lightnvm: pblk: fix race condition on GC

This patch fixes a race condition where a write is mapped to the last
sectors of a line. The write is synced to the device but the L2P is not
updated yet. When the line is garbage collected before the L2P update
is performed, the sectors are ignored by the GC logic and the line is
freed before all sectors are moved. When the L2P is finally updated, it
contains a mapping to a freed line, subsequent reads of the
corresponding LBAs fail.

This patch introduces a per line counter specifying the number of
sectors that are synced to the device but have not been updated in the
L2P. Lines with a counter of greater than zero will not be selected
for GC.

Signed-off-by: Heiner Litz <hlitz@ucsc.edu>
Reviewed-by: Hans Holmberg <hans.holmberg@cnexlabs.com>
Reviewed-by: Javier González <javier@javigon.com>
Signed-off-by: Matias Bjørling <mb@lightnvm.io>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

drivers/lightnvm/pblk-core.c

@@ -1278,6 +1278,7 @@ static int pblk_line_prepare(struct pblk *pblk, struct pblk_line *line)
 	spin_unlock(&line->lock);
 
 	kref_init(&line->ref);
+	atomic_set(&line->sec_to_update, 0);
 
 	return 0;
 }

drivers/lightnvm/pblk-gc.c

@@ -365,16 +365,22 @@ static struct pblk_line *pblk_gc_get_victim_line(struct pblk *pblk,
 						 struct list_head *group_list)
 {
 	struct pblk_line *line, *victim;
-	int line_vsc, victim_vsc;
+	unsigned int line_vsc = ~0x0L, victim_vsc = ~0x0L;
 
 	victim = list_first_entry(group_list, struct pblk_line, list);
+
 	list_for_each_entry(line, group_list, list) {
-		line_vsc = le32_to_cpu(*line->vsc);
+		if (!atomic_read(&line->sec_to_update))
-		victim_vsc = le32_to_cpu(*victim->vsc);
+			line_vsc = le32_to_cpu(*line->vsc);
-		if (line_vsc < victim_vsc)
+		if (line_vsc < victim_vsc) {
 			victim = line;
+			victim_vsc = le32_to_cpu(*victim->vsc);
+		}
 	}
 
+	if (victim_vsc == ~0x0)
+		return NULL;
+
 	return victim;
 }
 
@@ -448,13 +454,13 @@ next_gc_group:
 
 	do {
 		spin_lock(&l_mg->gc_lock);
-		if (list_empty(group_list)) {
+
+		line = pblk_gc_get_victim_line(pblk, group_list);
+		if (!line) {
 			spin_unlock(&l_mg->gc_lock);
 			break;
 		}
 
-		line = pblk_gc_get_victim_line(pblk, group_list);
-
 		spin_lock(&line->lock);
 		WARN_ON(line->state != PBLK_LINESTATE_CLOSED);
 		line->state = PBLK_LINESTATE_GC;

drivers/lightnvm/pblk-map.c

@@ -73,6 +73,7 @@ static int pblk_map_page_data(struct pblk *pblk, unsigned int sentry,
 		 */
 		if (i < valid_secs) {
 			kref_get(&line->ref);
+			atomic_inc(&line->sec_to_update);
 			w_ctx = pblk_rb_w_ctx(&pblk->rwb, sentry + i);
 			w_ctx->ppa = ppa_list[i];
 			meta->lba = cpu_to_le64(w_ctx->lba);

drivers/lightnvm/pblk-rb.c

@@ -260,6 +260,7 @@ static int __pblk_rb_update_l2p(struct pblk_rb *rb, unsigned int to_update)
 							entry->cacheline);
 
 		line = pblk_ppa_to_line(pblk, w_ctx->ppa);
+		atomic_dec(&line->sec_to_update);
 		kref_put(&line->ref, pblk_line_put);
 		clean_wctx(w_ctx);
 		rb->l2p_update = pblk_rb_ptr_wrap(rb, rb->l2p_update, 1);

drivers/lightnvm/pblk-write.c

@@ -177,6 +177,7 @@ static void pblk_prepare_resubmit(struct pblk *pblk, unsigned int sentry,
 		 * re-map these entries
 		 */
 		line = pblk_ppa_to_line(pblk, w_ctx->ppa);
+		atomic_dec(&line->sec_to_update);
 		kref_put(&line->ref, pblk_line_put);
 	}
 	spin_unlock(&pblk->trans_lock);

drivers/lightnvm/pblk.h

@@ -487,6 +487,7 @@ struct pblk_line {
 	__le32 *vsc;			/* Valid sector count in line */
 
 	struct kref ref;		/* Write buffer L2P references */
+	atomic_t sec_to_update;         /* Outstanding L2P updates to ppa */
 
 	struct pblk_w_err_gc *w_err_gc;	/* Write error gc recovery metadata */