> > The feature that was definitely the most helpful was crash dumps. Often
Fine. Actually if you want to dump your system usefully on a crash get
remote kernel debugging working. When you can crash in an interrupt handler
view the source line, fix the wrong pointer and continue then you'll
understand what being able to source level debug your kernel buys you -
and almost all the bits you need, all the hard stuff is already in
gdb now.
And how, this is the only way I was able to write an entire scsi
driver for the Sparc in one evening. It was most of the early silly
stuff that would have driven me nuts with scanning printk()'s on the
screen, after fixing these I am less apt to use the debugger at all
for the rest of the driver development unless I get completely stumped
and have to look at the state of the machine in a sophistaced manner
some time down the road.
Later,
David S. Miller
davem@caip.rutgers.edu