lightnvm: move bad block and chunk state logic to core

pblk implements two data paths for recovery line state. One for 1.2
and another for 2.0, instead of having pblk implement these, combine
them in the core to reduce complexity and make available to other
targets.

The new interface will adhere to the 2.0 chunk definition,
including managing open chunks with an active write pointer. To provide
this interface, a 1.2 device recovers the state of the chunks by
manually detecting if a chunk is either free/open/close/offline, and if
open, scanning the flash pages sequentially to find the next writeable
page. This process takes on average ~10 seconds on a device with 64 dies,
1024 blocks and 60us read access time. The process can be parallelized
but is left out for maintenance simplicity, as the 1.2 specification is
deprecated. For 2.0 devices, the logic is maintained internally in the
drive and retrieved through the 2.0 interface.

Signed-off-by: Matias Bjørling <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>

Author: MatiasBjorling

drivers/lightnvm/core.c

@@ -717,46 +717,6 @@ static void nvm_free_rqd_ppalist(struct nvm_tgt_dev *tgt_dev,
 	nvm_dev_dma_free(tgt_dev->parent, rqd->ppa_list, rqd->dma_ppa_list);
 }
 
-int nvm_get_chunk_meta(struct nvm_tgt_dev *tgt_dev, struct nvm_chk_meta *meta,
-		struct ppa_addr ppa, int nchks)
-{
-	struct nvm_dev *dev = tgt_dev->parent;
-
-	nvm_ppa_tgt_to_dev(tgt_dev, &ppa, 1);
-
-	return dev->ops->get_chk_meta(tgt_dev->parent, meta,
-						(sector_t)ppa.ppa, nchks);
-}
-EXPORT_SYMBOL(nvm_get_chunk_meta);
-
-int nvm_set_tgt_bb_tbl(struct nvm_tgt_dev *tgt_dev, struct ppa_addr *ppas,
-		       int nr_ppas, int type)
-{
-	struct nvm_dev *dev = tgt_dev->parent;
-	struct nvm_rq rqd;
-	int ret;
-
-	if (nr_ppas > NVM_MAX_VLBA) {
-		pr_err("nvm: unable to update all blocks atomically\n");
-		return -EINVAL;
-	}
-
-	memset(&rqd, 0, sizeof(struct nvm_rq));
-
-	nvm_set_rqd_ppalist(tgt_dev, &rqd, ppas, nr_ppas);
-	nvm_rq_tgt_to_dev(tgt_dev, &rqd);
-
-	ret = dev->ops->set_bb_tbl(dev, &rqd.ppa_addr, rqd.nr_ppas, type);
-	nvm_free_rqd_ppalist(tgt_dev, &rqd);
-	if (ret) {
-		pr_err("nvm: failed bb mark\n");
-		return -EINVAL;
-	}
-
-	return 0;
-}
-EXPORT_SYMBOL(nvm_set_tgt_bb_tbl);
-
 static int nvm_set_flags(struct nvm_geo *geo, struct nvm_rq *rqd)
 {
 	int flags = 0;
@@ -830,27 +790,159 @@ void nvm_end_io(struct nvm_rq *rqd)
 }
 EXPORT_SYMBOL(nvm_end_io);
 
+static int nvm_submit_io_sync_raw(struct nvm_dev *dev, struct nvm_rq *rqd)
+{
+	if (!dev->ops->submit_io_sync)
+		return -ENODEV;
+
+	rqd->flags = nvm_set_flags(&dev->geo, rqd);
+
+	return dev->ops->submit_io_sync(dev, rqd);
+}
+
+static int nvm_bb_chunk_sense(struct nvm_dev *dev, struct ppa_addr ppa)
+{
+	struct nvm_rq rqd = { NULL };
+	struct bio bio;
+	struct bio_vec bio_vec;
+	struct page *page;
+	int ret;
+
+	page = alloc_page(GFP_KERNEL);
+	if (!page)
+		return -ENOMEM;
+
+	bio_init(&bio, &bio_vec, 1);
+	bio_add_page(&bio, page, PAGE_SIZE, 0);
+	bio_set_op_attrs(&bio, REQ_OP_READ, 0);
+
+	rqd.bio = &bio;
+	rqd.opcode = NVM_OP_PREAD;
+	rqd.is_seq = 1;
+	rqd.nr_ppas = 1;
+	rqd.ppa_addr = generic_to_dev_addr(dev, ppa);
+
+	ret = nvm_submit_io_sync_raw(dev, &rqd);
+	if (ret)
+		return ret;
+
+	__free_page(page);
+
+	return rqd.error;
+}
+
 /*
- * folds a bad block list from its plane representation to its virtual
- * block representation. The fold is done in place and reduced size is
- * returned.
- *
- * If any of the planes status are bad or grown bad block, the virtual block
- * is marked bad. If not bad, the first plane state acts as the block state.
+ * Scans a 1.2 chunk first and last page to determine if its state.
+ * If the chunk is found to be open, also scan it to update the write
+ * pointer.
  */
-int nvm_bb_tbl_fold(struct nvm_dev *dev, u8 *blks, int nr_blks)
+static int nvm_bb_chunk_scan(struct nvm_dev *dev, struct ppa_addr ppa,
+			     struct nvm_chk_meta *meta)
 {
 	struct nvm_geo *geo = &dev->geo;
-	int blk, offset, pl, blktype;
+	int ret, pg, pl;
 
-	if (nr_blks != geo->num_chk * geo->pln_mode)
-		return -EINVAL;
+	/* sense first page */
+	ret = nvm_bb_chunk_sense(dev, ppa);
+	if (ret < 0) /* io error */
+		return ret;
+	else if (ret == 0) /* valid data */
+		meta->state = NVM_CHK_ST_OPEN;
+	else if (ret > 0) {
+		/*
+		 * If empty page, the chunk is free, else it is an
+		 * actual io error. In that case, mark it offline.
+		 */
+		switch (ret) {
+		case NVM_RSP_ERR_EMPTYPAGE:
+			meta->state = NVM_CHK_ST_FREE;
+			return 0;
+		case NVM_RSP_ERR_FAILCRC:
+		case NVM_RSP_ERR_FAILECC:
+		case NVM_RSP_WARN_HIGHECC:
+			meta->state = NVM_CHK_ST_OPEN;
+			goto scan;
+		default:
+			return -ret; /* other io error */
+		}
+	}
+
+	/* sense last page */
+	ppa.g.pg = geo->num_pg - 1;
+	ppa.g.pl = geo->num_pln - 1;
+
+	ret = nvm_bb_chunk_sense(dev, ppa);
+	if (ret < 0) /* io error */
+		return ret;
+	else if (ret == 0) { /* Chunk fully written */
+		meta->state = NVM_CHK_ST_CLOSED;
+		meta->wp = geo->clba;
+		return 0;
+	} else if (ret > 0) {
+		switch (ret) {
+		case NVM_RSP_ERR_EMPTYPAGE:
+		case NVM_RSP_ERR_FAILCRC:
+		case NVM_RSP_ERR_FAILECC:
+		case NVM_RSP_WARN_HIGHECC:
+			meta->state = NVM_CHK_ST_OPEN;
+			break;
+		default:
+			return -ret; /* other io error */
+		}
+	}
+
+scan:
+	/*
+	 * chunk is open, we scan sequentially to update the write pointer.
+	 * We make the assumption that targets write data across all planes
+	 * before moving to the next page.
+	 */
+	for (pg = 0; pg < geo->num_pg; pg++) {
+		for (pl = 0; pl < geo->num_pln; pl++) {
+			ppa.g.pg = pg;
+			ppa.g.pl = pl;
+
+			ret = nvm_bb_chunk_sense(dev, ppa);
+			if (ret < 0) /* io error */
+				return ret;
+			else if (ret == 0) {
+				meta->wp += geo->ws_min;
+			} else if (ret > 0) {
+				switch (ret) {
+				case NVM_RSP_ERR_EMPTYPAGE:
+					return 0;
+				case NVM_RSP_ERR_FAILCRC:
+				case NVM_RSP_ERR_FAILECC:
+				case NVM_RSP_WARN_HIGHECC:
+					meta->wp += geo->ws_min;
+					break;
+				default:
+					return -ret; /* other io error */
+				}
+			}
+		}
+	}
+
+	return 0;
+}
+
+/*
+ * folds a bad block list from its plane representation to its
+ * chunk representation.
+ *
+ * If any of the planes status are bad or grown bad, the chunk is marked
+ * offline. If not bad, the first plane state acts as the chunk state.
+ */
+static int nvm_bb_to_chunk(struct nvm_dev *dev, struct ppa_addr ppa,
+			   u8 *blks, int nr_blks, struct nvm_chk_meta *meta)
+{
+	struct nvm_geo *geo = &dev->geo;
+	int ret, blk, pl, offset, blktype;
 
 	for (blk = 0; blk < geo->num_chk; blk++) {
 		offset = blk * geo->pln_mode;
 		blktype = blks[offset];
 
-		/* Bad blocks on any planes take precedence over other types */
 		for (pl = 0; pl < geo->pln_mode; pl++) {
 			if (blks[offset + pl] &
 					(NVM_BLK_T_BAD|NVM_BLK_T_GRWN_BAD)) {
@@ -859,23 +951,124 @@ int nvm_bb_tbl_fold(struct nvm_dev *dev, u8 *blks, int nr_blks)
 			}
 		}
 
-		blks[blk] = blktype;
+		ppa.g.blk = blk;
+
+		meta->wp = 0;
+		meta->type = NVM_CHK_TP_W_SEQ;
+		meta->wi = 0;
+		meta->slba = generic_to_dev_addr(dev, ppa).ppa;
+		meta->cnlb = dev->geo.clba;
+
+		if (blktype == NVM_BLK_T_FREE) {
+			ret = nvm_bb_chunk_scan(dev, ppa, meta);
+			if (ret)
+				return ret;
+		} else {
+			meta->state = NVM_CHK_ST_OFFLINE;
+		}
+
+		meta++;
 	}
 
-	return geo->num_chk;
+	return 0;
+}
+
+static int nvm_get_bb_meta(struct nvm_dev *dev, sector_t slba,
+			   int nchks, struct nvm_chk_meta *meta)
+{
+	struct nvm_geo *geo = &dev->geo;
+	struct ppa_addr ppa;
+	u8 *blks;
+	int ch, lun, nr_blks;
+	int ret;
+
+	ppa.ppa = slba;
+	ppa = dev_to_generic_addr(dev, ppa);
+
+	if (ppa.g.blk != 0)
+		return -EINVAL;
+
+	if ((nchks % geo->num_chk) != 0)
+		return -EINVAL;
+
+	nr_blks = geo->num_chk * geo->pln_mode;
+
+	blks = kmalloc(nr_blks, GFP_KERNEL);
+	if (!blks)
+		return -ENOMEM;
+
+	for (ch = ppa.g.ch; ch < geo->num_ch; ch++) {
+		for (lun = ppa.g.lun; lun < geo->num_lun; lun++) {
+			struct ppa_addr ppa_gen, ppa_dev;
+
+			if (!nchks)
+				goto done;
+
+			ppa_gen.ppa = 0;
+			ppa_gen.g.ch = ch;
+			ppa_gen.g.lun = lun;
+			ppa_dev = generic_to_dev_addr(dev, ppa_gen);
+
+			ret = dev->ops->get_bb_tbl(dev, ppa_dev, blks);
+			if (ret)
+				goto done;
+
+			ret = nvm_bb_to_chunk(dev, ppa_gen, blks, nr_blks,
+									meta);
+			if (ret)
+				goto done;
+
+			meta += geo->num_chk;
+			nchks -= geo->num_chk;
+		}
+	}
+done:
+	kfree(blks);
+	return ret;
 }
-EXPORT_SYMBOL(nvm_bb_tbl_fold);
 
-int nvm_get_tgt_bb_tbl(struct nvm_tgt_dev *tgt_dev, struct ppa_addr ppa,
-		       u8 *blks)
+int nvm_get_chunk_meta(struct nvm_tgt_dev *tgt_dev, struct ppa_addr ppa,
+		       int nchks, struct nvm_chk_meta *meta)
 {
 	struct nvm_dev *dev = tgt_dev->parent;
 
 	nvm_ppa_tgt_to_dev(tgt_dev, &ppa, 1);
 
-	return dev->ops->get_bb_tbl(dev, ppa, blks);
+	if (dev->geo.version == NVM_OCSSD_SPEC_12)
+		return nvm_get_bb_meta(dev, (sector_t)ppa.ppa, nchks, meta);
+
+	return dev->ops->get_chk_meta(dev, (sector_t)ppa.ppa, nchks, meta);
+}
+EXPORT_SYMBOL_GPL(nvm_get_chunk_meta);
+
+int nvm_set_chunk_meta(struct nvm_tgt_dev *tgt_dev, struct ppa_addr *ppas,
+		       int nr_ppas, int type)
+{
+	struct nvm_dev *dev = tgt_dev->parent;
+	struct nvm_rq rqd;
+	int ret;
+
+	if (dev->geo.version == NVM_OCSSD_SPEC_20)
+		return 0;
+
+	if (nr_ppas > NVM_MAX_VLBA) {
+		pr_err("nvm: unable to update all blocks atomically\n");
+		return -EINVAL;
+	}
+
+	memset(&rqd, 0, sizeof(struct nvm_rq));
+
+	nvm_set_rqd_ppalist(tgt_dev, &rqd, ppas, nr_ppas);
+	nvm_rq_tgt_to_dev(tgt_dev, &rqd);
+
+	ret = dev->ops->set_bb_tbl(dev, &rqd.ppa_addr, rqd.nr_ppas, type);
+	nvm_free_rqd_ppalist(tgt_dev, &rqd);
+	if (ret)
+		return -EINVAL;
+
+	return 0;
 }
-EXPORT_SYMBOL(nvm_get_tgt_bb_tbl);
+EXPORT_SYMBOL_GPL(nvm_set_chunk_meta);
 
 static int nvm_core_init(struct nvm_dev *dev)
 {

drivers/lightnvm/pblk-core.c

@@ -27,7 +27,7 @@ static void pblk_line_mark_bb(struct work_struct *work)
 	struct ppa_addr *ppa = line_ws->priv;
 	int ret;
 
-	ret = nvm_set_tgt_bb_tbl(dev, ppa, 1, NVM_BLK_T_GRWN_BAD);
+	ret = nvm_set_chunk_meta(dev, ppa, 1, NVM_BLK_T_GRWN_BAD);
 	if (ret) {
 		struct pblk_line *line;
 		int pos;
@@ -110,7 +110,7 @@ static void pblk_end_io_erase(struct nvm_rq *rqd)
  *
  * The caller is responsible for freeing the returned structure
  */
-struct nvm_chk_meta *pblk_chunk_get_info(struct pblk *pblk)
+struct nvm_chk_meta *pblk_get_chunk_meta(struct pblk *pblk)
 {
 	struct nvm_tgt_dev *dev = pblk->dev;
 	struct nvm_geo *geo = &dev->geo;
@@ -126,7 +126,7 @@ struct nvm_chk_meta *pblk_chunk_get_info(struct pblk *pblk)
 	if (!meta)
 		return ERR_PTR(-ENOMEM);
 
-	ret = nvm_get_chunk_meta(dev, meta, ppa, geo->all_chunks);
+	ret = nvm_get_chunk_meta(dev, ppa, geo->all_chunks, meta);
 	if (ret) {
 		kfree(meta);
 		return ERR_PTR(-EIO);