libdatadog 11.0.0.1.0-x86_64-linux → 12.0.0.1.0-x86_64-linux

data/vendor/libdatadog-12.0.0/x86_64-linux-musl/libdatadog-x86_64-alpine-linux-musl/include/datadog/crashtracker.h

@@ -0,0 +1,419 @@
+// Copyright 2021-Present Datadog, Inc. https://www.datadoghq.com/
+// SPDX-License-Identifier: Apache-2.0
+
+
+#ifndef DDOG_CRASHTRACKER_H
+#define DDOG_CRASHTRACKER_H
+
+#pragma once
+
+#include <stdbool.h>
+#include <stddef.h>
+#include <stdint.h>
+#include "common.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif // __cplusplus
+
+/**
+ * Cleans up after the crash-tracker:
+ * Unregister the crash handler, restore the previous handler (if any), and
+ * shut down the receiver.  Note that the use of this function is optional:
+ * the receiver will automatically shutdown when the pipe is closed on program
+ * exit.
+ *
+ * # Preconditions
+ *   This function assumes that the crashtracker has previously been
+ *   initialized.
+ * # Safety
+ *   Crash-tracking functions are not reentrant.
+ *   No other crash-handler functions should be called concurrently.
+ * # Atomicity
+ *   This function is not atomic. A crash during its execution may lead to
+ *   unexpected crash-handling behaviour.
+ */
+DDOG_CHECK_RETURN struct ddog_crasht_Result ddog_crasht_shutdown(void);
+
+/**
+ * Reinitialize the crash-tracking infrastructure after a fork.
+ * This should be one of the first things done after a fork, to minimize the
+ * chance that a crash occurs between the fork, and this call.
+ * In particular, reset the counters that track the profiler state machine,
+ * and start a new receiver to collect data from this fork.
+ * NOTE: An alternative design would be to have a 1:many sidecar listening on a
+ * socket instead of 1:1 receiver listening on a pipe, but the only real
+ * advantage would be to have fewer processes in `ps -a`.
+ *
+ * # Preconditions
+ *   This function assumes that the crash-tracker has previously been
+ *   initialized.
+ * # Safety
+ *   Crash-tracking functions are not reentrant.
+ *   No other crash-handler functions should be called concurrently.
+ * # Atomicity
+ *   This function is not atomic. A crash during its execution may lead to
+ *   unexpected crash-handling behaviour.
+ */
+DDOG_CHECK_RETURN
+struct ddog_crasht_Result ddog_crasht_update_on_fork(struct ddog_crasht_Config config,
+                                                     struct ddog_crasht_ReceiverConfig receiver_config,
+                                                     struct ddog_crasht_Metadata metadata);
+
+/**
+ * Initialize the crash-tracking infrastructure.
+ *
+ * # Preconditions
+ *   None.
+ * # Safety
+ *   Crash-tracking functions are not reentrant.
+ *   No other crash-handler functions should be called concurrently.
+ * # Atomicity
+ *   This function is not atomic. A crash during its execution may lead to
+ *   unexpected crash-handling behaviour.
+ */
+DDOG_CHECK_RETURN
+struct ddog_crasht_Result ddog_crasht_init_with_receiver(struct ddog_crasht_Config config,
+                                                         struct ddog_crasht_ReceiverConfig receiver_config,
+                                                         struct ddog_crasht_Metadata metadata);
+
+/**
+ * Resets all counters to 0.
+ * Expected to be used after a fork, to reset the counters on the child
+ * ATOMICITY:
+ *     This is NOT ATOMIC.
+ *     Should only be used when no conflicting updates can occur,
+ *     e.g. after a fork but before profiling ops start on the child.
+ * # Safety
+ * No safety concerns.
+ */
+DDOG_CHECK_RETURN struct ddog_crasht_Result ddog_crasht_reset_counters(void);
+
+/**
+ * Atomically increments the count associated with `op`.
+ * Useful for tracking what operations were occuring when a crash occurred.
+ *
+ * # Safety
+ * No safety concerns.
+ */
+DDOG_CHECK_RETURN struct ddog_crasht_Result ddog_crasht_begin_op(enum ddog_crasht_OpTypes op);
+
+/**
+ * Atomically decrements the count associated with `op`.
+ * Useful for tracking what operations were occuring when a crash occurred.
+ *
+ * # Safety
+ * No safety concerns.
+ */
+DDOG_CHECK_RETURN struct ddog_crasht_Result ddog_crasht_end_op(enum ddog_crasht_OpTypes op);
+
+/**
+ * Resets all stored spans to 0.
+ * Expected to be used after a fork, to reset the spans on the child
+ * ATOMICITY:
+ *     This is NOT ATOMIC.
+ *     Should only be used when no conflicting updates can occur,
+ *     e.g. after a fork but before profiling ops start on the child.
+ * # Safety
+ * No safety concerns.
+ */
+DDOG_CHECK_RETURN struct ddog_crasht_Result ddog_crasht_clear_span_ids(void);
+
+/**
+ * Resets all stored traces to 0.
+ * Expected to be used after a fork, to reset the traces on the child
+ * ATOMICITY:
+ *     This is NOT ATOMIC.
+ *     Should only be used when no conflicting updates can occur,
+ *     e.g. after a fork but before profiling ops start on the child.
+ * # Safety
+ * No safety concerns.
+ */
+DDOG_CHECK_RETURN struct ddog_crasht_Result ddog_crasht_clear_trace_ids(void);
+
+/**
+ * Atomically registers an active traceId.
+ * Useful for tracking what operations were occurring when a crash occurred.
+ * 0 is reserved for "NoId"
+ * The set does not check for duplicates.  Adding the same id twice is an error.
+ *
+ * Inputs:
+ * id<high/low>: the 128 bit id, broken into 2 64 bit chunks (see note)
+ *
+ * Returns:
+ *   Ok(handle) on success.  The handle is needed to later remove the id;
+ *   Err() on failure. The most likely cause of failure is that the underlying set is full.
+ *
+ * Note: 128 bit ints in FFI were not stabilized until Rust 1.77
+ * https://blog.rust-lang.org/2024/03/30/i128-layout-update.html
+ * We're currently locked into 1.71, have to do an ugly workaround involving 2 64 bit ints
+ * until we can upgrade.
+ *
+ * # Safety
+ * No safety concerns.
+ */
+DDOG_CHECK_RETURN
+struct ddog_crasht_UsizeResult ddog_crasht_insert_trace_id(uint64_t id_high,
+                                                           uint64_t id_low);
+
+/**
+ * Atomically registers an active SpanId.
+ * Useful for tracking what operations were occurring when a crash occurred.
+ * 0 is reserved for "NoId".
+ * The set does not check for duplicates.  Adding the same id twice is an error.
+ *
+ * Inputs:
+ * id<high/low>: the 128 bit id, broken into 2 64 bit chunks (see note)
+ *
+ * Returns:
+ *   Ok(handle) on success.  The handle is needed to later remove the id;
+ *   Err() on failure. The most likely cause of failure is that the underlying set is full.
+ *
+ * Note: 128 bit ints in FFI were not stabilized until Rust 1.77
+ * https://blog.rust-lang.org/2024/03/30/i128-layout-update.html
+ * We're currently locked into 1.71, have to do an ugly workaround involving 2 64 bit ints
+ * until we can upgrade.
+ *
+ * # Safety
+ * No safety concerns.
+ */
+DDOG_CHECK_RETURN
+struct ddog_crasht_UsizeResult ddog_crasht_insert_span_id(uint64_t id_high,
+                                                          uint64_t id_low);
+
+/**
+ * Atomically removes a completed SpanId.
+ * Useful for tracking what operations were occurring when a crash occurred.
+ * 0 is reserved for "NoId"
+ *
+ * Inputs:
+ * id<high/low>: the 128 bit id, broken into 2 64 bit chunks (see note)
+ * idx: The handle for the id, from a previous successful call to `insert_span_id`.
+ *      Attempting to remove the same element twice is an error.
+ * Returns:
+ *   `Ok` on success.
+ *   `Err` on failure.  If `id` is not found at `idx`, `Err` will be returned and the set will not
+ *                      be modified.
+ *
+ * Note: 128 bit ints in FFI were not stabilized until Rust 1.77
+ * https://blog.rust-lang.org/2024/03/30/i128-layout-update.html
+ * We're currently locked into 1.71, have to do an ugly workaround involving 2 64 bit ints
+ * until we can upgrade.
+ *
+ * # Safety
+ * No safety concerns.
+ */
+DDOG_CHECK_RETURN
+struct ddog_crasht_Result ddog_crasht_remove_span_id(uint64_t id_high,
+                                                     uint64_t id_low,
+                                                     uintptr_t idx);
+
+/**
+ * Atomically removes a completed TraceId.
+ * Useful for tracking what operations were occurring when a crash occurred.
+ * 0 is reserved for "NoId"
+ *
+ * Inputs:
+ * id<high/low>: the 128 bit id, broken into 2 64 bit chunks (see note)
+ * idx: The handle for the id, from a previous successful call to `insert_span_id`.
+ *      Attempting to remove the same element twice is an error.
+ * Returns:
+ *   `Ok` on success.
+ *   `Err` on failure.  If `id` is not found at `idx`, `Err` will be returned and the set will not
+ *                      be modified.
+ *
+ * Note: 128 bit ints in FFI were not stabilized until Rust 1.77
+ * https://blog.rust-lang.org/2024/03/30/i128-layout-update.html
+ * We're currently locked into 1.71, have to do an ugly workaround involving 2 64 bit ints
+ * until we can upgrade.
+ *
+ * # Safety
+ * No safety concerns.
+ */
+DDOG_CHECK_RETURN
+struct ddog_crasht_Result ddog_crasht_remove_trace_id(uint64_t id_high,
+                                                      uint64_t id_low,
+                                                      uintptr_t idx);
+
+/**
+ * Create a new crashinfo, and returns an opaque reference to it.
+ * # Safety
+ * No safety issues.
+ */
+DDOG_CHECK_RETURN struct ddog_crasht_CrashInfoNewResult ddog_crasht_CrashInfo_new(void);
+
+/**
+ * # Safety
+ * The `crash_info` can be null, but if non-null it must point to a CrashInfo
+ * made by this module, which has not previously been dropped.
+ */
+void ddog_crasht_CrashInfo_drop(struct ddog_crasht_CrashInfo *crashinfo);
+
+/**
+ * Best effort attempt to normalize all `ip` on the stacktrace.
+ * `pid` must be the pid of the currently active process where the ips came from.
+ *
+ * # Safety
+ * `crashinfo` must be a valid pointer to a `CrashInfo` object.
+ */
+DDOG_CHECK_RETURN
+struct ddog_crasht_Result ddog_crasht_CrashInfo_normalize_ips(struct ddog_crasht_CrashInfo *crashinfo,
+                                                              uint32_t pid);
+
+/**
+ * Adds a "counter" variable, with the given value.  Useful for determining if
+ * "interesting" operations were occurring when the crash did.
+ *
+ * # Safety
+ * `crashinfo` must be a valid pointer to a `CrashInfo` object.
+ * `name` should be a valid reference to a utf8 encoded String.
+ * The string is copied into the crashinfo, so it does not need to outlive this
+ * call.
+ */
+DDOG_CHECK_RETURN
+struct ddog_crasht_Result ddog_crasht_CrashInfo_add_counter(struct ddog_crasht_CrashInfo *crashinfo,
+                                                            ddog_CharSlice name,
+                                                            int64_t val);
+
+/**
+ * Adds the contents of "file" to the crashinfo
+ *
+ * # Safety
+ * `crashinfo` must be a valid pointer to a `CrashInfo` object.
+ * `name` should be a valid reference to a utf8 encoded String.
+ * The string is copied into the crashinfo, so it does not need to outlive this
+ * call.
+ */
+DDOG_CHECK_RETURN
+struct ddog_crasht_Result ddog_crasht_CrashInfo_add_file(struct ddog_crasht_CrashInfo *crashinfo,
+                                                         ddog_CharSlice filename);
+
+/**
+ * Adds the tag with given "key" and "value" to the crashinfo
+ *
+ * # Safety
+ * `crashinfo` must be a valid pointer to a `CrashInfo` object.
+ * `key` should be a valid reference to a utf8 encoded String.
+ * `value` should be a valid reference to a utf8 encoded String.
+ * The string is copied into the crashinfo, so it does not need to outlive this
+ * call.
+ */
+DDOG_CHECK_RETURN
+struct ddog_crasht_Result ddog_crasht_CrashInfo_add_tag(struct ddog_crasht_CrashInfo *crashinfo,
+                                                        ddog_CharSlice key,
+                                                        ddog_CharSlice value);
+
+/**
+ * Sets the crashinfo metadata
+ *
+ * # Safety
+ * `crashinfo` must be a valid pointer to a `CrashInfo` object.
+ * All references inside `metadata` must be valid.
+ * Strings are copied into the crashinfo, and do not need to outlive this call.
+ */
+DDOG_CHECK_RETURN
+struct ddog_crasht_Result ddog_crasht_CrashInfo_set_metadata(struct ddog_crasht_CrashInfo *crashinfo,
+                                                             struct ddog_crasht_Metadata metadata);
+
+/**
+ * Sets the crashinfo siginfo
+ *
+ * # Safety
+ * `crashinfo` must be a valid pointer to a `CrashInfo` object.
+ * All references inside `metadata` must be valid.
+ * Strings are copied into the crashinfo, and do not need to outlive this call.
+ */
+DDOG_CHECK_RETURN
+struct ddog_crasht_Result ddog_crasht_CrashInfo_set_siginfo(struct ddog_crasht_CrashInfo *crashinfo,
+                                                            struct ddog_crasht_SigInfo siginfo);
+
+/**
+ * If `thread_id` is empty, sets `stacktrace` as the default stacktrace.
+ * Otherwise, adds an additional stacktrace with id "thread_id".
+ *
+ * # Safety
+ * `crashinfo` must be a valid pointer to a `CrashInfo` object.
+ * All references inside `stacktraces` must be valid.
+ * Strings are copied into the crashinfo, and do not need to outlive this call.
+ */
+DDOG_CHECK_RETURN
+struct ddog_crasht_Result ddog_crasht_CrashInfo_set_stacktrace(struct ddog_crasht_CrashInfo *crashinfo,
+                                                               ddog_CharSlice thread_id,
+                                                               struct ddog_crasht_Slice_StackFrame stacktrace);
+
+/**
+ * Sets the timestamp to the given unix timestamp
+ *
+ * # Safety
+ * `crashinfo` must be a valid pointer to a `CrashInfo` object.
+ */
+DDOG_CHECK_RETURN
+struct ddog_crasht_Result ddog_crasht_CrashInfo_set_timestamp(struct ddog_crasht_CrashInfo *crashinfo,
+                                                              struct ddog_Timespec ts);
+
+/**
+ * Sets the timestamp to the current time
+ *
+ * # Safety
+ * `crashinfo` must be a valid pointer to a `CrashInfo` object.
+ */
+DDOG_CHECK_RETURN
+struct ddog_crasht_Result ddog_crasht_CrashInfo_set_timestamp_to_now(struct ddog_crasht_CrashInfo *crashinfo);
+
+/**
+ * Exports `crashinfo` to the backend at `endpoint`
+ * Note that we support the "file://" endpoint for local file output.
+ * # Safety
+ * `crashinfo` must be a valid pointer to a `CrashInfo` object.
+ */
+DDOG_CHECK_RETURN
+struct ddog_crasht_Result ddog_crasht_CrashInfo_upload_to_endpoint(struct ddog_crasht_CrashInfo *crashinfo,
+                                                                   const struct ddog_Endpoint *endpoint);
+
+/**
+ * Demangles the string "name".
+ * If demangling fails, returns an empty string ""
+ *
+ * # Safety
+ * `name` should be a valid reference to a utf8 encoded String.
+ * The string is copied into the result, and does not need to outlive this call
+ */
+DDOG_CHECK_RETURN
+struct ddog_crasht_StringWrapperResult ddog_crasht_demangle(ddog_CharSlice name,
+                                                            enum ddog_crasht_DemangleOptions options);
+
+/**
+ * Receives data from a crash collector via a pipe on `stdin`, formats it into
+ * `CrashInfo` json, and emits it to the endpoint/file defined in `config`.
+ *
+ * At a high-level, this exists because doing anything in a
+ * signal handler is dangerous, so we fork a sidecar to do the stuff we aren't
+ * allowed to do in the handler.
+ *
+ * See comments in [crashtracker/lib.rs] for a full architecture description.
+ * # Safety
+ * No safety concerns
+ */
+DDOG_CHECK_RETURN struct ddog_crasht_Result ddog_crasht_receiver_entry_point_stdin(void);
+
+/**
+ * Receives data from a crash collector via a pipe on `stdin`, formats it into
+ * `CrashInfo` json, and emits it to the endpoint/file defined in `config`.
+ *
+ * At a high-level, this exists because doing anything in a
+ * signal handler is dangerous, so we fork a sidecar to do the stuff we aren't
+ * allowed to do in the handler.
+ *
+ * See comments in [profiling/crashtracker/mod.rs] for a full architecture
+ * description.
+ * # Safety
+ * No safety concerns
+ */
+DDOG_CHECK_RETURN
+struct ddog_crasht_Result ddog_crasht_receiver_entry_point_unix_socket(ddog_CharSlice socket_path);
+
+#ifdef __cplusplus
+} // extern "C"
+#endif // __cplusplus
+
+#endif /* DDOG_CRASHTRACKER_H */