>Ha ok. So I looked back a bit more in cvs, and in my mail archives from last
>year. I think the patch which solved your problem was:

Yes. Bingo. 

Except back then, the r_st_bmask was 0, and now its WDCS_DRDY.
Could that have anything to do with it?

(I've turned on more debugging, and added private debuggging patches
to the Ata code, and the symptoms seem to be the same as last time...
CMD_ERR from ata_get_params() returns CMD_ERR because the wdc_c.flags
are bad.)