| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-Id: <1504695071-25928-17-git-send-email-javier@cnexlabs.com>
Date: Wed, 6 Sep 2017 12:51:09 +0200
From: "Javier González" <jg@...htnvm.io>
To: mb@...htnvm.io, axboe@...com
Cc: linux-block@...r.kernel.org, linux-kernel@...r.kernel.org,
Javier González <javier@...xlabs.com>,
Matias Bjørling <matias@...xlabs.com>
Subject: [PATCH 16/18] lightnvm: pblk: enable 1 LUN configuration
Metadata I/Os are scheduled to minimize their impact on user data I/Os.
When there are enough LUNs instantiated (i.e., enought bandwidth), it is
easy to interleave metadata and data one after the other so that
metadata I/Os are the ones being blocked and not viceversa.
We do this by calculating the distance between the I/Os in terms of the
LUNs that are not in used, and selecting a free LUN that satisfies a
the simple heuristic that medatata is scheduled behind. The per-LUN
semaphores guarantee consistency. This works fine on >1 LUN
configuration. However, when a signle LUN is instantiated, this design
leads to a deadlock, where metadata waits to be scheduled on a free LUN.
This patch implements the 1 LUN case by simply scheduling the medatada
I/O after the data I/O. In the process, we refactor the way a line is
replaced to ensure that metadata writes are submitted after data writes
in order to guarantee block sequentiality. Note that, since there is
only one LUN, both I/Os will block each other by design. However, such
configuration only pursues tight read latencies, not write bandwidth.
Signed-off-by: Javier González <javier@...xlabs.com>
Signed-off-by: Matias Bjørling <matias@...xlabs.com>
---
drivers/lightnvm/pblk-core.c | 17 ++++++++++-------
drivers/lightnvm/pblk-init.c | 8 ++++++--
drivers/lightnvm/pblk-map.c | 21 ++++++++++++---------
drivers/lightnvm/pblk-write.c | 8 +++++---
drivers/lightnvm/pblk.h | 2 +-
5 files changed, 34 insertions(+), 22 deletions(-)
diff --git a/drivers/lightnvm/pblk-core.c b/drivers/lightnvm/pblk-core.c
index 1f746e31e379..873b66200678 100644
--- a/drivers/lightnvm/pblk-core.c
+++ b/drivers/lightnvm/pblk-core.c
@@ -1326,17 +1326,17 @@ void pblk_pipeline_stop(struct pblk *pblk)
spin_unlock(&l_mg->free_lock);
}
-void pblk_line_replace_data(struct pblk *pblk)
+struct pblk_line *pblk_line_replace_data(struct pblk *pblk)
{
struct pblk_line_mgmt *l_mg = &pblk->l_mg;
- struct pblk_line *cur, *new;
+ struct pblk_line *cur, *new = NULL;
unsigned int left_seblks;
int is_next = 0;
cur = l_mg->data_line;
new = l_mg->data_next;
if (!new)
- return;
+ goto out;
l_mg->data_line = new;
spin_lock(&l_mg->free_lock);
@@ -1344,7 +1344,7 @@ void pblk_line_replace_data(struct pblk *pblk)
l_mg->data_line = NULL;
l_mg->data_next = NULL;
spin_unlock(&l_mg->free_lock);
- return;
+ goto out;
}
pblk_line_setup_metadata(new, l_mg, &pblk->lm);
@@ -1356,7 +1356,7 @@ void pblk_line_replace_data(struct pblk *pblk)
/* If line is not fully erased, erase it */
if (atomic_read(&new->left_eblks)) {
if (pblk_line_erase(pblk, new))
- return;
+ goto out;
} else {
io_schedule();
}
@@ -1367,7 +1367,7 @@ void pblk_line_replace_data(struct pblk *pblk)
if (!pblk_line_init_metadata(pblk, new, cur)) {
new = pblk_line_retry(pblk, new);
if (!new)
- return;
+ goto out;
goto retry_setup;
}
@@ -1375,7 +1375,7 @@ void pblk_line_replace_data(struct pblk *pblk)
if (!pblk_line_init_bb(pblk, new, 1)) {
new = pblk_line_retry(pblk, new);
if (!new)
- return;
+ goto out;
goto retry_setup;
}
@@ -1399,6 +1399,9 @@ void pblk_line_replace_data(struct pblk *pblk)
if (is_next)
pblk_rl_free_lines_dec(&pblk->rl, l_mg->data_next);
+
+out:
+ return new;
}
void pblk_line_free(struct pblk *pblk, struct pblk_line *line)
diff --git a/drivers/lightnvm/pblk-init.c b/drivers/lightnvm/pblk-init.c
index 12c05aebac16..0409839cc8fc 100644
--- a/drivers/lightnvm/pblk-init.c
+++ b/drivers/lightnvm/pblk-init.c
@@ -714,8 +714,12 @@ static int pblk_lines_init(struct pblk *pblk)
}
lm->emeta_bb = geo->nr_luns - i;
- lm->min_blk_line = 1 + DIV_ROUND_UP(lm->smeta_sec + lm->emeta_sec[0],
- geo->sec_per_blk);
+
+ lm->min_blk_line = 1;
+ if (geo->nr_luns > 1)
+ lm->min_blk_line += DIV_ROUND_UP(lm->smeta_sec +
+ lm->emeta_sec[0], geo->sec_per_blk);
+
if (lm->min_blk_line > lm->blk_per_line) {
pr_err("pblk: config. not supported. Min. LUN in line:%d\n",
lm->blk_per_line);
diff --git a/drivers/lightnvm/pblk-map.c b/drivers/lightnvm/pblk-map.c
index fddb924f6dde..3bc4c94f9cf2 100644
--- a/drivers/lightnvm/pblk-map.c
+++ b/drivers/lightnvm/pblk-map.c
@@ -25,13 +25,23 @@ static void pblk_map_page_data(struct pblk *pblk, unsigned int sentry,
unsigned int valid_secs)
{
struct pblk_line *line = pblk_line_get_data(pblk);
- struct pblk_emeta *emeta = line->emeta;
+ struct pblk_emeta *emeta;
struct pblk_w_ctx *w_ctx;
- __le64 *lba_list = emeta_to_lbas(pblk, emeta->buf);
+ __le64 *lba_list;
u64 paddr;
int nr_secs = pblk->min_write_pgs;
int i;
+ if (pblk_line_is_full(line)) {
+ struct pblk_line *prev_line = line;
+
+ line = pblk_line_replace_data(pblk);
+ pblk_line_close_meta(pblk, prev_line);
+ }
+
+ emeta = line->emeta;
+ lba_list = emeta_to_lbas(pblk, emeta->buf);
+
paddr = pblk_alloc_page(pblk, line, nr_secs);
for (i = 0; i < nr_secs; i++, paddr++) {
@@ -60,13 +70,6 @@ static void pblk_map_page_data(struct pblk *pblk, unsigned int sentry,
}
}
- if (pblk_line_is_full(line)) {
- struct pblk_line *prev_line = line;
-
- pblk_line_replace_data(pblk);
- pblk_line_close_meta(pblk, prev_line);
- }
-
pblk_down_rq(pblk, ppa_list, nr_secs, lun_bitmap);
}
diff --git a/drivers/lightnvm/pblk-write.c b/drivers/lightnvm/pblk-write.c
index fb4d9c3983db..73300c059784 100644
--- a/drivers/lightnvm/pblk-write.c
+++ b/drivers/lightnvm/pblk-write.c
@@ -188,7 +188,8 @@ static void pblk_end_io_write_meta(struct nvm_rq *rqd)
struct pblk_emeta *emeta = line->emeta;
int sync;
- pblk_up_page(pblk, rqd->ppa_list, rqd->nr_ppas);
+ if (dev->geo.nr_luns > 1)
+ pblk_up_page(pblk, rqd->ppa_list, rqd->nr_ppas);
if (rqd->error) {
pblk_log_write_err(pblk, rqd);
@@ -332,7 +333,7 @@ static inline int pblk_valid_meta_ppa(struct pblk *pblk,
* the distance to not be optimal, but allow metadata I/Os to succeed.
*/
ppa_opt = addr_to_gen_ppa(pblk, paddr + data_line->meta_distance, 0);
- if (unlikely(ppa_opt.ppa == ppa.ppa)) {
+ if (geo->nr_luns > 1 && unlikely(ppa_opt.ppa == ppa.ppa)) {
data_line->meta_distance--;
return 0;
}
@@ -414,7 +415,8 @@ int pblk_submit_meta_io(struct pblk *pblk, struct pblk_line *meta_line)
spin_unlock(&l_mg->close_lock);
}
- pblk_down_page(pblk, rqd->ppa_list, rqd->nr_ppas);
+ if (geo->nr_luns > 1)
+ pblk_down_page(pblk, rqd->ppa_list, rqd->nr_ppas);
ret = pblk_submit_io(pblk, rqd);
if (ret) {
diff --git a/drivers/lightnvm/pblk.h b/drivers/lightnvm/pblk.h
index 2ae9d94e9852..f6d2e1e72057 100644
--- a/drivers/lightnvm/pblk.h
+++ b/drivers/lightnvm/pblk.h
@@ -716,7 +716,7 @@ struct bio *pblk_bio_map_addr(struct pblk *pblk, void *data,
int alloc_type, gfp_t gfp_mask, int reading);
struct pblk_line *pblk_line_get(struct pblk *pblk);
struct pblk_line *pblk_line_get_first_data(struct pblk *pblk);
-void pblk_line_replace_data(struct pblk *pblk);
+struct pblk_line *pblk_line_replace_data(struct pblk *pblk);
int pblk_line_recov_alloc(struct pblk *pblk, struct pblk_line *line);
void pblk_line_recov_close(struct pblk *pblk, struct pblk_line *line);
struct pblk_line *pblk_line_get_data(struct pblk *pblk);
--
2.7.4
Powered by blists - more mailing lists