2001-12-14 15:07:03

by Stephen Cameron

[permalink] [raw]
Subject: [PATCH] cciss 2.5.0 for 2.5.1-pre11

Christoph Hellwig ([email protected]) wrote:

> In article <[email protected]> you wrote:
> >
> > Here's a patch for the cciss driver in the 2.5.1-pre8 tree
> > (patch also applies to 2.5.1-pre9): <
> > http://geocities.com/dotslashstar/cciss_2.5.0_for_2.5.1-pre8.txt
> >
> > This patch:
> >
> > * adds support for SCSI tape drives.
> > * adds support for dynamically adding and removing
> > logical volumes on the fly.
>
> * sets hardsectsizes to '0' for invalid volumes, causing devisions by zero
> in ll_rw_block().

I think I fixed this. (thanks for the patch).

> * backs out random fixes done in the mainline

I can't see what you mean here. What "random fixes" are
you referring to?
.
Here is a new patch against 2.5.1-pre11:
http://www.geocities.com/smcameron/cciss_2.5.0_for_2.5.1-pre11.patch.gz

-- steve





__________________________________________________
Do You Yahoo!?
Check out Yahoo! Shopping and Yahoo! Auctions for all of
your unique holiday gifts! Buy at http://shopping.yahoo.com
or bid at http://auctions.yahoo.com


2001-12-19 16:12:11

by Stephen Cameron

[permalink] [raw]
Subject: Re:[PATCH] cciss 2.5.0 for 2.5.1-pre11

I wrote:
[...]
> Here is a new patch against 2.5.1-pre11:
> http://www.geocities.com/smcameron/cciss_2.5.0_for_2.5.1-pre11.patch.gz

Having played with it some more, I see it is seriously flawed
in the area of SCSI tape support and locking. I was probably already pushing
my luck a bit trying to make a hybrid scsi+block driver, and now that
there are per-adapter locks rather than one io_request_lock, it's seems
quite a lot trickier (perhaps even impossible?) to get such a strange
beast right.

Anyway, the patch seemed to work upon first trying it, but trying this,
once I managed to get my hands on more than 1 tape drive again,

tar cvf /dev/st0 /etc > /dev/null & tar cvf /dev/st1 /etc > /dev/null

(with /dev/st[01] mapping to tape drives on a cciss controller of course)
would quickly demonstrate that patch has big problems. In case anyone is
trying to use that patch for 2.5.x tree, you should stop using it. I must
apologize for distributing that kind of garbage, I really should have caught
that earlier. That the quickly moving kernel tree makes extensive
testing difficult would be my only and insufficient excuse.

-- steve



__________________________________________________
Do You Yahoo!?
Check out Yahoo! Shopping and Yahoo! Auctions for all of
your unique holiday gifts! Buy at http://shopping.yahoo.com
or bid at http://auctions.yahoo.com