Hi Isaku,
Thanks for tesing. Please see my comments below.
Isaku Yamahata wrote:
> Hi Yu. I tested it and have some comments.
>
> - When I run xenpm, xen panics
> It panics at 0xf40000000406ba91 =
> xen-unstable.hg/xen/drivers/acpi/pmstat.c:98 (addr2line told)
> It looks like pxpt->u.pt[pxpt->u.cur] wasn't allocated yet resulting
> in NULL pointer reference.
As Guanquan replied, this is a separate issue of xenpm logic, he will post
another patch to fix this issue.
>
> - From the boot message, ondemand governor fails to load.
> I'm not sure this error is expected because of my hardware.
> I guess this error case haven't been tested and that it caused
> the above panic.
>From the boot message, the BIOS _PSS information is totally messed up. E.g.
>1048575us latency is wrong which cause ondemand driver fail to load, also
>1048575MHz freq is obviously not correct. There are several possible reasons
>of this:
- BIOS itself is not correct
- Dom0 ACPICA parsing logic is not correct.
If the native linux (e.g RHEL5) works, then I tend to believe this is dut to
the second reason. The dom0 kernel is bit old and bunch of ACPICA fixes in
mainline kernel are not included in dom0. In my case, I also find the _PSS
parsing result in my Hitachi itanium 2 box is not correct. This issue is fixed
after I pulling two patch from mainline kernel (changeset 677 and 678 in
http://xenbits.xensource.com/linux-2.6.18-xen.hg)
Anyway, this is a separate issue. we will try to find other ia64 box to see if
this issue can be reproduced and then fixed.
BTW, is it possible to check in these patch first?
Best Regards
Ke
>
> - I added work around to avoid the above panic, but
> xenpm needs the following patch to get the following result.
> NOTE: hyperthreading is enabled.
>
>
> ---
> panic log
>
> (XEN) *** xen_handle_domain_access: exception table lookup failed,
> iip=0xf40000000406ba90, addr=0x8, spinning... (XEN) $$$$$ PANIC in
> domain 0 (k6=0xf0000002f6480000): *** xen_handle_domain_access:
> exception table lookup failed, iip=0xf40000000406ba90, addr=0x8,
> spinning... (XEN) d 0xf000000007fec100 domid 0 (XEN) vcpu
> 0xf0000002f6480000 vcpu 1 (XEN) (XEN) CPU 1
> (XEN) psr : 0000121008226038 ifs : 8000000000000691 ip :
> [<f40000000406ba91>] (XEN) ip is at do_get_pm_info+0x331/0x540
> (XEN) unat: 0000000000000000 pfs : 0000000000000691 rsc :
> 0000000000000003 (XEN) rnat: 0000000000000000 bsps: 0000000000000000
> pr : 000000000555aa99 (XEN) ldrs: 0000000000000000 ccv :
> 0000000000000000 fpsr: 0009804c0270033f (XEN) csd : 0000000000000000
> ssd : 0000000000000000 (XEN) b0 : f40000000406ba30 b6 :
> f400000004049cf0 b7 : f400000004002e30 (XEN) f6 :
> 1003e00037c21a350d836 f7 : 1003e0000000281be72cb (XEN) f8 :
> 000000000000000000000 f9 : 000000000000000000000 (XEN) f10 :
> 000000000000000000000 f11 : 000000000000000000000 (XEN) r1 :
> f4000000043e7df0 r2 : f0000002f6487d20 r3 : f0000002dce13cc1 (XEN)
> r8 : 00000032595ac662 r9 : 0000000000000003 r10 : 0000000000000000
> (XEN) r11 : 0000000005550559 r12 : f0000002f6487d20 r13 :
> f0000002f6480000 (XEN) r14 : 0000000000000003 r15 : 0000000000000008
> r16 : 00000032595ac662 (XEN) r17 : 0000000000000000 r18 :
> f000000004163430 r19 : f000000004163898 (XEN) r20 : f000000004163888
> r21 : f0000002f6487d98 r22 : 6000000000004060 (XEN) r23 :
> 2000000000043098 r24 : 00000000001c0600 r25 : 0000000000000003 (XEN)
> r26 : a000000100ab2ab8 r27 : 0000000000000000 r28 : 0000000000000000
> (XEN) r29 : 0000000000000004 r30 : 0000000000000000 r31 :
> f400000004190268 (XEN) (XEN) Call Trace: (XEN) [<f4000000040eb400>]
> show_stack+0x90/0xb0 (XEN)
> sp=f0000002f64877d0 bsp=f0000002f6481980 (XEN) [<f4000000040ebec0>]
> show_registers+0xaa0/0xac0 (XEN)
> sp=f0000002f64879a0 bsp=f0000002f6481940 (XEN) [<f4000000040b1b60>]
> panic_domain+0x120/0x170 (XEN)
> sp=f0000002f64879a0 bsp=f0000002f64818d0 (XEN) [<f4000000040a2930>]
> ia64_do_page_fault+0x710/0x720 (XEN)
> sp=f0000002f6487ae0 bsp=f0000002f6481838 (XEN) [<f4000000040e28c0>]
> ia64_leave_kernel+0x0/0x300 (XEN)
> sp=f0000002f6487b20 bsp=f0000002f6481838 (XEN) [<f40000000406ba90>]
> do_get_pm_info+0x330/0x540 (XEN)
> sp=f0000002f6487d20 bsp=f0000002f64817a8 (XEN) [<f400000004049d00>]
> do_sysctl+0x810/0x840 (XEN)
> sp=f0000002f6487d20 bsp=f0000002f6481730 (XEN) [<f400000004002e60>]
> fast_hypercall+0x170/0x310 (XEN)
> sp=f0000002f6487e00 bsp=f0000002f6481730
>
>
> ---
> cited from boot log
>
> (XEN) Set CPU acpi_id(3) cpuid(1) Px State info:
> (XEN) _PSS:
> (XEN) State0: 1048575MHz 1048575mW 1048575us 1048575us 0xfffff
> 0xfffff (XEN) State1: 1048575MHz 1048575mW 1048575us 1048575us
> 0xfffff 0xfffff (XEN) State2: 1048575MHz 1048575mW 1048575us
> 1048575us 0xfffff 0xfffff (XEN) _PSD: num_entries=5 rev=0 domain=3
> coord_type=252 num_processors=1 (XEN) Current freq of CPU 1 is
> 1048575000 (XEN) CPU 1 initialization completed
> (XEN) ondemand governor failed to load due to too long transition
> latency latency 0x1048575000 limit 0x10000000 (XEN) Set CPU
> acpi_id(11) cpuid(5) Px State info: (XEN) _PSS:
> (XEN) State0: 1048575MHz 1048575mW 1048575us 1048575us 0xfffff
> 0xfffff (XEN) State1: 1048575MHz 1048575mW 1048575us 1048575us
> 0xfffff 0xfffff (XEN) State2: 1048575MHz 1048575mW 1048575us
> 1048575us 0xfffff 0xfffff (XEN) _PSD: num_entries=5 rev=0 domain=11
> coord_type=252 num_processors=1 (XEN) Current freq of CPU 5 is
> 1048575000 (XEN) CPU 5 initialization completed
> (XEN) ondemand governor failed to load due to too long transition
> latency latency 0x1048575000 limit 0x10000000 (XEN) Set CPU
> acpi_id(7) cpuid(3) Px State info: (XEN) _PSS:
> (XEN) State0: 1048575MHz 1048575mW 1048575us 1048575us 0xfffff
> 0xfffff (XEN) State1: 1048575MHz 1048575mW 1048575us 1048575us
> 0xfffff 0xfffff (XEN) State2: 1048575MHz 1048575mW 1048575us
> 1048575us 0xfffff 0xfffff (XEN) _PSD: num_entries=5 rev=0 domain=7
> coord_type=252 num_processors=1 (XEN) Current freq of CPU 3 is
> 1048575000 (XEN) CPU 3 initialization completed
> (XEN) ondemand governor failed to load due to too long transition
> latency latency 0x1048575000 limit 0x10000000 (XEN) Set CPU
> acpi_id(15) cpuid(7) Px State info: (XEN) _PSS:
> (XEN) State0: 1048575MHz 1048575mW 1048575us 1048575us 0xfffff
> 0xfffff (XEN) State1: 1048575MHz 1048575mW 1048575us 1048575us
> 0xfffff 0xfffff (XEN) State2: 1048575MHz 1048575mW 1048575us
> 1048575us 0xfffff 0xfffff (XEN) _PSD: num_entries=5 rev=0 domain=15
> coord_type=252 num_processors=1 (XEN) Current freq of CPU 7 is
> 1048575000 (XEN) CPU 7 initialization completed
> (XEN) ondemand governor failed to load due to too long transition
> latency latency 0x1048575000 limit 0x10000000
>
>
> ---
> patch to xenpm
>
> diff -r ccb87aa7c243 tools/misc/xenpm.c
> --- a/tools/misc/xenpm.c Fri Oct 03 13:12:42 2008 +0900
> +++ b/tools/misc/xenpm.c Tue Oct 07 18:54:42 2008 +0900
> @@ -85,7 +85,7 @@
> else
> fprintf(stderr, "failed to get max C-state\n");
>
> - break;
> + continue;
> }
>
> cxstat->triggers = malloc(max_cx_num * sizeof(uint64_t));
> @@ -108,7 +108,7 @@
> fprintf(stderr, "failed to get C-states statistics
> information\n"); free(cxstat->triggers);
> free(cxstat->residencies);
> - break;
> + continue;
> }
>
> printf("cpu id : %d\n", i);
> @@ -144,7 +144,7 @@
> else
> fprintf(stderr, "failed to get max P-state\n");
>
> - break;
> + continue;
> }
>
> pxstat->trans_pt = malloc(max_px_num * max_px_num *
> sizeof(uint64_t)); @@ -161,12 +161,12 @@
> break;
> }
>
> - ret = xc_pm_get_pxstat(xc_fd, 0, pxstat);
> + ret = xc_pm_get_pxstat(xc_fd, i, pxstat);
> if( ret ) {
> fprintf(stderr, "failed to get P-states statistics
> information\n"); free(pxstat->trans_pt);
> free(pxstat->pt);
> - break;
> + continue;
> }
>
> printf("cpu id : %d\n", i);
>
>
> ---
> result of xenpm
>
> # ./xenpm
> Xen cpuidle is not enabled!
> Xen cpuidle is not enabled!
> Xen cpuidle is not enabled!
> Xen cpuidle is not enabled!
> Xen cpuidle is not enabled!
> Xen cpuidle is not enabled!
> Xen cpuidle is not enabled!
> Xen cpuidle is not enabled!
> failed to get max P-state
> cpu id : 1
> total P-states : 3
> usable P-states : 3
> current frequency : 1048575 MHz
> *P0 : freq [1048575 MHz]
> transition [00000000000000000000]
> residency [00000000000000004361 ms]
> P1 : freq [1048575 MHz]
> transition [00000000000000000000]
> residency [00000000000000000000 ms]
> P2 : freq [1048575 MHz]
> transition [00000000000000000000]
> residency [00000000000000000000 ms]
>
> failed to get max P-state
> cpu id : 3
> total P-states : 3
> usable P-states : 3
> current frequency : 1048575 MHz
> *P0 : freq [1048575 MHz]
> transition [00000000000000000000]
> residency [00000000000000004781 ms]
> P1 : freq [1048575 MHz]
> transition [00000000000000000000]
> residency [00000000000000000000 ms]
> P2 : freq [1048575 MHz]
> transition [00000000000000000000]
> residency [00000000000000000000 ms]
>
> failed to get max P-state
> cpu id : 5
> total P-states : 3
> usable P-states : 3
> current frequency : 1048575 MHz
> *P0 : freq [1048575 MHz]
> transition [00000000000000000000]
> residency [00000000000000004507 ms]
> P1 : freq [1048575 MHz]
> transition [00000000000000000000]
> residency [00000000000000000000 ms]
> P2 : freq [1048575 MHz]
> transition [00000000000000000000]
> residency [00000000000000000000 ms]
>
> failed to get max P-state
> cpu id : 7
> total P-states : 3
> usable P-states : 3
> current frequency : 1048575 MHz
> *P0 : freq [1048575 MHz]
> transition [00000000000000000000]
> residency [00000000000000005799 ms]
> P1 : freq [1048575 MHz]
> transition [00000000000000000000]
> residency [00000000000000000000 ms]
> P2 : freq [1048575 MHz]
> transition [00000000000000000000]
> residency [00000000000000000000 ms]
>
>
>
>
> On Fri, Oct 03, 2008 at 02:16:54PM +0800, Yu, Ke wrote:
>> Hi Isaku,
>>
>> I tried your updated patch, and xenpm works in my IPF box. the
>> attached is the xenpm output.
>>
>>> From you xenpm output, one possible reason is that the DBS (alias
>>> of Px) feature is not enabled in BIOS, thus xen can not get enough
>>> Px information to initialize cpufreq driver. You can double confirm
>>> by:
>> - check BIOS to see if DBS is enabled.
>> - or try the native linux (e.g. RHEL5), to see if the cpufreq works.
>>
>> Also, it will be helpful if you can attach the xen serial log.
>> please add "loglvl=info" to elilo xen option, to make sure cpufreq
>> information is printed.
>>
>> Best Regards
>> Ke
>>
>> Isaku Yamahata wrote:
>>> Hi, Yu.
>>> Thnaks for updating the patches while you're off.
>>>
>>> I applied your patch with some style fixes and tried xenpm.
>>> (I attached the fixed patch. Please confirm I didn't breake them.)
>>> Then I got the followings. (NOTE: I added cpufreq=xen to xen boot
>>> option) Doesn't xenpm work on ia64? Or is there another way to test
>>> them?
>>>
>>> # xenpm
>>> Xen cpuidle is not enabled!
>>> failed to get max P-state
>>>
>>> thanks,
>>>
>>> On Sat, Sep 27, 2008 at 10:11:29AM +0800, Yu, Ke wrote:
>>>> This patchset add cpufreq support for ia64 platform. The logic is
>>>> borrowed from linux kernel and ported to xen. The common cpufreq
>>>> infrastructure is already in xen and works in x86 side, so this
>>>> patchset reuse the common infrastructure and add ia64 specific
>>>> driver and cpufreq notify hypercall.
>>>>
>>>> Hypervisor side:
>>>> 1. px-xen-ipf-driver: add cpufreq ia64 driver
>>>> 2 .px-xen-ipf-hypercall: implement ia64 cpufreq notify hypercall
>>>>
>>>> Dom0 side:
>>>> 1. px-dom0-add-ia64-hypercall.patch: add ia64 cpufreq notify
>>>> hypercall
>>>>
>>>> Signed-off-by: Liu Jinsong <jinsong.liu@xxxxxxxxx>
>>>> Yu Ke <ke.yu@xxxxxxxxx>
>>>>
>>>> BTW, this patchset depends on the xen-unstable tree
>>>> (http://xenbits.xensource.com/staging/xen-unstable.hg) cset 18550,
>>>> 18551, 18552, and linux xen tree
>>>> (http://xenbits.xensource.com/linux-2.6.18-xen.hg) cset 681, which
>>>> restructure the common cpufreq code for ia64. so ia64 tree need to
>>>> sync before applying this patchset.
>>>>
>>>>
>>>> _______________________________________________
>>>> Xen-ia64-devel mailing list
>>>> Xen-ia64-devel@xxxxxxxxxxxxxxxxxxx
>>>> http://lists.xensource.com/xen-ia64-devel
>>
>
>
>> _______________________________________________
>> Xen-devel mailing list
>> Xen-devel@xxxxxxxxxxxxxxxxxxx
>> http://lists.xensource.com/xen-devel
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel
|
Home