> These aren't the only ways.
That's probably true.
> I think it would be good to let a crash dump handler open a device file
> to get control. Before dumping core, the kernel checks for a crash dump
> handler running with the same UID. If one is found, the dying process
> sleeps and the crash dump handler is asked to respond.
That would remove process invocation from the kernel path.
You could do a lot of *this* in user space by having a library function
set the default value for all signals, or just SIGSEGV, to a function
that uses some kind of IPC, such as writing to a fifo whose name includes
the uid. I wonder if it's possible to use preloading to inject that
snippet of code into an existing executable without relinking it and
without touching libc.
Michael Elizabeth Chastain
"love without fear"
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to email@example.com
Please read the FAQ at http://www.tux.org/lkml/