Bug 1324
| Summary: | telinit 1 boots into the default normal runlevel | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Product: | Sisyphus | Reporter: | imz <vanyaz> | ||||||
| Component: | startup | Assignee: | Alexey Gladkov <legion> | ||||||
| Status: | NEW --- | QA Contact: | |||||||
| Severity: | major | ||||||||
| Priority: | P4 | CC: | glebfm, ldv, legion, mike | ||||||
| Version: | unstable | ||||||||
| Hardware: | all | ||||||||
| OS: | Linux | ||||||||
| Attachments: |
|
||||||||
Surprisingly, removing \'exec \' in /etc/init.d/single made it work correctly. (The patch is attached.) Surprisingly, removing \'exec \' in /etc/init.d/single made it work correctly. (The patch is attached.) Applied patch in 5.49-ipl45mdk Applied patch in 5.49-ipl45mdk Can this be an error in /sbin/init? And the patch only implemets a work-araound. Can this be an error in /sbin/init? And the patch only implemets a work-araound. Unable to reproduce all these strange things. I\'m going to revert this patch and add \"exec\" back to /etc/init.d/single. Please check this \"exec\" thing again with SysVinit-2.84-alt5. Unable to reproduce all these strange things. I\'m going to revert this patch and add \"exec\" back to /etc/init.d/single. Please check this \"exec\" thing again with SysVinit-2.84-alt5. I\'m able to duplicate it with SysVinit-2.84-alt5, initscripts-5.49-ipl48mdk. It happens always. This time the behaviour is the same, no matter whether \'exec \' is present in /etc/init.d/single: I\'m not able to boot into runlevel 1 (either with telinit 1 or appending 1 in GRUB). I\'m able to duplicate it with SysVinit-2.84-alt5, initscripts-5.49-ipl48mdk. It happens always. This time the behaviour is the same, no matter whether \'exec \' is present in /etc/init.d/single: I\'m not able to boot into runlevel 1 (either with telinit 1 or appending 1 in GRUB). Please provide as much information as possible, including your /etc/inittab Please provide as much information as possible, including your /etc/inittab Tested with SysVinit-2.82-alt2. With \'exec telinit -t1 S\' in /etc/init.d/single: after telinit 1, it stops all services, writes \'telling INIT to go into single user mode\' and hangs, does nothing. SysRQ + T shows there 2 init processes and some k*. After appending 1 in GRUB, it writes after the minimal start-up about going into single user mode, no more start up messages appear, and finally, the system has been booted into runlevel 3 (all the services started). Without \'exec \': telinit 1 boots into runlevel 3 (as described: without messages). The same when appending 1 in GRUB. Tested with SysVinit-2.82-alt2. With \'exec telinit -t1 S\' in /etc/init.d/single: after telinit 1, it stops all services, writes \'telling INIT to go into single user mode\' and hangs, does nothing. SysRQ + T shows there 2 init processes and some k*. After appending 1 in GRUB, it writes after the minimal start-up about going into single user mode, no more start up messages appear, and finally, the system has been booted into runlevel 3 (all the services started). Without \'exec \': telinit 1 boots into runlevel 3 (as described: without messages). The same when appending 1 in GRUB. Tested with SysVinit-2.84-alt4 again. The bahaviour is the same, no matter whether \'exec \' is there. It writes \'Telling INIT to go to single user mode\' and after a while runlevel 3 is there (the login prompt), no start-up messages appeared before the login screen appeared. Tested with SysVinit-2.84-alt4 again. The bahaviour is the same, no matter whether \'exec \' is there. It writes \'Telling INIT to go to single user mode\' and after a while runlevel 3 is there (the login prompt), no start-up messages appeared before the login screen appeared. 1. Does -alt4 and -alt5 behaviour differ? -alt5 is preferred for testing. 2. Please restore CTRL-ALT-DEL default action for a while. 1. Does -alt4 and -alt5 behaviour differ? -alt5 is preferred for testing. 2. Please restore CTRL-ALT-DEL default action for a while. 1. It seems, no. 2. will do it 1. It seems, no. 2. will do it Now I\'m using the /etc/inittab from initscripts-5.49-ipl48mdk and SysVinit-2.84-alt5. Nothing changed: after \'Telling INIT to go to single user mode\' no messages appear, but runlevel 3 is booted. Now I\'m using the /etc/inittab from initscripts-5.49-ipl48mdk and SysVinit-2.84-alt5. Nothing changed: after \'Telling INIT to go to single user mode\' no messages appear, but runlevel 3 is booted. /etc/rc.d/init.d/single: remove exec (if any) and add smth like echo \"rc=$?\" right after telinit command. What\'s the rc? /etc/rc.d/init.d/single: remove exec (if any) and add smth like echo \"rc=$?\" right after telinit command. What\'s the rc? rc=0 But the result became less stable: in 1 of 5 tests, it went into single user mode! A kind of race? So I appended these lines: $ tail -5 /etc/init.d/single # Now go to the single user level. echo \"Telling INIT to go to single user mode.\" telinit -t1 S echo \"rc=$?\" sleep 5 && echo slept and now the behaviour is stable and correct. telinit 1 brings the services down, prints \"Telling INIT to go to single user mode.\", \"rc=0\", then prints smth like \"Going SINGLE\" and asks for the password. After a while, \"slept\" is printed. After I give \"exit\" command in the single user mode shell, the system goes into runlevel 3, with all the messages this time. Appending 1 in GRUB also works fine. rc=0 But the result became less stable: in 1 of 5 tests, it went into single user mode! A kind of race? So I appended these lines: $ tail -5 /etc/init.d/single # Now go to the single user level. echo \"Telling INIT to go to single user mode.\" telinit -t1 S echo \"rc=$?\" sleep 5 && echo slept and now the behaviour is stable and correct. telinit 1 brings the services down, prints \"Telling INIT to go to single user mode.\", \"rc=0\", then prints smth like \"Going SINGLE\" and asks for the password. After a while, \"slept\" is printed. After I give \"exit\" command in the single user mode shell, the system goes into runlevel 3, with all the messages this time. Appending 1 in GRUB also works fine. Looks like a race. Could you provide some info from your /proc/cpuinfo? Looks like a race. Could you provide some info from your /proc/cpuinfo? $ cat /proc/cpuinfo processor : 0 vendor_id : GenuineIntel cpu family : 6 model : 5 model name : Pentium II (Deschutes) stepping : 2 cpu MHz : 400.917 cache size : 512 KB fdiv_bug : no hlt_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 2 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 sep mtrr pge mca cmov pat pse36 mmx fxsr bogomips : 799.53 $ cat /proc/cpuinfo processor : 0 vendor_id : GenuineIntel cpu family : 6 model : 5 model name : Pentium II (Deschutes) stepping : 2 cpu MHz : 400.917 cache size : 512 KB fdiv_bug : no hlt_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 2 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 sep mtrr pge mca cmov pat pse36 mmx fxsr bogomips : 799.53 Never saw that... Перевешиваю на startup: баг всё ещё актуален, и чинить его надо там (или в init.c sysvinit). По коду: 1) startup/rc.d/init.d/single — логика не менялась с 2002 года, последней строкой telinit -t1 S без какой-либо синхронизации. Workaround из обсуждения выше (sleep после telinit, с ним поведение стабильно корректное) так и не был добавлен. 2) sysvinit 3.00, src/init.c, fifo_new_level(): код тот же, что в 2.84, CHANGE_WAIT по-прежнему 0 — init обрабатывает запрос на переход в S во время работы wait-процесса (самого скрипта single). Гонка на месте. Варианты фикса: ожидание/sleep после telinit -t1 S в скрипте single (минимально), либо корректная синхронизация смены уровня в init.c. |
Description
imz 2002-09-28 17:12:40 MSD