summaryrefslogtreecommitdiffstats
path: root/fs
diff options
context:
space:
mode:
authorJeff Layton <jlayton@redhat.com>2013-02-04 08:18:02 -0500
committerJ. Bruce Fields <bfields@redhat.com>2013-02-04 17:19:10 -0500
commit0338dd157282c19696d3c32614a748d0ba814b12 (patch)
tree4a42757386eab303fd723bcfbcadf843c921ebc8 /fs
parent0ee0bf7ee5b55f232b645c4af0b0c37d4e115a32 (diff)
downloadop-kernel-dev-0338dd157282c19696d3c32614a748d0ba814b12.zip
op-kernel-dev-0338dd157282c19696d3c32614a748d0ba814b12.tar.gz
nfsd: dynamically allocate DRC entries
The existing code keeps a fixed-size cache of 1024 entries. This is much too small for a busy server, and wastes memory on an idle one. This patch changes the code to dynamically allocate and free these cache entries. A cap on the number of entries is retained, but it's much larger than the existing value and now scales with the amount of low memory in the machine. Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Diffstat (limited to 'fs')
-rw-r--r--fs/nfsd/nfscache.c125
1 files changed, 75 insertions, 50 deletions
diff --git a/fs/nfsd/nfscache.c b/fs/nfsd/nfscache.c
index c0c5847..d213e6e 100644
--- a/fs/nfsd/nfscache.c
+++ b/fs/nfsd/nfscache.c
@@ -10,17 +10,13 @@
#include <linux/slab.h>
#include <linux/sunrpc/clnt.h>
+#include <linux/highmem.h>
#include "nfsd.h"
#include "cache.h"
-/* Size of reply cache. Common values are:
- * 4.3BSD: 128
- * 4.4BSD: 256
- * Solaris2: 1024
- * DEC Unix: 512-4096
- */
-#define CACHESIZE 1024
+#define NFSDDBG_FACILITY NFSDDBG_REPCACHE
+
#define HASHSIZE 64
static struct hlist_head * cache_hash;
@@ -28,6 +24,7 @@ static struct list_head lru_head;
static int cache_disabled = 1;
static struct kmem_cache *drc_slab;
static unsigned int num_drc_entries;
+static unsigned int max_drc_entries;
/*
* Calculate the hash index from an XID.
@@ -48,6 +45,34 @@ static int nfsd_cache_append(struct svc_rqst *rqstp, struct kvec *vec);
*/
static DEFINE_SPINLOCK(cache_lock);
+/*
+ * Put a cap on the size of the DRC based on the amount of available
+ * low memory in the machine.
+ *
+ * 64MB: 8192
+ * 128MB: 11585
+ * 256MB: 16384
+ * 512MB: 23170
+ * 1GB: 32768
+ * 2GB: 46340
+ * 4GB: 65536
+ * 8GB: 92681
+ * 16GB: 131072
+ *
+ * ...with a hard cap of 256k entries. In the worst case, each entry will be
+ * ~1k, so the above numbers should give a rough max of the amount of memory
+ * used in k.
+ */
+static unsigned int
+nfsd_cache_size_limit(void)
+{
+ unsigned int limit;
+ unsigned long low_pages = totalram_pages - totalhigh_pages;
+
+ limit = (16 * int_sqrt(low_pages)) << (PAGE_SHIFT-10);
+ return min_t(unsigned int, limit, 256*1024);
+}
+
static struct svc_cacherep *
nfsd_reply_cache_alloc(void)
{
@@ -68,6 +93,7 @@ nfsd_reply_cache_free_locked(struct svc_cacherep *rp)
{
if (rp->c_type == RC_REPLBUFF)
kfree(rp->c_replvec.iov_base);
+ hlist_del(&rp->c_hash);
list_del(&rp->c_lru);
--num_drc_entries;
kmem_cache_free(drc_slab, rp);
@@ -75,30 +101,18 @@ nfsd_reply_cache_free_locked(struct svc_cacherep *rp)
int nfsd_reply_cache_init(void)
{
- int i;
- struct svc_cacherep *rp;
-
drc_slab = kmem_cache_create("nfsd_drc", sizeof(struct svc_cacherep),
0, 0, NULL);
if (!drc_slab)
goto out_nomem;
- INIT_LIST_HEAD(&lru_head);
- i = CACHESIZE;
- num_drc_entries = 0;
- while (i) {
- rp = nfsd_reply_cache_alloc();
- if (!rp)
- goto out_nomem;
- ++num_drc_entries;
- list_add(&rp->c_lru, &lru_head);
- i--;
- }
-
- cache_hash = kcalloc (HASHSIZE, sizeof(struct hlist_head), GFP_KERNEL);
+ cache_hash = kcalloc(HASHSIZE, sizeof(struct hlist_head), GFP_KERNEL);
if (!cache_hash)
goto out_nomem;
+ INIT_LIST_HEAD(&lru_head);
+ max_drc_entries = nfsd_cache_size_limit();
+ num_drc_entries = 0;
cache_disabled = 0;
return 0;
out_nomem:
@@ -191,7 +205,7 @@ nfsd_cache_search(struct svc_rqst *rqstp)
int
nfsd_cache_lookup(struct svc_rqst *rqstp)
{
- struct svc_cacherep *rp;
+ struct svc_cacherep *rp, *found;
__be32 xid = rqstp->rq_xid;
u32 proto = rqstp->rq_prot,
vers = rqstp->rq_vers,
@@ -210,38 +224,48 @@ nfsd_cache_lookup(struct svc_rqst *rqstp)
rtn = RC_DOIT;
rp = nfsd_cache_search(rqstp);
- if (rp) {
- nfsdstats.rchits++;
+ if (rp)
goto found_entry;
+
+ /* Try to use the first entry on the LRU */
+ if (!list_empty(&lru_head)) {
+ rp = list_first_entry(&lru_head, struct svc_cacherep, c_lru);
+ if (nfsd_cache_entry_expired(rp) ||
+ num_drc_entries >= max_drc_entries)
+ goto setup_entry;
}
- nfsdstats.rcmisses++;
- /* This loop shouldn't take more than a few iterations normally */
- {
- int safe = 0;
- list_for_each_entry(rp, &lru_head, c_lru) {
- if (rp->c_state != RC_INPROG)
- break;
- if (safe++ > CACHESIZE) {
- printk("nfsd: loop in repcache LRU list\n");
- cache_disabled = 1;
- goto out;
- }
+ spin_unlock(&cache_lock);
+ rp = nfsd_reply_cache_alloc();
+ if (!rp) {
+ dprintk("nfsd: unable to allocate DRC entry!\n");
+ return RC_DOIT;
}
+ spin_lock(&cache_lock);
+ ++num_drc_entries;
+
+ /*
+ * Must search again just in case someone inserted one
+ * after we dropped the lock above.
+ */
+ found = nfsd_cache_search(rqstp);
+ if (found) {
+ nfsd_reply_cache_free_locked(rp);
+ rp = found;
+ goto found_entry;
}
- /* All entries on the LRU are in-progress. This should not happen */
- if (&rp->c_lru == &lru_head) {
- static int complaints;
-
- printk(KERN_WARNING "nfsd: all repcache entries locked!\n");
- if (++complaints > 5) {
- printk(KERN_WARNING "nfsd: disabling repcache.\n");
- cache_disabled = 1;
- }
- goto out;
- }
+ /*
+ * We're keeping the one we just allocated. Are we now over the
+ * limit? Prune one off the tip of the LRU in trade for the one we
+ * just allocated if so.
+ */
+ if (num_drc_entries >= max_drc_entries)
+ nfsd_reply_cache_free_locked(list_first_entry(&lru_head,
+ struct svc_cacherep, c_lru));
+setup_entry:
+ nfsdstats.rcmisses++;
rqstp->rq_cacherep = rp;
rp->c_state = RC_INPROG;
rp->c_xid = xid;
@@ -265,6 +289,7 @@ nfsd_cache_lookup(struct svc_rqst *rqstp)
return rtn;
found_entry:
+ nfsdstats.rchits++;
/* We found a matching entry which is either in progress or done. */
age = jiffies - rp->c_timestamp;
lru_put_end(rp);
@@ -295,7 +320,7 @@ found_entry:
break;
default:
printk(KERN_WARNING "nfsd: bad repcache type %d\n", rp->c_type);
- rp->c_state = RC_UNUSED;
+ nfsd_reply_cache_free_locked(rp);
}
goto out;
OpenPOWER on IntegriCloud