I was confused by Erik's earlier mail- I thought he'd said "NVRAM
error"- this is accessed via u_int16_t, but needs to be swizzled
anyway.

"Ram Checksum Failure" is what happens after you download new firmware
into the QLogic SRAM. There are usually only three possible reasons for
this kind of failure: you've set up the SRAM wrong (which I've never
actually seen), or the SRAM is bad, or in the case of some very old
QLogic 2100s you try and load f/w larger than > 0x8000 words (QLogic
PROM bug).

-matt


On Fri, 29 Jun 2001, Izumi Tsutsui wrote:

> In <20010629071516.D13931-100000@wonky.feral.com>
> mjacob@feral.com wrote:
>
> > Hmm. I haven't run across this myself. Does it work otherwise?
> > It might be a !)$*)!$ endian issue again, or possibly something
> > else.
>
> Since 93c46/56 EEPROM stores all data in u_int16_t,
> it is better to access them always via u_int16_t
> rather than via u_int8_t pointer, which causes nightmare ;-)
> In this case, no byteswapping is needed if we access data
> in proper size.
> (See dev/ic/iha.c, dev/ic/rtl81x9.c etc.)
> ---
> Izumi Tsutsui
> tsutsui@ceres.dti.ne.jp
>