On Thu, 15 Mar 2001, Manuel Bouyer wrote:

> On Thu, Mar 15, 2001 at 10:07:51AM +0100, P=E5l Halvorsen wrote:
> > OK, I might have a problem getting read sizes larger than 64 KB, but
> > in my system (having created a zero-copy data path for streams), I
> > allocate my own buf structures with my own memory region. I do not use =
the
> > buffer cache - but an separate buffer manager carrying out the file sys=
tem
> > functions. I "only" use=20
> > =09VOP_BMAP(vp, bp->b_lblkno, NULL, &bp->b_blkno, &num_ra);
> > to see the amount of data to read ahead and
> > =09VOP_STRATEGY(bp);
> > to call the disk driver (SCSI) with my buf structure.=20
> >=20
> > In such a case, is it possible, without too much work - especially in t=
he
> > driver code which I do not know much about, to get > 64 KB read
> > operations?
>=20
> It depends on the driver :)
> For IDE disks for example, it's hardware - an IDE disk can't do more than
> 64K (in fact I think it can do 127.5K - but not 128K it would overflow th=
e
> nsect register). For SCSI there may be limitations in some hardware,
> at the adapter level (or underlying bus level - I think on pmax you can't
> DMA more than 8k for example, but my memory may be failing :).
> So it really depends on the hardware. For example in siop it's just a mat=
ter
> of bumping a constant, and adding more slots in the script (it's even
> ready for 128K on hardware with 8k pages I think :).
> Other drivers may adapt automatically to MAXPHYS ...

So I could really just try to increase MAXPHYS and see what happends?

Will other devices be affected by an increase of MAXPHYS (making the
kernel panic'ing due to a too large MAXPHYS constant)?=20

-ph