kernel vm question

[Jacques Fourie]

Hi,

Yes, I am trying to patch a piece of code in the kernel. The strange
thing is that this code works without a problem on FreeBSD 4.8 - has
the VM system changed to such an extent between 4.8 and 4.9 that the
pages in the kernel code segment are not writeable anymore?

Thanks for the bit about the compiler optimization - at least I
understand that part now :)

regards,
jacques


On Thu, 27 Jan 2005 11:09:14 -0500, David Schultz <das at freebsd.org> wrote:
> On Thu, Jan 27, 2005, Jacques Fourie wrote:
> > Hi,
> >
> > I have a kernel module with the following entry point :
> >
> > static int test_modevent(module_t mod, int type, void *unused)
> > {
> >   int s;
> >   unsigned char *p = NULL;
> >   unsigned char v = 0x55;
> >
> >   switch (type)
> >     {
> >     case MOD_LOAD:
> >       p = (unsigned char *)ip_output;
> >
> >       s = splhigh();
> >
> >       v = p[0];                   /* Page fault without this line */
> >       p[0] = v;
> [...]
> > If I remove the line "Page fault without this line" line, I get a page
> > fault when loading this module. This crash seems to be hardware as well
> > as version specific - I can not reproduce the crash on 4.8 but on one
> > particular piece of hardware it crashes consistently on 4.9 and 4.10.
> 
> When the line is there, the compiler is probably smart enough to
> realize that 'x=y; y=x' is (usually) a no-op, so it optimizes away
> both statements.  Otherwise, you get a page fault because you're
> trying to write to a non-writable page (namely, one in the
> kernel's code segment).  If you're trying to do what I think
> you're trying to do, you need to poke around in the VM system and
> adjust the protection bits in the page you want to write to.
> _______________________________________________
> freebsd-hackers at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-hackers
> To unsubscribe, send any mail to "freebsd-hackers-unsubscribe at freebsd.org"
>