Re: current & V5 crashes with non-raw sata reads

[Thilo Jeremias <thilo%nispuk.com@localhost>]

Thor,

Thanks for telling me I'm doing silly things (maybe I would like to focus on the problem?).

I am trying to find why this system that runs other os's, consistently crashes (with the same strange ohci messages)
when it has to do a fsck on a 1tb sata raid array.

It is predictable.  And NO READ of whatever size from either device should cause any system corruption.
(and it does not on other boxes I have netbsd running on), and it does not crash with Netbsd 4.1 either.
 
Reading small blocks from the buffered block device works fine.

I know that fsck probably doesn't use the buffered device, but it still crashes with the same symptoms.


thilo


Thor Lancelot Simon wrote:

On Mon, Jun 08, 2009 at 12:48:03AM +1000, Thilo Jeremias wrote:
  

Hi all,
is this worth a PR?
(or where could I start searching for the problem?)

I get: a crash if I try to read bigger blocks (~1m or bigger) from a
sata drive on an asus m4a78-vm mobo.
    

No, you get: a crash if you try to read 1M at a time _through the
metadata cache_, which is a basically silly thing to do.  This should
not crash the system, but the fix you'll probably see soonest will be
the removal of the block device nodes entirely -- they're not really
useful for anything and just cause confusion.

Even if what you are trying to do didn't crash the system (which it
obviously shouldn't) it would be a very bad idea.  Why are you not
using the raw device?

Thor