cache: get rid of the global lock

Add a table of vnode locks and use them along with bucketlocks to provide concurrent modification support. The approach taken is to preserve the current behaviour of the namecache and just lock all relevant parts before any changes are made. Lookups still require the relevant bucket to be locked. Discussed with: kib Tested by: pho
svn path=/head/; revision=306224
2016-09-23 04:45:11 +00:00 · 2016-09-23 04:45:11 +00:00 · 1d2541fd1a · 2020-12-20 02:59:44 +00:00
commit 1d2541fd1a
parent 36f5d07745
2 changed files with 742 additions and 239 deletions
--- a/sys/kern/subr_witness.c
+++ b/sys/kern/subr_witness.c
@ -625,7 +625,7 @@ static struct witness_order_list_entry order_lists[] = {
 	/*
 	 * VFS namecache
 	 */
-	{ "ncglobal", &lock_class_rw },
+	{ "ncvn", &lock_class_mtx_sleep },
 	{ "ncbuc", &lock_class_rw },
 	{ "vnode interlock", &lock_class_mtx_sleep },
 	{ "ncneg", &lock_class_mtx_sleep },
--- a/sys/kern/vfs_cache.c
+++ b/sys/kern/vfs_cache.c