Vojtech Pavlik wrote:

>The later (lv_disk_t) struct isn't used anywhere in the kernel -
>probably defined for userspace only? That's weird! And also many other
>structs in lvm.h are nowhere to be found used. Guess we could swipe them
>out as well.
>
>The first lv_read_ahead (in lv_t) removed. And references to it as well.
>
Yes I know the lvm coders where too deaf to separate user level 
structure layout properly
from on disk and kernel space by using just different header files for 
different purposes.
And then they tryed apparently to embarce anything they could think off, 
without really
thinking hard about what should be there and what shouldn't. It was too 
hard for
them to have a sneak view on for example Solaris to recognize what's 
really needed.

But they promise perpetually once in a while that the next version will 
be "coming real soon" and
be wonderfull...  Perhaps someone should just stump over them and clean 
this mess up,
with the disrespect those pseudo maintainers do deserve....

Anyway, you apparently still missed to kill:

     int *bs, max_ra; in ide-probe.c

as well as:
    xpram_rahead and friends in s390 code

The attached patch is fixing this.

BTW.> Since there is no longer any difference about the request head 
handling
between IDE and SCSI, what about the idea of moving the whole ide 
interface stuff under
the umbrella of SCSI host adapter? This would be a true cleanup and make
the whole ide-scsi and ide-atapi mess go away. IDE is moving fast toward 
SCSI on
the logical level anyway and it would make the hwif macro/lookup crap in 
the ide code go
magically way! At least this generic device handler search stuff should 
be merged
between them (I'm trully tempted to give it a shoot this afternoon.)
The only thing it could result in, which would maybe surprise some would 
be the fact
that the major of his root device could just go suddenly away... But 
hey! What's the heck - we
are in odd kernel series anyway ;-).