Hmm. You might be right. It should work without that first alignment
operation because the (size) is already aligned. But for safety's sake I
would probably want to change the if() into a while() to allow count to be
adjusted twice if necessary, just to be sure. The init function is not in
a critical path. I'll test it out a bit and commit with a credit to you.
Thanks!
-Matt
diff --git a/sys/kern/kern_kmalloc.c b/sys/kern/kern_kmalloc.c
index d222e053869..27ae4a81ac9 100644
--- a/sys/kern/kern_kmalloc.c
+++ b/sys/kern/kern_kmalloc.c
@@ -288,21 +288,23 @@ malloc_mgt_init(struct malloc_type *type __unused,
/*
* Figure out the count by taking into account the size of the fobjs[]
- * array by adding it to the object size.
+ * array by adding it to the object size. This initial calculation
+ * ignores alignment edge-cases that might require the count to be
+ * reduced.
*/
offset = offsetof(struct kmalloc_slab, fobjs[0]);
- offset = __VM_CACHELINE_ALIGN(offset);
count = (KMALLOC_SLAB_SIZE - offset) / (size + sizeof(void *));
/*
- * However, the fobj[] array itself must be aligned, so we might
- * have to reduce the count by 1. (We can do this becaues 'size'
- * is already aligned as well).
+ * Recalculate the offset of the first object, this time including
+ * the required alignment. (size) should already be aligned. This
+ * may push the last object beyond the slab so check and loop with
+ * a reduced count as necessary.
*/
offset = offsetof(struct kmalloc_slab, fobjs[count]);
offset = __VM_CACHELINE_ALIGN(offset);
- if (offset + size * count > KMALLOC_SLAB_SIZE) {
+ while (offset + size * count > KMALLOC_SLAB_SIZE) {
--count;
offset = offsetof(struct kmalloc_slab, fobjs[count]);
offset = __VM_CACHELINE_ALIGN(offset);
@@ -311,6 +313,10 @@ malloc_mgt_init(struct malloc_type *type __unused,
mgt->slab_offset = offset;
mgt->slab_count = count;
+
+ kprintf("XXX mgt_init %s base offset %zx count %zd (end %08zx)\n",
+ type->ks_shortdesc, offset, count, offset + size * count);
+
}
void
