perf tools: Add coresight etm PMU record capabilities
authorMathieu Poirier <mathieu.poirier@linaro.org>
Fri, 16 Sep 2016 15:50:00 +0000 (09:50 -0600)
committerArnaldo Carvalho de Melo <acme@redhat.com>
Thu, 22 Sep 2016 15:19:40 +0000 (12:19 -0300)
Coresight ETMs are IP blocks used to perform HW assisted tracing on a
CPU core.  This patch introduce the required auxiliary API functions
allowing the perf core to interact with a tracer.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/1474041004-13956-4-git-send-email-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
MAINTAINERS
tools/perf/arch/arm/util/Build
tools/perf/arch/arm/util/auxtrace.c [new file with mode: 0644]
tools/perf/arch/arm/util/cs-etm.c [new file with mode: 0644]
tools/perf/arch/arm/util/cs-etm.h [new file with mode: 0644]
tools/perf/arch/arm64/util/Build
tools/perf/util/auxtrace.c
tools/perf/util/auxtrace.h
tools/perf/util/cs-etm.h [new file with mode: 0644]

index 7407fe779053fece24ecea2e76bebd9dceb89d1c..ac29d54537508a86e0437fe8a43a6611fd0f2d9c 100644 (file)
@@ -1124,6 +1124,10 @@ F:       Documentation/trace/coresight.txt
 F:     Documentation/devicetree/bindings/arm/coresight.txt
 F:     Documentation/ABI/testing/sysfs-bus-coresight-devices-*
 F:     tools/perf/arch/arm/util/pmu.c
+F:     tools/perf/arch/arm/util/auxtrace.c
+F:     tools/perf/arch/arm/util/cs-etm.c
+F:     tools/perf/arch/arm/util/cs-etm.h
+F:     tools/perf/util/cs-etm.h
 
 ARM/CORGI MACHINE SUPPORT
 M:     Richard Purdie <rpurdie@rpsys.net>
index 4093fd146f467e9f9e368595a693c2a2413eee4c..e64c5f216448069f4f184ffbb391963b49be6856 100644 (file)
@@ -3,4 +3,4 @@ libperf-$(CONFIG_DWARF) += dwarf-regs.o
 libperf-$(CONFIG_LOCAL_LIBUNWIND)    += unwind-libunwind.o
 libperf-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o
 
-libperf-$(CONFIG_AUXTRACE) += pmu.o
+libperf-$(CONFIG_AUXTRACE) += pmu.o auxtrace.o cs-etm.o
diff --git a/tools/perf/arch/arm/util/auxtrace.c b/tools/perf/arch/arm/util/auxtrace.c
new file mode 100644 (file)
index 0000000..8edf2cb
--- /dev/null
@@ -0,0 +1,54 @@
+/*
+ * Copyright(C) 2015 Linaro Limited. All rights reserved.
+ * Author: Mathieu Poirier <mathieu.poirier@linaro.org>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <stdbool.h>
+#include <linux/coresight-pmu.h>
+
+#include "../../util/auxtrace.h"
+#include "../../util/evlist.h"
+#include "../../util/pmu.h"
+#include "cs-etm.h"
+
+struct auxtrace_record
+*auxtrace_record__init(struct perf_evlist *evlist, int *err)
+{
+       struct perf_pmu *cs_etm_pmu;
+       struct perf_evsel *evsel;
+       bool found_etm = false;
+
+       cs_etm_pmu = perf_pmu__find(CORESIGHT_ETM_PMU_NAME);
+
+       if (evlist) {
+               evlist__for_each_entry(evlist, evsel) {
+                       if (cs_etm_pmu &&
+                           evsel->attr.type == cs_etm_pmu->type)
+                               found_etm = true;
+               }
+       }
+
+       if (found_etm)
+               return cs_etm_record_init(err);
+
+       /*
+        * Clear 'err' even if we haven't found a cs_etm event - that way perf
+        * record can still be used even if tracers aren't present.  The NULL
+        * return value will take care of telling the infrastructure HW tracing
+        * isn't available.
+        */
+       *err = 0;
+       return NULL;
+}
diff --git a/tools/perf/arch/arm/util/cs-etm.c b/tools/perf/arch/arm/util/cs-etm.c
new file mode 100644 (file)
index 0000000..829c479
--- /dev/null
@@ -0,0 +1,559 @@
+/*
+ * Copyright(C) 2015 Linaro Limited. All rights reserved.
+ * Author: Mathieu Poirier <mathieu.poirier@linaro.org>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <api/fs/fs.h>
+#include <linux/bitops.h>
+#include <linux/coresight-pmu.h>
+#include <linux/kernel.h>
+#include <linux/log2.h>
+#include <linux/types.h>
+
+#include "cs-etm.h"
+#include "../../perf.h"
+#include "../../util/auxtrace.h"
+#include "../../util/cpumap.h"
+#include "../../util/evlist.h"
+#include "../../util/pmu.h"
+#include "../../util/thread_map.h"
+#include "../../util/cs-etm.h"
+
+#include <stdlib.h>
+
+struct cs_etm_recording {
+       struct auxtrace_record  itr;
+       struct perf_pmu         *cs_etm_pmu;
+       struct perf_evlist      *evlist;
+       bool                    snapshot_mode;
+       size_t                  snapshot_size;
+};
+
+static bool cs_etm_is_etmv4(struct auxtrace_record *itr, int cpu);
+
+static int cs_etm_parse_snapshot_options(struct auxtrace_record *itr,
+                                        struct record_opts *opts,
+                                        const char *str)
+{
+       struct cs_etm_recording *ptr =
+                               container_of(itr, struct cs_etm_recording, itr);
+       unsigned long long snapshot_size = 0;
+       char *endptr;
+
+       if (str) {
+               snapshot_size = strtoull(str, &endptr, 0);
+               if (*endptr || snapshot_size > SIZE_MAX)
+                       return -1;
+       }
+
+       opts->auxtrace_snapshot_mode = true;
+       opts->auxtrace_snapshot_size = snapshot_size;
+       ptr->snapshot_size = snapshot_size;
+
+       return 0;
+}
+
+static int cs_etm_recording_options(struct auxtrace_record *itr,
+                                   struct perf_evlist *evlist,
+                                   struct record_opts *opts)
+{
+       struct cs_etm_recording *ptr =
+                               container_of(itr, struct cs_etm_recording, itr);
+       struct perf_pmu *cs_etm_pmu = ptr->cs_etm_pmu;
+       struct perf_evsel *evsel, *cs_etm_evsel = NULL;
+       const struct cpu_map *cpus = evlist->cpus;
+       bool privileged = (geteuid() == 0 || perf_event_paranoid() < 0);
+
+       ptr->evlist = evlist;
+       ptr->snapshot_mode = opts->auxtrace_snapshot_mode;
+
+       evlist__for_each_entry(evlist, evsel) {
+               if (evsel->attr.type == cs_etm_pmu->type) {
+                       if (cs_etm_evsel) {
+                               pr_err("There may be only one %s event\n",
+                                      CORESIGHT_ETM_PMU_NAME);
+                               return -EINVAL;
+                       }
+                       evsel->attr.freq = 0;
+                       evsel->attr.sample_period = 1;
+                       cs_etm_evsel = evsel;
+                       opts->full_auxtrace = true;
+               }
+       }
+
+       /* no need to continue if at least one event of interest was found */
+       if (!cs_etm_evsel)
+               return 0;
+
+       if (opts->use_clockid) {
+               pr_err("Cannot use clockid (-k option) with %s\n",
+                      CORESIGHT_ETM_PMU_NAME);
+               return -EINVAL;
+       }
+
+       /* we are in snapshot mode */
+       if (opts->auxtrace_snapshot_mode) {
+               /*
+                * No size were given to '-S' or '-m,', so go with
+                * the default
+                */
+               if (!opts->auxtrace_snapshot_size &&
+                   !opts->auxtrace_mmap_pages) {
+                       if (privileged) {
+                               opts->auxtrace_mmap_pages = MiB(4) / page_size;
+                       } else {
+                               opts->auxtrace_mmap_pages =
+                                                       KiB(128) / page_size;
+                               if (opts->mmap_pages == UINT_MAX)
+                                       opts->mmap_pages = KiB(256) / page_size;
+                       }
+               } else if (!opts->auxtrace_mmap_pages && !privileged &&
+                                               opts->mmap_pages == UINT_MAX) {
+                       opts->mmap_pages = KiB(256) / page_size;
+               }
+
+               /*
+                * '-m,xyz' was specified but no snapshot size, so make the
+                * snapshot size as big as the auxtrace mmap area.
+                */
+               if (!opts->auxtrace_snapshot_size) {
+                       opts->auxtrace_snapshot_size =
+                               opts->auxtrace_mmap_pages * (size_t)page_size;
+               }
+
+               /*
+                * -Sxyz was specified but no auxtrace mmap area, so make the
+                * auxtrace mmap area big enough to fit the requested snapshot
+                * size.
+                */
+               if (!opts->auxtrace_mmap_pages) {
+                       size_t sz = opts->auxtrace_snapshot_size;
+
+                       sz = round_up(sz, page_size) / page_size;
+                       opts->auxtrace_mmap_pages = roundup_pow_of_two(sz);
+               }
+
+               /* Snapshost size can't be bigger than the auxtrace area */
+               if (opts->auxtrace_snapshot_size >
+                               opts->auxtrace_mmap_pages * (size_t)page_size) {
+                       pr_err("Snapshot size %zu must not be greater than AUX area tracing mmap size %zu\n",
+                              opts->auxtrace_snapshot_size,
+                              opts->auxtrace_mmap_pages * (size_t)page_size);
+                       return -EINVAL;
+               }
+
+               /* Something went wrong somewhere - this shouldn't happen */
+               if (!opts->auxtrace_snapshot_size ||
+                   !opts->auxtrace_mmap_pages) {
+                       pr_err("Failed to calculate default snapshot size and/or AUX area tracing mmap pages\n");
+                       return -EINVAL;
+               }
+       }
+
+       /* We are in full trace mode but '-m,xyz' wasn't specified */
+       if (opts->full_auxtrace && !opts->auxtrace_mmap_pages) {
+               if (privileged) {
+                       opts->auxtrace_mmap_pages = MiB(4) / page_size;
+               } else {
+                       opts->auxtrace_mmap_pages = KiB(128) / page_size;
+                       if (opts->mmap_pages == UINT_MAX)
+                               opts->mmap_pages = KiB(256) / page_size;
+               }
+
+       }
+
+       /* Validate auxtrace_mmap_pages provided by user */
+       if (opts->auxtrace_mmap_pages) {
+               unsigned int max_page = (KiB(128) / page_size);
+               size_t sz = opts->auxtrace_mmap_pages * (size_t)page_size;
+
+               if (!privileged &&
+                   opts->auxtrace_mmap_pages > max_page) {
+                       opts->auxtrace_mmap_pages = max_page;
+                       pr_err("auxtrace too big, truncating to %d\n",
+                              max_page);
+               }
+
+               if (!is_power_of_2(sz)) {
+                       pr_err("Invalid mmap size for %s: must be a power of 2\n",
+                              CORESIGHT_ETM_PMU_NAME);
+                       return -EINVAL;
+               }
+       }
+
+       if (opts->auxtrace_snapshot_mode)
+               pr_debug2("%s snapshot size: %zu\n", CORESIGHT_ETM_PMU_NAME,
+                         opts->auxtrace_snapshot_size);
+
+       if (cs_etm_evsel) {
+               /*
+                * To obtain the auxtrace buffer file descriptor, the auxtrace
+                * event must come first.
+                */
+               perf_evlist__to_front(evlist, cs_etm_evsel);
+               /*
+                * In the case of per-cpu mmaps, we need the CPU on the
+                * AUX event.
+                */
+               if (!cpu_map__empty(cpus))
+                       perf_evsel__set_sample_bit(cs_etm_evsel, CPU);
+       }
+
+       /* Add dummy event to keep tracking */
+       if (opts->full_auxtrace) {
+               struct perf_evsel *tracking_evsel;
+               int err;
+
+               err = parse_events(evlist, "dummy:u", NULL);
+               if (err)
+                       return err;
+
+               tracking_evsel = perf_evlist__last(evlist);
+               perf_evlist__set_tracking_event(evlist, tracking_evsel);
+
+               tracking_evsel->attr.freq = 0;
+               tracking_evsel->attr.sample_period = 1;
+
+               /* In per-cpu case, always need the time of mmap events etc */
+               if (!cpu_map__empty(cpus))
+                       perf_evsel__set_sample_bit(tracking_evsel, TIME);
+       }
+
+       return 0;
+}
+
+static u64 cs_etm_get_config(struct auxtrace_record *itr)
+{
+       u64 config = 0;
+       struct cs_etm_recording *ptr =
+                       container_of(itr, struct cs_etm_recording, itr);
+       struct perf_pmu *cs_etm_pmu = ptr->cs_etm_pmu;
+       struct perf_evlist *evlist = ptr->evlist;
+       struct perf_evsel *evsel;
+
+       evlist__for_each_entry(evlist, evsel) {
+               if (evsel->attr.type == cs_etm_pmu->type) {
+                       /*
+                        * Variable perf_event_attr::config is assigned to
+                        * ETMv3/PTM.  The bit fields have been made to match
+                        * the ETMv3.5 ETRMCR register specification.  See the
+                        * PMU_FORMAT_ATTR() declarations in
+                        * drivers/hwtracing/coresight/coresight-perf.c for
+                        * details.
+                        */
+                       config = evsel->attr.config;
+                       break;
+               }
+       }
+
+       return config;
+}
+
+static size_t
+cs_etm_info_priv_size(struct auxtrace_record *itr __maybe_unused,
+                     struct perf_evlist *evlist __maybe_unused)
+{
+       int i;
+       int etmv3 = 0, etmv4 = 0;
+       const struct cpu_map *cpus = evlist->cpus;
+
+       /* cpu map is not empty, we have specific CPUs to work with */
+       if (!cpu_map__empty(cpus)) {
+               for (i = 0; i < cpu_map__nr(cpus); i++) {
+                       if (cs_etm_is_etmv4(itr, cpus->map[i]))
+                               etmv4++;
+                       else
+                               etmv3++;
+               }
+       } else {
+               /* get configuration for all CPUs in the system */
+               for (i = 0; i < cpu__max_cpu(); i++) {
+                       if (cs_etm_is_etmv4(itr, i))
+                               etmv4++;
+                       else
+                               etmv3++;
+               }
+       }
+
+       return (CS_ETM_HEADER_SIZE +
+              (etmv4 * CS_ETMV4_PRIV_SIZE) +
+              (etmv3 * CS_ETMV3_PRIV_SIZE));
+}
+
+static const char *metadata_etmv3_ro[CS_ETM_PRIV_MAX] = {
+       [CS_ETM_ETMCCER]        = "mgmt/etmccer",
+       [CS_ETM_ETMIDR]         = "mgmt/etmidr",
+};
+
+static const char *metadata_etmv4_ro[CS_ETMV4_PRIV_MAX] = {
+       [CS_ETMV4_TRCIDR0]              = "trcidr/trcidr0",
+       [CS_ETMV4_TRCIDR1]              = "trcidr/trcidr1",
+       [CS_ETMV4_TRCIDR2]              = "trcidr/trcidr2",
+       [CS_ETMV4_TRCIDR8]              = "trcidr/trcidr8",
+       [CS_ETMV4_TRCAUTHSTATUS]        = "mgmt/trcauthstatus",
+};
+
+static bool cs_etm_is_etmv4(struct auxtrace_record *itr, int cpu)
+{
+       bool ret = false;
+       char path[PATH_MAX];
+       int scan;
+       unsigned int val;
+       struct cs_etm_recording *ptr =
+                       container_of(itr, struct cs_etm_recording, itr);
+       struct perf_pmu *cs_etm_pmu = ptr->cs_etm_pmu;
+
+       /* Take any of the RO files for ETMv4 and see if it present */
+       snprintf(path, PATH_MAX, "cpu%d/%s",
+                cpu, metadata_etmv4_ro[CS_ETMV4_TRCIDR0]);
+       scan = perf_pmu__scan_file(cs_etm_pmu, path, "%x", &val);
+
+       /* The file was read successfully, we have a winner */
+       if (scan == 1)
+               ret = true;
+
+       return ret;
+}
+
+static int cs_etm_get_ro(struct perf_pmu *pmu, int cpu, const char *path)
+{
+       char pmu_path[PATH_MAX];
+       int scan;
+       unsigned int val = 0;
+
+       /* Get RO metadata from sysfs */
+       snprintf(pmu_path, PATH_MAX, "cpu%d/%s", cpu, path);
+
+       scan = perf_pmu__scan_file(pmu, pmu_path, "%x", &val);
+       if (scan != 1)
+               pr_err("%s: error reading: %s\n", __func__, pmu_path);
+
+       return val;
+}
+
+static void cs_etm_get_metadata(int cpu, u32 *offset,
+                               struct auxtrace_record *itr,
+                               struct auxtrace_info_event *info)
+{
+       u32 increment;
+       u64 magic;
+       struct cs_etm_recording *ptr =
+                       container_of(itr, struct cs_etm_recording, itr);
+       struct perf_pmu *cs_etm_pmu = ptr->cs_etm_pmu;
+
+       /* first see what kind of tracer this cpu is affined to */
+       if (cs_etm_is_etmv4(itr, cpu)) {
+               magic = __perf_cs_etmv4_magic;
+               /* Get trace configuration register */
+               info->priv[*offset + CS_ETMV4_TRCCONFIGR] =
+                                               cs_etm_get_config(itr);
+               /* Get traceID from the framework */
+               info->priv[*offset + CS_ETMV4_TRCTRACEIDR] =
+                                               coresight_get_trace_id(cpu);
+               /* Get read-only information from sysFS */
+               info->priv[*offset + CS_ETMV4_TRCIDR0] =
+                       cs_etm_get_ro(cs_etm_pmu, cpu,
+                                     metadata_etmv4_ro[CS_ETMV4_TRCIDR0]);
+               info->priv[*offset + CS_ETMV4_TRCIDR1] =
+                       cs_etm_get_ro(cs_etm_pmu, cpu,
+                                     metadata_etmv4_ro[CS_ETMV4_TRCIDR1]);
+               info->priv[*offset + CS_ETMV4_TRCIDR2] =
+                       cs_etm_get_ro(cs_etm_pmu, cpu,
+                                     metadata_etmv4_ro[CS_ETMV4_TRCIDR2]);
+               info->priv[*offset + CS_ETMV4_TRCIDR8] =
+                       cs_etm_get_ro(cs_etm_pmu, cpu,
+                                     metadata_etmv4_ro[CS_ETMV4_TRCIDR8]);
+               info->priv[*offset + CS_ETMV4_TRCAUTHSTATUS] =
+                       cs_etm_get_ro(cs_etm_pmu, cpu,
+                                     metadata_etmv4_ro
+                                     [CS_ETMV4_TRCAUTHSTATUS]);
+
+               /* How much space was used */
+               increment = CS_ETMV4_PRIV_MAX;
+       } else {
+               magic = __perf_cs_etmv3_magic;
+               /* Get configuration register */
+               info->priv[*offset + CS_ETM_ETMCR] = cs_etm_get_config(itr);
+               /* Get traceID from the framework */
+               info->priv[*offset + CS_ETM_ETMTRACEIDR] =
+                                               coresight_get_trace_id(cpu);
+               /* Get read-only information from sysFS */
+               info->priv[*offset + CS_ETM_ETMCCER] =
+                       cs_etm_get_ro(cs_etm_pmu, cpu,
+                                     metadata_etmv3_ro[CS_ETM_ETMCCER]);
+               info->priv[*offset + CS_ETM_ETMIDR] =
+                       cs_etm_get_ro(cs_etm_pmu, cpu,
+                                     metadata_etmv3_ro[CS_ETM_ETMIDR]);
+
+               /* How much space was used */
+               increment = CS_ETM_PRIV_MAX;
+       }
+
+       /* Build generic header portion */
+       info->priv[*offset + CS_ETM_MAGIC] = magic;
+       info->priv[*offset + CS_ETM_CPU] = cpu;
+       /* Where the next CPU entry should start from */
+       *offset += increment;
+}
+
+static int cs_etm_info_fill(struct auxtrace_record *itr,
+                           struct perf_session *session,
+                           struct auxtrace_info_event *info,
+                           size_t priv_size)
+{
+       int i;
+       u32 offset;
+       u64 nr_cpu, type;
+       const struct cpu_map *cpus = session->evlist->cpus;
+       struct cs_etm_recording *ptr =
+                       container_of(itr, struct cs_etm_recording, itr);
+       struct perf_pmu *cs_etm_pmu = ptr->cs_etm_pmu;
+
+       if (priv_size != cs_etm_info_priv_size(itr, session->evlist))
+               return -EINVAL;
+
+       if (!session->evlist->nr_mmaps)
+               return -EINVAL;
+
+       /* If the cpu_map is empty all CPUs are involved */
+       nr_cpu = cpu_map__empty(cpus) ? cpu__max_cpu() : cpu_map__nr(cpus);
+       /* Get PMU type as dynamically assigned by the core */
+       type = cs_etm_pmu->type;
+
+       /* First fill out the session header */
+       info->type = PERF_AUXTRACE_CS_ETM;
+       info->priv[CS_HEADER_VERSION_0] = 0;
+       info->priv[CS_PMU_TYPE_CPUS] = type << 32;
+       info->priv[CS_PMU_TYPE_CPUS] |= nr_cpu;
+       info->priv[CS_ETM_SNAPSHOT] = ptr->snapshot_mode;
+
+       offset = CS_ETM_SNAPSHOT + 1;
+
+       /* cpu map is not empty, we have specific CPUs to work with */
+       if (!cpu_map__empty(cpus)) {
+               for (i = 0; i < cpu_map__nr(cpus) && offset < priv_size; i++)
+                       cs_etm_get_metadata(cpus->map[i], &offset, itr, info);
+       } else {
+               /* get configuration for all CPUs in the system */
+               for (i = 0; i < cpu__max_cpu(); i++)
+                       cs_etm_get_metadata(i, &offset, itr, info);
+       }
+
+       return 0;
+}
+
+static int cs_etm_find_snapshot(struct auxtrace_record *itr __maybe_unused,
+                               int idx, struct auxtrace_mmap *mm,
+                               unsigned char *data __maybe_unused,
+                               u64 *head, u64 *old)
+{
+       pr_debug3("%s: mmap index %d old head %zu new head %zu size %zu\n",
+                 __func__, idx, (size_t)*old, (size_t)*head, mm->len);
+
+       *old = *head;
+       *head += mm->len;
+
+       return 0;
+}
+
+static int cs_etm_snapshot_start(struct auxtrace_record *itr)
+{
+       struct cs_etm_recording *ptr =
+                       container_of(itr, struct cs_etm_recording, itr);
+       struct perf_evsel *evsel;
+
+       evlist__for_each_entry(ptr->evlist, evsel) {
+               if (evsel->attr.type == ptr->cs_etm_pmu->type)
+                       return perf_evsel__disable(evsel);
+       }
+       return -EINVAL;
+}
+
+static int cs_etm_snapshot_finish(struct auxtrace_record *itr)
+{
+       struct cs_etm_recording *ptr =
+                       container_of(itr, struct cs_etm_recording, itr);
+       struct perf_evsel *evsel;
+
+       evlist__for_each_entry(ptr->evlist, evsel) {
+               if (evsel->attr.type == ptr->cs_etm_pmu->type)
+                       return perf_evsel__enable(evsel);
+       }
+       return -EINVAL;
+}
+
+static u64 cs_etm_reference(struct auxtrace_record *itr __maybe_unused)
+{
+       return (((u64) rand() <<  0) & 0x00000000FFFFFFFFull) |
+               (((u64) rand() << 32) & 0xFFFFFFFF00000000ull);
+}
+
+static void cs_etm_recording_free(struct auxtrace_record *itr)
+{
+       struct cs_etm_recording *ptr =
+                       container_of(itr, struct cs_etm_recording, itr);
+       free(ptr);
+}
+
+static int cs_etm_read_finish(struct auxtrace_record *itr, int idx)
+{
+       struct cs_etm_recording *ptr =
+                       container_of(itr, struct cs_etm_recording, itr);
+       struct perf_evsel *evsel;
+
+       evlist__for_each_entry(ptr->evlist, evsel) {
+               if (evsel->attr.type == ptr->cs_etm_pmu->type)
+                       return perf_evlist__enable_event_idx(ptr->evlist,
+                                                            evsel, idx);
+       }
+
+       return -EINVAL;
+}
+
+struct auxtrace_record *cs_etm_record_init(int *err)
+{
+       struct perf_pmu *cs_etm_pmu;
+       struct cs_etm_recording *ptr;
+
+       cs_etm_pmu = perf_pmu__find(CORESIGHT_ETM_PMU_NAME);
+
+       if (!cs_etm_pmu) {
+               *err = -EINVAL;
+               goto out;
+       }
+
+       ptr = zalloc(sizeof(struct cs_etm_recording));
+       if (!ptr) {
+               *err = -ENOMEM;
+               goto out;
+       }
+
+       ptr->cs_etm_pmu                 = cs_etm_pmu;
+       ptr->itr.parse_snapshot_options = cs_etm_parse_snapshot_options;
+       ptr->itr.recording_options      = cs_etm_recording_options;
+       ptr->itr.info_priv_size         = cs_etm_info_priv_size;
+       ptr->itr.info_fill              = cs_etm_info_fill;
+       ptr->itr.find_snapshot          = cs_etm_find_snapshot;
+       ptr->itr.snapshot_start         = cs_etm_snapshot_start;
+       ptr->itr.snapshot_finish        = cs_etm_snapshot_finish;
+       ptr->itr.reference              = cs_etm_reference;
+       ptr->itr.free                   = cs_etm_recording_free;
+       ptr->itr.read_finish            = cs_etm_read_finish;
+
+       *err = 0;
+       return &ptr->itr;
+out:
+       return NULL;
+}
diff --git a/tools/perf/arch/arm/util/cs-etm.h b/tools/perf/arch/arm/util/cs-etm.h
new file mode 100644 (file)
index 0000000..909f486
--- /dev/null
@@ -0,0 +1,23 @@
+/*
+ * Copyright(C) 2015 Linaro Limited. All rights reserved.
+ * Author: Mathieu Poirier <mathieu.poirier@linaro.org>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef INCLUDE__PERF_CS_ETM_H__
+#define INCLUDE__PERF_CS_ETM_H__
+
+struct auxtrace_record *cs_etm_record_init(int *err);
+
+#endif
index 3876dd05bb8b4e34ed8b1be16ae9e6fdba023e75..cef6fb38d17ebc7fdc5cc867758a43276f11617e 100644 (file)
@@ -1,4 +1,6 @@
 libperf-$(CONFIG_DWARF)     += dwarf-regs.o
 libperf-$(CONFIG_LOCAL_LIBUNWIND) += unwind-libunwind.o
 
-libperf-$(CONFIG_AUXTRACE) += ../../arm/util/pmu.o
+libperf-$(CONFIG_AUXTRACE) += ../../arm/util/pmu.o \
+                             ../../arm/util/auxtrace.o \
+                             ../../arm/util/cs-etm.o
index c9169011e55ef84bf52728f3a0853d68c8120702..c0aba8e839aaa1c6a604e5a634babf2daae38c94 100644 (file)
@@ -892,6 +892,7 @@ int perf_event__process_auxtrace_info(struct perf_tool *tool __maybe_unused,
                return intel_pt_process_auxtrace_info(event, session);
        case PERF_AUXTRACE_INTEL_BTS:
                return intel_bts_process_auxtrace_info(event, session);
+       case PERF_AUXTRACE_CS_ETM:
        case PERF_AUXTRACE_UNKNOWN:
        default:
                return -EINVAL;
index ac5f0d7167e657e505615736cd1875d3226d1317..09286f1935328d41106cd707f6e6cb809f853a56 100644 (file)
@@ -41,6 +41,7 @@ enum auxtrace_type {
        PERF_AUXTRACE_UNKNOWN,
        PERF_AUXTRACE_INTEL_PT,
        PERF_AUXTRACE_INTEL_BTS,
+       PERF_AUXTRACE_CS_ETM,
 };
 
 enum itrace_period_type {
diff --git a/tools/perf/util/cs-etm.h b/tools/perf/util/cs-etm.h
new file mode 100644 (file)
index 0000000..3cc6bc3
--- /dev/null
@@ -0,0 +1,74 @@
+/*
+ * Copyright(C) 2015 Linaro Limited. All rights reserved.
+ * Author: Mathieu Poirier <mathieu.poirier@linaro.org>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef INCLUDE__UTIL_PERF_CS_ETM_H__
+#define INCLUDE__UTIL_PERF_CS_ETM_H__
+
+/* Versionning header in case things need tro change in the future.  That way
+ * decoding of old snapshot is still possible.
+ */
+enum {
+       /* Starting with 0x0 */
+       CS_HEADER_VERSION_0,
+       /* PMU->type (32 bit), total # of CPUs (32 bit) */
+       CS_PMU_TYPE_CPUS,
+       CS_ETM_SNAPSHOT,
+       CS_HEADER_VERSION_0_MAX,
+};
+
+/* Beginning of header common to both ETMv3 and V4 */
+enum {
+       CS_ETM_MAGIC,
+       CS_ETM_CPU,
+};
+
+/* ETMv3/PTM metadata */
+enum {
+       /* Dynamic, configurable parameters */
+       CS_ETM_ETMCR = CS_ETM_CPU + 1,
+       CS_ETM_ETMTRACEIDR,
+       /* RO, taken from sysFS */
+       CS_ETM_ETMCCER,
+       CS_ETM_ETMIDR,
+       CS_ETM_PRIV_MAX,
+};
+
+/* ETMv4 metadata */
+enum {
+       /* Dynamic, configurable parameters */
+       CS_ETMV4_TRCCONFIGR = CS_ETM_CPU + 1,
+       CS_ETMV4_TRCTRACEIDR,
+       /* RO, taken from sysFS */
+       CS_ETMV4_TRCIDR0,
+       CS_ETMV4_TRCIDR1,
+       CS_ETMV4_TRCIDR2,
+       CS_ETMV4_TRCIDR8,
+       CS_ETMV4_TRCAUTHSTATUS,
+       CS_ETMV4_PRIV_MAX,
+};
+
+#define KiB(x) ((x) * 1024)
+#define MiB(x) ((x) * 1024 * 1024)
+
+#define CS_ETM_HEADER_SIZE (CS_HEADER_VERSION_0_MAX * sizeof(u64))
+
+static const u64 __perf_cs_etmv3_magic   = 0x3030303030303030ULL;
+static const u64 __perf_cs_etmv4_magic   = 0x4040404040404040ULL;
+#define CS_ETMV3_PRIV_SIZE (CS_ETM_PRIV_MAX * sizeof(u64))
+#define CS_ETMV4_PRIV_SIZE (CS_ETMV4_PRIV_MAX * sizeof(u64))
+
+#endif