On Tue, 2005-04-12 11:13:33 +0200, Peter 'p2' De Schrijver <p2@mind.be> wrote in message <[🔎] 20050412091333.GK16164@mind.be>: > It looks to me as if the soft reboot didn't properly shutdown PCI. > Looking at the cia init code > (arch/alpha/kernel/core_cia.c, verify_tb_operation) it looks as if the 5th > PCI TLB test somehow crashed the machine. Well, I only wrote it to nnnnn@bugs.d.o, but here it's again: ,------ | I just noticed this bug report on some debian mailing list and maybe I | can give a hint here. Some time ago, I did the srm_env.ko driver and | while I worked on the SRM stuff, I noticed that 2.4.x and 2.6.x have | subtle issues with SRM and PCI: | | PCI devices need to be configured. SRM will do that entirely on | power-up, but after a warm reboot, it heavily depends on the exact SRM | version what it'll to. Some early versions (eg. what you'll find on a | NoName (aka AXPpci33) board) won't reconfigure the PCI devices at all. | So if you tell the Linux kernel to soft-reboot, SRM will basically crash | because it cannot any longer access the SCSI devices etc. I guess that | this is an issue here, too. | | In 2.2.x time, Linux kernels recorded any changes done to the PCI config | space and reverted them before they dropped back to SRM. This is missing | these days and I guess this is what's causing trouble here. `------ So, yes, the current kernels are just missing to revert all changes done to the PCI config space... MfG, JBG -- Jan-Benedict Glaw jbglaw@lug-owl.de . +49-172-7608481 _ O _ "Eine Freie Meinung in einem Freien Kopf | Gegen Zensur | Gegen Krieg _ _ O fuer einen Freien Staat voll Freier Bürger" | im Internet! | im Irak! O O O ret = do_actions((curr | FREE_SPEECH) & ~(NEW_COPYRIGHT_LAW | DRM | TCPA));
Attachment:
signature.asc
Description: Digital signature