X-Git-Url: https://git.camperquake.de/gitweb.cgi?a=blobdiff_plain;f=module%2Fzfs%2Fvdev_cache.c;h=bf4ae7b2bc3351c2c6ecd8907d9dd8e00c9954b9;hb=refs%2Fheads%2Frertzinger%2Ffeature-zpool-get--p;hp=9b3a9f5a26031e14b10d6c0e8c1d115d95350132;hpb=d164b2093561a9771db07346e6fffc9ca19427a2;p=zfs.git

diff --git a/module/zfs/vdev_cache.c b/module/zfs/vdev_cache.c
index 9b3a9f5..bf4ae7b 100644
--- a/module/zfs/vdev_cache.c
+++ b/module/zfs/vdev_cache.c
@@ -71,9 +71,16 @@
  * 1<<zfs_vdev_cache_bshift byte reads by the vdev_cache (aka software
  * track buffer).  At most zfs_vdev_cache_size bytes will be kept in each
  * vdev's vdev_cache.
+ *
+ * TODO: Note that with the current ZFS code, it turns out that the
+ * vdev cache is not helpful, and in some cases actually harmful.  It
+ * is better if we disable this.  Once some time has passed, we should
+ * actually remove this to simplify the code.  For now we just disable
+ * it by setting the zfs_vdev_cache_size to zero.  Note that Solaris 11
+ * has made these same changes.
  */
 int zfs_vdev_cache_max = 1<<14;			/* 16KB */
-int zfs_vdev_cache_size = 10ULL << 20;		/* 10MB */
+int zfs_vdev_cache_size = 0;
 int zfs_vdev_cache_bshift = 16;
 
 #define	VCBS (1 << zfs_vdev_cache_bshift)	/* 64KB */
@@ -170,9 +177,9 @@ vdev_cache_allocate(zio_t *zio)
 		vdev_cache_evict(vc, ve);
 	}
 
-	ve = kmem_zalloc(sizeof (vdev_cache_entry_t), KM_SLEEP);
+	ve = kmem_zalloc(sizeof (vdev_cache_entry_t), KM_PUSHPAGE);
 	ve->ve_offset = offset;
-	ve->ve_lastused = lbolt;
+	ve->ve_lastused = ddi_get_lbolt();
 	ve->ve_data = zio_buf_alloc(VCBS);
 
 	avl_add(&vc->vc_offset_tree, ve);
@@ -189,9 +196,9 @@ vdev_cache_hit(vdev_cache_t *vc, vdev_cache_entry_t *ve, zio_t *zio)
 	ASSERT(MUTEX_HELD(&vc->vc_lock));
 	ASSERT(ve->ve_fill_io == NULL);
 
-	if (ve->ve_lastused != lbolt) {
+	if (ve->ve_lastused != ddi_get_lbolt()) {
 		avl_remove(&vc->vc_lastused_tree, ve);
-		ve->ve_lastused = lbolt;
+		ve->ve_lastused = ddi_get_lbolt();
 		avl_add(&vc->vc_lastused_tree, ve);
 	}
 
@@ -244,9 +251,9 @@ int
 vdev_cache_read(zio_t *zio)
 {
 	vdev_cache_t *vc = &zio->io_vd->vdev_cache;
-	vdev_cache_entry_t *ve, ve_search;
+	vdev_cache_entry_t *ve, *ve_search;
 	uint64_t cache_offset = P2ALIGN(zio->io_offset, VCBS);
-	uint64_t cache_phase = P2PHASE(zio->io_offset, VCBS);
+	ASSERTV(uint64_t cache_phase = P2PHASE(zio->io_offset, VCBS);)
 	zio_t *fio;
 
 	ASSERT(zio->io_type == ZIO_TYPE_READ);
@@ -267,8 +274,10 @@ vdev_cache_read(zio_t *zio)
 
 	mutex_enter(&vc->vc_lock);
 
-	ve_search.ve_offset = cache_offset;
-	ve = avl_find(&vc->vc_offset_tree, &ve_search, NULL);
+	ve_search = kmem_alloc(sizeof(vdev_cache_entry_t), KM_PUSHPAGE);
+	ve_search->ve_offset = cache_offset;
+	ve = avl_find(&vc->vc_offset_tree, ve_search, NULL);
+	kmem_free(ve_search, sizeof(vdev_cache_entry_t));
 
 	if (ve != NULL) {
 		if (ve->ve_missed_update) {
@@ -414,3 +423,14 @@ vdev_cache_stat_fini(void)
 		vdc_ksp = NULL;
 	}
 }
+
+#if defined(_KERNEL) && defined(HAVE_SPL)
+module_param(zfs_vdev_cache_max, int, 0644);
+MODULE_PARM_DESC(zfs_vdev_cache_max, "Inflate reads small than max");
+
+module_param(zfs_vdev_cache_size, int, 0444);
+MODULE_PARM_DESC(zfs_vdev_cache_size, "Total size of the per-disk cache");
+
+module_param(zfs_vdev_cache_bshift, int, 0644);
+MODULE_PARM_DESC(zfs_vdev_cache_bshift, "Shift size to inflate reads too");
+#endif