Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]
Groups > linux.kernel > #1525412 > unrolled thread
| Started by | "Javier González" <jg@lightnvm.io> |
|---|---|
| First post | 2016-11-18 15:50 +0100 |
| Last post | 2016-11-21 12:30 +0100 |
| Articles | 14 — 3 participants |
Back to article view | Back to linux.kernel
[PATCH 00/20] lightnvm: simplify media manager "Javier González" <jg@lightnvm.io> - 2016-11-18 15:50 +0100
[PATCH 10/20] lightnvm: use constant name instead of value "Javier González" <jg@lightnvm.io> - 2016-11-18 15:50 +0100
[PATCH 08/20] lightnvm: make address conversion functions global "Javier González" <jg@lightnvm.io> - 2016-11-18 15:50 +0100
[PATCH 09/20] lightnvm: remove unnecessary variables in rrpc "Javier González" <jg@lightnvm.io> - 2016-11-18 15:50 +0100
[PATCH 20/20] lightnvm: use target nvm on target-specific ops. "Javier González" <jg@lightnvm.io> - 2016-11-18 15:50 +0100
[PATCH 17/20] lightnvm: eliminate nvm_lun abstraction in mm "Javier González" <jg@lightnvm.io> - 2016-11-18 15:50 +0100
[PATCH 16/20] lightnvm: eliminate nvm_block abstraction on mm "Javier González" <jg@lightnvm.io> - 2016-11-18 15:50 +0100
Re: [PATCH 16/20] lightnvm: eliminate nvm_block abstraction on mm kbuild test robot <lkp@intel.com> - 2016-11-18 19:50 +0100
[PATCH 18/20] lightnvm: introduce helpers for generic ops in rrpc "Javier González" <jg@lightnvm.io> - 2016-11-18 15:50 +0100
[PATCH 03/20] lightnvm: export set bad block table "Javier González" <jg@lightnvm.io> - 2016-11-18 16:00 +0100
[PATCH 02/20] lightnvm: do not protect block 0 "Javier González" <jg@lightnvm.io> - 2016-11-18 16:00 +0100
[PATCH 01/20] lightnvm: enable to send hint to erase command "Javier González" <jg@lightnvm.io> - 2016-11-18 16:00 +0100
[PATCH 19/20] lightnvm: introduce max_phys_sects helper function "Javier González" <jg@lightnvm.io> - 2016-11-18 16:00 +0100
Re: [PATCH 00/20] lightnvm: simplify media manager Matias Bjørling <mb@lightnvm.io> - 2016-11-21 12:30 +0100
| From | "Javier González" <jg@lightnvm.io> |
|---|---|
| Date | 2016-11-18 15:50 +0100 |
| Subject | [PATCH 00/20] lightnvm: simplify media manager |
| Message-ID | <sESEh-8qT-3@gated-at.bofh.it> |
This patchset simplifies the generic media manager interface and moves lun and block functionality to lightnvm targets. This makes that rrpc-specific functionality is not exposed on the media manager. This is done in preparation for the pblk target. Matias: patches 0 - 4 have been posted before. Patches apply to you for-4.10/core Javier González (20): lightnvm: enable to send hint to erase command lightnvm: do not protect block 0 lightnvm: export set bad block table lightnvm: add ECC error codes lightnvm: rrpc: split bios of size > 256kb lightnvm: remove sysfs configuration interface lightnvm: cleanup unused target operations lightnvm: make address conversion functions global lightnvm: remove unnecessary variables in rrpc lightnvm: use constant name instead of value lightnvm: remove gen_lun abstraction lightnvm: manage lun partitions internally in mm lightnvm: move block provisioning to targets lightnvm: remove get_lun operation on gennvm lightnvm: remove debug lun statistics from gennvm lightnvm: eliminate nvm_block abstraction on mm lightnvm: eliminate nvm_lun abstraction in mm lightnvm: introduce helpers for generic ops in rrpc lightnvm: introduce max_phys_sects helper function lightnvm: use target nvm on target-specific ops. drivers/lightnvm/core.c | 306 ++++++++------------ drivers/lightnvm/gennvm.c | 664 ++++++++++++++++++++----------------------- drivers/lightnvm/gennvm.h | 34 ++- drivers/lightnvm/rrpc.c | 512 +++++++++++++++++++++------------ drivers/lightnvm/rrpc.h | 65 ++++- drivers/lightnvm/sysblk.c | 98 +++---- drivers/nvme/host/lightnvm.c | 15 +- include/linux/lightnvm.h | 231 +++++++-------- 8 files changed, 1005 insertions(+), 920 deletions(-) -- 2.7.4
[toc] | [next] | [standalone]
| From | "Javier González" <jg@lightnvm.io> |
|---|---|
| Date | 2016-11-18 15:50 +0100 |
| Subject | [PATCH 10/20] lightnvm: use constant name instead of value |
| Message-ID | <sESEi-8qT-35@gated-at.bofh.it> |
| In reply to | #1525412 |
There is a constant to refer to free blocks. Use it when marking bad
blocks instead of using a constant value
Signed-off-by: Javier González <javier@cnexlabs.com>
---
drivers/lightnvm/gennvm.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/lightnvm/gennvm.c b/drivers/lightnvm/gennvm.c
index e969e3a..aee5b72 100644
--- a/drivers/lightnvm/gennvm.c
+++ b/drivers/lightnvm/gennvm.c
@@ -279,7 +279,7 @@ static int gen_block_bb(struct gen_dev *gn, struct ppa_addr ppa,
lun = &gn->luns[(dev->luns_per_chnl * ppa.g.ch) + ppa.g.lun];
for (i = 0; i < nr_blks; i++) {
- if (blks[i] == 0)
+ if (blks[i] == NVM_BLK_T_FREE)
continue;
blk = &lun->vlun.blocks[i];
--
2.7.4
[toc] | [prev] | [next] | [standalone]
| From | "Javier González" <jg@lightnvm.io> |
|---|---|
| Date | 2016-11-18 15:50 +0100 |
| Subject | [PATCH 08/20] lightnvm: make address conversion functions global |
| Message-ID | <sESEi-8qT-37@gated-at.bofh.it> |
| In reply to | #1525412 |
Targets are assumed to used the same generic ppa format, where the
address is partitioned on ch:lun:block:pg:pl:sec. Thus, make the
function in charge of transforming the ppa address from a linear format
to the generic one available to all targets.
This function will be needed by the media manager in order to do target
mapping translations when targets are divided on different physical
partitions.
Signed-off-by: Javier González <javier@cnexlabs.com>
---
drivers/lightnvm/rrpc.c | 30 ------------------------------
include/linux/lightnvm.h | 30 ++++++++++++++++++++++++++++++
2 files changed, 30 insertions(+), 30 deletions(-)
diff --git a/drivers/lightnvm/rrpc.c b/drivers/lightnvm/rrpc.c
index 60ca9d4..9253acc 100644
--- a/drivers/lightnvm/rrpc.c
+++ b/drivers/lightnvm/rrpc.c
@@ -136,36 +136,6 @@ static u64 block_to_addr(struct rrpc *rrpc, struct rrpc_block *rblk)
return blk->id * rrpc->dev->sec_per_blk;
}
-static struct ppa_addr linear_to_generic_addr(struct nvm_dev *dev,
- struct ppa_addr r)
-{
- struct ppa_addr l;
- int secs, pgs, blks, luns;
- sector_t ppa = r.ppa;
-
- l.ppa = 0;
-
- div_u64_rem(ppa, dev->sec_per_pg, &secs);
- l.g.sec = secs;
-
- sector_div(ppa, dev->sec_per_pg);
- div_u64_rem(ppa, dev->pgs_per_blk, &pgs);
- l.g.pg = pgs;
-
- sector_div(ppa, dev->pgs_per_blk);
- div_u64_rem(ppa, dev->blks_per_lun, &blks);
- l.g.blk = blks;
-
- sector_div(ppa, dev->blks_per_lun);
- div_u64_rem(ppa, dev->luns_per_chnl, &luns);
- l.g.lun = luns;
-
- sector_div(ppa, dev->luns_per_chnl);
- l.g.ch = ppa;
-
- return l;
-}
-
static struct ppa_addr rrpc_ppa_to_gaddr(struct nvm_dev *dev, u64 addr)
{
struct ppa_addr paddr;
diff --git a/include/linux/lightnvm.h b/include/linux/lightnvm.h
index e598308..98278a9 100644
--- a/include/linux/lightnvm.h
+++ b/include/linux/lightnvm.h
@@ -361,6 +361,36 @@ struct nvm_dev {
spinlock_t lock;
};
+static inline struct ppa_addr linear_to_generic_addr(struct nvm_dev *dev,
+ struct ppa_addr r)
+{
+ struct ppa_addr l;
+ int secs, pgs, blks, luns;
+ sector_t ppa = r.ppa;
+
+ l.ppa = 0;
+
+ div_u64_rem(ppa, dev->sec_per_pg, &secs);
+ l.g.sec = secs;
+
+ sector_div(ppa, dev->sec_per_pg);
+ div_u64_rem(ppa, dev->pgs_per_blk, &pgs);
+ l.g.pg = pgs;
+
+ sector_div(ppa, dev->pgs_per_blk);
+ div_u64_rem(ppa, dev->blks_per_lun, &blks);
+ l.g.blk = blks;
+
+ sector_div(ppa, dev->blks_per_lun);
+ div_u64_rem(ppa, dev->luns_per_chnl, &luns);
+ l.g.lun = luns;
+
+ sector_div(ppa, dev->luns_per_chnl);
+ l.g.ch = ppa;
+
+ return l;
+}
+
static inline struct ppa_addr generic_to_dev_addr(struct nvm_dev *dev,
struct ppa_addr r)
{
--
2.7.4
[toc] | [prev] | [next] | [standalone]
| From | "Javier González" <jg@lightnvm.io> |
|---|---|
| Date | 2016-11-18 15:50 +0100 |
| Subject | [PATCH 09/20] lightnvm: remove unnecessary variables in rrpc |
| Message-ID | <sESEi-8qT-39@gated-at.bofh.it> |
| In reply to | #1525412 |
Before vectored I/Os were supported on rrpc, the physical address was
stored as part of the nvm_rqd request. This variable become obsolete
when the ppa_list was introduced. Cleanup this variable.
Signed-off-by: Javier González <javier@cnexlabs.com>
---
drivers/lightnvm/rrpc.c | 6 ------
drivers/lightnvm/rrpc.h | 3 ---
2 files changed, 9 deletions(-)
diff --git a/drivers/lightnvm/rrpc.c b/drivers/lightnvm/rrpc.c
index 9253acc..34d2ebf 100644
--- a/drivers/lightnvm/rrpc.c
+++ b/drivers/lightnvm/rrpc.c
@@ -758,7 +758,6 @@ static int rrpc_read_ppalist_rq(struct rrpc *rrpc, struct bio *bio,
static int rrpc_read_rq(struct rrpc *rrpc, struct bio *bio, struct nvm_rq *rqd,
unsigned long flags)
{
- struct rrpc_rq *rrqd = nvm_rq_to_pdu(rqd);
int is_gc = flags & NVM_IOTYPE_GC;
sector_t laddr = rrpc_get_laddr(bio);
struct rrpc_addr *gp;
@@ -778,7 +777,6 @@ static int rrpc_read_rq(struct rrpc *rrpc, struct bio *bio, struct nvm_rq *rqd,
}
rqd->opcode = NVM_OP_HBREAD;
- rrqd->addr = gp;
return NVM_IO_OK;
}
@@ -821,7 +819,6 @@ static int rrpc_write_ppalist_rq(struct rrpc *rrpc, struct bio *bio,
static int rrpc_write_rq(struct rrpc *rrpc, struct bio *bio,
struct nvm_rq *rqd, unsigned long flags)
{
- struct rrpc_rq *rrqd = nvm_rq_to_pdu(rqd);
struct rrpc_addr *p;
int is_gc = flags & NVM_IOTYPE_GC;
sector_t laddr = rrpc_get_laddr(bio);
@@ -839,7 +836,6 @@ static int rrpc_write_rq(struct rrpc *rrpc, struct bio *bio,
rqd->ppa_addr = rrpc_ppa_to_gaddr(rrpc->dev, p->addr);
rqd->opcode = NVM_OP_HBWRITE;
- rrqd->addr = p;
return NVM_IO_OK;
}
@@ -1389,7 +1385,6 @@ static void *rrpc_init(struct nvm_dev *dev, struct gendisk *tdisk,
INIT_WORK(&rrpc->ws_requeue, rrpc_requeue);
rrpc->nr_luns = lun_end - lun_begin + 1;
- rrpc->total_blocks = (unsigned long)dev->blks_per_lun * rrpc->nr_luns;
rrpc->nr_sects = (unsigned long long)dev->sec_per_lun * rrpc->nr_luns;
/* simple round-robin strategy */
@@ -1409,7 +1404,6 @@ static void *rrpc_init(struct nvm_dev *dev, struct gendisk *tdisk,
}
rrpc->poffset = dev->sec_per_lun * lun_begin;
- rrpc->lun_offset = lun_begin;
ret = rrpc_core_init(rrpc);
if (ret) {
diff --git a/drivers/lightnvm/rrpc.h b/drivers/lightnvm/rrpc.h
index 5e87d52..bed49ba 100644
--- a/drivers/lightnvm/rrpc.h
+++ b/drivers/lightnvm/rrpc.h
@@ -48,7 +48,6 @@ struct rrpc_inflight_rq {
struct rrpc_rq {
struct rrpc_inflight_rq inflight_rq;
- struct rrpc_addr *addr;
unsigned long flags;
};
@@ -92,14 +91,12 @@ struct rrpc {
sector_t soffset; /* logical sector offset */
u64 poffset; /* physical page offset */
- int lun_offset;
int nr_luns;
struct rrpc_lun *luns;
/* calculated values */
unsigned long long nr_sects;
- unsigned long total_blocks;
/* Write strategy variables. Move these into each for structure for each
* strategy
--
2.7.4
[toc] | [prev] | [next] | [standalone]
| From | "Javier González" <jg@lightnvm.io> |
|---|---|
| Date | 2016-11-18 15:50 +0100 |
| Subject | [PATCH 20/20] lightnvm: use target nvm on target-specific ops. |
| Message-ID | <sESEi-8qT-25@gated-at.bofh.it> |
| In reply to | #1525412 |
On target-specific operations pass on nvm_tgt_dev instead of the generic
nvm device.
Signed-off-by: Javier González <javier@cnexlabs.com>
---
drivers/lightnvm/core.c | 15 ++++++++++-----
drivers/lightnvm/rrpc.c | 14 +++++++-------
drivers/lightnvm/rrpc.h | 2 +-
include/linux/lightnvm.h | 6 +++---
4 files changed, 21 insertions(+), 16 deletions(-)
diff --git a/drivers/lightnvm/core.c b/drivers/lightnvm/core.c
index 3d8eaac..07bf989 100644
--- a/drivers/lightnvm/core.c
+++ b/drivers/lightnvm/core.c
@@ -86,8 +86,7 @@ void *nvm_dev_dma_alloc(struct nvm_dev *dev, gfp_t mem_flags,
}
EXPORT_SYMBOL(nvm_dev_dma_alloc);
-void nvm_dev_dma_free(struct nvm_dev *dev, void *addr,
- dma_addr_t dma_handler)
+void nvm_dev_dma_free(struct nvm_dev *dev, void *addr, dma_addr_t dma_handler)
{
dev->ops->dev_dma_free(dev->dma_pool, addr, dma_handler);
}
@@ -227,9 +226,11 @@ int nvm_erase_blk(struct nvm_tgt_dev *tgt_dev, struct ppa_addr *p, int flags)
}
EXPORT_SYMBOL(nvm_erase_blk);
-int nvm_get_l2p_tbl(struct nvm_dev *dev, u64 slba, u32 nlb,
+int nvm_get_l2p_tbl(struct nvm_tgt_dev *tgt_dev, u64 slba, u32 nlb,
nvm_l2p_update_fn *update_l2p, void *priv)
{
+ struct nvm_dev *dev = tgt_dev->parent;
+
if (!dev->ops->get_l2p_tbl)
return 0;
@@ -237,14 +238,18 @@ int nvm_get_l2p_tbl(struct nvm_dev *dev, u64 slba, u32 nlb,
}
EXPORT_SYMBOL(nvm_get_l2p_tbl);
-int nvm_get_area(struct nvm_dev *dev, sector_t *lba, sector_t len)
+int nvm_get_area(struct nvm_tgt_dev *tgt_dev, sector_t *lba, sector_t len)
{
+ struct nvm_dev *dev = tgt_dev->parent;
+
return dev->mt->get_area(dev, lba, len);
}
EXPORT_SYMBOL(nvm_get_area);
-void nvm_put_area(struct nvm_dev *dev, sector_t lba)
+void nvm_put_area(struct nvm_tgt_dev *tgt_dev, sector_t lba)
{
+ struct nvm_dev *dev = tgt_dev->parent;
+
dev->mt->put_area(dev, lba);
}
EXPORT_SYMBOL(nvm_put_area);
diff --git a/drivers/lightnvm/rrpc.c b/drivers/lightnvm/rrpc.c
index 150b5f30..a9fe89d 100644
--- a/drivers/lightnvm/rrpc.c
+++ b/drivers/lightnvm/rrpc.c
@@ -983,8 +983,8 @@ static int rrpc_submit_io(struct rrpc *rrpc, struct bio *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);
+ nvm_dev_dma_free(dev->parent, rqd->ppa_list,
+ rqd->dma_ppa_list);
}
return NVM_IO_ERR;
}
@@ -1116,7 +1116,7 @@ static int rrpc_l2p_update(u64 slba, u32 nlb, __le64 *entries, void *private)
div_u64_rem(pba, rrpc->nr_sects, &mod);
- gaddr = rrpc_recov_addr(dev->parent, pba);
+ gaddr = rrpc_recov_addr(dev, pba);
rlun = rrpc_ppa_to_lun(rrpc, gaddr);
if (!rlun) {
pr_err("rrpc: unmapped 'ch:%d,lun:%d' l2p corruption\n",
@@ -1168,8 +1168,8 @@ static int rrpc_map_init(struct rrpc *rrpc)
}
/* Bring up the mapping table from device */
- ret = nvm_get_l2p_tbl(dev->parent, rrpc->soffset, rrpc->nr_sects,
- rrpc_l2p_update, rrpc);
+ 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;
@@ -1374,7 +1374,7 @@ static int rrpc_area_init(struct rrpc *rrpc, sector_t *begin)
size >>= 9;
- ret = nvm_get_area(dev->parent, begin, size);
+ ret = nvm_get_area(dev, begin, size);
if (!ret)
*begin >>= (ilog2(dev->geo.sec_size) - 9);
@@ -1386,7 +1386,7 @@ 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->parent, begin);
+ nvm_put_area(dev, begin);
}
static void rrpc_free(struct rrpc *rrpc)
diff --git a/drivers/lightnvm/rrpc.h b/drivers/lightnvm/rrpc.h
index bc8adba..94e4d73 100644
--- a/drivers/lightnvm/rrpc.h
+++ b/drivers/lightnvm/rrpc.h
@@ -184,7 +184,7 @@ static inline struct ppa_addr rrpc_linear_to_generic_addr(struct nvm_geo *geo,
return l;
}
-static inline struct ppa_addr rrpc_recov_addr(struct nvm_dev *dev, u64 pba)
+static inline struct ppa_addr rrpc_recov_addr(struct nvm_tgt_dev *dev, u64 pba)
{
return linear_to_generic_addr(&dev->geo, pba);
}
diff --git a/include/linux/lightnvm.h b/include/linux/lightnvm.h
index 44780ef..5a8217f 100644
--- a/include/linux/lightnvm.h
+++ b/include/linux/lightnvm.h
@@ -538,10 +538,10 @@ extern int nvm_set_rqd_ppalist(struct nvm_dev *, struct nvm_rq *,
extern void nvm_free_rqd_ppalist(struct nvm_dev *, struct nvm_rq *);
extern int nvm_erase_ppa(struct nvm_dev *, struct ppa_addr *, int, int);
extern int nvm_erase_blk(struct nvm_tgt_dev *, struct ppa_addr *, int);
-extern int nvm_get_l2p_tbl(struct nvm_dev *, u64, u32, nvm_l2p_update_fn *,
+extern int nvm_get_l2p_tbl(struct nvm_tgt_dev *, u64, u32, nvm_l2p_update_fn *,
void *);
-extern int nvm_get_area(struct nvm_dev *, sector_t *, sector_t);
-extern void nvm_put_area(struct nvm_dev *, sector_t);
+extern int nvm_get_area(struct nvm_tgt_dev *, sector_t *, sector_t);
+extern void nvm_put_area(struct nvm_tgt_dev *, sector_t);
extern void nvm_end_io(struct nvm_rq *, int);
extern int nvm_submit_ppa(struct nvm_dev *, struct ppa_addr *, int, int, int,
void *, int);
--
2.7.4
[toc] | [prev] | [next] | [standalone]
| From | "Javier González" <jg@lightnvm.io> |
|---|---|
| Date | 2016-11-18 15:50 +0100 |
| Subject | [PATCH 17/20] lightnvm: eliminate nvm_lun abstraction in mm |
| Message-ID | <sESEi-8qT-33@gated-at.bofh.it> |
| In reply to | #1525412 |
In order to naturally support multi-target instances on an Open-Channel
SSD, targets should own the LUNs they get blocks from and manage
provisioning internally. This is done in several steps.
Since targets own the LUNs the are instantiated on top of and manage the
free block list internally, there is no need for a LUN abstraction in
the media manager. LUNs are intrinsically managed as in the physical
layout (ch:0,lun:0, ..., ch:0,lun:n, ch:1,lun:0, ch:1,lun:n, ...,
ch:m,lun:0, ch:m,lun:n) and given to the targets based on the target
creation ioctl. This simplifies LUN management and clears the path for a
partition manager to sit directly underneath LightNVM targets.
Signed-off-by: Javier González <javier@cnexlabs.com>
---
drivers/lightnvm/core.c | 14 +-
drivers/lightnvm/gennvm.c | 336 ++++++++++++++++++++++++++++++++++---------
drivers/lightnvm/gennvm.h | 20 ++-
drivers/lightnvm/rrpc.c | 142 +++++++++---------
drivers/lightnvm/rrpc.h | 32 ++++-
drivers/nvme/host/lightnvm.c | 3 +
include/linux/lightnvm.h | 41 +++---
7 files changed, 422 insertions(+), 166 deletions(-)
diff --git a/drivers/lightnvm/core.c b/drivers/lightnvm/core.c
index 691b16f..23d582f 100644
--- a/drivers/lightnvm/core.c
+++ b/drivers/lightnvm/core.c
@@ -203,15 +203,19 @@ int nvm_set_bb_tbl(struct nvm_dev *dev, struct ppa_addr *ppas, int nr_ppas,
}
EXPORT_SYMBOL(nvm_set_bb_tbl);
-int nvm_submit_io(struct nvm_dev *dev, struct nvm_rq *rqd)
+int nvm_submit_io(struct nvm_tgt_dev *tgt_dev, struct nvm_rq *rqd)
{
- return dev->mt->submit_io(dev, rqd);
+ struct nvm_dev *dev = tgt_dev->parent;
+
+ return dev->mt->submit_io(tgt_dev, rqd);
}
EXPORT_SYMBOL(nvm_submit_io);
-int nvm_erase_blk(struct nvm_dev *dev, struct ppa_addr *p, int flags)
+int nvm_erase_blk(struct nvm_tgt_dev *tgt_dev, struct ppa_addr *p, int flags)
{
- return dev->mt->erase_blk(dev, p, flags);
+ struct nvm_dev *dev = tgt_dev->parent;
+
+ return dev->mt->erase_blk(tgt_dev, p, flags);
}
EXPORT_SYMBOL(nvm_erase_blk);
@@ -350,7 +354,7 @@ static int __nvm_submit_ppa(struct nvm_dev *dev, struct nvm_rq *rqd, int opcode,
nvm_generic_to_addr_mode(dev, rqd);
- rqd->dev = dev;
+ rqd->dev = NULL;
rqd->opcode = opcode;
rqd->flags = flags;
rqd->bio = bio;
diff --git a/drivers/lightnvm/gennvm.c b/drivers/lightnvm/gennvm.c
index 19c3924..5d7c8c4 100644
--- a/drivers/lightnvm/gennvm.c
+++ b/drivers/lightnvm/gennvm.c
@@ -38,8 +38,6 @@ static const struct block_device_operations gen_fops = {
static int gen_reserve_luns(struct nvm_dev *dev, struct nvm_target *t,
int lun_begin, int lun_end)
{
- struct gen_dev *gn = dev->mp;
- struct nvm_lun *lun;
int i;
for (i = lun_begin; i <= lun_end; i++) {
@@ -47,35 +45,50 @@ static int gen_reserve_luns(struct nvm_dev *dev, struct nvm_target *t,
pr_err("nvm: lun %d already allocated\n", i);
goto err;
}
-
- lun = &gn->luns[i];
- list_add_tail(&lun->list, &t->lun_list);
}
return 0;
err:
- while (--i > lun_begin) {
- lun = &gn->luns[i];
+ while (--i > lun_begin)
clear_bit(i, dev->lun_map);
- list_del(&lun->list);
- }
return -EBUSY;
}
-static void gen_release_luns(struct nvm_dev *dev, struct nvm_target *t)
+static void gen_release_luns_err(struct nvm_dev *dev, int lun_begin,
+ int lun_end)
{
- struct nvm_lun *lun, *tmp;
+ int i;
- list_for_each_entry_safe(lun, tmp, &t->lun_list, list) {
- WARN_ON(!test_and_clear_bit(lun->id, dev->lun_map));
- list_del(&lun->list);
- }
+ for (i = lun_begin; i <= lun_end; i++)
+ WARN_ON(!test_and_clear_bit(i, dev->lun_map));
}
static void gen_remove_tgt_dev(struct nvm_tgt_dev *tgt_dev)
{
+ struct nvm_dev *dev = tgt_dev->parent;
+ struct gen_dev_map *dev_map = tgt_dev->map;
+ int i, j;
+
+ for (i = 0; i < dev_map->nr_chnls; i++) {
+ struct gen_ch_map *ch_map = &dev_map->chnls[i];
+ int *lun_offs = ch_map->lun_offs;
+ int ch = i + ch_map->ch_off;
+
+ for (j = 0; j < ch_map->nr_luns; j++) {
+ int lun = j + lun_offs[j];
+ int lunid = (ch * dev->geo.luns_per_chnl) + lun;
+
+ WARN_ON(!test_and_clear_bit(lunid, dev->lun_map));
+ }
+
+ kfree(ch_map->lun_offs);
+ }
+
+ kfree(dev_map->chnls);
+ kfree(dev_map);
+ kfree(tgt_dev->luns);
kfree(tgt_dev);
}
@@ -83,24 +96,103 @@ static struct nvm_tgt_dev *gen_create_tgt_dev(struct nvm_dev *dev,
int lun_begin, int lun_end)
{
struct nvm_tgt_dev *tgt_dev = NULL;
+ struct gen_dev_map *dev_rmap = dev->rmap;
+ struct gen_dev_map *dev_map;
+ struct ppa_addr *luns;
int nr_luns = lun_end - lun_begin + 1;
+ int luns_left = nr_luns;
+ int nr_chnls = nr_luns / dev->geo.luns_per_chnl;
+ int nr_chnls_mod = nr_luns % dev->geo.luns_per_chnl;
+ int bch = lun_begin / dev->geo.luns_per_chnl;
+ int blun = lun_begin % dev->geo.luns_per_chnl;
+ int lunid = 0;
+ int lun_balanced = 1;
+ int prev_nr_luns;
+ int i, j;
+
+ nr_chnls = nr_luns / dev->geo.luns_per_chnl;
+ nr_chnls = (nr_chnls_mod == 0) ? nr_chnls : nr_chnls + 1;
+
+ dev_map = kmalloc(sizeof(struct gen_dev_map), GFP_KERNEL);
+ if (!dev_map)
+ goto err_dev;
+
+ dev_map->chnls = kcalloc(nr_chnls, sizeof(struct gen_ch_map),
+ GFP_KERNEL);
+ if (!dev_map->chnls)
+ goto err_chnls;
+
+ luns = kcalloc(nr_luns, sizeof(struct ppa_addr), GFP_KERNEL);
+ if (!luns)
+ goto err_luns;
+
+ prev_nr_luns = (luns_left > dev->geo.luns_per_chnl) ?
+ dev->geo.luns_per_chnl : luns_left;
+ for (i = 0; i < nr_chnls; i++) {
+ struct gen_ch_map *ch_rmap = &dev_rmap->chnls[i + bch];
+ int *lun_roffs = ch_rmap->lun_offs;
+ struct gen_ch_map *ch_map = &dev_map->chnls[i];
+ int *lun_offs;
+ int luns_in_chnl = (luns_left > dev->geo.luns_per_chnl) ?
+ dev->geo.luns_per_chnl : luns_left;
+
+ if (lun_balanced && prev_nr_luns != luns_in_chnl)
+ lun_balanced = 0;
+
+ ch_map->ch_off = ch_rmap->ch_off = bch;
+ ch_map->nr_luns = luns_in_chnl;
+
+ lun_offs = kcalloc(luns_in_chnl, sizeof(int), GFP_KERNEL);
+ if (!lun_offs)
+ goto err_ch;
+
+ for (j = 0; j < luns_in_chnl; j++) {
+ luns[lunid].ppa = 0;
+ luns[lunid].g.ch = i;
+ luns[lunid++].g.lun = j;
+
+ lun_offs[j] = blun;
+ lun_roffs[j + blun] = blun;
+ }
+
+ ch_map->lun_offs = lun_offs;
+
+ /* when starting a new channel, lun offset is reset */
+ blun = 0;
+ luns_left -= luns_in_chnl;
+ }
+
+ dev_map->nr_chnls = nr_chnls;
tgt_dev = kmalloc(sizeof(struct nvm_tgt_dev), GFP_KERNEL);
if (!tgt_dev)
- goto out;
+ goto err_ch;
memcpy(&tgt_dev->geo, &dev->geo, sizeof(struct nvm_geo));
- tgt_dev->geo.nr_chnls = (nr_luns / (dev->geo.luns_per_chnl + 1)) + 1;
+ /* Target device only owns a portion of the physical device */
+ tgt_dev->geo.nr_chnls = nr_chnls;
tgt_dev->geo.nr_luns = nr_luns;
+ tgt_dev->geo.luns_per_chnl = (lun_balanced) ? prev_nr_luns : -1;
tgt_dev->total_secs = nr_luns * tgt_dev->geo.sec_per_lun;
tgt_dev->q = dev->q;
tgt_dev->ops = dev->ops;
tgt_dev->mt = dev->mt;
+ tgt_dev->map = dev_map;
+ tgt_dev->luns = luns;
memcpy(&tgt_dev->identity, &dev->identity, sizeof(struct nvm_id));
tgt_dev->parent = dev;
-out:
+ return tgt_dev;
+err_ch:
+ while (--i > 0)
+ kfree(dev_map->chnls[i].lun_offs);
+ kfree(luns);
+err_luns:
+ kfree(dev_map->chnls);
+err_chnls:
+ kfree(dev_map);
+err_dev:
return tgt_dev;
}
@@ -134,14 +226,14 @@ static int gen_create_tgt(struct nvm_dev *dev, struct nvm_ioctl_create *create)
if (!t)
return -ENOMEM;
- INIT_LIST_HEAD(&t->lun_list);
-
if (gen_reserve_luns(dev, t, s->lun_begin, s->lun_end))
goto err_t;
tgt_dev = gen_create_tgt_dev(dev, s->lun_begin, s->lun_end);
- if (!tgt_dev)
+ if (!tgt_dev) {
+ pr_err("nvm: could not create target device\n");
goto err_reserve;
+ }
tqueue = blk_alloc_queue_node(GFP_KERNEL, dev->q->node);
if (!tqueue)
@@ -159,7 +251,7 @@ static int gen_create_tgt(struct nvm_dev *dev, struct nvm_ioctl_create *create)
tdisk->fops = &gen_fops;
tdisk->queue = tqueue;
- targetdata = tt->init(tgt_dev, tdisk, &t->lun_list);
+ targetdata = tt->init(tgt_dev, tdisk);
if (IS_ERR(targetdata))
goto err_init;
@@ -187,7 +279,7 @@ static int gen_create_tgt(struct nvm_dev *dev, struct nvm_ioctl_create *create)
err_dev:
kfree(tgt_dev);
err_reserve:
- gen_release_luns(dev, t);
+ gen_release_luns_err(dev, s->lun_begin, s->lun_end);
err_t:
kfree(t);
return -ENOMEM;
@@ -205,7 +297,6 @@ static void __gen_remove_target(struct nvm_target *t)
if (tt->exit)
tt->exit(tdisk->private_data);
- gen_release_luns(t->dev->parent, t);
gen_remove_tgt_dev(t->dev);
put_disk(tdisk);
@@ -306,51 +397,54 @@ static void gen_put_area(struct nvm_dev *dev, sector_t begin)
spin_unlock(&dev->lock);
}
-static void gen_luns_free(struct nvm_dev *dev)
-{
- struct gen_dev *gn = dev->mp;
-
- kfree(gn->luns);
-}
-
-static int gen_luns_init(struct nvm_dev *dev, struct gen_dev *gn)
-{
- struct nvm_geo *geo = &dev->geo;
- struct nvm_lun *lun;
- int i;
-
- gn->luns = kcalloc(geo->nr_luns, sizeof(struct nvm_lun), GFP_KERNEL);
- if (!gn->luns)
- return -ENOMEM;
-
- gen_for_each_lun(gn, lun, i) {
- INIT_LIST_HEAD(&lun->list);
-
- lun->id = i;
- lun->lun_id = i % geo->luns_per_chnl;
- lun->chnl_id = i / geo->luns_per_chnl;
- }
- return 0;
-}
-
static void gen_free(struct nvm_dev *dev)
{
- gen_luns_free(dev);
kfree(dev->mp);
+ kfree(dev->rmap);
dev->mp = NULL;
}
static int gen_register(struct nvm_dev *dev)
{
struct gen_dev *gn;
- int ret;
+ struct gen_dev_map *dev_rmap;
+ int i, j;
if (!try_module_get(THIS_MODULE))
return -ENODEV;
gn = kzalloc(sizeof(struct gen_dev), GFP_KERNEL);
if (!gn)
- return -ENOMEM;
+ goto err_gn;
+
+ dev_rmap = kmalloc(sizeof(struct gen_dev_map), GFP_KERNEL);
+ if (!dev_rmap)
+ goto err_rmap;
+
+ dev_rmap->chnls = kcalloc(dev->geo.nr_chnls, sizeof(struct gen_ch_map),
+ GFP_KERNEL);
+ if (!dev_rmap->chnls)
+ goto err_chnls;
+
+ for (i = 0; i < dev->geo.nr_chnls; i++) {
+ struct gen_ch_map *ch_rmap;
+ int *lun_roffs;
+ int luns_in_chnl = dev->geo.luns_per_chnl;
+
+ ch_rmap = &dev_rmap->chnls[i];
+
+ ch_rmap->ch_off = -1;
+ ch_rmap->nr_luns = luns_in_chnl;
+
+ lun_roffs = kcalloc(luns_in_chnl, sizeof(int), GFP_KERNEL);
+ if (!lun_roffs)
+ goto err_ch;
+
+ for (j = 0; j < luns_in_chnl; j++)
+ lun_roffs[j] = -1;
+
+ ch_rmap->lun_offs = lun_roffs;
+ }
gn->dev = dev;
gn->nr_luns = dev->geo.nr_luns;
@@ -358,18 +452,19 @@ static int gen_register(struct nvm_dev *dev)
mutex_init(&gn->lock);
INIT_LIST_HEAD(&gn->targets);
dev->mp = gn;
-
- ret = gen_luns_init(dev, gn);
- if (ret) {
- pr_err("gen: could not initialize luns\n");
- goto err;
- }
+ dev->rmap = dev_rmap;
return 1;
-err:
+err_ch:
+ while (--i >= 0)
+ kfree(dev_rmap->chnls[i].lun_offs);
+err_chnls:
+ kfree(dev_rmap);
+err_rmap:
gen_free(dev);
+err_gn:
module_put(THIS_MODULE);
- return ret;
+ return -ENOMEM;
}
static void gen_unregister(struct nvm_dev *dev)
@@ -389,29 +484,137 @@ static void gen_unregister(struct nvm_dev *dev)
module_put(THIS_MODULE);
}
+enum {
+ TRANS_TGT_TO_DEV = 0x0,
+ TRANS_DEV_TO_TGT = 0x1,
+};
+
+
+static int gen_map_to_dev(struct nvm_tgt_dev *tgt_dev, struct ppa_addr *p)
+{
+ struct gen_dev_map *dev_map = tgt_dev->map;
+ struct gen_ch_map *ch_map = &dev_map->chnls[p->g.ch];
+ int lun_off = ch_map->lun_offs[p->g.lun];
+ struct nvm_dev *dev = tgt_dev->parent;
+ struct gen_dev_map *dev_rmap = dev->rmap;
+ struct gen_ch_map *ch_rmap;
+ int lun_roff;
+
+ p->g.ch += ch_map->ch_off;
+ p->g.lun += lun_off;
+
+ ch_rmap = &dev_rmap->chnls[p->g.ch];
+ lun_roff = ch_rmap->lun_offs[p->g.lun];
+
+ if (unlikely(ch_rmap->ch_off < 0 || lun_roff < 0)) {
+ pr_err("nvm: corrupted device partition table\n");
+ return -EINVAL;
+ }
+
+ return 0;
+}
+
+static int gen_map_to_tgt(struct nvm_tgt_dev *tgt_dev, struct ppa_addr *p)
+{
+ struct nvm_dev *dev = tgt_dev->parent;
+ struct gen_dev_map *dev_rmap = dev->rmap;
+ struct gen_ch_map *ch_rmap = &dev_rmap->chnls[p->g.ch];
+ int lun_roff = ch_rmap->lun_offs[p->g.lun];
+
+ p->g.ch -= ch_rmap->ch_off;
+ p->g.lun -= lun_roff;
+
+ return 0;
+}
+
+static int gen_trans_rq(struct nvm_tgt_dev *tgt_dev, struct nvm_rq *rqd,
+ int flag)
+{
+ gen_trans_fn *f;
+ int i;
+ int ret = 0;
+
+ f = (flag == TRANS_TGT_TO_DEV) ? gen_map_to_dev : gen_map_to_tgt;
+
+ if (rqd->nr_ppas == 1)
+ return f(tgt_dev, &rqd->ppa_addr);
+
+ for (i = 0; i < rqd->nr_ppas; i++) {
+ ret = f(tgt_dev, &rqd->ppa_list[i]);
+ if (ret)
+ goto out;
+ }
+
+out:
+ return ret;
+}
+
static void gen_end_io(struct nvm_rq *rqd)
{
+ struct nvm_tgt_dev *tgt_dev = rqd->dev;
struct nvm_tgt_instance *ins = rqd->ins;
+ /* Convert address space */
+ if (tgt_dev)
+ gen_trans_rq(tgt_dev, rqd, TRANS_DEV_TO_TGT);
+
ins->tt->end_io(rqd);
}
-static int gen_submit_io(struct nvm_dev *dev, struct nvm_rq *rqd)
+static int gen_submit_io(struct nvm_tgt_dev *tgt_dev, struct nvm_rq *rqd)
{
+ struct nvm_dev *dev = tgt_dev->parent;
+
if (!dev->ops->submit_io)
return -ENODEV;
/* Convert address space */
+ gen_trans_rq(tgt_dev, rqd, TRANS_TGT_TO_DEV);
nvm_generic_to_addr_mode(dev, rqd);
- rqd->dev = dev;
+ rqd->dev = tgt_dev;
rqd->end_io = gen_end_io;
return dev->ops->submit_io(dev, rqd);
}
-static int gen_erase_blk(struct nvm_dev *dev, struct ppa_addr *p, int flags)
+static int gen_erase_blk(struct nvm_tgt_dev *tgt_dev, struct ppa_addr *p,
+ int flags)
{
- return nvm_erase_ppa(dev, p, 1, flags);
+ /* Convert address space */
+ gen_map_to_dev(tgt_dev, p);
+
+ return nvm_erase_ppa(tgt_dev->parent, p, 1, flags);
+}
+
+static void gen_part_to_tgt(struct nvm_dev *dev, sector_t *entries,
+ int len)
+{
+ struct nvm_geo *geo = &dev->geo;
+ struct gen_dev_map *dev_rmap = dev->rmap;
+ u64 i;
+
+ for (i = 0; i < len; i++) {
+ struct gen_ch_map *ch_rmap;
+ int *lun_roffs;
+ struct ppa_addr gaddr;
+ u64 pba = le64_to_cpu(entries[i]);
+ int off;
+ u64 diff;
+
+ if (!pba)
+ continue;
+
+ gaddr = linear_to_generic_addr(geo, pba);
+ ch_rmap = &dev_rmap->chnls[gaddr.g.ch];
+ lun_roffs = ch_rmap->lun_offs;
+
+ off = gaddr.g.ch * geo->luns_per_chnl + gaddr.g.lun;
+
+ diff = ((ch_rmap->ch_off * geo->luns_per_chnl) +
+ (lun_roffs[gaddr.g.lun])) * geo->sec_per_lun;
+
+ entries[i] -= cpu_to_le64(diff);
+ }
}
static struct nvmm_type gen = {
@@ -430,6 +633,7 @@ static struct nvmm_type gen = {
.get_area = gen_get_area,
.put_area = gen_put_area,
+ .part_to_tgt = gen_part_to_tgt,
};
static int __init gen_module_init(void)
diff --git a/drivers/lightnvm/gennvm.h b/drivers/lightnvm/gennvm.h
index d167f39..6a4b3f3 100644
--- a/drivers/lightnvm/gennvm.h
+++ b/drivers/lightnvm/gennvm.h
@@ -24,19 +24,37 @@ struct gen_dev {
struct nvm_dev *dev;
int nr_luns;
- struct nvm_lun *luns;
struct list_head area_list;
struct mutex lock;
struct list_head targets;
};
+/* Map between virtual and physical channel and lun */
+struct gen_ch_map {
+ int ch_off;
+ int nr_luns;
+ int *lun_offs;
+};
+
+struct gen_dev_map {
+ struct gen_ch_map *chnls;
+ int nr_chnls;
+};
+
struct gen_area {
struct list_head list;
sector_t begin;
sector_t end; /* end is excluded */
};
+static inline void *ch_map_to_lun_offs(struct gen_ch_map *ch_map)
+{
+ return ch_map + 1;
+}
+
+typedef int (gen_trans_fn)(struct nvm_tgt_dev *, struct ppa_addr *);
+
#define gen_for_each_lun(bm, lun, i) \
for ((i) = 0, lun = &(bm)->luns[0]; \
(i) < (bm)->nr_luns; (i)++, lun = &(bm)->luns[(i)])
diff --git a/drivers/lightnvm/rrpc.c b/drivers/lightnvm/rrpc.c
index fad117b..08a633c 100644
--- a/drivers/lightnvm/rrpc.c
+++ b/drivers/lightnvm/rrpc.c
@@ -45,7 +45,7 @@ static void rrpc_page_invalidate(struct rrpc *rrpc, struct rrpc_addr *a)
spin_unlock(&rblk->lock);
- rrpc->rev_trans_map[a->addr - rrpc->poffset].addr = ADDR_EMPTY;
+ rrpc->rev_trans_map[a->addr].addr = ADDR_EMPTY;
}
static void rrpc_invalidate_range(struct rrpc *rrpc, sector_t slba,
@@ -127,28 +127,25 @@ 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;
- struct nvm_lun *lun = rlun->parent;
- return lun->id * dev->geo.sec_per_blk;
+ return rlun->id * dev->geo.sec_per_blk;
}
-/* Calculate global addr for the given block */
-static u64 block_to_addr(struct rrpc *rrpc, struct rrpc_block *rblk)
+static struct ppa_addr rrpc_ppa_to_gaddr(struct nvm_tgt_dev *dev,
+ struct rrpc_addr *gp)
{
- struct nvm_tgt_dev *dev = rrpc->dev;
- struct nvm_geo *geo = &dev->geo;
+ struct rrpc_block *rblk = gp->rblk;
struct rrpc_lun *rlun = rblk->rlun;
- struct nvm_lun *lun = rlun->parent;
-
- return lun->id * geo->sec_per_lun + rblk->id * geo->sec_per_blk;
-}
-
-static struct ppa_addr rrpc_ppa_to_gaddr(struct nvm_tgt_dev *dev, u64 addr)
-{
+ u64 addr = gp->addr;
struct ppa_addr paddr;
paddr.ppa = addr;
- return linear_to_generic_addr(&dev->geo, paddr);
+ 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 */
@@ -216,7 +213,6 @@ static struct rrpc_block *rrpc_get_blk(struct rrpc *rrpc, struct rrpc_lun *rlun,
static void rrpc_put_blk(struct rrpc *rrpc, struct rrpc_block *rblk)
{
struct rrpc_lun *rlun = rblk->rlun;
- struct nvm_lun *lun = rlun->parent;
spin_lock(&rlun->lock);
if (rblk->state & NVM_BLK_ST_TGT) {
@@ -229,8 +225,8 @@ static void rrpc_put_blk(struct rrpc *rrpc, struct rrpc_block *rblk)
} else {
WARN_ON_ONCE(1);
pr_err("rrpc: erroneous type (ch:%d,lun:%d,blk%d-> %u)\n",
- lun->chnl_id, lun->lun_id,
- rblk->id, rblk->state);
+ rlun->bppa.g.ch, rlun->bppa.g.lun,
+ rblk->id, rblk->state);
list_move_tail(&rblk->list, &rlun->bb_list);
}
spin_unlock(&rlun->lock);
@@ -336,7 +332,7 @@ static int rrpc_move_valid_pages(struct rrpc *rrpc, struct rrpc_block *rblk)
try:
spin_lock(&rrpc->rev_lock);
/* Get logical address from physical to logical table */
- rev = &rrpc->rev_trans_map[phys_addr - rrpc->poffset];
+ rev = &rrpc->rev_trans_map[phys_addr];
/* already updated by previous regular write */
if (rev->addr == ADDR_EMPTY) {
spin_unlock(&rrpc->rev_lock);
@@ -423,18 +419,18 @@ static void rrpc_block_gc(struct work_struct *work)
mempool_free(gcb, rrpc->gcb_pool);
pr_debug("nvm: block 'ch:%d,lun:%d,blk:%d' being reclaimed\n",
- rlun->parent->chnl_id, rlun->parent->lun_id,
+ 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->parent->chnl_id;
- ppa.g.lun = rlun->parent->lun_id;
+ ppa.g.ch = rlun->bppa.g.ch;
+ ppa.g.lun = rlun->bppa.g.lun;
ppa.g.blk = rblk->id;
- if (nvm_erase_blk(dev->parent, &ppa, 0))
+ if (nvm_erase_blk(dev, &ppa, 0))
goto put_back;
rrpc_put_blk(rrpc, rblk);
@@ -506,8 +502,7 @@ static void rrpc_lun_gc(struct work_struct *work)
WARN_ON(!block_is_full(rrpc, rblk));
pr_debug("rrpc: selected block 'ch:%d,lun:%d,blk:%d' for GC\n",
- rlun->parent->chnl_id,
- rlun->parent->lun_id,
+ rlun->bppa.g.ch, rlun->bppa.g.lun,
rblk->id);
gcb->rrpc = rrpc;
@@ -537,8 +532,7 @@ static void rrpc_gc_queue(struct work_struct *work)
mempool_free(gcb, rrpc->gcb_pool);
pr_debug("nvm: block 'ch:%d,lun:%d,blk:%d' full, allow GC (sched)\n",
- rlun->parent->chnl_id,
- rlun->parent->lun_id,
+ rlun->bppa.g.ch, rlun->bppa.g.lun,
rblk->id);
}
@@ -586,7 +580,7 @@ static struct rrpc_addr *rrpc_update_map(struct rrpc *rrpc, sector_t laddr,
gp->addr = paddr;
gp->rblk = rblk;
- rev = &rrpc->rev_trans_map[gp->addr - rrpc->poffset];
+ rev = &rrpc->rev_trans_map[gp->addr];
rev->addr = laddr;
spin_unlock(&rrpc->rev_lock);
@@ -601,7 +595,7 @@ static u64 rrpc_alloc_addr(struct rrpc *rrpc, struct rrpc_block *rblk)
if (block_is_full(rrpc, rblk))
goto out;
- addr = block_to_addr(rrpc, rblk) + rblk->next_page;
+ addr = rblk->next_page;
rblk->next_page++;
out:
@@ -615,18 +609,22 @@ static u64 rrpc_alloc_addr(struct rrpc *rrpc, struct rrpc_block *rblk)
* Returns rrpc_addr with the physical address and block. Returns NULL if no
* blocks in the next rlun are available.
*/
-static struct rrpc_addr *rrpc_map_page(struct rrpc *rrpc, sector_t laddr,
+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 NULL;
+ return ppa;
/*
* page allocation steps:
@@ -683,10 +681,15 @@ static struct rrpc_addr *rrpc_map_page(struct rrpc *rrpc, sector_t laddr,
}
pr_err("rrpc: failed to allocate new block\n");
- return NULL;
+ return ppa;
done:
spin_unlock(&rlun->lock);
- return rrpc_update_map(rrpc, laddr, rblk, paddr);
+ 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)
@@ -712,8 +715,8 @@ static struct rrpc_lun *rrpc_ppa_to_lun(struct rrpc *rrpc, struct ppa_addr p)
int i;
for (i = 0; i < rrpc->nr_luns; i++) {
- if (rrpc->luns[i].parent->chnl_id == p.g.ch &&
- rrpc->luns[i].parent->lun_id == p.g.lun) {
+ if (rrpc->luns[i].bppa.g.ch == p.g.ch &&
+ rrpc->luns[i].bppa.g.lun == p.g.lun) {
rlun = &rrpc->luns[i];
break;
}
@@ -823,7 +826,7 @@ static int rrpc_read_ppalist_rq(struct rrpc *rrpc, struct bio *bio,
gp = &rrpc->trans_map[laddr + i];
if (gp->rblk) {
- rqd->ppa_list[i] = rrpc_ppa_to_gaddr(dev, gp->addr);
+ rqd->ppa_list[i] = rrpc_ppa_to_gaddr(dev, gp);
} else {
BUG_ON(is_gc);
rrpc_unlock_laddr(rrpc, r);
@@ -852,7 +855,7 @@ static int rrpc_read_rq(struct rrpc *rrpc, struct bio *bio, struct nvm_rq *rqd,
gp = &rrpc->trans_map[laddr];
if (gp->rblk) {
- rqd->ppa_addr = rrpc_ppa_to_gaddr(rrpc->dev, gp->addr);
+ rqd->ppa_addr = rrpc_ppa_to_gaddr(rrpc->dev, gp);
} else {
BUG_ON(is_gc);
rrpc_unlock_rq(rrpc, rqd);
@@ -869,7 +872,7 @@ static int rrpc_write_ppalist_rq(struct rrpc *rrpc, struct bio *bio,
{
struct nvm_tgt_dev *dev = rrpc->dev;
struct rrpc_inflight_rq *r = rrpc_get_inflight_rq(rqd);
- struct rrpc_addr *p;
+ struct ppa_addr p;
sector_t laddr = rrpc_get_laddr(bio);
int is_gc = flags & NVM_IOTYPE_GC;
int i;
@@ -882,7 +885,7 @@ static int rrpc_write_ppalist_rq(struct rrpc *rrpc, struct bio *bio,
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) {
+ if (p.ppa == ADDR_EMPTY) {
BUG_ON(is_gc);
rrpc_unlock_laddr(rrpc, r);
nvm_dev_dma_free(dev->parent, rqd->ppa_list,
@@ -891,7 +894,7 @@ static int rrpc_write_ppalist_rq(struct rrpc *rrpc, struct bio *bio,
return NVM_IO_REQUEUE;
}
- rqd->ppa_list[i] = rrpc_ppa_to_gaddr(dev, p->addr);
+ rqd->ppa_list[i] = p;
}
rqd->opcode = NVM_OP_HBWRITE;
@@ -902,7 +905,7 @@ static int rrpc_write_ppalist_rq(struct rrpc *rrpc, struct bio *bio,
static int rrpc_write_rq(struct rrpc *rrpc, struct bio *bio,
struct nvm_rq *rqd, unsigned long flags)
{
- struct rrpc_addr *p;
+ struct ppa_addr p;
int is_gc = flags & NVM_IOTYPE_GC;
sector_t laddr = rrpc_get_laddr(bio);
@@ -910,14 +913,14 @@ static int rrpc_write_rq(struct rrpc *rrpc, struct bio *bio,
return NVM_IO_REQUEUE;
p = rrpc_map_page(rrpc, laddr, is_gc);
- if (!p) {
+ 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 = rrpc_ppa_to_gaddr(rrpc->dev, p->addr);
+ rqd->ppa_addr = p;
rqd->opcode = NVM_OP_HBWRITE;
return NVM_IO_OK;
@@ -973,7 +976,7 @@ static int rrpc_submit_io(struct rrpc *rrpc, struct bio *bio,
rqd->nr_ppas = nr_pages;
rrq->flags = flags;
- err = nvm_submit_io(dev->parent, rqd);
+ err = nvm_submit_io(dev, rqd);
if (err) {
pr_err("rrpc: I/O submission failed: %d\n", err);
bio_put(bio);
@@ -1113,14 +1116,12 @@ static int rrpc_l2p_update(u64 slba, u32 nlb, __le64 *entries, void *private)
div_u64_rem(pba, rrpc->nr_sects, &mod);
- addr[i].addr = pba;
- raddr[mod].addr = slba + i;
-
- gaddr = rrpc_ppa_to_gaddr(dev, pba);
+ gaddr = rrpc_recov_addr(dev->parent, pba);
rlun = rrpc_ppa_to_lun(rrpc, gaddr);
if (!rlun) {
- pr_err("rrpc: unmapped lun %d l2p corruption\n",
- rlun->parent->id);
+ pr_err("rrpc: unmapped 'ch:%d,lun:%d' l2p corruption\n",
+ rlun->bppa.g.ch,
+ rlun->bppa.g.lun);
return -EINVAL;
}
@@ -1134,6 +1135,10 @@ static int rrpc_l2p_update(u64 slba, u32 nlb, __le64 *entries, void *private)
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;
@@ -1230,7 +1235,6 @@ static void rrpc_core_free(struct rrpc *rrpc)
static void rrpc_luns_free(struct rrpc *rrpc)
{
- struct nvm_lun *lun;
struct rrpc_lun *rlun;
int i;
@@ -1239,9 +1243,6 @@ static void rrpc_luns_free(struct rrpc *rrpc)
for (i = 0; i < rrpc->nr_luns; i++) {
rlun = &rrpc->luns[i];
- lun = rlun->parent;
- if (!lun)
- break;
vfree(rlun->blocks);
}
@@ -1264,8 +1265,8 @@ static int rrpc_bb_discovery(struct nvm_tgt_dev *dev, struct rrpc_lun *rlun)
return -ENOMEM;
ppa.ppa = 0;
- ppa.g.ch = rlun->parent->chnl_id;
- ppa.g.lun = rlun->parent->lun_id;
+ ppa.g.ch = rlun->bppa.g.ch;
+ ppa.g.lun = rlun->bppa.g.lun;
ret = nvm_get_bb_tbl(dev->parent, ppa, blks);
if (ret) {
@@ -1294,11 +1295,17 @@ static int rrpc_bb_discovery(struct nvm_tgt_dev *dev, struct rrpc_lun *rlun)
return ret;
}
-static int rrpc_luns_init(struct rrpc *rrpc, struct list_head *lun_list)
+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 nvm_lun *lun;
struct rrpc_lun *rlun;
int i, j, ret = -EINVAL;
@@ -1314,12 +1321,11 @@ static int rrpc_luns_init(struct rrpc *rrpc, struct list_head *lun_list)
if (!rrpc->luns)
return -ENOMEM;
- i = 0;
-
/* 1:1 mapping */
- list_for_each_entry(lun, lun_list, list) {
- rlun = &rrpc->luns[i++];
- rlun->parent = lun;
+ 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) {
@@ -1357,8 +1363,6 @@ static int rrpc_luns_init(struct rrpc *rrpc, struct list_head *lun_list)
spin_lock_init(&rlun->lock);
}
- WARN_ON(i != rrpc->nr_luns);
-
return 0;
err:
return ret;
@@ -1512,14 +1516,12 @@ static int rrpc_luns_configure(struct rrpc *rrpc)
static struct nvm_tgt_type tt_rrpc;
-static void *rrpc_init(struct nvm_tgt_dev *dev, struct gendisk *tdisk,
- struct list_head *lun_list)
+static void *rrpc_init(struct nvm_tgt_dev *dev, struct gendisk *tdisk)
{
struct request_queue *bqueue = dev->q;
struct request_queue *tqueue = tdisk->queue;
struct nvm_geo *geo = &dev->geo;
struct rrpc *rrpc;
- int lun_begin = (list_first_entry(lun_list, struct nvm_lun, list))->id;
sector_t soffset;
int ret;
@@ -1554,14 +1556,12 @@ static void *rrpc_init(struct nvm_tgt_dev *dev, struct gendisk *tdisk,
}
rrpc->soffset = soffset;
- ret = rrpc_luns_init(rrpc, lun_list);
+ ret = rrpc_luns_init(rrpc, dev->luns);
if (ret) {
pr_err("nvm: rrpc: could not initialize luns\n");
goto err;
}
- rrpc->poffset = geo->sec_per_lun * lun_begin;
-
ret = rrpc_core_init(rrpc);
if (ret) {
pr_err("nvm: rrpc: could not initialize core\n");
diff --git a/drivers/lightnvm/rrpc.h b/drivers/lightnvm/rrpc.h
index c55cec9..bc8adba 100644
--- a/drivers/lightnvm/rrpc.h
+++ b/drivers/lightnvm/rrpc.h
@@ -74,7 +74,9 @@ struct rrpc_block {
struct rrpc_lun {
struct rrpc *rrpc;
- struct nvm_lun *parent;
+
+ int id;
+ struct ppa_addr bppa;
struct rrpc_block *cur, *gc_cur;
struct rrpc_block *blocks; /* Reference to block allocation */
@@ -107,7 +109,6 @@ struct rrpc {
struct gendisk *disk;
sector_t soffset; /* logical sector offset */
- u64 poffset; /* physical page offset */
int nr_luns;
struct rrpc_lun *luns;
@@ -164,14 +165,37 @@ 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_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;
- struct nvm_lun *lun = rlun->parent;
- return (lun->id * geo->sec_per_lun) + (rblk->id * geo->sec_per_blk);
+ return (rlun->id * geo->sec_per_lun) + (rblk->id * geo->sec_per_blk);
}
static inline sector_t rrpc_get_laddr(struct bio *bio)
diff --git a/drivers/nvme/host/lightnvm.c b/drivers/nvme/host/lightnvm.c
index 1cdc812..588d4a3 100644
--- a/drivers/nvme/host/lightnvm.c
+++ b/drivers/nvme/host/lightnvm.c
@@ -371,6 +371,9 @@ static int nvme_nvm_get_l2p_tbl(struct nvm_dev *nvmdev, u64 slba, u32 nlb,
return -EINVAL;
}
+ /* Transform physical address to target address space */
+ nvmdev->mt->part_to_tgt(nvmdev, entries, cmd_nlb);
+
if (update_l2p(cmd_slba, cmd_nlb, entries, priv)) {
ret = -EINTR;
goto out;
diff --git a/include/linux/lightnvm.h b/include/linux/lightnvm.h
index cc210cc..2222853 100644
--- a/include/linux/lightnvm.h
+++ b/include/linux/lightnvm.h
@@ -47,6 +47,7 @@ struct ppa_addr {
struct nvm_rq;
struct nvm_id;
struct nvm_dev;
+struct nvm_tgt_dev;
typedef int (nvm_l2p_update_fn)(u64, u32, __le64 *, void *);
typedef int (nvm_id_fn)(struct nvm_dev *, struct nvm_id *);
@@ -210,7 +211,6 @@ struct nvm_id {
struct nvm_target {
struct list_head list;
- struct list_head lun_list;
struct nvm_tgt_dev *dev;
struct nvm_tgt_type *type;
struct gendisk *disk;
@@ -231,7 +231,7 @@ typedef void (nvm_end_io_fn)(struct nvm_rq *);
struct nvm_rq {
struct nvm_tgt_instance *ins;
- struct nvm_dev *dev;
+ struct nvm_tgt_dev *dev;
struct bio *bio;
@@ -266,15 +266,6 @@ static inline void *nvm_rq_to_pdu(struct nvm_rq *rqdata)
return rqdata + 1;
}
-struct nvm_lun {
- int id;
-
- int lun_id;
- int chnl_id;
-
- struct list_head list;
-};
-
enum {
NVM_BLK_ST_FREE = 0x1, /* Free block */
NVM_BLK_ST_TGT = 0x2, /* Block in use by target */
@@ -321,6 +312,9 @@ struct nvm_tgt_dev {
/* Device information */
struct nvm_geo geo;
+ /* Base ppas for target LUNs */
+ struct ppa_addr *luns;
+
sector_t total_secs;
struct nvm_id identity;
@@ -330,6 +324,7 @@ struct nvm_tgt_dev {
struct nvm_dev_ops *ops;
void *parent;
+ void *map;
};
struct nvm_dev {
@@ -363,16 +358,18 @@ struct nvm_dev {
char name[DISK_NAME_LEN];
void *private_data;
+ void *rmap;
+
struct mutex mlock;
spinlock_t lock;
};
static inline struct ppa_addr linear_to_generic_addr(struct nvm_geo *geo,
- struct ppa_addr r)
+ u64 pba)
{
struct ppa_addr l;
int secs, pgs, blks, luns;
- sector_t ppa = r.ppa;
+ sector_t ppa = pba;
l.ppa = 0;
@@ -465,8 +462,7 @@ static inline int ppa_to_slc(struct nvm_dev *dev, int slc_pg)
typedef blk_qc_t (nvm_tgt_make_rq_fn)(struct request_queue *, struct bio *);
typedef sector_t (nvm_tgt_capacity_fn)(void *);
-typedef void *(nvm_tgt_init_fn)(struct nvm_tgt_dev *, struct gendisk *,
- struct list_head *lun_list);
+typedef void *(nvm_tgt_init_fn)(struct nvm_tgt_dev *, struct gendisk *);
typedef void (nvm_tgt_exit_fn)(void *);
struct nvm_tgt_type {
@@ -499,10 +495,11 @@ typedef void (nvmm_unregister_fn)(struct nvm_dev *);
typedef int (nvmm_create_tgt_fn)(struct nvm_dev *, struct nvm_ioctl_create *);
typedef int (nvmm_remove_tgt_fn)(struct nvm_dev *, struct nvm_ioctl_remove *);
-typedef int (nvmm_submit_io_fn)(struct nvm_dev *, struct nvm_rq *);
-typedef int (nvmm_erase_blk_fn)(struct nvm_dev *, struct ppa_addr *, int);
+typedef int (nvmm_submit_io_fn)(struct nvm_tgt_dev *, struct nvm_rq *);
+typedef int (nvmm_erase_blk_fn)(struct nvm_tgt_dev *, struct ppa_addr *, int);
typedef int (nvmm_get_area_fn)(struct nvm_dev *, sector_t *, sector_t);
typedef void (nvmm_put_area_fn)(struct nvm_dev *, sector_t);
+typedef void (nvmm_part_to_tgt_fn)(struct nvm_dev *, sector_t*, int);
struct nvmm_type {
const char *name;
@@ -520,6 +517,8 @@ struct nvmm_type {
nvmm_get_area_fn *get_area;
nvmm_put_area_fn *put_area;
+ nvmm_part_to_tgt_fn *part_to_tgt;
+
struct list_head list;
};
@@ -533,14 +532,18 @@ extern void nvm_unregister(struct nvm_dev *);
extern int nvm_set_bb_tbl(struct nvm_dev *dev, struct ppa_addr *ppas,
int nr_ppas, int type);
-extern int nvm_submit_io(struct nvm_dev *, struct nvm_rq *);
+extern int nvm_submit_io(struct nvm_tgt_dev *, struct nvm_rq *);
extern void nvm_generic_to_addr_mode(struct nvm_dev *, struct nvm_rq *);
extern void nvm_addr_to_generic_mode(struct nvm_dev *, struct nvm_rq *);
extern int nvm_set_rqd_ppalist(struct nvm_dev *, struct nvm_rq *,
const struct ppa_addr *, int, int);
extern void nvm_free_rqd_ppalist(struct nvm_dev *, struct nvm_rq *);
extern int nvm_erase_ppa(struct nvm_dev *, struct ppa_addr *, int, int);
-extern int nvm_erase_blk(struct nvm_dev *, struct ppa_addr *, int);
+extern int nvm_erase_blk(struct nvm_tgt_dev *, struct ppa_addr *, int);
+extern int nvm_get_l2p_tbl(struct nvm_dev *, u64, u32, nvm_l2p_update_fn *,
+ void *);
+extern int nvm_get_area(struct nvm_dev *, sector_t *, sector_t);
+extern void nvm_put_area(struct nvm_dev *, sector_t);
extern void nvm_end_io(struct nvm_rq *, int);
extern int nvm_submit_ppa(struct nvm_dev *, struct ppa_addr *, int, int, int,
void *, int);
--
2.7.4
[toc] | [prev] | [next] | [standalone]
| From | "Javier González" <jg@lightnvm.io> |
|---|---|
| Date | 2016-11-18 15:50 +0100 |
| Subject | [PATCH 16/20] lightnvm: eliminate nvm_block abstraction on mm |
| Message-ID | <sESEj-8qT-51@gated-at.bofh.it> |
| In reply to | #1525412 |
In order to naturally support multi-target instances on an Open-Channel
SSD, targets should own the LUNs they get blocks from and manage
provisioning internally. This is done in several steps.
A part of this transformation is that targets manage their blocks
internally. This patch eliminates the nvm_block abstraction and moves
block management to the target logic. The rrpc target is transformed.
Signed-off-by: Javier González <javier@cnexlabs.com>
---
drivers/lightnvm/core.c | 10 +-
drivers/lightnvm/gennvm.c | 205 +---------------------------------------
drivers/lightnvm/rrpc.c | 235 +++++++++++++++++++++++++++++++++-------------
drivers/lightnvm/rrpc.h | 23 +++--
include/linux/lightnvm.h | 48 +---------
5 files changed, 192 insertions(+), 329 deletions(-)
diff --git a/drivers/lightnvm/core.c b/drivers/lightnvm/core.c
index 493ce03..691b16f 100644
--- a/drivers/lightnvm/core.c
+++ b/drivers/lightnvm/core.c
@@ -176,12 +176,6 @@ static struct nvm_dev *nvm_find_nvm_dev(const char *name)
return NULL;
}
-void nvm_mark_blk(struct nvm_dev *dev, struct ppa_addr ppa, int type)
-{
- return dev->mt->mark_blk(dev, ppa, type);
-}
-EXPORT_SYMBOL(nvm_mark_blk);
-
int nvm_set_bb_tbl(struct nvm_dev *dev, struct ppa_addr *ppas, int nr_ppas,
int type)
{
@@ -215,9 +209,9 @@ int nvm_submit_io(struct nvm_dev *dev, struct nvm_rq *rqd)
}
EXPORT_SYMBOL(nvm_submit_io);
-int nvm_erase_blk(struct nvm_dev *dev, struct nvm_block *blk, int flags)
+int nvm_erase_blk(struct nvm_dev *dev, struct ppa_addr *p, int flags)
{
- return dev->mt->erase_blk(dev, blk, flags);
+ return dev->mt->erase_blk(dev, p, flags);
}
EXPORT_SYMBOL(nvm_erase_blk);
diff --git a/drivers/lightnvm/gennvm.c b/drivers/lightnvm/gennvm.c
index dd9afd7..19c3924 100644
--- a/drivers/lightnvm/gennvm.c
+++ b/drivers/lightnvm/gennvm.c
@@ -306,19 +306,6 @@ static void gen_put_area(struct nvm_dev *dev, sector_t begin)
spin_unlock(&dev->lock);
}
-static void gen_blocks_free(struct nvm_dev *dev)
-{
- struct gen_dev *gn = dev->mp;
- struct nvm_lun *lun;
- int i;
-
- gen_for_each_lun(gn, lun, i) {
- if (!lun->blocks)
- break;
- vfree(lun->blocks);
- }
-}
-
static void gen_luns_free(struct nvm_dev *dev)
{
struct gen_dev *gn = dev->mp;
@@ -337,167 +324,17 @@ static int gen_luns_init(struct nvm_dev *dev, struct gen_dev *gn)
return -ENOMEM;
gen_for_each_lun(gn, lun, i) {
- INIT_LIST_HEAD(&lun->free_list);
- INIT_LIST_HEAD(&lun->used_list);
- INIT_LIST_HEAD(&lun->bb_list);
INIT_LIST_HEAD(&lun->list);
- spin_lock_init(&lun->lock);
-
lun->id = i;
lun->lun_id = i % geo->luns_per_chnl;
lun->chnl_id = i / geo->luns_per_chnl;
- lun->nr_free_blocks = geo->blks_per_lun;
}
return 0;
}
-static int gen_block_bb(struct gen_dev *gn, struct ppa_addr ppa,
- u8 *blks, int nr_blks)
-{
- struct nvm_dev *dev = gn->dev;
- struct nvm_lun *lun;
- struct nvm_block *blk;
- int i;
-
- nr_blks = nvm_bb_tbl_fold(dev, blks, nr_blks);
- if (nr_blks < 0)
- return nr_blks;
-
- lun = &gn->luns[(dev->geo.luns_per_chnl * ppa.g.ch) + ppa.g.lun];
-
- for (i = 0; i < nr_blks; i++) {
- if (blks[i] == NVM_BLK_T_FREE)
- continue;
-
- blk = &lun->blocks[i];
- list_move_tail(&blk->list, &lun->bb_list);
- blk->state = NVM_BLK_ST_BAD;
- lun->nr_free_blocks--;
- }
-
- return 0;
-}
-
-static int gen_block_map(u64 slba, u32 nlb, __le64 *entries, void *private)
-{
- struct nvm_dev *dev = private;
- struct nvm_geo *geo = &dev->geo;
- struct gen_dev *gn = dev->mp;
- u64 elba = slba + nlb;
- struct nvm_lun *lun;
- struct nvm_block *blk;
- u64 i;
- int lun_id;
-
- if (unlikely(elba > dev->total_secs)) {
- pr_err("gen: L2P data from device is out of bounds!\n");
- return -EINVAL;
- }
-
- for (i = 0; i < nlb; i++) {
- u64 pba = le64_to_cpu(entries[i]);
-
- if (unlikely(pba >= dev->total_secs && pba != U64_MAX)) {
- pr_err("gen: L2P data entry is out of bounds!\n");
- return -EINVAL;
- }
-
- /* Address zero is a special one. The first page on a disk is
- * protected. It often holds internal device boot
- * information.
- */
- if (!pba)
- continue;
-
- /* resolve block from physical address */
- lun_id = div_u64(pba, geo->sec_per_lun);
- lun = &gn->luns[lun_id];
-
- /* Calculate block offset into lun */
- pba = pba - (geo->sec_per_lun * lun_id);
- blk = &lun->blocks[div_u64(pba, geo->sec_per_blk)];
-
- if (!blk->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(&blk->list, &lun->used_list);
- blk->state = NVM_BLK_ST_TGT;
- lun->nr_free_blocks--;
- }
- }
-
- return 0;
-}
-
-static int gen_blocks_init(struct nvm_dev *dev, struct gen_dev *gn)
-{
- struct nvm_geo *geo = &dev->geo;
- struct nvm_lun *lun;
- struct nvm_block *block;
- sector_t lun_iter, blk_iter, cur_block_id = 0;
- int ret, nr_blks;
- u8 *blks;
-
- nr_blks = geo->blks_per_lun * geo->plane_mode;
- blks = kmalloc(nr_blks, GFP_KERNEL);
- if (!blks)
- return -ENOMEM;
-
- gen_for_each_lun(gn, lun, lun_iter) {
- lun->blocks = vzalloc(sizeof(struct nvm_block) *
- geo->blks_per_lun);
- if (!lun->blocks) {
- kfree(blks);
- return -ENOMEM;
- }
-
- for (blk_iter = 0; blk_iter < geo->blks_per_lun; blk_iter++) {
- block = &lun->blocks[blk_iter];
-
- INIT_LIST_HEAD(&block->list);
-
- block->lun = lun;
- block->id = cur_block_id++;
-
- list_add_tail(&block->list, &lun->free_list);
- }
-
- if (dev->ops->get_bb_tbl) {
- struct ppa_addr ppa;
-
- ppa.ppa = 0;
- ppa.g.ch = lun->chnl_id;
- ppa.g.lun = lun->lun_id;
-
- ret = nvm_get_bb_tbl(dev, ppa, blks);
- if (ret)
- pr_err("gen: could not get BB table\n");
-
- ret = gen_block_bb(gn, ppa, blks, nr_blks);
- if (ret)
- pr_err("gen: BB table map failed\n");
- }
- }
-
- if ((dev->identity.dom & NVM_RSP_L2P) && dev->ops->get_l2p_tbl) {
- ret = dev->ops->get_l2p_tbl(dev, 0, dev->total_secs,
- gen_block_map, dev);
- if (ret) {
- pr_err("gen: could not read L2P table.\n");
- pr_warn("gen: default block initialization");
- }
- }
-
- kfree(blks);
- return 0;
-}
-
static void gen_free(struct nvm_dev *dev)
{
- gen_blocks_free(dev);
gen_luns_free(dev);
kfree(dev->mp);
dev->mp = NULL;
@@ -528,12 +365,6 @@ static int gen_register(struct nvm_dev *dev)
goto err;
}
- ret = gen_blocks_init(dev, gn);
- if (ret) {
- pr_err("gen: could not initialize blocks\n");
- goto err;
- }
-
return 1;
err:
gen_free(dev);
@@ -558,34 +389,6 @@ static void gen_unregister(struct nvm_dev *dev)
module_put(THIS_MODULE);
}
-static void gen_mark_blk(struct nvm_dev *dev, struct ppa_addr ppa, int type)
-{
- struct nvm_geo *geo = &dev->geo;
- struct gen_dev *gn = dev->mp;
- struct nvm_lun *lun;
- struct nvm_block *blk;
-
- pr_debug("gen: ppa (ch: %u lun: %u blk: %u pg: %u) -> %u\n",
- ppa.g.ch, ppa.g.lun, ppa.g.blk, ppa.g.pg, type);
-
- if (unlikely(ppa.g.ch > geo->nr_chnls ||
- ppa.g.lun > geo->luns_per_chnl ||
- ppa.g.blk > geo->blks_per_lun)) {
- WARN_ON_ONCE(1);
- pr_err("gen: ppa broken (ch: %u > %u lun: %u > %u blk: %u > %u",
- ppa.g.ch, geo->nr_chnls,
- ppa.g.lun, geo->luns_per_chnl,
- ppa.g.blk, geo->blks_per_lun);
- return;
- }
-
- lun = &gn->luns[(geo->luns_per_chnl * ppa.g.ch) + ppa.g.lun];
- blk = &lun->blocks[ppa.g.blk];
-
- /* will be moved to bb list on put_blk from target */
- blk->state = type;
-}
-
static void gen_end_io(struct nvm_rq *rqd)
{
struct nvm_tgt_instance *ins = rqd->ins;
@@ -606,11 +409,9 @@ static int gen_submit_io(struct nvm_dev *dev, struct nvm_rq *rqd)
return dev->ops->submit_io(dev, rqd);
}
-static int gen_erase_blk(struct nvm_dev *dev, struct nvm_block *blk, int flags)
+static int gen_erase_blk(struct nvm_dev *dev, struct ppa_addr *p, int flags)
{
- struct ppa_addr addr = block_to_ppa(dev, blk);
-
- return nvm_erase_ppa(dev, &addr, 1, flags);
+ return nvm_erase_ppa(dev, p, 1, flags);
}
static struct nvmm_type gen = {
@@ -626,8 +427,6 @@ static struct nvmm_type gen = {
.submit_io = gen_submit_io,
.erase_blk = gen_erase_blk,
- .mark_blk = gen_mark_blk,
-
.get_area = gen_get_area,
.put_area = gen_put_area,
diff --git a/drivers/lightnvm/rrpc.c b/drivers/lightnvm/rrpc.c
index 165b9d3..fad117b 100644
--- a/drivers/lightnvm/rrpc.c
+++ b/drivers/lightnvm/rrpc.c
@@ -126,19 +126,21 @@ static int block_is_full(struct rrpc *rrpc, struct rrpc_block *rblk)
static u64 block_to_rel_addr(struct rrpc *rrpc, struct rrpc_block *rblk)
{
struct nvm_tgt_dev *dev = rrpc->dev;
- struct nvm_block *blk = rblk->parent;
- int lun_blk = blk->id % (dev->geo.blks_per_lun * rrpc->nr_luns);
+ struct rrpc_lun *rlun = rblk->rlun;
+ struct nvm_lun *lun = rlun->parent;
- return lun_blk * dev->geo.sec_per_blk;
+ return lun->id * dev->geo.sec_per_blk;
}
/* Calculate global addr for the given block */
static u64 block_to_addr(struct rrpc *rrpc, struct rrpc_block *rblk)
{
struct nvm_tgt_dev *dev = rrpc->dev;
- struct nvm_block *blk = rblk->parent;
+ struct nvm_geo *geo = &dev->geo;
+ struct rrpc_lun *rlun = rblk->rlun;
+ struct nvm_lun *lun = rlun->parent;
- return blk->id * dev->geo.sec_per_blk;
+ return lun->id * geo->sec_per_lun + rblk->id * geo->sec_per_blk;
}
static struct ppa_addr rrpc_ppa_to_gaddr(struct nvm_tgt_dev *dev, u64 addr)
@@ -163,51 +165,46 @@ static void rrpc_set_lun_cur(struct rrpc_lun *rlun, struct rrpc_block *new_rblk,
*cur_rblk = new_rblk;
}
-static struct nvm_block *__rrpc_get_blk(struct rrpc *rrpc,
+static struct rrpc_block *__rrpc_get_blk(struct rrpc *rrpc,
struct rrpc_lun *rlun)
{
- struct nvm_lun *lun = rlun->parent;
- struct nvm_block *blk = NULL;
+ struct rrpc_block *rblk = NULL;
- if (list_empty(&lun->free_list))
+ if (list_empty(&rlun->free_list))
goto out;
- blk = list_first_entry(&lun->free_list, struct nvm_block, list);
+ rblk = list_first_entry(&rlun->free_list, struct rrpc_block, list);
- list_move_tail(&blk->list, &lun->used_list);
- blk->state = NVM_BLK_ST_TGT;
- lun->nr_free_blocks--;
+ list_move_tail(&rblk->list, &rlun->used_list);
+ rblk->state = NVM_BLK_ST_TGT;
+ rlun->nr_free_blocks--;
out:
- return blk;
+ 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 nvm_lun *lun = rlun->parent;
- struct nvm_block *blk;
struct rrpc_block *rblk;
int is_gc = flags & NVM_IOTYPE_GC;
spin_lock(&rlun->lock);
- if (!is_gc && lun->nr_free_blocks < rlun->reserved_blocks) {
+ 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;
}
- blk = __rrpc_get_blk(rrpc, rlun);
- if (!blk) {
+ 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);
- rblk = rrpc_get_rblk(rlun, blk->id);
- blk->priv = rblk;
bitmap_zero(rblk->invalid_pages, dev->geo.sec_per_blk);
rblk->next_page = 0;
rblk->nr_invalid_pages = 0;
@@ -218,23 +215,23 @@ static struct rrpc_block *rrpc_get_blk(struct rrpc *rrpc, struct rrpc_lun *rlun,
static void rrpc_put_blk(struct rrpc *rrpc, struct rrpc_block *rblk)
{
- struct nvm_block *blk = rblk->parent;
struct rrpc_lun *rlun = rblk->rlun;
struct nvm_lun *lun = rlun->parent;
spin_lock(&rlun->lock);
- if (blk->state & NVM_BLK_ST_TGT) {
- list_move_tail(&blk->list, &lun->free_list);
- lun->nr_free_blocks++;
- blk->state = NVM_BLK_ST_FREE;
- } else if (blk->state & NVM_BLK_ST_BAD) {
- list_move_tail(&blk->list, &lun->bb_list);
- blk->state = NVM_BLK_ST_BAD;
+ 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 block type (%lu -> %u)\n",
- blk->id, blk->state);
- list_move_tail(&blk->list, &lun->bb_list);
+ pr_err("rrpc: erroneous type (ch:%d,lun:%d,blk%d-> %u)\n",
+ lun->chnl_id, lun->lun_id,
+ rblk->id, rblk->state);
+ list_move_tail(&rblk->list, &rlun->bb_list);
}
spin_unlock(&rlun->lock);
}
@@ -334,7 +331,7 @@ static int rrpc_move_valid_pages(struct rrpc *rrpc, struct rrpc_block *rblk)
nr_sec_per_blk)) < nr_sec_per_blk) {
/* Lock laddr */
- phys_addr = rblk->parent->id * nr_sec_per_blk + slot;
+ phys_addr = rrpc_blk_to_ppa(rrpc, rblk) + slot;
try:
spin_lock(&rrpc->rev_lock);
@@ -422,14 +419,22 @@ static void rrpc_block_gc(struct work_struct *work)
struct rrpc_block *rblk = gcb->rblk;
struct rrpc_lun *rlun = rblk->rlun;
struct nvm_tgt_dev *dev = rrpc->dev;
+ struct ppa_addr ppa;
mempool_free(gcb, rrpc->gcb_pool);
- pr_debug("nvm: block '%lu' being reclaimed\n", rblk->parent->id);
+ pr_debug("nvm: block 'ch:%d,lun:%d,blk:%d' being reclaimed\n",
+ rlun->parent->chnl_id, rlun->parent->lun_id,
+ rblk->id);
if (rrpc_move_valid_pages(rrpc, rblk))
goto put_back;
- if (nvm_erase_blk(dev->parent, rblk->parent, 0))
+ ppa.ppa = 0;
+ ppa.g.ch = rlun->parent->chnl_id;
+ ppa.g.lun = rlun->parent->lun_id;
+ ppa.g.blk = rblk->id;
+
+ if (nvm_erase_blk(dev->parent, &ppa, 0))
goto put_back;
rrpc_put_blk(rrpc, rblk);
@@ -445,7 +450,7 @@ static void rrpc_block_gc(struct work_struct *work)
/* the block with highest number of invalid pages, will be in the beginning
* of the list
*/
-static struct rrpc_block *rblock_max_invalid(struct rrpc_block *ra,
+static struct rrpc_block *rblk_max_invalid(struct rrpc_block *ra,
struct rrpc_block *rb)
{
if (ra->nr_invalid_pages == rb->nr_invalid_pages)
@@ -460,13 +465,13 @@ static struct rrpc_block *rblock_max_invalid(struct rrpc_block *ra,
static struct rrpc_block *block_prio_find_max(struct rrpc_lun *rlun)
{
struct list_head *prio_list = &rlun->prio_list;
- struct rrpc_block *rblock, *max;
+ 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(rblock, prio_list, prio)
- max = rblock_max_invalid(max, rblock);
+ list_for_each_entry(rblk, prio_list, prio)
+ max = rblk_max_invalid(max, rblk);
return max;
}
@@ -476,7 +481,6 @@ 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 nvm_lun *lun = rlun->parent;
struct rrpc_block_gc *gcb;
unsigned int nr_blocks_need;
@@ -486,26 +490,28 @@ static void rrpc_lun_gc(struct work_struct *work)
nr_blocks_need = rrpc->nr_luns;
spin_lock(&rlun->lock);
- while (nr_blocks_need > lun->nr_free_blocks &&
+ while (nr_blocks_need > rlun->nr_free_blocks &&
!list_empty(&rlun->prio_list)) {
- struct rrpc_block *rblock = block_prio_find_max(rlun);
- struct nvm_block *block = rblock->parent;
+ struct rrpc_block *rblk = block_prio_find_max(rlun);
- if (!rblock->nr_invalid_pages)
+ if (!rblk->nr_invalid_pages)
break;
gcb = mempool_alloc(rrpc->gcb_pool, GFP_ATOMIC);
if (!gcb)
break;
- list_del_init(&rblock->prio);
+ list_del_init(&rblk->prio);
- BUG_ON(!block_is_full(rrpc, rblock));
+ WARN_ON(!block_is_full(rrpc, rblk));
- pr_debug("rrpc: selected block '%lu' for GC\n", block->id);
+ pr_debug("rrpc: selected block 'ch:%d,lun:%d,blk:%d' for GC\n",
+ rlun->parent->chnl_id,
+ rlun->parent->lun_id,
+ rblk->id);
gcb->rrpc = rrpc;
- gcb->rblk = rblock;
+ gcb->rblk = rblk;
INIT_WORK(&gcb->ws_gc, rrpc_block_gc);
queue_work(rrpc->kgc_wq, &gcb->ws_gc);
@@ -530,8 +536,10 @@ static void rrpc_gc_queue(struct work_struct *work)
spin_unlock(&rlun->lock);
mempool_free(gcb, rrpc->gcb_pool);
- pr_debug("nvm: block '%lu' is full, allow GC (sched)\n",
- rblk->parent->id);
+ pr_debug("nvm: block 'ch:%d,lun:%d,blk:%d' full, allow GC (sched)\n",
+ rlun->parent->chnl_id,
+ rlun->parent->lun_id,
+ rblk->id);
}
static const struct block_device_operations rrpc_fops = {
@@ -555,8 +563,7 @@ static struct rrpc_lun *rrpc_get_lun_rr(struct rrpc *rrpc, int is_gc)
* estimate.
*/
rrpc_for_each_lun(rrpc, rlun, i) {
- if (rlun->parent->nr_free_blocks >
- max_free->parent->nr_free_blocks)
+ if (rlun->nr_free_blocks > max_free->nr_free_blocks)
max_free = rlun;
}
@@ -613,14 +620,12 @@ static struct rrpc_addr *rrpc_map_page(struct rrpc *rrpc, sector_t laddr,
{
struct rrpc_lun *rlun;
struct rrpc_block *rblk, **cur_rblk;
- struct nvm_lun *lun;
u64 paddr;
int gc_force = 0;
rlun = rrpc_get_lun_rr(rrpc, is_gc);
- lun = rlun->parent;
- if (!is_gc && lun->nr_free_blocks < rrpc->nr_luns * 4)
+ if (!is_gc && rlun->nr_free_blocks < rrpc->nr_luns * 4)
return NULL;
/*
@@ -701,22 +706,44 @@ static void rrpc_run_gc(struct rrpc *rrpc, struct rrpc_block *rblk)
queue_work(rrpc->kgc_wq, &gcb->ws_gc);
}
-static void __rrpc_mark_bad_block(struct nvm_tgt_dev *dev, struct ppa_addr *ppa)
+static struct rrpc_lun *rrpc_ppa_to_lun(struct rrpc *rrpc, struct ppa_addr p)
{
- nvm_mark_blk(dev->parent, *ppa, NVM_BLK_ST_BAD);
- nvm_set_bb_tbl(dev->parent, ppa, 1, NVM_BLK_T_GRWN_BAD);
+ struct rrpc_lun *rlun = NULL;
+ int i;
+
+ for (i = 0; i < rrpc->nr_luns; i++) {
+ if (rrpc->luns[i].parent->chnl_id == p.g.ch &&
+ rrpc->luns[i].parent->lun_id == 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_bb_tbl(dev->parent, &ppa, 1, NVM_BLK_T_GRWN_BAD);
}
static void rrpc_mark_bad_block(struct rrpc *rrpc, struct nvm_rq *rqd)
{
- struct nvm_tgt_dev *dev = rrpc->dev;
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(dev, &rqd->ppa_addr);
+ __rrpc_mark_bad_block(rrpc, rqd->ppa_addr);
ppa_set_empty(&prev_ppa);
bit = -1;
@@ -725,7 +752,7 @@ static void rrpc_mark_bad_block(struct rrpc *rrpc, struct nvm_rq *rqd)
if (ppa_cmp_blk(ppa, prev_ppa))
continue;
- __rrpc_mark_bad_block(dev, &ppa);
+ __rrpc_mark_bad_block(rrpc, ppa);
}
}
@@ -735,13 +762,11 @@ static void rrpc_end_io_write(struct rrpc *rrpc, struct rrpc_rq *rrqd,
struct nvm_tgt_dev *dev = rrpc->dev;
struct rrpc_addr *p;
struct rrpc_block *rblk;
- struct nvm_lun *lun;
int cmnt_size, i;
for (i = 0; i < npages; i++) {
p = &rrpc->trans_map[laddr + i];
rblk = p->rblk;
- lun = rblk->parent->lun;
cmnt_size = atomic_inc_return(&rblk->data_cmnt_size);
if (unlikely(cmnt_size == dev->geo.sec_per_blk))
@@ -1061,16 +1086,21 @@ static int rrpc_l2p_update(u64 slba, u32 nlb, __le64 *entries, void *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;
}
@@ -1085,6 +1115,25 @@ static int rrpc_l2p_update(u64 slba, u32 nlb, __le64 *entries, void *private)
addr[i].addr = pba;
raddr[mod].addr = slba + i;
+
+ gaddr = rrpc_ppa_to_gaddr(dev, pba);
+ rlun = rrpc_ppa_to_lun(rrpc, gaddr);
+ if (!rlun) {
+ pr_err("rrpc: unmapped lun %d l2p corruption\n",
+ rlun->parent->id);
+ 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--;
+ }
}
return 0;
@@ -1199,6 +1248,52 @@ static void rrpc_luns_free(struct rrpc *rrpc)
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;
+
+ 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->parent->chnl_id;
+ ppa.g.lun = rlun->parent->lun_id;
+
+ ret = nvm_get_bb_tbl(dev->parent, ppa, blks);
+ if (ret) {
+ pr_err("rrpc: could not get BB table\n");
+ kfree(blks);
+ goto out;
+ }
+
+ nr_blks = nvm_bb_tbl_fold(dev->parent, blks, nr_blks);
+ if (nr_blks < 0)
+ return nr_blks;
+
+ rlun->nr_free_blocks = geo->blks_per_lun;
+ 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 int rrpc_luns_init(struct rrpc *rrpc, struct list_head *lun_list)
{
struct nvm_tgt_dev *dev = rrpc->dev;
@@ -1232,16 +1327,26 @@ static int rrpc_luns_init(struct rrpc *rrpc, struct list_head *lun_list)
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];
- struct nvm_block *blk = &lun->blocks[j];
- rblk->parent = blk;
+ 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);
}
+ if (rrpc_bb_discovery(dev, rlun))
+ goto err;
+
rlun->reserved_blocks = 2; /* for GC only */
rlun->rrpc = rrpc;
diff --git a/drivers/lightnvm/rrpc.h b/drivers/lightnvm/rrpc.h
index 242d4c1..c55cec9 100644
--- a/drivers/lightnvm/rrpc.h
+++ b/drivers/lightnvm/rrpc.h
@@ -52,8 +52,7 @@ struct rrpc_rq {
};
struct rrpc_block {
- unsigned long id;
- struct nvm_block *parent;
+ int id; /* id inside of LUN */
struct rrpc_lun *rlun;
struct list_head prio; /* LUN CG list */
@@ -83,6 +82,16 @@ struct rrpc_lun {
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;
@@ -155,14 +164,14 @@ struct rrpc_rev_addr {
u64 addr;
};
-static inline struct rrpc_block *rrpc_get_rblk(struct rrpc_lun *rlun,
- int blk_id)
+static inline u64 rrpc_blk_to_ppa(struct rrpc *rrpc, struct rrpc_block *rblk)
{
- struct rrpc *rrpc = rlun->rrpc;
struct nvm_tgt_dev *dev = rrpc->dev;
- int lun_blk = blk_id % dev->geo.blks_per_lun;
+ struct nvm_geo *geo = &dev->geo;
+ struct rrpc_lun *rlun = rblk->rlun;
+ struct nvm_lun *lun = rlun->parent;
- return &rlun->blocks[lun_blk];
+ return (lun->id * geo->sec_per_lun) + (rblk->id * geo->sec_per_blk);
}
static inline sector_t rrpc_get_laddr(struct bio *bio)
diff --git a/include/linux/lightnvm.h b/include/linux/lightnvm.h
index ed04fa6..cc210cc 100644
--- a/include/linux/lightnvm.h
+++ b/include/linux/lightnvm.h
@@ -266,8 +266,6 @@ static inline void *nvm_rq_to_pdu(struct nvm_rq *rqdata)
return rqdata + 1;
}
-struct nvm_block;
-
struct nvm_lun {
int id;
@@ -275,19 +273,6 @@ struct nvm_lun {
int chnl_id;
struct list_head list;
- spinlock_t lock;
-
- /* 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 nvm_block *blocks;
};
enum {
@@ -296,15 +281,6 @@ enum {
NVM_BLK_ST_BAD = 0x8, /* Bad block */
};
-struct nvm_block {
- struct list_head list;
- struct nvm_lun *lun;
- unsigned long id;
-
- void *priv;
- int state;
-};
-
/* system block cpu representation */
struct nvm_sb_info {
unsigned long seqnr;
@@ -473,21 +449,6 @@ static inline void ppa_set_empty(struct ppa_addr *ppa_addr)
ppa_addr->ppa = ADDR_EMPTY;
}
-static inline struct ppa_addr block_to_ppa(struct nvm_dev *dev,
- struct nvm_block *blk)
-{
- struct nvm_geo *geo = &dev->geo;
- struct ppa_addr ppa;
- struct nvm_lun *lun = blk->lun;
-
- ppa.ppa = 0;
- ppa.g.blk = blk->id % geo->blks_per_lun;
- ppa.g.lun = lun->lun_id;
- ppa.g.ch = lun->chnl_id;
-
- return ppa;
-}
-
static inline int ppa_cmp_blk(struct ppa_addr ppa1, struct ppa_addr ppa2)
{
if (ppa_empty(ppa1) || ppa_empty(ppa2))
@@ -539,8 +500,7 @@ typedef void (nvmm_unregister_fn)(struct nvm_dev *);
typedef int (nvmm_create_tgt_fn)(struct nvm_dev *, struct nvm_ioctl_create *);
typedef int (nvmm_remove_tgt_fn)(struct nvm_dev *, struct nvm_ioctl_remove *);
typedef int (nvmm_submit_io_fn)(struct nvm_dev *, struct nvm_rq *);
-typedef int (nvmm_erase_blk_fn)(struct nvm_dev *, struct nvm_block *, int);
-typedef void (nvmm_mark_blk_fn)(struct nvm_dev *, struct ppa_addr, int);
+typedef int (nvmm_erase_blk_fn)(struct nvm_dev *, struct ppa_addr *, int);
typedef int (nvmm_get_area_fn)(struct nvm_dev *, sector_t *, sector_t);
typedef void (nvmm_put_area_fn)(struct nvm_dev *, sector_t);
@@ -557,9 +517,6 @@ struct nvmm_type {
nvmm_submit_io_fn *submit_io;
nvmm_erase_blk_fn *erase_blk;
- /* Bad block mgmt */
- nvmm_mark_blk_fn *mark_blk;
-
nvmm_get_area_fn *get_area;
nvmm_put_area_fn *put_area;
@@ -573,7 +530,6 @@ extern struct nvm_dev *nvm_alloc_dev(int);
extern int nvm_register(struct nvm_dev *);
extern void nvm_unregister(struct nvm_dev *);
-extern void nvm_mark_blk(struct nvm_dev *dev, struct ppa_addr ppa, int type);
extern int nvm_set_bb_tbl(struct nvm_dev *dev, struct ppa_addr *ppas,
int nr_ppas, int type);
@@ -584,7 +540,7 @@ extern int nvm_set_rqd_ppalist(struct nvm_dev *, struct nvm_rq *,
const struct ppa_addr *, int, int);
extern void nvm_free_rqd_ppalist(struct nvm_dev *, struct nvm_rq *);
extern int nvm_erase_ppa(struct nvm_dev *, struct ppa_addr *, int, int);
-extern int nvm_erase_blk(struct nvm_dev *, struct nvm_block *, int);
+extern int nvm_erase_blk(struct nvm_dev *, struct ppa_addr *, int);
extern void nvm_end_io(struct nvm_rq *, int);
extern int nvm_submit_ppa(struct nvm_dev *, struct ppa_addr *, int, int, int,
void *, int);
--
2.7.4
[toc] | [prev] | [next] | [standalone]
| From | kbuild test robot <lkp@intel.com> |
|---|---|
| Date | 2016-11-18 19:50 +0100 |
| Subject | Re: [PATCH 16/20] lightnvm: eliminate nvm_block abstraction on mm |
| Message-ID | <sEWox-2qv-1@gated-at.bofh.it> |
| In reply to | #1525420 |
Hi Javier,
[auto build test WARNING on linus/master]
[also build test WARNING on v4.9-rc5]
[cannot apply to next-20161117]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]
url: https://github.com/0day-ci/linux/commits/Javier-Gonz-lez/lightnvm-simplify-media-manager/20161119-005946
coccinelle warnings: (new ones prefixed by >>)
>> drivers/lightnvm/rrpc.c:1123:13-19: ERROR: rlun is NULL but dereferenced.
--
>> drivers/lightnvm/rrpc.c:1293:7-11: ERROR: reference preceded by free on line 1273
vim +1123 drivers/lightnvm/rrpc.c
1117 raddr[mod].addr = slba + i;
1118
1119 gaddr = rrpc_ppa_to_gaddr(dev, pba);
1120 rlun = rrpc_ppa_to_lun(rrpc, gaddr);
1121 if (!rlun) {
1122 pr_err("rrpc: unmapped lun %d l2p corruption\n",
> 1123 rlun->parent->id);
1124 return -EINVAL;
1125 }
1126
1127 rblk = &rlun->blocks[gaddr.g.blk];
1128 if (!rblk->state) {
1129 /* at this point, we don't know anything about the
1130 * block. It's up to the FTL on top to re-etablish the
1131 * block state. The block is assumed to be open.
1132 */
1133 list_move_tail(&rblk->list, &rlun->used_list);
1134 rblk->state = NVM_BLK_ST_TGT;
1135 rlun->nr_free_blocks--;
1136 }
1137 }
1138
1139 return 0;
1140 }
1141
1142 static int rrpc_map_init(struct rrpc *rrpc)
1143 {
1144 struct nvm_tgt_dev *dev = rrpc->dev;
1145 sector_t i;
1146 int ret;
1147
1148 rrpc->trans_map = vzalloc(sizeof(struct rrpc_addr) * rrpc->nr_sects);
1149 if (!rrpc->trans_map)
1150 return -ENOMEM;
1151
1152 rrpc->rev_trans_map = vmalloc(sizeof(struct rrpc_rev_addr)
1153 * rrpc->nr_sects);
1154 if (!rrpc->rev_trans_map)
1155 return -ENOMEM;
1156
1157 for (i = 0; i < rrpc->nr_sects; i++) {
1158 struct rrpc_addr *p = &rrpc->trans_map[i];
1159 struct rrpc_rev_addr *r = &rrpc->rev_trans_map[i];
1160
1161 p->addr = ADDR_EMPTY;
1162 r->addr = ADDR_EMPTY;
1163 }
1164
1165 if (!dev->ops->get_l2p_tbl)
1166 return 0;
1167
1168 /* Bring up the mapping table from device */
1169 ret = dev->ops->get_l2p_tbl(dev->parent, rrpc->soffset, rrpc->nr_sects,
1170 rrpc_l2p_update, rrpc);
1171 if (ret) {
1172 pr_err("nvm: rrpc: could not read L2P table.\n");
1173 return -EINVAL;
1174 }
1175
1176 return 0;
1177 }
1178
1179 /* Minimum pages needed within a lun */
1180 #define PAGE_POOL_SIZE 16
1181 #define ADDR_POOL_SIZE 64
1182
1183 static int rrpc_core_init(struct rrpc *rrpc)
1184 {
1185 down_write(&rrpc_lock);
1186 if (!rrpc_gcb_cache) {
1187 rrpc_gcb_cache = kmem_cache_create("rrpc_gcb",
1188 sizeof(struct rrpc_block_gc), 0, 0, NULL);
1189 if (!rrpc_gcb_cache) {
1190 up_write(&rrpc_lock);
1191 return -ENOMEM;
1192 }
1193
1194 rrpc_rq_cache = kmem_cache_create("rrpc_rq",
1195 sizeof(struct nvm_rq) + sizeof(struct rrpc_rq),
1196 0, 0, NULL);
1197 if (!rrpc_rq_cache) {
1198 kmem_cache_destroy(rrpc_gcb_cache);
1199 up_write(&rrpc_lock);
1200 return -ENOMEM;
1201 }
1202 }
1203 up_write(&rrpc_lock);
1204
1205 rrpc->page_pool = mempool_create_page_pool(PAGE_POOL_SIZE, 0);
1206 if (!rrpc->page_pool)
1207 return -ENOMEM;
1208
1209 rrpc->gcb_pool = mempool_create_slab_pool(rrpc->dev->geo.nr_luns,
1210 rrpc_gcb_cache);
1211 if (!rrpc->gcb_pool)
1212 return -ENOMEM;
1213
1214 rrpc->rq_pool = mempool_create_slab_pool(64, rrpc_rq_cache);
1215 if (!rrpc->rq_pool)
1216 return -ENOMEM;
1217
1218 spin_lock_init(&rrpc->inflights.lock);
1219 INIT_LIST_HEAD(&rrpc->inflights.reqs);
1220
1221 return 0;
1222 }
1223
1224 static void rrpc_core_free(struct rrpc *rrpc)
1225 {
1226 mempool_destroy(rrpc->page_pool);
1227 mempool_destroy(rrpc->gcb_pool);
1228 mempool_destroy(rrpc->rq_pool);
1229 }
1230
1231 static void rrpc_luns_free(struct rrpc *rrpc)
1232 {
1233 struct nvm_lun *lun;
1234 struct rrpc_lun *rlun;
1235 int i;
1236
1237 if (!rrpc->luns)
1238 return;
1239
1240 for (i = 0; i < rrpc->nr_luns; i++) {
1241 rlun = &rrpc->luns[i];
1242 lun = rlun->parent;
1243 if (!lun)
1244 break;
1245 vfree(rlun->blocks);
1246 }
1247
1248 kfree(rrpc->luns);
1249 }
1250
1251 static int rrpc_bb_discovery(struct nvm_tgt_dev *dev, struct rrpc_lun *rlun)
1252 {
1253 struct nvm_geo *geo = &dev->geo;
1254 struct rrpc_block *rblk;
1255 struct ppa_addr ppa;
1256 u8 *blks;
1257 int nr_blks;
1258 int i;
1259 int ret;
1260
1261 nr_blks = geo->blks_per_lun * geo->plane_mode;
1262 blks = kmalloc(nr_blks, GFP_KERNEL);
1263 if (!blks)
1264 return -ENOMEM;
1265
1266 ppa.ppa = 0;
1267 ppa.g.ch = rlun->parent->chnl_id;
1268 ppa.g.lun = rlun->parent->lun_id;
1269
1270 ret = nvm_get_bb_tbl(dev->parent, ppa, blks);
1271 if (ret) {
1272 pr_err("rrpc: could not get BB table\n");
> 1273 kfree(blks);
1274 goto out;
1275 }
1276
1277 nr_blks = nvm_bb_tbl_fold(dev->parent, blks, nr_blks);
1278 if (nr_blks < 0)
1279 return nr_blks;
1280
1281 rlun->nr_free_blocks = geo->blks_per_lun;
1282 for (i = 0; i < nr_blks; i++) {
1283 if (blks[i] == NVM_BLK_T_FREE)
1284 continue;
1285
1286 rblk = &rlun->blocks[i];
1287 list_move_tail(&rblk->list, &rlun->bb_list);
1288 rblk->state = NVM_BLK_ST_BAD;
1289 rlun->nr_free_blocks--;
1290 }
1291
1292 out:
> 1293 kfree(blks);
1294 return ret;
1295 }
1296
---
0-DAY kernel test infrastructure Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all Intel Corporation
[toc] | [prev] | [next] | [standalone]
| From | "Javier González" <jg@lightnvm.io> |
|---|---|
| Date | 2016-11-18 15:50 +0100 |
| Subject | [PATCH 18/20] lightnvm: introduce helpers for generic ops in rrpc |
| Message-ID | <sESEj-8qT-53@gated-at.bofh.it> |
| In reply to | #1525412 |
Avoid calling media manager and device-specific operations directly from
rrpc. Create helper functions on lightnvm's core instead.
Signed-off-by: Javier González <javier@cnexlabs.com>
---
drivers/lightnvm/core.c | 22 ++++++++++++++++++++++
drivers/lightnvm/gennvm.c | 2 --
drivers/lightnvm/rrpc.c | 11 +++--------
include/linux/lightnvm.h | 3 ---
4 files changed, 25 insertions(+), 13 deletions(-)
diff --git a/drivers/lightnvm/core.c b/drivers/lightnvm/core.c
index 23d582f..69f261e 100644
--- a/drivers/lightnvm/core.c
+++ b/drivers/lightnvm/core.c
@@ -219,6 +219,28 @@ int nvm_erase_blk(struct nvm_tgt_dev *tgt_dev, struct ppa_addr *p, int flags)
}
EXPORT_SYMBOL(nvm_erase_blk);
+int nvm_get_l2p_tbl(struct nvm_dev *dev, u64 slba, u32 nlb,
+ nvm_l2p_update_fn *update_l2p, void *priv)
+{
+ if (!dev->ops->get_l2p_tbl)
+ return 0;
+
+ return dev->ops->get_l2p_tbl(dev, slba, nlb, update_l2p, priv);
+}
+EXPORT_SYMBOL(nvm_get_l2p_tbl);
+
+int nvm_get_area(struct nvm_dev *dev, sector_t *lba, sector_t len)
+{
+ return dev->mt->get_area(dev, lba, len);
+}
+EXPORT_SYMBOL(nvm_get_area);
+
+void nvm_put_area(struct nvm_dev *dev, sector_t lba)
+{
+ dev->mt->put_area(dev, lba);
+}
+EXPORT_SYMBOL(nvm_put_area);
+
void nvm_addr_to_generic_mode(struct nvm_dev *dev, struct nvm_rq *rqd)
{
int i;
diff --git a/drivers/lightnvm/gennvm.c b/drivers/lightnvm/gennvm.c
index 5d7c8c4..befa828 100644
--- a/drivers/lightnvm/gennvm.c
+++ b/drivers/lightnvm/gennvm.c
@@ -175,8 +175,6 @@ static struct nvm_tgt_dev *gen_create_tgt_dev(struct nvm_dev *dev,
tgt_dev->geo.luns_per_chnl = (lun_balanced) ? prev_nr_luns : -1;
tgt_dev->total_secs = nr_luns * tgt_dev->geo.sec_per_lun;
tgt_dev->q = dev->q;
- tgt_dev->ops = dev->ops;
- tgt_dev->mt = dev->mt;
tgt_dev->map = dev_map;
tgt_dev->luns = luns;
memcpy(&tgt_dev->identity, &dev->identity, sizeof(struct nvm_id));
diff --git a/drivers/lightnvm/rrpc.c b/drivers/lightnvm/rrpc.c
index 08a633c..150b5f30 100644
--- a/drivers/lightnvm/rrpc.c
+++ b/drivers/lightnvm/rrpc.c
@@ -1167,11 +1167,8 @@ static int rrpc_map_init(struct rrpc *rrpc)
r->addr = ADDR_EMPTY;
}
- if (!dev->ops->get_l2p_tbl)
- return 0;
-
/* Bring up the mapping table from device */
- ret = dev->ops->get_l2p_tbl(dev->parent, rrpc->soffset, rrpc->nr_sects,
+ ret = nvm_get_l2p_tbl(dev->parent, rrpc->soffset, rrpc->nr_sects,
rrpc_l2p_update, rrpc);
if (ret) {
pr_err("nvm: rrpc: could not read L2P table.\n");
@@ -1372,13 +1369,12 @@ static int rrpc_luns_init(struct rrpc *rrpc, struct ppa_addr *luns)
static int rrpc_area_init(struct rrpc *rrpc, sector_t *begin)
{
struct nvm_tgt_dev *dev = rrpc->dev;
- struct nvmm_type *mt = dev->mt;
sector_t size = rrpc->nr_sects * dev->geo.sec_size;
int ret;
size >>= 9;
- ret = mt->get_area(dev->parent, begin, size);
+ ret = nvm_get_area(dev->parent, begin, size);
if (!ret)
*begin >>= (ilog2(dev->geo.sec_size) - 9);
@@ -1388,10 +1384,9 @@ static int rrpc_area_init(struct rrpc *rrpc, sector_t *begin)
static void rrpc_area_free(struct rrpc *rrpc)
{
struct nvm_tgt_dev *dev = rrpc->dev;
- struct nvmm_type *mt = dev->mt;
sector_t begin = rrpc->soffset << (ilog2(dev->geo.sec_size) - 9);
- mt->put_area(dev->parent, begin);
+ nvm_put_area(dev->parent, begin);
}
static void rrpc_free(struct rrpc *rrpc)
diff --git a/include/linux/lightnvm.h b/include/linux/lightnvm.h
index 2222853..b3ca603 100644
--- a/include/linux/lightnvm.h
+++ b/include/linux/lightnvm.h
@@ -320,9 +320,6 @@ struct nvm_tgt_dev {
struct nvm_id identity;
struct request_queue *q;
- struct nvmm_type *mt;
- struct nvm_dev_ops *ops;
-
void *parent;
void *map;
};
--
2.7.4
[toc] | [prev] | [next] | [standalone]
| From | "Javier González" <jg@lightnvm.io> |
|---|---|
| Date | 2016-11-18 16:00 +0100 |
| Subject | [PATCH 03/20] lightnvm: export set bad block table |
| Message-ID | <sESNX-8ud-7@gated-at.bofh.it> |
| In reply to | #1525412 |
From: Javier González <jg@lightnvm.io>
Bad blocks should be managed by block owners. This would be either
targets for data blocks or sysblk for system blocks.
In order to support this, export two functions: One to mark a block as
an specific type (e.g., bad block) and another to update the bad block
table on the device.
Move bad block management to rrpc.
Signed-off-by: Javier González <javier@cnexlabs.com>
Signed-off-by: Matias Bjørling <m@bjorling.me>
---
drivers/lightnvm/core.c | 27 +++++++++++++++++++++++++++
drivers/lightnvm/gennvm.c | 24 ------------------------
drivers/lightnvm/rrpc.c | 34 +++++++++++++++++++++++++++++++++-
drivers/lightnvm/sysblk.c | 29 +++++------------------------
include/linux/lightnvm.h | 13 ++++++++++++-
5 files changed, 77 insertions(+), 50 deletions(-)
diff --git a/drivers/lightnvm/core.c b/drivers/lightnvm/core.c
index 8664fe0..6527cf6 100644
--- a/drivers/lightnvm/core.c
+++ b/drivers/lightnvm/core.c
@@ -196,6 +196,33 @@ void nvm_mark_blk(struct nvm_dev *dev, struct ppa_addr ppa, int type)
}
EXPORT_SYMBOL(nvm_mark_blk);
+int nvm_set_bb_tbl(struct nvm_dev *dev, struct ppa_addr *ppas, int nr_ppas,
+ int type)
+{
+ struct nvm_rq rqd;
+ int ret;
+
+ if (nr_ppas > dev->ops->max_phys_sect) {
+ pr_err("nvm: unable to update all sysblocks atomically\n");
+ return -EINVAL;
+ }
+
+ memset(&rqd, 0, sizeof(struct nvm_rq));
+
+ nvm_set_rqd_ppalist(dev, &rqd, ppas, nr_ppas, 1);
+ nvm_generic_to_addr_mode(dev, &rqd);
+
+ ret = dev->ops->set_bb_tbl(dev, &rqd.ppa_addr, rqd.nr_ppas, type);
+ nvm_free_rqd_ppalist(dev, &rqd);
+ if (ret) {
+ pr_err("nvm: sysblk failed bb mark\n");
+ return -EINVAL;
+ }
+
+ return 0;
+}
+EXPORT_SYMBOL(nvm_set_bb_tbl);
+
int nvm_submit_io(struct nvm_dev *dev, struct nvm_rq *rqd)
{
return dev->mt->submit_io(dev, rqd);
diff --git a/drivers/lightnvm/gennvm.c b/drivers/lightnvm/gennvm.c
index a7e17fa..e969e3a 100644
--- a/drivers/lightnvm/gennvm.c
+++ b/drivers/lightnvm/gennvm.c
@@ -543,34 +543,10 @@ static void gen_mark_blk(struct nvm_dev *dev, struct ppa_addr ppa, int type)
blk->state = type;
}
-/*
- * mark block bad in gen. It is expected that the target recovers separately
- */
-static void gen_mark_blk_bad(struct nvm_dev *dev, struct nvm_rq *rqd)
-{
- int bit = -1;
- int max_secs = dev->ops->max_phys_sect;
- void *comp_bits = &rqd->ppa_status;
-
- nvm_addr_to_generic_mode(dev, rqd);
-
- /* look up blocks and mark them as bad */
- if (rqd->nr_ppas == 1) {
- gen_mark_blk(dev, rqd->ppa_addr, NVM_BLK_ST_BAD);
- return;
- }
-
- while ((bit = find_next_bit(comp_bits, max_secs, bit + 1)) < max_secs)
- gen_mark_blk(dev, rqd->ppa_list[bit], NVM_BLK_ST_BAD);
-}
-
static void gen_end_io(struct nvm_rq *rqd)
{
struct nvm_tgt_instance *ins = rqd->ins;
- if (rqd->error == NVM_RSP_ERR_FAILWRITE)
- gen_mark_blk_bad(rqd->dev, rqd);
-
ins->tt->end_io(rqd);
}
diff --git a/drivers/lightnvm/rrpc.c b/drivers/lightnvm/rrpc.c
index 067e890..2b71b7e 100644
--- a/drivers/lightnvm/rrpc.c
+++ b/drivers/lightnvm/rrpc.c
@@ -675,6 +675,34 @@ static void rrpc_run_gc(struct rrpc *rrpc, struct rrpc_block *rblk)
queue_work(rrpc->kgc_wq, &gcb->ws_gc);
}
+static void __rrpc_mark_bad_block(struct nvm_dev *dev, struct ppa_addr *ppa)
+{
+ nvm_mark_blk(dev, *ppa, NVM_BLK_ST_BAD);
+ nvm_set_bb_tbl(dev, ppa, 1, NVM_BLK_T_GRWN_BAD);
+}
+
+static void rrpc_mark_bad_block(struct rrpc *rrpc, struct nvm_rq *rqd)
+{
+ struct nvm_dev *dev = rrpc->dev;
+ 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(dev, &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(dev, &ppa);
+ }
+}
+
static void rrpc_end_io_write(struct rrpc *rrpc, struct rrpc_rq *rrqd,
sector_t laddr, uint8_t npages)
{
@@ -701,8 +729,12 @@ static void rrpc_end_io(struct nvm_rq *rqd)
uint8_t npages = rqd->nr_ppas;
sector_t laddr = rrpc_get_laddr(rqd->bio) - npages;
- if (bio_data_dir(rqd->bio) == WRITE)
+ 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);
diff --git a/drivers/lightnvm/sysblk.c b/drivers/lightnvm/sysblk.c
index d229067..fa644af 100644
--- a/drivers/lightnvm/sysblk.c
+++ b/drivers/lightnvm/sysblk.c
@@ -267,29 +267,10 @@ static int nvm_scan_block(struct nvm_dev *dev, struct ppa_addr *ppa,
return found;
}
-static int nvm_set_bb_tbl(struct nvm_dev *dev, struct sysblk_scan *s, int type)
+static int nvm_sysblk_set_bb_tbl(struct nvm_dev *dev, struct sysblk_scan *s,
+ int type)
{
- struct nvm_rq rqd;
- int ret;
-
- if (s->nr_ppas > dev->ops->max_phys_sect) {
- pr_err("nvm: unable to update all sysblocks atomically\n");
- return -EINVAL;
- }
-
- memset(&rqd, 0, sizeof(struct nvm_rq));
-
- nvm_set_rqd_ppalist(dev, &rqd, s->ppas, s->nr_ppas, 1);
- nvm_generic_to_addr_mode(dev, &rqd);
-
- ret = dev->ops->set_bb_tbl(dev, &rqd.ppa_addr, rqd.nr_ppas, type);
- nvm_free_rqd_ppalist(dev, &rqd);
- if (ret) {
- pr_err("nvm: sysblk failed bb mark\n");
- return -EINVAL;
- }
-
- return 0;
+ return nvm_set_bb_tbl(dev, s->ppas, s->nr_ppas, type);
}
static int nvm_write_and_verify(struct nvm_dev *dev, struct nvm_sb_info *info,
@@ -573,7 +554,7 @@ int nvm_init_sysblock(struct nvm_dev *dev, struct nvm_sb_info *info)
if (ret)
goto err_mark;
- ret = nvm_set_bb_tbl(dev, &s, NVM_BLK_T_HOST);
+ ret = nvm_sysblk_set_bb_tbl(dev, &s, NVM_BLK_T_HOST);
if (ret)
goto err_mark;
@@ -733,7 +714,7 @@ int nvm_dev_factory(struct nvm_dev *dev, int flags)
mutex_lock(&dev->mlock);
ret = nvm_get_all_sysblks(dev, &s, sysblk_ppas, 0);
if (!ret)
- ret = nvm_set_bb_tbl(dev, &s, NVM_BLK_T_FREE);
+ ret = nvm_sysblk_set_bb_tbl(dev, &s, NVM_BLK_T_FREE);
mutex_unlock(&dev->mlock);
}
err_ppas:
diff --git a/include/linux/lightnvm.h b/include/linux/lightnvm.h
index d87be02..4480d1c 100644
--- a/include/linux/lightnvm.h
+++ b/include/linux/lightnvm.h
@@ -423,6 +423,15 @@ static inline struct ppa_addr block_to_ppa(struct nvm_dev *dev,
return ppa;
}
+static inline int ppa_cmp_blk(struct ppa_addr ppa1, struct ppa_addr ppa2)
+{
+ if (ppa_empty(ppa1) || ppa_empty(ppa2))
+ return 0;
+
+ return ((ppa1.g.ch == ppa2.g.ch) && (ppa1.g.lun == ppa2.g.lun) &&
+ (ppa1.g.blk == ppa2.g.blk));
+}
+
static inline int ppa_to_slc(struct nvm_dev *dev, int slc_pg)
{
return dev->lptbl[slc_pg];
@@ -528,7 +537,9 @@ extern struct nvm_dev *nvm_alloc_dev(int);
extern int nvm_register(struct nvm_dev *);
extern void nvm_unregister(struct nvm_dev *);
-void nvm_mark_blk(struct nvm_dev *dev, struct ppa_addr ppa, int type);
+extern void nvm_mark_blk(struct nvm_dev *dev, struct ppa_addr ppa, int type);
+extern int nvm_set_bb_tbl(struct nvm_dev *dev, struct ppa_addr *ppas,
+ int nr_ppas, int type);
extern int nvm_submit_io(struct nvm_dev *, struct nvm_rq *);
extern void nvm_generic_to_addr_mode(struct nvm_dev *, struct nvm_rq *);
--
2.7.4
[toc] | [prev] | [next] | [standalone]
| From | "Javier González" <jg@lightnvm.io> |
|---|---|
| Date | 2016-11-18 16:00 +0100 |
| Subject | [PATCH 02/20] lightnvm: do not protect block 0 |
| Message-ID | <sESNX-8ud-15@gated-at.bofh.it> |
| In reply to | #1525412 |
From: Javier González <jg@lightnvm.io>
Device blocks should be marked by the device and considered as bad
blocks by the media manager. Thus, do not make assumptions on which
blocks are going to be used by the device. In doing so we might lose
valid blocks from the free list.
Signed-off-by: Javier González <javier@cnexlabs.com>
Signed-off-by: Matias Bjørling <m@bjorling.me>
---
drivers/lightnvm/gennvm.c | 6 ------
1 file changed, 6 deletions(-)
diff --git a/drivers/lightnvm/gennvm.c b/drivers/lightnvm/gennvm.c
index 730d736..a7e17fa 100644
--- a/drivers/lightnvm/gennvm.c
+++ b/drivers/lightnvm/gennvm.c
@@ -371,12 +371,6 @@ static int gen_blocks_init(struct nvm_dev *dev, struct gen_dev *gn)
block->lun = &lun->vlun;
block->id = cur_block_id++;
- /* First block is reserved for device */
- if (unlikely(lun_iter == 0 && blk_iter == 0)) {
- lun->vlun.nr_free_blocks--;
- continue;
- }
-
list_add_tail(&block->list, &lun->free_list);
}
--
2.7.4
[toc] | [prev] | [next] | [standalone]
| From | "Javier González" <jg@lightnvm.io> |
|---|---|
| Date | 2016-11-18 16:00 +0100 |
| Subject | [PATCH 01/20] lightnvm: enable to send hint to erase command |
| Message-ID | <sESNX-8ud-23@gated-at.bofh.it> |
| In reply to | #1525412 |
From: Javier González <jg@lightnvm.io>
Erases might be subject to host hints. An example is multi-plane
programming to erase blocks in parallel. Enable targets to specify this
hints.
Signed-off-by: Javier González <javier@cnexlabs.com>
Signed-off-by: Matias Bjørling <m@bjorling.me>
---
drivers/lightnvm/core.c | 9 ++++++---
drivers/lightnvm/gennvm.c | 5 ++---
drivers/lightnvm/rrpc.c | 2 +-
drivers/lightnvm/sysblk.c | 4 ++--
drivers/nvme/host/lightnvm.c | 1 +
include/linux/lightnvm.h | 7 +++----
6 files changed, 15 insertions(+), 13 deletions(-)
diff --git a/drivers/lightnvm/core.c b/drivers/lightnvm/core.c
index 1111740..8664fe0 100644
--- a/drivers/lightnvm/core.c
+++ b/drivers/lightnvm/core.c
@@ -202,9 +202,9 @@ int nvm_submit_io(struct nvm_dev *dev, struct nvm_rq *rqd)
}
EXPORT_SYMBOL(nvm_submit_io);
-int nvm_erase_blk(struct nvm_dev *dev, struct nvm_block *blk)
+int nvm_erase_blk(struct nvm_dev *dev, struct nvm_block *blk, int flags)
{
- return dev->mt->erase_blk(dev, blk, 0);
+ return dev->mt->erase_blk(dev, blk, flags);
}
EXPORT_SYMBOL(nvm_erase_blk);
@@ -285,7 +285,8 @@ void nvm_free_rqd_ppalist(struct nvm_dev *dev, struct nvm_rq *rqd)
}
EXPORT_SYMBOL(nvm_free_rqd_ppalist);
-int nvm_erase_ppa(struct nvm_dev *dev, struct ppa_addr *ppas, int nr_ppas)
+int nvm_erase_ppa(struct nvm_dev *dev, struct ppa_addr *ppas, int nr_ppas,
+ int flags)
{
struct nvm_rq rqd;
int ret;
@@ -301,6 +302,8 @@ int nvm_erase_ppa(struct nvm_dev *dev, struct ppa_addr *ppas, int nr_ppas)
nvm_generic_to_addr_mode(dev, &rqd);
+ rqd.flags = flags;
+
ret = dev->ops->erase_block(dev, &rqd);
nvm_free_rqd_ppalist(dev, &rqd);
diff --git a/drivers/lightnvm/gennvm.c b/drivers/lightnvm/gennvm.c
index b74174c..730d736 100644
--- a/drivers/lightnvm/gennvm.c
+++ b/drivers/lightnvm/gennvm.c
@@ -593,12 +593,11 @@ static int gen_submit_io(struct nvm_dev *dev, struct nvm_rq *rqd)
return dev->ops->submit_io(dev, rqd);
}
-static int gen_erase_blk(struct nvm_dev *dev, struct nvm_block *blk,
- unsigned long flags)
+static int gen_erase_blk(struct nvm_dev *dev, struct nvm_block *blk, int flags)
{
struct ppa_addr addr = block_to_ppa(dev, blk);
- return nvm_erase_ppa(dev, &addr, 1);
+ return nvm_erase_ppa(dev, &addr, 1, flags);
}
static int gen_reserve_lun(struct nvm_dev *dev, int lunid)
diff --git a/drivers/lightnvm/rrpc.c b/drivers/lightnvm/rrpc.c
index 37fcaad..067e890 100644
--- a/drivers/lightnvm/rrpc.c
+++ b/drivers/lightnvm/rrpc.c
@@ -404,7 +404,7 @@ static void rrpc_block_gc(struct work_struct *work)
if (rrpc_move_valid_pages(rrpc, rblk))
goto put_back;
- if (nvm_erase_blk(dev, rblk->parent))
+ if (nvm_erase_blk(dev, rblk->parent, 0))
goto put_back;
rrpc_put_blk(rrpc, rblk);
diff --git a/drivers/lightnvm/sysblk.c b/drivers/lightnvm/sysblk.c
index a75bd28..d229067 100644
--- a/drivers/lightnvm/sysblk.c
+++ b/drivers/lightnvm/sysblk.c
@@ -379,7 +379,7 @@ static int nvm_prepare_new_sysblks(struct nvm_dev *dev, struct sysblk_scan *s)
ppa = &s->ppas[scan_ppa_idx(i, nxt_blk)];
ppa->g.pg = ppa_to_slc(dev, 0);
- ret = nvm_erase_ppa(dev, ppa, 1);
+ ret = nvm_erase_ppa(dev, ppa, 1, 0);
if (ret)
return ret;
@@ -725,7 +725,7 @@ int nvm_dev_factory(struct nvm_dev *dev, int flags)
/* continue to erase until list of blks until empty */
while ((ppa_cnt =
nvm_fact_get_blks(dev, ppas, max_ppas, blk_bitmap)) > 0)
- nvm_erase_ppa(dev, ppas, ppa_cnt);
+ nvm_erase_ppa(dev, ppas, ppa_cnt, 0);
/* mark host reserved blocks free */
if (flags & NVM_FACTORY_RESET_HOST_BLKS) {
diff --git a/drivers/nvme/host/lightnvm.c b/drivers/nvme/host/lightnvm.c
index f23e6fe..037dff5 100644
--- a/drivers/nvme/host/lightnvm.c
+++ b/drivers/nvme/host/lightnvm.c
@@ -526,6 +526,7 @@ static int nvme_nvm_erase_block(struct nvm_dev *dev, struct nvm_rq *rqd)
c.erase.nsid = cpu_to_le32(ns->ns_id);
c.erase.spba = cpu_to_le64(rqd->ppa_addr.ppa);
c.erase.length = cpu_to_le16(rqd->nr_ppas - 1);
+ c.erase.control = cpu_to_le16(rqd->flags);
return nvme_submit_sync_cmd(q, (struct nvme_command *)&c, NULL, 0);
}
diff --git a/include/linux/lightnvm.h b/include/linux/lightnvm.h
index fb2e601..d87be02 100644
--- a/include/linux/lightnvm.h
+++ b/include/linux/lightnvm.h
@@ -470,8 +470,7 @@ typedef int (nvmm_open_blk_fn)(struct nvm_dev *, struct nvm_block *);
typedef int (nvmm_close_blk_fn)(struct nvm_dev *, struct nvm_block *);
typedef void (nvmm_flush_blk_fn)(struct nvm_dev *, struct nvm_block *);
typedef int (nvmm_submit_io_fn)(struct nvm_dev *, struct nvm_rq *);
-typedef int (nvmm_erase_blk_fn)(struct nvm_dev *, struct nvm_block *,
- unsigned long);
+typedef int (nvmm_erase_blk_fn)(struct nvm_dev *, struct nvm_block *, int);
typedef void (nvmm_mark_blk_fn)(struct nvm_dev *, struct ppa_addr, int);
typedef struct nvm_lun *(nvmm_get_lun_fn)(struct nvm_dev *, int);
typedef int (nvmm_reserve_lun)(struct nvm_dev *, int);
@@ -537,8 +536,8 @@ extern void nvm_addr_to_generic_mode(struct nvm_dev *, struct nvm_rq *);
extern int nvm_set_rqd_ppalist(struct nvm_dev *, struct nvm_rq *,
const struct ppa_addr *, int, int);
extern void nvm_free_rqd_ppalist(struct nvm_dev *, struct nvm_rq *);
-extern int nvm_erase_ppa(struct nvm_dev *, struct ppa_addr *, int);
-extern int nvm_erase_blk(struct nvm_dev *, struct nvm_block *);
+extern int nvm_erase_ppa(struct nvm_dev *, struct ppa_addr *, int, int);
+extern int nvm_erase_blk(struct nvm_dev *, struct nvm_block *, int);
extern void nvm_end_io(struct nvm_rq *, int);
extern int nvm_submit_ppa(struct nvm_dev *, struct ppa_addr *, int, int, int,
void *, int);
--
2.7.4
[toc] | [prev] | [next] | [standalone]
| From | "Javier González" <jg@lightnvm.io> |
|---|---|
| Date | 2016-11-18 16:00 +0100 |
| Subject | [PATCH 19/20] lightnvm: introduce max_phys_sects helper function |
| Message-ID | <sESNY-8ud-43@gated-at.bofh.it> |
| In reply to | #1525412 |
Target devices do not have access to the device driver operations.
Introduce a helper function that exposes the max. number of physical
sectors supported by the underlying device.
Signed-off-by: Javier González <javier@cnexlabs.com>
---
drivers/lightnvm/core.c | 8 ++++++++
include/linux/lightnvm.h | 1 +
2 files changed, 9 insertions(+)
diff --git a/drivers/lightnvm/core.c b/drivers/lightnvm/core.c
index 69f261e..3d8eaac 100644
--- a/drivers/lightnvm/core.c
+++ b/drivers/lightnvm/core.c
@@ -203,6 +203,14 @@ int nvm_set_bb_tbl(struct nvm_dev *dev, struct ppa_addr *ppas, int nr_ppas,
}
EXPORT_SYMBOL(nvm_set_bb_tbl);
+int nvm_max_phys_sects(struct nvm_tgt_dev *tgt_dev)
+{
+ struct nvm_dev *dev = tgt_dev->parent;
+
+ return dev->ops->max_phys_sect;
+}
+EXPORT_SYMBOL(nvm_max_phys_sects);
+
int nvm_submit_io(struct nvm_tgt_dev *tgt_dev, struct nvm_rq *rqd)
{
struct nvm_dev *dev = tgt_dev->parent;
diff --git a/include/linux/lightnvm.h b/include/linux/lightnvm.h
index b3ca603..44780ef 100644
--- a/include/linux/lightnvm.h
+++ b/include/linux/lightnvm.h
@@ -529,6 +529,7 @@ extern void nvm_unregister(struct nvm_dev *);
extern int nvm_set_bb_tbl(struct nvm_dev *dev, struct ppa_addr *ppas,
int nr_ppas, int type);
+extern int nvm_max_phys_sects(struct nvm_tgt_dev *);
extern int nvm_submit_io(struct nvm_tgt_dev *, struct nvm_rq *);
extern void nvm_generic_to_addr_mode(struct nvm_dev *, struct nvm_rq *);
extern void nvm_addr_to_generic_mode(struct nvm_dev *, struct nvm_rq *);
--
2.7.4
[toc] | [prev] | [next] | [standalone]
| From | Matias Bjørling <mb@lightnvm.io> |
|---|---|
| Date | 2016-11-21 12:30 +0100 |
| Message-ID | <sFUXn-vK-15@gated-at.bofh.it> |
| In reply to | #1525412 |
On 11/18/2016 03:43 PM, Javier González wrote: > This patchset simplifies the generic media manager interface and moves > lun and block functionality to lightnvm targets. This makes that > rrpc-specific functionality is not exposed on the media manager. This is > done in preparation for the pblk target. > > Matias: patches 0 - 4 have been posted before. Patches apply to you > for-4.10/core > > Javier González (20): > lightnvm: enable to send hint to erase command > lightnvm: do not protect block 0 > lightnvm: export set bad block table > lightnvm: add ECC error codes > lightnvm: rrpc: split bios of size > 256kb > lightnvm: remove sysfs configuration interface > lightnvm: cleanup unused target operations > lightnvm: make address conversion functions global > lightnvm: remove unnecessary variables in rrpc > lightnvm: use constant name instead of value > lightnvm: remove gen_lun abstraction > lightnvm: manage lun partitions internally in mm > lightnvm: move block provisioning to targets > lightnvm: remove get_lun operation on gennvm > lightnvm: remove debug lun statistics from gennvm > lightnvm: eliminate nvm_block abstraction on mm > lightnvm: eliminate nvm_lun abstraction in mm > lightnvm: introduce helpers for generic ops in rrpc > lightnvm: introduce max_phys_sects helper function > lightnvm: use target nvm on target-specific ops. > > drivers/lightnvm/core.c | 306 ++++++++------------ > drivers/lightnvm/gennvm.c | 664 ++++++++++++++++++++----------------------- > drivers/lightnvm/gennvm.h | 34 ++- > drivers/lightnvm/rrpc.c | 512 +++++++++++++++++++++------------ > drivers/lightnvm/rrpc.h | 65 ++++- > drivers/lightnvm/sysblk.c | 98 +++---- > drivers/nvme/host/lightnvm.c | 15 +- > include/linux/lightnvm.h | 231 +++++++-------- > 8 files changed, 1005 insertions(+), 920 deletions(-) > Thanks Javier. I have picked 1-12 (including "manage lun partitions internally in mm"). Will wait on repost of 13-20 to pick the rest.
[toc] | [prev] | [standalone]
Back to top | Article view | linux.kernel
csiph-web