Denis Vlasenko | 0d05836 | 2007-04-11 16:16:41 +0000 | [diff] [blame] | 1 | NOEXEC and NOFORK applets. |
| 2 | |
| 3 | Unix shells traditionally execute some commands internally in the attempt |
| 4 | to dramatically speed up execution. It will be slow as hell if for every |
Denis Vlasenko | 8fd3712 | 2007-04-11 19:27:22 +0000 | [diff] [blame^] | 5 | "echo blah" shell will fork and exec /bin/echo. For this end, shells |
Denis Vlasenko | 0d05836 | 2007-04-11 16:16:41 +0000 | [diff] [blame] | 6 | have to _reimplement_ these commands internally. |
| 7 | |
| 8 | Busybox is unique in this regard because it already is a collection |
| 9 | of reimplemented Unix commands, and we can do the same trick |
| 10 | for speeding up busybox shells, and more. NOEXEC and NOFORK applets |
| 11 | are exactly those applets which are eligible for these tricks. |
| 12 | |
| 13 | Applet will be subject to NOFORK/NOEXEC tricks if it is marked as such |
| 14 | in applets.h. CONFIG_FEATURE_PREFER_APPLETS is a config option which |
| 15 | globally enables usage of NOFORK/NOEXEC tricks. |
| 16 | |
| 17 | If you want to call a program and wait for it, use spawn_and_wait(argv). |
| 18 | It will check whether argv[0] is an applet name and will optionally |
| 19 | do NOFORK/NOEXEC thing. |
| 20 | |
| 21 | NOEXEC |
| 22 | |
| 23 | NOEXEC applet should work correctly if another applet forks and then |
| 24 | executes exit(<applet>_main(argc,argv)) in the child. The rules |
| 25 | roughly are: |
| 26 | |
| 27 | * do not expect shared global variables/buffers to be in their |
| 28 | "initialized" state. Examples: xfunc_error_retval can be != 1, |
| 29 | bb_common_bufsiz1 can be scribbled over, ... |
| 30 | * do not expect that stdio wasn't used before. Calling set[v]buf() |
| 31 | can be disastrous. |
| 32 | * ... |
| 33 | |
| 34 | NOEXEC applets save only one half of fork+exec overhead. |
| 35 | NOEXEC trick is disabled for NOMMU compile. |
| 36 | |
| 37 | NOFORK |
| 38 | |
| 39 | NOFORK applet should work correctly if another applet simply runs |
| 40 | <applet>_main(argc,argv) and then continues with its business (xargs, |
| 41 | find, shells can do it). This poses much more serious limitations |
| 42 | on what applet can/cannot do: |
| 43 | |
| 44 | * all NOEXEC limitations apply. |
| 45 | * do not ever exit() or exec(). |
| 46 | - xfuncs are okay. They are using special trick to return |
| 47 | to the caller applet instead of dying when they detect "x" condition. |
| 48 | - you may "exit" to caller applet by calling xfunc_die(). Return value |
| 49 | is taken from xfunc_error_retval. |
| 50 | - fflush_stdout_and_exit(n) is ok to use. |
| 51 | * do not use shared global data, or save/restore shared global data |
| 52 | prior to returning. (e.g. bb_common_bufsiz1 is off-limits). |
| 53 | - getopt32() is ok to use. You do not need to save/restore option_mask32, |
| 54 | it is already done by core code. |
| 55 | * if you allocate memory, you can use xmalloc() only on the very first |
| 56 | allocation. All other allocations should use malloc[_or_warn](). |
| 57 | After first allocation, you cannot use any xfuncs. |
| 58 | * All allocated data, opened files, signal handlers, termios settings, |
| 59 | O_NONBLOCK flags etc should be freed/closed/restored prior to return. |
| 60 | * ... |
| 61 | |
| 62 | NOFORK applets give the most of speed advantage, but are trickiest |
| 63 | to implement. In order to minimize amount of bugs and maintenance, |
| 64 | prime candidates for NOFORK-ification are those applets which |
| 65 | are small and easy to audit, and those which are more likely to be |
| 66 | frequently executed from shell/find/xargs, particularly in shell |
| 67 | script loops. Applets which mess with signal handlers, termios etc |
| 68 | are probably not worth the effort. |
| 69 | |
| 70 | Any NOFORK applet is also a NOEXEC applet. |