Distribution:Ubuntu 5.04 Hardware Environment: Dual P3/866MHz 768 megs of PC-133 RAM MSI 694-D2 Ver. 2 motherboard a 200 Gig IDE hdd, 10X IDE cdrom Via-Rhine NIC SYM- SCSI card. Software Environment: Using kernel 2.6.13-rc4 through latest rc6 (as of two days ago). gcc-3.3.5 Problem Description: I've booted the machine using 2.6.13-rc4 and/or later kernels and after boot, I've tried rebooting it (either immediately or after some normal-compilation, testing). At the end of reboot cycle, it just powers off. Earlier 2.6.13-rcX kernels panic during startup. 2.6.12.5 works just fine. Steps to reproduce: Boot the kernel, type 'reboot' on command line and watch the machine act as if you've called halt -p on it. It doesn't happen if you add 'reboot=w' or 'acpi=off' to the kernel command line before booting.
Apparently, another guy has also the same problem, with a different hardware: http://lkml.org/lkml/2005/8/20/57
Created attachment 5725 [details] Kernel config file This is the config file used to compile the kernel
Created attachment 5726 [details] kernel messages This is dmesg after boot, with 'reboot=w' added to commandline.
Created attachment 5727 [details] lspci -vv output
Created attachment 5728 [details] lspci -vv output
Created attachment 5729 [details] lspci -vv output Output of lspci -vv
can you isolate where between 2.6.12 aznd 2.6.12-rc3 the regression begain? Alternatively, can you see if the latest ACPI patch caused this regression? It is located here and you can apply it to 2.6.12.*: http://ftp.kernel.org/pub/linux/kernel/people/lenb/acpi/patches/release/2.6.12/
2.6.12.5 works. 2.6.13-rc1 through 2.6.13-rc3 crash at boot (I'll recompile them and post the log messages here.) 2.6.13-rc4 through 2.6.13-rc6 shutdown on crash. I shall also try the patch you mentioned.
Yes. The patch at http://ftp.kernel.org/pub/linux/kernel/people/lenb/acpi/patches/release/2.6.12/ acpi-20050729-2.6.12.diff.gz shuts down the machine when executing reboot.
The patch at http://ftp.kernel.org/pub/linux/kernel/people/lenb/acpi/patches/release/2.6.12/ broken-out/acpi-4-poweroff-alexey.y.starikovskiy@intel.com.patch, when added to 2.6.12.5 also powers off the system on reboot.
I modified portions of the code touched by the patch so that the functions print their name on the way on reboot: * Rebooting... <--- This is the init (or initscripts) message. acpi_shutdown: called. acpi_sleep_prepare: state=5 Restarting system. On a separate note, Machine shuts off even with CPU triple faulting. I commented out the code handing reset via keyboard controller (since the machine is attached to a KVM, and most of the time, the other machine is using the keyboard). It still turned itself off at the end of reboot.
There is a patch in http://lkml.org/lkml/2005/8/27/8 (also included in 2.6.13 final) that fixes this issue: diff --git a/drivers/acpi/sleep/poweroff.c b/drivers/acpi/sleep/poweroff.c --- a/drivers/acpi/sleep/poweroff.c +++ b/drivers/acpi/sleep/poweroff.c @@ -55,7 +55,11 @@ void acpi_power_off(void) static int acpi_shutdown(struct sys_device *x) { - return acpi_sleep_prepare(ACPI_STATE_S5); + if (system_state == SYSTEM_POWER_OFF) { + /* Prepare if we are going to power off the system */ + return acpi_sleep_prepare(ACPI_STATE_S5); + } + return 0; }
Created attachment 5843 [details] set system_state at S4 power-off Could you please verify that this patch does not break your setup?
And what should I apply this patch to?
Please apply over patch in #12, e.g. to 2.6.13 kernel. Thanks.
Unfortunately, I can't boot 2.6.13 on this machine: See bug 5171.
The bug seems to be fixed in 2.6.14-rc2; It reboots without any external command line switch.