When I call \'telinit 1\' (as root), all the services are brought down as exepcted. Then there is a pause, and after it some messages appear, among them \'Telling INIT to go to single user mode\'. And after a short while a login prompt appears: the system has been booted into runlevel 3 (the default). The OK messages for the started services doesn\'t show up, but the service are started (syslog logged this). The same situation when I append \'1\' to the kernel command for GRUB: it boots into runlevel 3 finally. When \'single\' is appended, it is booted into the sungle user mode as expected. \'init S\' called from a root shell also asks for the root\'s password and starts a root shell (I expect this to be the final step of \'telinit 1\'). --- --- $ rpm -q SysVinit initscripts SysVinit-2.84-alt4 initscripts-5.49-ipl44mdk
Surprisingly, removing \'exec \' in /etc/init.d/single made it work correctly. (The patch is attached.)
Applied patch in 5.49-ipl45mdk
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.
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
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.
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
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?
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?
$ 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.
