Re: 2.1.118 Tons of oopes

Michael H. Warfield (mhw@wittsend.com)
Wed, 26 Aug 1998 11:46:13 -0400 (EDT)


Richard Gooch enscribed thusly:
> Garst R. Reese writes:
> > When I started pcmcia-cs-3.0.5 with only a serial card active I got a
> > flood of unable to dereference null pointer oopes. If anybody wants them
> > I'll run em thru ksymoops, but I suspect it's a matter of telling dhinds
> > what got changed (all those added NULLS?).
> > Got another one shutting down, but 2.1.117 + bill's patches came up
> > clean.

> I think this is because of the incompatible source change to struct
> file_operations introduced in 2.1.118. A new flush() method was
> inserted between the open() and release() methods instead of being
> placed at the bottom (where it should have been placed).

> If you look at modules/ds.c you'll see the use of ds_fops is
> incompatible with the kernel: include/linux/fs.h

> A warning to all: anyone with a driver which isn't distributed with
> the kernel sources will have similar problems.

I'm afraid it may be worse than that. I just saw a compiler warning
in joystick.c that I hadn't seen before about an improper structure
initialization and, having read these messages previously, got suspicious.
Sure enough, the joystick driver in the kernel sources themselves seems
to be busted in this regard. It may not be the only one. I just happen
to be watching at that moment and caught the compiler warning. (Who SAYS
those warnings aren't usefull!?!?!?)

> Linus: for god's sake, back out the particular patch that did this!
> Surely we can live with having flush() at the end of the structure?!?!

Yeah... This one looks like is could be prone to random acts
of terrorism for some time to come. Anyone with external drivers like
pcmcia and such would have to conditionalize that structure on this
kernel version...

Now, the flip side to all of that is that, even with the new
additions at the end of the structure, all of those drivers STILL should
be updated. If they are not, you STILL end up with an uninitiallized
structure element. Yes I realize it gets set to NULL in this case, but
we really should NOT depend on that.

Hmmm... Alan came out with an AC3 patch just "moments" before
2.1.118 was released and I saw some Joliet patches that showed up on the
list "moments" after the release. Looks like 2.1.118 goes in the dumper
at this end. I think I'll go with 2.1.117ac3 until 2.1.119 shows up... :-)

> Regards,

> Richard....

Mike

-- 
 Michael H. Warfield    |  (770) 985-6132   |  mhw@WittsEnd.com
  (The Mad Wizard)      |  (770) 925-8248   |  http://www.wittsend.com/mhw/
  NIC whois:  MHW9      |  An optimist believes we live in the best of all
 PGP Key: 0xDF1DD471    |  possible worlds.  A pessimist is sure of it!

- 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.altern.org/andrebalsa/doc/lkml-faq.html