Hello Hao,
Answers to your comments:
1. Patch 5 fixes the issue with accessing SD as well as eMMC files.
2. Using Private variable for controller version would mean I can only access
it in functions having an instance of the installed protocol which is not
possible in some functions for example where we set clock dividers. We either
need to use what I have or pass Private or controller version as an argument to
functions using it.
3. I think there is a bigger issue here. During host controller initialization
we need to setup 64b addressing which happens before we initialize 64b DMA in
PCI layer. So what you are suggesting may not be that helpful. Also, what we
are doing right now is that we go through all slots and if controller on any
slot does not support 64b, we do not enable 64b DMA in PCI layer. What is the
likelihood of all controllers on an SOC not supporting similar address width?
Also, shouldn't we be enabling 64b DMA in PCI layer if any of the controller
supports 64b addressing? Wouldn't 64b DMA enabled mean 32b is enabled by
default?
4. Patch 5 has been rebased on tip of edk2.
Thanks
Ashish
-----Original Message-----
From: Wu, Hao A <hao.a...@intel.com>
Sent: Wednesday, November 28, 2018 12:25 AM
To: Ashish Singhal <ashishsin...@nvidia.com>; edk2-devel@lists.01.org
Subject: RE: [edk2] [PATCH v3 2/2] MdeModulePkg/SdMmcPciHcDxe: Add V4 64bit
SDMA and ADMA2 support.
Hello,
Sorry for the delayed response.
Apart from inserting the Bugzilla tracker information, several more general
level
comments:
1. Cannot access the files (content) on SD & eMMC devices
After applying (rebasing onto the latest codes) your patches, I found that
though the SD & eMMC devices can be detected, yet the content cannot be
accessed.
There are 1 SD card and 1 embedded eMMC device within the system. Under Shell,
I can see 2 file systems on SD & eMMC devices being detected, but when I try to
access the files on those file systems by using the 'ls' command, it fails,
saying "ls: File Not Found - 'FS0:\'". I can confirm that files can be listed
successfully without applying this patch.
I also tried the 'dblk' command for display the data via BlockIO protocols
created on those devices. I found that for SD, the command works fine. But for
eMMC, the command fails with message "dblk: Read file error - 'BlockIo'".
For the hardware I own, the controllers are all version 3.0. I tested on both
IA32 and X64 arch image. The results were the same as described above.
Unfortunately, I do not have access to boards with SD controller with version
newer than 3.0. So I am not able to perform those tests on my side for Version
4.0 or greater SD controllers.
Do you have any hardware with SD controller of version 3.0? Is it working on
your side?
Please let me know if additional information is needed.
2. On SdMmcHcGetControllerVersion()
I found that there is a 'ControllerVersion' version in the data structure
SD_MMC_HC_PRIVATE_DATA. But currently, this field seems never used.
I think we can get the controller version information only once and use this
field to store it. Hence, this new function can be eliminated and also can
avoid calling it multiple times.
3. Setting the SD_MMC_HC_64_ADDR_EN bit in SdMmcHcV4Init64BitSupport()
The setting of the SD_MMC_HC_64_ADDR_EN bit is decided by the 'SysBus64V4'
field in the CAP register. For me, additional dependency on the 64-bit DMA
support in the PCI layer is needed as well.
In function SdMmcPciHcDriverBindingStart():
//
// Enable 64-bit DMA support in the PCI layer if this controller
// supports it.
//
if (Support64BitDma) {
Status = PciIo->Attributes (
PciIo,
EfiPciIoAttributeOperationEnable,
EFI_PCI_IO_ATTRIBUTE_DUAL_ADDRESS_CYCLE,
NULL
);
if (EFI_ERROR (Status)) {
DEBUG ((DEBUG_WARN, "SdMmcPciHcDriverBindingStart: failed to enable
64-bit DMA (%r)\n", Status));
}
}
If the above PCI attribute setting fails, the PCI layer will not support the
64-bit DMA.
Hence, I am thinking to introduce a new BOOLEAN field in the
"SD_MMC_HC_PRIVATE_DATA" data structure (maybe putting the 'Support64BitDma'
into the data structure). If the above PCI operation succeeds, the BOOLEAN
field will be set to TRUE, otherwise, FALSE.
And the setting of the SD_MMC_HC_64_ADDR_EN bit should depend on the BOOLEAN
field as well.
4. Please help to rebase the patch upon the latest master branch.
Some changes within the SdMmcPciHcDxe have been pushed recently. Could you help
to rebase this patch upon the latest changes. Thanks in advance.
Also, some inline comments below.
> -----Original Message-----
> From: edk2-devel [mailto:edk2-devel-boun...@lists.01.org] On Behalf Of
> Ashish Singhal
> Sent: Tuesday, November 20, 2018 4:59 AM
> To: edk2-devel@lists.01.org
> Cc: Ashish Singhal
> Subject: [edk2] [PATCH v3 2/2] MdeModulePkg/SdMmcPciHcDxe: Add V4
> 64bit SDMA and ADMA2 support.
>
> If V4 64 bit address mode is enabled in compatibility register,
> program controller to enable V4 host mode.
> Use appropriate ADMA2 descriptors supporting 64 bit addresses.
> Use appropriate registers for SDMA mode operation.
>
> Contributed-under: TianoCore Contribution Agreement 1.1
> Signed-off-by: Ashish Singhal <ashishsin...@nvidia.com>
> ---
> MdeModulePkg/Bus/Pci/SdMmcPciHcDxe/SdMmcPciHcDxe.h | 4 +-
> MdeModulePkg/Bus/Pci/SdMmcPciHcDxe/SdMmcPciHci.c | 273
> +++++++++++++++++----
> MdeModulePkg/Bus/Pci/SdMmcPciHcDxe/SdMmcPciHci.h | 28 ++-
> 3 files changed, 260 insertions(+), 45 deletions(-)
>
> diff --git a/MdeModulePkg/Bus/Pci/SdMmcPciHcDxe/SdMmcPciHcDxe.h
> b/MdeModulePkg/Bus/Pci/SdMmcPciHcDxe/SdMmcPciHcDxe.h
> index c683600..22795df 100644
> --- a/MdeModulePkg/Bus/Pci/SdMmcPciHcDxe/SdMmcPciHcDxe.h
> +++ b/MdeModulePkg/Bus/Pci/SdMmcPciHcDxe/SdMmcPciHcDxe.h
> @@ -2,6 +2,7 @@
>
> Provides some data structure definitions used by the SD/MMC host
> controller driver.
>
> +Copyright (c) 2018, NVIDIA CORPORATION. All rights reserved.
> Copyright (c) 2015, Intel Corporation. All rights reserved.<BR> This
> program and the accompanying materials are licensed and made
> available under the terms and conditions of the BSD License @@ -144,7
> +145,8 @@ typedef struct {
> BOOLEAN Started;
> UINT64 Timeout;
>
> - SD_MMC_HC_ADMA_DESC_LINE *AdmaDesc;
> + SD_MMC_HC_ADMA_32_DESC_LINE *Adma32Desc;
> + SD_MMC_HC_ADMA_64_DESC_LINE *Adma64Desc;
> EFI_PHYSICAL_ADDRESS AdmaDescPhy;
> VOID *AdmaMap;
> UINT32 AdmaPages;
> diff --git a/MdeModulePkg/Bus/Pci/SdMmcPciHcDxe/SdMmcPciHci.c
> b/MdeModulePkg/Bus/Pci/SdMmcPciHcDxe/SdMmcPciHci.c
> index e506875..9fef3fb 100644
> --- a/MdeModulePkg/Bus/Pci/SdMmcPciHcDxe/SdMmcPciHci.c
> +++ b/MdeModulePkg/Bus/Pci/SdMmcPciHcDxe/SdMmcPciHci.c
> @@ -4,6 +4,7 @@
>
> It would expose EFI_SD_MMC_PASS_THRU_PROTOCOL for upper layer use.
>
> + Copyright (c) 2018, NVIDIA CORPORATION. All rights reserved.
> Copyright (c) 2015 - 2017, Intel Corporation. All rights reserved.<BR>
> This program and the accompanying materials
> are licensed and made available under the terms and conditions of
> the BSD License @@ -418,6 +419,36 @@ SdMmcHcWaitMmioSet ( }
>
> /**
> + Get the controller version information from the specified slot.
> +
> + @param[in] PciIo The PCI IO protocol instance.
> + @param[in] Slot The slot number of the SD card to send the
> command to.
> + @param[out] Version The buffer to store the version information.
> +
> + @retval EFI_SUCCESS The operation executes successfully.
> + @retval Others The operation fails.
> +
> +**/
> +EFI_STATUS
> +SdMmcHcGetControllerVersion (
> + IN EFI_PCI_IO_PROTOCOL *PciIo,
> + IN UINT8 Slot,
> + OUT UINT16 *Version
> + )
> +{
> + EFI_STATUS Status;
> +
> + Status = SdMmcHcRwMmio (PciIo, Slot, SD_MMC_HC_CTRL_VER, TRUE,
> sizeof (UINT16), Version);
> + if (EFI_ERROR (Status)) {
> + return Status;
> + }
> +
> + *Version &= 0xFF;
> +
> + return EFI_SUCCESS;
> +}
> +
> +/**
> Software reset the specified SD/MMC host controller and enable all
> interrupts.
>
> @param[in] Private A pointer to the SD_MMC_HC_PRIVATE_DATA
> instance.
> @@ -776,18 +807,18 @@ SdMmcHcClockSupply (
>
> DEBUG ((DEBUG_INFO, "BaseClkFreq %dMHz Divisor %d ClockFreq
> %dKhz\n", BaseClkFreq, Divisor, ClockFreq));
>
> - Status = SdMmcHcRwMmio (PciIo, Slot, SD_MMC_HC_CTRL_VER, TRUE,
> sizeof (ControllerVer), &ControllerVer);
> + Status = SdMmcHcGetControllerVersion (PciIo, Slot, &ControllerVer);
> if (EFI_ERROR (Status)) {
> return Status;
> }
> //
> // Set SDCLK Frequency Select and Internal Clock Enable fields in
> Clock Control register.
> //
> - if (((ControllerVer & 0xFF) >= SD_MMC_HC_CTRL_VER_300) &&
> - ((ControllerVer & 0xFF) <= SD_MMC_HC_CTRL_VER_420)) {
> + if ((ControllerVer >= SD_MMC_HC_CTRL_VER_300) &&
> + (ControllerVer <= SD_MMC_HC_CTRL_VER_420)) {
> ASSERT (Divisor <= 0x3FF);
> ClockCtrl = ((Divisor & 0xFF) << 8) | ((Divisor & 0x300) >> 2);
> - } else if (((ControllerVer & 0xFF) == 0) || ((ControllerVer & 0xFF)
> == 1)) {
> + } else if ((ControllerVer == 0) || (ControllerVer == 1)) {
Could you help to update the above 'else if' statement to use the below version
definitions?
SD_MMC_HC_CTRL_VER_100
SD_MMC_HC_CTRL_VER_200
> //
> // Only the most significant bit can be used as divisor.
> //
> @@ -935,6 +966,54 @@ SdMmcHcSetBusWidth ( }
>
> /**
> + Configure V4 64 bit system address support at initialization.
> +
> + @param[in] PciIo The PCI IO protocol instance.
> + @param[in] Slot The slot number of the SD card to send the
> command to.
> + @param[in] Capability The capability of the slot.
> +
> + @retval EFI_SUCCESS The clock is supplied successfully.
> +
> +**/
> +EFI_STATUS
> +SdMmcHcV4Init64BitSupport (
> + IN EFI_PCI_IO_PROTOCOL *PciIo,
> + IN UINT8 Slot,
> + IN SD_MMC_HC_SLOT_CAP Capability
> + )
> +{
> + EFI_STATUS Status;
> + UINT16 ControllerVer;
> + UINT16 HostCtrl2;
> +
> + //
> + // Check if V4 64bit support is available // Status =
> + SdMmcHcGetControllerVersion (PciIo, Slot, &ControllerVer); if
> + (EFI_ERROR (Status)) {
> + return Status;
> + }
> +
> + if (ControllerVer >= SD_MMC_HC_CTRL_VER_400) {
> + HostCtrl2 = SD_MMC_HC_V4_EN;
> + //
> + // Check if V4 64bit support is available
> + //
> + if (Capability.SysBus64V4 == TRUE) {
Apart from the additional check needed comments above, could you help to refine
the above check to:
if (Capability.SysBus64V4 != 0) {
> + HostCtrl2 |= SD_MMC_HC_64_ADDR_EN;
> + DEBUG ((DEBUG_INFO, "Enabled V4 64 bit system bus support\n"));
> + }
> + Status = SdMmcHcOrMmio (PciIo, Slot, SD_MMC_HC_HOST_CTRL2, sizeof
> (HostCtrl2), &HostCtrl2);
> + if (EFI_ERROR (Status)) {
> + return Status;
> + }
> + DEBUG ((DEBUG_INFO, "Enabled V4 64 bit system bus support\n"));
> + }
> +
> + return EFI_SUCCESS;
> +}
> +
> +/**
> Supply SD/MMC card with lowest clock frequency at initialization.
>
> @param[in] PciIo The PCI IO protocol instance.
> @@ -1101,6 +1180,11 @@ SdMmcHcInitHost (
> PciIo = Private->PciIo;
> Capability = Private->Capability[Slot];
>
> + Status = SdMmcHcV4Init64BitSupport (PciIo, Slot, Capability); if
> + (EFI_ERROR (Status)) {
> + return Status;
> + }
> +
> Status = SdMmcHcInitClockFreq (PciIo, Slot, Capability);
> if (EFI_ERROR (Status)) {
> return Status;
> @@ -1169,7 +1253,7 @@ SdMmcHcLedOnOff (
> /**
> Build ADMA descriptor table for transfer.
>
> - Refer to SD Host Controller Simplified spec 3.0 Section 1.13 for details.
> + Refer to SD Host Controller Simplified spec 4.2 Section 1.13 for details.
>
> @param[in] Trb The pointer to the SD_MMC_HC_TRB instance.
>
> @@ -1187,49 +1271,69 @@ BuildAdmaDescTable (
> UINT64 Entries;
> UINT32 Index;
> UINT64 Remaining;
> - UINT32 Address;
> + UINT64 Address;
> UINTN TableSize;
> EFI_PCI_IO_PROTOCOL *PciIo;
> EFI_STATUS Status;
> UINTN Bytes;
> + UINT16 ControllerVer;
> + BOOLEAN AddressingMode64 = FALSE;
> + UINTN DescSize = sizeof (SD_MMC_HC_ADMA_32_DESC_LINE);
> + VOID *AdmaDesc = NULL;
Please help to separate the variable declaration and initial value assignment.
Also, please update the type of variable 'DescSize' from UINTN to UINT32. The
Visual Studio 2015 complier complains for:
" TableSize = (UINTN)MultU64x32 (Entries, DescSize); ":
conversion from 'UINTN' to 'UINT32', possible loss of data.
>
> Data = Trb->DataPhy;
> DataLen = Trb->DataLen;
> PciIo = Trb->Private->PciIo;
> +
> + //
> + // Detect whether 64bit addressing is supported.
> //
> - // Only support 32bit ADMA Descriptor Table
> + Status = SdMmcHcGetControllerVersion (PciIo, Trb->Slot,
> + &ControllerVer); if (EFI_ERROR (Status)) {
> + return Status;
> + }
> + if (ControllerVer >= SD_MMC_HC_CTRL_VER_400) {
> + Status = SdMmcHcCheckMmioSet(PciIo, Trb->Slot,
> SD_MMC_HC_HOST_CTRL2, 0x2,
> +
> + SD_MMC_HC_V4_EN|SD_MMC_HC_64_ADDR_EN,
> SD_MMC_HC_V4_EN|SD_MMC_HC_64_ADDR_EN);
> + if (!EFI_ERROR (Status)) {
> + AddressingMode64 = TRUE;
> + DescSize = sizeof (SD_MMC_HC_ADMA_64_DESC_LINE);
> + }
> + }
> //
> - if ((Data >= 0x100000000ul) || ((Data + DataLen) > 0x100000000ul))
> {
> + // Check for valid ranges in 32bit ADMA Descriptor Table // if
> + (AddressingMode64 == FALSE &&
> + ((Data >= 0x100000000ul) || ((Data + DataLen) >
> + 0x100000000ul))) {
Please help to update the above check to:
if ((!AddressingMode64) &&
((Data >= 0x100000000ul) || ((Data + DataLen) > 0x100000000ul))) {
> return EFI_INVALID_PARAMETER;
> }
> //
> // Address field shall be set on 32-bit boundary (Lower 2-bit is
> always set to
> 0)
> - // for 32-bit address descriptor table.
> //
> if ((Data & (BIT0 | BIT1)) != 0) {
> DEBUG ((DEBUG_INFO, "The buffer [0x%x] to construct ADMA desc is
> not aligned to 4 bytes boundary!\n", Data));
> }
I found that the above check (also the comments) should be updated. For 32-bit
addressing the address of ADMA desc is 4-bytes aligned; for 64-bit case, the
requirement is 8-byte.
Seems the align check for 64-bit addressing is missing here.
>
> Entries = DivU64x32 ((DataLen + ADMA_MAX_DATA_PER_LINE - 1),
> ADMA_MAX_DATA_PER_LINE);
> - TableSize = (UINTN)MultU64x32 (Entries, sizeof
> (SD_MMC_HC_ADMA_DESC_LINE));
> + TableSize = (UINTN)MultU64x32 (Entries, DescSize);
> Trb->AdmaPages = (UINT32)EFI_SIZE_TO_PAGES (TableSize);
> Status = PciIo->AllocateBuffer (
> PciIo,
> AllocateAnyPages,
> EfiBootServicesData,
> EFI_SIZE_TO_PAGES (TableSize),
> - (VOID **)&Trb->AdmaDesc,
> + (VOID **)&AdmaDesc,
> 0
> );
> if (EFI_ERROR (Status)) {
> return EFI_OUT_OF_RESOURCES;
> }
> - ZeroMem (Trb->AdmaDesc, TableSize);
> + ZeroMem (AdmaDesc, TableSize);
> Bytes = TableSize;
> Status = PciIo->Map (
> PciIo,
> EfiPciIoOperationBusMasterCommonBuffer,
> - Trb->AdmaDesc,
> + AdmaDesc,
> &Bytes,
> &Trb->AdmaDescPhy,
> &Trb->AdmaMap
> @@ -1242,12 +1346,13 @@ BuildAdmaDescTable (
> PciIo->FreeBuffer (
> PciIo,
> EFI_SIZE_TO_PAGES (TableSize),
> - Trb->AdmaDesc
> + AdmaDesc
> );
> return EFI_OUT_OF_RESOURCES;
> }
>
> - if ((UINT64)(UINTN)Trb->AdmaDescPhy > 0x100000000ul) {
> + if ((AddressingMode64 == FALSE) &&
> + (UINT64)(UINTN)Trb->AdmaDescPhy > 0x100000000ul) {
Please update the above check to:
if ((!AddressingMode64) &&
(UINT64)(UINTN)Trb->AdmaDescPhy > 0x100000000ul) {
> //
> // The ADMA doesn't support 64bit addressing.
> //
> @@ -1258,25 +1363,49 @@ BuildAdmaDescTable (
> PciIo->FreeBuffer (
> PciIo,
> EFI_SIZE_TO_PAGES (TableSize),
> - Trb->AdmaDesc
> + AdmaDesc
> );
> return EFI_DEVICE_ERROR;
> }
>
> Remaining = DataLen;
> - Address = (UINT32)Data;
> + Address = Data;
> + if (AddressingMode64 == FALSE) {
Please help to update the above check to:
if (!AddressingMode64) {
> + Trb->Adma32Desc = AdmaDesc;
> + Trb->Adma64Desc = NULL;
> + } else {
> + Trb->Adma64Desc = AdmaDesc;
> + Trb->Adma32Desc = NULL;
> + }
> for (Index = 0; Index < Entries; Index++) {
> - if (Remaining <= ADMA_MAX_DATA_PER_LINE) {
> - Trb->AdmaDesc[Index].Valid = 1;
> - Trb->AdmaDesc[Index].Act = 2;
> - Trb->AdmaDesc[Index].Length = (UINT16)Remaining;
> - Trb->AdmaDesc[Index].Address = Address;
> - break;
> + if (AddressingMode64 == FALSE) {
Please help to update the above check to:
if (!AddressingMode64) {
> + if (Remaining < ADMA_MAX_DATA_PER_LINE) {
> + Trb->Adma32Desc[Index].Valid = 1;
> + Trb->Adma32Desc[Index].Act = 2;
> + Trb->Adma32Desc[Index].Length = (UINT16)Remaining;
> + Trb->Adma32Desc[Index].Address = (UINT32)Address;
> + break;
> + } else {
> + Trb->Adma32Desc[Index].Valid = 1;
> + Trb->Adma32Desc[Index].Act = 2;
> + Trb->Adma32Desc[Index].Length = 0;
> + Trb->Adma32Desc[Index].Address = (UINT32)Address;
> + }
> } else {
> - Trb->AdmaDesc[Index].Valid = 1;
> - Trb->AdmaDesc[Index].Act = 2;
> - Trb->AdmaDesc[Index].Length = 0;
> - Trb->AdmaDesc[Index].Address = Address;
> + if (Remaining < ADMA_MAX_DATA_PER_LINE) {
> + Trb->Adma64Desc[Index].Valid = 1;
> + Trb->Adma64Desc[Index].Act = 2;
> + Trb->Adma64Desc[Index].Length = (UINT16)Remaining;
> + Trb->Adma64Desc[Index].LowerAddress = (UINT32)(Address &
> MAX_UINT32);
> + Trb->Adma64Desc[Index].UpperAddress = (UINT32)(Address>>32);
> + break;
> + } else {
> + Trb->Adma64Desc[Index].Valid = 1;
> + Trb->Adma64Desc[Index].Act = 2;
> + Trb->Adma64Desc[Index].Length = 0;
> + Trb->Adma64Desc[Index].LowerAddress = (UINT32)(Address &
> MAX_UINT32);
I think the above 2 " & MAX_UINT32" can be dropped.
> + Trb->Adma64Desc[Index].UpperAddress = (UINT32)(Address>>32);
> + }
> }
>
> Remaining -= ADMA_MAX_DATA_PER_LINE; @@ -1286,7 +1415,7 @@
> BuildAdmaDescTable (
> //
> // Set the last descriptor line as end of descriptor table
> //
> - Trb->AdmaDesc[Index].End = 1;
> + AddressingMode64 ? (Trb->Adma64Desc[Index].End = 1) : (Trb-
> >Adma32Desc[Index].End = 1);
> return EFI_SUCCESS;
> }
>
> @@ -1430,11 +1559,18 @@ SdMmcFreeTrb (
> Trb->AdmaMap
> );
> }
> - if (Trb->AdmaDesc != NULL) {
> + if (Trb->Adma32Desc != NULL) {
> PciIo->FreeBuffer (
> PciIo,
> Trb->AdmaPages,
> - Trb->AdmaDesc
> + Trb->Adma32Desc
> + );
> + }
> + if (Trb->Adma64Desc != NULL) {
> + PciIo->FreeBuffer (
> + PciIo,
> + Trb->AdmaPages,
> + Trb->Adma64Desc
> );
> }
> if (Trb->DataMap != NULL) {
> @@ -1574,12 +1710,14 @@ SdMmcExecTrb (
> UINT16 Cmd;
> UINT16 IntStatus;
> UINT32 Argument;
> - UINT16 BlkCount;
> + UINT32 BlkCount;
> UINT16 BlkSize;
> UINT16 TransMode;
> UINT8 HostCtrl1;
> - UINT32 SdmaAddr;
> + UINT64 SdmaAddr;
> UINT64 AdmaAddr;
> + UINT16 ControllerVer;
> + BOOLEAN AddressingMode64 = FALSE;
Please help to separate the variable declaration and initial value assignment.
>
> Packet = Trb->Packet;
> PciIo = Trb->Private->PciIo;
> @@ -1612,13 +1750,33 @@ SdMmcExecTrb (
>
> SdMmcHcLedOnOff (PciIo, Trb->Slot, TRUE);
>
> + Status = SdMmcHcGetControllerVersion (PciIo, Trb->Slot,
> + &ControllerVer); if (EFI_ERROR (Status)) {
> + return Status;
> + }
> + if (ControllerVer >= SD_MMC_HC_CTRL_VER_400) {
> + Status = SdMmcHcCheckMmioSet(PciIo, Trb->Slot,
> SD_MMC_HC_HOST_CTRL2, 0x2,
> +
> + SD_MMC_HC_V4_EN|SD_MMC_HC_64_ADDR_EN,
> SD_MMC_HC_V4_EN|SD_MMC_HC_64_ADDR_EN);
> + if (!EFI_ERROR (Status)) {
> + AddressingMode64 = TRUE;
> + }
> + }
> +
> if (Trb->Mode == SdMmcSdmaMode) {
> - if ((UINT64)(UINTN)Trb->DataPhy >= 0x100000000ul) {
> + if ((AddressingMode64 == FALSE) &&
> + ((UINT64)(UINTN)Trb->DataPhy >= 0x100000000ul)) {
Please help to update the above check to:
if ((!AddressingMode64) &&
((UINT64)(UINTN)Trb->DataPhy >= 0x100000000ul)) {
> return EFI_INVALID_PARAMETER;
> }
>
> - SdmaAddr = (UINT32)(UINTN)Trb->DataPhy;
> - Status = SdMmcHcRwMmio (PciIo, Trb->Slot, SD_MMC_HC_SDMA_ADDR,
> FALSE, sizeof (SdmaAddr), &SdmaAddr);
> + SdmaAddr = (UINT64)(UINTN)Trb->DataPhy;
> +
> + if (ControllerVer >= SD_MMC_HC_CTRL_VER_400) {
> + Status = SdMmcHcRwMmio (PciIo, Trb->Slot,
> SD_MMC_HC_ADMA_SYS_ADDR, FALSE, sizeof (UINT64), &SdmaAddr);
> + }
> + else {
Minor coding style comment for the above line:
} else {
> + Status = SdMmcHcRwMmio (PciIo, Trb->Slot,
> SD_MMC_HC_SDMA_ADDR, FALSE, sizeof (UINT32), &SdmaAddr);
> + }
> +
> if (EFI_ERROR (Status)) {
> return Status;
> }
> @@ -1648,9 +1806,14 @@ SdMmcExecTrb (
> //
> // Calcuate Block Count.
> //
> - BlkCount = (UINT16)(Trb->DataLen / Trb->BlockSize);
> + BlkCount = (Trb->DataLen / Trb->BlockSize); }
For the below 'Block Count' settings, I think it would be better to add
comments that quote the SD Host Controller Spec to mention that the 32-bit
block count support for all operations in SD mode was introduced at Version
4.10.
> + if (ControllerVer >= SD_MMC_HC_CTRL_VER_410) {
> + Status = SdMmcHcRwMmio (PciIo, Trb->Slot, SD_MMC_HC_SDMA_ADDR,
> FALSE, sizeof (UINT32), &BlkCount);
> + }
> + else {
Minor coding style comment for the above line:
} else {
> + Status = SdMmcHcRwMmio (PciIo, Trb->Slot, SD_MMC_HC_BLK_COUNT,
> FALSE, sizeof (UINT16), &BlkCount);
> }
> - Status = SdMmcHcRwMmio (PciIo, Trb->Slot, SD_MMC_HC_BLK_COUNT,
> FALSE, sizeof (BlkCount), &BlkCount);
> if (EFI_ERROR (Status)) {
> return Status;
> }
> @@ -1746,10 +1909,11 @@ SdMmcCheckTrbResult (
> EFI_SD_MMC_PASS_THRU_COMMAND_PACKET *Packet;
> UINT16 IntStatus;
> UINT32 Response[4];
> - UINT32 SdmaAddr;
> + UINT64 SdmaAddr;
> UINT8 Index;
> UINT8 SwReset;
> UINT32 PioLength;
> + UINT16 ControllerVer;
>
> SwReset = 0;
> Packet = Trb->Packet;
> @@ -1870,19 +2034,42 @@ SdMmcCheckTrbResult (
> //
> // Update SDMA Address register.
> //
> - SdmaAddr = SD_MMC_SDMA_ROUND_UP ((UINT32)(UINTN)Trb-
> >DataPhy, SD_MMC_SDMA_BOUNDARY);
> - Status = SdMmcHcRwMmio (
> + SdmaAddr = SD_MMC_SDMA_ROUND_UP ((UINTN)Trb->DataPhy,
> SD_MMC_SDMA_BOUNDARY);
> +
> + Status = SdMmcHcGetControllerVersion (
> + Private->PciIo,
> + Trb->Slot,
> + &ControllerVer
> + );
> + if (EFI_ERROR (Status)) {
> + return Status;
> + }
> +
> + if (ControllerVer >= SD_MMC_HC_CTRL_VER_400) {
> + Status = SdMmcHcRwMmio (
> Private->PciIo,
> Trb->Slot,
> - SD_MMC_HC_SDMA_ADDR,
> + SD_MMC_HC_ADMA_SYS_ADDR,
> FALSE,
> - sizeof (UINT32),
> + sizeof (UINT64),
> &SdmaAddr
> );
> + }
> + else {
> + Status = SdMmcHcRwMmio (
> + Private->PciIo,
> + Trb->Slot,
> + SD_MMC_HC_SDMA_ADDR,
> + FALSE,
> + sizeof (UINT32),
> + &SdmaAddr
> + );
> + }
Minor coding style comments for the above line:
1. } else {
2. Please help to refine the space indent of the above SdMmcHcRwMmio() call
Best Regards,
Hao Wu
> +
> if (EFI_ERROR (Status)) {
> goto Done;
> }
> - Trb->DataPhy = (UINT32)(UINTN)SdmaAddr;
> + Trb->DataPhy = (UINT64)(UINTN)SdmaAddr;
> }
>
> if ((Packet->SdMmcCmdBlk->CommandType != SdMmcCommandTypeAdtc) &&
> diff --git a/MdeModulePkg/Bus/Pci/SdMmcPciHcDxe/SdMmcPciHci.h
> b/MdeModulePkg/Bus/Pci/SdMmcPciHcDxe/SdMmcPciHci.h
> index cc138fc..a6234f1 100644
> --- a/MdeModulePkg/Bus/Pci/SdMmcPciHcDxe/SdMmcPciHci.h
> +++ b/MdeModulePkg/Bus/Pci/SdMmcPciHcDxe/SdMmcPciHci.h
> @@ -2,6 +2,7 @@
>
> Provides some data structure definitions used by the SD/MMC host
> controller driver.
>
> +Copyright (c) 2018, NVIDIA CORPORATION. All rights reserved.
> Copyright (c) 2015, Intel Corporation. All rights reserved.<BR> This
> program and the accompanying materials are licensed and made
> available under the terms and conditions of the BSD License @@ -78,6
> +79,9 @@ typedef enum { //
> #define ADMA_MAX_DATA_PER_LINE 0x10000
>
> +//
> +// ADMA descriptor for 32b addressing.
> +//
> typedef struct {
> UINT32 Valid:1;
> UINT32 End:1;
> @@ -87,7 +91,23 @@ typedef struct {
> UINT32 Reserved1:10;
> UINT32 Length:16;
> UINT32 Address;
> -} SD_MMC_HC_ADMA_DESC_LINE;
> +} SD_MMC_HC_ADMA_32_DESC_LINE;
> +
> +//
> +// ADMA descriptor for 64b addressing.
> +//
> +typedef struct {
> + UINT32 Valid:1;
> + UINT32 End:1;
> + UINT32 Int:1;
> + UINT32 Reserved:1;
> + UINT32 Act:2;
> + UINT32 Reserved1:10;
> + UINT32 Length:16;
> + UINT32 LowerAddress;
> + UINT32 UpperAddress;
> + UINT32 Reserved2;
> +} SD_MMC_HC_ADMA_64_DESC_LINE;
>
> #define SD_MMC_SDMA_BOUNDARY 512 * 1024
> #define SD_MMC_SDMA_ROUND_UP(x, n) (((x) + n) & ~(n - 1))
> @@ -145,6 +165,12 @@ typedef struct {
> #define SD_MMC_HC_CTRL_VER_410 0x04
> #define SD_MMC_HC_CTRL_VER_420 0x05
>
> +//
> +// SD Host controller V4 Support
> +//
> +#define SD_MMC_HC_V4_EN BIT12
> +#define SD_MMC_HC_64_ADDR_EN BIT13
> +
> /**
> Dump the content of SD/MMC host controller's Capability Register.
>
> --
> 2.7.4
>
> _______________________________________________
> edk2-devel mailing list
> edk2-devel@lists.01.org
> https://lists.01.org/mailman/listinfo/edk2-devel
-----------------------------------------------------------------------------------
This email message is for the sole use of the intended recipient(s) and may
contain
confidential information. Any unauthorized review, use, disclosure or
distribution
is prohibited. If you are not the intended recipient, please contact the
sender by
reply email and destroy all copies of the original message.
-----------------------------------------------------------------------------------
_______________________________________________
edk2-devel mailing list
edk2-devel@lists.01.org
https://lists.01.org/mailman/listinfo/edk2-devel
