Skip to contentSkip to conversation

changes to drivers/scsi/sym53c8xx_defs.h in the mips tree

DISCUSSION Christoph Hellwig 2 participants 3 messages Jan 29, 2005 – Jan 30, 2005 MIPS

lore
Share & tools
lore
#0 lore raw

Hi Carsten,

do you remember what your changes to drivers/scsi/sym53c8xx_defs.h in the mips tree are supposed to help with?

sym53c8xx_defs.h is only used for the ncr53c8xx driver, which despite the name is only for NCR53c720 chips these days (supported on parisc, and eisa/mca). It has correct big endian support for parisc alrady.

Also the patch duplicates your changelog and copyright entry that's in the file already.

#1 lore raw
Quoted 1 line · Christoph Hellwig

Hi Carsten,

I'm afraid Carsten may no longer be reachable, but I may try to help in
his stead.

Quoted 2 lines

do you remember what your changes to drivers/scsi/sym53c8xx_defs.h in the mips tree are supposed to help with?

It must have been the MIPS Technologies' Atlas development board which
has a SYM53C810A incorporated, even though the resulting changes are specific to the board. There appears to have been some sort of confusion about the semantics of i/o and mmio operations in mixed-endian configurations, which for the MIPS port actually is currently being resolved.

Quoted 3 lines

sym53c8xx_defs.h is only used for the ncr53c8xx driver, which despite the name is only for NCR53c720 chips these days (supported on parisc, and eisa/mca). It has correct big endian support for parisc alrady.

I can verify whether the other driver operates correctly in the four
possible configurations with an Atlas board.

BTW, that SCSI_NCR_BIG_ENDIAN macro looks scary -- is the chip wired
incorrectly (i.e. with byte lanes swapped at PCI) or are i/o and/or mmio operations simply broken for parisc?

Quoted 2 lines

Also the patch duplicates your changelog and copyright entry that's in the file already.

It looks like some sort of a merging mistake.

Maciej

#2 lore raw
Quoted 3 lines · Maciej W. Rozycki

BTW, that SCSI_NCR_BIG_ENDIAN macro looks scary -- is the chip wired
incorrectly (i.e. with byte lanes swapped at PCI) or are i/o and/or mmio operations simply broken for parisc?

Many parisc systems have a NCR720 on the HP GSC bus, which actually is accessed as big endian.