mmc: sunxi: Use new timing mode for A64 eMMC controller
diff mbox

Message ID 20180712030225.15681-1-wens@csie.org
State New, archived
Headers show

Commit Message

Chen-Yu Tsai July 12, 2018, 3:02 a.m. UTC
The eMMC controller is also a new timing mode controller, but it doesn't
have the timing mode switch. It does however have signal delay and
calibration controls, typical of Allwinner MMC controllers that support
the new timing mode.

Enable the new timing mode setting for the A64 eMMC controller. This
also enables MMC HS-DDR modes, which gives higher throughput for eMMC
chips that support it, and can deliver such throughput.

Signed-off-by: Chen-Yu Tsai <wens@csie.org>
---
 drivers/mmc/host/sunxi-mmc.c | 1 +
 1 file changed, 1 insertion(+)

Comments

Maxime Ripard July 12, 2018, 7:19 a.m. UTC | #1
Hi,

On Thu, Jul 12, 2018 at 11:02:25AM +0800, Chen-Yu Tsai wrote:
> The eMMC controller is also a new timing mode controller, but it doesn't
> have the timing mode switch. It does however have signal delay and
> calibration controls, typical of Allwinner MMC controllers that support
> the new timing mode.
> 
> Enable the new timing mode setting for the A64 eMMC controller. This
> also enables MMC HS-DDR modes, which gives higher throughput for eMMC
> chips that support it, and can deliver such throughput.
> 
> Signed-off-by: Chen-Yu Tsai <wens@csie.org>

That doesn't look right. The datasheet explicitly mentions that this
bit doesn't apply to the eMMC controller, and the BSP is doing the same:
https://github.com/longsleep/linux-pine64/blob/lichee-dev-v3.10.65/drivers/mmc/host/sunxi-mmc-sun50iw1p1-1.c

vs
https://github.com/longsleep/linux-pine64/blob/lichee-dev-v3.10.65/drivers/mmc/host/sunxi-mmc-sun50iw1p1-2.c

And I definitely remember having HS-DDR working back when I added the
a64 eMMC support.

Maxime
Chen-Yu Tsai July 12, 2018, 10:17 a.m. UTC | #2
On Thu, Jul 12, 2018 at 3:19 PM, Maxime Ripard
<maxime.ripard@bootlin.com> wrote:
> Hi,
>
> On Thu, Jul 12, 2018 at 11:02:25AM +0800, Chen-Yu Tsai wrote:
>> The eMMC controller is also a new timing mode controller, but it doesn't
>> have the timing mode switch. It does however have signal delay and
>> calibration controls, typical of Allwinner MMC controllers that support
>> the new timing mode.
>>
>> Enable the new timing mode setting for the A64 eMMC controller. This
>> also enables MMC HS-DDR modes, which gives higher throughput for eMMC
>> chips that support it, and can deliver such throughput.
>>
>> Signed-off-by: Chen-Yu Tsai <wens@csie.org>
>
> That doesn't look right. The datasheet explicitly mentions that this
> bit doesn't apply to the eMMC controller, and the BSP is doing the same:
> https://github.com/longsleep/linux-pine64/blob/lichee-dev-v3.10.65/drivers/mmc/host/sunxi-mmc-sun50iw1p1-1.c
>
> vs
> https://github.com/longsleep/linux-pine64/blob/lichee-dev-v3.10.65/drivers/mmc/host/sunxi-mmc-sun50iw1p1-2.c

You mean the bit in SDXC_REG_SD_NTSR? Yes I know that doesn't exist for
the eMMC controller. I mentioned this in the commit message. It doesn't
exist, and writes to it become a no-op.

Would a comment, or comments, help with making this clear?

> And I definitely remember having HS-DDR working back when I added the
> a64 eMMC support.

Well it doesn't at the moment. My BPI-M64 reports:

    [    1.634276] mmc2: new high speed MMC card at address 0001

And with the patch:

    [    1.632552] mmc2: new DDR MMC card at address 0001

Regards
ChenYu
Maxime Ripard July 17, 2018, 3:15 p.m. UTC | #3
On Thu, Jul 12, 2018 at 06:17:23PM +0800, Chen-Yu Tsai wrote:
> On Thu, Jul 12, 2018 at 3:19 PM, Maxime Ripard
> <maxime.ripard@bootlin.com> wrote:
> > Hi,
> >
> > On Thu, Jul 12, 2018 at 11:02:25AM +0800, Chen-Yu Tsai wrote:
> >> The eMMC controller is also a new timing mode controller, but it doesn't
> >> have the timing mode switch. It does however have signal delay and
> >> calibration controls, typical of Allwinner MMC controllers that support
> >> the new timing mode.
> >>
> >> Enable the new timing mode setting for the A64 eMMC controller. This
> >> also enables MMC HS-DDR modes, which gives higher throughput for eMMC
> >> chips that support it, and can deliver such throughput.
> >>
> >> Signed-off-by: Chen-Yu Tsai <wens@csie.org>
> >
> > That doesn't look right. The datasheet explicitly mentions that this
> > bit doesn't apply to the eMMC controller, and the BSP is doing the same:
> > https://github.com/longsleep/linux-pine64/blob/lichee-dev-v3.10.65/drivers/mmc/host/sunxi-mmc-sun50iw1p1-1.c
> >
> > vs
> > https://github.com/longsleep/linux-pine64/blob/lichee-dev-v3.10.65/drivers/mmc/host/sunxi-mmc-sun50iw1p1-2.c
> 
> You mean the bit in SDXC_REG_SD_NTSR? Yes I know that doesn't exist
> for the eMMC controller. I mentioned this in the commit message. It
> doesn't exist, and writes to it become a no-op.
> 
> Would a comment, or comments, help with making this clear?

Ah right. Maybe we should move the calibration under can_calibrate
though, or create another boolean for this?

Putting it under has_new_timings while the SoC doesn't use it looks
very confusing.

Maxime
Chen-Yu Tsai July 17, 2018, 3:43 p.m. UTC | #4
On Tue, Jul 17, 2018 at 11:15 PM, Maxime Ripard
<maxime.ripard@bootlin.com> wrote:
> On Thu, Jul 12, 2018 at 06:17:23PM +0800, Chen-Yu Tsai wrote:
>> On Thu, Jul 12, 2018 at 3:19 PM, Maxime Ripard
>> <maxime.ripard@bootlin.com> wrote:
>> > Hi,
>> >
>> > On Thu, Jul 12, 2018 at 11:02:25AM +0800, Chen-Yu Tsai wrote:
>> >> The eMMC controller is also a new timing mode controller, but it doesn't
>> >> have the timing mode switch. It does however have signal delay and
>> >> calibration controls, typical of Allwinner MMC controllers that support
>> >> the new timing mode.
>> >>
>> >> Enable the new timing mode setting for the A64 eMMC controller. This
>> >> also enables MMC HS-DDR modes, which gives higher throughput for eMMC
>> >> chips that support it, and can deliver such throughput.
>> >>
>> >> Signed-off-by: Chen-Yu Tsai <wens@csie.org>
>> >
>> > That doesn't look right. The datasheet explicitly mentions that this
>> > bit doesn't apply to the eMMC controller, and the BSP is doing the same:
>> > https://github.com/longsleep/linux-pine64/blob/lichee-dev-v3.10.65/drivers/mmc/host/sunxi-mmc-sun50iw1p1-1.c
>> >
>> > vs
>> > https://github.com/longsleep/linux-pine64/blob/lichee-dev-v3.10.65/drivers/mmc/host/sunxi-mmc-sun50iw1p1-2.c
>>
>> You mean the bit in SDXC_REG_SD_NTSR? Yes I know that doesn't exist
>> for the eMMC controller. I mentioned this in the commit message. It
>> doesn't exist, and writes to it become a no-op.
>>
>> Would a comment, or comments, help with making this clear?
>
> Ah right. Maybe we should move the calibration under can_calibrate
> though, or create another boolean for this?
>
> Putting it under has_new_timings while the SoC doesn't use it looks
> very confusing.

IIRC we don't support calibration anyway. This boolean simply signals
the usage of the new timing mode, whether by choice, or because it is
the only mode the controller supports.

ChenYu
Maxime Ripard July 18, 2018, 3:22 p.m. UTC | #5
On Tue, Jul 17, 2018 at 11:43:03PM +0800, Chen-Yu Tsai wrote:
> On Tue, Jul 17, 2018 at 11:15 PM, Maxime Ripard
> <maxime.ripard@bootlin.com> wrote:
> > On Thu, Jul 12, 2018 at 06:17:23PM +0800, Chen-Yu Tsai wrote:
> >> On Thu, Jul 12, 2018 at 3:19 PM, Maxime Ripard
> >> <maxime.ripard@bootlin.com> wrote:
> >> > Hi,
> >> >
> >> > On Thu, Jul 12, 2018 at 11:02:25AM +0800, Chen-Yu Tsai wrote:
> >> >> The eMMC controller is also a new timing mode controller, but it doesn't
> >> >> have the timing mode switch. It does however have signal delay and
> >> >> calibration controls, typical of Allwinner MMC controllers that support
> >> >> the new timing mode.
> >> >>
> >> >> Enable the new timing mode setting for the A64 eMMC controller. This
> >> >> also enables MMC HS-DDR modes, which gives higher throughput for eMMC
> >> >> chips that support it, and can deliver such throughput.
> >> >>
> >> >> Signed-off-by: Chen-Yu Tsai <wens@csie.org>
> >> >
> >> > That doesn't look right. The datasheet explicitly mentions that this
> >> > bit doesn't apply to the eMMC controller, and the BSP is doing the same:
> >> > https://github.com/longsleep/linux-pine64/blob/lichee-dev-v3.10.65/drivers/mmc/host/sunxi-mmc-sun50iw1p1-1.c
> >> >
> >> > vs
> >> > https://github.com/longsleep/linux-pine64/blob/lichee-dev-v3.10.65/drivers/mmc/host/sunxi-mmc-sun50iw1p1-2.c
> >>
> >> You mean the bit in SDXC_REG_SD_NTSR? Yes I know that doesn't exist
> >> for the eMMC controller. I mentioned this in the commit message. It
> >> doesn't exist, and writes to it become a no-op.
> >>
> >> Would a comment, or comments, help with making this clear?
> >
> > Ah right. Maybe we should move the calibration under can_calibrate
> > though, or create another boolean for this?
> >
> > Putting it under has_new_timings while the SoC doesn't use it looks
> > very confusing.
> 
> IIRC we don't support calibration anyway. This boolean simply signals
> the usage of the new timing mode, whether by choice, or because it is
> the only mode the controller supports.

This is not the semantic I had in mind when I introduced it. The
original intent was to set the new timing bit all the time for
SoCs. If we want to change that semantic, then we also need to make
sure what this bit means is documented properly.

Maxime
Chen-Yu Tsai July 30, 2018, 9:27 a.m. UTC | #6
On Wed, Jul 18, 2018 at 11:22 PM, Maxime Ripard
<maxime.ripard@bootlin.com> wrote:
> On Tue, Jul 17, 2018 at 11:43:03PM +0800, Chen-Yu Tsai wrote:
>> On Tue, Jul 17, 2018 at 11:15 PM, Maxime Ripard
>> <maxime.ripard@bootlin.com> wrote:
>> > On Thu, Jul 12, 2018 at 06:17:23PM +0800, Chen-Yu Tsai wrote:
>> >> On Thu, Jul 12, 2018 at 3:19 PM, Maxime Ripard
>> >> <maxime.ripard@bootlin.com> wrote:
>> >> > Hi,
>> >> >
>> >> > On Thu, Jul 12, 2018 at 11:02:25AM +0800, Chen-Yu Tsai wrote:
>> >> >> The eMMC controller is also a new timing mode controller, but it doesn't
>> >> >> have the timing mode switch. It does however have signal delay and
>> >> >> calibration controls, typical of Allwinner MMC controllers that support
>> >> >> the new timing mode.
>> >> >>
>> >> >> Enable the new timing mode setting for the A64 eMMC controller. This
>> >> >> also enables MMC HS-DDR modes, which gives higher throughput for eMMC
>> >> >> chips that support it, and can deliver such throughput.
>> >> >>
>> >> >> Signed-off-by: Chen-Yu Tsai <wens@csie.org>
>> >> >
>> >> > That doesn't look right. The datasheet explicitly mentions that this
>> >> > bit doesn't apply to the eMMC controller, and the BSP is doing the same:
>> >> > https://github.com/longsleep/linux-pine64/blob/lichee-dev-v3.10.65/drivers/mmc/host/sunxi-mmc-sun50iw1p1-1.c
>> >> >
>> >> > vs
>> >> > https://github.com/longsleep/linux-pine64/blob/lichee-dev-v3.10.65/drivers/mmc/host/sunxi-mmc-sun50iw1p1-2.c
>> >>
>> >> You mean the bit in SDXC_REG_SD_NTSR? Yes I know that doesn't exist
>> >> for the eMMC controller. I mentioned this in the commit message. It
>> >> doesn't exist, and writes to it become a no-op.
>> >>
>> >> Would a comment, or comments, help with making this clear?
>> >
>> > Ah right. Maybe we should move the calibration under can_calibrate
>> > though, or create another boolean for this?
>> >
>> > Putting it under has_new_timings while the SoC doesn't use it looks
>> > very confusing.
>>
>> IIRC we don't support calibration anyway. This boolean simply signals
>> the usage of the new timing mode, whether by choice, or because it is
>> the only mode the controller supports.
>
> This is not the semantic I had in mind when I introduced it. The
> original intent was to set the new timing bit all the time for
> SoCs. If we want to change that semantic, then we also need to make
> sure what this bit means is documented properly.

In the driver:

    /* hardware only supports new timing mode */
    bool needs_new_timings;

    /* hardware can switch between old and new timing modes */
    bool has_timings_switch;

So if the A64 / H6 eMMC controller only supports / is stuck with the new
timing mode, that surely fits the description of the first one, right?
As for setting the new timing bit all the time, yes it is set, but it's
a no-op. Would a comment clarifying this at the point the hardware bit
is set suffice?

Thanks
ChenYu
Maxime Ripard July 31, 2018, 2:19 p.m. UTC | #7
On Mon, Jul 30, 2018 at 05:27:43PM +0800, Chen-Yu Tsai wrote:
> On Wed, Jul 18, 2018 at 11:22 PM, Maxime Ripard
> <maxime.ripard@bootlin.com> wrote:
> > On Tue, Jul 17, 2018 at 11:43:03PM +0800, Chen-Yu Tsai wrote:
> >> On Tue, Jul 17, 2018 at 11:15 PM, Maxime Ripard
> >> <maxime.ripard@bootlin.com> wrote:
> >> > On Thu, Jul 12, 2018 at 06:17:23PM +0800, Chen-Yu Tsai wrote:
> >> >> On Thu, Jul 12, 2018 at 3:19 PM, Maxime Ripard
> >> >> <maxime.ripard@bootlin.com> wrote:
> >> >> > Hi,
> >> >> >
> >> >> > On Thu, Jul 12, 2018 at 11:02:25AM +0800, Chen-Yu Tsai wrote:
> >> >> >> The eMMC controller is also a new timing mode controller, but it doesn't
> >> >> >> have the timing mode switch. It does however have signal delay and
> >> >> >> calibration controls, typical of Allwinner MMC controllers that support
> >> >> >> the new timing mode.
> >> >> >>
> >> >> >> Enable the new timing mode setting for the A64 eMMC controller. This
> >> >> >> also enables MMC HS-DDR modes, which gives higher throughput for eMMC
> >> >> >> chips that support it, and can deliver such throughput.
> >> >> >>
> >> >> >> Signed-off-by: Chen-Yu Tsai <wens@csie.org>
> >> >> >
> >> >> > That doesn't look right. The datasheet explicitly mentions that this
> >> >> > bit doesn't apply to the eMMC controller, and the BSP is doing the same:
> >> >> > https://github.com/longsleep/linux-pine64/blob/lichee-dev-v3.10.65/drivers/mmc/host/sunxi-mmc-sun50iw1p1-1.c
> >> >> >
> >> >> > vs
> >> >> > https://github.com/longsleep/linux-pine64/blob/lichee-dev-v3.10.65/drivers/mmc/host/sunxi-mmc-sun50iw1p1-2.c
> >> >>
> >> >> You mean the bit in SDXC_REG_SD_NTSR? Yes I know that doesn't exist
> >> >> for the eMMC controller. I mentioned this in the commit message. It
> >> >> doesn't exist, and writes to it become a no-op.
> >> >>
> >> >> Would a comment, or comments, help with making this clear?
> >> >
> >> > Ah right. Maybe we should move the calibration under can_calibrate
> >> > though, or create another boolean for this?
> >> >
> >> > Putting it under has_new_timings while the SoC doesn't use it looks
> >> > very confusing.
> >>
> >> IIRC we don't support calibration anyway. This boolean simply signals
> >> the usage of the new timing mode, whether by choice, or because it is
> >> the only mode the controller supports.
> >
> > This is not the semantic I had in mind when I introduced it. The
> > original intent was to set the new timing bit all the time for
> > SoCs. If we want to change that semantic, then we also need to make
> > sure what this bit means is documented properly.
> 
> In the driver:
> 
>     /* hardware only supports new timing mode */
>     bool needs_new_timings;
> 
>     /* hardware can switch between old and new timing modes */
>     bool has_timings_switch;
> 
> So if the A64 / H6 eMMC controller only supports / is stuck with the new
> timing mode, that surely fits the description of the first one, right?

I guess so, yes

> As for setting the new timing bit all the time, yes it is set, but it's
> a no-op. Would a comment clarifying this at the point the hardware bit
> is set suffice?

Yep. Thanks!
Maxime

Patch
diff mbox

diff --git a/drivers/mmc/host/sunxi-mmc.c b/drivers/mmc/host/sunxi-mmc.c
index 8e7f3e35ee3d..13201bddfcc3 100644
--- a/drivers/mmc/host/sunxi-mmc.c
+++ b/drivers/mmc/host/sunxi-mmc.c
@@ -1166,6 +1166,7 @@  static const struct sunxi_mmc_cfg sun50i_a64_emmc_cfg = {
 	.idma_des_size_bits = 13,
 	.clk_delays = NULL,
 	.can_calibrate = true,
+	.needs_new_timings = true,
 };
 
 static const struct of_device_id sunxi_mmc_of_match[] = {