_NAKS_RCVD]         = "rnr_naks_rcvd",
77 	[BNXT_RE_MISSING_RESP]          = "missing_resp",
78 	[BNXT_RE_UNRECOVERABLE_ERR]     = "unrecoverable_err",
79 	[BNXT_RE_BAD_RESP_ERR]          = "bad_resp_err",
80 	[BNXT_RE_LOCAL_QP_OP_ERR]       = "local_qp_op_err",
81 	[BNXT_RE_LOCAL_PROTECTION_ERR]  = "local_protection_err",
82 	[BNXT_RE_MEM_MGMT_OP_ERR]       = "mem_mgmt_op_err",
83 	[BNXT_RE_REMOTE_INVALID_REQ_ERR] = "remote_invalid_req_err",
84 	[BNXT_RE_REMOTE_ACCESS_ERR]     = "remote_access_err",
85 	[BNXT_RE_REMOTE_OP_ERR]         = "remote_op_err",
86 	[BNXT_RE_DUP_REQ]               = "dup_req",
87 	[BNXT_RE_RES_EXCEED_MAX]        = "res_exceed_max",
88 	[BNXT_RE_RES_LENGTH_MISMATCH]   = "res_length_mismatch",
89 	[BNXT_RE_RES_EXCEEDS_WQE]       = "res_exceeds_wqe",
90 	[BNXT_RE_RES_OPCODE_ERR]        = "res_opcode_err",
91 	[BNXT_RE_RES_RX_INVALID_RKEY]   = "res_rx_invalid_rkey",
92 	[BNXT_RE_RES_RX_DOMAIN_ERR]     = "res_rx_domain_err",
93 	[BNXT_RE_RES_RX_NO_PERM]        = "res_rx_no_perm",
94 	[BNXT_RE_RES_RX_RANGE_ERR]      = "res_rx_range_err",
95 	[BNXT_RE_RES_TX_INVALID_RKEY]   = "res_tx_invalid_rkey",
96 	[BNXT_RE_RES_TX_DOMAIN_ERR]     = "res_tx_domain_err",
97 	[BNXT_RE_RES_TX_NO_PERM]        = "res_tx_no_perm",
98 	[BNXT_RE_RES_TX_RANGE_ERR]      = "res_tx_range_err",
99 	[BNXT_RE_RES_IRRQ_OFLOW]        = "res_irrq_oflow",
100 	[BNXT_RE_RES_UNSUP_OPCODE]      = "res_unsup_opcode",
101 	[BNXT_RE_RES_UNALIGNED_ATOMIC]  = "res_unaligned_atomic",
102 	[BNXT_RE_RES_REM_INV_ERR]       = "res_rem_inv_err",
103 	[BNXT_RE_RES_MEM_ERROR]         = "res_mem_err",
104 	[BNXT_RE_RES_SRQ_ERR]           = "res_srq_err",
105 	[BNXT_RE_RES_CMP_ERR]           = "res_cmp_err",
106 	[BNXT_RE_RES_INVALID_DUP_RKEY]  = "res_invalid_dup_rkey",
107 	[BNXT_RE_RES_WQE_FORMAT_ERR]    = "res_wqe_format_err",
108 	[BNXT_RE_RES_CQ_LOAD_ERR]       = "res_cq_load_err",
109 	[BNXT_RE_RES_SRQ_LOAD_ERR]      = "res_srq_load_err",
110 	[BNXT_RE_RES_TX_PCI_ERR]        = "res_tx_pci_err",
111 	[BNXT_RE_RES_RX_PCI_ERR]        = "res_rx_pci_err",
112 	[BNXT_RE_OUT_OF_SEQ_ERR]        = "oos_drop_count"
113 };
114 
115 int bnxt_re_ib_get_hw_stats(struct ib_device *ibdev,
116 			    struct rdma_hw_stats *stats,
117 			    u32 port, int index)
118 {
119 	struct bnxt_re_dev *rdev = to_bnxt_re_dev(ibdev, ibdev);
120 	struct ctx_hw_stats *bnxt_re_stats = rdev->qplib_ctx.stats.dma;
121 	int rc  = 0;
122 
123 	if (!port || !stats)
124 		return -EINVAL;
125 
126 	stats->value[BNXT_RE_ACTIVE_QP] = atomic_read(&rdev->qp_count);
127 	stats->value[BNXT_RE_ACTIVE_SRQ] = atomic_read(&rdev->srq_count);
128 	stats->value[BNXT_RE_ACTIVE_CQ] = atomic_read(&rdev->cq_count);
129 	stats->value[BNXT_RE_ACTIVE_MR] = atomic_read(&rdev->mr_count);
130 	stats->value[BNXT_RE_ACTIVE_MW] = atomic_read(&rdev->mw_count);
131 	if (bnxt_re_stats) {
132 		stats->value[BNXT_RE_RECOVERABLE_ERRORS] =
133 			le64_to_cpu(bnxt_re_stats->tx_bcast_pkts);
134 		stats->value[BNXT_RE_RX_DROPS] =
135 			le64_to_cpu(bnxt_re_stats->rx_error_pkts);
136 		stats->value[BNXT_RE_RX_DISCARDS] =
137 			le64_to_cpu(bnxt_re_stats->rx_discard_pkts);
138 		stats->value[BNXT_RE_RX_PKTS] =
139 			le64_to_cpu(bnxt_re_stats->rx_ucast_pkts);
140 		stats->value[BNXT_RE_RX_BYTES] =
141 			le64_to_cpu(bnxt_re_stats->rx_ucast_bytes);
142 		stats->value[BNXT_RE_TX_PKTS] =
143 			le64_to_cpu(bnxt_re_stats->tx_ucast_pkts);
144 		stats->value[BNXT_RE_TX_BYTES] =
145 			le64_to_cpu(bnxt_re_stats->tx_ucast_bytes);
146 	}
147 	if (test_bit(BNXT_RE_FLAG_ISSUE_ROCE_STATS, &rdev->flags)) {
148 		rc = bnxt_qplib_get_roce_stats(&rdev->rcfw, &rdev->stats);
149 		if (rc)
150 			clear_bit(BNXT_RE_FLAG_ISSUE_ROCE_STATS,
151 				  &rdev->flags);
152 		stats->value[BNXT_RE_TO_RETRANSMITS] =
153 					rdev->stats.to_retransmits;
154 		stats->value[BNXT_RE_SEQ_ERR_NAKS_RCVD] =
155 					rdev->stats.seq_err_naks_rcvd;
156 		stats->value[BNXT_RE_MAX_RETRY_EXCEEDED] =
157 					rdev->stats.max_retry_exceeded;
158 		stats->value[BNXT_RE_RNR_NAKS_RCVD] =
159 					rdev->stats.rnr_naks_rcvd;
160 		stats->value[BNXT_RE_MISSING_RESP] =
161 					rdev->stats.missing_resp;
162 		stats->value[BNXT_RE_UNRECOVERABLE_ERR] =
163 					rdev->stats.unrecoverable_err;
164 		stats->value[BNXT_RE_BAD_RESP_ERR] =
165 					rdev->stats.bad_resp_err;
166 		stats->value[BNXT_RE_LOCAL_QP_OP_ERR]	=
167 				rdev->stats.local_qp_op_err;
168 		stats->value[BNXT_RE_LOCAL_PROTECTION_ERR] =
169 				rdev->stats.local_protection_err;
170 		stats->value[BNXT_RE_MEM_MGMT_OP_ERR] =
171 				rdev->stats.mem_mgmt_op_err;
172 		stats->value[BNXT_RE_REMOTE_INVALID_REQ_ERR] =
173 				rdev->stats.remote_invalid_req_err;
174 		stats->value[BNXT_RE_REMOTE_ACCESS_ERR] =
175 				rdev->stats.remote_access_err;
176 		stats->value[BNXT_RE_REMOTE_OP_ERR] =
177 				rdev->stats.remote_op_err;
178 		stats->value[BNXT_RE_DUP_REQ] =
179 				rdev->stats.dup_req;
180 		stats->value[BNXT_RE_RES_EXCEED_MAX] =
181 				rdev->stats.res_exceed_max;
182 		stats->value[BNXT_RE_RES_LENGTH_MISMATCH] =
183 				rdev->stats.res_length_mismatch;
184 		stats->value[BNXT_RE_RES_EXCEEDS_WQE] =
185 				rdev->stats.res_exceeds_wqe;
186 		stats->value[BNXT_RE_RES_OPCODE_ERR] =
187 				rdev->stats.res_opcode_err;
188 		stats->value[BNXT_RE_RES_RX_INVALID_RKEY] =
189 				rdev->stats.res_rx_invalid_rkey;
190 		stats->value[BNXT_RE_RES_RX_DOMAIN_ERR] =
191 				rdev->stats.res_rx_domain_err;
192 		stats->value[BNXT_RE_RES_RX_NO_PERM] =
193 				rdev->stats.res_rx_no_perm;
194 		stats->value[BNXT_RE_RES_RX_RANGE_ERR]  =
195 				rdev->stats.res_rx_range_err;
196 		stats->value[BNXT_RE_RES_TX_INVALID_RKEY] =
197 				rdev->stats.res_tx_invalid_rkey;
198 		stats->value[BNXT_RE_RES_TX_DOMAIN_ERR] =
199 				rdev->stats.res_tx_domain_err;
200 		stats->value[BNXT_RE_RES_TX_NO_PERM] =
201 				rdev->stats.res_tx_no_perm;
202 		stats->value[BNXT_RE_RES_TX_RANGE_ERR]  =
203 				rdev->stats.res_tx_range_err;
204 		stats->value[BNXT_RE_RES_IRRQ_OFLOW] =
205 				rdev->stats.res_irrq_oflow;
206 		stats->value[BNXT_RE_RES_UNSUP_OPCODE]  =
207 				rdev->stats.res_unsup_opcode;
208 		stats->value[BNXT_RE_RES_UNALIGNED_ATOMIC] =
209 				rdev->stats.res_unaligned_atomic;
210 		stats->value[BNXT_RE_RES_REM_INV_ERR]   =
211 				rdev->stats.res_rem_inv_err;
212 		stats->value[BNXT_RE_RES_MEM_ERROR] =
213 				rdev->stats.res_mem_error;
214 		stats->value[BNXT_RE_RES_SRQ_ERR] =
215 				rdev->stats.res_srq_err;
216 		stats->value[BNXT_RE_RES_CMP_ERR] =
217 				rdev->stats.res_cmp_err;
218 		stats->value[BNXT_RE_RES_INVALID_DUP_RKEY] =
219 				rdev->stats.res_invalid_dup_rkey;
220 		stats->value[BNXT_RE_RES_WQE_FORMAT_ERR] =
221 				rdev->stats.res_wqe_format_err;
222 		stats->value[BNXT_RE_RES_CQ_LOAD_ERR]   =
223 				rdev->stats.res_cq_load_err;
224 		stats->value[BNXT_RE_RES_SRQ_LOAD_ERR]  =
225 				rdev->stats.res_srq_load_err;
226 		stats->value[BNXT_RE_RES_TX_PCI_ERR]    =
227 				rdev->stats.res_tx_pci_err;
228 		stats->value[BNXT_RE_RES_RX_PCI_ERR]    =
229 				rdev->stats.res_rx_pci_err;
230 		stats->value[BNXT_RE_OUT_OF_SEQ_ERR]    =
231 				rdev->stats.res_oos_drop_count;
232 	}
233 
234 	return ARRAY_SIZE(bnxt_re_stat_name);
235 }
236 
237 struct rdma_hw_stats *bnxt_re_ib_alloc_hw_port_stats(struct ib_device *ibdev,
238 						     u32 port_num)
239 {
240 	BUILD_BUG_ON(ARRAY_SIZE(bnxt_re_stat_name) != BNXT_RE_NUM_COUNTERS);
241 
242 	return rdma_alloc_hw_stats_struct(bnxt_re_stat_name,
243 					  ARRAY_SIZE(bnxt_re_stat_name),
244 					  RDMA_HW_STATS_DEFAULT_LIFESPAN);
245 }
246