2017-09-26 03:41:45

by Masahiro Yamada

Subject: [PATCH] mtd: nand: wait for tWHR after NAND_CMD_STATUS / NAND_CMD_READID

Read Status and Read ID require tWHR before reading the first data.
Insert a very short wait to make sure to meet the spec.

I have not seen any problem report for now, but nand_command() and
nand_command_lP() are generic hooks, so it makes sense to implement
fail-safe code here.

Signed-off-by: Masahiro Yamada <[email protected]>
---

drivers/mtd/nand/nand_base.c | 10 ++++++++--
1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/drivers/mtd/nand/nand_base.c b/drivers/mtd/nand/nand_base.c
index b1cf32c..55c0ba5 100644
--- a/drivers/mtd/nand/nand_base.c
+++ b/drivers/mtd/nand/nand_base.c
@@ -742,9 +742,12 @@ static void nand_command(struct mtd_info *mtd, unsigned int command,
case NAND_CMD_ERASE1:
case NAND_CMD_ERASE2:
case NAND_CMD_SEQIN:
+ case NAND_CMD_SET_FEATURES:
+ return;
+
case NAND_CMD_STATUS:
case NAND_CMD_READID:
- case NAND_CMD_SET_FEATURES:
+ ndelay(200); /* tWHR */
return;

case NAND_CMD_RESET:
@@ -871,9 +874,12 @@ static void nand_command_lp(struct mtd_info *mtd, unsigned int command,
case NAND_CMD_ERASE1:
case NAND_CMD_ERASE2:
case NAND_CMD_SEQIN:
+ case NAND_CMD_SET_FEATURES:
+ return;
+
case NAND_CMD_STATUS:
case NAND_CMD_READID:
- case NAND_CMD_SET_FEATURES:
+ ndelay(200); /* tWHR */
return;

case NAND_CMD_RNDIN:
--
2.7.4

2017-09-26 06:44:07

by Boris Brezillon

Subject: Re: [PATCH] mtd: nand: wait for tWHR after NAND_CMD_STATUS / NAND_CMD_READID

On Tue, 26 Sep 2017 12:39:24 +0900
Masahiro Yamada <[email protected]> wrote:

> Read Status and Read ID require tWHR before reading the first data.
> Insert a very short wait to make sure to meet the spec.
>
> I have not seen any problem report for now, but nand_command() and
> nand_command_lP() are generic hooks, so it makes sense to implement
> fail-safe code here.
>
> Signed-off-by: Masahiro Yamada <[email protected]>
> ---
>
> drivers/mtd/nand/nand_base.c | 10 ++++++++--
> 1 file changed, 8 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/mtd/nand/nand_base.c b/drivers/mtd/nand/nand_base.c
> index b1cf32c..55c0ba5 100644
> --- a/drivers/mtd/nand/nand_base.c
> +++ b/drivers/mtd/nand/nand_base.c
> @@ -742,9 +742,12 @@ static void nand_command(struct mtd_info *mtd, unsigned int command,
> case NAND_CMD_ERASE1:
> case NAND_CMD_ERASE2:
> case NAND_CMD_SEQIN:
> + case NAND_CMD_SET_FEATURES:
> + return;
> +
> case NAND_CMD_STATUS:
> case NAND_CMD_READID:
> - case NAND_CMD_SET_FEATURES:
> + ndelay(200); /* tWHR */
> return;
>
> case NAND_CMD_RESET:
> @@ -871,9 +874,12 @@ static void nand_command_lp(struct mtd_info *mtd, unsigned int command,
> case NAND_CMD_ERASE1:
> case NAND_CMD_ERASE2:
> case NAND_CMD_SEQIN:
> + case NAND_CMD_SET_FEATURES:
> + return;
> +
> case NAND_CMD_STATUS:
> case NAND_CMD_READID:
> - case NAND_CMD_SET_FEATURES:
> + ndelay(200); /* tWHR */

Can you do something like nand_ccs_delay() [1] instead of
unconditionally adding this 200ns delay here? I'm not worried about the
overhead introduced by this ndelay(), but I don't want to risk
introducing a regression.


> return;
>
> case NAND_CMD_RNDIN:

[1]http://elixir.free-electrons.com/linux/v4.14-rc2/source/drivers/mtd/nand/nand_base.c#L793

2017-09-26 08:18:45

by Masahiro Yamada

Subject: Re: [PATCH] mtd: nand: wait for tWHR after NAND_CMD_STATUS / NAND_CMD_READID

2017-09-26 15:43 GMT+09:00 Boris Brezillon <[email protected]>:
> On Tue, 26 Sep 2017 12:39:24 +0900
> Masahiro Yamada <[email protected]> wrote:
>
>> Read Status and Read ID require tWHR before reading the first data.
>> Insert a very short wait to make sure to meet the spec.
>>
>> I have not seen any problem report for now, but nand_command() and
>> nand_command_lP() are generic hooks, so it makes sense to implement
>> fail-safe code here.
>>
>> Signed-off-by: Masahiro Yamada <[email protected]>
>> ---
>


OK, will do.



BTW, I see unconditional wait for tWB a few lines below,
but it gives no performance regression because we will wait
much longer in nand_wait_ready().


/*
* Apply this short delay always to ensure that we do wait tWB in
* any case on any machine.
*/
ndelay(100);

nand_wait_ready(mtd);




--
Best Regards
Masahiro Yamada

2017-09-26 08:23:49

by Boris Brezillon

Subject: Re: [PATCH] mtd: nand: wait for tWHR after NAND_CMD_STATUS / NAND_CMD_READID

On Tue, 26 Sep 2017 17:17:49 +0900
Masahiro Yamada <[email protected]> wrote:

> 2017-09-26 15:43 GMT+09:00 Boris Brezillon <[email protected]>:
> > On Tue, 26 Sep 2017 12:39:24 +0900
> > Masahiro Yamada <[email protected]> wrote:
> >
> >> Read Status and Read ID require tWHR before reading the first data.
> >> Insert a very short wait to make sure to meet the spec.
> >>
> >> I have not seen any problem report for now, but nand_command() and
> >> nand_command_lP() are generic hooks, so it makes sense to implement
> >> fail-safe code here.
> >>
> >> Signed-off-by: Masahiro Yamada <[email protected]>
> >> ---
> >
>
>
> OK, will do.
>
>
>
> BTW, I see unconditional wait for tWB a few lines below,
> but it gives no performance regression because we will wait
> much longer in nand_wait_ready().

Yep, but this one is here for quite some time. Normally there should be
no problem with your ndelay(200), but I don't want to take the risk
and have someone complain that his NAND controller driver is broken
because of this extra delay ;-).

>
>
> /*
> * Apply this short delay always to ensure that we do wait tWB in
> * any case on any machine.
> */
> ndelay(100);
>
> nand_wait_ready(mtd);
>
>
>
>