lightnvm: remove rrpc
authorMatias Bjørling <m@bjorling.me>
Fri, 5 Jan 2018 13:15:58 +0000 (14:15 +0100)
committerJens Axboe <axboe@kernel.dk>
Fri, 5 Jan 2018 15:50:12 +0000 (08:50 -0700)
The hybrid mode for 1.2 revision was deprecated, and have
no users. Remove to make it easier to move to the 2.0 revision.

Signed-off-by: Matias Bjørling <m@bjorling.me>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
drivers/lightnvm/Kconfig
drivers/lightnvm/Makefile
drivers/lightnvm/rrpc.c [deleted file]
drivers/lightnvm/rrpc.h [deleted file]

index 2a953efec4e196b8bbac2203845954e2a3163fe6..10c08982185a572ff05683461d514e86c8920f96 100644 (file)
@@ -27,13 +27,6 @@ config NVM_DEBUG
 
        It is required to create/remove targets without IOCTLs.
 
-config NVM_RRPC
-       tristate "Round-robin Hybrid Open-Channel SSD target"
-       ---help---
-       Allows an open-channel SSD to be exposed as a block device to the
-       host. The target is implemented using a linear mapping table and
-       cost-based garbage collection. It is optimized for 4K IO sizes.
-
 config NVM_PBLK
        tristate "Physical Block Device Open-Channel SSD target"
        ---help---
index 2c3fd9d2c08c5e03477677eac57e188ed4fda427..97d9d7c71550cac4bbccba965ffee5df1b0c7cd8 100644 (file)
@@ -4,7 +4,6 @@
 #
 
 obj-$(CONFIG_NVM)              := core.o
-obj-$(CONFIG_NVM_RRPC)         += rrpc.o
 obj-$(CONFIG_NVM_PBLK)         += pblk.o
 pblk-y                         := pblk-init.o pblk-core.o pblk-rb.o \
                                   pblk-write.o pblk-cache.o pblk-read.o \
diff --git a/drivers/lightnvm/rrpc.c b/drivers/lightnvm/rrpc.c
deleted file mode 100644 (file)
index 0993c14..0000000
+++ /dev/null
@@ -1,1625 +0,0 @@
-/*
- * Copyright (C) 2015 IT University of Copenhagen
- * Initial release: Matias Bjorling <m@bjorling.me>
- *
- * This program is free software; you can redistribute it and/or
- * modify it under the terms of the GNU General Public License version
- * 2 as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License for more details.
- *
- * Implementation of a Round-robin page-based Hybrid FTL for Open-channel SSDs.
- */
-
-#include "rrpc.h"
-
-static struct kmem_cache *rrpc_gcb_cache, *rrpc_rq_cache;
-static DECLARE_RWSEM(rrpc_lock);
-
-static int rrpc_submit_io(struct rrpc *rrpc, struct bio *bio,
-                               struct nvm_rq *rqd, unsigned long flags);
-
-#define rrpc_for_each_lun(rrpc, rlun, i) \
-               for ((i) = 0, rlun = &(rrpc)->luns[0]; \
-                       (i) < (rrpc)->nr_luns; (i)++, rlun = &(rrpc)->luns[(i)])
-
-static void rrpc_page_invalidate(struct rrpc *rrpc, struct rrpc_addr *a)
-{
-       struct nvm_tgt_dev *dev = rrpc->dev;
-       struct rrpc_block *rblk = a->rblk;
-       unsigned int pg_offset;
-
-       lockdep_assert_held(&rrpc->rev_lock);
-
-       if (a->addr == ADDR_EMPTY || !rblk)
-               return;
-
-       spin_lock(&rblk->lock);
-
-       div_u64_rem(a->addr, dev->geo.sec_per_blk, &pg_offset);
-       WARN_ON(test_and_set_bit(pg_offset, rblk->invalid_pages));
-       rblk->nr_invalid_pages++;
-
-       spin_unlock(&rblk->lock);
-
-       rrpc->rev_trans_map[a->addr].addr = ADDR_EMPTY;
-}
-
-static void rrpc_invalidate_range(struct rrpc *rrpc, sector_t slba,
-                                                       unsigned int len)
-{
-       sector_t i;
-
-       spin_lock(&rrpc->rev_lock);
-       for (i = slba; i < slba + len; i++) {
-               struct rrpc_addr *gp = &rrpc->trans_map[i];
-
-               rrpc_page_invalidate(rrpc, gp);
-               gp->rblk = NULL;
-       }
-       spin_unlock(&rrpc->rev_lock);
-}
-
-static struct nvm_rq *rrpc_inflight_laddr_acquire(struct rrpc *rrpc,
-                                       sector_t laddr, unsigned int pages)
-{
-       struct nvm_rq *rqd;
-       struct rrpc_inflight_rq *inf;
-
-       rqd = mempool_alloc(rrpc->rq_pool, GFP_ATOMIC);
-       if (!rqd)
-               return ERR_PTR(-ENOMEM);
-
-       inf = rrpc_get_inflight_rq(rqd);
-       if (rrpc_lock_laddr(rrpc, laddr, pages, inf)) {
-               mempool_free(rqd, rrpc->rq_pool);
-               return NULL;
-       }
-
-       return rqd;
-}
-
-static void rrpc_inflight_laddr_release(struct rrpc *rrpc, struct nvm_rq *rqd)
-{
-       struct rrpc_inflight_rq *inf = rrpc_get_inflight_rq(rqd);
-
-       rrpc_unlock_laddr(rrpc, inf);
-
-       mempool_free(rqd, rrpc->rq_pool);
-}
-
-static void rrpc_discard(struct rrpc *rrpc, struct bio *bio)
-{
-       sector_t slba = bio->bi_iter.bi_sector / NR_PHY_IN_LOG;
-       sector_t len = bio->bi_iter.bi_size / RRPC_EXPOSED_PAGE_SIZE;
-       struct nvm_rq *rqd;
-
-       while (1) {
-               rqd = rrpc_inflight_laddr_acquire(rrpc, slba, len);
-               if (rqd)
-                       break;
-
-               schedule();
-       }
-
-       if (IS_ERR(rqd)) {
-               pr_err("rrpc: unable to acquire inflight IO\n");
-               bio_io_error(bio);
-               return;
-       }
-
-       rrpc_invalidate_range(rrpc, slba, len);
-       rrpc_inflight_laddr_release(rrpc, rqd);
-}
-
-static int block_is_full(struct rrpc *rrpc, struct rrpc_block *rblk)
-{
-       struct nvm_tgt_dev *dev = rrpc->dev;
-
-       return (rblk->next_page == dev->geo.sec_per_blk);
-}
-
-/* Calculate relative addr for the given block, considering instantiated LUNs */
-static u64 block_to_rel_addr(struct rrpc *rrpc, struct rrpc_block *rblk)
-{
-       struct nvm_tgt_dev *dev = rrpc->dev;
-       struct rrpc_lun *rlun = rblk->rlun;
-
-       return rlun->id * dev->geo.sec_per_blk;
-}
-
-static struct ppa_addr rrpc_ppa_to_gaddr(struct nvm_tgt_dev *dev,
-                                        struct rrpc_addr *gp)
-{
-       struct rrpc_block *rblk = gp->rblk;
-       struct rrpc_lun *rlun = rblk->rlun;
-       u64 addr = gp->addr;
-       struct ppa_addr paddr;
-
-       paddr.ppa = addr;
-       paddr = rrpc_linear_to_generic_addr(&dev->geo, paddr);
-       paddr.g.ch = rlun->bppa.g.ch;
-       paddr.g.lun = rlun->bppa.g.lun;
-       paddr.g.blk = rblk->id;
-
-       return paddr;
-}
-
-/* requires lun->lock taken */
-static void rrpc_set_lun_cur(struct rrpc_lun *rlun, struct rrpc_block *new_rblk,
-                                               struct rrpc_block **cur_rblk)
-{
-       struct rrpc *rrpc = rlun->rrpc;
-
-       if (*cur_rblk) {
-               spin_lock(&(*cur_rblk)->lock);
-               WARN_ON(!block_is_full(rrpc, *cur_rblk));
-               spin_unlock(&(*cur_rblk)->lock);
-       }
-       *cur_rblk = new_rblk;
-}
-
-static struct rrpc_block *__rrpc_get_blk(struct rrpc *rrpc,
-                                                       struct rrpc_lun *rlun)
-{
-       struct rrpc_block *rblk = NULL;
-
-       if (list_empty(&rlun->free_list))
-               goto out;
-
-       rblk = list_first_entry(&rlun->free_list, struct rrpc_block, list);
-
-       list_move_tail(&rblk->list, &rlun->used_list);
-       rblk->state = NVM_BLK_ST_TGT;
-       rlun->nr_free_blocks--;
-
-out:
-       return rblk;
-}
-
-static struct rrpc_block *rrpc_get_blk(struct rrpc *rrpc, struct rrpc_lun *rlun,
-                                                       unsigned long flags)
-{
-       struct nvm_tgt_dev *dev = rrpc->dev;
-       struct rrpc_block *rblk;
-       int is_gc = flags & NVM_IOTYPE_GC;
-
-       spin_lock(&rlun->lock);
-       if (!is_gc && rlun->nr_free_blocks < rlun->reserved_blocks) {
-               pr_err("nvm: rrpc: cannot give block to non GC request\n");
-               spin_unlock(&rlun->lock);
-               return NULL;
-       }
-
-       rblk = __rrpc_get_blk(rrpc, rlun);
-       if (!rblk) {
-               pr_err("nvm: rrpc: cannot get new block\n");
-               spin_unlock(&rlun->lock);
-               return NULL;
-       }
-       spin_unlock(&rlun->lock);
-
-       bitmap_zero(rblk->invalid_pages, dev->geo.sec_per_blk);
-       rblk->next_page = 0;
-       rblk->nr_invalid_pages = 0;
-       atomic_set(&rblk->data_cmnt_size, 0);
-
-       return rblk;
-}
-
-static void rrpc_put_blk(struct rrpc *rrpc, struct rrpc_block *rblk)
-{
-       struct rrpc_lun *rlun = rblk->rlun;
-
-       spin_lock(&rlun->lock);
-       if (rblk->state & NVM_BLK_ST_TGT) {
-               list_move_tail(&rblk->list, &rlun->free_list);
-               rlun->nr_free_blocks++;
-               rblk->state = NVM_BLK_ST_FREE;
-       } else if (rblk->state & NVM_BLK_ST_BAD) {
-               list_move_tail(&rblk->list, &rlun->bb_list);
-               rblk->state = NVM_BLK_ST_BAD;
-       } else {
-               WARN_ON_ONCE(1);
-               pr_err("rrpc: erroneous type (ch:%d,lun:%d,blk%d-> %u)\n",
-                                       rlun->bppa.g.ch, rlun->bppa.g.lun,
-                                       rblk->id, rblk->state);
-               list_move_tail(&rblk->list, &rlun->bb_list);
-       }
-       spin_unlock(&rlun->lock);
-}
-
-static void rrpc_put_blks(struct rrpc *rrpc)
-{
-       struct rrpc_lun *rlun;
-       int i;
-
-       for (i = 0; i < rrpc->nr_luns; i++) {
-               rlun = &rrpc->luns[i];
-               if (rlun->cur)
-                       rrpc_put_blk(rrpc, rlun->cur);
-               if (rlun->gc_cur)
-                       rrpc_put_blk(rrpc, rlun->gc_cur);
-       }
-}
-
-static struct rrpc_lun *get_next_lun(struct rrpc *rrpc)
-{
-       int next = atomic_inc_return(&rrpc->next_lun);
-
-       return &rrpc->luns[next % rrpc->nr_luns];
-}
-
-static void rrpc_gc_kick(struct rrpc *rrpc)
-{
-       struct rrpc_lun *rlun;
-       unsigned int i;
-
-       for (i = 0; i < rrpc->nr_luns; i++) {
-               rlun = &rrpc->luns[i];
-               queue_work(rrpc->krqd_wq, &rlun->ws_gc);
-       }
-}
-
-/*
- * timed GC every interval.
- */
-static void rrpc_gc_timer(struct timer_list *t)
-{
-       struct rrpc *rrpc = from_timer(rrpc, t, gc_timer);
-
-       rrpc_gc_kick(rrpc);
-       mod_timer(&rrpc->gc_timer, jiffies + msecs_to_jiffies(10));
-}
-
-static void rrpc_end_sync_bio(struct bio *bio)
-{
-       struct completion *waiting = bio->bi_private;
-
-       if (bio->bi_status)
-               pr_err("nvm: gc request failed (%u).\n", bio->bi_status);
-
-       complete(waiting);
-}
-
-/*
- * rrpc_move_valid_pages -- migrate live data off the block
- * @rrpc: the 'rrpc' structure
- * @block: the block from which to migrate live pages
- *
- * Description:
- *   GC algorithms may call this function to migrate remaining live
- *   pages off the block prior to erasing it. This function blocks
- *   further execution until the operation is complete.
- */
-static int rrpc_move_valid_pages(struct rrpc *rrpc, struct rrpc_block *rblk)
-{
-       struct nvm_tgt_dev *dev = rrpc->dev;
-       struct request_queue *q = dev->q;
-       struct rrpc_rev_addr *rev;
-       struct nvm_rq *rqd;
-       struct bio *bio;
-       struct page *page;
-       int slot;
-       int nr_sec_per_blk = dev->geo.sec_per_blk;
-       u64 phys_addr;
-       DECLARE_COMPLETION_ONSTACK(wait);
-
-       if (bitmap_full(rblk->invalid_pages, nr_sec_per_blk))
-               return 0;
-
-       bio = bio_alloc(GFP_NOIO, 1);
-       if (!bio) {
-               pr_err("nvm: could not alloc bio to gc\n");
-               return -ENOMEM;
-       }
-
-       page = mempool_alloc(rrpc->page_pool, GFP_NOIO);
-
-       while ((slot = find_first_zero_bit(rblk->invalid_pages,
-                                           nr_sec_per_blk)) < nr_sec_per_blk) {
-
-               /* Lock laddr */
-               phys_addr = rrpc_blk_to_ppa(rrpc, rblk) + slot;
-
-try:
-               spin_lock(&rrpc->rev_lock);
-               /* Get logical address from physical to logical table */
-               rev = &rrpc->rev_trans_map[phys_addr];
-               /* already updated by previous regular write */
-               if (rev->addr == ADDR_EMPTY) {
-                       spin_unlock(&rrpc->rev_lock);
-                       continue;
-               }
-
-               rqd = rrpc_inflight_laddr_acquire(rrpc, rev->addr, 1);
-               if (IS_ERR_OR_NULL(rqd)) {
-                       spin_unlock(&rrpc->rev_lock);
-                       schedule();
-                       goto try;
-               }
-
-               spin_unlock(&rrpc->rev_lock);
-
-               /* Perform read to do GC */
-               bio->bi_iter.bi_sector = rrpc_get_sector(rev->addr);
-               bio_set_op_attrs(bio,  REQ_OP_READ, 0);
-               bio->bi_private = &wait;
-               bio->bi_end_io = rrpc_end_sync_bio;
-
-               /* TODO: may fail when EXP_PG_SIZE > PAGE_SIZE */
-               bio_add_pc_page(q, bio, page, RRPC_EXPOSED_PAGE_SIZE, 0);
-
-               if (rrpc_submit_io(rrpc, bio, rqd, NVM_IOTYPE_GC)) {
-                       pr_err("rrpc: gc read failed.\n");
-                       rrpc_inflight_laddr_release(rrpc, rqd);
-                       goto finished;
-               }
-               wait_for_completion_io(&wait);
-               if (bio->bi_status) {
-                       rrpc_inflight_laddr_release(rrpc, rqd);
-                       goto finished;
-               }
-
-               bio_reset(bio);
-               reinit_completion(&wait);
-
-               bio->bi_iter.bi_sector = rrpc_get_sector(rev->addr);
-               bio_set_op_attrs(bio, REQ_OP_WRITE, 0);
-               bio->bi_private = &wait;
-               bio->bi_end_io = rrpc_end_sync_bio;
-
-               bio_add_pc_page(q, bio, page, RRPC_EXPOSED_PAGE_SIZE, 0);
-
-               /* turn the command around and write the data back to a new
-                * address
-                */
-               if (rrpc_submit_io(rrpc, bio, rqd, NVM_IOTYPE_GC)) {
-                       pr_err("rrpc: gc write failed.\n");
-                       rrpc_inflight_laddr_release(rrpc, rqd);
-                       goto finished;
-               }
-               wait_for_completion_io(&wait);
-
-               rrpc_inflight_laddr_release(rrpc, rqd);
-               if (bio->bi_status)
-                       goto finished;
-
-               bio_reset(bio);
-       }
-
-finished:
-       mempool_free(page, rrpc->page_pool);
-       bio_put(bio);
-
-       if (!bitmap_full(rblk->invalid_pages, nr_sec_per_blk)) {
-               pr_err("nvm: failed to garbage collect block\n");
-               return -EIO;
-       }
-
-       return 0;
-}
-
-static void rrpc_block_gc(struct work_struct *work)
-{
-       struct rrpc_block_gc *gcb = container_of(work, struct rrpc_block_gc,
-                                                                       ws_gc);
-       struct rrpc *rrpc = gcb->rrpc;
-       struct rrpc_block *rblk = gcb->rblk;
-       struct rrpc_lun *rlun = rblk->rlun;
-       struct ppa_addr ppa;
-
-       mempool_free(gcb, rrpc->gcb_pool);
-       pr_debug("nvm: block 'ch:%d,lun:%d,blk:%d' being reclaimed\n",
-                       rlun->bppa.g.ch, rlun->bppa.g.lun,
-                       rblk->id);
-
-       if (rrpc_move_valid_pages(rrpc, rblk))
-               goto put_back;
-
-       ppa.ppa = 0;
-       ppa.g.ch = rlun->bppa.g.ch;
-       ppa.g.lun = rlun->bppa.g.lun;
-       ppa.g.blk = rblk->id;
-
-       if (nvm_erase_sync(rrpc->dev, &ppa, 1))
-               goto put_back;
-
-       rrpc_put_blk(rrpc, rblk);
-
-       return;
-
-put_back:
-       spin_lock(&rlun->lock);
-       list_add_tail(&rblk->prio, &rlun->prio_list);
-       spin_unlock(&rlun->lock);
-}
-
-/* the block with highest number of invalid pages, will be in the beginning
- * of the list
- */
-static struct rrpc_block *rblk_max_invalid(struct rrpc_block *ra,
-                                                       struct rrpc_block *rb)
-{
-       if (ra->nr_invalid_pages == rb->nr_invalid_pages)
-               return ra;
-
-       return (ra->nr_invalid_pages < rb->nr_invalid_pages) ? rb : ra;
-}
-
-/* linearly find the block with highest number of invalid pages
- * requires lun->lock
- */
-static struct rrpc_block *block_prio_find_max(struct rrpc_lun *rlun)
-{
-       struct list_head *prio_list = &rlun->prio_list;
-       struct rrpc_block *rblk, *max;
-
-       BUG_ON(list_empty(prio_list));
-
-       max = list_first_entry(prio_list, struct rrpc_block, prio);
-       list_for_each_entry(rblk, prio_list, prio)
-               max = rblk_max_invalid(max, rblk);
-
-       return max;
-}
-
-static void rrpc_lun_gc(struct work_struct *work)
-{
-       struct rrpc_lun *rlun = container_of(work, struct rrpc_lun, ws_gc);
-       struct rrpc *rrpc = rlun->rrpc;
-       struct nvm_tgt_dev *dev = rrpc->dev;
-       struct rrpc_block_gc *gcb;
-       unsigned int nr_blocks_need;
-
-       nr_blocks_need = dev->geo.blks_per_lun / GC_LIMIT_INVERSE;
-
-       if (nr_blocks_need < rrpc->nr_luns)
-               nr_blocks_need = rrpc->nr_luns;
-
-       spin_lock(&rlun->lock);
-       while (nr_blocks_need > rlun->nr_free_blocks &&
-                                       !list_empty(&rlun->prio_list)) {
-               struct rrpc_block *rblk = block_prio_find_max(rlun);
-
-               if (!rblk->nr_invalid_pages)
-                       break;
-
-               gcb = mempool_alloc(rrpc->gcb_pool, GFP_ATOMIC);
-               if (!gcb)
-                       break;
-
-               list_del_init(&rblk->prio);
-
-               WARN_ON(!block_is_full(rrpc, rblk));
-
-               pr_debug("rrpc: selected block 'ch:%d,lun:%d,blk:%d' for GC\n",
-                                       rlun->bppa.g.ch, rlun->bppa.g.lun,
-                                       rblk->id);
-
-               gcb->rrpc = rrpc;
-               gcb->rblk = rblk;
-               INIT_WORK(&gcb->ws_gc, rrpc_block_gc);
-
-               queue_work(rrpc->kgc_wq, &gcb->ws_gc);
-
-               nr_blocks_need--;
-       }
-       spin_unlock(&rlun->lock);
-
-       /* TODO: Hint that request queue can be started again */
-}
-
-static void rrpc_gc_queue(struct work_struct *work)
-{
-       struct rrpc_block_gc *gcb = container_of(work, struct rrpc_block_gc,
-                                                                       ws_gc);
-       struct rrpc *rrpc = gcb->rrpc;
-       struct rrpc_block *rblk = gcb->rblk;
-       struct rrpc_lun *rlun = rblk->rlun;
-
-       spin_lock(&rlun->lock);
-       list_add_tail(&rblk->prio, &rlun->prio_list);
-       spin_unlock(&rlun->lock);
-
-       mempool_free(gcb, rrpc->gcb_pool);
-       pr_debug("nvm: block 'ch:%d,lun:%d,blk:%d' full, allow GC (sched)\n",
-                                       rlun->bppa.g.ch, rlun->bppa.g.lun,
-                                       rblk->id);
-}
-
-static const struct block_device_operations rrpc_fops = {
-       .owner          = THIS_MODULE,
-};
-
-static struct rrpc_lun *rrpc_get_lun_rr(struct rrpc *rrpc, int is_gc)
-{
-       unsigned int i;
-       struct rrpc_lun *rlun, *max_free;
-
-       if (!is_gc)
-               return get_next_lun(rrpc);
-
-       /* during GC, we don't care about RR, instead we want to make
-        * sure that we maintain evenness between the block luns.
-        */
-       max_free = &rrpc->luns[0];
-       /* prevent GC-ing lun from devouring pages of a lun with
-        * little free blocks. We don't take the lock as we only need an
-        * estimate.
-        */
-       rrpc_for_each_lun(rrpc, rlun, i) {
-               if (rlun->nr_free_blocks > max_free->nr_free_blocks)
-                       max_free = rlun;
-       }
-
-       return max_free;
-}
-
-static struct rrpc_addr *rrpc_update_map(struct rrpc *rrpc, sector_t laddr,
-                                       struct rrpc_block *rblk, u64 paddr)
-{
-       struct rrpc_addr *gp;
-       struct rrpc_rev_addr *rev;
-
-       BUG_ON(laddr >= rrpc->nr_sects);
-
-       gp = &rrpc->trans_map[laddr];
-       spin_lock(&rrpc->rev_lock);
-       if (gp->rblk)
-               rrpc_page_invalidate(rrpc, gp);
-
-       gp->addr = paddr;
-       gp->rblk = rblk;
-
-       rev = &rrpc->rev_trans_map[gp->addr];
-       rev->addr = laddr;
-       spin_unlock(&rrpc->rev_lock);
-
-       return gp;
-}
-
-static u64 rrpc_alloc_addr(struct rrpc *rrpc, struct rrpc_block *rblk)
-{
-       u64 addr = ADDR_EMPTY;
-
-       spin_lock(&rblk->lock);
-       if (block_is_full(rrpc, rblk))
-               goto out;
-
-       addr = rblk->next_page;
-
-       rblk->next_page++;
-out:
-       spin_unlock(&rblk->lock);
-       return addr;
-}
-
-/* Map logical address to a physical page. The mapping implements a round robin
- * approach and allocates a page from the next lun available.
- *
- * Returns rrpc_addr with the physical address and block. Returns NULL if no
- * blocks in the next rlun are available.
- */
-static struct ppa_addr rrpc_map_page(struct rrpc *rrpc, sector_t laddr,
-                                                               int is_gc)
-{
-       struct nvm_tgt_dev *tgt_dev = rrpc->dev;
-       struct rrpc_lun *rlun;
-       struct rrpc_block *rblk, **cur_rblk;
-       struct rrpc_addr *p;
-       struct ppa_addr ppa;
-       u64 paddr;
-       int gc_force = 0;
-
-       ppa.ppa = ADDR_EMPTY;
-       rlun = rrpc_get_lun_rr(rrpc, is_gc);
-
-       if (!is_gc && rlun->nr_free_blocks < rrpc->nr_luns * 4)
-               return ppa;
-
-       /*
-        * page allocation steps:
-        * 1. Try to allocate new page from current rblk
-        * 2a. If succeed, proceed to map it in and return
-        * 2b. If fail, first try to allocate a new block from media manger,
-        *     and then retry step 1. Retry until the normal block pool is
-        *     exhausted.
-        * 3. If exhausted, and garbage collector is requesting the block,
-        *    go to the reserved block and retry step 1.
-        *    In the case that this fails as well, or it is not GC
-        *    requesting, report not able to retrieve a block and let the
-        *    caller handle further processing.
-        */
-
-       spin_lock(&rlun->lock);
-       cur_rblk = &rlun->cur;
-       rblk = rlun->cur;
-retry:
-       paddr = rrpc_alloc_addr(rrpc, rblk);
-
-       if (paddr != ADDR_EMPTY)
-               goto done;
-
-       if (!list_empty(&rlun->wblk_list)) {
-new_blk:
-               rblk = list_first_entry(&rlun->wblk_list, struct rrpc_block,
-                                                                       prio);
-               rrpc_set_lun_cur(rlun, rblk, cur_rblk);
-               list_del(&rblk->prio);
-               goto retry;
-       }
-       spin_unlock(&rlun->lock);
-
-       rblk = rrpc_get_blk(rrpc, rlun, gc_force);
-       if (rblk) {
-               spin_lock(&rlun->lock);
-               list_add_tail(&rblk->prio, &rlun->wblk_list);
-               /*
-                * another thread might already have added a new block,
-                * Therefore, make sure that one is used, instead of the
-                * one just added.
-                */
-               goto new_blk;
-       }
-
-       if (unlikely(is_gc) && !gc_force) {
-               /* retry from emergency gc block */
-               cur_rblk = &rlun->gc_cur;
-               rblk = rlun->gc_cur;
-               gc_force = 1;
-               spin_lock(&rlun->lock);
-               goto retry;
-       }
-
-       pr_err("rrpc: failed to allocate new block\n");
-       return ppa;
-done:
-       spin_unlock(&rlun->lock);
-       p = rrpc_update_map(rrpc, laddr, rblk, paddr);
-       if (!p)
-               return ppa;
-
-       /* return global address */
-       return rrpc_ppa_to_gaddr(tgt_dev, p);
-}
-
-static void rrpc_run_gc(struct rrpc *rrpc, struct rrpc_block *rblk)
-{
-       struct rrpc_block_gc *gcb;
-
-       gcb = mempool_alloc(rrpc->gcb_pool, GFP_ATOMIC);
-       if (!gcb) {
-               pr_err("rrpc: unable to queue block for gc.");
-               return;
-       }
-
-       gcb->rrpc = rrpc;
-       gcb->rblk = rblk;
-
-       INIT_WORK(&gcb->ws_gc, rrpc_gc_queue);
-       queue_work(rrpc->kgc_wq, &gcb->ws_gc);
-}
-
-static struct rrpc_lun *rrpc_ppa_to_lun(struct rrpc *rrpc, struct ppa_addr p)
-{
-       struct rrpc_lun *rlun = NULL;
-       int i;
-
-       for (i = 0; i < rrpc->nr_luns; i++) {
-               if (rrpc->luns[i].bppa.g.ch == p.g.ch &&
-                               rrpc->luns[i].bppa.g.lun == p.g.lun) {
-                       rlun = &rrpc->luns[i];
-                       break;
-               }
-       }
-
-       return rlun;
-}
-
-static void __rrpc_mark_bad_block(struct rrpc *rrpc, struct ppa_addr ppa)
-{
-       struct nvm_tgt_dev *dev = rrpc->dev;
-       struct rrpc_lun *rlun;
-       struct rrpc_block *rblk;
-
-       rlun = rrpc_ppa_to_lun(rrpc, ppa);
-       rblk = &rlun->blocks[ppa.g.blk];
-       rblk->state = NVM_BLK_ST_BAD;
-
-       nvm_set_tgt_bb_tbl(dev, &ppa, 1, NVM_BLK_T_GRWN_BAD);
-}
-
-static void rrpc_mark_bad_block(struct rrpc *rrpc, struct nvm_rq *rqd)
-{
-       void *comp_bits = &rqd->ppa_status;
-       struct ppa_addr ppa, prev_ppa;
-       int nr_ppas = rqd->nr_ppas;
-       int bit;
-
-       if (rqd->nr_ppas == 1)
-               __rrpc_mark_bad_block(rrpc, rqd->ppa_addr);
-
-       ppa_set_empty(&prev_ppa);
-       bit = -1;
-       while ((bit = find_next_bit(comp_bits, nr_ppas, bit + 1)) < nr_ppas) {
-               ppa = rqd->ppa_list[bit];
-               if (ppa_cmp_blk(ppa, prev_ppa))
-                       continue;
-
-               __rrpc_mark_bad_block(rrpc, ppa);
-       }
-}
-
-static void rrpc_end_io_write(struct rrpc *rrpc, struct rrpc_rq *rrqd,
-                                               sector_t laddr, uint8_t npages)
-{
-       struct nvm_tgt_dev *dev = rrpc->dev;
-       struct rrpc_addr *p;
-       struct rrpc_block *rblk;
-       int cmnt_size, i;
-
-       for (i = 0; i < npages; i++) {
-               p = &rrpc->trans_map[laddr + i];
-               rblk = p->rblk;
-
-               cmnt_size = atomic_inc_return(&rblk->data_cmnt_size);
-               if (unlikely(cmnt_size == dev->geo.sec_per_blk))
-                       rrpc_run_gc(rrpc, rblk);
-       }
-}
-
-static void rrpc_end_io(struct nvm_rq *rqd)
-{
-       struct rrpc *rrpc = rqd->private;
-       struct nvm_tgt_dev *dev = rrpc->dev;
-       struct rrpc_rq *rrqd = nvm_rq_to_pdu(rqd);
-       uint8_t npages = rqd->nr_ppas;
-       sector_t laddr = rrpc_get_laddr(rqd->bio) - npages;
-
-       if (bio_data_dir(rqd->bio) == WRITE) {
-               if (rqd->error == NVM_RSP_ERR_FAILWRITE)
-                       rrpc_mark_bad_block(rrpc, rqd);
-
-               rrpc_end_io_write(rrpc, rrqd, laddr, npages);
-       }
-
-       bio_put(rqd->bio);
-
-       if (rrqd->flags & NVM_IOTYPE_GC)
-               return;
-
-       rrpc_unlock_rq(rrpc, rqd);
-
-       if (npages > 1)
-               nvm_dev_dma_free(dev->parent, rqd->ppa_list, rqd->dma_ppa_list);
-
-       mempool_free(rqd, rrpc->rq_pool);
-}
-
-static int rrpc_read_ppalist_rq(struct rrpc *rrpc, struct bio *bio,
-                       struct nvm_rq *rqd, unsigned long flags, int npages)
-{
-       struct nvm_tgt_dev *dev = rrpc->dev;
-       struct rrpc_inflight_rq *r = rrpc_get_inflight_rq(rqd);
-       struct rrpc_addr *gp;
-       sector_t laddr = rrpc_get_laddr(bio);
-       int is_gc = flags & NVM_IOTYPE_GC;
-       int i;
-
-       if (!is_gc && rrpc_lock_rq(rrpc, bio, rqd)) {
-               nvm_dev_dma_free(dev->parent, rqd->ppa_list, rqd->dma_ppa_list);
-               return NVM_IO_REQUEUE;
-       }
-
-       for (i = 0; i < npages; i++) {
-               /* We assume that mapping occurs at 4KB granularity */
-               BUG_ON(!(laddr + i < rrpc->nr_sects));
-               gp = &rrpc->trans_map[laddr + i];
-
-               if (gp->rblk) {
-                       rqd->ppa_list[i] = rrpc_ppa_to_gaddr(dev, gp);
-               } else {
-                       BUG_ON(is_gc);
-                       rrpc_unlock_laddr(rrpc, r);
-                       nvm_dev_dma_free(dev->parent, rqd->ppa_list,
-                                                       rqd->dma_ppa_list);
-                       return NVM_IO_DONE;
-               }
-       }
-
-       rqd->opcode = NVM_OP_HBREAD;
-
-       return NVM_IO_OK;
-}
-
-static int rrpc_read_rq(struct rrpc *rrpc, struct bio *bio, struct nvm_rq *rqd,
-                                                       unsigned long flags)
-{
-       int is_gc = flags & NVM_IOTYPE_GC;
-       sector_t laddr = rrpc_get_laddr(bio);
-       struct rrpc_addr *gp;
-
-       if (!is_gc && rrpc_lock_rq(rrpc, bio, rqd))
-               return NVM_IO_REQUEUE;
-
-       BUG_ON(!(laddr < rrpc->nr_sects));
-       gp = &rrpc->trans_map[laddr];
-
-       if (gp->rblk) {
-               rqd->ppa_addr = rrpc_ppa_to_gaddr(rrpc->dev, gp);
-       } else {
-               BUG_ON(is_gc);
-               rrpc_unlock_rq(rrpc, rqd);
-               return NVM_IO_DONE;
-       }
-
-       rqd->opcode = NVM_OP_HBREAD;
-
-       return NVM_IO_OK;
-}
-
-static int rrpc_write_ppalist_rq(struct rrpc *rrpc, struct bio *bio,
-                       struct nvm_rq *rqd, unsigned long flags, int npages)
-{
-       struct nvm_tgt_dev *dev = rrpc->dev;
-       struct rrpc_inflight_rq *r = rrpc_get_inflight_rq(rqd);
-       struct ppa_addr p;
-       sector_t laddr = rrpc_get_laddr(bio);
-       int is_gc = flags & NVM_IOTYPE_GC;
-       int i;
-
-       if (!is_gc && rrpc_lock_rq(rrpc, bio, rqd)) {
-               nvm_dev_dma_free(dev->parent, rqd->ppa_list, rqd->dma_ppa_list);
-               return NVM_IO_REQUEUE;
-       }
-
-       for (i = 0; i < npages; i++) {
-               /* We assume that mapping occurs at 4KB granularity */
-               p = rrpc_map_page(rrpc, laddr + i, is_gc);
-               if (p.ppa == ADDR_EMPTY) {
-                       BUG_ON(is_gc);
-                       rrpc_unlock_laddr(rrpc, r);
-                       nvm_dev_dma_free(dev->parent, rqd->ppa_list,
-                                                       rqd->dma_ppa_list);
-                       rrpc_gc_kick(rrpc);
-                       return NVM_IO_REQUEUE;
-               }
-
-               rqd->ppa_list[i] = p;
-       }
-
-       rqd->opcode = NVM_OP_HBWRITE;
-
-       return NVM_IO_OK;
-}
-
-static int rrpc_write_rq(struct rrpc *rrpc, struct bio *bio,
-                               struct nvm_rq *rqd, unsigned long flags)
-{
-       struct ppa_addr p;
-       int is_gc = flags & NVM_IOTYPE_GC;
-       sector_t laddr = rrpc_get_laddr(bio);
-
-       if (!is_gc && rrpc_lock_rq(rrpc, bio, rqd))
-               return NVM_IO_REQUEUE;
-
-       p = rrpc_map_page(rrpc, laddr, is_gc);
-       if (p.ppa == ADDR_EMPTY) {
-               BUG_ON(is_gc);
-               rrpc_unlock_rq(rrpc, rqd);
-               rrpc_gc_kick(rrpc);
-               return NVM_IO_REQUEUE;
-       }
-
-       rqd->ppa_addr = p;
-       rqd->opcode = NVM_OP_HBWRITE;
-
-       return NVM_IO_OK;
-}
-
-static int rrpc_setup_rq(struct rrpc *rrpc, struct bio *bio,
-                       struct nvm_rq *rqd, unsigned long flags, uint8_t npages)
-{
-       struct nvm_tgt_dev *dev = rrpc->dev;
-
-       if (npages > 1) {
-               rqd->ppa_list = nvm_dev_dma_alloc(dev->parent, GFP_KERNEL,
-                                                       &rqd->dma_ppa_list);
-               if (!rqd->ppa_list) {
-                       pr_err("rrpc: not able to allocate ppa list\n");
-                       return NVM_IO_ERR;
-               }
-
-               if (bio_op(bio) == REQ_OP_WRITE)
-                       return rrpc_write_ppalist_rq(rrpc, bio, rqd, flags,
-                                                                       npages);
-
-               return rrpc_read_ppalist_rq(rrpc, bio, rqd, flags, npages);
-       }
-
-       if (bio_op(bio) == REQ_OP_WRITE)
-               return rrpc_write_rq(rrpc, bio, rqd, flags);
-
-       return rrpc_read_rq(rrpc, bio, rqd, flags);
-}
-
-static int rrpc_submit_io(struct rrpc *rrpc, struct bio *bio,
-                               struct nvm_rq *rqd, unsigned long flags)
-{
-       struct nvm_tgt_dev *dev = rrpc->dev;
-       struct rrpc_rq *rrq = nvm_rq_to_pdu(rqd);
-       uint8_t nr_pages = rrpc_get_pages(bio);
-       int bio_size = bio_sectors(bio) << 9;
-       int err;
-
-       if (bio_size < dev->geo.sec_size)
-               return NVM_IO_ERR;
-       else if (bio_size > dev->geo.max_rq_size)
-               return NVM_IO_ERR;
-
-       err = rrpc_setup_rq(rrpc, bio, rqd, flags, nr_pages);
-       if (err)
-               return err;
-
-       bio_get(bio);
-       rqd->bio = bio;
-       rqd->private = rrpc;
-       rqd->nr_ppas = nr_pages;
-       rqd->end_io = rrpc_end_io;
-       rrq->flags = flags;
-
-       err = nvm_submit_io(dev, rqd);
-       if (err) {
-               pr_err("rrpc: I/O submission failed: %d\n", err);
-               bio_put(bio);
-               if (!(flags & NVM_IOTYPE_GC)) {
-                       rrpc_unlock_rq(rrpc, rqd);
-                       if (rqd->nr_ppas > 1)
-                               nvm_dev_dma_free(dev->parent, rqd->ppa_list,
-                                                       rqd->dma_ppa_list);
-               }
-               return NVM_IO_ERR;
-       }
-
-       return NVM_IO_OK;
-}
-
-static blk_qc_t rrpc_make_rq(struct request_queue *q, struct bio *bio)
-{
-       struct rrpc *rrpc = q->queuedata;
-       struct nvm_rq *rqd;
-       int err;
-
-       blk_queue_split(q, &bio);
-
-       if (bio_op(bio) == REQ_OP_DISCARD) {
-               rrpc_discard(rrpc, bio);
-               return BLK_QC_T_NONE;
-       }
-
-       rqd = mempool_alloc(rrpc->rq_pool, GFP_KERNEL);
-       memset(rqd, 0, sizeof(struct nvm_rq));
-
-       err = rrpc_submit_io(rrpc, bio, rqd, NVM_IOTYPE_NONE);
-       switch (err) {
-       case NVM_IO_OK:
-               return BLK_QC_T_NONE;
-       case NVM_IO_ERR:
-               bio_io_error(bio);
-               break;
-       case NVM_IO_DONE:
-               bio_endio(bio);
-               break;
-       case NVM_IO_REQUEUE:
-               spin_lock(&rrpc->bio_lock);
-               bio_list_add(&rrpc->requeue_bios, bio);
-               spin_unlock(&rrpc->bio_lock);
-               queue_work(rrpc->kgc_wq, &rrpc->ws_requeue);
-               break;
-       }
-
-       mempool_free(rqd, rrpc->rq_pool);
-       return BLK_QC_T_NONE;
-}
-
-static void rrpc_requeue(struct work_struct *work)
-{
-       struct rrpc *rrpc = container_of(work, struct rrpc, ws_requeue);
-       struct bio_list bios;
-       struct bio *bio;
-
-       bio_list_init(&bios);
-
-       spin_lock(&rrpc->bio_lock);
-       bio_list_merge(&bios, &rrpc->requeue_bios);
-       bio_list_init(&rrpc->requeue_bios);
-       spin_unlock(&rrpc->bio_lock);
-
-       while ((bio = bio_list_pop(&bios)))
-               rrpc_make_rq(rrpc->disk->queue, bio);
-}
-
-static void rrpc_gc_free(struct rrpc *rrpc)
-{
-       if (rrpc->krqd_wq)
-               destroy_workqueue(rrpc->krqd_wq);
-
-       if (rrpc->kgc_wq)
-               destroy_workqueue(rrpc->kgc_wq);
-}
-
-static int rrpc_gc_init(struct rrpc *rrpc)
-{
-       rrpc->krqd_wq = alloc_workqueue("rrpc-lun", WQ_MEM_RECLAIM|WQ_UNBOUND,
-                                                               rrpc->nr_luns);
-       if (!rrpc->krqd_wq)
-               return -ENOMEM;
-
-       rrpc->kgc_wq = alloc_workqueue("rrpc-bg", WQ_MEM_RECLAIM, 1);
-       if (!rrpc->kgc_wq)
-               return -ENOMEM;
-
-       timer_setup(&rrpc->gc_timer, rrpc_gc_timer, 0);
-
-       return 0;
-}
-
-static void rrpc_map_free(struct rrpc *rrpc)
-{
-       vfree(rrpc->rev_trans_map);
-       vfree(rrpc->trans_map);
-}
-
-static int rrpc_l2p_update(u64 slba, u32 nlb, __le64 *entries, void *private)
-{
-       struct rrpc *rrpc = (struct rrpc *)private;
-       struct nvm_tgt_dev *dev = rrpc->dev;
-       struct rrpc_addr *addr = rrpc->trans_map + slba;
-       struct rrpc_rev_addr *raddr = rrpc->rev_trans_map;
-       struct rrpc_lun *rlun;
-       struct rrpc_block *rblk;
-       u64 i;
-
-       for (i = 0; i < nlb; i++) {
-               struct ppa_addr gaddr;
-               u64 pba = le64_to_cpu(entries[i]);
-               unsigned int mod;
-
-               /* LNVM treats address-spaces as silos, LBA and PBA are
-                * equally large and zero-indexed.
-                */
-               if (unlikely(pba >= dev->total_secs && pba != U64_MAX)) {
-                       pr_err("nvm: L2P data entry is out of bounds!\n");
-                       pr_err("nvm: Maybe loaded an old target L2P\n");
-                       return -EINVAL;
-               }
-
-               /* Address zero is a special one. The first page on a disk is
-                * protected. As it often holds internal device boot
-                * information.
-                */
-               if (!pba)
-                       continue;
-
-               div_u64_rem(pba, rrpc->nr_sects, &mod);
-
-               gaddr = rrpc_recov_addr(dev, pba);
-               rlun = rrpc_ppa_to_lun(rrpc, gaddr);
-               if (!rlun) {
-                       pr_err("rrpc: l2p corruption on lba %llu\n",
-                                                       slba + i);
-                       return -EINVAL;
-               }
-
-               rblk = &rlun->blocks[gaddr.g.blk];
-               if (!rblk->state) {
-                       /* at this point, we don't know anything about the
-                        * block. It's up to the FTL on top to re-etablish the
-                        * block state. The block is assumed to be open.
-                        */
-                       list_move_tail(&rblk->list, &rlun->used_list);
-                       rblk->state = NVM_BLK_ST_TGT;
-                       rlun->nr_free_blocks--;
-               }
-
-               addr[i].addr = pba;
-               addr[i].rblk = rblk;
-               raddr[mod].addr = slba + i;
-       }
-
-       return 0;
-}
-
-static int rrpc_map_init(struct rrpc *rrpc)
-{
-       struct nvm_tgt_dev *dev = rrpc->dev;
-       sector_t i;
-       int ret;
-
-       rrpc->trans_map = vzalloc(sizeof(struct rrpc_addr) * rrpc->nr_sects);
-       if (!rrpc->trans_map)
-               return -ENOMEM;
-
-       rrpc->rev_trans_map = vmalloc(sizeof(struct rrpc_rev_addr)
-                                                       * rrpc->nr_sects);
-       if (!rrpc->rev_trans_map)
-               return -ENOMEM;
-
-       for (i = 0; i < rrpc->nr_sects; i++) {
-               struct rrpc_addr *p = &rrpc->trans_map[i];
-               struct rrpc_rev_addr *r = &rrpc->rev_trans_map[i];
-
-               p->addr = ADDR_EMPTY;
-               r->addr = ADDR_EMPTY;
-       }
-
-       /* Bring up the mapping table from device */
-       ret = nvm_get_l2p_tbl(dev, rrpc->soffset, rrpc->nr_sects,
-                                                       rrpc_l2p_update, rrpc);
-       if (ret) {
-               pr_err("nvm: rrpc: could not read L2P table.\n");
-               return -EINVAL;
-       }
-
-       return 0;
-}
-
-/* Minimum pages needed within a lun */
-#define PAGE_POOL_SIZE 16
-#define ADDR_POOL_SIZE 64
-
-static int rrpc_core_init(struct rrpc *rrpc)
-{
-       down_write(&rrpc_lock);
-       if (!rrpc_gcb_cache) {
-               rrpc_gcb_cache = kmem_cache_create("rrpc_gcb",
-                               sizeof(struct rrpc_block_gc), 0, 0, NULL);
-               if (!rrpc_gcb_cache) {
-                       up_write(&rrpc_lock);
-                       return -ENOMEM;
-               }
-
-               rrpc_rq_cache = kmem_cache_create("rrpc_rq",
-                               sizeof(struct nvm_rq) + sizeof(struct rrpc_rq),
-                               0, 0, NULL);
-               if (!rrpc_rq_cache) {
-                       kmem_cache_destroy(rrpc_gcb_cache);
-                       up_write(&rrpc_lock);
-                       return -ENOMEM;
-               }
-       }
-       up_write(&rrpc_lock);
-
-       rrpc->page_pool = mempool_create_page_pool(PAGE_POOL_SIZE, 0);
-       if (!rrpc->page_pool)
-               return -ENOMEM;
-
-       rrpc->gcb_pool = mempool_create_slab_pool(rrpc->dev->geo.nr_luns,
-                                                               rrpc_gcb_cache);
-       if (!rrpc->gcb_pool)
-               return -ENOMEM;
-
-       rrpc->rq_pool = mempool_create_slab_pool(64, rrpc_rq_cache);
-       if (!rrpc->rq_pool)
-               return -ENOMEM;
-
-       spin_lock_init(&rrpc->inflights.lock);
-       INIT_LIST_HEAD(&rrpc->inflights.reqs);
-
-       return 0;
-}
-
-static void rrpc_core_free(struct rrpc *rrpc)
-{
-       mempool_destroy(rrpc->page_pool);
-       mempool_destroy(rrpc->gcb_pool);
-       mempool_destroy(rrpc->rq_pool);
-}
-
-static void rrpc_luns_free(struct rrpc *rrpc)
-{
-       struct rrpc_lun *rlun;
-       int i;
-
-       if (!rrpc->luns)
-               return;
-
-       for (i = 0; i < rrpc->nr_luns; i++) {
-               rlun = &rrpc->luns[i];
-               vfree(rlun->blocks);
-       }
-
-       kfree(rrpc->luns);
-}
-
-static int rrpc_bb_discovery(struct nvm_tgt_dev *dev, struct rrpc_lun *rlun)
-{
-       struct nvm_geo *geo = &dev->geo;
-       struct rrpc_block *rblk;
-       struct ppa_addr ppa;
-       u8 *blks;
-       int nr_blks;
-       int i;
-       int ret;
-
-       if (!dev->parent->ops->get_bb_tbl)
-               return 0;
-
-       nr_blks = geo->blks_per_lun * geo->plane_mode;
-       blks = kmalloc(nr_blks, GFP_KERNEL);
-       if (!blks)
-               return -ENOMEM;
-
-       ppa.ppa = 0;
-       ppa.g.ch = rlun->bppa.g.ch;
-       ppa.g.lun = rlun->bppa.g.lun;
-
-       ret = nvm_get_tgt_bb_tbl(dev, ppa, blks);
-       if (ret) {
-               pr_err("rrpc: could not get BB table\n");
-               goto out;
-       }
-
-       nr_blks = nvm_bb_tbl_fold(dev->parent, blks, nr_blks);
-       if (nr_blks < 0) {
-               ret = nr_blks;
-               goto out;
-       }
-
-       for (i = 0; i < nr_blks; i++) {
-               if (blks[i] == NVM_BLK_T_FREE)
-                       continue;
-
-               rblk = &rlun->blocks[i];
-               list_move_tail(&rblk->list, &rlun->bb_list);
-               rblk->state = NVM_BLK_ST_BAD;
-               rlun->nr_free_blocks--;
-       }
-
-out:
-       kfree(blks);
-       return ret;
-}
-
-static void rrpc_set_lun_ppa(struct rrpc_lun *rlun, struct ppa_addr ppa)
-{
-       rlun->bppa.ppa = 0;
-       rlun->bppa.g.ch = ppa.g.ch;
-       rlun->bppa.g.lun = ppa.g.lun;
-}
-
-static int rrpc_luns_init(struct rrpc *rrpc, struct ppa_addr *luns)
-{
-       struct nvm_tgt_dev *dev = rrpc->dev;
-       struct nvm_geo *geo = &dev->geo;
-       struct rrpc_lun *rlun;
-       int i, j, ret = -EINVAL;
-
-       if (geo->sec_per_blk > MAX_INVALID_PAGES_STORAGE * BITS_PER_LONG) {
-               pr_err("rrpc: number of pages per block too high.");
-               return -EINVAL;
-       }
-
-       spin_lock_init(&rrpc->rev_lock);
-
-       rrpc->luns = kcalloc(rrpc->nr_luns, sizeof(struct rrpc_lun),
-                                                               GFP_KERNEL);
-       if (!rrpc->luns)
-               return -ENOMEM;
-
-       /* 1:1 mapping */
-       for (i = 0; i < rrpc->nr_luns; i++) {
-               rlun = &rrpc->luns[i];
-               rlun->id = i;
-               rrpc_set_lun_ppa(rlun, luns[i]);
-               rlun->blocks = vzalloc(sizeof(struct rrpc_block) *
-                                                       geo->blks_per_lun);
-               if (!rlun->blocks) {
-                       ret = -ENOMEM;
-                       goto err;
-               }
-
-               INIT_LIST_HEAD(&rlun->free_list);
-               INIT_LIST_HEAD(&rlun->used_list);
-               INIT_LIST_HEAD(&rlun->bb_list);
-
-               for (j = 0; j < geo->blks_per_lun; j++) {
-                       struct rrpc_block *rblk = &rlun->blocks[j];
-
-                       rblk->id = j;
-                       rblk->rlun = rlun;
-                       rblk->state = NVM_BLK_T_FREE;
-                       INIT_LIST_HEAD(&rblk->prio);
-                       INIT_LIST_HEAD(&rblk->list);
-                       spin_lock_init(&rblk->lock);
-
-                       list_add_tail(&rblk->list, &rlun->free_list);
-               }
-
-               rlun->rrpc = rrpc;
-               rlun->nr_free_blocks = geo->blks_per_lun;
-               rlun->reserved_blocks = 2; /* for GC only */
-
-               INIT_LIST_HEAD(&rlun->prio_list);
-               INIT_LIST_HEAD(&rlun->wblk_list);
-
-               INIT_WORK(&rlun->ws_gc, rrpc_lun_gc);
-               spin_lock_init(&rlun->lock);
-
-               if (rrpc_bb_discovery(dev, rlun))
-                       goto err;
-
-       }
-
-       return 0;
-err:
-       return ret;
-}
-
-/* returns 0 on success and stores the beginning address in *begin */
-static int rrpc_area_init(struct rrpc *rrpc, sector_t *begin)
-{
-       struct nvm_tgt_dev *dev = rrpc->dev;
-       sector_t size = rrpc->nr_sects * dev->geo.sec_size;
-       int ret;
-
-       size >>= 9;
-
-       ret = nvm_get_area(dev, begin, size);
-       if (!ret)
-               *begin >>= (ilog2(dev->geo.sec_size) - 9);
-
-       return ret;
-}
-
-static void rrpc_area_free(struct rrpc *rrpc)
-{
-       struct nvm_tgt_dev *dev = rrpc->dev;
-       sector_t begin = rrpc->soffset << (ilog2(dev->geo.sec_size) - 9);
-
-       nvm_put_area(dev, begin);
-}
-
-static void rrpc_free(struct rrpc *rrpc)
-{
-       rrpc_gc_free(rrpc);
-       rrpc_map_free(rrpc);
-       rrpc_core_free(rrpc);
-       rrpc_luns_free(rrpc);
-       rrpc_area_free(rrpc);
-
-       kfree(rrpc);
-}
-
-static void rrpc_exit(void *private)
-{
-       struct rrpc *rrpc = private;
-
-       del_timer(&rrpc->gc_timer);
-
-       flush_workqueue(rrpc->krqd_wq);
-       flush_workqueue(rrpc->kgc_wq);
-
-       rrpc_free(rrpc);
-}
-
-static sector_t rrpc_capacity(void *private)
-{
-       struct rrpc *rrpc = private;
-       struct nvm_tgt_dev *dev = rrpc->dev;
-       sector_t reserved, provisioned;
-
-       /* cur, gc, and two emergency blocks for each lun */
-       reserved = rrpc->nr_luns * dev->geo.sec_per_blk * 4;
-       provisioned = rrpc->nr_sects - reserved;
-
-       if (reserved > rrpc->nr_sects) {
-               pr_err("rrpc: not enough space available to expose storage.\n");
-               return 0;
-       }
-
-       sector_div(provisioned, 10);
-       return provisioned * 9 * NR_PHY_IN_LOG;
-}
-
-/*
- * Looks up the logical address from reverse trans map and check if its valid by
- * comparing the logical to physical address with the physical address.
- * Returns 0 on free, otherwise 1 if in use
- */
-static void rrpc_block_map_update(struct rrpc *rrpc, struct rrpc_block *rblk)
-{
-       struct nvm_tgt_dev *dev = rrpc->dev;
-       int offset;
-       struct rrpc_addr *laddr;
-       u64 bpaddr, paddr, pladdr;
-
-       bpaddr = block_to_rel_addr(rrpc, rblk);
-       for (offset = 0; offset < dev->geo.sec_per_blk; offset++) {
-               paddr = bpaddr + offset;
-
-               pladdr = rrpc->rev_trans_map[paddr].addr;
-               if (pladdr == ADDR_EMPTY)
-                       continue;
-
-               laddr = &rrpc->trans_map[pladdr];
-
-               if (paddr == laddr->addr) {
-                       laddr->rblk = rblk;
-               } else {
-                       set_bit(offset, rblk->invalid_pages);
-                       rblk->nr_invalid_pages++;
-               }
-       }
-}
-
-static int rrpc_blocks_init(struct rrpc *rrpc)
-{
-       struct nvm_tgt_dev *dev = rrpc->dev;
-       struct rrpc_lun *rlun;
-       struct rrpc_block *rblk;
-       int lun_iter, blk_iter;
-
-       for (lun_iter = 0; lun_iter < rrpc->nr_luns; lun_iter++) {
-               rlun = &rrpc->luns[lun_iter];
-
-               for (blk_iter = 0; blk_iter < dev->geo.blks_per_lun;
-                                                               blk_iter++) {
-                       rblk = &rlun->blocks[blk_iter];
-                       rrpc_block_map_update(rrpc, rblk);
-               }
-       }
-
-       return 0;
-}
-
-static int rrpc_luns_configure(struct rrpc *rrpc)
-{
-       struct rrpc_lun *rlun;
-       struct rrpc_block *rblk;
-       int i;
-
-       for (i = 0; i < rrpc->nr_luns; i++) {
-               rlun = &rrpc->luns[i];
-
-               rblk = rrpc_get_blk(rrpc, rlun, 0);
-               if (!rblk)
-                       goto err;
-               rrpc_set_lun_cur(rlun, rblk, &rlun->cur);
-
-               /* Emergency gc block */
-               rblk = rrpc_get_blk(rrpc, rlun, 1);
-               if (!rblk)
-                       goto err;
-               rrpc_set_lun_cur(rlun, rblk, &rlun->gc_cur);
-       }
-
-       return 0;
-err:
-       rrpc_put_blks(rrpc);
-       return -EINVAL;
-}
-
-static struct nvm_tgt_type tt_rrpc;
-
-static void *rrpc_init(struct nvm_tgt_dev *dev, struct gendisk *tdisk,
-                      int flags)
-{
-       struct request_queue *bqueue = dev->q;
-       struct request_queue *tqueue = tdisk->queue;
-       struct nvm_geo *geo = &dev->geo;
-       struct rrpc *rrpc;
-       sector_t soffset;
-       int ret;
-
-       if (!(dev->identity.dom & NVM_RSP_L2P)) {
-               pr_err("nvm: rrpc: device does not support l2p (%x)\n",
-                                                       dev->identity.dom);
-               return ERR_PTR(-EINVAL);
-       }
-
-       rrpc = kzalloc(sizeof(struct rrpc), GFP_KERNEL);
-       if (!rrpc)
-               return ERR_PTR(-ENOMEM);
-
-       rrpc->dev = dev;
-       rrpc->disk = tdisk;
-
-       bio_list_init(&rrpc->requeue_bios);
-       spin_lock_init(&rrpc->bio_lock);
-       INIT_WORK(&rrpc->ws_requeue, rrpc_requeue);
-
-       rrpc->nr_luns = geo->nr_luns;
-       rrpc->nr_sects = (unsigned long long)geo->sec_per_lun * rrpc->nr_luns;
-
-       /* simple round-robin strategy */
-       atomic_set(&rrpc->next_lun, -1);
-
-       ret = rrpc_area_init(rrpc, &soffset);
-       if (ret < 0) {
-               pr_err("nvm: rrpc: could not initialize area\n");
-               return ERR_PTR(ret);
-       }
-       rrpc->soffset = soffset;
-
-       ret = rrpc_luns_init(rrpc, dev->luns);
-       if (ret) {
-               pr_err("nvm: rrpc: could not initialize luns\n");
-               goto err;
-       }
-
-       ret = rrpc_core_init(rrpc);
-       if (ret) {
-               pr_err("nvm: rrpc: could not initialize core\n");
-               goto err;
-       }
-
-       ret = rrpc_map_init(rrpc);
-       if (ret) {
-               pr_err("nvm: rrpc: could not initialize maps\n");
-               goto err;
-       }
-
-       ret = rrpc_blocks_init(rrpc);
-       if (ret) {
-               pr_err("nvm: rrpc: could not initialize state for blocks\n");
-               goto err;
-       }
-
-       ret = rrpc_luns_configure(rrpc);
-       if (ret) {
-               pr_err("nvm: rrpc: not enough blocks available in LUNs.\n");
-               goto err;
-       }
-
-       ret = rrpc_gc_init(rrpc);
-       if (ret) {
-               pr_err("nvm: rrpc: could not initialize gc\n");
-               goto err;
-       }
-
-       /* inherit the size from the underlying device */
-       blk_queue_logical_block_size(tqueue, queue_physical_block_size(bqueue));
-       blk_queue_max_hw_sectors(tqueue, queue_max_hw_sectors(bqueue));
-
-       pr_info("nvm: rrpc initialized with %u luns and %llu pages.\n",
-                       rrpc->nr_luns, (unsigned long long)rrpc->nr_sects);
-
-       mod_timer(&rrpc->gc_timer, jiffies + msecs_to_jiffies(10));
-
-       return rrpc;
-err:
-       rrpc_free(rrpc);
-       return ERR_PTR(ret);
-}
-
-/* round robin, page-based FTL, and cost-based GC */
-static struct nvm_tgt_type tt_rrpc = {
-       .name           = "rrpc",
-       .version        = {1, 0, 0},
-
-       .make_rq        = rrpc_make_rq,
-       .capacity       = rrpc_capacity,
-
-       .init           = rrpc_init,
-       .exit           = rrpc_exit,
-};
-
-static int __init rrpc_module_init(void)
-{
-       return nvm_register_tgt_type(&tt_rrpc);
-}
-
-static void rrpc_module_exit(void)
-{
-       nvm_unregister_tgt_type(&tt_rrpc);
-}
-
-module_init(rrpc_module_init);
-module_exit(rrpc_module_exit);
-MODULE_LICENSE("GPL v2");
-MODULE_DESCRIPTION("Block-Device Target for Open-Channel SSDs");
diff --git a/drivers/lightnvm/rrpc.h b/drivers/lightnvm/rrpc.h
deleted file mode 100644 (file)
index fdb6ff9..0000000
+++ /dev/null
@@ -1,290 +0,0 @@
-/*
- * Copyright (C) 2015 IT University of Copenhagen
- * Initial release: Matias Bjorling <m@bjorling.me>
- *
- * This program is free software; you can redistribute it and/or
- * modify it under the terms of the GNU General Public License version
- * 2 as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License for more details.
- *
- * Implementation of a Round-robin page-based Hybrid FTL for Open-channel SSDs.
- */
-
-#ifndef RRPC_H_
-#define RRPC_H_
-
-#include <linux/blkdev.h>
-#include <linux/blk-mq.h>
-#include <linux/bio.h>
-#include <linux/module.h>
-#include <linux/kthread.h>
-#include <linux/vmalloc.h>
-
-#include <linux/lightnvm.h>
-
-/* Run only GC if less than 1/X blocks are free */
-#define GC_LIMIT_INVERSE 10
-#define GC_TIME_SECS 100
-
-#define RRPC_SECTOR (512)
-#define RRPC_EXPOSED_PAGE_SIZE (4096)
-
-#define NR_PHY_IN_LOG (RRPC_EXPOSED_PAGE_SIZE / RRPC_SECTOR)
-
-struct rrpc_inflight {
-       struct list_head reqs;
-       spinlock_t lock;
-};
-
-struct rrpc_inflight_rq {
-       struct list_head list;
-       sector_t l_start;
-       sector_t l_end;
-};
-
-struct rrpc_rq {
-       struct rrpc_inflight_rq inflight_rq;
-       unsigned long flags;
-};
-
-struct rrpc_block {
-       int id;                         /* id inside of LUN */
-       struct rrpc_lun *rlun;
-
-       struct list_head prio;          /* LUN CG list */
-       struct list_head list;          /* LUN free, used, bb list */
-
-#define MAX_INVALID_PAGES_STORAGE 8
-       /* Bitmap for invalid page intries */
-       unsigned long invalid_pages[MAX_INVALID_PAGES_STORAGE];
-       /* points to the next writable page within a block */
-       unsigned int next_page;
-       /* number of pages that are invalid, wrt host page size */
-       unsigned int nr_invalid_pages;
-
-       int state;
-
-       spinlock_t lock;
-       atomic_t data_cmnt_size; /* data pages committed to stable storage */
-};
-
-struct rrpc_lun {
-       struct rrpc *rrpc;
-
-       int id;
-       struct ppa_addr bppa;
-
-       struct rrpc_block *cur, *gc_cur;
-       struct rrpc_block *blocks;      /* Reference to block allocation */
-
-       struct list_head prio_list;     /* Blocks that may be GC'ed */
-       struct list_head wblk_list;     /* Queued blocks to be written to */
-
-       /* lun block lists */
-       struct list_head used_list;     /* In-use blocks */
-       struct list_head free_list;     /* Not used blocks i.e. released
-                                        * and ready for use
-                                        */
-       struct list_head bb_list;       /* Bad blocks. Mutually exclusive with
-                                        * free_list and used_list
-                                        */
-       unsigned int nr_free_blocks;    /* Number of unused blocks */
-
-       struct work_struct ws_gc;
-
-       int reserved_blocks;
-
-       spinlock_t lock;
-};
-
-struct rrpc {
-       struct nvm_tgt_dev *dev;
-       struct gendisk *disk;
-
-       sector_t soffset; /* logical sector offset */
-
-       int nr_luns;
-       struct rrpc_lun *luns;
-
-       /* calculated values */
-       unsigned long long nr_sects;
-
-       /* Write strategy variables. Move these into each for structure for each
-        * strategy
-        */
-       atomic_t next_lun; /* Whenever a page is written, this is updated
-                           * to point to the next write lun
-                           */
-
-       spinlock_t bio_lock;
-       struct bio_list requeue_bios;
-       struct work_struct ws_requeue;
-
-       /* Simple translation map of logical addresses to physical addresses.
-        * The logical addresses is known by the host system, while the physical
-        * addresses are used when writing to the disk block device.
-        */
-       struct rrpc_addr *trans_map;
-       /* also store a reverse map for garbage collection */
-       struct rrpc_rev_addr *rev_trans_map;
-       spinlock_t rev_lock;
-
-       struct rrpc_inflight inflights;
-
-       mempool_t *addr_pool;
-       mempool_t *page_pool;
-       mempool_t *gcb_pool;
-       mempool_t *rq_pool;
-
-       struct timer_list gc_timer;
-       struct workqueue_struct *krqd_wq;
-       struct workqueue_struct *kgc_wq;
-};
-
-struct rrpc_block_gc {
-       struct rrpc *rrpc;
-       struct rrpc_block *rblk;
-       struct work_struct ws_gc;
-};
-
-/* Logical to physical mapping */
-struct rrpc_addr {
-       u64 addr;
-       struct rrpc_block *rblk;
-};
-
-/* Physical to logical mapping */
-struct rrpc_rev_addr {
-       u64 addr;
-};
-
-static inline struct ppa_addr rrpc_linear_to_generic_addr(struct nvm_geo *geo,
-                                                         struct ppa_addr r)
-{
-       struct ppa_addr l;
-       int secs, pgs;
-       sector_t ppa = r.ppa;
-
-       l.ppa = 0;
-
-       div_u64_rem(ppa, geo->sec_per_pg, &secs);
-       l.g.sec = secs;
-
-       sector_div(ppa, geo->sec_per_pg);
-       div_u64_rem(ppa, geo->pgs_per_blk, &pgs);
-       l.g.pg = pgs;
-
-       return l;
-}
-
-static inline struct ppa_addr rrpc_recov_addr(struct nvm_tgt_dev *dev, u64 pba)
-{
-       return linear_to_generic_addr(&dev->geo, pba);
-}
-
-static inline u64 rrpc_blk_to_ppa(struct rrpc *rrpc, struct rrpc_block *rblk)
-{
-       struct nvm_tgt_dev *dev = rrpc->dev;
-       struct nvm_geo *geo = &dev->geo;
-       struct rrpc_lun *rlun = rblk->rlun;
-
-       return (rlun->id * geo->sec_per_lun) + (rblk->id * geo->sec_per_blk);
-}
-
-static inline sector_t rrpc_get_laddr(struct bio *bio)
-{
-       return bio->bi_iter.bi_sector / NR_PHY_IN_LOG;
-}
-
-static inline unsigned int rrpc_get_pages(struct bio *bio)
-{
-       return  bio->bi_iter.bi_size / RRPC_EXPOSED_PAGE_SIZE;
-}
-
-static inline sector_t rrpc_get_sector(sector_t laddr)
-{
-       return laddr * NR_PHY_IN_LOG;
-}
-
-static inline int request_intersects(struct rrpc_inflight_rq *r,
-                               sector_t laddr_start, sector_t laddr_end)
-{
-       return (laddr_end >= r->l_start) && (laddr_start <= r->l_end);
-}
-
-static int __rrpc_lock_laddr(struct rrpc *rrpc, sector_t laddr,
-                            unsigned int pages, struct rrpc_inflight_rq *r)
-{
-       sector_t laddr_end = laddr + pages - 1;
-       struct rrpc_inflight_rq *rtmp;
-
-       WARN_ON(irqs_disabled());
-
-       spin_lock_irq(&rrpc->inflights.lock);
-       list_for_each_entry(rtmp, &rrpc->inflights.reqs, list) {
-               if (unlikely(request_intersects(rtmp, laddr, laddr_end))) {
-                       /* existing, overlapping request, come back later */
-                       spin_unlock_irq(&rrpc->inflights.lock);
-                       return 1;
-               }
-       }
-
-       r->l_start = laddr;
-       r->l_end = laddr_end;
-
-       list_add_tail(&r->list, &rrpc->inflights.reqs);
-       spin_unlock_irq(&rrpc->inflights.lock);
-       return 0;
-}
-
-static inline int rrpc_lock_laddr(struct rrpc *rrpc, sector_t laddr,
-                                unsigned int pages,
-                                struct rrpc_inflight_rq *r)
-{
-       BUG_ON((laddr + pages) > rrpc->nr_sects);
-
-       return __rrpc_lock_laddr(rrpc, laddr, pages, r);
-}
-
-static inline struct rrpc_inflight_rq *rrpc_get_inflight_rq(struct nvm_rq *rqd)
-{
-       struct rrpc_rq *rrqd = nvm_rq_to_pdu(rqd);
-
-       return &rrqd->inflight_rq;
-}
-
-static inline int rrpc_lock_rq(struct rrpc *rrpc, struct bio *bio,
-                                                       struct nvm_rq *rqd)
-{
-       sector_t laddr = rrpc_get_laddr(bio);
-       unsigned int pages = rrpc_get_pages(bio);
-       struct rrpc_inflight_rq *r = rrpc_get_inflight_rq(rqd);
-
-       return rrpc_lock_laddr(rrpc, laddr, pages, r);
-}
-
-static inline void rrpc_unlock_laddr(struct rrpc *rrpc,
-                                               struct rrpc_inflight_rq *r)
-{
-       unsigned long flags;
-
-       spin_lock_irqsave(&rrpc->inflights.lock, flags);
-       list_del_init(&r->list);
-       spin_unlock_irqrestore(&rrpc->inflights.lock, flags);
-}
-
-static inline void rrpc_unlock_rq(struct rrpc *rrpc, struct nvm_rq *rqd)
-{
-       struct rrpc_inflight_rq *r = rrpc_get_inflight_rq(rqd);
-       uint8_t pages = rqd->nr_ppas;
-
-       BUG_ON((r->l_start + pages) > rrpc->nr_sects);
-
-       rrpc_unlock_laddr(rrpc, r);
-}
-
-#endif /* RRPC_H_ */