Re: kernel upgrade
On Wed 18 Aug 1999, Robb Romans wrote:
> On Tue, 17 Aug 1999, Paul Slootman wrote:
>
> > On Sun 15 Aug 1999, Ryan Kirkpatrick wrote:
> >
> > > > Don't use 2.2.11. It's br0ken. 2.2.12 will be out on monday or tuesday.
> > >
> > > I will second this. The best it did on my XLT was kernel panic
> > > when trying to initialize my SCSI controllers. I would recommend 2.2.5 or
> >
> > I tried this myself yesterday (2.2.11 that is), and the ncr53c8xx and
> > sym53c8xx drivers think that there's some problem with the cache, as it
> > detects that the script isn't running (the script that gets executed by
> > the 810 chip that is). I tried the 53c7,8xx driver, same problem only
> > worded differently :-(
>
> I saw this exact thing on an XL300 with a 53c810a and 53c875.
I've since got a fix via the kernel list after I posted all the details
of my crash:
: From: Gerard Roudier <groudier@club-internet.fr>
:
: > I just verified that I still have this problem with 2.2.12-final.
: > The symptom:
: >
: > sym53c8xx: 53c810a detected
: > sym53c810a-0: rev=0x11, base=0xa000000, io_port=0x9800, irq=28
: > sym53c810a-0: ID=7, Fast-IO, Parity Checking
: > CACHE TEST FAILED: script exection failed
: > start=80292d08, pc=80292d08, end=80292d28
:
: The DMA area is supposed to start at 1GB as seen from the BUS but the bus
: addresses above (start, pc and end) are beyond 2 GB. If your machine has
: less than 1 GB, then it is quite abnormal.
:
: In earlier versions both dma base and size were 1GB. Only the dma size has
: been changed to 2GB but not the dma base.
:
: There is some dubious section in core_cia.h that seems to confuse dma base
: and dma size. If it is so, this just break virt_to_bus() translations and
: explains the breakage and the offending bus addresses above as well:
:
: You should give a try to the below simple changes (do it by hand):
:
: #if defined(CONFIG_ALPHA_GENERIC) || defined(CONFIG_ALPHA_SRM_SETUP)
: #define CIA_DMA_WIN_BASE alpha_mv.dma_win_base
: #define CIA_DMA_WIN_SIZE alpha_mv.dma_win_size
: #else
: -#define CIA_DMA_WIN_BASE CIA_DMA_WIN_SIZE_DEFAULT
: +#define CIA_DMA_WIN_BASE CIA_DMA_WIN_BASE_DEFAULT
: #define CIA_DMA_WIN_SIZE CIA_DMA_WIN_SIZE_DEFAULT
: #endif
:
: By the way, the core_pyxis stuff seems fine in this regard.
The last line is the reason why other people had this problem fixed in
2.2.11, and I still had problems... And yes, I didn't choose the generic
system type. What are the advantages of either? I could imagine that a
kernel built for the generic system type might be suboptimal...
Paul Slootman
--
home: paul@wurtel.demon.nl http://www.wurtel.demon.nl/
work: paul@murphy.nl http://www.murphy.nl/
debian: paul@debian.org http://www.debian.org/
isdn4linux: paul@isdn4linux.de http://www.isdn4linux.de/
Reply to: