Blob Blame History Raw
From: "J. Bruce Fields" <bfields@redhat.com>
Date: Tue, 19 Sep 2017 20:51:31 -0400
Subject: [PATCH] nfsd: increase DRC cache limit
Git-commit: 44d8660d3bb0a1c8363ebcb906af2343ea8e15f6
Patch-mainline: v4.15
References: bsc#1150381

An NFSv4.1+ client negotiates the size of its duplicate reply cache size
in the initial CREATE_SESSION request.  The server preallocates the
memory for the duplicate reply cache to ensure that we'll never fail to
record the response to a nonidempotent operation.

To prevent a few CREATE_SESSIONs from consuming all of memory we set an
upper limit based on nr_free_buffer_pages().  1/2^10 has been too
limiting in practice; 1/2^7 is still less than one percent.

Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Acked-by: NeilBrown <neilb@suse.com>

---
 fs/nfsd/nfssvc.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

--- a/fs/nfsd/nfssvc.c
+++ b/fs/nfsd/nfssvc.c
@@ -446,7 +446,7 @@ void nfsd_reset_versions(void)
  */
 static void set_max_drc(void)
 {
-	#define NFSD_DRC_SIZE_SHIFT	10
+	#define NFSD_DRC_SIZE_SHIFT	7
 	nfsd_drc_max_mem = (nr_free_buffer_pages()
 					>> NFSD_DRC_SIZE_SHIFT) * PAGE_SIZE;
 	nfsd_drc_mem_used = 0;