[Linux-ia64] kacpidpc process takes 100% of CPU

[Linux-ia64] kacpidpc process takes 100% of CPU

[Paris, DavidX]

I've just built 2.4.13 and 2.4.14 with and without the KDB patch and in all
cases when the system comes up there is a kacpidpc process (PID 10) taking
100% of one of the cpus (changes which CPU it's on every boot). The system
is a 4 processor Lion (733 Mhz B3) with 1 gig of ram. This process isn't
even running when I build the same kernels on a BigSpur. I've even tried
copying over the very kernel that works fine on the BigSpur, but when I boot
it on the Lion I get this kacpidpc process. Naturally the process ignores my
'kill -9's. I've tried booting the kernel on single proc Lions (733 C0s),
but the kernel won't even boot on those systems (hangs right after it loads
the SCSI driver is begins looking for devices). Does anyone have any idea
why this kacpidpc process is hogging one of the CPUs, and how I go about
getting rid of it? Unlike in the IA32 kernels, it doesn't look like you can
turn off ACPI in the IA64 kernel. Thanks,

David

Re: [Linux-ia64] kacpidpc process takes 100% of CPU

[David Mosberger]

>>>>> On Tue, 13 Nov 2001 09:43:52 -0800, "Paris, DavidX" <davidx.paris@intel.com> said:

  >> I've just built 2.4.13 and 2.4.14 with and without the KDB patch
  >> and in all cases when the system comes up there is a kacpidpc
  >> process (PID 10) taking 100% of one of the cpus (changes which
  >> CPU it's on every boot). The system is a 4 processor Lion (733
  >> Mhz B3) with 1 gig of ram. This process isn't even running when I
  >> build the same kernels on a BigSpur. I've even tried copying over
  >> the very kernel that works fine on the BigSpur, but when I boot
  >> it on the Lion I get this kacpidpc process. Naturally the process
  >> ignores my 'kill -9's. I've tried booting the kernel on single
  >> proc Lions (733 C0s), but the kernel won't even boot on those
  >> systems (hangs right after it loads the SCSI driver is begins
  >> looking for devices). Does anyone have any idea why this kacpidpc
  >> process is hogging one of the CPUs, and how I go about getting
  >> rid of it? Unlike in the IA32 kernels, it doesn't look like you
  >> can turn off ACPI in the IA64 kernel. Thanks,

What ACPI configuration do you use?  In my kernels, I turn on the
following ACPI related configs:

CONFIG_ACPI=y
CONFIG_ACPI_EFI=y
CONFIG_ACPI_INTERPRETER=y
CONFIG_ACPI_KERNEL_CONFIG=y

All other ACPI stuff I turn off.

What firmware are you using?  In my experience, Lions need build 99 or
newer.

	--david

RE: [Linux-ia64] kacpidpc process takes 100% of CPU

[Paris, DavidX]

I have the exact same ACPI configuration you do, with just those four things
turned on. I have also tried commenting all of those things out, but then
the kernel doesn't build (almost does, just barfs at the end when it's
linking everything). I think the bios I'm using is the latest development
one, v1.1 (release date was October 9th). Any ideas? I did a lot of
searching on this and I'm afraid that I'm the only person in the universe
having this problem.

David

-----Original Message-----
From: David Mosberger [mailto:davidm@hpl.hp.com]
Sent: Tuesday, November 13, 2001 9:57 AM
To: Paris, DavidX
Cc: 'linux-ia64@linuxia64.org'
Subject: Re: [Linux-ia64] kacpidpc process takes 100% of CPU


>>>>> On Tue, 13 Nov 2001 09:43:52 -0800, "Paris, DavidX"
<davidx.paris@intel.com> said:

  >> I've just built 2.4.13 and 2.4.14 with and without the KDB patch
  >> and in all cases when the system comes up there is a kacpidpc
  >> process (PID 10) taking 100% of one of the cpus (changes which
  >> CPU it's on every boot). The system is a 4 processor Lion (733
  >> Mhz B3) with 1 gig of ram. This process isn't even running when I
  >> build the same kernels on a BigSpur. I've even tried copying over
  >> the very kernel that works fine on the BigSpur, but when I boot
  >> it on the Lion I get this kacpidpc process. Naturally the process
  >> ignores my 'kill -9's. I've tried booting the kernel on single
  >> proc Lions (733 C0s), but the kernel won't even boot on those
  >> systems (hangs right after it loads the SCSI driver is begins
  >> looking for devices). Does anyone have any idea why this kacpidpc
  >> process is hogging one of the CPUs, and how I go about getting
  >> rid of it? Unlike in the IA32 kernels, it doesn't look like you
  >> can turn off ACPI in the IA64 kernel. Thanks,

What ACPI configuration do you use?  In my kernels, I turn on the
following ACPI related configs:

CONFIG_ACPI=y
CONFIG_ACPI_EFI=y
CONFIG_ACPI_INTERPRETER=y
CONFIG_ACPI_KERNEL_CONFIG=y

All other ACPI stuff I turn off.

What firmware are you using?  In my experience, Lions need build 99 or
newer.

	--david

RE: [Linux-ia64] kacpidpc process takes 100% of CPU

[Grover, Andrew]

The ACPI interpreter starts threads to run control methods, and these are
labeled kacpidpc. When nothing is wrong, these are short-lived, and then
terminate.

Can you please use kdb, break in, and get a stack trace to find out where
we're stuck?

Thanks -- Regards -- Andy

> -----Original Message-----
> From: Paris, DavidX [mailto:davidx.paris@intel.com]
> Sent: Tuesday, November 13, 2001 9:44 AM
> To: 'linux-ia64@linuxia64.org'
> Subject: [Linux-ia64] kacpidpc process takes 100% of CPU
> 
> 
> I've just built 2.4.13 and 2.4.14 with and without the KDB 
> patch and in all
> cases when the system comes up there is a kacpidpc process 
> (PID 10) taking
> 100% of one of the cpus (changes which CPU it's on every 
> boot). The system
> is a 4 processor Lion (733 Mhz B3) with 1 gig of ram. This 
> process isn't
> even running when I build the same kernels on a BigSpur. I've 
> even tried
> copying over the very kernel that works fine on the BigSpur, 
> but when I boot
> it on the Lion I get this kacpidpc process. Naturally the 
> process ignores my
> 'kill -9's. I've tried booting the kernel on single proc 
> Lions (733 C0s),
> but the kernel won't even boot on those systems (hangs right 
> after it loads
> the SCSI driver is begins looking for devices). Does anyone 
> have any idea
> why this kacpidpc process is hogging one of the CPUs, and how 
> I go about
> getting rid of it? Unlike in the IA32 kernels, it doesn't 
> look like you can
> turn off ACPI in the IA64 kernel. Thanks,
> 
> David
> 
> _______________________________________________
> Linux-IA64 mailing list
> Linux-IA64@linuxia64.org
> http://lists.linuxia64.org/lists/listinfo/linux-ia64
> 

RE: [Linux-ia64] kacpidpc process takes 100% of CPU

[Paris, DavidX]

Here's the back trace... I hope it's useful:

[0]kdb> cpu 1

Entering kdb (current=0xe00000003f940000, pid 10) on processor 1 due to cpu
switch
[1]kdb> bt
0xe0000000045cce40 acpi_ut_release_mutex+0xc0
        args (0x9, 0xe0000000045d63b0, 0x307)
        kernel .text 0xe000000004400000 0xe0000000045ccd80
0xe0000000045ccee0
0xe0000000045d63b0 acpi_ds_delete_walk_state+0x190
        args (0xe00000000004b140, 0xe00000000004b4b8, 0x30a,
0xe00000000004b4b8, 0xe0000000045f5720)
        kernel .text 0xe000000004400000 0xe0000000045d6220
0xe0000000045d6420
0xe0000000045f5720 acpi_ps_delete_parse_tree+0x160
        args (0xe00000000004b4d8, 0x793, 0xe00000000004b140,
0xe00000000004b4d8, 0xe0000000045f29b0)
        kernel .text 0xe000000004400000 0xe0000000045f55c0
0xe0000000045f5740
0xe0000000045f29b0 acpi_ps_complete_this_op+0x1b0
        args (0xe00000003f96dd90, 0xe00000003f96dd80, 0x0,
0xe0000000045f3410, 0x793)
        kernel .text 0xe000000004400000 0xe0000000045f2800
0xe0000000045f2ba0
0xe0000000045f3410 acpi_ps_parse_loop+0x5f0
        args (0xe000000000049940, 0xe0000000048f7df0, 0xe000000000049bf8,
0x4002, 0x0)
        kernel .text 0xe000000004400000 0xe0000000045f2e20
0xe0000000045f4120
0xe0000000045f4420 acpi_ps_parse_aml+0x300
        args (0xe000000000049940, 0xe00000000004a540, 0x0,
0xe00000003fb3fd30, 0x4003)
        kernel .text 0xe000000004400000 0xe0000000045f4120
0xe0000000045f4440
0xe0000000045f5a70 acpi_psx_execute+0x330
        args (0xe00000003f9649c0, 0x0, 0xe00000003f947de0, 0x309,
0xe000000000049d40)
        kernel .text 0xe000000004400000 0xe0000000045f5740
0xe0000000045f5b40
0xe0000000045eceb0 acpi_ns_execute_control_method+0xb0
[1]more>
        args (0xe00000003f9649c0, 0x0, 0xe00000003f947de0,
0xe000000004a614c0, 0xe0000000045ecdc0)
        kernel .text 0xe000000004400000 0xe0000000045ece00
0xe0000000045ecf00
0xe0000000045ecdc0 acpi_ns_evaluate_by_handle+0x1a0
        args (0xe00000003f9649c0, 0x0, 0x0, 0xe0000000045d75a0, 0x186)
        kernel .text 0xe000000004400000 0xe0000000045ecc20
0xe0000000045ece00
0xe0000000045d75a0 acpi_ev_asynch_execute_gpe_method+0xe0
        args (0x16, 0xe0000000045c8170, 0x205)
        kernel .text 0xe000000004400000 0xe0000000045d74c0
0xe0000000045d7600
0xe0000000045c8170 acpi_os_queue_exec+0xd0
        args (0xe00000003fb5e940, 0xe0000000049cda80, 0xe000000004498a10,
0x30c)        kernel .text 0xe000000004400000 0xe0000000045c80a0
0xe0000000045c81c0
0xe000000004498a10 kernel_thread+0x70
        args (0xe0000000048fafd8, 0xe0000000049cda80, 0x0,
0xe00000003fb88000, 0xe0000000045c8200)
        kernel .text 0xe000000004400000 0xe0000000044989a0
0xe000000004498a60
0xe0000000045c8200 acpi_os_schedule_exec+0x40
        args (0xe00000003fb5e940, 0xe0000000044f6730, 0x205)
        kernel .text 0xe000000004400000 0xe0000000045c81c0
0xe0000000045c8240
0xe0000000044f6730 __run_task_queue+0x1b0
        args (0xe00000003fb8fe10, 0xe0000000049cda80, 0xe00000000450bc80,
0x612)        kernel .text 0xe000000004400000 0xe0000000044f6580
0xe0000000044f6780
0xe00000000450bc80 context_thread+0x2c0
        args (0xe0000000049e79f8, 0xe00000003fb88000, 0xe00000003fb88ef0,
0xe00000003fb88010, 0xe00000003fb88ef0)
        kernel .text 0xe000000004400000 0xe00000000450b9c0
0xe00000000450be00
[1]more>
0xe000000004498a10 kernel_thread+0x70
        args (0xe0000000048fb958, 0xe0000000049cda80, 0x0,
0xe00000003fb38000, 0xe00000000450bf60)
        kernel .text 0xe000000004400000 0xe0000000044989a0
0xe000000004498a60
0xe00000000450bf60 start_context_thread+0x40
        args (0xe000000004929e28, 0xe000000004901610, 0x102)
        kernel .text 0xe000000004400000 0xe00000000450bf20
0xe00000000450bfa0
0xe000000004901610 do_basic_setup+0x50
        args (0xe000000004491450, 0x20a)
        kernel .text.init 0xe000000004900000 0xe0000000049015c0
0xe000000004901640
0xe000000004491450 init+0x50
        args (0x0, 0xe000000004a5d930, 0xe000000004498a10, 0x30c)
        kernel .text 0xe000000004400000 0xe000000004491400
0xe0000000044916e0
0xe000000004498a10 kernel_thread+0x70
        args (0xe0000000048fd978, 0xe0000000049cda80, 0x0,
0xe00000000495c000, 0xe000000004490f30)
        kernel .text 0xe000000004400000 0xe0000000044989a0
0xe000000004498a60
0xe000000004490f30 rest_init+0x30
        args (0xe000000004901420, 0x58e)
        kernel .text 0xe000000004400000 0xe000000004490f00
0xe000000004490fc0
0xe000000004901420 start_kernel+0x400
        args (0x3fb53d08, 0x120e, 0x3fa68770, 0x713, 0x5843)
        kernel .text.init 0xe000000004900000 0xe000000004901020
0xe000000004901520
0xe0000000044903d0 start_ap+0x1b0
        args (0x3fc92000, 0x3fb52008, 0x1, 0x1, 0x3fb53d08)
[1]more>
        kernel .text 0xe000000004400000 0xe000000004490220
0xe0000000044903f0
[1]kdb> go

-----Original Message-----
From: Grover, Andrew 
Sent: Tuesday, November 13, 2001 11:20 AM
To: Paris, DavidX; 'linux-ia64@linuxia64.org'
Subject: RE: [Linux-ia64] kacpidpc process takes 100% of CPU


The ACPI interpreter starts threads to run control methods, and these are
labeled kacpidpc. When nothing is wrong, these are short-lived, and then
terminate.

Can you please use kdb, break in, and get a stack trace to find out where
we're stuck?

Thanks -- Regards -- Andy

> -----Original Message-----
> From: Paris, DavidX [mailto:davidx.paris@intel.com]
> Sent: Tuesday, November 13, 2001 9:44 AM
> To: 'linux-ia64@linuxia64.org'
> Subject: [Linux-ia64] kacpidpc process takes 100% of CPU
> 
> 
> I've just built 2.4.13 and 2.4.14 with and without the KDB 
> patch and in all
> cases when the system comes up there is a kacpidpc process 
> (PID 10) taking
> 100% of one of the cpus (changes which CPU it's on every 
> boot). The system
> is a 4 processor Lion (733 Mhz B3) with 1 gig of ram. This 
> process isn't
> even running when I build the same kernels on a BigSpur. I've 
> even tried
> copying over the very kernel that works fine on the BigSpur, 
> but when I boot
> it on the Lion I get this kacpidpc process. Naturally the 
> process ignores my
> 'kill -9's. I've tried booting the kernel on single proc 
> Lions (733 C0s),
> but the kernel won't even boot on those systems (hangs right 
> after it loads
> the SCSI driver is begins looking for devices). Does anyone 
> have any idea
> why this kacpidpc process is hogging one of the CPUs, and how 
> I go about
> getting rid of it? Unlike in the IA32 kernels, it doesn't 
> look like you can
> turn off ACPI in the IA64 kernel. Thanks,
> 
> David
> 
> _______________________________________________
> Linux-IA64 mailing list
> Linux-IA64@linuxia64.org
> http://lists.linuxia64.org/lists/listinfo/linux-ia64
>