Most recent kernel where this bug did not occur: 2.6.7 Distribution: Debian unstable Hardware Environment: Pentium 4, SiS chipset (Asus a2500h notebook) Software Environment: bash, Icewm window manager Problem Description: Icewm's battery monitor and "acpi -V" can't report the battery status. The kernel logs the following 3 lines every time someone tries to read the battery status: ACPI-0508: *** Error: Method execution failed [\ECWR] (Node c145ecc0),AE_AML_OPERAND_TYPE ACPI-0508: *** Error: Method execution failed [\RBAT] (Node c145ec80), AE_AML_OPERAND_TYPE ACPI-0508: *** Error: Method execution failed [\_SB_.BAT0._BST] (Node c145da60), AE_AML_OPERAND_TYPE or even: ACPI-0508: *** Error: Method execution failed [\WEIE] (Node deba7ce0), AE_AML_OPERAND_TYPE ACPI-0508: *** Error: Method execution failed [\ECWR] (Node deba7c80), AE_AML_OPERAND_TYPE ACPI-0508: *** Error: Method execution failed [\RBAT] (Node deba7c40), AE_AML_OPERAND_TYPE ACPI-0508: *** Error: Method execution failed [\_SB_.BAT0._BST] (Node deba6a20), AE_AML_OPERAND_TYPE Steps to reproduce: The problem happened twice in 3 days, out of the blue, while the computer was on AC power. The first time I killed X, restarted it, and had the whole system freeze (had to shutdown using hardware switch). The second time I killed X and managed to reboot the system properly (and the problem didn't show up at reboot).
Created attachment 6715 [details] acpidump output Here is the output of acpidump, when the system is working properly (I'll attach a dump from a "sick" system as soon as the problem shows up again)
Created attachment 6716 [details] lspci and .config Here you are the output of lspci -vv and the .config for the 2.6.14 kernel
Hopefully, you can find a way to consistently reproduce this problem. There is a potentially very serious bug in the ASL code for this machine, the vendor should be notified: Method (GPLV, 0, NotSerialized) { Acquire (MTXE, 0xFFFF) Store (0x59, CMOI) Store (0x00, CMOP) Store (0x02, CAPC) ISMI (0x49) Return (CMOD) Release (MTXE) } dsdt.dsl 1570: Release (MTXE) Warning 2097 - Statement is unreachable ^ The global mutex is never released by this method.
Created attachment 6776 [details] acpidump output, after the event Here you are the output of acpidump, right after the problem happened
acpidump just dumps the ACPI tables for the machine, these don't change at runtime. We need an idea of the sequence of events that results in the failure.
Some more information (I hope it helps): - The problem happens also with kernel 2.6.15-rc5 - The problem seems to happen everytime I reach an uptime of about 6 hours; I did some analysis on my syslog: 20060112 problem happened at 20:38:29 (boot at 14:32:46) 20060110 problem happened at 15:35:46 (boot at 09:24:04) 20060109 problem happened at 15:35:32 (boot at 09:16:32) 20060105 problem happened at 16:26:16 (boot at 10:22:02) 20040105 problem happened at 18:27:42 (boot at 12:17:59)
Please try the latest stable kernel version (2.6.17.1 is available now) and the latest mm version (2.6.17-mm1 is available now). It will give us more information.
I compiled 2.6.17.1 and this time I get the following errors on the kernel log: ACPI Warning (utdelete-0378): Large Reference Count (401) in object c145d6b4 [20060127] ACPI Warning (utdelete-0378): Large Reference Count (401) in object c145d6b4 [20060127] ACPI Warning (utdelete-0378): Large Reference Count (402) in object c145d6b4 [20060127] ACPI Warning (utdelete-0378): Large Reference Count (401) in object c145d6b4 [20060127] ACPI Warning (utdelete-0378): Large Reference Count (401) in object c145d6b4 [20060127] ACPI Warning (utdelete-0378): Large Reference Count (402) in object c145d6b4 [20060127] ... ACPI Warning (utdelete-0378): Large Reference Count (406) in object c145d6b4 [20060127] ACPI Warning (utdelete-0378): Large Reference Count (407) in object c145d6b4 [20060127] ACPI Warning (utdelete-0378): Large Reference Count (408) in object c145d6b4 [20060127] ACPI Warning (utdelete-0378): Large Reference Count (407) in object c145d6b4 [20060127] ACPI Warning (utdelete-0378): Large Reference Count (406) in object c145d6b4 [20060127] ACPI Warning (utdelete-0378): Large Reference Count (407) in object c145d6b4 [20060127] ... and so on Some observations: - The annoying messages start to appear after about 80 times that the battery status is checked (using "acpi -V" or my window manager's facility) - I get about 40 of the above messages every time the battery status is checked - If I don't check the battery, nothing bad happens (at least for a couple of hours, I haven't (yet) done further tests) Please let me know if I can be of any help
And what about 2.6.17-mm* tree (2.6.17-mm3 is available now)?
Great. With 2.6.17-mm3 all the errors disappeared: both the "ACPI error" I reported in my first message and the latest "ACPI warning". The downside of it (although I think it's unrelated :) is that with 2.6.17-mm3 the keyboard lacks any autorepeat functionality. There is no way to set it with kbdrate; this happens both in X and in the console, and passing "atkbd_softrepeat=1" at boot doesn't have any effect. I searched bugzilla about it but I couldn't find any open bugs, maybe this is due to a configuration error on my side? I used the same .config I used for 2.6.17.1...
I have the same problem, but I think it is not ACPI bug; some time is needed for investigation. Maybe next kernel mm version will resolve this new problem.
Try mm4, the problem is resolved - at least on my laptop.
Do you have any problems on latest kernel versions, or bug can be closed? Thanks. Vladimir.
Ok, I tested 2.6.17.4 and I got the same behaviour as with 2.6.17.1: ACPI Warning (utdelete-0378): Large Reference Count (401) in object c145d6b4 [20060127] Then I tested 2.6.18-rc1 and everything was just fine, no more errors even after 8hrs+ uptime (and no more keyboard errors)