Bug 8706
Summary: | ACPI errors (possibly) make cpu frequency scaling hangs - Dell i5150 | ||
---|---|---|---|
Product: | ACPI | Reporter: | Federico Flego (flego.federico) |
Component: | ACPICA-Core | Assignee: | Len Brown (lenb) |
Status: | CLOSED PATCH_ALREADY_AVAILABLE | ||
Severity: | high | CC: | acpi-bugzilla, trenn, venki |
Priority: | P1 | ||
Hardware: | All | ||
OS: | Linux | ||
Kernel Version: | 2.6.20 | Subsystem: | |
Regression: | --- | Bisected commit-id: | |
Attachments: |
output of acpidump
acpidump from BIOS A38 dmesg-2.6.22-rc7 dmesg 2.6.20 acpidump bios A38 dmesg 2.6.22_1 acpidump bios A38 kernel 2.6.22.1 kernel 2.6.22.1 config output of dmesg with acpi_debug=on |
Description
Federico Flego
2007-07-02 12:45:58 UTC
Created attachment 11924 [details]
output of acpidump
The acpidump file looks corrupt. Have you e.g. done: acpidump >/tmp/acpidump Then uploaded /tmp/acpidump as attachement? Best is you also set the mime type to text/plain. Subject: Re: ACPI errors (possibly) make cpu frequency scaling hangs up the machine Yes I attached the file acpidump. But I forgot to get rid of lines: > acpidump >/tmp/acpidump > cat /tmp/acpidump which I wrote because I didn't know how to attach a file at the beginning. The file I attached seems readable, if it is not, please tell me. Thank you very much, Federico. bugme-daemon@bugzilla.kernel.org wrote: > http://bugzilla.kernel.org/show_bug.cgi?id=8706 > > > > > > ------- Comment #2 from trenn@suse.de 2007-07-02 14:10 ------- > The acpidump file looks corrupt. > Have you e.g. done: acpidump >/tmp/acpidump > Then uploaded /tmp/acpidump as attachement? > Best is you also set the mime type to text/plain. > > Created attachment 11931 [details]
acpidump from BIOS A38
Created attachment 11932 [details]
dmesg-2.6.22-rc7
please upgrade to the latest BIOS please try disabling HT in the BIOS (as a test) please don't run i8kfan (as a test) if you still have a problem please re-open. Created attachment 11947 [details]
dmesg 2.6.20
Created attachment 11948 [details]
acpidump bios A38
I upgraded to BIOS A38 and got rid of i8kfan. So far no more hangs on my machine :) Still ACPI errors/exceptions appear in dmesg though. I attached whole output of dmesg and acpidump. Thanx, Federico. Doing: acpixtract -a acpidump acpiexec -x 0x1F DSDT.dat gives: Completing Region/Field/Buffer/Package initialization:..dsobject-0472 [11] DsBuildInternalPackage: Package List length larger than NumElements count (2), truncated .dsobject-0472 [11] DsBuildInternalPackage: Package List length larger than NumElements count (D), truncated .dsobject-0472 [11] DsBuildInternalPackage: Package List length larger than NumElements count (1), truncated and a lot more of these... later it's hanging when trying to execute _INI methods. This was tested with acpica AML Execution/Debug Utility version 20060912 Work has been done in this area, see commit: commit 8f9337c88335846b01801b1047a4caf10527a320 Author: Bob Moore <robert.moore@intel.com> Date: Fri Feb 2 19:48:18 2007 +0300 ACPICA: Handle case NumElements > Package length Additional update for NumElements fix. Must handle case where NumElements > Package list length, pad package with null elements. Signed-off-by: Alexey Starikovskiy <alexey.y.starikovskiy@intel.com> Signed-off-by: Len Brown <len.brown@intel.com> Argghh, can someone tell me how I can fetch the -rcX-gitY when this patch got committed? Ok, I manually checked, it seems to not yet be in 2.6.21. I wonder how you can boot at all, better try 2.6.22-rc7. Sorry, I was wrong. On Linus git tree, one could see the patch going in: Follows: v2.6.20-rc7 Precedes: v2.6.21-rc1 The methods are very similar and I looked at the wrong one. This patch is already in 2.6.21. However, the current official acpica version is quite old (09-11-2006, Bob wanted to publish another one, but seems he hasn't come to it before holidays...), it's not worth trying with an old acpica package. Sorry I can't help here, Intel people need to have a look at this. > I upgraded to BIOS A38 and got rid of i8kfan. > So far no more hangs on my machine :) Okay, so the primary issue that this bug is filed against (a hang) is gone. It would be interesting if you re-enable i8kfan and hangs come back. That, of course, would be an i8kfan bug... > Still ACPI errors/exceptions appear in dmesg though. So this is what remains of this bug report. ACPI Error (psparse-0537): Method parse/execution failed [\_PR_.CPU0._PDC] (Node c1903cfc), AE_ALREADY_EXISTS But your dmesg is from 2.6.20. The dmesg from 2.6.22 I posed above has these gone, so it appears this issue has already been addressed, and I believe that this bug should be closed. Re: DsBuildInternalPackage: Package List length larger than NumElements count (D), truncated I'm not sure why this is suspected to be related to the subject of this bug report -- which was apparently the repeated _PDC issue. Also, one would not expect a raw DSDT to necessarily survive acpiexec, as the system hardware that the DSDT talks to is not emulated. But to answer your question, git checkout -b foo 8f9337c88335846b01801b1047a4caf10527a320 would give you a tree with that commit on the top. Alternatively, you can get just that patch with git show 8f9337c88335846b01801b1047a4caf10527a320 and gitk 8f9337c88335846b01801b1047a4caf10527a320 will bring up the GUI which will have links indicating the preceeding and next tags. this one shipped in 2.6.21. closed Thank you all for all your efforts! Actually I had no more hangs and that's very positive. Just to provide an extra feed-back, I tested kernel 2.6.22.1 and still have kernel error messages. I include as attachments the .config, dmesg and acpidump of this last test. Thank's once more, Federico. Created attachment 12039 [details]
dmesg 2.6.22_1
Created attachment 12040 [details]
acpidump bios A38 kernel 2.6.22.1
Created attachment 12041 [details]
kernel 2.6.22.1 config
>ACPI: SSDT 3FFF320B, 066C (r1 DELL CPU0HTCI 1001 MSFT 100000E)
>ACPI Error (dswload-0334): [SMIX] Namespace lookup failure, AE_ALREADY_EXISTS
>ACPI Exception (psloop-0225): AE_ALREADY_EXISTS, During name lookup/catalog
>[20070126]
>ACPI Error (psparse-0537): Method parse/execution failed [\_PR_.CPU0._PDC]
>(Node c1908cfc), AE_ALREADY_EXISTS
>ACPI: Marking method _PDC as Serialized
>ACPI: Processor [CPU0] (supports 8 throttling states)
>ACPI: SSDT 3FFF3877, 00C6 (r1 DELL CPU1HTCI 1001 MSFT 100000E)
>ACPI Error (dswload-0334): [GETC] Namespace lookup failure, AE_ALREADY_EXISTS
>ACPI Exception (psloop-0225): AE_ALREADY_EXISTS, During name lookup/catalog
>[20070126]
>ACPI Error (psparse-0537): Method parse/execution failed [\_PR_.CPU1._PDC]
>(Node c1908cd4), AE_ALREADY_EXISTS
>ACPI: Marking method _PDC as Serialized
This problem is not clear yet.
Could you please boot with acpi.debug_level=0xf00fffff acpi.debug_layer=0xffff3ff0 and attach the dmesg output?
About comment 17: If you increase the debug output the kernel's ring buffer might get filled before everything got logged and log parts might get overridden (dmesg does not show the whole boot logs anymore, the beginning might be missing) You can increase the buffer with the boot param: log_buf_len= (The value is in bytes, e.g. log_buf_len=34359738368 increases the buffer to 32MB). I run kernel 2.6.22 with acpi_debug and other options. When I tried dmesg > dmesg_6_22_1_acpi_debug but I obtained just a truncated version of the whole thing. So I took /var/log/kernel.log and excerpted the last log session. Hope is useful... let me know. Thank you! Federico. Created attachment 12125 [details]
output of dmesg with acpi_debug=on
|