When installing on a Toshiba Satellite Pro 480CDT, the installer dies while inserting the yenta_socket module from the PCMCIA driver disk.
After some further work, I've discovered that yenta_socket blows up the machine if, and only if, usb-ohci has already been added.
Pete, this is with the 2.4.2-0.1.22 kernel that has the irq fix for yenta that fixes some toshiba and Dell laptops, so I am wondering if ohci might be more directly involved.
ohci didn't survive slab-poison AT ALL, so maybe it corrupts something
Trond, a couple of questions: 1. Can you reproduce it without an installer? 2. Can you remind me, what kernel version is there, was it 2.4.1-0.1.9?
1. Yes. I did that by loading/unloading modules. Hence the conclusion about the ohci driver 2. -22 - the qa-tree on 2000-03-07 (yes, the one from the day before as well)
Trond, have you had a chance to try a non-poisoned kernel yet?
No, this is not a laptop I usually have access to. Glen dug it up for me along with a couple of others. Glen, can you find it again?
I ended up with this machine -- it is available for testing as necessary. Stock 7.1 has the same problem. The latest rawhide kernel (2.4.3-7) and the latest rawhide kernel-pcmcia-cs (3.1.24-4) do not help -- same issue. Booting normally gives this error (?) message: PCI: No IRQ known for interrupt pin A of device 00:02.0. Please try using pci=biosirq. PCI: No IRQ known for interrupt pin B of device 00:02.1. Please try using pci=biosirq Yenta IRQ list 06b8, PCI irq0 Socket status: 30000007 <boot locks up at this point> Booting with "nousb pci=biosirq" yields this: PCI: No IRQ known for interrupt pin A of device 00:02.0. PCI: No IRQ known for interrupt pin B of device 00:02.1. Yenta IRQ list 06b8, PCI irq0 Socket status: 30000007 Yenta IRQ list 0eb8, PCI irq0 Socket status 300000011 Which seems to work okay with an older pcnet_cs PCMCIA network card I have, except for the occasional 'spurious interrupt" message that I get while using the card. Something that might be useful is that the load order matters; if you add the ohci module *after* pcmcia is started, everything seems to work fine. I'll leave it running overnight to check -- feel free to grab the machine for a few days if this interests you.
May I have an output of lspci -v, and dmesg of a working machine (where usb-ohci was loaded after the start of pcmcia), please.
forgot to CC: myself on this, sorry about the delay. The problem still exists in beta1. I'll attach the files from beta1.
Created attachment 23247 [details] lspci -v on Satellite Pro 480CDT
Created attachment 23248 [details] dmesg with usb-ohci commented out in /etc/modules.conf
pcmcia boot for an NFS install fails as well (at the "Initializing PCMICA devices" screen, well before you've got a console), same solution.
Please try changing the bios setting regarding "PCMCIA" to "PCIC" or "Cardbus" but not "Auto"
Changing the bios setting to PCIC-compatible gave this message on PCMCIA startup: ds: no socket drivers loaded! and PCMCIA did not come up (even though there was an OK on bootup). Changing the bios setting to "Cardbus/16-bit" ... worked. :) Would someone care to explain what the difference is? Is "Auto-selected" some kind of PnP thing?
May we have the status with 2.4.9-21 ? Also, this does not look like ohci issue anymore. Can I close as "having a workaround in BIOS settings"?
Probably. I don't have access to the machine, and glen/mdrew no longer work here.