https://bugzilla.kernel.org/show_bug.cgi?id=15946
--- Comment #107 from Xavier Hourcade <[email protected]> 2013-04-16 14:57:46 --- (In reply to comment #104) > ping ... Thank you Lan for your patch, which I will try for sure then report here, but not before a few weeks from now. It seems to me there are several, distinct issues. Here is a paste of my notes to date, if this may help as yet? How to reproduce, typically: - dual-head, allocating one CPU core to a VM guest - closing/re-opening the Fatest-Firefox-profile-ever™ several times - if needed, start eg. building a kernel and/or heavy LUKS write/swap. Hardware observations - cooling system is common to CPU/GPU on this hardware - nouveau kernel module gets the GPU to heat more, and brings many more failures than nvidia driver (by far). - the higher ambient temperature, the more occurrences - at ambient temperature of 16ºC or under, issue seems to entirely disappear, no matter the system usage - at any time and under any OS, maximum "true" temperatures seen were 86/75ºC (CPU/GPU), which matches specifications. Shutdown observations - may occur at grub screen - may occur under Vista's boot-time "memory test" software - may occur from GNU/Linux pre-boot or at any later stage - may occur disregarding to the "thermal.nocrt" parameter, - may follow bogus value report from kernel itself - may follow bogus value report from monitor script, reading from /sys during several seconds, sometimes in the absence of kernel messaging - may occur while any of these reports are printed to screen, but before they could be written to disk (via logger) - may follow none of these reports at all - most often occurs in series, same day - series often start while CPU is under heavy load - pm-suspend clearly extends "good mood" by several weeks - when EC is "in a good moon", overall fan quieter from boot Applies to - all Fedora stock kernels at least until 3.6.11 (later kernels not yet tested on this hardware) Does NOT seem to apply to (only a few test so far) - older Debian kernels e.g. 2.6 (Squeeze) or even 3.4 (Wheezy) - Vista "normal" session Hence there might be several distinct issues (?) - a built-in EC defect (soft-fixed by Asus/Vista ?) - kernel /sys reading clearly not consistent - kernel messaging absent while monitor script succeeds in reporting and is even logging to disk over several seconds (via logger) - kernel ignoring thermal.crt and shutting down nevertheless (?) OS Since Linux Distro Video Occurrences -------------------------------------------------------- 4 : 2013-01-16 2.6.32 5.deb6 nouveau 0 (3 cold boots) 5 : 2013-01-16 3.2.35 .deb7 nouveau 0 (2 cold boots) 1 : 2012-07-28 3.4.6 1.fc16 nvidia 0 1 : 2012-08-07 3.4.7 1.fc16 nvidia 2 (over 2 days) 1 : 2012-08-19 3.4.9 1.fc16 nvidia 2 (over 2 days) 1 : 2012-08-31 3.4.9 2.fc16 nvidia 15 (over 7 days) 1 : 2012-09-23 3.4.11 1.fc16 nvidia 2 (over 2 days) 2 : 2012-08-12 3.5.1 1.fc17 nouveau Dies within secs 3 : 2012-09-05 3.6.0 x.fc18 nouveau 8 (20+ cold boots) 1 : 2012-10-24 3.6.2 1.fc16 nvidia 0 (cold boot) 1 : 2012-11-06 3.6.5 2.fc16 nvidia 0 (warm boot) 1 : 2012-11-09 3.6.6 1.fc16 nvidia 0 (warm boot) 1 : 2012-11-10 3.6.6 1.fc16 nouveau 1 (warm boot) 1 : 2012-11-29 3.6.7 4.fc16 nouveau 9 (over 3 days) 1 : 2012-12-18 3.6.7 4.fc16 nvidia 2 (over 1 day) 1 : 2012-12-18 3.6.10 2.fc16 nvidia 13 (over 7 days) 3 : 2013-01-05 3.6.10 4.fc18 nouveau 2 (3rd cold boot) 1 : 2013-01-05 3.6.11 1.fc16 nvidia 12 (over 6 days) 1 : 2013-01-28 3.6.11 4.fc16 nvidia 9 (over 7 days) OS notes 1 : Fedora 16 stable HD install, intensive use 365d/y! 3.4.6 to 3.4.11: hotest ambient temperature 3.4.9: more cold boots (between OS 2&3 tests) 3.6.2 to 3.6.5 : did not try nouveau 3.6.2 to 3.6.6 : 16ºC - ambient temperature 3.6.7 to 3.6.11: warmer ambient temperature 2 : Fedora 17 stable HD install, for test, light usage 3 : Fedora 18 Live, just opening many apps at once 4 : Squeeze 6.0.6 live, heavy testing, no issues (building kernel and other tasks all at once) 5 : Wheezy 7.b3 live, heavy testing, no issues Other Sometimes (but not always) immediately after installing a new kernel with yum, or running plymouth-theme -R, firmware gets upset - system becomes "clunky" (half of keystokes would get lost while typing) while messages log is filled up with the following pattern : Dec 18 14:45:33 hostname kernel: [50913.287022] ACPI Exception: AE_TIME, Returned by Handler for [EmbeddedControl] (20120711/evregion-501) Dec 18 14:45:33 hostname kernel: [50913.287040] ACPI Error: Method parse/execution failed [\_SB_.PCI0.SBRG.EC0_.RRAM] (Node ffff880131dccc58), AE_TIME (20120711/psparse-536) Dec 18 14:45:33 hostname kernel: [50913.287054] ACPI Error: Method parse/execution failed [\ODTS] (Node ffff880131de4f00), AE_TIME (20120711/psparse-536) Dec 18 14:45:33 hostname kernel: [50913.287062] ACPI Error: Method parse/execution failed [\_GPE._L02] (Node ffff880131db8280), AE_TIME (20120711/psparse-536) Dec 18 14:45:33 hostname kernel: [50913.287073] ACPI Exception: AE_TIME, while evaluating GPE method [_L02] (20120711/evgpe-560) Dec 18 14:45:34 hostname kernel: [50914.498033] ACPI: EC: input buffer is not empty, aborting transaction Dec 18 14:45:34 hostname kernel: [50914.498041] ACPI Exception: AE_TIME, Returned by Handler for [EmbeddedControl] (20120711/evregion-501) -- Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug. ------------------------------------------------------------------------------ Precog is a next-generation analytics platform capable of advanced analytics on semi-structured data. The platform includes APIs for building apps and a phenomenal toolset for data science. Developers can use our toolset for easy data analysis & visualization. Get a free account! http://www2.precog.com/precogplatform/slashdotnewsletter _______________________________________________ acpi-bugzilla mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/acpi-bugzilla
