public inbox for devel@edk2.groups.io
 help / color / mirror / Atom feed
From: Ard Biesheuvel <ard.biesheuvel@linaro.org>
To: Laszlo Ersek <lersek@redhat.com>
Cc: edk2-devel-01 <edk2-devel@lists.01.org>,
	Gerd Hoffmann <kraxel@redhat.com>,
	Igor Mammedov <imammedo@redhat.com>,
	Jordan Justen <jordan.l.justen@intel.com>
Subject: Re: [PATCH 1/1] OvmfPkg/AcpiPlatformDxe: alloc blobs from 64-bit space unless restricted
Date: Sat, 3 Jun 2017 18:33:04 +0000	[thread overview]
Message-ID: <CAKv+Gu_QaVGXq6EYkzT+C5cwLXQbMRQqYsQsO0yxxfZFwcbo4A@mail.gmail.com> (raw)
In-Reply-To: <20170603154203.29907-2-lersek@redhat.com>

On 3 June 2017 at 15:42, Laszlo Ersek <lersek@redhat.com> wrote:
> ... by narrower than 8-byte ADD_POINTER references.
>
> Introduce the CollectRestrictedAllocations() function, which iterates over
> the linker/loader script, and collects the names of the fw_cfg blobs that
> are referenced by QEMU_LOADER_ADD_POINTER.PointeeFile fields, such that
> QEMU_LOADER_ADD_POINTER.PointerSize is less than 8. This means that the
> pointee blob's address will have to be patched into a narrower-than-8 byte
> pointer field, hence the pointee blob must not be allocated from 64-bit
> address space.
>
> In ProcessCmdAllocate(), consult these restrictions when setting the
> maximum address for gBS->AllocatePages(). The default is now MAX_UINT64,
> unless restricted like described above to the pre-patch MAX_UINT32 limit.
>
> In combination with Ard's QEMU commit cb51ac2ffe36 ("hw/arm/virt: generate
> 64-bit addressable ACPI objects", 2017-04-10), this patch enables
> OvmfPkg/AcpiPlatformDxe to work entirely above the 4GB mark.
>
> (An upcoming / planned aarch64 QEMU machine type will have no RAM under
> 4GB at all. Plus, moving the allocations higher is beneficial to the
> current "virt" machine type as well; in Ard's words: "having all firmware
> allocations inside the same 1 GB (or 512 MB for 64k pages) frame reduces
> the TLB footprint".)
>
> Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
> Cc: Gerd Hoffmann <kraxel@redhat.com>
> Cc: Igor Mammedov <imammedo@redhat.com>
> Cc: Jordan Justen <jordan.l.justen@intel.com>
> Suggested-by: Igor Mammedov <imammedo@redhat.com>
> Suggested-by: Gerd Hoffmann <kraxel@redhat.com>
> Contributed-under: TianoCore Contribution Agreement 1.0
> Signed-off-by: Laszlo Ersek <lersek@redhat.com>

Thanks Laszlo

This works as expected for me. The ACPI reclaim regions happily reside
beyond the 4 GB mark on mach-virt. I haven't tried the 'enterprise'
flavor machine, but given that it does not actually exist yet, there
are no compatibility concerns, and so the behavior observed on
mach-virt should extrapolate to that machine as well.

Tested-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>

I am quite pleased with this solution, given that it should seamlessly
interoperate with QEMU versions with or without patch cb51ac2ffe36.
One realization I did make due to my involvement in this discussion is
that ACPI reclaim regions are omitted from the linear mapping on arm64
for no good reason: even though we preserve such regions (for kexec or
other 'late' references to their contents), the fact that the firmware
allows them to be reclaimed implies that they have no special
significance to the firmware, and so there is really no good reason to
punch them out of the linear mapping. So while the [modest] TLB
footprint concern* does apply to mach-virt currently, I intend to
propose a change to the arm64 kernel to deal with ACPI reclaim regions
differently, after which this benefit no longer exists.

Thanks for all the work.

Ard.


* The default arm64 defconfig defines 4 levels of stage 1 page tables,
and KVM uses 3 levels with concatenation at level one, resulting in a
48-bit VA space and a 40-bit IPA space. Under virt, each level of
lookup at stage 1 may result in a full page table walk at stage 2, and
so it is generally wise to use as few levels as possible. ARM is
finicky about aliased mappings with mismatched attributes, and so we
omit 'special' memory regions from the linear mapping, which means the
remaining DRAM in the same 1 GB frame cannot be mapped as efficiently.
This is why it is preferable to keep all such regions close together.


> ---
>  OvmfPkg/AcpiPlatformDxe/QemuFwCfgAcpi.c | 190 +++++++++++++++++++-
>  1 file changed, 183 insertions(+), 7 deletions(-)
>
> diff --git a/OvmfPkg/AcpiPlatformDxe/QemuFwCfgAcpi.c b/OvmfPkg/AcpiPlatformDxe/QemuFwCfgAcpi.c
> index 1bc5fe297a96..7aeb6472188b 100644
> --- a/OvmfPkg/AcpiPlatformDxe/QemuFwCfgAcpi.c
> +++ b/OvmfPkg/AcpiPlatformDxe/QemuFwCfgAcpi.c
> @@ -132,14 +132,169 @@ PointerCompare (
>  }
>
>
>  /**
> +  Comparator function for two ASCII strings. Can be used as both Key and
> +  UserStruct comparator.
> +
> +  This function exists solely so we can avoid casting &AsciiStrCmp to
> +  ORDERED_COLLECTION_USER_COMPARE and ORDERED_COLLECTION_KEY_COMPARE.
> +
> +  @param[in] AsciiString1  Pointer to the first ASCII string.
> +
> +  @param[in] AsciiString2  Pointer to the second ASCII string.
> +
> +  @return  The return value of AsciiStrCmp (AsciiString1, AsciiString2).
> +**/
> +STATIC
> +INTN
> +EFIAPI
> +AsciiStringCompare (
> +  IN CONST VOID *AsciiString1,
> +  IN CONST VOID *AsciiString2
> +  )
> +{
> +  return AsciiStrCmp (AsciiString1, AsciiString2);
> +}
> +
> +
> +/**
> +  Release the ORDERED_COLLECTION structure populated by
> +  CollectRestrictedAllocations() (below).
> +
> +  This function may be called by CollectRestrictedAllocations() itself, on
> +  the error path.
> +
> +  @param[in] RestrictedAllocations  The ORDERED_COLLECTION structure to
> +                                    release.
> +**/
> +STATIC
> +VOID
> +ReleaseRestrictedAllocations (
> +  IN ORDERED_COLLECTION *RestrictedAllocations
> +)
> +{
> +  ORDERED_COLLECTION_ENTRY *Entry, *Entry2;
> +
> +  for (Entry = OrderedCollectionMin (RestrictedAllocations);
> +       Entry != NULL;
> +       Entry = Entry2) {
> +    Entry2 = OrderedCollectionNext (Entry);
> +    OrderedCollectionDelete (RestrictedAllocations, Entry, NULL);
> +  }
> +  OrderedCollectionUninit (RestrictedAllocations);
> +}
> +
> +
> +/**
> +  Iterate over the linker/loader script, and collect the names of the fw_cfg
> +  blobs that are referenced by QEMU_LOADER_ADD_POINTER.PointeeFile fields, such
> +  that QEMU_LOADER_ADD_POINTER.PointerSize is less than 8. This means that the
> +  pointee blob's address will have to be patched into a narrower-than-8 byte
> +  pointer field, hence the pointee blob must not be allocated from 64-bit
> +  address space.
> +
> +  @param[out] RestrictedAllocations  The ORDERED_COLLECTION structure linking
> +                                     (not copying / owning) such
> +                                     QEMU_LOADER_ADD_POINTER.PointeeFile fields
> +                                     that name the blobs restricted from 64-bit
> +                                     allocation.
> +
> +  @param[in] LoaderStart             Points to the first entry in the
> +                                     linker/loader script.
> +
> +  @param[in] LoaderEnd               Points one past the last entry in the
> +                                     linker/loader script.
> +
> +  @retval EFI_SUCCESS           RestrictedAllocations has been populated.
> +
> +  @retval EFI_OUT_OF_RESOURCES  Memory allocation failed.
> +
> +  @retval EFI_PROTOCOL_ERROR    Invalid linker/loader script contents.
> +**/
> +STATIC
> +EFI_STATUS
> +CollectRestrictedAllocations (
> +  OUT ORDERED_COLLECTION     **RestrictedAllocations,
> +  IN CONST QEMU_LOADER_ENTRY *LoaderStart,
> +  IN CONST QEMU_LOADER_ENTRY *LoaderEnd
> +)
> +{
> +  ORDERED_COLLECTION      *Collection;
> +  CONST QEMU_LOADER_ENTRY *LoaderEntry;
> +  EFI_STATUS              Status;
> +
> +  Collection = OrderedCollectionInit (AsciiStringCompare, AsciiStringCompare);
> +  if (Collection == NULL) {
> +    return EFI_OUT_OF_RESOURCES;
> +  }
> +
> +  for (LoaderEntry = LoaderStart; LoaderEntry < LoaderEnd; ++LoaderEntry) {
> +    CONST QEMU_LOADER_ADD_POINTER *AddPointer;
> +
> +    if (LoaderEntry->Type != QemuLoaderCmdAddPointer) {
> +      continue;
> +    }
> +    AddPointer = &LoaderEntry->Command.AddPointer;
> +
> +    if (AddPointer->PointerSize >= 8) {
> +      continue;
> +    }
> +
> +    if (AddPointer->PointeeFile[QEMU_LOADER_FNAME_SIZE - 1] != '\0') {
> +      DEBUG ((DEBUG_ERROR, "%a: malformed file name\n", __FUNCTION__));
> +      Status = EFI_PROTOCOL_ERROR;
> +      goto RollBack;
> +    }
> +
> +    Status = OrderedCollectionInsert (
> +               Collection,
> +               NULL,                           // Entry
> +               (VOID *)AddPointer->PointeeFile
> +               );
> +    switch (Status) {
> +    case EFI_SUCCESS:
> +      DEBUG ((
> +        DEBUG_VERBOSE,
> +        "%a: restricting blob \"%a\" from 64-bit allocation\n",
> +        __FUNCTION__,
> +        AddPointer->PointeeFile
> +        ));
> +      break;
> +    case EFI_ALREADY_STARTED:
> +      //
> +      // The restriction has been recorded already.
> +      //
> +      break;
> +    case EFI_OUT_OF_RESOURCES:
> +      goto RollBack;
> +    default:
> +      ASSERT (FALSE);
> +    }
> +  }
> +
> +  *RestrictedAllocations = Collection;
> +  return EFI_SUCCESS;
> +
> +RollBack:
> +  ReleaseRestrictedAllocations (Collection);
> +  return Status;
> +}
> +
> +
> +/**
>    Process a QEMU_LOADER_ALLOCATE command.
>
> -  @param[in] Allocate     The QEMU_LOADER_ALLOCATE command to process.
> +  @param[in] Allocate               The QEMU_LOADER_ALLOCATE command to
> +                                    process.
>
> -  @param[in,out] Tracker  The ORDERED_COLLECTION tracking the BLOB user
> -                          structures created thus far.
> +  @param[in,out] Tracker            The ORDERED_COLLECTION tracking the BLOB
> +                                    user structures created thus far.
> +
> +  @param[in] RestrictedAllocations  The ORDERED_COLLECTION populated by
> +                                    CollectRestrictedAllocations(), naming the
> +                                    fw_cfg blobs that must not be allocated
> +                                    from 64-bit address space.
>
>    @retval EFI_SUCCESS           An area of whole AcpiNVS pages has been
>                                  allocated for the blob contents, and the
>                                  contents have been saved. A BLOB object (user
> @@ -163,9 +318,10 @@ STATIC
>  EFI_STATUS
>  EFIAPI
>  ProcessCmdAllocate (
>    IN CONST QEMU_LOADER_ALLOCATE *Allocate,
> -  IN OUT ORDERED_COLLECTION     *Tracker
> +  IN OUT ORDERED_COLLECTION     *Tracker,
> +  IN ORDERED_COLLECTION         *RestrictedAllocations
>    )
>  {
>    FIRMWARE_CONFIG_ITEM FwCfgItem;
>    UINTN                FwCfgSize;
> @@ -192,9 +348,12 @@ ProcessCmdAllocate (
>      return Status;
>    }
>
>    NumPages = EFI_SIZE_TO_PAGES (FwCfgSize);
> -  Address = 0xFFFFFFFF;
> +  Address = MAX_UINT64;
> +  if (OrderedCollectionFind (RestrictedAllocations, Allocate->File) != NULL) {
> +    Address = MAX_UINT32;
> +  }
>    Status = gBS->AllocatePages (AllocateMaxAddress, EfiACPIMemoryNVS, NumPages,
>                    &Address);
>    if (EFI_ERROR (Status)) {
>      return Status;
> @@ -805,8 +964,9 @@ InstallQemuFwCfgTables (
>    CONST QEMU_LOADER_ENTRY  *LoaderEntry, *LoaderEnd;
>    CONST QEMU_LOADER_ENTRY  *WritePointerSubsetEnd;
>    ORIGINAL_ATTRIBUTES      *OriginalPciAttributes;
>    UINTN                    OriginalPciAttributesCount;
> +  ORDERED_COLLECTION       *RestrictedAllocations;
>    S3_CONTEXT               *S3Context;
>    ORDERED_COLLECTION       *Tracker;
>    UINTN                    *InstalledKey;
>    INT32                    Installed;
> @@ -833,17 +993,26 @@ InstallQemuFwCfgTables (
>    QemuFwCfgReadBytes (FwCfgSize, LoaderStart);
>    RestorePciDecoding (OriginalPciAttributes, OriginalPciAttributesCount);
>    LoaderEnd = LoaderStart + FwCfgSize / sizeof *LoaderEntry;
>
> +  Status = CollectRestrictedAllocations (
> +             &RestrictedAllocations,
> +             LoaderStart,
> +             LoaderEnd
> +             );
> +  if (EFI_ERROR (Status)) {
> +    goto FreeLoader;
> +  }
> +
>    S3Context = NULL;
>    if (QemuFwCfgS3Enabled ()) {
>      //
>      // Size the allocation pessimistically, assuming that all commands in the
>      // script are QEMU_LOADER_WRITE_POINTER commands.
>      //
>      Status = AllocateS3Context (&S3Context, LoaderEnd - LoaderStart);
>      if (EFI_ERROR (Status)) {
> -      goto FreeLoader;
> +      goto FreeRestrictedAllocations;
>      }
>    }
>
>    Tracker = OrderedCollectionInit (BlobCompare, BlobKeyCompare);
> @@ -862,9 +1031,13 @@ InstallQemuFwCfgTables (
>    WritePointerSubsetEnd = LoaderStart;
>    for (LoaderEntry = LoaderStart; LoaderEntry < LoaderEnd; ++LoaderEntry) {
>      switch (LoaderEntry->Type) {
>      case QemuLoaderCmdAllocate:
> -      Status = ProcessCmdAllocate (&LoaderEntry->Command.Allocate, Tracker);
> +      Status = ProcessCmdAllocate (
> +                 &LoaderEntry->Command.Allocate,
> +                 Tracker,
> +                 RestrictedAllocations
> +                 );
>        break;
>
>      case QemuLoaderCmdAddPointer:
>        Status = ProcessCmdAddPointer (&LoaderEntry->Command.AddPointer,
> @@ -1009,8 +1182,11 @@ FreeS3Context:
>    if (S3Context != NULL) {
>      ReleaseS3Context (S3Context);
>    }
>
> +FreeRestrictedAllocations:
> +  ReleaseRestrictedAllocations (RestrictedAllocations);
> +
>  FreeLoader:
>    FreePool (LoaderStart);
>
>    return Status;
> --
> 2.9.3
>


  reply	other threads:[~2017-06-03 18:32 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-06-03 15:42 [PATCH 0/1] OvmfPkg/AcpiPlatformDxe: alloc blobs from 64-bit space unless restricted Laszlo Ersek
2017-06-03 15:42 ` [PATCH 1/1] " Laszlo Ersek
2017-06-03 18:33   ` Ard Biesheuvel [this message]
2017-06-05 17:47   ` Jordan Justen
2017-06-06 18:16     ` Laszlo Ersek
2017-06-07 23:10       ` Laszlo Ersek
2017-06-08 10:11         ` Ard Biesheuvel
2017-06-08 18:40           ` Laszlo Ersek
2017-06-08 19:05             ` Ard Biesheuvel

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=CAKv+Gu_QaVGXq6EYkzT+C5cwLXQbMRQqYsQsO0yxxfZFwcbo4A@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