[ros-kernel] AMD PCNet driver support broken

Mike Nordell tamlin at algonet.se
Fri Feb 20 09:13:06 CET 2004


Hartmut Birr wrote:

> it seems that someone writes over the end of a allocated block. Ros
> boots successful, if I add some dummy bytes at the begin of BLOCK_HDR
> (npool.c).

One way to narrow this down (though it'll make the kernel dreadfully slow,
why it's a local-tree-only change), would be to walk all the allocations on
both alloc and free and check for red-zone violations. Dumping address and
size (of the allocated offending block) it should be possible to pinpoint
the violator (bug) and terminate its existance.

/Mike



More information about the Ros-kernel mailing list