2018-08-17 10:23:38

by Javier González

Subject: [PATCH] lightnvm: pblk: fix race condition on metadata I/O

In pblk, when a new line is allocated, metadata for the previously
written line is scheduled. This is done through a fixed memory region
that is shared through time and contexts across different lines and
therefore protected by a lock. Unfortunately, this lock is not properly
covering all the metadata used for sharing this memory regions,
resulting on a race condition.

This patch fixes this race condition by protecting this metadata
properly.

Fixes: dd2a43437337 ("lightnvm: pblk: sched. metadata on write thread")

Signed-off-by: Javier González <[email protected]>
---
drivers/lightnvm/pblk-write.c | 14 +++++++-------
1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/drivers/lightnvm/pblk-write.c b/drivers/lightnvm/pblk-write.c
index 508c63701eda..df99c45778d4 100644
--- a/drivers/lightnvm/pblk-write.c
+++ b/drivers/lightnvm/pblk-write.c
@@ -417,12 +417,11 @@ int pblk_submit_meta_io(struct pblk *pblk, struct pblk_line *meta_line)
rqd->ppa_list[i] = addr_to_gen_ppa(pblk, paddr, id);
}

+ spin_lock(&l_mg->close_lock);
emeta->mem += rq_len;
- if (emeta->mem >= lm->emeta_len[0]) {
- spin_lock(&l_mg->close_lock);
+ if (emeta->mem >= lm->emeta_len[0])
list_del(&meta_line->list);
- spin_unlock(&l_mg->close_lock);
- }
+ spin_unlock(&l_mg->close_lock);

pblk_down_page(pblk, rqd->ppa_list, rqd->nr_ppas);

@@ -491,14 +490,15 @@ static struct pblk_line *pblk_should_submit_meta_io(struct pblk *pblk,
struct pblk_line *meta_line;

spin_lock(&l_mg->close_lock);
-retry:
if (list_empty(&l_mg->emeta_list)) {
spin_unlock(&l_mg->close_lock);
return NULL;
}
meta_line = list_first_entry(&l_mg->emeta_list, struct pblk_line, list);
- if (meta_line->emeta->mem >= lm->emeta_len[0])
- goto retry;
+ if (meta_line->emeta->mem >= lm->emeta_len[0]) {
+ spin_unlock(&l_mg->close_lock);
+ return NULL;
+ }
spin_unlock(&l_mg->close_lock);

if (!pblk_valid_meta_ppa(pblk, meta_line, data_rqd))
--
2.7.4

2018-08-17 10:38:42

by Matias Bjørling

Subject: Re: [PATCH] lightnvm: pblk: fix race condition on metadata I/O

On 08/17/2018 12:22 PM, Javier González wrote:
> In pblk, when a new line is allocated, metadata for the previously
> written line is scheduled. This is done through a fixed memory region
> that is shared through time and contexts across different lines and
> therefore protected by a lock. Unfortunately, this lock is not properly
> covering all the metadata used for sharing this memory regions,
> resulting on a race condition.
>
> This patch fixes this race condition by protecting this metadata
> properly.
>
> Fixes: dd2a43437337 ("lightnvm: pblk: sched. metadata on write thread")
>
> Signed-off-by: Javier González <[email protected]>
> ---
> drivers/lightnvm/pblk-write.c | 14 +++++++-------
> 1 file changed, 7 insertions(+), 7 deletions(-)
>
> diff --git a/drivers/lightnvm/pblk-write.c b/drivers/lightnvm/pblk-write.c
> index 508c63701eda..df99c45778d4 100644
> --- a/drivers/lightnvm/pblk-write.c
> +++ b/drivers/lightnvm/pblk-write.c
> @@ -417,12 +417,11 @@ int pblk_submit_meta_io(struct pblk *pblk, struct pblk_line *meta_line)
> rqd->ppa_list[i] = addr_to_gen_ppa(pblk, paddr, id);
> }
>
> + spin_lock(&l_mg->close_lock);
> emeta->mem += rq_len;
> - if (emeta->mem >= lm->emeta_len[0]) {
> - spin_lock(&l_mg->close_lock);
> + if (emeta->mem >= lm->emeta_len[0])
> list_del(&meta_line->list);
> - spin_unlock(&l_mg->close_lock);
> - }
> + spin_unlock(&l_mg->close_lock);
>
> pblk_down_page(pblk, rqd->ppa_list, rqd->nr_ppas);
>
> @@ -491,14 +490,15 @@ static struct pblk_line *pblk_should_submit_meta_io(struct pblk *pblk,
> struct pblk_line *meta_line;
>
> spin_lock(&l_mg->close_lock);
> -retry:
> if (list_empty(&l_mg->emeta_list)) {
> spin_unlock(&l_mg->close_lock);
> return NULL;
> }
> meta_line = list_first_entry(&l_mg->emeta_list, struct pblk_line, list);
> - if (meta_line->emeta->mem >= lm->emeta_len[0])
> - goto retry;
> + if (meta_line->emeta->mem >= lm->emeta_len[0]) {
> + spin_unlock(&l_mg->close_lock);
> + return NULL;
> + } > spin_unlock(&l_mg->close_lock);
>
> if (!pblk_valid_meta_ppa(pblk, meta_line, data_rqd))
>

Thanks. Applied for 4.20.