I'm having some issues with getting a Rosewill RC-301 (Moschip NM9835-based) 2-port PCI serial card working under squeeze (2.6.32-5-sparc64; stock kernel) on a Netra T1-105 and would very much appreciate assistance getting it up and running. Here's the full gamut of what I've gone through with it:
Installed the card in the box. The machine would not boot with it installed - basically, it would pass LOM then drop to OpenFirmware; issuing any form of boot command would simply cause it to hang. This behaviour continued until `setenv pcia-probe-list 0` was set in OpenFirmware. Note that I can see where this might be problematic further down the line, but, moving on:
The machine now boots as expected. pci_stub, 8250, and 8250_pci were added to /etc/modules and are loading; they can also be manually removed from and inserted into the kernel without error.
Entries are automatically created in /dev for ttyS2, ttyS3, ttyS4, and ttyS5 at boot (ttyS0 and ttyS1 are also present, but are the onboard serial ports). This is the first odd thing that I've noticed: despite being a two-port card, four port devices are defined. Doing a `setserial -g /dev/ttyS*` gives the following:
It's not clear to me as to why all port addresses are set to 0x0000, nor why ttyS2 through ttyS5 have no IRQ but appear to have been detected by the system. Attempting to configure ports other than ttyS0 or ttyS1 with either stty or setserial (predictably) fails.
As you can see, the serial card does not appear in the list of installed devices. At this point, my best guess is that by disabling probing of the PCI port in OpenFirmware I've made the OS unable to detect its base address - but if I re-enable it, boot hangs. Complete Catch-22.
Note that it's been several years since I've had to work with this era of Sun hardware, so it's entirely possible that I've missed out something that needs to be set in OpenFirmware in order to make it all work. However, I'm really not sure where the problem might be at this point.