Here's another clue: Loading the thermal and fan modules allows the fan to turn on and operate normally (but only after it's booted and starts loading modules). With thermal and fan compiled in-kernal, there's no lag-time after boot-up and the fan works immediately.
Last night, I checked and the acpi package wasn't loaded. I installed it, and it loads the proper modules and turns on the fan with the stock kernel packages. Unfortunately, there's a lag-time between booting and getting the fan running. If something happens that stalls the system, the fan is off and the system can still overheat.
I can't override this behavior with the BIOS settings. My system has a setting for "fan always on," but once a kernel with ACPI support loads, it simply overrides this and turns the fan off somehow.
Will the output with my custom kernel be okay? If thermal and fan don't have to be modules, I can do this after the system has been running for a while, without having to worry about overheating.... Otherwise, if I have to use the stock kernels, I'll have to let the system cool off a bit first, just as a precaution.
Francesco P. Lovergine wrote:
severity 259481 normal thanks On Wed, Jul 14, 2004 at 05:04:14PM -0700, Daniel Burton wrote:Package: kernel-image-2.6.7-1-686 Version: 2.6.7-2 Severity: important File: kernel-image-2.6.7As soon as I boot up and ACPI loads, the fan on my laptop turns off, allowing it to overheat. This is a Compaq Evo N620c. It also does this in kernel-image-2.6.6-1-686. I can't produce a boot log, because the overheating happens so fast that it literally won't boot up again afterwards (the automatic safety mechanisms kick in and shut it down), and I don't want to damage the hardware.I'll try to see if I can change the BIOS settings to override the the fan state and get a boot log.Unfortunately there are tons of laptops with buggy bios out there.Compaq ones do not appear particularly brilliant on this regards. First of all, load acpi package and run acpi -t, this could activate the fan. Unfortunately it should probably stay onforever in this manner. Please provide a DSDT dump of your boxcat /proc/acpi/dsdt > dsdt.aml iasl -d dsdt.aml > dsdt.aslThat's the possibility that it needs hacking to work.