Has anyone every used the sym53c8xx driver in an 8xx system? I seem to be having several problems - if anyone can offer any insight perhaps I won't have to re-invent the wheel - though I'm guessing I'll have a good chunk of work to do anyway.
The first problem seems to be with the way the driver assumes that the PCI chipset offers access to the (L2 I'm guessing) cache. My take on this is that the driver stores "SCRIPTS" in main memory which the controller directly accesses and executes. The only problem is that the scripts are actually cached so when the SCSI controller reads the main memory (through the PCI bridge - in my case QSPAN II) it doesn't see the data that it should. I tested this by adding a flush_dcache_range to the driver before the SCSI chip accessed that memory - after doing so it read the right data. So is it true that in a desktop system the PCI chipset would normally be smart enought to know what is cached and to fetch data from the cache if appropriate? Do I have to remove this assumption from the driver for 8xx? My second problem is with endianness. The PCI bridge is doing byte reversing, so in the driver I've made sure that I/O functions are not byte reversed. That works fine - registers on the SCSI chip get the correct values - except when the value that a register is getting is an address - for instance into main memory. It appears that, for instance the dsp register on the sym53c8xx controllers is stored big endian though the rest of the chip is little endian. So although the driver uses the same I/O macros for acceses to the controllers registers - regardless of whether it a regular data register or a register containing an address into PPC memory - it appears to me that it should be doing something different for these two cases. If anyone is familiar with this chip (particulary the 53c895A) and can offer insight into this I would appreciate it. Thanks! Steve -- ------------------------------------------------------- Steven K. Rossi srossi at ccrl.mot.com Staff Engineer Multimedia Communications Research Laboratory Motorola Labs ------------------------------------------------------- ** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/
