Re: [PATCH v3] lightnvm: pblk: add asynchronous partial read
Thank you Igor! Nice catch. If this is the only issue I'll send another patch.
On Fri, Jul 6, 2018 at 7:51 PM Igor Konopko wrote:
>
>
>
> On 06.07.2018 05:18, Matias Bjørling wrote:
> > On 07/06/2018 12:12 PM, Heiner Litz wrote:
> >> In the read path, partial reads are currently performed synchronously
> >> which affects performance for workloads that generate many partial
> >> reads. This patch adds an asynchronous partial read path as well as
> >> the required partial read ctx.
> >>
> >> Signed-off-by: Heiner Litz
> >>
> >> ---
> >>
> >> v3: rebase to head, incorporate 64-bit read bitmap
> >>
> >> ---
> >> drivers/lightnvm/pblk-read.c | 183
> >> ---
> >> drivers/lightnvm/pblk.h | 10 +++
> >> 2 files changed, 130 insertions(+), 63 deletions(-)
> >>
> >> diff --git a/drivers/lightnvm/pblk-read.c b/drivers/lightnvm/pblk-read.c
> >> index 9c9362b..4a44076 100644
> >> --- a/drivers/lightnvm/pblk-read.c
> >> +++ b/drivers/lightnvm/pblk-read.c
> >> @@ -231,74 +231,36 @@ static void pblk_end_io_read(struct nvm_rq *rqd)
> >> __pblk_end_io_read(pblk, rqd, true);
> >> }
> >> -static int pblk_partial_read(struct pblk *pblk, struct nvm_rq *rqd,
> >> - struct bio *orig_bio, unsigned int bio_init_idx,
> >> - unsigned long *read_bitmap)
> >> +static void pblk_end_partial_read(struct nvm_rq *rqd)
> >> {
> >> -struct pblk_sec_meta *meta_list = rqd->meta_list;
> >> -struct bio *new_bio;
> >> +struct pblk *pblk = rqd->private;
> >> +struct pblk_g_ctx *r_ctx = nvm_rq_to_pdu(rqd);
> >> +struct pblk_pr_ctx *pr_ctx = r_ctx->private;
> >> +struct bio *new_bio = rqd->bio;
> >> +struct bio *bio = pr_ctx->orig_bio;
> >> struct bio_vec src_bv, dst_bv;
> >> -void *ppa_ptr = NULL;
> >> -void *src_p, *dst_p;
> >> -dma_addr_t dma_ppa_list = 0;
> >> -__le64 *lba_list_mem, *lba_list_media;
> >> -int nr_secs = rqd->nr_ppas;
> >> +struct pblk_sec_meta *meta_list = rqd->meta_list;
> >> +int bio_init_idx = pr_ctx->bio_init_idx;
> >> +unsigned long *read_bitmap = _ctx->bitmap;
> >> +int nr_secs = pr_ctx->orig_nr_secs;
> >> int nr_holes = nr_secs - bitmap_weight(read_bitmap, nr_secs);
> >> -int i, ret, hole;
> >> -
> >> -/* Re-use allocated memory for intermediate lbas */
> >> -lba_list_mem = (((void *)rqd->ppa_list) + pblk_dma_ppa_size);
> >> -lba_list_media = (((void *)rqd->ppa_list) + 2 * pblk_dma_ppa_size);
> >> -
> >> -new_bio = bio_alloc(GFP_KERNEL, nr_holes);
> >> -
> >> -if (pblk_bio_add_pages(pblk, new_bio, GFP_KERNEL, nr_holes))
> >> -goto fail_add_pages;
> >> -
> >> -if (nr_holes != new_bio->bi_vcnt) {
> >> -pblk_err(pblk, "malformed bio\n");
> >> -goto fail;
> >> -}
> >> -
> >> -for (i = 0; i < nr_secs; i++)
> >> -lba_list_mem[i] = meta_list[i].lba;
> >> -
> >> -new_bio->bi_iter.bi_sector = 0; /* internal bio */
> >> -bio_set_op_attrs(new_bio, REQ_OP_READ, 0);
> >> -
> >> -rqd->bio = new_bio;
> >> -rqd->nr_ppas = nr_holes;
> >> -rqd->flags = pblk_set_read_mode(pblk, PBLK_READ_RANDOM);
> >> -
> >> -if (unlikely(nr_holes == 1)) {
> >> -ppa_ptr = rqd->ppa_list;
> >> -dma_ppa_list = rqd->dma_ppa_list;
> >> -rqd->ppa_addr = rqd->ppa_list[0];
> >> -}
> >> -
> >> -ret = pblk_submit_io_sync(pblk, rqd);
> >> -if (ret) {
> >> -bio_put(rqd->bio);
> >> -pblk_err(pblk, "sync read IO submission failed\n");
> >> -goto fail;
> >> -}
> >> -
> >> -if (rqd->error) {
> >> -atomic_long_inc(>read_failed);
> >> -#ifdef CONFIG_NVM_PBLK_DEBUG
> >> -pblk_print_failed_rqd(pblk, rqd, rqd->error);
> >> -#endif
> >> -}
> >> +__le64 *lba_list_mem, *lba_list_media;
> >> +void *src_p, *dst_p;
> >> +int hole, i;
> >> if (unlikely(nr_holes == 1)) {
> >> struct ppa_addr ppa;
> >> ppa = rqd->ppa_addr;
> >> -rqd->ppa_list = ppa_ptr;
> >> -rqd->dma_ppa_list = dma_ppa_list;
> >> +rqd->ppa_list = pr_ctx->ppa_ptr;
> >> +rqd->dma_ppa_list = pr_ctx->dma_ppa_list;
> >> rqd->ppa_list[0] = ppa;
> >> }
> >> +/* Re-use allocated memory for intermediate lbas */
> >> +lba_list_mem = (((void *)rqd->ppa_list) + pblk_dma_ppa_size);
> >> +lba_list_media = (((void *)rqd->ppa_list) + 2 * pblk_dma_ppa_size);
> >> +
> >> for (i = 0; i < nr_secs; i++) {
> >> lba_list_media[i] = meta_list[i].lba;
> >> meta_list[i].lba = lba_list_mem[i];
> >> @@ -316,7 +278,7 @@ static int pblk_partial_read(struct pblk *pblk,
> >> struct nvm_rq *rqd,
> >> meta_list[hole].lba = lba_list_media[i];
> >> src_bv = new_bio->bi_io_vec[i++];
> >> -dst_bv = orig_bio->bi_io_vec[bio_init_idx + hole];
> >> +dst_bv = bio->bi_io_vec[bio_init_idx + hole];
> >> src_p = kmap_atomic(src_bv.bv_page);
> >>
Re: [PATCH v3] lightnvm: pblk: add asynchronous partial read
On 06.07.2018 05:18, Matias Bjørling wrote:
On 07/06/2018 12:12 PM, Heiner Litz wrote:
In the read path, partial reads are currently performed synchronously
which affects performance for workloads that generate many partial
reads. This patch adds an asynchronous partial read path as well as
the required partial read ctx.
Signed-off-by: Heiner Litz
---
v3: rebase to head, incorporate 64-bit read bitmap
---
drivers/lightnvm/pblk-read.c | 183
---
drivers/lightnvm/pblk.h | 10 +++
2 files changed, 130 insertions(+), 63 deletions(-)
diff --git a/drivers/lightnvm/pblk-read.c b/drivers/lightnvm/pblk-read.c
index 9c9362b..4a44076 100644
--- a/drivers/lightnvm/pblk-read.c
+++ b/drivers/lightnvm/pblk-read.c
@@ -231,74 +231,36 @@ static void pblk_end_io_read(struct nvm_rq *rqd)
__pblk_end_io_read(pblk, rqd, true);
}
-static int pblk_partial_read(struct pblk *pblk, struct nvm_rq *rqd,
- struct bio *orig_bio, unsigned int bio_init_idx,
- unsigned long *read_bitmap)
+static void pblk_end_partial_read(struct nvm_rq *rqd)
{
- struct pblk_sec_meta *meta_list = rqd->meta_list;
- struct bio *new_bio;
+ struct pblk *pblk = rqd->private;
+ struct pblk_g_ctx *r_ctx = nvm_rq_to_pdu(rqd);
+ struct pblk_pr_ctx *pr_ctx = r_ctx->private;
+ struct bio *new_bio = rqd->bio;
+ struct bio *bio = pr_ctx->orig_bio;
struct bio_vec src_bv, dst_bv;
- void *ppa_ptr = NULL;
- void *src_p, *dst_p;
- dma_addr_t dma_ppa_list = 0;
- __le64 *lba_list_mem, *lba_list_media;
- int nr_secs = rqd->nr_ppas;
+ struct pblk_sec_meta *meta_list = rqd->meta_list;
+ int bio_init_idx = pr_ctx->bio_init_idx;
+ unsigned long *read_bitmap = _ctx->bitmap;
+ int nr_secs = pr_ctx->orig_nr_secs;
int nr_holes = nr_secs - bitmap_weight(read_bitmap, nr_secs);
- int i, ret, hole;
-
- /* Re-use allocated memory for intermediate lbas */
- lba_list_mem = (((void *)rqd->ppa_list) + pblk_dma_ppa_size);
- lba_list_media = (((void *)rqd->ppa_list) + 2 * pblk_dma_ppa_size);
-
- new_bio = bio_alloc(GFP_KERNEL, nr_holes);
-
- if (pblk_bio_add_pages(pblk, new_bio, GFP_KERNEL, nr_holes))
- goto fail_add_pages;
-
- if (nr_holes != new_bio->bi_vcnt) {
- pblk_err(pblk, "malformed bio\n");
- goto fail;
- }
-
- for (i = 0; i < nr_secs; i++)
- lba_list_mem[i] = meta_list[i].lba;
-
- new_bio->bi_iter.bi_sector = 0; /* internal bio */
- bio_set_op_attrs(new_bio, REQ_OP_READ, 0);
-
- rqd->bio = new_bio;
- rqd->nr_ppas = nr_holes;
- rqd->flags = pblk_set_read_mode(pblk, PBLK_READ_RANDOM);
-
- if (unlikely(nr_holes == 1)) {
- ppa_ptr = rqd->ppa_list;
- dma_ppa_list = rqd->dma_ppa_list;
- rqd->ppa_addr = rqd->ppa_list[0];
- }
-
- ret = pblk_submit_io_sync(pblk, rqd);
- if (ret) {
- bio_put(rqd->bio);
- pblk_err(pblk, "sync read IO submission failed\n");
- goto fail;
- }
-
- if (rqd->error) {
- atomic_long_inc(>read_failed);
-#ifdef CONFIG_NVM_PBLK_DEBUG
- pblk_print_failed_rqd(pblk, rqd, rqd->error);
-#endif
- }
+ __le64 *lba_list_mem, *lba_list_media;
+ void *src_p, *dst_p;
+ int hole, i;
if (unlikely(nr_holes == 1)) {
struct ppa_addr ppa;
ppa = rqd->ppa_addr;
- rqd->ppa_list = ppa_ptr;
- rqd->dma_ppa_list = dma_ppa_list;
+ rqd->ppa_list = pr_ctx->ppa_ptr;
+ rqd->dma_ppa_list = pr_ctx->dma_ppa_list;
rqd->ppa_list[0] = ppa;
}
+ /* Re-use allocated memory for intermediate lbas */
+ lba_list_mem = (((void *)rqd->ppa_list) + pblk_dma_ppa_size);
+ lba_list_media = (((void *)rqd->ppa_list) + 2 * pblk_dma_ppa_size);
+
for (i = 0; i < nr_secs; i++) {
lba_list_media[i] = meta_list[i].lba;
meta_list[i].lba = lba_list_mem[i];
@@ -316,7 +278,7 @@ static int pblk_partial_read(struct pblk *pblk,
struct nvm_rq *rqd,
meta_list[hole].lba = lba_list_media[i];
src_bv = new_bio->bi_io_vec[i++];
- dst_bv = orig_bio->bi_io_vec[bio_init_idx + hole];
+ dst_bv = bio->bi_io_vec[bio_init_idx + hole];
src_p = kmap_atomic(src_bv.bv_page);
dst_p = kmap_atomic(dst_bv.bv_page);
@@ -334,19 +296,107 @@ static int pblk_partial_read(struct pblk *pblk,
struct nvm_rq *rqd,
} while (hole < nr_secs);
bio_put(new_bio);
+ kfree(pr_ctx);
/* restore original request */
rqd->bio = NULL;
rqd->nr_ppas = nr_secs;
+ bio_endio(bio);
__pblk_end_io_read(pblk, rqd, false);
- return NVM_IO_DONE;
+}
-fail:
- /* Free allocated pages in new bio */
+static int pblk_setup_partial_read(struct pblk *pblk, struct nvm_rq
*rqd,
+ unsigned int bio_init_idx,
+ unsigned long *read_bitmap,
+ int nr_holes)
+{
+ struct pblk_sec_meta *meta_list = rqd->meta_list;
+ struct pblk_g_ctx *r_ctx
Re: [PATCH v3] lightnvm: pblk: add asynchronous partial read
On 07/06/2018 12:12 PM, Heiner Litz wrote:
In the read path, partial reads are currently performed synchronously
which affects performance for workloads that generate many partial
reads. This patch adds an asynchronous partial read path as well as
the required partial read ctx.
Signed-off-by: Heiner Litz
---
v3: rebase to head, incorporate 64-bit read bitmap
---
drivers/lightnvm/pblk-read.c | 183 ---
drivers/lightnvm/pblk.h | 10 +++
2 files changed, 130 insertions(+), 63 deletions(-)
diff --git a/drivers/lightnvm/pblk-read.c b/drivers/lightnvm/pblk-read.c
index 9c9362b..4a44076 100644
--- a/drivers/lightnvm/pblk-read.c
+++ b/drivers/lightnvm/pblk-read.c
@@ -231,74 +231,36 @@ static void pblk_end_io_read(struct nvm_rq *rqd)
__pblk_end_io_read(pblk, rqd, true);
}
-static int pblk_partial_read(struct pblk *pblk, struct nvm_rq *rqd,
-struct bio *orig_bio, unsigned int bio_init_idx,
-unsigned long *read_bitmap)
+static void pblk_end_partial_read(struct nvm_rq *rqd)
{
- struct pblk_sec_meta *meta_list = rqd->meta_list;
- struct bio *new_bio;
+ struct pblk *pblk = rqd->private;
+ struct pblk_g_ctx *r_ctx = nvm_rq_to_pdu(rqd);
+ struct pblk_pr_ctx *pr_ctx = r_ctx->private;
+ struct bio *new_bio = rqd->bio;
+ struct bio *bio = pr_ctx->orig_bio;
struct bio_vec src_bv, dst_bv;
- void *ppa_ptr = NULL;
- void *src_p, *dst_p;
- dma_addr_t dma_ppa_list = 0;
- __le64 *lba_list_mem, *lba_list_media;
- int nr_secs = rqd->nr_ppas;
+ struct pblk_sec_meta *meta_list = rqd->meta_list;
+ int bio_init_idx = pr_ctx->bio_init_idx;
+ unsigned long *read_bitmap = _ctx->bitmap;
+ int nr_secs = pr_ctx->orig_nr_secs;
int nr_holes = nr_secs - bitmap_weight(read_bitmap, nr_secs);
- int i, ret, hole;
-
- /* Re-use allocated memory for intermediate lbas */
- lba_list_mem = (((void *)rqd->ppa_list) + pblk_dma_ppa_size);
- lba_list_media = (((void *)rqd->ppa_list) + 2 * pblk_dma_ppa_size);
-
- new_bio = bio_alloc(GFP_KERNEL, nr_holes);
-
- if (pblk_bio_add_pages(pblk, new_bio, GFP_KERNEL, nr_holes))
- goto fail_add_pages;
-
- if (nr_holes != new_bio->bi_vcnt) {
- pblk_err(pblk, "malformed bio\n");
- goto fail;
- }
-
- for (i = 0; i < nr_secs; i++)
- lba_list_mem[i] = meta_list[i].lba;
-
- new_bio->bi_iter.bi_sector = 0; /* internal bio */
- bio_set_op_attrs(new_bio, REQ_OP_READ, 0);
-
- rqd->bio = new_bio;
- rqd->nr_ppas = nr_holes;
- rqd->flags = pblk_set_read_mode(pblk, PBLK_READ_RANDOM);
-
- if (unlikely(nr_holes == 1)) {
- ppa_ptr = rqd->ppa_list;
- dma_ppa_list = rqd->dma_ppa_list;
- rqd->ppa_addr = rqd->ppa_list[0];
- }
-
- ret = pblk_submit_io_sync(pblk, rqd);
- if (ret) {
- bio_put(rqd->bio);
- pblk_err(pblk, "sync read IO submission failed\n");
- goto fail;
- }
-
- if (rqd->error) {
- atomic_long_inc(>read_failed);
-#ifdef CONFIG_NVM_PBLK_DEBUG
- pblk_print_failed_rqd(pblk, rqd, rqd->error);
-#endif
- }
+ __le64 *lba_list_mem, *lba_list_media;
+ void *src_p, *dst_p;
+ int hole, i;
if (unlikely(nr_holes == 1)) {
struct ppa_addr ppa;
ppa = rqd->ppa_addr;
- rqd->ppa_list = ppa_ptr;
- rqd->dma_ppa_list = dma_ppa_list;
+ rqd->ppa_list = pr_ctx->ppa_ptr;
+ rqd->dma_ppa_list = pr_ctx->dma_ppa_list;
rqd->ppa_list[0] = ppa;
}
+ /* Re-use allocated memory for intermediate lbas */
+ lba_list_mem = (((void *)rqd->ppa_list) + pblk_dma_ppa_size);
+ lba_list_media = (((void *)rqd->ppa_list) + 2 * pblk_dma_ppa_size);
+
for (i = 0; i < nr_secs; i++) {
lba_list_media[i] = meta_list[i].lba;
meta_list[i].lba = lba_list_mem[i];
@@ -316,7 +278,7 @@ static int pblk_partial_read(struct pblk *pblk, struct
nvm_rq *rqd,
meta_list[hole].lba = lba_list_media[i];
src_bv = new_bio->bi_io_vec[i++];
- dst_bv = orig_bio->bi_io_vec[bio_init_idx + hole];
+ dst_bv = bio->bi_io_vec[bio_init_idx + hole];
src_p = kmap_atomic(src_bv.bv_page);
dst_p = kmap_atomic(dst_bv.bv_page);
@@ -334,19 +296,107 @@ static int pblk_partial_read(struct pblk *pblk, struct
nvm_rq *rqd,
} while (hole < nr_secs);
bio_put(new_bio);
+ kfree(pr_ctx);
/* restore original request */
rqd->bio = NULL;
rqd->nr_ppas = nr_secs;
+ bio_endio(bio);
__pblk_end_io_read(pblk, rqd, false);
- return NVM_IO_DONE;
+}
-fail:
- /* Free allocated pages in new
[PATCH v3] lightnvm: pblk: add asynchronous partial read
In the read path, partial reads are currently performed synchronously
which affects performance for workloads that generate many partial
reads. This patch adds an asynchronous partial read path as well as
the required partial read ctx.
Signed-off-by: Heiner Litz
---
v3: rebase to head, incorporate 64-bit read bitmap
---
drivers/lightnvm/pblk-read.c | 183 ---
drivers/lightnvm/pblk.h | 10 +++
2 files changed, 130 insertions(+), 63 deletions(-)
diff --git a/drivers/lightnvm/pblk-read.c b/drivers/lightnvm/pblk-read.c
index 9c9362b..4a44076 100644
--- a/drivers/lightnvm/pblk-read.c
+++ b/drivers/lightnvm/pblk-read.c
@@ -231,74 +231,36 @@ static void pblk_end_io_read(struct nvm_rq *rqd)
__pblk_end_io_read(pblk, rqd, true);
}
-static int pblk_partial_read(struct pblk *pblk, struct nvm_rq *rqd,
-struct bio *orig_bio, unsigned int bio_init_idx,
-unsigned long *read_bitmap)
+static void pblk_end_partial_read(struct nvm_rq *rqd)
{
- struct pblk_sec_meta *meta_list = rqd->meta_list;
- struct bio *new_bio;
+ struct pblk *pblk = rqd->private;
+ struct pblk_g_ctx *r_ctx = nvm_rq_to_pdu(rqd);
+ struct pblk_pr_ctx *pr_ctx = r_ctx->private;
+ struct bio *new_bio = rqd->bio;
+ struct bio *bio = pr_ctx->orig_bio;
struct bio_vec src_bv, dst_bv;
- void *ppa_ptr = NULL;
- void *src_p, *dst_p;
- dma_addr_t dma_ppa_list = 0;
- __le64 *lba_list_mem, *lba_list_media;
- int nr_secs = rqd->nr_ppas;
+ struct pblk_sec_meta *meta_list = rqd->meta_list;
+ int bio_init_idx = pr_ctx->bio_init_idx;
+ unsigned long *read_bitmap = _ctx->bitmap;
+ int nr_secs = pr_ctx->orig_nr_secs;
int nr_holes = nr_secs - bitmap_weight(read_bitmap, nr_secs);
- int i, ret, hole;
-
- /* Re-use allocated memory for intermediate lbas */
- lba_list_mem = (((void *)rqd->ppa_list) + pblk_dma_ppa_size);
- lba_list_media = (((void *)rqd->ppa_list) + 2 * pblk_dma_ppa_size);
-
- new_bio = bio_alloc(GFP_KERNEL, nr_holes);
-
- if (pblk_bio_add_pages(pblk, new_bio, GFP_KERNEL, nr_holes))
- goto fail_add_pages;
-
- if (nr_holes != new_bio->bi_vcnt) {
- pblk_err(pblk, "malformed bio\n");
- goto fail;
- }
-
- for (i = 0; i < nr_secs; i++)
- lba_list_mem[i] = meta_list[i].lba;
-
- new_bio->bi_iter.bi_sector = 0; /* internal bio */
- bio_set_op_attrs(new_bio, REQ_OP_READ, 0);
-
- rqd->bio = new_bio;
- rqd->nr_ppas = nr_holes;
- rqd->flags = pblk_set_read_mode(pblk, PBLK_READ_RANDOM);
-
- if (unlikely(nr_holes == 1)) {
- ppa_ptr = rqd->ppa_list;
- dma_ppa_list = rqd->dma_ppa_list;
- rqd->ppa_addr = rqd->ppa_list[0];
- }
-
- ret = pblk_submit_io_sync(pblk, rqd);
- if (ret) {
- bio_put(rqd->bio);
- pblk_err(pblk, "sync read IO submission failed\n");
- goto fail;
- }
-
- if (rqd->error) {
- atomic_long_inc(>read_failed);
-#ifdef CONFIG_NVM_PBLK_DEBUG
- pblk_print_failed_rqd(pblk, rqd, rqd->error);
-#endif
- }
+ __le64 *lba_list_mem, *lba_list_media;
+ void *src_p, *dst_p;
+ int hole, i;
if (unlikely(nr_holes == 1)) {
struct ppa_addr ppa;
ppa = rqd->ppa_addr;
- rqd->ppa_list = ppa_ptr;
- rqd->dma_ppa_list = dma_ppa_list;
+ rqd->ppa_list = pr_ctx->ppa_ptr;
+ rqd->dma_ppa_list = pr_ctx->dma_ppa_list;
rqd->ppa_list[0] = ppa;
}
+ /* Re-use allocated memory for intermediate lbas */
+ lba_list_mem = (((void *)rqd->ppa_list) + pblk_dma_ppa_size);
+ lba_list_media = (((void *)rqd->ppa_list) + 2 * pblk_dma_ppa_size);
+
for (i = 0; i < nr_secs; i++) {
lba_list_media[i] = meta_list[i].lba;
meta_list[i].lba = lba_list_mem[i];
@@ -316,7 +278,7 @@ static int pblk_partial_read(struct pblk *pblk, struct
nvm_rq *rqd,
meta_list[hole].lba = lba_list_media[i];
src_bv = new_bio->bi_io_vec[i++];
- dst_bv = orig_bio->bi_io_vec[bio_init_idx + hole];
+ dst_bv = bio->bi_io_vec[bio_init_idx + hole];
src_p = kmap_atomic(src_bv.bv_page);
dst_p = kmap_atomic(dst_bv.bv_page);
@@ -334,19 +296,107 @@ static int pblk_partial_read(struct pblk *pblk, struct
nvm_rq *rqd,
} while (hole < nr_secs);
bio_put(new_bio);
+ kfree(pr_ctx);
/* restore original request */
rqd->bio = NULL;
rqd->nr_ppas = nr_secs;
+ bio_endio(bio);
__pblk_end_io_read(pblk, rqd, false);
- return NVM_IO_DONE;
+}
-fail:
- /* Free allocated pages in new bio */
