[POWERPC] 4xx: Deal with 44x virtually tagged icache

The 44x family has an interesting "feature" which is a virtually
tagged instruction cache (yuck !). So far, we haven't dealt with
it properly, which means we've been mostly lucky or people didn't
report the problems, unless people have been running custom patches
in their distro...

This is an attempt at fixing it properly. I chose to do it by
setting a global flag whenever we change a PTE that was previously
marked executable, and flush the entire instruction cache upon
return to user space when that happens.

This is a bit heavy handed, but it's hard to do more fine grained
flushes as the icbi instruction, on those processor, for some very
strange reasons (since the cache is virtually mapped) still requires
a valid TLB entry for reading in the target address space, which
isn't something I want to deal with.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Josh Boyer <jwboyer@linux.vnet.ibm.com>

1 files changed, 9 insertions, 0 deletions

diff --git a/arch/ppc/kernel/misc.S b/arch/ppc/kernel/misc.S
index 2b81e71d6b2..e0c850d85c5 100644
--- a/arch/ppc/kernel/misc.S
+++ b/arch/ppc/kernel/misc.S
@@ -499,12 +499,21 @@ END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
 	addi	r3,r3,L1_CACHE_BYTES
 	bdnz	0b
 	sync
+#ifndef CONFIG_44x
+	/* We don't flush the icache on 44x. Those have a virtual icache
+	 * and we don't have access to the virtual address here (it's
+	 * not the page vaddr but where it's mapped in user space). The
+	 * flushing of the icache on these is handled elsewhere, when
+	 * a change in the address space occurs, before returning to
+	 * user space
+	 */
 	mtctr	r4
 1:	icbi	0,r6
 	addi	r6,r6,L1_CACHE_BYTES
 	bdnz	1b
 	sync
 	isync
+#endif /* CONFIG_44x */
 	blr
 
 /*