Hello Marcin,

Do you mean by only reverting as below:

-  SdMmcCmdBlk.CommandArgument = (AccessMode & 0xF) | ((CommandSystem & 0xF) << 4) | \

-                                ((DriverStrength & 0xF) << 8) | ((PowerLimit & 0xF) << 12) | \

+  SdMmcCmdBlk.CommandArgument = (AccessMode & 0xF) | ((PowerLimit & 0xF) << 4) | \^M

+                                ((DriverStrength & 0xF) << 8) | ((DriverStrength & 0xF) << 12) | \^M

ModeValue;

All your devices work fine?

Since the origin code is clearly not correct (DriveStrength used 2 times,

the offset of PowerLimit is not 4, should be 12 according to SD spec).

Best Regards,

Hao Wu

From: Marcin Wojtas [mailto:mw@semihalf.com]
Sent: Friday, June 28, 2019 2:33 PM
To: Wu, Hao A; Albecki, Mateusz
Cc: Sumit Garg; Ard Biesheuvel; edk2-devel-groups-io
Subject: Re: [edk2-devel] [PATCH v4 0/2] MdeModulePkg/SdMmcHcDxe: Implement revision 3 of SdMmcOverrideProtocol

Hi,

I was almost happily sending you email with 'tested-by' information, but I checked 3 boards:

Board 1 (out of tree): SD - OK, MMC - OK

Board 2: (Armada80x0McBin): SD - OK, MMC - OK

Board 3: (Armada70x0Db): SD - problems, MMC - OK

In the latter case I get stall and booting takes around 3 minutes.

Without "MdeModulePkg/SdMmcHcDxe: Implement revision 3 of SdMmcOverrideProtocol" patch it works as before.

I enabled debugs, and in theory everything seems fine, the DriverStrength is set to EDKII_SD_MMC_DRIVER_STRENGTH_IGNORE.

SdCardIdentification: Found a SD device at slot [0]

SdCardSetBusMode: Target bus mode: bus timing = 1, bus width = 4, clock freq[MHz] = 50, driver strength = 255

However right after Csd register dump the booting stalls until printing following and continuing:

FatOpenDevice: read of part_lba failed Time out

This is absent from the prints I dumped from vanilla kernel. Despite I currently really have no time to additional debug, I checked and with following diff, everything works as before:

--- a/MdeModulePkg/Bus/Pci/SdMmcPciHcDxe/SdDevice.c

+++ b/MdeModulePkg/Bus/Pci/SdMmcPciHcDxe/SdDevice.c

@@ -536,8 +536,8 @@ SdCardSwitch (

AccessMode = 0xF;

   }

-  SdMmcCmdBlk.CommandArgument = (AccessMode & 0xF) | ((CommandSystem & 0xF) << 4) | \

-                                ((DriverStrength & 0xF) << 8) | ((PowerLimit & 0xF) << 12) | \

+  SdMmcCmdBlk.CommandArgument = (AccessMode & 0xF) | ((PowerLimit & 0xF) << 4) | \^M

+                                ((DriverStrength & 0xF) << 8) | ((DriverStrength & 0xF) << 12) | \^M

ModeValue;

Above is restoring SdMmcCmdBlk.CommandArgument to the state from before the patch in question. Now the question - why the layout of the command changed? CommandSystem was unused before, and PowerLimit changed its position. Is this change really related to the rest of the patch? What is the justification?

Best regards,

Marcin

pt., 28 cze 2019 o 02:57 Wu, Hao A <hao.a.wu@intel.com> napisał(a):

> -----Original Message-----
> From: Sumit Garg [mailto:sumit.garg@linaro.org]
> Sent: Thursday, June 27, 2019 9:39 PM
> To: Ard Biesheuvel
> Cc: edk2-devel-groups-io; Wu, Hao A; Marcin Wojtas; Albecki, Mateusz
> Subject: Re: [edk2-devel] [PATCH v4 0/2] MdeModulePkg/SdMmcHcDxe:
> Implement revision 3 of SdMmcOverrideProtocol
>
> On Thu, 27 Jun 2019 at 13:40, Ard Biesheuvel <ard.biesheuvel@linaro.org>
> wrote:
> >
> > (+ Sumit)
> >
> > On Thu, 27 Jun 2019 at 08:29, Wu, Hao A <hao.a.wu@intel.com> wrote:
> > >
> > > > -----Original Message-----
> > > > From: Marcin Wojtas [mailto:mw@semihalf.com]
> > > > Sent: Thursday, June 27, 2019 2:25 PM
> > > > To: Albecki, Mateusz
> > > > Cc: edk2-devel-groups-io; Wu, Hao A
> > > > Subject: Re: [edk2-devel] [PATCH v4 0/2]
> MdeModulePkg/SdMmcHcDxe:
> > > > Implement revision 3 of SdMmcOverrideProtocol
> > > >
> > > > Hi Mateusz,
> > > >
> > > > Can you please push those patches somewhere (github?) so that I can
> > > > easily do a quick check for regression?
> > > >
> > > > Thanks,
> > > > Marcin
> > >
> > >
> > > Hello Marcin,
> > >
> > > I have just pushed the series at:
> > > https://github.com/hwu25/edk2/tree/sdmmc_override_extend_v4
> > >
> > > Please help to check.
> > >
> >
> > I have cc'ed my colleague Sumit, who has kindly agreed to regression
> > test this branch on Socionext SynQuacer, which also uses the SD/MMC
> > override infrastructure.
> >
> > Sumit, please reply here with your results. And thanks again!
>
> I did picked this patch-set and applied on top of edk2 master branch.
> It works well on SynQuacer with eMMC device enumerated properly and
> all three eMMC partitions are visible:
>
>      BLK4: Alias(s):
>           VenHw(0D51905B-B77E-452A-A2C0-
> ECA0CC8D514A,000030520000000000)/eMMC(0x
> 0)/Ctrl(0x0)
>      BLK5: Alias(s):
>           VenHw(0D51905B-B77E-452A-A2C0-
> ECA0CC8D514A,000030520000000000)/eMMC(0x
> 0)/Ctrl(0x1)
>      BLK6: Alias(s):
>           VenHw(0D51905B-B77E-452A-A2C0-
> ECA0CC8D514A,000030520000000000)/eMMC(0x
> 0)/Ctrl(0x2)
>
> Shell> devices
> <snip>
>   E9 D - -  1  1   0 VenHw(0D51905B-B77E-452A-A2C0-
> ECA0CC8D514A,0000305200000000
> 00)/eMMC(0x0)/Ctrl(0x0)
>   EA D - -  1  1   0 VenHw(0D51905B-B77E-452A-A2C0-
> ECA0CC8D514A,0000305200000000
> 00)/eMMC(0x0)/Ctrl(0x1)
>   EB D - -  1  1   0 VenHw(0D51905B-B77E-452A-A2C0-
> ECA0CC8D514A,0000305200000000
> 00)/eMMC(0x0)/Ctrl(0x2)
>
> So you can add following:
>
> Regression-tested-by: Sumit Garg <sumit.garg@linaro.org>


Thanks a lot for the testing.

Best Regards,
Hao Wu


>
> -Sumit