Possible regression due to "arm64: renesas: salvator: set driver type for eMMC"

Possible regression due to "arm64: renesas: salvator: set driver type for eMMC"

[Simon Horman]

Hi Wolfram,

I have observed the following behaviour in renesas-devel-20171130-v4.15-rc1
which I have bisected down to being caused by d3decc2c34c8 ("arm64: dts:
renesas: salvator: set driver type for eMMC"). The patch in question is
based on v4.15-rc1.

I am compiling using renesas_defconfig

The problem manifests on Salvator-X / H3 ES1.0.

[    2.567304] renesas_sdhi_internal_dmac ee140000.sd: mmc0 base at 0xee140000 max clock rate 200 MHz
[    2.700110] mmc0: tuning execution failed: -5
[    2.704591] mmc0: error -5 whilst initialising MMC card
[    2.783042] mmc0: tuning execution failed: -5
[    2.783057] mmc0: error -5 whilst initialising MMC card
[    2.935219] mmc0: tuning execution failed: -5
[    2.939676] mmc0: error -5 whilst initialising MMC card
[    3.103876] renesas_sdhi_internal_dmac ee100000.sd: mmc1 base at 0xee100000 max clock rate 200 MHz
[    3.174033] mmc0: tuning execution failed: -5

The behaviour on Salvator-XS / H3 ES2.0 seems inconsistent.

And on Salvator-X / M3-W ES1.0 I do not see the problem regardless of
the presence of the patch.

Re: Possible regression due to "arm64: renesas: salvator: set driver type for eMMC"

[Wolfram Sang]

[-- Attachment #1: Type: text/plain, Size: 2720 bytes --]

On Thu, Nov 30, 2017 at 05:18:03PM +0100, Simon Horman wrote:
> Hi Wolfram,
> 
> I have observed the following behaviour in renesas-devel-20171130-v4.15-rc1
> which I have bisected down to being caused by d3decc2c34c8 ("arm64: dts:
> renesas: salvator: set driver type for eMMC"). The patch in question is
> based on v4.15-rc1.
> 
> I am compiling using renesas_defconfig
> 
> The problem manifests on Salvator-X / H3 ES1.0.
> 
> [    2.567304] renesas_sdhi_internal_dmac ee140000.sd: mmc0 base at 0xee140000 max clock rate 200 MHz
> [    2.700110] mmc0: tuning execution failed: -5
> [    2.704591] mmc0: error -5 whilst initialising MMC card
> [    2.783042] mmc0: tuning execution failed: -5
> [    2.783057] mmc0: error -5 whilst initialising MMC card
> [    2.935219] mmc0: tuning execution failed: -5
> [    2.939676] mmc0: error -5 whilst initialising MMC card
> [    3.103876] renesas_sdhi_internal_dmac ee100000.sd: mmc1 base at 0xee100000 max clock rate 200 MHz
> [    3.174033] mmc0: tuning execution failed: -5

I can't reproduce this:

[    2.567789] mmc0: new HS200 MMC card at address 0001
[    2.568765] mmcblk0: mmc0:0001 BGSD3R 29.1 GiB 
[    2.569028] mmcblk0boot0: mmc0:0001 BGSD3R partition 1 16.0 MiB
[    2.569256] mmcblk0boot1: mmc0:0001 BGSD3R partition 2 16.0 MiB
[    2.569587] mmcblk0rpmb: mmc0:0001 BGSD3R partition 3 4.00 MiB, chardev (243:0)
[    2.571586]  mmcblk0: p1

I tested with your above settings, but also with plain v4.15-rc1 and a custom
config. That matches my test results when developing the patchset; I tested on
all Gen3 boards without problems.

While I think it could be that the early ES1.0 boards might have different eMMC
chips...

> The behaviour on Salvator-XS / H3 ES2.0 seems inconsistent.

... I am surprised by this behaviour. I would have assumed the eMMC chips would
have been fixed meanwhile.

So, the BSP had the new drive_strength setting for all Gen3 boards. See
following commits in periupport:

arm64: dts: r8a7795-es1-salvator-x: Add card driver strength property in SDHI2 (5086b3ae0b4667)
arm64: dts: r8a7795-salvator-x: Add card driver strength property in SDHI2 (8314f72db0b861)
arm64: dts: r8a7795-salvator-xs: Add card driver strength property in SDHI2 (01b074f6f59312)
arm64: dts: r8a7796-salvator-x: Add card driver strength property in SDHI2 (e4a3cc1bbe1c7b)

However, there is a comment in periupport for the patches implementing the new
drive strength feature:

mmc: sh_mobile_sdhi: Add the driver strength setting (3d277f92925504)

says

"eMMC operation becomes unstable in M3ws1.1+Salvator-xs."

So, maybe revert this change for other Gen3 boards now?

Regards,

   Wolfram


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: Possible regression due to "arm64: renesas: salvator: set driver type for eMMC"

[Geert Uytterhoeven]

On Fri, Dec 1, 2017 at 12:06 PM, Wolfram Sang <wsa@the-dreams.de> wrote:
> While I think it could be that the early ES1.0 boards might have different eMMC
> chips...

FTR, my boards have:

Salvator-X with H3 ES1.0:
    mmc0: new HS200 MMC card at address 0001
    mmcblk0: mmc0:0001 BGSD3R 29.1 GiB

Salvator-X with M3-W ES1.0:

    mmc0: new HS200 MMC card at address 0001
    mmcblk0: mmc0:0001 eMMC   28.8 GiB

Salvator-XS with H3 ES2.0:

    mmc0: new HS200 MMC card at address 0001
    mmcblk0: mmc0:0001 BGSD3R 29.1 GiB

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds