On Sat Aug 02, 2003 at 10:06:19PM +0100, Alan Cox wrote:
> On Sad, 2003-08-02 at 18:42, Andries Brouwer wrote:
> > OK, so we have to investigate. This strange test was inserted
> > in 2.4 and 2.5 via Alan, and google gives me Alan's changelog:
> > 
> > Linux 2.5.66-ac1
> > o Don't issue WIN_SET_MAX on older drivers (Jens Axboe)
> >   (Breaks some Samsung)
> 
> Some older Samsung drives don't abort WIN_SET_MAX but the firmware
> hangs hence the check.

Ok, I think I can actually test that one.

    <rummages in ye olde box of hardware>

Cool, found it, I have an ancient Samsung SHD-3212A (426MB)
drive that will hopefully show the problem.

    <sound of testing in the distance>

Ok, found the problem.  The current code (in addition to being
badly written) does not even bother to test if the drive supports
the HPA feature set before issuing a WIN_SET_MAX call.  In my
case, it didn't crash my Samsung drive, but it certainly did make
it complain rather loudly.

I have rewritten the init_idedisk_capacity() function and taught
it to behave itself.  It is now much cleaner IMHO, and will only
issues SET_MAX* calls to drives that claim they support such
things.  I've tested this patch with a 200GB drive, a 120GB
drive, an 80GB drive and my ancient Samsung drive and in each
case (48bit LBA, 28bit LBA, 28bit CHS w/o support for HPA), my
new version appears to the Right Thing(tm).

Attached is a patch vs 2.4.22-pre10, and a patch vs 2.6.0-pre2. 
Please apply,

 -Erik

--
Erik B. Andersen             http://codepoet-consulting.com/
--This message was written using 73% post-consumer electrons--