[Replying to a sort of related message, since I've already deleted the
original commit mail]
On Mon, Mar 15, 2010 at 22:44:11 +0900, Izumi Tsutsui wrote:
> > > I've tested with an i386 kernel. I'd rather not make those changes
> > > conditional so I'll see if I can work out why the Atari kernel would
> > > have problems. If no CTF section is found then ksyms should behave
> > > exactly as it did before the change, just with an empty CTF section.
> >
> > Ok, I'll check what happens on initialization,
> > but should it work even with old bootloaders?
>
> It looks more sanity checks are required in ksyms_addsyms_elf().
>
> ---
> #ifdef KDTRACE_HOOKS
> /* Find the CTF section */
> shdr = (Elf_Shdr *)((uint8_t *)start + ehdr->e_shoff);
> if (ehdr->e_shstrndx != 0) {
> char *shstr = (uint8_t*)start +
> shdr[ehdr->e_shstrndx].sh_offset;
> for (i = 1; i < ehdr->e_shnum; i++) {
> if (shdr[i].sh_type != SHT_PROGBITS)
> continue;
> if (strncmp(".SUNW_ctf", &shstr[shdr[i].sh_name] ,10)
> != 0)
> continue;
> ---
>
> My atari kernel hangs in this strncmp().
> shstr is 0x4c60f4 and it looks invalid because
> kernel end is 0x277d48 and shdr is 0x277d7c.
>
> (BTW, please also use KNF properly)
BTW, a counterpart strncmp(".SUNW_ctf", ...) in libsa broke at least
http://releng.netbsd.org/builds/HEAD/201003140000Z/mvme68k.build.failed
http://releng.netbsd.org/builds/HEAD/201003140000Z/mvmeppc.build.failed
with
loadfile_elf32.c:(.text+0x...): undefined reference to `strncmp'
I guess you need to add strncmp.c to SRCS (see libsa/Makefile:35)
It's also suspicious that the same check is not in loadfile_elf64.c
SY, Uwe
--
u...@stderr.spb.ru | Zu Grunde kommen
http://snark.ptc.spbu.ru/~uwe/ | Ist zu Grunde gehen
