/root/doris/be/src/runtime/thread_context.cpp
Line | Count | Source |
1 | | // Licensed to the Apache Software Foundation (ASF) under one |
2 | | // or more contributor license agreements. See the NOTICE file |
3 | | // distributed with this work for additional information |
4 | | // regarding copyright ownership. The ASF licenses this file |
5 | | // to you under the Apache License, Version 2.0 (the |
6 | | // "License"); you may not use this file except in compliance |
7 | | // with the License. You may obtain a copy of the License at |
8 | | // |
9 | | // http://www.apache.org/licenses/LICENSE-2.0 |
10 | | // |
11 | | // Unless required by applicable law or agreed to in writing, |
12 | | // software distributed under the License is distributed on an |
13 | | // "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY |
14 | | // KIND, either express or implied. See the License for the |
15 | | // specific language governing permissions and limitations |
16 | | // under the License. |
17 | | |
18 | | #include "runtime/thread_context.h" |
19 | | |
20 | | #include <mutex> |
21 | | |
22 | | #include "common/signal_handler.h" |
23 | | #include "runtime/query_context.h" |
24 | | #include "runtime/runtime_state.h" |
25 | | |
26 | | namespace doris { |
27 | | class MemTracker; |
28 | | |
29 | | bthread_key_t btls_key; |
30 | | |
31 | | namespace { |
32 | | |
33 | 56 | void thread_context_deleter(void* data) { |
34 | 56 | delete static_cast<ThreadContext*>(data); |
35 | 56 | } |
36 | | |
37 | | } // namespace |
38 | | |
39 | 2.07M | void init_thread_context_btls_key() { |
40 | 2.07M | static std::once_flag btls_key_once; |
41 | 2.07M | std::call_once(btls_key_once, |
42 | 2.07M | []() { CHECK_EQ(0, bthread_key_create(&btls_key, thread_context_deleter)); }); |
43 | 2.07M | } |
44 | | |
45 | 611 | void AttachTask::init(const std::shared_ptr<ResourceContext>& rc) { |
46 | | // Validate the ResourceContext chain before mutating any thread-local |
47 | | // or signal state. If any link is null we throw immediately, so the |
48 | | // caller's stack-unwind sees a clean state (no thread-local handle is |
49 | | // acquired and no signal task id is set). Without these the previous |
50 | | // code would silently propagate a null mem_tracker into |
51 | | // ThreadMemTrackerMgr and crash much later inside the allocator. |
52 | 611 | if (UNLIKELY(rc == nullptr)) { |
53 | 0 | throw Exception( |
54 | 0 | Status::FatalError("AttachTask::init: rc is null. signal_query_id={:x}-{:x}", |
55 | 0 | signal::query_id_hi, signal::query_id_lo)); |
56 | 0 | } |
57 | 611 | if (UNLIKELY(rc->memory_context() == nullptr)) { |
58 | 0 | throw Exception(Status::FatalError( |
59 | 0 | "AttachTask::init: rc->memory_context() is null. signal_query_id={:x}-{:x}", |
60 | 0 | signal::query_id_hi, signal::query_id_lo)); |
61 | 0 | } |
62 | 611 | if (UNLIKELY(rc->memory_context()->mem_tracker() == nullptr)) { |
63 | 0 | throw Exception(Status::FatalError( |
64 | 0 | "AttachTask::init: rc->memory_context()->mem_tracker() is null. " |
65 | 0 | "ResourceContext was created but set_mem_tracker has not been called yet " |
66 | 0 | "(likely a half-initialized QueryContext used before _init_query_mem_tracker). " |
67 | 0 | "signal_query_id={:x}-{:x}", |
68 | 0 | signal::query_id_hi, signal::query_id_lo)); |
69 | 0 | } |
70 | 611 | if (UNLIKELY(rc->task_controller() == nullptr)) { |
71 | 0 | throw Exception(Status::FatalError( |
72 | 0 | "AttachTask::init: rc->task_controller() is null. signal_query_id={:x}-{:x}", |
73 | 0 | signal::query_id_hi, signal::query_id_lo)); |
74 | 0 | } |
75 | 611 | ThreadLocalHandle::create_thread_local_if_not_exits(); |
76 | 611 | signal::set_signal_task_id(rc->task_controller()->task_id()); |
77 | 611 | thread_context()->attach_task(rc); |
78 | 611 | } |
79 | | |
80 | 216 | AttachTask::AttachTask(const std::shared_ptr<ResourceContext>& rc) { |
81 | 216 | init(rc); |
82 | 216 | } |
83 | | |
84 | 321 | AttachTask::AttachTask(const std::shared_ptr<MemTrackerLimiter>& mem_tracker) { |
85 | | // if parameter is `orphan_mem_tracker`, if you do not switch thraed mem tracker afterwards, |
86 | | // alloc or free memory from Allocator will fail DCHECK. unless you know for sure that |
87 | | // the thread will not alloc or free memory from Allocator later. |
88 | | // |
89 | | // Validate before constructing the ResourceContext: MemoryContext::set_mem_tracker() |
90 | | // immediately dereferences mem_tracker->limit(), so a null shared_ptr would |
91 | | // crash there before reaching init()'s diagnostics. |
92 | 321 | if (UNLIKELY(mem_tracker == nullptr)) { |
93 | 0 | throw Exception(Status::FatalError( |
94 | 0 | "AttachTask(MemTrackerLimiter): mem_tracker is null. signal_query_id={:x}-{:x}", |
95 | 0 | signal::query_id_hi, signal::query_id_lo)); |
96 | 0 | } |
97 | 321 | std::shared_ptr<ResourceContext> rc = ResourceContext::create_shared(); |
98 | 321 | rc->memory_context()->set_mem_tracker(mem_tracker); |
99 | 321 | init(rc); |
100 | 321 | } |
101 | | |
102 | 74 | AttachTask::AttachTask(RuntimeState* runtime_state) { |
103 | | // Walk the chain `runtime_state -> get_query_ctx() -> resource_ctx()` |
104 | | // step by step so that an unexpected null pinpoints exactly which link |
105 | | // failed instead of crashing with a generic NPE inside attach_task() or |
106 | | // even later inside the allocator. |
107 | 74 | if (UNLIKELY(runtime_state == nullptr)) { |
108 | 0 | throw Exception(Status::FatalError( |
109 | 0 | "AttachTask(RuntimeState*): runtime_state is null. signal_query_id={:x}-{:x}", |
110 | 0 | signal::query_id_hi, signal::query_id_lo)); |
111 | 0 | } |
112 | 74 | if (UNLIKELY(runtime_state->get_query_ctx() == nullptr)) { |
113 | 0 | throw Exception(Status::FatalError( |
114 | 0 | "AttachTask(RuntimeState*): runtime_state->get_query_ctx() is null. " |
115 | 0 | "signal_query_id={:x}-{:x}", |
116 | 0 | signal::query_id_hi, signal::query_id_lo)); |
117 | 0 | } |
118 | 74 | if (UNLIKELY(runtime_state->get_query_ctx()->resource_ctx() == nullptr)) { |
119 | 0 | throw Exception( |
120 | 0 | Status::FatalError("AttachTask(RuntimeState*): query_ctx->resource_ctx() is null. " |
121 | 0 | "signal_query_id={:x}-{:x}", |
122 | 0 | signal::query_id_hi, signal::query_id_lo)); |
123 | 0 | } |
124 | 74 | signal::set_signal_is_nereids(runtime_state->is_nereids()); |
125 | 74 | init(runtime_state->get_query_ctx()->resource_ctx()); |
126 | 74 | } |
127 | | |
128 | 0 | AttachTask::AttachTask(QueryContext* query_ctx) { |
129 | 0 | if (UNLIKELY(query_ctx == nullptr)) { |
130 | 0 | throw Exception(Status::FatalError( |
131 | 0 | "AttachTask(QueryContext*): query_ctx is null. signal_query_id={:x}-{:x}", |
132 | 0 | signal::query_id_hi, signal::query_id_lo)); |
133 | 0 | } |
134 | 0 | init(query_ctx->resource_ctx()); |
135 | 0 | } |
136 | | |
137 | 611 | AttachTask::~AttachTask() { |
138 | 611 | signal::set_signal_task_id(TUniqueId()); |
139 | 611 | thread_context()->detach_task(); |
140 | 611 | ThreadLocalHandle::del_thread_local_if_count_is_zero(); |
141 | 611 | } |
142 | | |
143 | 5 | SwitchResourceContext::SwitchResourceContext(const std::shared_ptr<ResourceContext>& rc) { |
144 | 5 | DCHECK(rc != nullptr); |
145 | | // Validate the chain before mutating any thread-local or signal state, |
146 | | // symmetric to AttachTask::init(). Throwing after the thread-local |
147 | | // handle was acquired or the signal task id was set would skip this |
148 | | // object's destructor (because construction failed) and leak the |
149 | | // handle / leave a stale signal task id behind. |
150 | 5 | if (UNLIKELY(rc == nullptr)) { |
151 | 0 | throw Exception( |
152 | 0 | Status::FatalError("SwitchResourceContext: rc is null. signal_query_id={:x}-{:x}", |
153 | 0 | signal::query_id_hi, signal::query_id_lo)); |
154 | 0 | } |
155 | 5 | if (UNLIKELY(rc->memory_context() == nullptr)) { |
156 | 0 | throw Exception(Status::FatalError( |
157 | 0 | "SwitchResourceContext: rc->memory_context() is null. signal_query_id={:x}-{:x}", |
158 | 0 | signal::query_id_hi, signal::query_id_lo)); |
159 | 0 | } |
160 | 5 | if (UNLIKELY(rc->memory_context()->mem_tracker() == nullptr)) { |
161 | 0 | throw Exception(Status::FatalError( |
162 | 0 | "SwitchResourceContext: rc->memory_context()->mem_tracker() is null. " |
163 | 0 | "ResourceContext was switched in before _init_query_mem_tracker ran. " |
164 | 0 | "signal_query_id={:x}-{:x}", |
165 | 0 | signal::query_id_hi, signal::query_id_lo)); |
166 | 0 | } |
167 | 5 | if (UNLIKELY(rc->task_controller() == nullptr)) { |
168 | 0 | throw Exception(Status::FatalError( |
169 | 0 | "SwitchResourceContext: rc->task_controller() is null. signal_query_id={:x}-{:x}", |
170 | 0 | signal::query_id_hi, signal::query_id_lo)); |
171 | 0 | } |
172 | 5 | doris::ThreadLocalHandle::create_thread_local_if_not_exits(); |
173 | 5 | DCHECK(thread_context()->is_attach_task()); |
174 | 5 | old_resource_ctx_ = thread_context()->resource_ctx(); |
175 | 5 | if (rc != old_resource_ctx_) { |
176 | 4 | signal::set_signal_task_id(rc->task_controller()->task_id()); |
177 | 4 | thread_context()->resource_ctx_ = rc; |
178 | 4 | thread_context()->thread_mem_tracker_mgr->attach_limiter_tracker( |
179 | 4 | rc->memory_context()->mem_tracker(), rc->workload_group()); |
180 | 4 | } |
181 | 5 | } |
182 | | |
183 | 5 | SwitchResourceContext::~SwitchResourceContext() { |
184 | 5 | if (old_resource_ctx_ != thread_context()->resource_ctx()) { |
185 | 4 | DCHECK(old_resource_ctx_ != nullptr); |
186 | 4 | signal::set_signal_task_id(old_resource_ctx_->task_controller()->task_id()); |
187 | 4 | thread_context()->resource_ctx_ = old_resource_ctx_; |
188 | 4 | thread_context()->thread_mem_tracker_mgr->detach_limiter_tracker(); |
189 | 4 | } |
190 | 5 | doris::ThreadLocalHandle::del_thread_local_if_count_is_zero(); |
191 | 5 | } |
192 | | |
193 | | SwitchThreadMemTrackerLimiter::SwitchThreadMemTrackerLimiter( |
194 | 1.07M | const std::shared_ptr<doris::MemTrackerLimiter>& mem_tracker) { |
195 | 1.07M | DCHECK(mem_tracker); |
196 | | // Third entry point that calls attach_limiter_tracker(). Without this |
197 | | // null guard a null mem_tracker silently propagates and the next |
198 | | // allocation on this thread would NPE deep inside the allocator. Throw |
199 | | // before acquiring the thread-local handle / doing any side effect so |
200 | | // the destructor (which is noexcept) never runs in a dirty state. |
201 | 1.07M | if (UNLIKELY(mem_tracker == nullptr)) { |
202 | 0 | throw Exception(Status::FatalError( |
203 | 0 | "SwitchThreadMemTrackerLimiter: mem_tracker is null. signal_query_id={:x}-{:x}", |
204 | 0 | signal::query_id_hi, signal::query_id_lo)); |
205 | 0 | } |
206 | 1.07M | doris::ThreadLocalHandle::create_thread_local_if_not_exits(); |
207 | 1.07M | if (mem_tracker != thread_context()->thread_mem_tracker_mgr->limiter_mem_tracker_sptr()) { |
208 | 992k | thread_context()->thread_mem_tracker_mgr->attach_limiter_tracker(mem_tracker); |
209 | 992k | is_switched_ = true; |
210 | 992k | } |
211 | 1.07M | } |
212 | | |
213 | 1.07M | SwitchThreadMemTrackerLimiter::~SwitchThreadMemTrackerLimiter() { |
214 | 1.07M | if (is_switched_) { |
215 | 992k | thread_context()->thread_mem_tracker_mgr->detach_limiter_tracker(); |
216 | 992k | } |
217 | 1.07M | doris::ThreadLocalHandle::del_thread_local_if_count_is_zero(); |
218 | 1.07M | } |
219 | | |
220 | 96 | AddThreadMemTrackerConsumer::AddThreadMemTrackerConsumer(MemTracker* mem_tracker) { |
221 | 96 | ThreadLocalHandle::create_thread_local_if_not_exits(); |
222 | 96 | if (mem_tracker) { |
223 | 96 | _need_pop = thread_context()->thread_mem_tracker_mgr->push_consumer_tracker(mem_tracker); |
224 | 96 | } |
225 | 96 | } |
226 | | |
227 | | AddThreadMemTrackerConsumer::AddThreadMemTrackerConsumer( |
228 | | const std::shared_ptr<MemTracker>& mem_tracker) |
229 | 76 | : _mem_tracker(mem_tracker) { |
230 | 76 | ThreadLocalHandle::create_thread_local_if_not_exits(); |
231 | 76 | if (_mem_tracker) { |
232 | 76 | _need_pop = |
233 | 76 | thread_context()->thread_mem_tracker_mgr->push_consumer_tracker(_mem_tracker.get()); |
234 | 76 | } |
235 | 76 | } |
236 | | |
237 | 172 | AddThreadMemTrackerConsumer::~AddThreadMemTrackerConsumer() { |
238 | 172 | if (_need_pop) { |
239 | 171 | thread_context()->thread_mem_tracker_mgr->pop_consumer_tracker(); |
240 | 171 | } |
241 | 172 | ThreadLocalHandle::del_thread_local_if_count_is_zero(); |
242 | 172 | } |
243 | | |
244 | | } // namespace doris |