Re: 2.4.22 ACPI power off via sysrq not working

Re: 2.4.22 ACPI power off via sysrq not working

[Andi Kleen]

Sylvain Pasche <sylvain_pasche@yahoo.fr> writes:

> Hi, 
> 
> If I want to halt the system using sys-rq - o key, I get an oops instead
> of a power down.
> Inside pm.c:159, there is:
> 	
> 	if (in_interrupt())
> 		BUG();
> 
> But if we look at the trace, we are in the interrupt of the keyboard
> handler.
> One fix would be to comment out the BUG line, but there's certainly "a
> better way to do it".

Use this patch.

diff -u linux/drivers/acpi/system.c-o linux/drivers/acpi/system.c
--- linux/drivers/acpi/system.c-o	2003-09-07 16:20:44.000000000 +0200
+++ linux/drivers/acpi/system.c	2003-09-08 21:04:46.000000000 +0200
@@ -1192,11 +1192,21 @@
 
 #if defined(CONFIG_MAGIC_SYSRQ) && defined(CONFIG_PM)
 
+static int po_cb_active; 
+
+static void acpi_po_tramp(void *x)
+{ 
+	acpi_power_off();	
+} 
+
 /* Simple wrapper calling power down function. */
 static void acpi_sysrq_power_off(int key, struct pt_regs *pt_regs,
 	struct kbd_struct *kbd, struct tty_struct *tty)
-{
-	acpi_power_off();
+{	
+	static struct tq_struct tq = { .routine = acpi_po_tramp };
+	if (po_cb_active++)
+		return;
+	schedule_task(&tq); 
 }
 
 struct sysrq_key_op sysrq_acpi_poweroff_op = {


-Andi

Re: 2.4.22 ACPI power off via sysrq not working

[Sylvain Pasche]

Le ven 03/10/2003 à 10:24, Andi Kleen a écrit :
> Sylvain Pasche <sylvain_pasche@yahoo.fr> writes:

> > If I want to halt the system using sys-rq - o key, I get an oops instead
> > of a power down.

> Use this patch.
> 
> diff -u linux/drivers/acpi/system.c-o linux/drivers/acpi/system.c
> --- linux/drivers/acpi/system.c-o	2003-09-07 16:20:44.000000000 +0200
> +++ linux/drivers/acpi/system.c	2003-09-08 21:04:46.000000000 +0200
> @@ -1192,11 +1192,21 @@
>  
>  #if defined(CONFIG_MAGIC_SYSRQ) && defined(CONFIG_PM)
>  
> +static int po_cb_active; 
> +
> +static void acpi_po_tramp(void *x)
> +{ 
> +	acpi_power_off();	
> +} 
> +
>  /* Simple wrapper calling power down function. */
>  static void acpi_sysrq_power_off(int key, struct pt_regs *pt_regs,
>  	struct kbd_struct *kbd, struct tty_struct *tty)
> -{
> -	acpi_power_off();
> +{	
> +	static struct tq_struct tq = { .routine = acpi_po_tramp };
> +	if (po_cb_active++)
> +		return;
> +	schedule_task(&tq); 
>  }
>  
>  struct sysrq_key_op sysrq_acpi_poweroff_op = {
> 
> 
> -Andi

Yes, it works perfectly with it, thanks.
Do you forward it to the acpi maintainer ?

Cheers,

Sylvain

Re: 2.4.22 ACPI power off via sysrq not working

[Marcelo Tosatti]

On 3 Oct 2003, Andi Kleen wrote:

> Sylvain Pasche <sylvain_pasche@yahoo.fr> writes:
> 
> > Hi, 
> > 
> > If I want to halt the system using sys-rq - o key, I get an oops instead
> > of a power down.
> > Inside pm.c:159, there is:
> > 	
> > 	if (in_interrupt())
> > 		BUG();
> > 
> > But if we look at the trace, we are in the interrupt of the keyboard
> > handler.
> > One fix would be to comment out the BUG line, but there's certainly "a
> > better way to do it".
> 
> Use this patch.
> 
> diff -u linux/drivers/acpi/system.c-o linux/drivers/acpi/system.c
> --- linux/drivers/acpi/system.c-o	2003-09-07 16:20:44.000000000 +0200
> +++ linux/drivers/acpi/system.c	2003-09-08 21:04:46.000000000 +0200
> @@ -1192,11 +1192,21 @@
>  
>  #if defined(CONFIG_MAGIC_SYSRQ) && defined(CONFIG_PM)
>  
> +static int po_cb_active; 
> +
> +static void acpi_po_tramp(void *x)
> +{ 
> +	acpi_power_off();	
> +} 
> +
>  /* Simple wrapper calling power down function. */
>  static void acpi_sysrq_power_off(int key, struct pt_regs *pt_regs,
>  	struct kbd_struct *kbd, struct tty_struct *tty)
> -{
> -	acpi_power_off();
> +{	
> +	static struct tq_struct tq = { .routine = acpi_po_tramp };
> +	if (po_cb_active++)
> +		return;
> +	schedule_task(&tq); 
>  }
>  
>  struct sysrq_key_op sysrq_acpi_poweroff_op = {

Fix applied to 2.4 mainline.

2.4.22 ACPI power off via sysrq not working

[Sylvain Pasche]

Hi, 

If I want to halt the system using sys-rq - o key, I get an oops instead
of a power down.
Inside pm.c:159, there is:
	
	if (in_interrupt())
		BUG();

But if we look at the trace, we are in the interrupt of the keyboard
handler.
One fix would be to comment out the BUG line, but there's certainly "a
better way to do it".

Sylvain


kernel BUG at pm.c:159!
invalid operand: 0000
CPU:    0
EIP:    0010:[<c012880e>]    Not tainted
Using defaults from ksymoops -t elf32-i386 -a i386
EFLAGS: 00010002
eax: 00000001   ebx: f7a9cb20   ecx: 00000000   edx: 00000002
esi: 00000001   edi: 00000003   ebp: c02f6808   esp: c0337e68
ds: 0018   es: 0018   ss: 0018
Process swapper (pid: 0, stackpage=c0337000)
Stack: 0000000b 00002a8d c011a795 f7a9cb3c f7a9cb20 00000005 c0128958
f7a9cb20 
       00000002 00000003 00000000 00000005 00000005 0000006f c01fa60e
00000002 
       00000003 c02fd208 0000006f c01fa725 00000005 c02fd208 0000006f
00000007 
Call Trace:    [<c011a795>] [<c0128958>] [<c01fa60e>] [<c01fa725>]
[<c01fa59b>]
  [<c021864b>] [<c02185a9>] [<c02164b6>] [<c02175b7>] [<c02175cf>]
[<c010a6a5>]
  [<c010a824>] [<c0107080>] [<c010cfa8>] [<c0107080>] [<c01070a3>]
[<c0107132>]
  [<c0105000>]
Code: 0f 0b 9f 00 6f 37 2b c0 83 fa 01 77 45 8b 73 14 39 fe 74 34 


>>EIP; c012880e <pm_send+2e/a0>   <=====

>>ebx; f7a9cb20 <_end+3770bfb4/385784f4>
>>ebp; c02f6808 <pm_devs_lock+0/10>
>>esp; c0337e68 <init_task_union+1e68/2000>

Trace; c011a795 <call_console_drivers+65/120>
Trace; c0128958 <pm_send_all+78/b0>
Trace; c01fa60e <acpi_system_save_state+16/91>
Trace; c01fa725 <acpi_suspend+5e/9d>
Trace; c01fa59b <acpi_power_off+7/9>
Trace; c021864b <__handle_sysrq_nolock+7b/f0>
Trace; c02185a9 <handle_sysrq+59/80>
Trace; c02164b6 <handle_scancode+276/310>
Trace; c02175b7 <handle_kbd_event+b7/c0>
Trace; c02175cf <keyboard_interrupt+f/20>
Trace; c010a6a5 <handle_IRQ_event+45/70>
Trace; c010a824 <do_IRQ+64/a0>
Trace; c0107080 <default_idle+0/40>
Trace; c010cfa8 <call_do_IRQ+5/d>
Trace; c0107080 <default_idle+0/40>
Trace; c01070a3 <default_idle+23/40>
Trace; c0107132 <cpu_idle+52/70>
Trace; c0105000 <_stext+0/0>