Hello, I'm sorry for this very late reply :-( It turned out that I was busier than I considered myself to be :-(
On Thu, Jul 02, 2009 at 08:51:11PM +0200, Samuel Thibault wrote: > Thanks for your investigation, it's clearly useful! I actually enjoy doing this. It has been my dream for a long time to go down to the kernel level, so it's very exciting for me to actually do something to the kernel itself (although it's only adding printk's so far :-) ). > Sergiu Ivanov, le Mon 29 Jun 2009 16:31:19 +0300, a écrit : > > Since ide_probe_promise_20246 prints a message in case it detects what > > it probes for and nothing gets printed to my screen, I guess that Mach > > does not detect my Promise drives anyway. > > The serial number doesn't exactly match, so there is little surprise I > guess. Yep, this is a clear matter. > > ide_probe_for_cmd640x hangs in the call to probe_for_cmd640_pci1. > > Ok. That's odd as there is no infinite loop there :) > > > This latter function calls match_pci_cmd640_device in a loop and what > > my output tells me is that ide_probe_for_cmd640x calls > > probe_for_cmd640_pci1 several times > > Oh? That's not supposed to happen, as the code shows as well as callers > of these. Could you investigate in that direction? No, surely ide_probe_for_cmd640x calls probe_for_cmd640_pci1 only once. I'm sorry :-( Unfourtunately, I'm mostly studying the matter at odd times, which results in attention bugs :-( > > In the first calls match_pci_cmd640_device gets called > > only once per call to probe_for_cmd640_pci1. > > That's odd too, could you check the value returned by > match_pci_cmd640_device and the value of cmd640_key? My replica about match_pci_cmd640_device being called once per call is, obviously, rubbish. Also, I have noticed another ``attention bug'': gnumach hangs in probe_for_cmd640_pci2. cmd640_key in this function goes from 0xC000 up to 0xCC00 and match_pci_cmd640_device hangs there. At all times the i-loop inside the match_pci_cmd640_device function does not go beyond i = 0. The function get_cmd640_reg_pci2 hangs when cmd640_key is 0xCC00 at the line ``b = inb_p(cmd640_key + reg);'' (sorry, I don't give line numbers, because mine are distorted by debug printk's). I very much hope I will not find any big mistakes in this description when I will look at it next time... > > Other from FDC, nothing looks like a hard drive here. > > Indeed. I guess your controler is a SATA one. You need to enable legacy > IDE/ATA emulation in the bios for gnumach to be able to use it. If there > is no such option, you're hosed :/ Well, I don't think my controller is a SATA one :-) The reasoning is simple: Wikipedia tells that SATA was created in 2003, and I bought my old box in 2001 :-) Unfourtunately, I have nothing similar to IDE/ATA emulation in my BIOS :-( Great thanks for your helpfulness! :-) Regards, scolobb
