1 /*
2 * Copyright 2020 The Android Open Source Project
3 *
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at
7 *
8 * http://www.apache.org/licenses/LICENSE-2.0
9 *
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
15 */
16
17 #undef LOG_TAG
18 #define LOG_TAG "GpuMem"
19 #define ATRACE_TAG ATRACE_TAG_GRAPHICS
20
21 #include "gpumem/GpuMem.h"
22
23 #include <android-base/stringprintf.h>
24 #include <libbpf.h>
25 #include <bpf/WaitForProgsLoaded.h>
26 #include <log/log.h>
27 #include <unistd.h>
28 #include <utils/Timers.h>
29 #include <utils/Trace.h>
30
31 #include <unordered_map>
32 #include <vector>
33
34 namespace android {
35
36 using base::StringAppendF;
37
~GpuMem()38 GpuMem::~GpuMem() {
39 bpf_detach_tracepoint(kGpuMemTraceGroup, kGpuMemTotalTracepoint);
40 }
41
initialize()42 void GpuMem::initialize() {
43 // Make sure bpf programs are loaded
44 bpf::waitForProgsLoaded();
45
46 errno = 0;
47 int fd = bpf::retrieveProgram(kGpuMemTotalProgPath);
48 if (fd < 0) {
49 ALOGE("Failed to retrieve pinned program from %s [%d(%s)]", kGpuMemTotalProgPath, errno,
50 strerror(errno));
51 return;
52 }
53
54 // Attach the program to the tracepoint, and the tracepoint is automatically enabled here.
55 errno = 0;
56 int count = 0;
57 while (bpf_attach_tracepoint(fd, kGpuMemTraceGroup, kGpuMemTotalTracepoint) < 0) {
58 if (++count > kGpuWaitTimeout) {
59 ALOGE("Failed to attach bpf program to %s/%s tracepoint [%d(%s)]", kGpuMemTraceGroup,
60 kGpuMemTotalTracepoint, errno, strerror(errno));
61 return;
62 }
63 // Retry until GPU driver loaded or timeout.
64 if (mStop.load()) return;
65 sleep(1);
66 }
67
68 // Use the read-only wrapper BpfMapRO to properly retrieve the read-only map.
69 errno = 0;
70 auto map = bpf::BpfMapRO<uint64_t, uint64_t>(kGpuMemTotalMapPath);
71 if (!map.isValid()) {
72 ALOGE("Failed to create bpf map from %s [%d(%s)]", kGpuMemTotalMapPath, errno,
73 strerror(errno));
74 return;
75 }
76 setGpuMemTotalMap(map);
77
78 mInitialized.store(true);
79 }
80
setGpuMemTotalMap(bpf::BpfMapRO<uint64_t,uint64_t> & map)81 void GpuMem::setGpuMemTotalMap(bpf::BpfMapRO<uint64_t, uint64_t>& map) {
82 mGpuMemTotalMap = std::move(map);
83 }
84
85 // Dump the snapshots of global and per process memory usage on all gpus
dump(const Vector<String16> &,std::string * result)86 void GpuMem::dump(const Vector<String16>& /* args */, std::string* result) {
87 ATRACE_CALL();
88
89 if (!mInitialized.load() || !mGpuMemTotalMap.isValid()) {
90 result->append("Failed to initialize GPU memory eBPF\n");
91 return;
92 }
93
94 auto res = mGpuMemTotalMap.getFirstKey();
95 if (!res.ok()) {
96 result->append("GPU memory total usage map is empty\n");
97 return;
98 }
99 uint64_t key = res.value();
100 // unordered_map<gpu_id, vector<pair<pid, size>>>
101 std::unordered_map<uint32_t, std::vector<std::pair<uint32_t, uint64_t>>> dumpMap;
102 while (true) {
103 uint32_t gpu_id = key >> 32;
104 uint32_t pid = key;
105
106 res = mGpuMemTotalMap.readValue(key);
107 if (!res.ok()) break;
108 uint64_t size = res.value();
109
110 dumpMap[gpu_id].emplace_back(pid, size);
111
112 res = mGpuMemTotalMap.getNextKey(key);
113 if (!res.ok()) break;
114 key = res.value();
115 }
116
117 for (auto& gpu : dumpMap) {
118 if (gpu.second.empty()) continue;
119 StringAppendF(result, "Memory snapshot for GPU %u:\n", gpu.first);
120
121 std::sort(gpu.second.begin(), gpu.second.end(),
122 [](auto& l, auto& r) { return l.first < r.first; });
123
124 int i = 0;
125 if (gpu.second[0].first != 0) {
126 StringAppendF(result, "Global total: N/A\n");
127 } else {
128 StringAppendF(result, "Global total: %" PRIu64 "\n", gpu.second[0].second);
129 i++;
130 }
131 for (; i < gpu.second.size(); i++) {
132 StringAppendF(result, "Proc %u total: %" PRIu64 "\n", gpu.second[i].first,
133 gpu.second[i].second);
134 }
135 }
136 }
137
traverseGpuMemTotals(const std::function<void (int64_t ts,uint32_t gpuId,uint32_t pid,uint64_t size)> & callback)138 void GpuMem::traverseGpuMemTotals(const std::function<void(int64_t ts, uint32_t gpuId, uint32_t pid,
139 uint64_t size)>& callback) {
140 auto res = mGpuMemTotalMap.getFirstKey();
141 if (!res.ok()) return;
142 uint64_t key = res.value();
143 while (true) {
144 uint32_t gpu_id = key >> 32;
145 uint32_t pid = key;
146
147 res = mGpuMemTotalMap.readValue(key);
148 if (!res.ok()) break;
149 uint64_t size = res.value();
150
151 callback(systemTime(), gpu_id, pid, size);
152 res = mGpuMemTotalMap.getNextKey(key);
153 if (!res.ok()) break;
154 key = res.value();
155 }
156 }
157
158 } // namespace android
159