From: "Pete Batard" <pete@akeo.ie>
To: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: edk2-devel-groups-io <devel@edk2.groups.io>,
Leif Lindholm <leif.lindholm@linaro.org>,
philmd@redhat.com, andrey.warkentin@gmail.com
Subject: Re: [edk2-platforms][PATCH 5/6] Platform/RPi4: Add XHCI and MCFG ACPI tables
Date: Wed, 18 Dec 2019 16:31:28 +0000 [thread overview]
Message-ID: <b1143318-2d8e-08a9-ac49-e2afd19e1e8c@akeo.ie> (raw)
In-Reply-To: <CAKv+Gu_w+c-OVvYq3BRNoRPs0GQ2HXzj7Hu3a5PJj4CGkn9Q3g@mail.gmail.com>
On 2019.12.18 14:55, Ard Biesheuvel wrote:
> On Wed, 18 Dec 2019 at 13:42, Pete Batard <pete@akeo.ie> wrote:
>>
>> From: Andrei Warkentin <andrey.warkentin@gmail.com>
>>
>> Since the RPi4 PCIe host bridge is not ECAM compliant, we cannot
>> expose it as a host bridge to the OS via ACPI, so we add a dummy
>> MCFG table.
>
> I don't think the MCFG table is mandatory, so if the OS complains
> about this, we should fix the OS, not add dummy tables to work around
> it.
No objection here. I'll remove the MCFG altogether from v2.
>
>> However, given the hardwired nature of this platform,
>> we can expose the xHCI controller that is guaranteed to live at
>> the base of the MMIO32 BAR window as a platform device directly.
>>
>> It should be noted that the xHCI table is not finalized at this
>> stage, as Windows xHCI support is still a major question mark.
>>
>> Signed-off-by: Pete Batard <pete@akeo.ie>
>> ---
>> Platform/RaspberryPi/RPi4/AcpiTables/AcpiTables.inf | 5 +
>> Platform/RaspberryPi/RPi4/AcpiTables/Dsdt.asl | 1 +
>> Platform/RaspberryPi/RPi4/AcpiTables/Mcfg.aslc | 81 +++++++++++
>> Platform/RaspberryPi/RPi4/AcpiTables/Xhci.asl | 140 ++++++++++++++++++++
>> 4 files changed, 227 insertions(+)
>>
>> diff --git a/Platform/RaspberryPi/RPi4/AcpiTables/AcpiTables.inf b/Platform/RaspberryPi/RPi4/AcpiTables/AcpiTables.inf
>> index 6b1155d66444..69cb43a2f982 100644
>> --- a/Platform/RaspberryPi/RPi4/AcpiTables/AcpiTables.inf
>> +++ b/Platform/RaspberryPi/RPi4/AcpiTables/AcpiTables.inf
>> @@ -29,6 +29,7 @@ [Sources]
>> Fadt.aslc
>> Dbg2.aslc
>> Gtdt.aslc
>> + Mcfg.aslc
>> Dsdt.asl
>> Csrt.aslc
>> Spcr.aslc
>> @@ -39,6 +40,8 @@ [Packages]
>> EmbeddedPkg/EmbeddedPkg.dec
>> MdeModulePkg/MdeModulePkg.dec
>> MdePkg/MdePkg.dec
>> + Silicon/Broadcom/Bcm27xx/Bcm27xx.dec
>> + Silicon/Broadcom/Bcm283x/Bcm283x.dec
>>
>> [FixedPcd]
>> gArmTokenSpaceGuid.PcdArmArchTimerIntrNum
>> @@ -47,6 +50,8 @@ [FixedPcd]
>> gArmTokenSpaceGuid.PcdArmArchTimerVirtIntrNum
>> gArmTokenSpaceGuid.PcdGicInterruptInterfaceBase
>> gArmTokenSpaceGuid.PcdGicDistributorBase
>> + gBcm27xxTokenSpaceGuid.PcdBcm27xxPciCpuMmioAdr
>> + gBcm27xxTokenSpaceGuid.PcdBcm27xxPciRegBase
>> gEfiMdeModulePkgTokenSpaceGuid.PcdSerialRegisterBase
>> gEfiMdePkgTokenSpaceGuid.PcdUartDefaultBaudRate
>> gEmbeddedTokenSpaceGuid.PcdInterruptBaseAddress
>> diff --git a/Platform/RaspberryPi/RPi4/AcpiTables/Dsdt.asl b/Platform/RaspberryPi/RPi4/AcpiTables/Dsdt.asl
>> index 42e650a3ef29..b2f1d3439211 100644
>> --- a/Platform/RaspberryPi/RPi4/AcpiTables/Dsdt.asl
>> +++ b/Platform/RaspberryPi/RPi4/AcpiTables/Dsdt.asl
>> @@ -22,6 +22,7 @@ DefinitionBlock ("Dsdt.aml", "DSDT", 5, "MSFT", "EDK2", 2)
>> {
>> include ("Sdhc.asl")
>> include ("Pep.asl")
>> + include ("Xhci.asl")
>>
>> Device (CPU0)
>> {
>> diff --git a/Platform/RaspberryPi/RPi4/AcpiTables/Mcfg.aslc b/Platform/RaspberryPi/RPi4/AcpiTables/Mcfg.aslc
>> new file mode 100644
>> index 000000000000..7315e2a88c91
>> --- /dev/null
>> +++ b/Platform/RaspberryPi/RPi4/AcpiTables/Mcfg.aslc
>> @@ -0,0 +1,81 @@
>> +/** @file
>> + *
>> + * Memory Mapped Configuration Address Space table (MCFG)
>> + *
>> + * Copyright (c) 2019, Pete Batard <pete@akeo.ie>
>> + * Copyright (c) 2013-2015 Intel Corporation.
>> + *
>> + * SPDX-License-Identifier: BSD-2-Clause-Patent
>> + *
>> + **/
>> +
>> +#include <IndustryStandard/Acpi.h>
>> +#include <IndustryStandard/MemoryMappedConfigurationSpaceAccessTable.h>
>> +
>> +#include "AcpiTables.h"
>> +
>> +//
>> +// Multiple APIC Description Table
>> +//
>> +#pragma pack (1)
>> +
>> +#define RPI_ACPI_OEM_MCFG_REVISION 0x00000001
>> +//
>> +// The Pi 4 is not ECAM compliant so, ideally, we would just skip populating
>> +// an allocation structure. However, GenFw throws 'MCFG length check failed'
>> +// when the number of allocation structures is zero, so we need at least one.
>> +//
>> +#define RPI_ACPI_ALLOCATION_STRUCTURE_COUNT 1
>> +
>> +typedef struct {
>> + EFI_ACPI_MEMORY_MAPPED_CONFIGURATION_BASE_ADDRESS_TABLE_HEADER Header;
>> + EFI_ACPI_MEMORY_MAPPED_ENHANCED_CONFIGURATION_SPACE_BASE_ADDRESS_ALLOCATION_STRUCTURE AllocationStructure[RPI_ACPI_ALLOCATION_STRUCTURE_COUNT];
>> +} EFI_ACPI_MEMORY_MAPPED_CONFIGURATION_SPACE_ACCESS_DESCRIPTION_TABLE;
>> +
>> +STATIC EFI_ACPI_MEMORY_MAPPED_CONFIGURATION_SPACE_ACCESS_DESCRIPTION_TABLE Mcfg =
>> +{
>> + {
>> + //------------------------------------------------------------------------
>> + // ACPI Table Header
>> + //------------------------------------------------------------------------
>> + {
>> + EFI_ACPI_3_0_PCI_EXPRESS_MEMORY_MAPPED_CONFIGURATION_SPACE_BASE_ADDRESS_DESCRIPTION_TABLE_SIGNATURE,
>> + sizeof (EFI_ACPI_MEMORY_MAPPED_CONFIGURATION_SPACE_ACCESS_DESCRIPTION_TABLE),
>> + EFI_ACPI_MEMORY_MAPPED_CONFIGURATION_SPACE_ACCESS_TABLE_REVISION,
>> + 0x00, // Checksum calculated at runtime.
>> + EFI_ACPI_OEM_ID, // OEMID is a 6 bytes long field "BC2836"
>> + EFI_ACPI_OEM_TABLE_ID, // OEM table identification(8 bytes long) "RPI4EDK2"
>> + EFI_ACPI_OEM_REVISION, // OEM revision number
>> + EFI_ACPI_CREATOR_ID, // ASL compiler vendor ID
>> + EFI_ACPI_CREATOR_REVISION // ASL compiler revision number
>> + },
>> + //------------------------------------------------------------------------
>> + // Reserved
>> + //------------------------------------------------------------------------
>> + 0x0000000000000000, // Reserved
>> + },
>> +
>> + //------------------------------------------------------------------------
>> + // MCFG specific fields
>> + //------------------------------------------------------------------------
>> + {
>> + {
>> + //
>> + // Using (-1) as the base address tells the OS to ignore it.
>> + //
>> + 0xFFFFFFFFFFFFFFFFULL, // BaseAddress
>> + 0x0000, // PciSegmentGroupNumber
>> + 0x00, // StartBusNumber
>> + 0x00, // EndBusNumber
>> + 0x00000000 // Reserved
>> + }
>> + }
>> +};
>> +
>> +#pragma pack()
>> +
>> +//
>> +// Reference the table being generated to prevent the optimizer from removing
>> +// the data structure from the executable
>> +//
>> +VOID* CONST ReferenceAcpiTable = &Mcfg;
>> diff --git a/Platform/RaspberryPi/RPi4/AcpiTables/Xhci.asl b/Platform/RaspberryPi/RPi4/AcpiTables/Xhci.asl
>> new file mode 100644
>> index 000000000000..1eaeb7ba451b
>> --- /dev/null
>> +++ b/Platform/RaspberryPi/RPi4/AcpiTables/Xhci.asl
>> @@ -0,0 +1,140 @@
>> +/** @file
>> + *
>> + * Copyright (c) 2019 Linaro, Limited. All rights reserved.
>> + * Copyright (c) 2019 Andrei Warkentin <andrey.warkentin@gmail.com>
>> + *
>> + * SPDX-License-Identifier: BSD-2-Clause-Patent
>> + *
>> + **/
>> +
>> +#include <IndustryStandard/Bcm2711.h>
>> +
>> +/*
>> + * The following can be used to remove parenthesis from
>> + * defined macros that the compiler complains about.
>> + */
>> +#define _REMOVE_PAREN(...) __VA_ARGS__
>> +#define REMOVE_PAREN(x) _REMOVE_PAREN x
>> +
>> +/*
>> + * According to UEFI boot log for the VLI device on Pi 4.
>> + */
>> +#define XHCI_REG_LENGTH 0x1000
>> +
>> +Device (SCB0) {
>> + Name (_HID, "ACPI0004")
>> + Name (_UID, 0x0)
>> + Name (_CCA, 0x0)
>> +
>> + Method (_CRS, 0, Serialized) { // _CRS: Current Resource Settings
>> + /*
>> + * Container devices with _DMA must have _CRS, meaning SCB0
>> + * to provide all resources that XHC0 consumes (except
>> + * interrupts).
>> + */
>> + Name (RBUF, ResourceTemplate () {
>> + QWordMemory(ResourceProducer,
>> + ,
>> + MinFixed,
>> + MaxFixed,
>> + NonCacheable,
>> + ReadWrite,
>> + 0x0,
>> + 0xAAAA, // MIN
>> + 0xBBBA, // MAX
>> + 0x0,
>> + 0x1111, // LEN
>
> Could we make this slightly less ugly, by setting MIN and MAX to
> PCIE_CPU_MMIO_WINDOW and LEN to 1, and only updating MAX and LEN below
> by adding (XHCI_REG_LENGTH - 1) to them?
I'll do that in v2, for both entries.
Adding Andrei in CC just in case he wants to comment on that request.
Regards,
/Pete
>
>> + ,
>> + ,
>> + MMIO
>> + )
>> + })
>> + CreateQwordField (RBUF, MMIO._MIN, MMBA)
>> + CreateQwordField (RBUF, MMIO._MAX, MMBE)
>> + CreateQwordField (RBUF, MMIO._LEN, MMLE)
>> + Store (REMOVE_PAREN(PCIE_CPU_MMIO_WINDOW), MMBA)
>> + Store (REMOVE_PAREN(PCIE_CPU_MMIO_WINDOW), MMBE)
>> + Store (XHCI_REG_LENGTH, MMLE)
>> + Add (MMBA, MMLE, MMBE)
>> + Return (RBUF)
>> + }
>> +
>> + Name (_DMA, ResourceTemplate() {
>> + /*
>> + * XHC0 is limited to DMA to first 3GB. Note this
>> + * only applies to PCIe, not GENET or other devices
>> + * next to the A72.
>> + */
>> + QWordMemory(ResourceConsumer,
>> + ,
>> + MinFixed,
>> + MaxFixed,
>> + NonCacheable,
>> + ReadWrite,
>> + 0x0,
>> + 0x0, // MIN
>> + 0xbfffffff, // MAX
>> + 0x0, // TRA
>> + 0xc0000000, // LEN
>> + ,
>> + ,
>> + )
>> + })
>> +
>> + Device (XHC0)
>> + {
>> + Name (_HID, "11063483") // _HID: Hardware ID
>> + Name (_CID, "PNP0D10") // _CID: Hardware ID
>> + Name (_UID, 0x0) // _UID: Unique ID
>> + Name (_CCA, 0x0) // _CCA: Cache Coherency Attribute
>> +
>> + Method (_CRS, 0, Serialized) { // _CRS: Current Resource Settings
>> + Name (RBUF, ResourceTemplate () {
>> + QWordMemory(ResourceConsumer,
>> + ,
>> + MinFixed,
>> + MaxFixed,
>> + NonCacheable,
>> + ReadWrite,
>> + 0x0,
>> + 0xAAAA, // MIN
>> + 0xBBBA, // MAX
>> + 0x0,
>> + 0x1111, // LEN
>
> Same here.
>
>> + ,
>> + ,
>> + MMIO
>> + )
>> + Interrupt (ResourceConsumer, Level, ActiveHigh, Exclusive, ,, ) {
>> + 175
>> + }
>> + })
>> + CreateQwordField (RBUF, MMIO._MIN, MMBA)
>> + CreateQwordField (RBUF, MMIO._MAX, MMBE)
>> + CreateQwordField (RBUF, MMIO._LEN, MMLE)
>> + Store (REMOVE_PAREN(PCIE_CPU_MMIO_WINDOW), MMBA)
>> + Store (REMOVE_PAREN(PCIE_CPU_MMIO_WINDOW), MMBE)
>> + Store (XHCI_REG_LENGTH, MMLE)
>> + Add (MMBA, MMLE, MMBE)
>> + Return (RBUF)
>> + }
>> +
>> + Method (_INI, 0, Serialized) {
>> + OperationRegion (PCFG, SystemMemory, REMOVE_PAREN(PCIE_REG_BASE) + PCIE_EXT_CFG_DATA, 0x1000)
>> + Field (PCFG, AnyAcc, NoLock, Preserve) {
>> + Offset (0),
>> + VNID, 16, // Vendor ID
>> + DVID, 16, // Device ID
>> + CMND, 16, // Command register
>> + STAT, 16, // Status register
>> + }
>> +
>> + // Set command register to:
>> + // 1) decode MMIO (set bit 1)
>> + // 2) enable DMA (set bit 2)
>> + // 3) enable interrupts (clear bit 10)
>> + Debug = "xHCI enable"
>> + Store (0x6, CMND)
>> + }
>> + }
>> +}
>> --
>> 2.21.0.windows.1
>>
next prev parent reply other threads:[~2019-12-18 16:31 UTC|newest]
Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-12-18 11:41 [edk2-platforms][PATCH 0/6] Platform/RPi4: ACPI improvements Pete Batard
2019-12-18 11:41 ` [edk2-platforms][PATCH 1/6] Platform/RPi4: Clean up ACPI definitions Pete Batard
2019-12-18 11:41 ` [edk2-platforms][PATCH 2/6] Platform/RPi4: Improve FADT ACPI table generation Pete Batard
2019-12-18 14:46 ` Ard Biesheuvel
2019-12-18 16:31 ` Pete Batard
2019-12-18 11:41 ` [edk2-platforms][PATCH 3/6] Platform/RPi4: Improve SPCR and DBG2 " Pete Batard
2019-12-18 15:57 ` Philippe Mathieu-Daudé
2019-12-18 16:36 ` Pete Batard
2019-12-18 17:00 ` Philippe Mathieu-Daudé
2019-12-18 17:10 ` Pete Batard
2019-12-18 11:41 ` [edk2-platforms][PATCH 4/6] Platform/RPi4: Add switch to select between PL011 and miniUART Pete Batard
2019-12-18 16:05 ` Philippe Mathieu-Daudé
2019-12-18 16:59 ` [edk2-devel] " Pete Batard
2019-12-18 17:05 ` Philippe Mathieu-Daudé
2019-12-18 11:41 ` [edk2-platforms][PATCH 5/6] Platform/RPi4: Add XHCI and MCFG ACPI tables Pete Batard
2019-12-18 14:55 ` Ard Biesheuvel
2019-12-18 16:31 ` Pete Batard [this message]
2019-12-18 11:41 ` [edk2-platforms][PATCH 6/6] Platform/RPi4: Add ACPI basic mode build option Pete Batard
2019-12-18 12:10 ` Leif Lindholm
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-list from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=b1143318-2d8e-08a9-ac49-e2afd19e1e8c@akeo.ie \
--to=devel@edk2.groups.io \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox