DLPX-84995 NFSD: Never call nfsd_file_gc() in foreground paths (#25)

The checks in nfsd_file_acquire() and nfsd_file_put() that directly
invoke filecache garbage collection are intended to keep cache
occupancy between a low- and high-watermark. The reason to limit the
capacity of the filecache is to keep filecache lookups reasonably
fast.

However, invoking garbage collection at those points has some
undesirable negative impacts. Files that are held open by NFSv4
clients often push the occupancy of the filecache over these
watermarks. At that point:

- Every call to nfsd_file_acquire() and nfsd_file_put() results in
  an LRU walk. This has the same effect on lookup latency as long
  chains in the hash table.
- Garbage collection will then run on every nfsd thread, causing a
  lot of unnecessary lock contention.
- Limiting cache capacity pushes out files used only by NFSv3
  clients, which are the type of files the filecache is supposed to
  help.

To address those negative impacts, remove the direct calls to the
garbage collector.

Author: Don Brady

fs/nfsd/filecache.c

@@ -28,8 +28,6 @@
 #define NFSD_LAUNDRETTE_DELAY		     (2 * HZ)
 
 #define NFSD_FILE_SHUTDOWN		     (1)
-#define NFSD_FILE_LRU_THRESHOLD		     (4096UL)
-#define NFSD_FILE_LRU_LIMIT		     (NFSD_FILE_LRU_THRESHOLD << 2)
 
 /* We only care about NFSD_MAY_READ/WRITE for this cache */
 #define NFSD_FILE_MAY_MASK	(NFSD_MAY_READ|NFSD_MAY_WRITE)
@@ -65,8 +63,6 @@ static struct delayed_work		nfsd_filecache_laundrette;
 static DEFINE_SPINLOCK(laundrette_lock);
 static LIST_HEAD(laundrettes);
 
-static void nfsd_file_gc(void);
-
 static void
 nfsd_file_schedule_laundrette(void)
 {
@@ -306,8 +302,6 @@ nfsd_file_put(struct nfsd_file *nf)
 	nfsd_file_put_noref(nf);
 	if (is_hashed)
 		nfsd_file_schedule_laundrette();
-	if (atomic_long_read(&nfsd_filecache_count) >= NFSD_FILE_LRU_LIMIT)
-		nfsd_file_gc();
 }
 
 struct nfsd_file *
@@ -1007,8 +1001,7 @@ nfsd_file_acquire(struct svc_rqst *rqstp, struct svc_fh *fhp,
 	nfsd_file_hashtbl[hashval].nfb_maxcount = max(nfsd_file_hashtbl[hashval].nfb_maxcount,
 			nfsd_file_hashtbl[hashval].nfb_count);
 	spin_unlock(&nfsd_file_hashtbl[hashval].nfb_lock);
-	if (atomic_long_inc_return(&nfsd_filecache_count) >= NFSD_FILE_LRU_THRESHOLD)
-		nfsd_file_gc();
+	atomic_long_inc(&nfsd_filecache_count);
 
 	nf->nf_mark = nfsd_file_mark_find_or_create(nf);
 	if (nf->nf_mark)