]>
Commit | Line | Data |
---|---|---|
5ce8f203 UD |
1 | @node Resource Usage And Limitation, Non-Local Exits, Date and Time, Top |
2 | @c %MENU% Functions for examining resource usage and getting and setting limits | |
3 | @chapter Resource Usage And Limitation | |
4 | This chapter describes functions for examining how much of various kinds of | |
5 | resources (CPU time, memory, etc.) a process has used and getting and setting | |
6 | limits on future usage. | |
7 | ||
8 | @menu | |
9 | * Resource Usage:: Measuring various resources used. | |
10 | * Limits on Resources:: Specifying limits on resource usage. | |
11 | * Priority:: Reading or setting process run priority. | |
b642f101 UD |
12 | * Memory Resources:: Querying memory available resources. |
13 | * Processor Resources:: Learn about the processors available. | |
5ce8f203 UD |
14 | @end menu |
15 | ||
16 | ||
17 | @node Resource Usage | |
18 | @section Resource Usage | |
19 | ||
20 | @pindex sys/resource.h | |
21 | The function @code{getrusage} and the data type @code{struct rusage} | |
22 | are used to examine the resource usage of a process. They are declared | |
23 | in @file{sys/resource.h}. | |
24 | ||
25 | @comment sys/resource.h | |
26 | @comment BSD | |
27 | @deftypefun int getrusage (int @var{processes}, struct rusage *@var{rusage}) | |
c8ce789c AO |
28 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}} |
29 | @c On HURD, this calls task_info 3 times. On UNIX, it's a syscall. | |
5ce8f203 UD |
30 | This function reports resource usage totals for processes specified by |
31 | @var{processes}, storing the information in @code{*@var{rusage}}. | |
32 | ||
33 | In most systems, @var{processes} has only two valid values: | |
34 | ||
35 | @table @code | |
36 | @comment sys/resource.h | |
37 | @comment BSD | |
38 | @item RUSAGE_SELF | |
39 | Just the current process. | |
40 | ||
41 | @comment sys/resource.h | |
42 | @comment BSD | |
43 | @item RUSAGE_CHILDREN | |
44 | All child processes (direct and indirect) that have already terminated. | |
45 | @end table | |
46 | ||
5ce8f203 UD |
47 | The return value of @code{getrusage} is zero for success, and @code{-1} |
48 | for failure. | |
49 | ||
50 | @table @code | |
51 | @item EINVAL | |
52 | The argument @var{processes} is not valid. | |
53 | @end table | |
54 | @end deftypefun | |
55 | ||
56 | One way of getting resource usage for a particular child process is with | |
57 | the function @code{wait4}, which returns totals for a child when it | |
58 | terminates. @xref{BSD Wait Functions}. | |
59 | ||
60 | @comment sys/resource.h | |
61 | @comment BSD | |
62 | @deftp {Data Type} {struct rusage} | |
63 | This data type stores various resource usage statistics. It has the | |
64 | following members, and possibly others: | |
65 | ||
66 | @table @code | |
67 | @item struct timeval ru_utime | |
68 | Time spent executing user instructions. | |
69 | ||
70 | @item struct timeval ru_stime | |
71 | Time spent in operating system code on behalf of @var{processes}. | |
72 | ||
73 | @item long int ru_maxrss | |
74 | The maximum resident set size used, in kilobytes. That is, the maximum | |
75 | number of kilobytes of physical memory that @var{processes} used | |
76 | simultaneously. | |
77 | ||
78 | @item long int ru_ixrss | |
79 | An integral value expressed in kilobytes times ticks of execution, which | |
80 | indicates the amount of memory used by text that was shared with other | |
81 | processes. | |
82 | ||
83 | @item long int ru_idrss | |
84 | An integral value expressed the same way, which is the amount of | |
85 | unshared memory used for data. | |
86 | ||
87 | @item long int ru_isrss | |
88 | An integral value expressed the same way, which is the amount of | |
89 | unshared memory used for stack space. | |
90 | ||
91 | @item long int ru_minflt | |
92 | The number of page faults which were serviced without requiring any I/O. | |
93 | ||
94 | @item long int ru_majflt | |
95 | The number of page faults which were serviced by doing I/O. | |
96 | ||
97 | @item long int ru_nswap | |
98 | The number of times @var{processes} was swapped entirely out of main memory. | |
99 | ||
100 | @item long int ru_inblock | |
101 | The number of times the file system had to read from the disk on behalf | |
102 | of @var{processes}. | |
103 | ||
104 | @item long int ru_oublock | |
105 | The number of times the file system had to write to the disk on behalf | |
106 | of @var{processes}. | |
107 | ||
108 | @item long int ru_msgsnd | |
109 | Number of IPC messages sent. | |
110 | ||
111 | @item long int ru_msgrcv | |
112 | Number of IPC messages received. | |
113 | ||
114 | @item long int ru_nsignals | |
115 | Number of signals received. | |
116 | ||
117 | @item long int ru_nvcsw | |
118 | The number of times @var{processes} voluntarily invoked a context switch | |
119 | (usually to wait for some service). | |
120 | ||
121 | @item long int ru_nivcsw | |
122 | The number of times an involuntary context switch took place (because | |
123 | a time slice expired, or another process of higher priority was | |
124 | scheduled). | |
125 | @end table | |
126 | @end deftp | |
127 | ||
128 | @code{vtimes} is a historical function that does some of what | |
129 | @code{getrusage} does. @code{getrusage} is a better choice. | |
130 | ||
131 | @code{vtimes} and its @code{vtimes} data structure are declared in | |
132 | @file{sys/vtimes.h}. | |
133 | @pindex sys/vtimes.h | |
5ce8f203 | 134 | |
8ded91fb RM |
135 | @comment sys/vtimes.h |
136 | @deftypefun int vtimes (struct vtimes *@var{current}, struct vtimes *@var{child}) | |
c8ce789c AO |
137 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}} |
138 | @c Calls getrusage twice. | |
5ce8f203 UD |
139 | |
140 | @code{vtimes} reports resource usage totals for a process. | |
141 | ||
142 | If @var{current} is non-null, @code{vtimes} stores resource usage totals for | |
143 | the invoking process alone in the structure to which it points. If | |
144 | @var{child} is non-null, @code{vtimes} stores resource usage totals for all | |
145 | past children (which have terminated) of the invoking process in the structure | |
146 | to which it points. | |
147 | ||
148 | @deftp {Data Type} {struct vtimes} | |
149 | This data type contains information about the resource usage of a process. | |
150 | Each member corresponds to a member of the @code{struct rusage} data type | |
151 | described above. | |
152 | ||
153 | @table @code | |
154 | @item vm_utime | |
155 | User CPU time. Analogous to @code{ru_utime} in @code{struct rusage} | |
156 | @item vm_stime | |
157 | System CPU time. Analogous to @code{ru_stime} in @code{struct rusage} | |
158 | @item vm_idsrss | |
159 | Data and stack memory. The sum of the values that would be reported as | |
160 | @code{ru_idrss} and @code{ru_isrss} in @code{struct rusage} | |
161 | @item vm_ixrss | |
162 | Shared memory. Analogous to @code{ru_ixrss} in @code{struct rusage} | |
163 | @item vm_maxrss | |
164 | Maximent resident set size. Analogous to @code{ru_maxrss} in | |
165 | @code{struct rusage} | |
166 | @item vm_majflt | |
167 | Major page faults. Analogous to @code{ru_majflt} in @code{struct rusage} | |
168 | @item vm_minflt | |
169 | Minor page faults. Analogous to @code{ru_minflt} in @code{struct rusage} | |
170 | @item vm_nswap | |
171 | Swap count. Analogous to @code{ru_nswap} in @code{struct rusage} | |
172 | @item vm_inblk | |
173 | Disk reads. Analogous to @code{ru_inblk} in @code{struct rusage} | |
174 | @item vm_oublk | |
175 | Disk writes. Analogous to @code{ru_oublk} in @code{struct rusage} | |
176 | @end table | |
177 | @end deftp | |
178 | ||
179 | ||
180 | The return value is zero if the function succeeds; @code{-1} otherwise. | |
181 | ||
182 | ||
183 | ||
184 | @end deftypefun | |
185 | An additional historical function for examining resource usage, | |
186 | @code{vtimes}, is supported but not documented here. It is declared in | |
187 | @file{sys/vtimes.h}. | |
188 | ||
189 | @node Limits on Resources | |
190 | @section Limiting Resource Usage | |
191 | @cindex resource limits | |
192 | @cindex limits on resource usage | |
193 | @cindex usage limits | |
194 | ||
195 | You can specify limits for the resource usage of a process. When the | |
196 | process tries to exceed a limit, it may get a signal, or the system call | |
197 | by which it tried to do so may fail, depending on the resource. Each | |
198 | process initially inherits its limit values from its parent, but it can | |
199 | subsequently change them. | |
200 | ||
201 | There are two per-process limits associated with a resource: | |
202 | @cindex limit | |
203 | ||
204 | @table @dfn | |
205 | @item current limit | |
206 | The current limit is the value the system will not allow usage to | |
207 | exceed. It is also called the ``soft limit'' because the process being | |
208 | limited can generally raise the current limit at will. | |
209 | @cindex current limit | |
210 | @cindex soft limit | |
211 | ||
212 | @item maximum limit | |
213 | The maximum limit is the maximum value to which a process is allowed to | |
214 | set its current limit. It is also called the ``hard limit'' because | |
215 | there is no way for a process to get around it. A process may lower | |
216 | its own maximum limit, but only the superuser may increase a maximum | |
217 | limit. | |
218 | @cindex maximum limit | |
219 | @cindex hard limit | |
220 | @end table | |
221 | ||
222 | @pindex sys/resource.h | |
223 | The symbols for use with @code{getrlimit}, @code{setrlimit}, | |
0bc93a2f | 224 | @code{getrlimit64}, and @code{setrlimit64} are defined in |
5ce8f203 UD |
225 | @file{sys/resource.h}. |
226 | ||
227 | @comment sys/resource.h | |
228 | @comment BSD | |
229 | @deftypefun int getrlimit (int @var{resource}, struct rlimit *@var{rlp}) | |
c8ce789c AO |
230 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}} |
231 | @c Direct syscall on most systems. | |
5ce8f203 UD |
232 | Read the current and maximum limits for the resource @var{resource} |
233 | and store them in @code{*@var{rlp}}. | |
234 | ||
235 | The return value is @code{0} on success and @code{-1} on failure. The | |
236 | only possible @code{errno} error condition is @code{EFAULT}. | |
237 | ||
238 | When the sources are compiled with @code{_FILE_OFFSET_BITS == 64} on a | |
239 | 32-bit system this function is in fact @code{getrlimit64}. Thus, the | |
240 | LFS interface transparently replaces the old interface. | |
241 | @end deftypefun | |
242 | ||
243 | @comment sys/resource.h | |
244 | @comment Unix98 | |
245 | @deftypefun int getrlimit64 (int @var{resource}, struct rlimit64 *@var{rlp}) | |
c8ce789c AO |
246 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}} |
247 | @c Direct syscall on most systems, wrapper to getrlimit otherwise. | |
5ce8f203 UD |
248 | This function is similar to @code{getrlimit} but its second parameter is |
249 | a pointer to a variable of type @code{struct rlimit64}, which allows it | |
250 | to read values which wouldn't fit in the member of a @code{struct | |
251 | rlimit}. | |
252 | ||
253 | If the sources are compiled with @code{_FILE_OFFSET_BITS == 64} on a | |
254 | 32-bit machine, this function is available under the name | |
255 | @code{getrlimit} and so transparently replaces the old interface. | |
256 | @end deftypefun | |
257 | ||
258 | @comment sys/resource.h | |
259 | @comment BSD | |
260 | @deftypefun int setrlimit (int @var{resource}, const struct rlimit *@var{rlp}) | |
c8ce789c AO |
261 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}} |
262 | @c Direct syscall on most systems; lock-taking critical section on HURD. | |
5ce8f203 UD |
263 | Store the current and maximum limits for the resource @var{resource} |
264 | in @code{*@var{rlp}}. | |
265 | ||
266 | The return value is @code{0} on success and @code{-1} on failure. The | |
267 | following @code{errno} error condition is possible: | |
268 | ||
269 | @table @code | |
270 | @item EPERM | |
271 | @itemize @bullet | |
272 | @item | |
273 | The process tried to raise a current limit beyond the maximum limit. | |
274 | ||
275 | @item | |
276 | The process tried to raise a maximum limit, but is not superuser. | |
277 | @end itemize | |
278 | @end table | |
279 | ||
280 | When the sources are compiled with @code{_FILE_OFFSET_BITS == 64} on a | |
281 | 32-bit system this function is in fact @code{setrlimit64}. Thus, the | |
282 | LFS interface transparently replaces the old interface. | |
283 | @end deftypefun | |
284 | ||
285 | @comment sys/resource.h | |
286 | @comment Unix98 | |
287 | @deftypefun int setrlimit64 (int @var{resource}, const struct rlimit64 *@var{rlp}) | |
c8ce789c AO |
288 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}} |
289 | @c Wrapper for setrlimit or direct syscall. | |
5ce8f203 UD |
290 | This function is similar to @code{setrlimit} but its second parameter is |
291 | a pointer to a variable of type @code{struct rlimit64} which allows it | |
292 | to set values which wouldn't fit in the member of a @code{struct | |
293 | rlimit}. | |
294 | ||
295 | If the sources are compiled with @code{_FILE_OFFSET_BITS == 64} on a | |
296 | 32-bit machine this function is available under the name | |
297 | @code{setrlimit} and so transparently replaces the old interface. | |
298 | @end deftypefun | |
299 | ||
300 | @comment sys/resource.h | |
301 | @comment BSD | |
302 | @deftp {Data Type} {struct rlimit} | |
303 | This structure is used with @code{getrlimit} to receive limit values, | |
304 | and with @code{setrlimit} to specify limit values for a particular process | |
305 | and resource. It has two fields: | |
306 | ||
307 | @table @code | |
308 | @item rlim_t rlim_cur | |
309 | The current limit | |
310 | ||
311 | @item rlim_t rlim_max | |
312 | The maximum limit. | |
313 | @end table | |
314 | ||
315 | For @code{getrlimit}, the structure is an output; it receives the current | |
316 | values. For @code{setrlimit}, it specifies the new values. | |
317 | @end deftp | |
318 | ||
319 | For the LFS functions a similar type is defined in @file{sys/resource.h}. | |
320 | ||
321 | @comment sys/resource.h | |
322 | @comment Unix98 | |
323 | @deftp {Data Type} {struct rlimit64} | |
324 | This structure is analogous to the @code{rlimit} structure above, but | |
325 | its components have wider ranges. It has two fields: | |
326 | ||
327 | @table @code | |
328 | @item rlim64_t rlim_cur | |
329 | This is analogous to @code{rlimit.rlim_cur}, but with a different type. | |
330 | ||
331 | @item rlim64_t rlim_max | |
332 | This is analogous to @code{rlimit.rlim_max}, but with a different type. | |
333 | @end table | |
334 | ||
335 | @end deftp | |
336 | ||
337 | Here is a list of resources for which you can specify a limit. Memory | |
338 | and file sizes are measured in bytes. | |
339 | ||
340 | @table @code | |
341 | @comment sys/resource.h | |
342 | @comment BSD | |
343 | @item RLIMIT_CPU | |
344 | @vindex RLIMIT_CPU | |
345 | The maximum amount of CPU time the process can use. If it runs for | |
346 | longer than this, it gets a signal: @code{SIGXCPU}. The value is | |
347 | measured in seconds. @xref{Operation Error Signals}. | |
348 | ||
349 | @comment sys/resource.h | |
350 | @comment BSD | |
351 | @item RLIMIT_FSIZE | |
352 | @vindex RLIMIT_FSIZE | |
353 | The maximum size of file the process can create. Trying to write a | |
354 | larger file causes a signal: @code{SIGXFSZ}. @xref{Operation Error | |
355 | Signals}. | |
356 | ||
357 | @comment sys/resource.h | |
358 | @comment BSD | |
359 | @item RLIMIT_DATA | |
360 | @vindex RLIMIT_DATA | |
361 | The maximum size of data memory for the process. If the process tries | |
362 | to allocate data memory beyond this amount, the allocation function | |
363 | fails. | |
364 | ||
365 | @comment sys/resource.h | |
366 | @comment BSD | |
367 | @item RLIMIT_STACK | |
368 | @vindex RLIMIT_STACK | |
369 | The maximum stack size for the process. If the process tries to extend | |
370 | its stack past this size, it gets a @code{SIGSEGV} signal. | |
371 | @xref{Program Error Signals}. | |
372 | ||
373 | @comment sys/resource.h | |
374 | @comment BSD | |
375 | @item RLIMIT_CORE | |
376 | @vindex RLIMIT_CORE | |
377 | The maximum size core file that this process can create. If the process | |
378 | terminates and would dump a core file larger than this, then no core | |
379 | file is created. So setting this limit to zero prevents core files from | |
380 | ever being created. | |
381 | ||
382 | @comment sys/resource.h | |
383 | @comment BSD | |
384 | @item RLIMIT_RSS | |
385 | @vindex RLIMIT_RSS | |
386 | The maximum amount of physical memory that this process should get. | |
387 | This parameter is a guide for the system's scheduler and memory | |
388 | allocator; the system may give the process more memory when there is a | |
389 | surplus. | |
390 | ||
391 | @comment sys/resource.h | |
392 | @comment BSD | |
393 | @item RLIMIT_MEMLOCK | |
394 | The maximum amount of memory that can be locked into physical memory (so | |
395 | it will never be paged out). | |
396 | ||
397 | @comment sys/resource.h | |
398 | @comment BSD | |
399 | @item RLIMIT_NPROC | |
400 | The maximum number of processes that can be created with the same user ID. | |
401 | If you have reached the limit for your user ID, @code{fork} will fail | |
402 | with @code{EAGAIN}. @xref{Creating a Process}. | |
403 | ||
404 | @comment sys/resource.h | |
405 | @comment BSD | |
406 | @item RLIMIT_NOFILE | |
407 | @vindex RLIMIT_NOFILE | |
408 | @itemx RLIMIT_OFILE | |
409 | @vindex RLIMIT_OFILE | |
410 | The maximum number of files that the process can open. If it tries to | |
411 | open more files than this, its open attempt fails with @code{errno} | |
412 | @code{EMFILE}. @xref{Error Codes}. Not all systems support this limit; | |
413 | GNU does, and 4.4 BSD does. | |
414 | ||
415 | @comment sys/resource.h | |
416 | @comment Unix98 | |
417 | @item RLIMIT_AS | |
418 | @vindex RLIMIT_AS | |
419 | The maximum size of total memory that this process should get. If the | |
420 | process tries to allocate more memory beyond this amount with, for | |
421 | example, @code{brk}, @code{malloc}, @code{mmap} or @code{sbrk}, the | |
422 | allocation function fails. | |
423 | ||
424 | @comment sys/resource.h | |
425 | @comment BSD | |
426 | @item RLIM_NLIMITS | |
427 | @vindex RLIM_NLIMITS | |
428 | The number of different resource limits. Any valid @var{resource} | |
429 | operand must be less than @code{RLIM_NLIMITS}. | |
430 | @end table | |
431 | ||
432 | @comment sys/resource.h | |
433 | @comment BSD | |
8ded91fb | 434 | @deftypevr Constant rlim_t RLIM_INFINITY |
5ce8f203 UD |
435 | This constant stands for a value of ``infinity'' when supplied as |
436 | the limit value in @code{setrlimit}. | |
437 | @end deftypevr | |
438 | ||
439 | ||
440 | The following are historical functions to do some of what the functions | |
441 | above do. The functions above are better choices. | |
442 | ||
443 | @code{ulimit} and the command symbols are declared in @file{ulimit.h}. | |
444 | @pindex ulimit.h | |
5ce8f203 | 445 | |
b642f101 UD |
446 | @comment ulimit.h |
447 | @comment BSD | |
8ded91fb | 448 | @deftypefun {long int} ulimit (int @var{cmd}, @dots{}) |
c8ce789c AO |
449 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}} |
450 | @c Wrapper for getrlimit, setrlimit or | |
451 | @c sysconf(_SC_OPEN_MAX)->getdtablesize->getrlimit. | |
5ce8f203 UD |
452 | |
453 | @code{ulimit} gets the current limit or sets the current and maximum | |
454 | limit for a particular resource for the calling process according to the | |
455 | command @var{cmd}.a | |
456 | ||
457 | If you are getting a limit, the command argument is the only argument. | |
458 | If you are setting a limit, there is a second argument: | |
459 | @code{long int} @var{limit} which is the value to which you are setting | |
460 | the limit. | |
461 | ||
462 | The @var{cmd} values and the operations they specify are: | |
463 | @table @code | |
464 | ||
465 | @item GETFSIZE | |
466 | Get the current limit on the size of a file, in units of 512 bytes. | |
467 | ||
468 | @item SETFSIZE | |
469 | Set the current and maximum limit on the size of a file to @var{limit} * | |
470 | 512 bytes. | |
471 | ||
472 | @end table | |
473 | ||
474 | There are also some other @var{cmd} values that may do things on some | |
475 | systems, but they are not supported. | |
476 | ||
477 | Only the superuser may increase a maximum limit. | |
478 | ||
479 | When you successfully get a limit, the return value of @code{ulimit} is | |
480 | that limit, which is never negative. When you successfully set a limit, | |
481 | the return value is zero. When the function fails, the return value is | |
482 | @code{-1} and @code{errno} is set according to the reason: | |
483 | ||
484 | @table @code | |
485 | @item EPERM | |
486 | A process tried to increase a maximum limit, but is not superuser. | |
487 | @end table | |
488 | ||
489 | ||
490 | @end deftypefun | |
491 | ||
492 | @code{vlimit} and its resource symbols are declared in @file{sys/vlimit.h}. | |
5ce8f203 | 493 | @pindex sys/vlimit.h |
5ce8f203 | 494 | |
b642f101 UD |
495 | @comment sys/vlimit.h |
496 | @comment BSD | |
5ce8f203 | 497 | @deftypefun int vlimit (int @var{resource}, int @var{limit}) |
c8ce789c AO |
498 | @safety{@prelim{}@mtunsafe{@mtasurace{:setrlimit}}@asunsafe{}@acsafe{}} |
499 | @c It calls getrlimit and modifies the rlim_cur field before calling | |
500 | @c setrlimit. There's a window for a concurrent call to setrlimit that | |
501 | @c modifies e.g. rlim_max, which will be lost if running as super-user. | |
5ce8f203 UD |
502 | |
503 | @code{vlimit} sets the current limit for a resource for a process. | |
504 | ||
505 | @var{resource} identifies the resource: | |
506 | ||
507 | @table @code | |
508 | @item LIM_CPU | |
509 | Maximum CPU time. Same as @code{RLIMIT_CPU} for @code{setrlimit}. | |
510 | @item LIM_FSIZE | |
511 | Maximum file size. Same as @code{RLIMIT_FSIZE} for @code{setrlimit}. | |
512 | @item LIM_DATA | |
513 | Maximum data memory. Same as @code{RLIMIT_DATA} for @code{setrlimit}. | |
514 | @item LIM_STACK | |
515 | Maximum stack size. Same as @code{RLIMIT_STACK} for @code{setrlimit}. | |
516 | @item LIM_CORE | |
517 | Maximum core file size. Same as @code{RLIMIT_COR} for @code{setrlimit}. | |
518 | @item LIM_MAXRSS | |
519 | Maximum physical memory. Same as @code{RLIMIT_RSS} for @code{setrlimit}. | |
520 | @end table | |
521 | ||
522 | The return value is zero for success, and @code{-1} with @code{errno} set | |
523 | accordingly for failure: | |
524 | ||
525 | @table @code | |
526 | @item EPERM | |
527 | The process tried to set its current limit beyond its maximum limit. | |
528 | @end table | |
529 | ||
530 | @end deftypefun | |
531 | ||
532 | @node Priority | |
639c6286 | 533 | @section Process CPU Priority And Scheduling |
5ce8f203 | 534 | @cindex process priority |
639c6286 | 535 | @cindex cpu priority |
5ce8f203 UD |
536 | @cindex priority of a process |
537 | ||
639c6286 UD |
538 | When multiple processes simultaneously require CPU time, the system's |
539 | scheduling policy and process CPU priorities determine which processes | |
540 | get it. This section describes how that determination is made and | |
1f77f049 | 541 | @glibcadj{} functions to control it. |
639c6286 UD |
542 | |
543 | It is common to refer to CPU scheduling simply as scheduling and a | |
544 | process' CPU priority simply as the process' priority, with the CPU | |
545 | resource being implied. Bear in mind, though, that CPU time is not the | |
546 | only resource a process uses or that processes contend for. In some | |
547 | cases, it is not even particularly important. Giving a process a high | |
548 | ``priority'' may have very little effect on how fast a process runs with | |
549 | respect to other processes. The priorities discussed in this section | |
550 | apply only to CPU time. | |
551 | ||
552 | CPU scheduling is a complex issue and different systems do it in wildly | |
553 | different ways. New ideas continually develop and find their way into | |
554 | the intricacies of the various systems' scheduling algorithms. This | |
87b56f36 | 555 | section discusses the general concepts, some specifics of systems |
1f77f049 | 556 | that commonly use @theglibc{}, and some standards. |
639c6286 UD |
557 | |
558 | For simplicity, we talk about CPU contention as if there is only one CPU | |
559 | in the system. But all the same principles apply when a processor has | |
560 | multiple CPUs, and knowing that the number of processes that can run at | |
561 | any one time is equal to the number of CPUs, you can easily extrapolate | |
562 | the information. | |
563 | ||
564 | The functions described in this section are all defined by the POSIX.1 | |
95fdc6a0 | 565 | and POSIX.1b standards (the @code{sched@dots{}} functions are POSIX.1b). |
639c6286 UD |
566 | However, POSIX does not define any semantics for the values that these |
567 | functions get and set. In this chapter, the semantics are based on the | |
568 | Linux kernel's implementation of the POSIX standard. As you will see, | |
569 | the Linux implementation is quite the inverse of what the authors of the | |
570 | POSIX syntax had in mind. | |
571 | ||
572 | @menu | |
573 | * Absolute Priority:: The first tier of priority. Posix | |
574 | * Realtime Scheduling:: Scheduling among the process nobility | |
575 | * Basic Scheduling Functions:: Get/set scheduling policy, priority | |
576 | * Traditional Scheduling:: Scheduling among the vulgar masses | |
d9997a45 | 577 | * CPU Affinity:: Limiting execution to certain CPUs |
639c6286 UD |
578 | @end menu |
579 | ||
580 | ||
581 | ||
582 | @node Absolute Priority | |
583 | @subsection Absolute Priority | |
584 | @cindex absolute priority | |
585 | @cindex priority, absolute | |
586 | ||
587 | Every process has an absolute priority, and it is represented by a number. | |
588 | The higher the number, the higher the absolute priority. | |
589 | ||
590 | @cindex realtime CPU scheduling | |
591 | On systems of the past, and most systems today, all processes have | |
592 | absolute priority 0 and this section is irrelevant. In that case, | |
593 | @xref{Traditional Scheduling}. Absolute priorities were invented to | |
0bc93a2f | 594 | accommodate realtime systems, in which it is vital that certain processes |
639c6286 UD |
595 | be able to respond to external events happening in real time, which |
596 | means they cannot wait around while some other process that @emph{wants | |
597 | to}, but doesn't @emph{need to} run occupies the CPU. | |
598 | ||
599 | @cindex ready to run | |
600 | @cindex preemptive scheduling | |
601 | When two processes are in contention to use the CPU at any instant, the | |
602 | one with the higher absolute priority always gets it. This is true even if the | |
11bf311e | 603 | process with the lower priority is already using the CPU (i.e., the |
639c6286 UD |
604 | scheduling is preemptive). Of course, we're only talking about |
605 | processes that are running or ``ready to run,'' which means they are | |
606 | ready to execute instructions right now. When a process blocks to wait | |
607 | for something like I/O, its absolute priority is irrelevant. | |
608 | ||
609 | @cindex runnable process | |
48b22986 | 610 | @strong{NB:} The term ``runnable'' is a synonym for ``ready to run.'' |
639c6286 UD |
611 | |
612 | When two processes are running or ready to run and both have the same | |
613 | absolute priority, it's more interesting. In that case, who gets the | |
0bc93a2f | 614 | CPU is determined by the scheduling policy. If the processes have |
639c6286 UD |
615 | absolute priority 0, the traditional scheduling policy described in |
616 | @ref{Traditional Scheduling} applies. Otherwise, the policies described | |
617 | in @ref{Realtime Scheduling} apply. | |
618 | ||
619 | You normally give an absolute priority above 0 only to a process that | |
620 | can be trusted not to hog the CPU. Such processes are designed to block | |
621 | (or terminate) after relatively short CPU runs. | |
622 | ||
623 | A process begins life with the same absolute priority as its parent | |
624 | process. Functions described in @ref{Basic Scheduling Functions} can | |
625 | change it. | |
626 | ||
627 | Only a privileged process can change a process' absolute priority to | |
628 | something other than @code{0}. Only a privileged process or the | |
629 | target process' owner can change its absolute priority at all. | |
630 | ||
631 | POSIX requires absolute priority values used with the realtime | |
632 | scheduling policies to be consecutive with a range of at least 32. On | |
633 | Linux, they are 1 through 99. The functions | |
634 | @code{sched_get_priority_max} and @code{sched_set_priority_min} portably | |
635 | tell you what the range is on a particular system. | |
636 | ||
637 | ||
638 | @subsubsection Using Absolute Priority | |
639 | ||
640 | One thing you must keep in mind when designing real time applications is | |
641 | that having higher absolute priority than any other process doesn't | |
642 | guarantee the process can run continuously. Two things that can wreck a | |
87b56f36 | 643 | good CPU run are interrupts and page faults. |
639c6286 UD |
644 | |
645 | Interrupt handlers live in that limbo between processes. The CPU is | |
646 | executing instructions, but they aren't part of any process. An | |
647 | interrupt will stop even the highest priority process. So you must | |
648 | allow for slight delays and make sure that no device in the system has | |
649 | an interrupt handler that could cause too long a delay between | |
650 | instructions for your process. | |
651 | ||
652 | Similarly, a page fault causes what looks like a straightforward | |
653 | sequence of instructions to take a long time. The fact that other | |
654 | processes get to run while the page faults in is of no consequence, | |
655 | because as soon as the I/O is complete, the high priority process will | |
656 | kick them out and run again, but the wait for the I/O itself could be a | |
657 | problem. To neutralize this threat, use @code{mlock} or | |
658 | @code{mlockall}. | |
659 | ||
660 | There are a few ramifications of the absoluteness of this priority on a | |
661 | single-CPU system that you need to keep in mind when you choose to set a | |
662 | priority and also when you're working on a program that runs with high | |
663 | absolute priority. Consider a process that has higher absolute priority | |
664 | than any other process in the system and due to a bug in its program, it | |
665 | gets into an infinite loop. It will never cede the CPU. You can't run | |
666 | a command to kill it because your command would need to get the CPU in | |
667 | order to run. The errant program is in complete control. It controls | |
668 | the vertical, it controls the horizontal. | |
669 | ||
670 | There are two ways to avoid this: 1) keep a shell running somewhere with | |
671 | a higher absolute priority. 2) keep a controlling terminal attached to | |
672 | the high priority process group. All the priority in the world won't | |
673 | stop an interrupt handler from running and delivering a signal to the | |
674 | process if you hit Control-C. | |
675 | ||
95fdc6a0 | 676 | Some systems use absolute priority as a means of allocating a fixed |
0bc93a2f | 677 | percentage of CPU time to a process. To do this, a super high priority |
639c6286 UD |
678 | privileged process constantly monitors the process' CPU usage and raises |
679 | its absolute priority when the process isn't getting its entitled share | |
680 | and lowers it when the process is exceeding it. | |
681 | ||
48b22986 | 682 | @strong{NB:} The absolute priority is sometimes called the ``static |
639c6286 UD |
683 | priority.'' We don't use that term in this manual because it misses the |
684 | most important feature of the absolute priority: its absoluteness. | |
685 | ||
686 | ||
687 | @node Realtime Scheduling | |
688 | @subsection Realtime Scheduling | |
b642f101 | 689 | @cindex realtime scheduling |
639c6286 UD |
690 | |
691 | Whenever two processes with the same absolute priority are ready to run, | |
692 | the kernel has a decision to make, because only one can run at a time. | |
693 | If the processes have absolute priority 0, the kernel makes this decision | |
694 | as described in @ref{Traditional Scheduling}. Otherwise, the decision | |
695 | is as described in this section. | |
696 | ||
697 | If two processes are ready to run but have different absolute priorities, | |
698 | the decision is much simpler, and is described in @ref{Absolute | |
699 | Priority}. | |
700 | ||
87b56f36 | 701 | Each process has a scheduling policy. For processes with absolute |
639c6286 UD |
702 | priority other than zero, there are two available: |
703 | ||
704 | @enumerate | |
705 | @item | |
706 | First Come First Served | |
707 | @item | |
708 | Round Robin | |
709 | @end enumerate | |
710 | ||
711 | The most sensible case is where all the processes with a certain | |
712 | absolute priority have the same scheduling policy. We'll discuss that | |
713 | first. | |
714 | ||
715 | In Round Robin, processes share the CPU, each one running for a small | |
716 | quantum of time (``time slice'') and then yielding to another in a | |
717 | circular fashion. Of course, only processes that are ready to run and | |
718 | have the same absolute priority are in this circle. | |
719 | ||
720 | In First Come First Served, the process that has been waiting the | |
721 | longest to run gets the CPU, and it keeps it until it voluntarily | |
722 | relinquishes the CPU, runs out of things to do (blocks), or gets | |
723 | preempted by a higher priority process. | |
724 | ||
725 | First Come First Served, along with maximal absolute priority and | |
726 | careful control of interrupts and page faults, is the one to use when a | |
727 | process absolutely, positively has to run at full CPU speed or not at | |
728 | all. | |
729 | ||
730 | Judicious use of @code{sched_yield} function invocations by processes | |
731 | with First Come First Served scheduling policy forms a good compromise | |
732 | between Round Robin and First Come First Served. | |
733 | ||
734 | To understand how scheduling works when processes of different scheduling | |
735 | policies occupy the same absolute priority, you have to know the nitty | |
736 | gritty details of how processes enter and exit the ready to run list: | |
737 | ||
738 | In both cases, the ready to run list is organized as a true queue, where | |
739 | a process gets pushed onto the tail when it becomes ready to run and is | |
740 | popped off the head when the scheduler decides to run it. Note that | |
741 | ready to run and running are two mutually exclusive states. When the | |
742 | scheduler runs a process, that process is no longer ready to run and no | |
743 | longer in the ready to run list. When the process stops running, it | |
744 | may go back to being ready to run again. | |
745 | ||
746 | The only difference between a process that is assigned the Round Robin | |
747 | scheduling policy and a process that is assigned First Come First Serve | |
748 | is that in the former case, the process is automatically booted off the | |
749 | CPU after a certain amount of time. When that happens, the process goes | |
750 | back to being ready to run, which means it enters the queue at the tail. | |
751 | The time quantum we're talking about is small. Really small. This is | |
752 | not your father's timesharing. For example, with the Linux kernel, the | |
753 | round robin time slice is a thousand times shorter than its typical | |
754 | time slice for traditional scheduling. | |
755 | ||
756 | A process begins life with the same scheduling policy as its parent process. | |
757 | Functions described in @ref{Basic Scheduling Functions} can change it. | |
758 | ||
759 | Only a privileged process can set the scheduling policy of a process | |
760 | that has absolute priority higher than 0. | |
761 | ||
762 | @node Basic Scheduling Functions | |
763 | @subsection Basic Scheduling Functions | |
764 | ||
1f77f049 | 765 | This section describes functions in @theglibc{} for setting the |
639c6286 UD |
766 | absolute priority and scheduling policy of a process. |
767 | ||
768 | @strong{Portability Note:} On systems that have the functions in this | |
769 | section, the macro _POSIX_PRIORITY_SCHEDULING is defined in | |
770 | @file{<unistd.h>}. | |
771 | ||
772 | For the case that the scheduling policy is traditional scheduling, more | |
773 | functions to fine tune the scheduling are in @ref{Traditional Scheduling}. | |
774 | ||
775 | Don't try to make too much out of the naming and structure of these | |
776 | functions. They don't match the concepts described in this manual | |
777 | because the functions are as defined by POSIX.1b, but the implementation | |
1f77f049 | 778 | on systems that use @theglibc{} is the inverse of what the POSIX |
639c6286 UD |
779 | structure contemplates. The POSIX scheme assumes that the primary |
780 | scheduling parameter is the scheduling policy and that the priority | |
781 | value, if any, is a parameter of the scheduling policy. In the | |
782 | implementation, though, the priority value is king and the scheduling | |
783 | policy, if anything, only fine tunes the effect of that priority. | |
784 | ||
785 | The symbols in this section are declared by including file @file{sched.h}. | |
786 | ||
787 | @comment sched.h | |
788 | @comment POSIX | |
789 | @deftp {Data Type} {struct sched_param} | |
790 | This structure describes an absolute priority. | |
791 | @table @code | |
792 | @item int sched_priority | |
793 | absolute priority value | |
794 | @end table | |
795 | @end deftp | |
796 | ||
797 | @comment sched.h | |
798 | @comment POSIX | |
799 | @deftypefun int sched_setscheduler (pid_t @var{pid}, int @var{policy}, const struct sched_param *@var{param}) | |
c8ce789c AO |
800 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}} |
801 | @c Direct syscall, Linux only. | |
639c6286 UD |
802 | |
803 | This function sets both the absolute priority and the scheduling policy | |
804 | for a process. | |
805 | ||
806 | It assigns the absolute priority value given by @var{param} and the | |
807 | scheduling policy @var{policy} to the process with Process ID @var{pid}, | |
808 | or the calling process if @var{pid} is zero. If @var{policy} is | |
0bc93a2f | 809 | negative, @code{sched_setscheduler} keeps the existing scheduling policy. |
639c6286 UD |
810 | |
811 | The following macros represent the valid values for @var{policy}: | |
812 | ||
813 | @table @code | |
814 | @item SCHED_OTHER | |
815 | Traditional Scheduling | |
816 | @item SCHED_FIFO | |
87b56f36 | 817 | First In First Out |
639c6286 UD |
818 | @item SCHED_RR |
819 | Round Robin | |
820 | @end table | |
821 | ||
822 | @c The Linux kernel code (in sched.c) actually reschedules the process, | |
823 | @c but it puts it at the head of the run queue, so I'm not sure just what | |
824 | @c the effect is, but it must be subtle. | |
825 | ||
826 | On success, the return value is @code{0}. Otherwise, it is @code{-1} | |
827 | and @code{ERRNO} is set accordingly. The @code{errno} values specific | |
828 | to this function are: | |
829 | ||
830 | @table @code | |
831 | @item EPERM | |
832 | @itemize @bullet | |
833 | @item | |
834 | The calling process does not have @code{CAP_SYS_NICE} permission and | |
835 | @var{policy} is not @code{SCHED_OTHER} (or it's negative and the | |
836 | existing policy is not @code{SCHED_OTHER}. | |
837 | ||
838 | @item | |
839 | The calling process does not have @code{CAP_SYS_NICE} permission and its | |
11bf311e | 840 | owner is not the target process' owner. I.e., the effective uid of the |
639c6286 UD |
841 | calling process is neither the effective nor the real uid of process |
842 | @var{pid}. | |
843 | @c We need a cross reference to the capabilities section, when written. | |
844 | @end itemize | |
845 | ||
846 | @item ESRCH | |
847 | There is no process with pid @var{pid} and @var{pid} is not zero. | |
848 | ||
849 | @item EINVAL | |
850 | @itemize @bullet | |
851 | @item | |
852 | @var{policy} does not identify an existing scheduling policy. | |
853 | ||
854 | @item | |
855 | The absolute priority value identified by *@var{param} is outside the | |
856 | valid range for the scheduling policy @var{policy} (or the existing | |
857 | scheduling policy if @var{policy} is negative) or @var{param} is | |
858 | null. @code{sched_get_priority_max} and @code{sched_get_priority_min} | |
859 | tell you what the valid range is. | |
860 | ||
861 | @item | |
862 | @var{pid} is negative. | |
863 | @end itemize | |
864 | @end table | |
865 | ||
866 | @end deftypefun | |
867 | ||
868 | ||
869 | @comment sched.h | |
870 | @comment POSIX | |
871 | @deftypefun int sched_getscheduler (pid_t @var{pid}) | |
c8ce789c AO |
872 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}} |
873 | @c Direct syscall, Linux only. | |
639c6286 UD |
874 | |
875 | This function returns the scheduling policy assigned to the process with | |
876 | Process ID (pid) @var{pid}, or the calling process if @var{pid} is zero. | |
877 | ||
878 | The return value is the scheduling policy. See | |
879 | @code{sched_setscheduler} for the possible values. | |
880 | ||
881 | If the function fails, the return value is instead @code{-1} and | |
882 | @code{errno} is set accordingly. | |
883 | ||
884 | The @code{errno} values specific to this function are: | |
885 | ||
886 | @table @code | |
887 | ||
888 | @item ESRCH | |
889 | There is no process with pid @var{pid} and it is not zero. | |
890 | ||
891 | @item EINVAL | |
892 | @var{pid} is negative. | |
893 | ||
894 | @end table | |
895 | ||
896 | Note that this function is not an exact mate to @code{sched_setscheduler} | |
897 | because while that function sets the scheduling policy and the absolute | |
898 | priority, this function gets only the scheduling policy. To get the | |
899 | absolute priority, use @code{sched_getparam}. | |
900 | ||
901 | @end deftypefun | |
902 | ||
903 | ||
904 | @comment sched.h | |
905 | @comment POSIX | |
906 | @deftypefun int sched_setparam (pid_t @var{pid}, const struct sched_param *@var{param}) | |
c8ce789c AO |
907 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}} |
908 | @c Direct syscall, Linux only. | |
639c6286 UD |
909 | |
910 | This function sets a process' absolute priority. | |
911 | ||
912 | It is functionally identical to @code{sched_setscheduler} with | |
913 | @var{policy} = @code{-1}. | |
914 | ||
915 | @c in fact, that's how it's implemented in Linux. | |
916 | ||
917 | @end deftypefun | |
918 | ||
919 | @comment sched.h | |
920 | @comment POSIX | |
8ded91fb | 921 | @deftypefun int sched_getparam (pid_t @var{pid}, struct sched_param *@var{param}) |
c8ce789c AO |
922 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}} |
923 | @c Direct syscall, Linux only. | |
639c6286 UD |
924 | |
925 | This function returns a process' absolute priority. | |
926 | ||
927 | @var{pid} is the Process ID (pid) of the process whose absolute priority | |
928 | you want to know. | |
929 | ||
930 | @var{param} is a pointer to a structure in which the function stores the | |
931 | absolute priority of the process. | |
932 | ||
933 | On success, the return value is @code{0}. Otherwise, it is @code{-1} | |
934 | and @code{ERRNO} is set accordingly. The @code{errno} values specific | |
935 | to this function are: | |
936 | ||
937 | @table @code | |
938 | ||
939 | @item ESRCH | |
940 | There is no process with pid @var{pid} and it is not zero. | |
941 | ||
942 | @item EINVAL | |
943 | @var{pid} is negative. | |
944 | ||
945 | @end table | |
946 | ||
947 | @end deftypefun | |
948 | ||
949 | ||
950 | @comment sched.h | |
951 | @comment POSIX | |
8ded91fb | 952 | @deftypefun int sched_get_priority_min (int @var{policy}) |
c8ce789c AO |
953 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}} |
954 | @c Direct syscall, Linux only. | |
639c6286 UD |
955 | |
956 | This function returns the lowest absolute priority value that is | |
957 | allowable for a process with scheduling policy @var{policy}. | |
958 | ||
959 | On Linux, it is 0 for SCHED_OTHER and 1 for everything else. | |
960 | ||
961 | On success, the return value is @code{0}. Otherwise, it is @code{-1} | |
962 | and @code{ERRNO} is set accordingly. The @code{errno} values specific | |
963 | to this function are: | |
964 | ||
965 | @table @code | |
966 | @item EINVAL | |
967 | @var{policy} does not identify an existing scheduling policy. | |
968 | @end table | |
969 | ||
970 | @end deftypefun | |
971 | ||
972 | @comment sched.h | |
973 | @comment POSIX | |
8ded91fb | 974 | @deftypefun int sched_get_priority_max (int @var{policy}) |
c8ce789c AO |
975 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}} |
976 | @c Direct syscall, Linux only. | |
639c6286 UD |
977 | |
978 | This function returns the highest absolute priority value that is | |
979 | allowable for a process that with scheduling policy @var{policy}. | |
980 | ||
981 | On Linux, it is 0 for SCHED_OTHER and 99 for everything else. | |
982 | ||
983 | On success, the return value is @code{0}. Otherwise, it is @code{-1} | |
984 | and @code{ERRNO} is set accordingly. The @code{errno} values specific | |
985 | to this function are: | |
986 | ||
987 | @table @code | |
988 | @item EINVAL | |
989 | @var{policy} does not identify an existing scheduling policy. | |
990 | @end table | |
991 | ||
992 | @end deftypefun | |
993 | ||
994 | @comment sched.h | |
995 | @comment POSIX | |
996 | @deftypefun int sched_rr_get_interval (pid_t @var{pid}, struct timespec *@var{interval}) | |
c8ce789c AO |
997 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}} |
998 | @c Direct syscall, Linux only. | |
639c6286 | 999 | |
87b56f36 | 1000 | This function returns the length of the quantum (time slice) used with |
639c6286 UD |
1001 | the Round Robin scheduling policy, if it is used, for the process with |
1002 | Process ID @var{pid}. | |
1003 | ||
87b56f36 | 1004 | It returns the length of time as @var{interval}. |
639c6286 UD |
1005 | @c We need a cross-reference to where timespec is explained. But that |
1006 | @c section doesn't exist yet, and the time chapter needs to be slightly | |
1007 | @c reorganized so there is a place to put it (which will be right next | |
1008 | @c to timeval, which is presently misplaced). 2000.05.07. | |
1009 | ||
1010 | With a Linux kernel, the round robin time slice is always 150 | |
1011 | microseconds, and @var{pid} need not even be a real pid. | |
1012 | ||
1013 | The return value is @code{0} on success and in the pathological case | |
1014 | that it fails, the return value is @code{-1} and @code{errno} is set | |
1015 | accordingly. There is nothing specific that can go wrong with this | |
1016 | function, so there are no specific @code{errno} values. | |
1017 | ||
1018 | @end deftypefun | |
1019 | ||
1020 | @comment sched.h | |
1021 | @comment POSIX | |
3c44837c | 1022 | @deftypefun int sched_yield (void) |
c8ce789c AO |
1023 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}} |
1024 | @c Direct syscall on Linux; alias to swtch on HURD. | |
639c6286 UD |
1025 | |
1026 | This function voluntarily gives up the process' claim on the CPU. | |
1027 | ||
1028 | Technically, @code{sched_yield} causes the calling process to be made | |
1029 | immediately ready to run (as opposed to running, which is what it was | |
1030 | before). This means that if it has absolute priority higher than 0, it | |
1031 | gets pushed onto the tail of the queue of processes that share its | |
1032 | absolute priority and are ready to run, and it will run again when its | |
1033 | turn next arrives. If its absolute priority is 0, it is more | |
1034 | complicated, but still has the effect of yielding the CPU to other | |
1035 | processes. | |
1036 | ||
1037 | If there are no other processes that share the calling process' absolute | |
1038 | priority, this function doesn't have any effect. | |
1039 | ||
1040 | To the extent that the containing program is oblivious to what other | |
1041 | processes in the system are doing and how fast it executes, this | |
1042 | function appears as a no-op. | |
1043 | ||
1044 | The return value is @code{0} on success and in the pathological case | |
1045 | that it fails, the return value is @code{-1} and @code{errno} is set | |
1046 | accordingly. There is nothing specific that can go wrong with this | |
1047 | function, so there are no specific @code{errno} values. | |
1048 | ||
1049 | @end deftypefun | |
1050 | ||
1051 | @node Traditional Scheduling | |
1052 | @subsection Traditional Scheduling | |
1053 | @cindex scheduling, traditional | |
1054 | ||
1055 | This section is about the scheduling among processes whose absolute | |
1056 | priority is 0. When the system hands out the scraps of CPU time that | |
0bc93a2f | 1057 | are left over after the processes with higher absolute priority have |
639c6286 UD |
1058 | taken all they want, the scheduling described herein determines who |
1059 | among the great unwashed processes gets them. | |
1060 | ||
1061 | @menu | |
1062 | * Traditional Scheduling Intro:: | |
1063 | * Traditional Scheduling Functions:: | |
1064 | @end menu | |
1065 | ||
1066 | @node Traditional Scheduling Intro | |
1067 | @subsubsection Introduction To Traditional Scheduling | |
1068 | ||
1069 | Long before there was absolute priority (See @ref{Absolute Priority}), | |
1070 | Unix systems were scheduling the CPU using this system. When Posix came | |
0bc93a2f | 1071 | in like the Romans and imposed absolute priorities to accommodate the |
639c6286 UD |
1072 | needs of realtime processing, it left the indigenous Absolute Priority |
1073 | Zero processes to govern themselves by their own familiar scheduling | |
1074 | policy. | |
1075 | ||
1076 | Indeed, absolute priorities higher than zero are not available on many | |
1077 | systems today and are not typically used when they are, being intended | |
1078 | mainly for computers that do realtime processing. So this section | |
1079 | describes the only scheduling many programmers need to be concerned | |
1080 | about. | |
1081 | ||
1082 | But just to be clear about the scope of this scheduling: Any time a | |
9dcc8f11 | 1083 | process with an absolute priority of 0 and a process with an absolute |
639c6286 UD |
1084 | priority higher than 0 are ready to run at the same time, the one with |
1085 | absolute priority 0 does not run. If it's already running when the | |
1086 | higher priority ready-to-run process comes into existence, it stops | |
1087 | immediately. | |
1088 | ||
1089 | In addition to its absolute priority of zero, every process has another | |
1090 | priority, which we will refer to as "dynamic priority" because it changes | |
87b56f36 | 1091 | over time. The dynamic priority is meaningless for processes with |
639c6286 UD |
1092 | an absolute priority higher than zero. |
1093 | ||
1094 | The dynamic priority sometimes determines who gets the next turn on the | |
1095 | CPU. Sometimes it determines how long turns last. Sometimes it | |
1096 | determines whether a process can kick another off the CPU. | |
1097 | ||
1098 | In Linux, the value is a combination of these things, but mostly it is | |
1099 | just determines the length of the time slice. The higher a process' | |
1100 | dynamic priority, the longer a shot it gets on the CPU when it gets one. | |
1101 | If it doesn't use up its time slice before giving up the CPU to do | |
1102 | something like wait for I/O, it is favored for getting the CPU back when | |
1103 | it's ready for it, to finish out its time slice. Other than that, | |
1104 | selection of processes for new time slices is basically round robin. | |
1105 | But the scheduler does throw a bone to the low priority processes: A | |
1106 | process' dynamic priority rises every time it is snubbed in the | |
1107 | scheduling process. In Linux, even the fat kid gets to play. | |
1108 | ||
1109 | The fluctuation of a process' dynamic priority is regulated by another | |
1110 | value: The ``nice'' value. The nice value is an integer, usually in the | |
1111 | range -20 to 20, and represents an upper limit on a process' dynamic | |
1112 | priority. The higher the nice number, the lower that limit. | |
1113 | ||
1114 | On a typical Linux system, for example, a process with a nice value of | |
1115 | 20 can get only 10 milliseconds on the CPU at a time, whereas a process | |
1116 | with a nice value of -20 can achieve a high enough priority to get 400 | |
1117 | milliseconds. | |
1118 | ||
1119 | The idea of the nice value is deferential courtesy. In the beginning, | |
1120 | in the Unix garden of Eden, all processes shared equally in the bounty | |
1121 | of the computer system. But not all processes really need the same | |
1122 | share of CPU time, so the nice value gave a courteous process the | |
1123 | ability to refuse its equal share of CPU time that others might prosper. | |
1124 | Hence, the higher a process' nice value, the nicer the process is. | |
1125 | (Then a snake came along and offered some process a negative nice value | |
1126 | and the system became the crass resource allocation system we know | |
1127 | today). | |
1128 | ||
1129 | Dynamic priorities tend upward and downward with an objective of | |
1130 | smoothing out allocation of CPU time and giving quick response time to | |
1131 | infrequent requests. But they never exceed their nice limits, so on a | |
1132 | heavily loaded CPU, the nice value effectively determines how fast a | |
1133 | process runs. | |
1134 | ||
1135 | In keeping with the socialistic heritage of Unix process priority, a | |
1136 | process begins life with the same nice value as its parent process and | |
1137 | can raise it at will. A process can also raise the nice value of any | |
1138 | other process owned by the same user (or effective user). But only a | |
1139 | privileged process can lower its nice value. A privileged process can | |
1140 | also raise or lower another process' nice value. | |
1141 | ||
1f77f049 | 1142 | @glibcadj{} functions for getting and setting nice values are described in |
639c6286 UD |
1143 | @xref{Traditional Scheduling Functions}. |
1144 | ||
1145 | @node Traditional Scheduling Functions | |
1146 | @subsubsection Functions For Traditional Scheduling | |
1147 | ||
5ce8f203 | 1148 | @pindex sys/resource.h |
639c6286 UD |
1149 | This section describes how you can read and set the nice value of a |
1150 | process. All these symbols are declared in @file{sys/resource.h}. | |
1151 | ||
1152 | The function and macro names are defined by POSIX, and refer to | |
1153 | "priority," but the functions actually have to do with nice values, as | |
1154 | the terms are used both in the manual and POSIX. | |
1155 | ||
1156 | The range of valid nice values depends on the kernel, but typically it | |
1157 | runs from @code{-20} to @code{20}. A lower nice value corresponds to | |
1158 | higher priority for the process. These constants describe the range of | |
5ce8f203 UD |
1159 | priority values: |
1160 | ||
b642f101 | 1161 | @vtable @code |
5ce8f203 UD |
1162 | @comment sys/resource.h |
1163 | @comment BSD | |
1164 | @item PRIO_MIN | |
639c6286 | 1165 | The lowest valid nice value. |
5ce8f203 UD |
1166 | |
1167 | @comment sys/resource.h | |
1168 | @comment BSD | |
1169 | @item PRIO_MAX | |
639c6286 | 1170 | The highest valid nice value. |
b642f101 | 1171 | @end vtable |
5ce8f203 UD |
1172 | |
1173 | @comment sys/resource.h | |
639c6286 | 1174 | @comment BSD,POSIX |
5ce8f203 | 1175 | @deftypefun int getpriority (int @var{class}, int @var{id}) |
c8ce789c AO |
1176 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}} |
1177 | @c Direct syscall on UNIX. On HURD, calls _hurd_priority_which_map. | |
639c6286 | 1178 | Return the nice value of a set of processes; @var{class} and @var{id} |
5ce8f203 | 1179 | specify which ones (see below). If the processes specified do not all |
639c6286 | 1180 | have the same nice value, this returns the lowest value that any of them |
5ce8f203 UD |
1181 | has. |
1182 | ||
639c6286 UD |
1183 | On success, the return value is @code{0}. Otherwise, it is @code{-1} |
1184 | and @code{ERRNO} is set accordingly. The @code{errno} values specific | |
1185 | to this function are: | |
5ce8f203 UD |
1186 | |
1187 | @table @code | |
1188 | @item ESRCH | |
1189 | The combination of @var{class} and @var{id} does not match any existing | |
1190 | process. | |
1191 | ||
1192 | @item EINVAL | |
1193 | The value of @var{class} is not valid. | |
1194 | @end table | |
1195 | ||
639c6286 UD |
1196 | If the return value is @code{-1}, it could indicate failure, or it could |
1197 | be the nice value. The only way to make certain is to set @code{errno = | |
1198 | 0} before calling @code{getpriority}, then use @code{errno != 0} | |
1199 | afterward as the criterion for failure. | |
5ce8f203 UD |
1200 | @end deftypefun |
1201 | ||
1202 | @comment sys/resource.h | |
639c6286 UD |
1203 | @comment BSD,POSIX |
1204 | @deftypefun int setpriority (int @var{class}, int @var{id}, int @var{niceval}) | |
c8ce789c AO |
1205 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}} |
1206 | @c Direct syscall on UNIX. On HURD, calls _hurd_priority_which_map. | |
639c6286 | 1207 | Set the nice value of a set of processes to @var{niceval}; @var{class} |
5ce8f203 UD |
1208 | and @var{id} specify which ones (see below). |
1209 | ||
6a7a8b22 | 1210 | The return value is @code{0} on success, and @code{-1} on |
639c6286 UD |
1211 | failure. The following @code{errno} error condition are possible for |
1212 | this function: | |
5ce8f203 UD |
1213 | |
1214 | @table @code | |
1215 | @item ESRCH | |
1216 | The combination of @var{class} and @var{id} does not match any existing | |
1217 | process. | |
1218 | ||
1219 | @item EINVAL | |
1220 | The value of @var{class} is not valid. | |
1221 | ||
1222 | @item EPERM | |
639c6286 | 1223 | The call would set the nice value of a process which is owned by a different |
11bf311e | 1224 | user than the calling process (i.e., the target process' real or effective |
639c6286 UD |
1225 | uid does not match the calling process' effective uid) and the calling |
1226 | process does not have @code{CAP_SYS_NICE} permission. | |
5ce8f203 UD |
1227 | |
1228 | @item EACCES | |
639c6286 UD |
1229 | The call would lower the process' nice value and the process does not have |
1230 | @code{CAP_SYS_NICE} permission. | |
5ce8f203 | 1231 | @end table |
639c6286 | 1232 | |
5ce8f203 UD |
1233 | @end deftypefun |
1234 | ||
1235 | The arguments @var{class} and @var{id} together specify a set of | |
1236 | processes in which you are interested. These are the possible values of | |
1237 | @var{class}: | |
1238 | ||
b642f101 | 1239 | @vtable @code |
5ce8f203 UD |
1240 | @comment sys/resource.h |
1241 | @comment BSD | |
1242 | @item PRIO_PROCESS | |
639c6286 | 1243 | One particular process. The argument @var{id} is a process ID (pid). |
5ce8f203 UD |
1244 | |
1245 | @comment sys/resource.h | |
1246 | @comment BSD | |
1247 | @item PRIO_PGRP | |
639c6286 UD |
1248 | All the processes in a particular process group. The argument @var{id} is |
1249 | a process group ID (pgid). | |
5ce8f203 UD |
1250 | |
1251 | @comment sys/resource.h | |
1252 | @comment BSD | |
1253 | @item PRIO_USER | |
11bf311e | 1254 | All the processes owned by a particular user (i.e., whose real uid |
639c6286 | 1255 | indicates the user). The argument @var{id} is a user ID (uid). |
b642f101 | 1256 | @end vtable |
5ce8f203 | 1257 | |
639c6286 UD |
1258 | If the argument @var{id} is 0, it stands for the calling process, its |
1259 | process group, or its owner (real uid), according to @var{class}. | |
5ce8f203 | 1260 | |
b642f101 UD |
1261 | @comment unistd.h |
1262 | @comment BSD | |
5ce8f203 | 1263 | @deftypefun int nice (int @var{increment}) |
c8ce789c AO |
1264 | @safety{@prelim{}@mtunsafe{@mtasurace{:setpriority}}@asunsafe{}@acsafe{}} |
1265 | @c Calls getpriority before and after setpriority, using the result of | |
1266 | @c the first call to compute the argument for setpriority. This creates | |
1267 | @c a window for a concurrent setpriority (or nice) call to be lost or | |
1268 | @c exhibit surprising behavior. | |
639c6286 | 1269 | Increment the nice value of the calling process by @var{increment}. |
6a7a8b22 AJ |
1270 | The return value is the new nice value on success, and @code{-1} on |
1271 | failure. In the case of failure, @code{errno} will be set to the | |
1272 | same values as for @code{setpriority}. | |
1273 | ||
5ce8f203 UD |
1274 | |
1275 | Here is an equivalent definition of @code{nice}: | |
1276 | ||
1277 | @smallexample | |
1278 | int | |
1279 | nice (int increment) | |
1280 | @{ | |
6a7a8b22 AJ |
1281 | int result, old = getpriority (PRIO_PROCESS, 0); |
1282 | result = setpriority (PRIO_PROCESS, 0, old + increment); | |
1283 | if (result != -1) | |
1284 | return old + increment; | |
1285 | else | |
1286 | return -1; | |
5ce8f203 UD |
1287 | @} |
1288 | @end smallexample | |
1289 | @end deftypefun | |
b642f101 | 1290 | |
d9997a45 UD |
1291 | |
1292 | @node CPU Affinity | |
1293 | @subsection Limiting execution to certain CPUs | |
1294 | ||
1295 | On a multi-processor system the operating system usually distributes | |
1296 | the different processes which are runnable on all available CPUs in a | |
1297 | way which allows the system to work most efficiently. Which processes | |
1298 | and threads run can be to some extend be control with the scheduling | |
1299 | functionality described in the last sections. But which CPU finally | |
1300 | executes which process or thread is not covered. | |
1301 | ||
1302 | There are a number of reasons why a program might want to have control | |
1303 | over this aspect of the system as well: | |
1304 | ||
1305 | @itemize @bullet | |
1306 | @item | |
1307 | One thread or process is responsible for absolutely critical work | |
1308 | which under no circumstances must be interrupted or hindered from | |
1309 | making process by other process or threads using CPU resources. In | |
1310 | this case the special process would be confined to a CPU which no | |
1311 | other process or thread is allowed to use. | |
1312 | ||
1313 | @item | |
1314 | The access to certain resources (RAM, I/O ports) has different costs | |
1315 | from different CPUs. This is the case in NUMA (Non-Uniform Memory | |
11bf311e | 1316 | Architecture) machines. Preferably memory should be accessed locally |
d9997a45 UD |
1317 | but this requirement is usually not visible to the scheduler. |
1318 | Therefore forcing a process or thread to the CPUs which have local | |
1319 | access to the mostly used memory helps to significantly boost the | |
1320 | performance. | |
1321 | ||
1322 | @item | |
1323 | In controlled runtimes resource allocation and book-keeping work (for | |
1324 | instance garbage collection) is performance local to processors. This | |
1325 | can help to reduce locking costs if the resources do not have to be | |
1326 | protected from concurrent accesses from different processors. | |
1327 | @end itemize | |
1328 | ||
1329 | The POSIX standard up to this date is of not much help to solve this | |
1330 | problem. The Linux kernel provides a set of interfaces to allow | |
1331 | specifying @emph{affinity sets} for a process. The scheduler will | |
bbf70ae9 | 1332 | schedule the thread or process on CPUs specified by the affinity |
1f77f049 | 1333 | masks. The interfaces which @theglibc{} define follow to some |
d9997a45 UD |
1334 | extend the Linux kernel interface. |
1335 | ||
1336 | @comment sched.h | |
1337 | @comment GNU | |
1338 | @deftp {Data Type} cpu_set_t | |
1339 | This data set is a bitset where each bit represents a CPU. How the | |
1340 | system's CPUs are mapped to bits in the bitset is system dependent. | |
1341 | The data type has a fixed size; in the unlikely case that the number | |
1342 | of bits are not sufficient to describe the CPUs of the system a | |
1343 | different interface has to be used. | |
1344 | ||
1345 | This type is a GNU extension and is defined in @file{sched.h}. | |
1346 | @end deftp | |
1347 | ||
1348 | To manipulate the bitset, to set and reset bits, a number of macros is | |
1349 | defined. Some of the macros take a CPU number as a parameter. Here | |
1350 | it is important to never exceed the size of the bitset. The following | |
1351 | macro specifies the number of bits in the @code{cpu_set_t} bitset. | |
1352 | ||
1353 | @comment sched.h | |
1354 | @comment GNU | |
1355 | @deftypevr Macro int CPU_SETSIZE | |
1356 | The value of this macro is the maximum number of CPUs which can be | |
1357 | handled with a @code{cpu_set_t} object. | |
1358 | @end deftypevr | |
1359 | ||
1360 | The type @code{cpu_set_t} should be considered opaque; all | |
1361 | manipulation should happen via the next four macros. | |
1362 | ||
1363 | @comment sched.h | |
1364 | @comment GNU | |
1365 | @deftypefn Macro void CPU_ZERO (cpu_set_t *@var{set}) | |
c8ce789c AO |
1366 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}} |
1367 | @c CPU_ZERO ok | |
1368 | @c __CPU_ZERO_S ok | |
1369 | @c memset dup ok | |
d9997a45 UD |
1370 | This macro initializes the CPU set @var{set} to be the empty set. |
1371 | ||
1372 | This macro is a GNU extension and is defined in @file{sched.h}. | |
1373 | @end deftypefn | |
1374 | ||
1375 | @comment sched.h | |
1376 | @comment GNU | |
1377 | @deftypefn Macro void CPU_SET (int @var{cpu}, cpu_set_t *@var{set}) | |
c8ce789c AO |
1378 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}} |
1379 | @c CPU_SET ok | |
1380 | @c __CPU_SET_S ok | |
1381 | @c __CPUELT ok | |
1382 | @c __CPUMASK ok | |
d9997a45 UD |
1383 | This macro adds @var{cpu} to the CPU set @var{set}. |
1384 | ||
1385 | The @var{cpu} parameter must not have side effects since it is | |
1386 | evaluated more than once. | |
1387 | ||
1388 | This macro is a GNU extension and is defined in @file{sched.h}. | |
1389 | @end deftypefn | |
1390 | ||
1391 | @comment sched.h | |
1392 | @comment GNU | |
1393 | @deftypefn Macro void CPU_CLR (int @var{cpu}, cpu_set_t *@var{set}) | |
c8ce789c AO |
1394 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}} |
1395 | @c CPU_CLR ok | |
1396 | @c __CPU_CLR_S ok | |
1397 | @c __CPUELT dup ok | |
1398 | @c __CPUMASK dup ok | |
d9997a45 UD |
1399 | This macro removes @var{cpu} from the CPU set @var{set}. |
1400 | ||
1401 | The @var{cpu} parameter must not have side effects since it is | |
1402 | evaluated more than once. | |
1403 | ||
1404 | This macro is a GNU extension and is defined in @file{sched.h}. | |
1405 | @end deftypefn | |
1406 | ||
1407 | @comment sched.h | |
1408 | @comment GNU | |
1409 | @deftypefn Macro int CPU_ISSET (int @var{cpu}, const cpu_set_t *@var{set}) | |
c8ce789c AO |
1410 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}} |
1411 | @c CPU_ISSET ok | |
1412 | @c __CPU_ISSET_S ok | |
1413 | @c __CPUELT dup ok | |
1414 | @c __CPUMASK dup ok | |
d9997a45 UD |
1415 | This macro returns a nonzero value (true) if @var{cpu} is a member |
1416 | of the CPU set @var{set}, and zero (false) otherwise. | |
1417 | ||
1418 | The @var{cpu} parameter must not have side effects since it is | |
1419 | evaluated more than once. | |
1420 | ||
1421 | This macro is a GNU extension and is defined in @file{sched.h}. | |
1422 | @end deftypefn | |
1423 | ||
1424 | ||
1425 | CPU bitsets can be constructed from scratch or the currently installed | |
1426 | affinity mask can be retrieved from the system. | |
1427 | ||
1428 | @comment sched.h | |
1429 | @comment GNU | |
6f0b2e1f | 1430 | @deftypefun int sched_getaffinity (pid_t @var{pid}, size_t @var{cpusetsize}, cpu_set_t *@var{cpuset}) |
c8ce789c AO |
1431 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}} |
1432 | @c Wrapped syscall to zero out past the kernel cpu set size; Linux | |
1433 | @c only. | |
d9997a45 UD |
1434 | |
1435 | This functions stores the CPU affinity mask for the process or thread | |
6f0b2e1f RM |
1436 | with the ID @var{pid} in the @var{cpusetsize} bytes long bitmap |
1437 | pointed to by @var{cpuset}. If successful, the function always | |
1438 | initializes all bits in the @code{cpu_set_t} object and returns zero. | |
d9997a45 UD |
1439 | |
1440 | If @var{pid} does not correspond to a process or thread on the system | |
1441 | the or the function fails for some other reason, it returns @code{-1} | |
1442 | and @code{errno} is set to represent the error condition. | |
1443 | ||
1444 | @table @code | |
1445 | @item ESRCH | |
1446 | No process or thread with the given ID found. | |
1447 | ||
1448 | @item EFAULT | |
1449 | The pointer @var{cpuset} is does not point to a valid object. | |
1450 | @end table | |
1451 | ||
1452 | This function is a GNU extension and is declared in @file{sched.h}. | |
1453 | @end deftypefun | |
1454 | ||
1455 | Note that it is not portably possible to use this information to | |
1456 | retrieve the information for different POSIX threads. A separate | |
1457 | interface must be provided for that. | |
1458 | ||
1459 | @comment sched.h | |
1460 | @comment GNU | |
6f0b2e1f | 1461 | @deftypefun int sched_setaffinity (pid_t @var{pid}, size_t @var{cpusetsize}, const cpu_set_t *@var{cpuset}) |
c8ce789c AO |
1462 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}} |
1463 | @c Wrapped syscall to detect attempts to set bits past the kernel cpu | |
1464 | @c set size; Linux only. | |
d9997a45 | 1465 | |
6f0b2e1f RM |
1466 | This function installs the @var{cpusetsize} bytes long affinity mask |
1467 | pointed to by @var{cpuset} for the process or thread with the ID @var{pid}. | |
1468 | If successful the function returns zero and the scheduler will in future | |
1469 | take the affinity information into account. | |
d9997a45 UD |
1470 | |
1471 | If the function fails it will return @code{-1} and @code{errno} is set | |
1472 | to the error code: | |
1473 | ||
1474 | @table @code | |
1475 | @item ESRCH | |
1476 | No process or thread with the given ID found. | |
1477 | ||
1478 | @item EFAULT | |
1479 | The pointer @var{cpuset} is does not point to a valid object. | |
1480 | ||
1481 | @item EINVAL | |
1482 | The bitset is not valid. This might mean that the affinity set might | |
1483 | not leave a processor for the process or thread to run on. | |
1484 | @end table | |
1485 | ||
1486 | This function is a GNU extension and is declared in @file{sched.h}. | |
1487 | @end deftypefun | |
1488 | ||
1489 | ||
b642f101 UD |
1490 | @node Memory Resources |
1491 | @section Querying memory available resources | |
1492 | ||
1493 | The amount of memory available in the system and the way it is organized | |
1494 | determines oftentimes the way programs can and have to work. For | |
5a7eedfb | 1495 | functions like @code{mmap} it is necessary to know about the size of |
b642f101 UD |
1496 | individual memory pages and knowing how much memory is available enables |
1497 | a program to select appropriate sizes for, say, caches. Before we get | |
1498 | into these details a few words about memory subsystems in traditional | |
5a7eedfb | 1499 | Unix systems will be given. |
b642f101 UD |
1500 | |
1501 | @menu | |
1502 | * Memory Subsystem:: Overview about traditional Unix memory handling. | |
1503 | * Query Memory Parameters:: How to get information about the memory | |
1504 | subsystem? | |
1505 | @end menu | |
1506 | ||
1507 | @node Memory Subsystem | |
1508 | @subsection Overview about traditional Unix memory handling | |
1509 | ||
1510 | @cindex address space | |
1511 | @cindex physical memory | |
1512 | @cindex physical address | |
1513 | Unix systems normally provide processes virtual address spaces. This | |
1514 | means that the addresses of the memory regions do not have to correspond | |
1515 | directly to the addresses of the actual physical memory which stores the | |
1516 | data. An extra level of indirection is introduced which translates | |
1517 | virtual addresses into physical addresses. This is normally done by the | |
1518 | hardware of the processor. | |
1519 | ||
1520 | @cindex shared memory | |
1521 | Using a virtual address space has several advantage. The most important | |
1522 | is process isolation. The different processes running on the system | |
1523 | cannot interfere directly with each other. No process can write into | |
1524 | the address space of another process (except when shared memory is used | |
1525 | but then it is wanted and controlled). | |
1526 | ||
1527 | Another advantage of virtual memory is that the address space the | |
1528 | processes see can actually be larger than the physical memory available. | |
1529 | The physical memory can be extended by storage on an external media | |
1530 | where the content of currently unused memory regions is stored. The | |
1531 | address translation can then intercept accesses to these memory regions | |
1532 | and make memory content available again by loading the data back into | |
1533 | memory. This concept makes it necessary that programs which have to use | |
1534 | lots of memory know the difference between available virtual address | |
1535 | space and available physical memory. If the working set of virtual | |
1536 | memory of all the processes is larger than the available physical memory | |
1537 | the system will slow down dramatically due to constant swapping of | |
1538 | memory content from the memory to the storage media and back. This is | |
1539 | called ``thrashing''. | |
1540 | @cindex thrashing | |
1541 | ||
1542 | @cindex memory page | |
1543 | @cindex page, memory | |
1544 | A final aspect of virtual memory which is important and follows from | |
1545 | what is said in the last paragraph is the granularity of the virtual | |
1546 | address space handling. When we said that the virtual address handling | |
1547 | stores memory content externally it cannot do this on a byte-by-byte | |
1548 | basis. The administrative overhead does not allow this (leaving alone | |
1549 | the processor hardware). Instead several thousand bytes are handled | |
1550 | together and form a @dfn{page}. The size of each page is always a power | |
1551 | of two byte. The smallest page size in use today is 4096, with 8192, | |
1552 | 16384, and 65536 being other popular sizes. | |
1553 | ||
1554 | @node Query Memory Parameters | |
1555 | @subsection How to get information about the memory subsystem? | |
1556 | ||
1557 | The page size of the virtual memory the process sees is essential to | |
1558 | know in several situations. Some programming interface (e.g., | |
1559 | @code{mmap}, @pxref{Memory-mapped I/O}) require the user to provide | |
1560 | information adjusted to the page size. In the case of @code{mmap} is it | |
1561 | necessary to provide a length argument which is a multiple of the page | |
1562 | size. Another place where the knowledge about the page size is useful | |
1563 | is in memory allocation. If one allocates pieces of memory in larger | |
1564 | chunks which are then subdivided by the application code it is useful to | |
1565 | adjust the size of the larger blocks to the page size. If the total | |
1566 | memory requirement for the block is close (but not larger) to a multiple | |
1567 | of the page size the kernel's memory handling can work more effectively | |
1568 | since it only has to allocate memory pages which are fully used. (To do | |
1569 | this optimization it is necessary to know a bit about the memory | |
1570 | allocator which will require a bit of memory itself for each block and | |
1571 | this overhead must not push the total size over the page size multiple. | |
1572 | ||
1573 | The page size traditionally was a compile time constant. But recent | |
1574 | development of processors changed this. Processors now support | |
1575 | different page sizes and they can possibly even vary among different | |
1576 | processes on the same system. Therefore the system should be queried at | |
1577 | runtime about the current page size and no assumptions (except about it | |
1578 | being a power of two) should be made. | |
1579 | ||
1580 | @vindex _SC_PAGESIZE | |
1581 | The correct interface to query about the page size is @code{sysconf} | |
1582 | (@pxref{Sysconf Definition}) with the parameter @code{_SC_PAGESIZE}. | |
1583 | There is a much older interface available, too. | |
1584 | ||
1585 | @comment unistd.h | |
1586 | @comment BSD | |
1587 | @deftypefun int getpagesize (void) | |
c8ce789c AO |
1588 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}} |
1589 | @c Obtained from the aux vec at program startup time. GNU/Linux/m68k is | |
1590 | @c the exception, with the possibility of a syscall. | |
b642f101 UD |
1591 | The @code{getpagesize} function returns the page size of the process. |
1592 | This value is fixed for the runtime of the process but can vary in | |
1593 | different runs of the application. | |
1594 | ||
1595 | The function is declared in @file{unistd.h}. | |
1596 | @end deftypefun | |
1597 | ||
1598 | Widely available on @w{System V} derived systems is a method to get | |
1599 | information about the physical memory the system has. The call | |
1600 | ||
1601 | @vindex _SC_PHYS_PAGES | |
1602 | @cindex sysconf | |
1603 | @smallexample | |
1604 | sysconf (_SC_PHYS_PAGES) | |
1605 | @end smallexample | |
1606 | ||
cb4fe8a2 UD |
1607 | @noindent |
1608 | returns the total number of pages of physical the system has. | |
b642f101 UD |
1609 | This does not mean all this memory is available. This information can |
1610 | be found using | |
1611 | ||
1612 | @vindex _SC_AVPHYS_PAGES | |
1613 | @cindex sysconf | |
1614 | @smallexample | |
1615 | sysconf (_SC_AVPHYS_PAGES) | |
1616 | @end smallexample | |
1617 | ||
1618 | These two values help to optimize applications. The value returned for | |
1619 | @code{_SC_AVPHYS_PAGES} is the amount of memory the application can use | |
1620 | without hindering any other process (given that no other process | |
1621 | increases its memory usage). The value returned for | |
1622 | @code{_SC_PHYS_PAGES} is more or less a hard limit for the working set. | |
1623 | If all applications together constantly use more than that amount of | |
1624 | memory the system is in trouble. | |
1625 | ||
1f77f049 | 1626 | @Theglibc{} provides in addition to these already described way to |
cb4fe8a2 UD |
1627 | get this information two functions. They are declared in the file |
1628 | @file{sys/sysinfo.h}. Programmers should prefer to use the | |
1629 | @code{sysconf} method described above. | |
1630 | ||
1631 | @comment sys/sysinfo.h | |
1632 | @comment GNU | |
4c78249d | 1633 | @deftypefun {long int} get_phys_pages (void) |
c8ce789c AO |
1634 | @safety{@prelim{}@mtsafe{}@asunsafe{@ascuheap{} @asulock{}}@acunsafe{@aculock{} @acsfd{} @acsmem{}}} |
1635 | @c This fopens a /proc file and scans it for the requested information. | |
cb4fe8a2 UD |
1636 | The @code{get_phys_pages} function returns the total number of pages of |
1637 | physical the system has. To get the amount of memory this number has to | |
1638 | be multiplied by the page size. | |
1639 | ||
1640 | This function is a GNU extension. | |
1641 | @end deftypefun | |
1642 | ||
1643 | @comment sys/sysinfo.h | |
1644 | @comment GNU | |
4c78249d | 1645 | @deftypefun {long int} get_avphys_pages (void) |
c8ce789c | 1646 | @safety{@prelim{}@mtsafe{}@asunsafe{@ascuheap{} @asulock{}}@acunsafe{@aculock{} @acsfd{} @acsmem{}}} |
cd1fb604 | 1647 | The @code{get_avphys_pages} function returns the number of available pages of |
cb4fe8a2 UD |
1648 | physical the system has. To get the amount of memory this number has to |
1649 | be multiplied by the page size. | |
1650 | ||
1651 | This function is a GNU extension. | |
1652 | @end deftypefun | |
1653 | ||
b642f101 UD |
1654 | @node Processor Resources |
1655 | @section Learn about the processors available | |
1656 | ||
1657 | The use of threads or processes with shared memory allows an application | |
1658 | to take advantage of all the processing power a system can provide. If | |
1659 | the task can be parallelized the optimal way to write an application is | |
1660 | to have at any time as many processes running as there are processors. | |
1661 | To determine the number of processors available to the system one can | |
1662 | run | |
1663 | ||
1664 | @vindex _SC_NPROCESSORS_CONF | |
1665 | @cindex sysconf | |
1666 | @smallexample | |
1667 | sysconf (_SC_NPROCESSORS_CONF) | |
1668 | @end smallexample | |
1669 | ||
1670 | @noindent | |
1671 | which returns the number of processors the operating system configured. | |
1672 | But it might be possible for the operating system to disable individual | |
1673 | processors and so the call | |
1674 | ||
1675 | @vindex _SC_NPROCESSORS_ONLN | |
1676 | @cindex sysconf | |
1677 | @smallexample | |
1678 | sysconf (_SC_NPROCESSORS_ONLN) | |
1679 | @end smallexample | |
1680 | ||
1681 | @noindent | |
26428b7c | 1682 | returns the number of processors which are currently online (i.e., |
b642f101 | 1683 | available). |
e4cf5229 | 1684 | |
1f77f049 | 1685 | For these two pieces of information @theglibc{} also provides |
cb4fe8a2 UD |
1686 | functions to get the information directly. The functions are declared |
1687 | in @file{sys/sysinfo.h}. | |
1688 | ||
1689 | @comment sys/sysinfo.h | |
1690 | @comment GNU | |
1691 | @deftypefun int get_nprocs_conf (void) | |
c8ce789c AO |
1692 | @safety{@prelim{}@mtsafe{}@asunsafe{@ascuheap{} @asulock{}}@acunsafe{@aculock{} @acsfd{} @acsmem{}}} |
1693 | @c This function reads from from /sys using dir streams (single user, so | |
1694 | @c no @mtasurace issue), and on some arches, from /proc using streams. | |
cb4fe8a2 UD |
1695 | The @code{get_nprocs_conf} function returns the number of processors the |
1696 | operating system configured. | |
1697 | ||
1698 | This function is a GNU extension. | |
1699 | @end deftypefun | |
1700 | ||
1701 | @comment sys/sysinfo.h | |
1702 | @comment GNU | |
1703 | @deftypefun int get_nprocs (void) | |
c8ce789c AO |
1704 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{@acsfd{}}} |
1705 | @c This function reads from /proc using file descriptor I/O. | |
cb4fe8a2 UD |
1706 | The @code{get_nprocs} function returns the number of available processors. |
1707 | ||
1708 | This function is a GNU extension. | |
1709 | @end deftypefun | |
1710 | ||
e4cf5229 UD |
1711 | @cindex load average |
1712 | Before starting more threads it should be checked whether the processors | |
1713 | are not already overused. Unix systems calculate something called the | |
1714 | @dfn{load average}. This is a number indicating how many processes were | |
1715 | running. This number is average over different periods of times | |
1716 | (normally 1, 5, and 15 minutes). | |
1717 | ||
1718 | @comment stdlib.h | |
1719 | @comment BSD | |
1720 | @deftypefun int getloadavg (double @var{loadavg}[], int @var{nelem}) | |
c8ce789c AO |
1721 | @safety{@prelim{}@mtsafe{}@assafe{}@acsafe{@acsfd{}}} |
1722 | @c Calls host_info on HURD; on Linux, opens /proc/loadavg, reads from | |
1723 | @c it, closes it, without cancellation point, and calls strtod_l with | |
1724 | @c the C locale to convert the strings to doubles. | |
e4cf5229 | 1725 | This function gets the 1, 5 and 15 minute load averages of the |
cf822e3c | 1726 | system. The values are placed in @var{loadavg}. @code{getloadavg} will |
e4cf5229 UD |
1727 | place at most @var{nelem} elements into the array but never more than |
1728 | three elements. The return value is the number of elements written to | |
1729 | @var{loadavg}, or -1 on error. | |
1730 | ||
1731 | This function is declared in @file{stdlib.h}. | |
1732 | @end deftypefun |