On Thursday, June 24, 1999 4:24 PM, Chad Page [SMTP:[EMAIL PROTECTED]] wrote:
:
: There is (was?) some stack checking done at the system call level.
: However, it's not 100% foolproof - if the program gets sp above the
: low-water mark after dipping into bss before the next system call it won't
: be detected.
:
: Now, if you had a magic # right after bss and checked that also,
: that would really improve the stack checking.
:
That's a pretty neat idea, but it would require that the kernel make certain
assumptions about the image layout of process data, as well as the linker
emitting data after the bss section, both of which aren't too good...
Using this for in-process stack checking would work though, and would
tell whether the interrupt stack blew the process out.
