2019-03-07 12:19:53

by Javier González

Subject: [PATCH] pblk: fix max_io calculation

When calculating the maximun I/O size allowed into the buffer, consider
the write size (ws_opt) used by the write thread in order to cover the
case in which, due to flushes, the mem and subm pointers are disaligned
by (ws_opt - 1). This case currently translates into a stall when
an I/O of the largest possible size is submitted.

Fixes: f9f9d1ae2c66 ("lightnvm: pblk: prevent stall due to wb threshold")

Signed-off-by: Javier González <[email protected]>
---

Matias: Can you apply this as a fix to 5.1. This is a case I missed when fixing
the wb threshold, which is also scheduled for 5.1

Thanks,
Javier

drivers/lightnvm/pblk-rl.c | 7 ++++++-
1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/drivers/lightnvm/pblk-rl.c b/drivers/lightnvm/pblk-rl.c
index b014957dde0b..a5f8bc2defbc 100644
--- a/drivers/lightnvm/pblk-rl.c
+++ b/drivers/lightnvm/pblk-rl.c
@@ -233,10 +233,15 @@ void pblk_rl_init(struct pblk_rl *rl, int budget, int threshold)
/* To start with, all buffer is available to user I/O writers */
rl->rb_budget = budget;
rl->rb_user_max = budget;
- rl->rb_max_io = threshold ? (budget - threshold) : (budget - 1);
rl->rb_gc_max = 0;
rl->rb_state = PBLK_RL_HIGH;

+ /* Maximize I/O size and ansure that back threshold is respected */
+ if (threshold)
+ rl->rb_max_io = budget - pblk->min_write_pgs_data - threshold;
+ else
+ rl->rb_max_io = budget - pblk->min_write_pgs_data - 1;
+
atomic_set(&rl->rb_user_cnt, 0);
atomic_set(&rl->rb_gc_cnt, 0);
atomic_set(&rl->rb_space, -1);
--
2.17.1

2019-03-07 15:58:48

by Matias Bjørling

Subject: Re: [PATCH] pblk: fix max_io calculation

On 3/7/19 1:18 PM, Javier González wrote:
> When calculating the maximun I/O size allowed into the buffer, consider
> the write size (ws_opt) used by the write thread in order to cover the
> case in which, due to flushes, the mem and subm pointers are disaligned
> by (ws_opt - 1). This case currently translates into a stall when
> an I/O of the largest possible size is submitted.
>
> Fixes: f9f9d1ae2c66 ("lightnvm: pblk: prevent stall due to wb threshold")
>
> Signed-off-by: Javier González <[email protected]>
> ---
>
> Matias: Can you apply this as a fix to 5.1. This is a case I missed when fixing
> the wb threshold, which is also scheduled for 5.1
>
> Thanks,
> Javier
>
> drivers/lightnvm/pblk-rl.c | 7 ++++++-
> 1 file changed, 6 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/lightnvm/pblk-rl.c b/drivers/lightnvm/pblk-rl.c
> index b014957dde0b..a5f8bc2defbc 100644
> --- a/drivers/lightnvm/pblk-rl.c
> +++ b/drivers/lightnvm/pblk-rl.c
> @@ -233,10 +233,15 @@ void pblk_rl_init(struct pblk_rl *rl, int budget, int threshold)
> /* To start with, all buffer is available to user I/O writers */
> rl->rb_budget = budget;
> rl->rb_user_max = budget;
> - rl->rb_max_io = threshold ? (budget - threshold) : (budget - 1);
> rl->rb_gc_max = 0;
> rl->rb_state = PBLK_RL_HIGH;
>
> + /* Maximize I/O size and ansure that back threshold is respected */
> + if (threshold)
> + rl->rb_max_io = budget - pblk->min_write_pgs_data - threshold;
> + else
> + rl->rb_max_io = budget - pblk->min_write_pgs_data - 1;
> +
> atomic_set(&rl->rb_user_cnt, 0);
> atomic_set(&rl->rb_gc_cnt, 0);
> atomic_set(&rl->rb_space, -1);
>

Hi Jens,

If possible, could you please pick this one up for 5.1? It fixes a
previous patch that was introduced in 5.1 that should fix a stall, but
didn't quite catch it.

Thank you,
-Matias

2019-03-07 16:02:41

by Jens Axboe

Subject: Re: [PATCH] pblk: fix max_io calculation

On 3/7/19 8:57 AM, Matias Bjørling wrote:
> On 3/7/19 1:18 PM, Javier González wrote:
>> When calculating the maximun I/O size allowed into the buffer, consider
>> the write size (ws_opt) used by the write thread in order to cover the
>> case in which, due to flushes, the mem and subm pointers are disaligned
>> by (ws_opt - 1). This case currently translates into a stall when
>> an I/O of the largest possible size is submitted.
>>
>> Fixes: f9f9d1ae2c66 ("lightnvm: pblk: prevent stall due to wb threshold")
>>
>> Signed-off-by: Javier González <[email protected]>
>> ---
>>
>> Matias: Can you apply this as a fix to 5.1. This is a case I missed when fixing
>> the wb threshold, which is also scheduled for 5.1
>>
>> Thanks,
>> Javier
>>
>> drivers/lightnvm/pblk-rl.c | 7 ++++++-
>> 1 file changed, 6 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/lightnvm/pblk-rl.c b/drivers/lightnvm/pblk-rl.c
>> index b014957dde0b..a5f8bc2defbc 100644
>> --- a/drivers/lightnvm/pblk-rl.c
>> +++ b/drivers/lightnvm/pblk-rl.c
>> @@ -233,10 +233,15 @@ void pblk_rl_init(struct pblk_rl *rl, int budget, int threshold)
>> /* To start with, all buffer is available to user I/O writers */
>> rl->rb_budget = budget;
>> rl->rb_user_max = budget;
>> - rl->rb_max_io = threshold ? (budget - threshold) : (budget - 1);
>> rl->rb_gc_max = 0;
>> rl->rb_state = PBLK_RL_HIGH;
>>
>> + /* Maximize I/O size and ansure that back threshold is respected */
>> + if (threshold)
>> + rl->rb_max_io = budget - pblk->min_write_pgs_data - threshold;
>> + else
>> + rl->rb_max_io = budget - pblk->min_write_pgs_data - 1;
>> +
>> atomic_set(&rl->rb_user_cnt, 0);
>> atomic_set(&rl->rb_gc_cnt, 0);
>> atomic_set(&rl->rb_space, -1);
>>
>
> Hi Jens,
>
> If possible, could you please pick this one up for 5.1? It fixes a
> previous patch that was introduced in 5.1 that should fix a stall, but
> didn't quite catch it.

Applied

--
Jens Axboe