First retrieve length and hash values *before* advancing to the next entry.
On hashline collision it would check a random one next.
This commit is contained in:
parent
aa9cb3deaa
commit
43188548d8
|
@ -1,4 +1,4 @@
|
|||
/* $NetBSD: udf_subr.c,v 1.59 2008/07/17 19:10:22 reinoud Exp $ */
|
||||
/* $NetBSD: udf_subr.c,v 1.60 2008/07/18 13:34:05 reinoud Exp $ */
|
||||
|
||||
/*
|
||||
* Copyright (c) 2006, 2008 Reinoud Zandijk
|
||||
|
@ -29,7 +29,7 @@
|
|||
|
||||
#include <sys/cdefs.h>
|
||||
#ifndef lint
|
||||
__KERNEL_RCSID(0, "$NetBSD: udf_subr.c,v 1.59 2008/07/17 19:10:22 reinoud Exp $");
|
||||
__KERNEL_RCSID(0, "$NetBSD: udf_subr.c,v 1.60 2008/07/18 13:34:05 reinoud Exp $");
|
||||
#endif /* not lint */
|
||||
|
||||
|
||||
|
@ -4061,11 +4061,14 @@ udf_dirhash_lookup(struct udf_node *dir_node, const char *d_name, int d_namlen,
|
|||
/* start where we were */
|
||||
if (*result) {
|
||||
KASSERT(dir_node->dir_hash);
|
||||
dirh_e = LIST_NEXT(*result, next);
|
||||
dirh_e = *result;
|
||||
/* retrieve information */
|
||||
if (dirh_e) {
|
||||
hashvalue = dirh_e->hashvalue;
|
||||
d_namlen = dirh_e->d_namlen;
|
||||
}
|
||||
/* advance */
|
||||
dirh_e = LIST_NEXT(*result, next);
|
||||
} else {
|
||||
/* calculate our hash */
|
||||
hashvalue = udf_dirhash_hash(d_name, d_namlen);
|
||||
|
|
Loading…
Reference in New Issue