[PATCH 0/6] PowerPc 8xx TLB/MMU fixes

[Joakim Tjernlund]

Benjamin Herrenschmidt <benh at kernel.crashing.org> wrote on 06/10/2009 00:37:28:
>
> On Tue, 2009-10-06 at 00:31 +0200, Joakim Tjernlund wrote:
> >
> > regs or regs->nip is NULL? Either one does not make sense
> > In any case it might be a secondary problem as DAR is NULL already
> > when you
> > enter the page fault.
> > >
> > >                 insn = *((unsigned long *)regs->nip);
> > > c000e110:       80 a9 00 00     lwz     r5,0(r9)
> > >
> > > fails.
> >
> > hmm, I wonder if you managed to invalidate the a kernel TLB?
> > Are you using pinned kernel TLBs?
>
> You should not dereference a user address like that. Use get_user !

So how does this look? Does it change anything?
It should as the previous way was way off :(

diff --git a/arch/powerpc/mm/fault.c b/arch/powerpc/mm/fault.c
index c33c6de..08a392f 100644
--- a/arch/powerpc/mm/fault.c
+++ b/arch/powerpc/mm/fault.c
@@ -153,7 +153,7 @@ int __kprobes do_page_fault(struct pt_regs *regs, unsigned long address,
 #ifdef DEBUG_DCBX
 		const char *istr = NULL;

-		insn = *((unsigned long *)regs->nip);
+		__get_user(insn, (unsigned long __user *)regs->nip);
 		if (((insn >> (31-5)) & 0x3f) == 31) {
 			if (((insn >> 1) & 0x3ff) == 1014) /* dcbz ? 0x3f6 */
 				istr = "dcbz";
@@ -178,11 +178,12 @@ int __kprobes do_page_fault(struct pt_regs *regs, unsigned long address,
 					       ra, rb, dar);
 					is_write = 0;
 				}
-
+#if 0
 				if (trap == 0x300 && address != dar) {
 					__asm__ ("mtdar %0" : : "r" (dar));
 					return 0;
 				}
+#endif
 			}
 		}
 #endif
@@ -191,7 +192,7 @@ int __kprobes do_page_fault(struct pt_regs *regs, unsigned long address,

 			/* This is from a dcbX or icbi insn gone bad, these
 			 * insn do not set DAR so we have to do it here instead */
-			insn = *((unsigned long *)regs->nip);
+			__get_user(insn, (unsigned long __user *)regs->nip);

 			ra = (insn >> (31-15)) & 0x1f; /* Reg RA */
 			rb = (insn >> (31-20)) & 0x1f; /* Reg RB */