From: "Patrick Rudolph" <patrick.rudolph@9elements.com>
To: "Dong, Guo" <guo.dong@intel.com>
Cc: "devel@edk2.groups.io" <devel@edk2.groups.io>,
"Ma, Maurice" <maurice.ma@intel.com>,
"You, Benjamin" <benjamin.you@intel.com>
Subject: Re: [edk2-devel] [PATCH] UefiPayloadPkg/UefiPayloadEntry: Improve bootloader memrange parsing
Date: Wed, 16 Jun 2021 08:25:27 +0200 [thread overview]
Message-ID: <CALNFmy2utSJP9kjT+yWDvYSYC4Lr=T0Ho5d3SAL=J-rM96izTQ@mail.gmail.com> (raw)
In-Reply-To: <BYAPR11MB362208AB11F12539C85BA1439E309@BYAPR11MB3622.namprd11.prod.outlook.com>
Hi Guo,
The PciHostBridgeDxe expects the PCI Aperature and MMCONF to be marked
as EfiMemoryMappedIO,
however as the bootloader provides an e820 compatible memory map, it's
actually marked as
EfiReservedMemoryType. The PciHostBridgeDxe asserts on this incorrect
memory type.
This is what the patch tries to fix by marking MMIO (between TOLUD and
4GiB) as EfiMemoryMappedIO.
I don't think that the mentioned patch (UefiPayloadPkg: UefiPayload
retrieve PCI root bridge
from Guid Hob) would change that, as the same memory ranges are still
being incorrectly marked
reserved by UefiPayloadEntry.
Even with TOLUD guessing not working the OS wouldn't see a difference
as EfiMemoryMappedIO and
EfiReservedMemoryType both appear as reserved to the OS, so it's still
the same memory from OS point of view.
In case TOLUD is estimated too high, MMIO would be marked as Reserved,
so it's the same situation as now and
DXE drivers might fail.
In case TOLUD is estimated too low, Reserved DRAM would be marked as
MMIO, however I can't think of a DXE
that would actually care, it would skip the region as it would be
marked reserved.
Thanks,
Patrick
On Wed, Jun 16, 2021 at 12:20 AM Dong, Guo <guo.dong@intel.com> wrote:
>
>
> Could you provide more info on the issues in fixed by this patch?
> There is a UEFI payload patch (under code review) to support the bootloader to provide a HOB
> on the PCI resources, this way the UEFI payload doesn't need collect resources again by parsing
> all the PCI devices. Not sure if that patch (UefiPayloadPkg: UefiPayload retrieve PCI root bridge
> from Guid Hob) could avoid the issues.
>
> In this patch, as it mentioned some information (system memory or MMIO) is missing in the
> bootloader E820 style table. Guessing the TOLUD might work in most cases, but may fail since
> the assumptions.
>
> I saw this patch fixed the memory type for ACPI memory. I think we could have this fix if the
> TOLUD guessing could be avoided.
>
> Thanks,
> Guo
>
> > -----Original Message-----
> > From: devel@edk2.groups.io <devel@edk2.groups.io> On Behalf Of Patrick
> > Rudolph
> > Sent: Tuesday, June 15, 2021 6:23 AM
> > To: devel@edk2.groups.io
> > Cc: Ma, Maurice <maurice.ma@intel.com>; Dong, Guo
> > <guo.dong@intel.com>; You, Benjamin <benjamin.you@intel.com>
> > Subject: [edk2-devel] [PATCH] UefiPayloadPkg/UefiPayloadEntry: Improve
> > bootloader memrange parsing
> >
> > Currently several DXE crash due to invalid memory resource settings.
> > coreboot and slimbootloader provide an e820 compatible memory map,
> > which doesn't work well with EDK2 as the e820 spec is missing MMIO regions.
> > In e820 'reserved' could either mean "DRAM used by boot firmware" or
> > "MMIO
> > in use and not detectable by OS".
> >
> > Guess Top of lower usable DRAM (TOLUD) by walking memory ranges and
> > then
> > mark everything reserved below TOLUD as DRAM and everything reserved
> > above
> > TOLUD as MMIO.
> >
> > This fixes several assertions seen in PciHostBridgeDxe.
> >
> > Signed-off-by: Patrick Rudolph <patrick.rudolph@9elements.com>
> > ---
> > .../UefiPayloadEntry/UefiPayloadEntry.c | 187 +++++++++++++++++-
> > .../UefiPayloadEntry/UefiPayloadEntry.h | 10 +
> > 2 files changed, 194 insertions(+), 3 deletions(-)
> >
> > diff --git a/UefiPayloadPkg/UefiPayloadEntry/UefiPayloadEntry.c
> > b/UefiPayloadPkg/UefiPayloadEntry/UefiPayloadEntry.c
> > index 805f5448d9..d20e1a0862 100644
> > --- a/UefiPayloadPkg/UefiPayloadEntry/UefiPayloadEntry.c
> > +++ b/UefiPayloadPkg/UefiPayloadEntry/UefiPayloadEntry.c
> > @@ -7,10 +7,162 @@
> >
> >
> > #include "UefiPayloadEntry.h"
> >
> >
> >
> > +STATIC UINT32 TopOfLowerUsableDram = 0;
> >
> > +
> >
> > /**
> >
> > Callback function to build resource descriptor HOB
> >
> >
> >
> > This function build a HOB based on the memory map entry info.
> >
> > + It creates only EFI_RESOURCE_MEMORY_MAPPED_IO and
> > EFI_RESOURCE_MEMORY_RESERVED
> >
> > + resources.
> >
> > +
> >
> > + @param MemoryMapEntry Memory map entry info got from
> > bootloader.
> >
> > + @param Params A pointer to ACPI_BOARD_INFO.
> >
> > +
> >
> > + @retval RETURN_SUCCESS Successfully build a HOB.
> >
> > + @retval EFI_INVALID_PARAMETER Invalid parameter provided.
> >
> > +**/
> >
> > +EFI_STATUS
> >
> > +MemInfoCallbackMMIO (
> >
> > + IN MEMROY_MAP_ENTRY *MemoryMapEntry,
> >
> > + IN VOID *Params
> >
> > + )
> >
> > +{
> >
> > + EFI_PHYSICAL_ADDRESS Base;
> >
> > + EFI_RESOURCE_TYPE Type;
> >
> > + UINT64 Size;
> >
> > + EFI_RESOURCE_ATTRIBUTE_TYPE Attribue;
> >
> > + ACPI_BOARD_INFO *AcpiBoardInfo;
> >
> > +
> >
> > + AcpiBoardInfo = (ACPI_BOARD_INFO *)Params;
> >
> > + if (AcpiBoardInfo == NULL) {
> >
> > + return EFI_INVALID_PARAMETER;
> >
> > + }
> >
> > +
> >
> > + //
> >
> > + // Skip types already handled in MemInfoCallback
> >
> > + //
> >
> > + if (MemoryMapEntry->Type == E820_RAM || MemoryMapEntry->Type ==
> > E820_ACPI) {
> >
> > + return RETURN_SUCCESS;
> >
> > + }
> >
> > +
> >
> > + if (MemoryMapEntry->Base == AcpiBoardInfo->PcieBaseAddress) {
> >
> > + //
> >
> > + // MMCONF is always MMIO
> >
> > + //
> >
> > + Type = EFI_RESOURCE_MEMORY_MAPPED_IO;
> >
> > + } else if (MemoryMapEntry->Base < TopOfLowerUsableDram) {
> >
> > + //
> >
> > + // It's in DRAM and thus must be reserved
> >
> > + //
> >
> > + Type = EFI_RESOURCE_MEMORY_RESERVED;
> >
> > + } else if (MemoryMapEntry->Base < 0x100000000ULL &&
> >
> > + MemoryMapEntry->Base >= TopOfLowerUsableDram) {
> >
> > + //
> >
> > + // It's not in DRAM, must be MMIO
> >
> > + //
> >
> > + Type = EFI_RESOURCE_MEMORY_MAPPED_IO;
> >
> > + } else {
> >
> > + Type = EFI_RESOURCE_MEMORY_RESERVED;
> >
> > + }
> >
> > +
> >
> > + Base = MemoryMapEntry->Base;
> >
> > + Size = MemoryMapEntry->Size;
> >
> > +
> >
> > + Attribue = EFI_RESOURCE_ATTRIBUTE_PRESENT |
> >
> > + EFI_RESOURCE_ATTRIBUTE_INITIALIZED |
> >
> > + EFI_RESOURCE_ATTRIBUTE_TESTED |
> >
> > + EFI_RESOURCE_ATTRIBUTE_UNCACHEABLE |
> >
> > + EFI_RESOURCE_ATTRIBUTE_WRITE_COMBINEABLE |
> >
> > + EFI_RESOURCE_ATTRIBUTE_WRITE_THROUGH_CACHEABLE |
> >
> > + EFI_RESOURCE_ATTRIBUTE_WRITE_BACK_CACHEABLE;
> >
> > +
> >
> > + BuildResourceDescriptorHob (Type, Attribue,
> > (EFI_PHYSICAL_ADDRESS)Base, Size);
> >
> > + DEBUG ((DEBUG_INFO , "buildhob: base = 0x%lx, size = 0x%lx, type =
> > 0x%x\n", Base, Size, Type));
> >
> > +
> >
> > + if (MemoryMapEntry->Type == E820_NVS) {
> >
> > + BuildMemoryAllocationHob (Base, Size, EfiACPIMemoryNVS);
> >
> > + } else if (MemoryMapEntry->Type == E820_UNUSABLE ||
> >
> > + MemoryMapEntry->Type == E820_DISABLED) {
> >
> > + BuildMemoryAllocationHob (Base, Size, EfiUnusableMemory);
> >
> > + } else if (MemoryMapEntry->Type == E820_PMEM) {
> >
> > + BuildMemoryAllocationHob (Base, Size, EfiPersistentMemory);
> >
> > + }
> >
> > +
> >
> > + return RETURN_SUCCESS;
> >
> > +}
> >
> > +
> >
> > +
> >
> > +/**
> >
> > + Callback function to find TOLUD (Top of Lower Usable DRAM)
> >
> > +
> >
> > + Estimate where TOLUD (Top of Lower Usable DRAM) resides. The exact
> > position
> >
> > + would require platform specific code.
> >
> > +
> >
> > + @param MemoryMapEntry Memory map entry info got from
> > bootloader.
> >
> > + @param Params Not used for now.
> >
> > +
> >
> > + @retval RETURN_SUCCESS Successfully updated
> > TopOfLowerUsableDram.
> >
> > +**/
> >
> > +EFI_STATUS
> >
> > +FindToludCallback (
> >
> > + IN MEMROY_MAP_ENTRY *MemoryMapEntry,
> >
> > + IN VOID *Params
> >
> > + )
> >
> > +{
> >
> > + //
> >
> > + // This code assumes that the memory map on this x86 machine below
> > 4GiB is continous
> >
> > + // until TOLUD. In addition it assumes that the bootloader provided
> > memory tables have
> >
> > + // no "holes" and thus the first memory range not covered by e820 marks
> > the end of
> >
> > + // usable DRAM. In addition it's assumed that every reserved memory
> > region touching
> >
> > + // usable RAM is also covering DRAM, everything else that is marked
> > reserved thus must be
> >
> > + // MMIO not detectable by bootloader/OS
> >
> > + //
> >
> > +
> >
> > + //
> >
> > + // Skip memory types not RAM or reserved
> >
> > + //
> >
> > + if (MemoryMapEntry->Type == E820_NVS || MemoryMapEntry->Type ==
> > E820_UNUSABLE ||
> >
> > + MemoryMapEntry->Type == E820_DISABLED || MemoryMapEntry->Type
> > == E820_PMEM) {
> >
> > + return RETURN_SUCCESS;
> >
> > + }
> >
> > +
> >
> > + //
> >
> > + // Skip resources above 4GiB
> >
> > + //
> >
> > + if (MemoryMapEntry->Base >= 0x100000000ULL) {
> >
> > + return RETURN_SUCCESS;
> >
> > + }
> >
> > +
> >
> > + if ((MemoryMapEntry->Type == E820_RAM) ||
> >
> > + (MemoryMapEntry->Type == E820_ACPI)) {
> >
> > + //
> >
> > + // It's usable DRAM. Update TOLUD.
> >
> > + //
> >
> > + if (TopOfLowerUsableDram < (MemoryMapEntry->Base +
> > MemoryMapEntry->Size)) {
> >
> > + TopOfLowerUsableDram = MemoryMapEntry->Base +
> > MemoryMapEntry->Size;
> >
> > + }
> >
> > + } else {
> >
> > + //
> >
> > + // It might be reserved DRAM or MMIO.
> >
> > + //
> >
> > + // If it touches usable DRAM at Base assume it's DRAM as well,
> >
> > + // as it could be bootloader installed tables, TSEG, GTT, ...
> >
> > + //
> >
> > + if (TopOfLowerUsableDram == MemoryMapEntry->Base) {
> >
> > + TopOfLowerUsableDram = MemoryMapEntry->Base +
> > MemoryMapEntry->Size;
> >
> > + }
> >
> > + }
> >
> > +
> >
> > + return RETURN_SUCCESS;
> >
> > +}
> >
> > +
> >
> > +
> >
> > +/**
> >
> > + Callback function to build resource descriptor HOB
> >
> > +
> >
> > + This function build a HOB based on the memory map entry info.
> >
> > + Only add EFI_RESOURCE_SYSTEM_MEMORY.
> >
> >
> >
> > @param MemoryMapEntry Memory map entry info got from
> > bootloader.
> >
> > @param Params Not used for now.
> >
> > @@ -28,7 +180,15 @@ MemInfoCallback (
> > UINT64 Size;
> >
> > EFI_RESOURCE_ATTRIBUTE_TYPE Attribue;
> >
> >
> >
> > - Type = (MemoryMapEntry->Type == 1) ?
> > EFI_RESOURCE_SYSTEM_MEMORY : EFI_RESOURCE_MEMORY_RESERVED;
> >
> > + //
> >
> > + // Skip everything not known to be usable DRAM.
> >
> > + // It will be added later.
> >
> > + //
> >
> > + if (MemoryMapEntry->Type != E820_RAM && MemoryMapEntry->Type !=
> > E820_ACPI) {
> >
> > + return RETURN_SUCCESS;
> >
> > + }
> >
> > +
> >
> > + Type = EFI_RESOURCE_SYSTEM_MEMORY;
> >
> > Base = MemoryMapEntry->Base;
> >
> > Size = MemoryMapEntry->Size;
> >
> >
> >
> > @@ -40,7 +200,7 @@ MemInfoCallback (
> > EFI_RESOURCE_ATTRIBUTE_WRITE_THROUGH_CACHEABLE |
> >
> > EFI_RESOURCE_ATTRIBUTE_WRITE_BACK_CACHEABLE;
> >
> >
> >
> > - if (Base >= BASE_4GB ) {
> >
> > + if (Base >= BASE_4GB) {
> >
> > // Remove tested attribute to avoid DXE core to dispatch driver to memory
> > above 4GB
> >
> > Attribue &= ~EFI_RESOURCE_ATTRIBUTE_TESTED;
> >
> > }
> >
> > @@ -48,6 +208,10 @@ MemInfoCallback (
> > BuildResourceDescriptorHob (Type, Attribue,
> > (EFI_PHYSICAL_ADDRESS)Base, Size);
> >
> > DEBUG ((DEBUG_INFO , "buildhob: base = 0x%lx, size = 0x%lx, type =
> > 0x%x\n", Base, Size, Type));
> >
> >
> >
> > + if (MemoryMapEntry->Type == E820_ACPI) {
> >
> > + BuildMemoryAllocationHob (Base, Size, EfiACPIReclaimMemory);
> >
> > + }
> >
> > +
> >
> > return RETURN_SUCCESS;
> >
> > }
> >
> >
> >
> > @@ -236,7 +400,16 @@ BuildHobFromBl (
> > EFI_PEI_GRAPHICS_DEVICE_INFO_HOB *NewGfxDeviceInfo;
> >
> >
> >
> > //
> >
> > - // Parse memory info and build memory HOBs
> >
> > + // First find TOLUD
> >
> > + //
> >
> > + Status = ParseMemoryInfo (FindToludCallback, NULL);
> >
> > + if (EFI_ERROR(Status)) {
> >
> > + return Status;
> >
> > + }
> >
> > + DEBUG ((DEBUG_INFO , "Assuming TOLUD = 0x%x\n",
> > TopOfLowerUsableDram));
> >
> > +
> >
> > + //
> >
> > + // Parse memory info and build memory HOBs for Usable RAM
> >
> > //
> >
> > Status = ParseMemoryInfo (MemInfoCallback, NULL);
> >
> > if (EFI_ERROR(Status)) {
> >
> > @@ -289,6 +462,14 @@ BuildHobFromBl (
> > DEBUG ((DEBUG_INFO, "Create acpi board info guid hob\n"));
> >
> > }
> >
> >
> >
> > + //
> >
> > + // Parse memory info and build memory HOBs for reserved DRAM and
> > MMIO
> >
> > + //
> >
> > + Status = ParseMemoryInfo (MemInfoCallbackMMIO, &AcpiBoardInfo);
> >
> > + if (EFI_ERROR(Status)) {
> >
> > + return Status;
> >
> > + }
> >
> > +
> >
> > //
> >
> > // Parse platform specific information.
> >
> > //
> >
> > diff --git a/UefiPayloadPkg/UefiPayloadEntry/UefiPayloadEntry.h
> > b/UefiPayloadPkg/UefiPayloadEntry/UefiPayloadEntry.h
> > index 2c84d6ed53..35ea23d202 100644
> > --- a/UefiPayloadPkg/UefiPayloadEntry/UefiPayloadEntry.h
> > +++ b/UefiPayloadPkg/UefiPayloadEntry/UefiPayloadEntry.h
> > @@ -38,6 +38,16 @@
> > #define GET_OCCUPIED_SIZE(ActualSize, Alignment) \
> >
> > ((ActualSize) + (((Alignment) - ((ActualSize) & ((Alignment) - 1))) &
> > ((Alignment) - 1)))
> >
> >
> >
> > +
> >
> > +#define E820_RAM 1
> >
> > +#define E820_RESERVED 2
> >
> > +#define E820_ACPI 3
> >
> > +#define E820_NVS 4
> >
> > +#define E820_UNUSABLE 5
> >
> > +#define E820_DISABLED 6
> >
> > +#define E820_PMEM 7
> >
> > +#define E820_UNDEFINED 8
> >
> > +
> >
> > /**
> >
> > Auto-generated function that calls the library constructors for all of the
> > module's
> >
> > dependent libraries.
> >
> > --
> > 2.30.2
> >
> >
> >
> > -=-=-=-=-=-=
> > Groups.io Links: You receive all messages sent to this group.
> > View/Reply Online (#76517): https://edk2.groups.io/g/devel/message/76517
> > Mute This Topic: https://groups.io/mt/83555387/1781375
> > Group Owner: devel+owner@edk2.groups.io
> > Unsubscribe: https://edk2.groups.io/g/devel/unsub [guo.dong@intel.com]
> > -=-=-=-=-=-=
> >
>
next prev parent reply other threads:[~2021-06-16 6:25 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-06-15 13:22 [PATCH] UefiPayloadPkg/UefiPayloadEntry: Improve bootloader memrange parsing Patrick Rudolph
2021-06-15 22:20 ` [edk2-devel] " Guo Dong
2021-06-16 6:25 ` Patrick Rudolph [this message]
2021-06-17 0:32 ` Ma, Maurice
2021-06-17 13:20 ` Patrick Rudolph
2021-06-17 13:51 ` Ma, Maurice
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='CALNFmy2utSJP9kjT+yWDvYSYC4Lr=T0Ho5d3SAL=J-rM96izTQ@mail.gmail.com' \
--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