After switching mutexes to use lwsync, they no longer provide sufficient

guarantees on acquire for the tlbie mutex. Conversely, the TLB invalidation
sequence provides guarantees that do not need to be redundantly applied on
release. Roll a small custom lock that is just right. Simultaneously,
convert the SLB tree changes back to lwsync, as changing them to sync
was a misdiagnosis of the tlbie barrier problem this commit actually fixes.

1 files changed, 2 insertions, 2 deletions

diff --git a/sys/powerpc/aim/slb.c b/sys/powerpc/aim/slb.c
index 7f4b2ef..162c7fb 100644
--- a/sys/powerpc/aim/slb.c
+++ b/sys/powerpc/aim/slb.c
@@ -139,7 +139,7 @@ make_new_leaf(uint64_t esid, uint64_t slbv, struct slbtnode *parent)
 	 * that a lockless searcher always sees a valid path through
 	 * the tree.
 	 */
-	powerpc_sync();
+	mb();
 
 	idx = esid2idx(esid, parent->ua_level);
 	parent->u.ua_child[idx] = child;
@@ -187,7 +187,7 @@ make_intermediate(uint64_t esid, struct slbtnode *parent)
 	idx = esid2idx(child->ua_base, inter->ua_level);
 	inter->u.ua_child[idx] = child;
 	setbit(&inter->ua_alloc, idx);
-	powerpc_sync();
+	mb();
 
 	/* Set up parent to point to intermediate node ... */
 	idx = esid2idx(inter->ua_base, parent->ua_level);