Messages in this thread | | | Date | Fri, 16 Jul 1999 22:49:54 +0200 (MET DST) | From | Gerard Roudier <> | Subject | Re: [BUGREPORT] 2.2.10-ac10: oops+panic when interrupting cdrecord session |
| |
On Fri, 16 Jul 1999, Marc Mutz wrote:
> Bert Lindner wrote: > > > <snip> > > > And that would seem to be in the sym53c8xx driver when unloaded. Gerard? > > > I think that you will not have any trouble if you compile the driver > > > into your kernel instead of as a module. > > > > I just tried that, and the problem remained. Putting the sym53c8xx driver > <snip> > > Bert, > Maybe you should interchange the sym53c8xx with the older/more general > ncr53c8xx driver. If the problem remains, then one can rule out that > particular driver (except when both share a fair amount of code).
Indeed. But not the way they deal with memory allocations. The sym53c8xx driver incorporates its own memory allocator since I wanted naturally aligned blocks in order to make efficient address calculations using the 8 bit ALUs of the SCRIPTS processor. Linux didn't offered that unlike most other O/Ses.
There is something that amazed me since years in Linux. If we take into account all the hardware, firmare, software, bloatware, userware, overware wrongware and chineseware out of what a modern computer is made, the SCSI driver is probably less than 0,00..1 % of the total amount of that whateverware. But SCSI drivers are focused 99,99..9 % of the time when a machine using SCSI is experiencing problems.
I donnot mean that the driver is not candidate in the problem trigerring or that it cannot be the culprit. But, in my experience, people are often surprised of the real cause of a failure, especially those that arbitrarily designate a single component as the only possible candidate.
For the problem that have been reported, I didn't see any obvious correlation with some driver fault. In the initial post, it appeared that the kernel stack contains some frames that indicated that the driver had serviced an interrupt, but the last visible trace was scsi_old_done() that may indicate the driver was probably done with its interrupt handling and that the problem occurred later.
Gérard.
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |