1========================================= 2user_events: User-based Event Tracing 3========================================= 4 5:Author: Beau Belgrave 6 7Overview 8-------- 9User based trace events allow user processes to create events and trace data 10that can be viewed via existing tools, such as ftrace and perf. 11To enable this feature, build your kernel with CONFIG_USER_EVENTS=y. 12 13Programs can view status of the events via 14/sys/kernel/tracing/user_events_status and can both register and write 15data out via /sys/kernel/tracing/user_events_data. 16 17Typically programs will register a set of events that they wish to expose to 18tools that can read trace_events (such as ftrace and perf). The registration 19process tells the kernel which address and bit to reflect if any tool has 20enabled the event and data should be written. The registration will give back 21a write index which describes the data when a write() or writev() is called 22on the /sys/kernel/tracing/user_events_data file. 23 24The structures referenced in this document are contained within the 25/include/uapi/linux/user_events.h file in the source tree. 26 27**NOTE:** *Both user_events_status and user_events_data are under the tracefs 28filesystem and may be mounted at different paths than above.* 29 30Registering 31----------- 32Registering within a user process is done via ioctl() out to the 33/sys/kernel/tracing/user_events_data file. The command to issue is 34DIAG_IOCSREG. 35 36This command takes a packed struct user_reg as an argument:: 37 38 struct user_reg { 39 /* Input: Size of the user_reg structure being used */ 40 __u32 size; 41 42 /* Input: Bit in enable address to use */ 43 __u8 enable_bit; 44 45 /* Input: Enable size in bytes at address */ 46 __u8 enable_size; 47 48 /* Input: Flags for future use, set to 0 */ 49 __u16 flags; 50 51 /* Input: Address to update when enabled */ 52 __u64 enable_addr; 53 54 /* Input: Pointer to string with event name, description and flags */ 55 __u64 name_args; 56 57 /* Output: Index of the event to use when writing data */ 58 __u32 write_index; 59 } __attribute__((__packed__)); 60 61The struct user_reg requires all the above inputs to be set appropriately. 62 63+ size: This must be set to sizeof(struct user_reg). 64 65+ enable_bit: The bit to reflect the event status at the address specified by 66 enable_addr. 67 68+ enable_size: The size of the value specified by enable_addr. 69 This must be 4 (32-bit) or 8 (64-bit). 64-bit values are only allowed to be 70 used on 64-bit kernels, however, 32-bit can be used on all kernels. 71 72+ flags: The flags to use, if any. For the initial version this must be 0. 73 Callers should first attempt to use flags and retry without flags to ensure 74 support for lower versions of the kernel. If a flag is not supported -EINVAL 75 is returned. 76 77+ enable_addr: The address of the value to use to reflect event status. This 78 must be naturally aligned and write accessible within the user program. 79 80+ name_args: The name and arguments to describe the event, see command format 81 for details. 82 83Upon successful registration the following is set. 84 85+ write_index: The index to use for this file descriptor that represents this 86 event when writing out data. The index is unique to this instance of the file 87 descriptor that was used for the registration. See writing data for details. 88 89User based events show up under tracefs like any other event under the 90subsystem named "user_events". This means tools that wish to attach to the 91events need to use /sys/kernel/tracing/events/user_events/[name]/enable 92or perf record -e user_events:[name] when attaching/recording. 93 94**NOTE:** The event subsystem name by default is "user_events". Callers should 95not assume it will always be "user_events". Operators reserve the right in the 96future to change the subsystem name per-process to accomodate event isolation. 97 98Command Format 99^^^^^^^^^^^^^^ 100The command string format is as follows:: 101 102 name[:FLAG1[,FLAG2...]] [Field1[;Field2...]] 103 104Supported Flags 105^^^^^^^^^^^^^^^ 106None yet 107 108Field Format 109^^^^^^^^^^^^ 110:: 111 112 type name [size] 113 114Basic types are supported (__data_loc, u32, u64, int, char, char[20], etc). 115User programs are encouraged to use clearly sized types like u32. 116 117**NOTE:** *Long is not supported since size can vary between user and kernel.* 118 119The size is only valid for types that start with a struct prefix. 120This allows user programs to describe custom structs out to tools, if required. 121 122For example, a struct in C that looks like this:: 123 124 struct mytype { 125 char data[20]; 126 }; 127 128Would be represented by the following field:: 129 130 struct mytype myname 20 131 132Deleting 133-------- 134Deleting an event from within a user process is done via ioctl() out to the 135/sys/kernel/tracing/user_events_data file. The command to issue is 136DIAG_IOCSDEL. 137 138This command only requires a single string specifying the event to delete by 139its name. Delete will only succeed if there are no references left to the 140event (in both user and kernel space). User programs should use a separate file 141to request deletes than the one used for registration due to this. 142 143**NOTE:** By default events will auto-delete when there are no references left 144to the event. Flags in the future may change this logic. 145 146Unregistering 147------------- 148If after registering an event it is no longer wanted to be updated then it can 149be disabled via ioctl() out to the /sys/kernel/tracing/user_events_data file. 150The command to issue is DIAG_IOCSUNREG. This is different than deleting, where 151deleting actually removes the event from the system. Unregistering simply tells 152the kernel your process is no longer interested in updates to the event. 153 154This command takes a packed struct user_unreg as an argument:: 155 156 struct user_unreg { 157 /* Input: Size of the user_unreg structure being used */ 158 __u32 size; 159 160 /* Input: Bit to unregister */ 161 __u8 disable_bit; 162 163 /* Input: Reserved, set to 0 */ 164 __u8 __reserved; 165 166 /* Input: Reserved, set to 0 */ 167 __u16 __reserved2; 168 169 /* Input: Address to unregister */ 170 __u64 disable_addr; 171 } __attribute__((__packed__)); 172 173The struct user_unreg requires all the above inputs to be set appropriately. 174 175+ size: This must be set to sizeof(struct user_unreg). 176 177+ disable_bit: This must be set to the bit to disable (same bit that was 178 previously registered via enable_bit). 179 180+ disable_addr: This must be set to the address to disable (same address that was 181 previously registered via enable_addr). 182 183**NOTE:** Events are automatically unregistered when execve() is invoked. During 184fork() the registered events will be retained and must be unregistered manually 185in each process if wanted. 186 187Status 188------ 189When tools attach/record user based events the status of the event is updated 190in realtime. This allows user programs to only incur the cost of the write() or 191writev() calls when something is actively attached to the event. 192 193The kernel will update the specified bit that was registered for the event as 194tools attach/detach from the event. User programs simply check if the bit is set 195to see if something is attached or not. 196 197Administrators can easily check the status of all registered events by reading 198the user_events_status file directly via a terminal. The output is as follows:: 199 200 Name [# Comments] 201 ... 202 203 Active: ActiveCount 204 Busy: BusyCount 205 206For example, on a system that has a single event the output looks like this:: 207 208 test 209 210 Active: 1 211 Busy: 0 212 213If a user enables the user event via ftrace, the output would change to this:: 214 215 test # Used by ftrace 216 217 Active: 1 218 Busy: 1 219 220Writing Data 221------------ 222After registering an event the same fd that was used to register can be used 223to write an entry for that event. The write_index returned must be at the start 224of the data, then the remaining data is treated as the payload of the event. 225 226For example, if write_index returned was 1 and I wanted to write out an int 227payload of the event. Then the data would have to be 8 bytes (2 ints) in size, 228with the first 4 bytes being equal to 1 and the last 4 bytes being equal to the 229value I want as the payload. 230 231In memory this would look like this:: 232 233 int index; 234 int payload; 235 236User programs might have well known structs that they wish to use to emit out 237as payloads. In those cases writev() can be used, with the first vector being 238the index and the following vector(s) being the actual event payload. 239 240For example, if I have a struct like this:: 241 242 struct payload { 243 int src; 244 int dst; 245 int flags; 246 } __attribute__((__packed__)); 247 248It's advised for user programs to do the following:: 249 250 struct iovec io[2]; 251 struct payload e; 252 253 io[0].iov_base = &write_index; 254 io[0].iov_len = sizeof(write_index); 255 io[1].iov_base = &e; 256 io[1].iov_len = sizeof(e); 257 258 writev(fd, (const struct iovec*)io, 2); 259 260**NOTE:** *The write_index is not emitted out into the trace being recorded.* 261 262Example Code 263------------ 264See sample code in samples/user_events. 265