On Mon, Jul 13, 2015 at 05:40:35PM -0500, Josh Poimboeuf wrote:
> Also, it's late and this is still a half-formed thought, but... in the
> case of building livepatch modules using kbuild (rather than a more
> heavyweight tool like kpatch-build) I think it might also make it
> possible to have a C macro which allows you to specify unexported symbol
> addresses directly from the patch code source. That could be a big win
> for the kbuild approach to building patch modules. Need to think about
> that some more.
Ok, I think this could work out *really* nicely...
Something like:
struct klp_symbol __section(__klp_symbols) symbols[] = {
{
.objname = NULL,
.sym = "saved_command_line",
.addr = 0xffffffff81ebc008,
},
};
extern char *saved_command_line;
static int cmdline_proc_show__patched(struct seq_file *m, void *v)
{
seq_printf(m, "%s patched=1\n", saved_command_line);
return 0;
}
Then, as Jessica suggested, the module simplify_symbols() code could
call klp_resolve_symbol() for the SHN_UNDEF saved_command_line symbol.
klp_resolve_symbol() would read saved_command_line's address from the
__klp_symbols section and verify that it matches kallsyms. Then the
module code would relocate it normally in apply_relocations().
IMO this would take away one of the biggest reasons why we need a big
complex patch generation tool like kpatch-build.
--
Josh
--
To unsubscribe from this list: send the line "unsubscribe live-patching" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
