From: Leif Lindholm <leif.lindholm@linaro.org>
To: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: "edk2-devel@lists.01.org" <edk2-devel@lists.01.org>,
Laszlo Ersek <lersek@redhat.com>
Subject: Re: [PATCH v2 3/4] ArmPkg/CpuDxe ARM: honour RO/XP attributes in SetMemoryAttributes()
Date: Mon, 6 Mar 2017 16:40:47 +0000 [thread overview]
Message-ID: <20170306164047.GC16034@bivouac.eciton.net> (raw)
In-Reply-To: <CAKv+Gu8pVyitCPbOYY2s6eXKvziVz-gsFH7Up8dNT8hb6B4gKg@mail.gmail.com>
On Mon, Mar 06, 2017 at 04:11:50PM +0100, Ard Biesheuvel wrote:
> On 6 March 2017 at 15:48, Leif Lindholm <leif.lindholm@linaro.org> wrote:
> > On Thu, Mar 02, 2017 at 10:36:15AM +0000, Ard Biesheuvel wrote:
> >> Enable the use of strict memory permissions on ARM by processing the
> >> EFI_MEMORY_RO and EFI_MEMORY_XP rather than ignoring them. As before,
> >> calls to CpuArchProtocol::SetMemoryAttributes that only set RO/XP
> >> bits will preserve the cacheability attributes. Permissions attributes
> >> are not preserved when setting the memory type only: the way the memory
> >> permission attributes are defined does not allows for that, and so this
> >> situation does not deviate from other architectures.
> >>
> >> Contributed-under: TianoCore Contribution Agreement 1.0
> >> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
> >> ---
> >> ArmPkg/Drivers/CpuDxe/Arm/Mmu.c | 151 ++++++++------------
> >> 1 file changed, 62 insertions(+), 89 deletions(-)
> >>
> >> diff --git a/ArmPkg/Drivers/CpuDxe/Arm/Mmu.c b/ArmPkg/Drivers/CpuDxe/Arm/Mmu.c
> >> index 26b637e7658f..6dd749dadf8b 100644
> >> --- a/ArmPkg/Drivers/CpuDxe/Arm/Mmu.c
> >> +++ b/ArmPkg/Drivers/CpuDxe/Arm/Mmu.c
> >> @@ -374,50 +374,41 @@ UpdatePageEntries (
> >>
> >> // EntryMask: bitmask of values to change (1 = change this value, 0 = leave alone)
> >> // EntryValue: values at bit positions specified by EntryMask
> >> - EntryMask = TT_DESCRIPTOR_PAGE_TYPE_MASK;
> >> - EntryValue = TT_DESCRIPTOR_PAGE_TYPE_PAGE;
> >> + EntryMask = TT_DESCRIPTOR_PAGE_TYPE_MASK | TT_DESCRIPTOR_PAGE_AP_MASK;
> >> + if ((Attributes & EFI_MEMORY_XP) != 0) {
> >> + EntryValue = TT_DESCRIPTOR_PAGE_TYPE_PAGE_XN;
> >> + } else {
> >> + EntryValue = TT_DESCRIPTOR_PAGE_TYPE_PAGE;
> >> + }
> >> +
> >> // Although the PI spec is unclear on this the GCD guarantees that only
> >> // one Attribute bit is set at a time, so we can safely use a switch statement
> >
> > But the switch statement is going away.
> >
> > The change effectively introduces a guaranteed priority of
> > interpretation if the spec is violated. Say something about this order
> > being arbitrarily decided due to PI spce guarantee instead?
> >
>
> Indeed.
OK, I'll hold back to see what you come up with :)
> >> - switch (Attributes) {
> >> - case EFI_MEMORY_UC:
> >> - // modify cacheability attributes
> >> - EntryMask |= TT_DESCRIPTOR_PAGE_CACHE_POLICY_MASK;
> >> - // map to strongly ordered
> >> - EntryValue |= TT_DESCRIPTOR_PAGE_CACHE_POLICY_STRONGLY_ORDERED; // TEX[2:0] = 0, C=0, B=0
> >> - break;
> >> -
> >> - case EFI_MEMORY_WC:
> >> - // modify cacheability attributes
> >> - EntryMask |= TT_DESCRIPTOR_PAGE_CACHE_POLICY_MASK;
> >> - // map to normal non-cachable
> >> - EntryValue |= TT_DESCRIPTOR_PAGE_CACHE_POLICY_NON_CACHEABLE; // TEX [2:0]= 001 = 0x2, B=0, C=0
> >> - break;
> >> -
> >> - case EFI_MEMORY_WT:
> >> - // modify cacheability attributes
> >> - EntryMask |= TT_DESCRIPTOR_PAGE_CACHE_POLICY_MASK;
> >> - // write through with no-allocate
> >> - EntryValue |= TT_DESCRIPTOR_PAGE_CACHE_POLICY_WRITE_THROUGH_NO_ALLOC; // TEX [2:0] = 0, C=1, B=0
> >> - break;
> >> -
> >> - case EFI_MEMORY_WB:
> >> - // modify cacheability attributes
> >> - EntryMask |= TT_DESCRIPTOR_PAGE_CACHE_POLICY_MASK;
> >> - // write back (with allocate)
> >> - EntryValue |= TT_DESCRIPTOR_PAGE_CACHE_POLICY_WRITE_BACK_ALLOC; // TEX [2:0] = 001, C=1, B=1
> >> - break;
> >> -
> >> - case EFI_MEMORY_WP:
> >> - case EFI_MEMORY_XP:
> >> - case EFI_MEMORY_UCE:
> >> - // cannot be implemented UEFI definition unclear for ARM
> >> - // Cause a page fault if these ranges are accessed.
> >> - EntryValue = TT_DESCRIPTOR_PAGE_TYPE_FAULT;
> >> - DEBUG ((EFI_D_PAGE, "SetMemoryAttributes(): setting page %lx with unsupported attribute %x will page fault on access\n", BaseAddress, Attributes));
> >> - break;
> >> + if ((Attributes & EFI_MEMORY_UC) != 0) {
> >
> > Or should these be "Attributes & Mask" == "ATTRIBUTE"?
> >
>
> I know this is more idiomatic for EDK2, but the mask *is* the
> attribute in this case, and so the former implies the latter. If the
> mask were
OK ... in that case, could you just drop the != 0?
> (EFI_MEMORY_WB | EFI_MEMORY_WT | etc
>
> it would make more sense do to the latter I think
>
> >> + // modify cacheability attributes
> >> + EntryMask |= TT_DESCRIPTOR_PAGE_CACHE_POLICY_MASK;
> >> + // map to strongly ordered
> >> + EntryValue |= TT_DESCRIPTOR_PAGE_CACHE_POLICY_STRONGLY_ORDERED; // TEX[2:0] = 0, C=0, B=0
> >> + } else if ((Attributes & EFI_MEMORY_WC) != 0) {
> >> + // modify cacheability attributes
> >> + EntryMask |= TT_DESCRIPTOR_PAGE_CACHE_POLICY_MASK;
> >> + // map to normal non-cachable
> >> + EntryValue |= TT_DESCRIPTOR_PAGE_CACHE_POLICY_NON_CACHEABLE; // TEX [2:0]= 001 = 0x2, B=0, C=0
> >> + } else if ((Attributes & EFI_MEMORY_WT) != 0) {
> >> + // modify cacheability attributes
> >> + EntryMask |= TT_DESCRIPTOR_PAGE_CACHE_POLICY_MASK;
> >> + // write through with no-allocate
> >> + EntryValue |= TT_DESCRIPTOR_PAGE_CACHE_POLICY_WRITE_THROUGH_NO_ALLOC; // TEX [2:0] = 0, C=1, B=0
> >> + } else if ((Attributes & EFI_MEMORY_WB) != 0) {
> >> + // modify cacheability attributes
> >> + EntryMask |= TT_DESCRIPTOR_PAGE_CACHE_POLICY_MASK;
> >> + // write back (with allocate)
> >> + EntryValue |= TT_DESCRIPTOR_PAGE_CACHE_POLICY_WRITE_BACK_ALLOC; // TEX [2:0] = 001, C=1, B=1
> >> + }
> >>
> >> - default:
> >> - return EFI_UNSUPPORTED;
> >
> > Do we not want a fallback handling for the if-form as well?
> >
>
> No, and that is actually the point of this patch. SetMemoryAttributes
> may be called without a memory type argument, in which case it only
> modifies permission attributes, and leaves the memory type attributes
> only.
But should there not be a final "else if (Attributes &
CACHE_ATTRIBUTE_MASK)" error path then?
/
Leif
> >> + if ((Attributes & EFI_MEMORY_RO) != 0) {
> >> + EntryValue |= TT_DESCRIPTOR_PAGE_AP_RO_RO;
> >> + } else {
> >> + EntryValue |= TT_DESCRIPTOR_PAGE_AP_RW_RW;
> >> }
> >>
> >> // Obtain page table base
> >> @@ -520,53 +511,42 @@ UpdateSectionEntries (
> >> // EntryValue: values at bit positions specified by EntryMask
> >>
> >> // Make sure we handle a section range that is unmapped
> >> - EntryMask = TT_DESCRIPTOR_SECTION_TYPE_MASK;
> >> + EntryMask = TT_DESCRIPTOR_SECTION_TYPE_MASK | TT_DESCRIPTOR_SECTION_XN_MASK |
> >> + TT_DESCRIPTOR_SECTION_AP_MASK;
> >> EntryValue = TT_DESCRIPTOR_SECTION_TYPE_SECTION;
> >>
> >> // Although the PI spec is unclear on this the GCD guarantees that only
> >> // one Attribute bit is set at a time, so we can safely use a switch statement
> >
> > Repeat of above.
> >
> >> - switch(Attributes) {
> >> - case EFI_MEMORY_UC:
> >> - // modify cacheability attributes
> >> - EntryMask |= TT_DESCRIPTOR_SECTION_CACHE_POLICY_MASK;
> >> - // map to strongly ordered
> >> - EntryValue |= TT_DESCRIPTOR_SECTION_CACHE_POLICY_STRONGLY_ORDERED; // TEX[2:0] = 0, C=0, B=0
> >> - break;
> >> -
> >> - case EFI_MEMORY_WC:
> >> - // modify cacheability attributes
> >> - EntryMask |= TT_DESCRIPTOR_SECTION_CACHE_POLICY_MASK;
> >> - // map to normal non-cachable
> >> - EntryValue |= TT_DESCRIPTOR_SECTION_CACHE_POLICY_NON_CACHEABLE; // TEX [2:0]= 001 = 0x2, B=0, C=0
> >> - break;
> >> -
> >> - case EFI_MEMORY_WT:
> >> - // modify cacheability attributes
> >> - EntryMask |= TT_DESCRIPTOR_SECTION_CACHE_POLICY_MASK;
> >> - // write through with no-allocate
> >> - EntryValue |= TT_DESCRIPTOR_SECTION_CACHE_POLICY_WRITE_THROUGH_NO_ALLOC; // TEX [2:0] = 0, C=1, B=0
> >> - break;
> >> -
> >> - case EFI_MEMORY_WB:
> >> - // modify cacheability attributes
> >> - EntryMask |= TT_DESCRIPTOR_SECTION_CACHE_POLICY_MASK;
> >> - // write back (with allocate)
> >> - EntryValue |= TT_DESCRIPTOR_SECTION_CACHE_POLICY_WRITE_BACK_ALLOC; // TEX [2:0] = 001, C=1, B=1
> >> - break;
> >> -
> >> - case EFI_MEMORY_WP:
> >> - case EFI_MEMORY_XP:
> >> - case EFI_MEMORY_RP:
> >> - case EFI_MEMORY_UCE:
> >> - // cannot be implemented UEFI definition unclear for ARM
> >> - // Cause a page fault if these ranges are accessed.
> >> - EntryValue = TT_DESCRIPTOR_SECTION_TYPE_FAULT;
> >> - DEBUG ((EFI_D_PAGE, "SetMemoryAttributes(): setting section %lx with unsupported attribute %x will page fault on access\n", BaseAddress, Attributes));
> >> - break;
> >> + if ((Attributes & EFI_MEMORY_UC) != 0) {
> >> + // modify cacheability attributes
> >> + EntryMask |= TT_DESCRIPTOR_SECTION_CACHE_POLICY_MASK;
> >> + // map to strongly ordered
> >> + EntryValue |= TT_DESCRIPTOR_SECTION_CACHE_POLICY_STRONGLY_ORDERED; // TEX[2:0] = 0, C=0, B=0
> >> + } else if ((Attributes & EFI_MEMORY_WC) != 0) {
> >> + // modify cacheability attributes
> >> + EntryMask |= TT_DESCRIPTOR_SECTION_CACHE_POLICY_MASK;
> >> + // map to normal non-cachable
> >> + EntryValue |= TT_DESCRIPTOR_SECTION_CACHE_POLICY_NON_CACHEABLE; // TEX [2:0]= 001 = 0x2, B=0, C=0
> >> + } else if ((Attributes & EFI_MEMORY_WT) != 0) {
> >> + // modify cacheability attributes
> >> + EntryMask |= TT_DESCRIPTOR_SECTION_CACHE_POLICY_MASK;
> >> + // write through with no-allocate
> >> + EntryValue |= TT_DESCRIPTOR_SECTION_CACHE_POLICY_WRITE_THROUGH_NO_ALLOC; // TEX [2:0] = 0, C=1, B=0
> >> + } else if ((Attributes & EFI_MEMORY_WB) != 0) {
> >> + // modify cacheability attributes
> >> + EntryMask |= TT_DESCRIPTOR_SECTION_CACHE_POLICY_MASK;
> >> + // write back (with allocate)
> >> + EntryValue |= TT_DESCRIPTOR_SECTION_CACHE_POLICY_WRITE_BACK_ALLOC; // TEX [2:0] = 001, C=1, B=1
> >> + }
> >
> > (Again, same questions as above.)
> >
> >>
> >> + if ((Attributes & EFI_MEMORY_RO) != 0) {
> >> + EntryValue |= TT_DESCRIPTOR_SECTION_AP_RO_RO;
> >> + } else {
> >> + EntryValue |= TT_DESCRIPTOR_SECTION_AP_RW_RW;
> >> + }
> >>
> >> - default:
> >> - return EFI_UNSUPPORTED;
> >> + if ((Attributes & EFI_MEMORY_XP) != 0) {
> >> + EntryValue |= TT_DESCRIPTOR_SECTION_XN_MASK;
> >> }
> >>
> >> // obtain page table base
> >> @@ -693,13 +673,6 @@ SetMemoryAttributes (
> >> UINT64 ChunkLength;
> >> BOOLEAN FlushTlbs;
> >>
> >> - //
> >> - // Ignore invocations that only modify permission bits
> >> - //
> >> - if ((Attributes & EFI_MEMORY_CACHETYPE_MASK) == 0) {
> >> - return EFI_SUCCESS;
> >> - }
> >> -
> >> FlushTlbs = FALSE;
> >> while (Length > 0) {
> >> if ((BaseAddress % TT_DESCRIPTOR_SECTION_SIZE == 0) &&
> >> --
> >> 2.7.4
> >>
next prev parent reply other threads:[~2017-03-06 16:40 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-03-02 10:36 [PATCH v2 0/4] ArmPkg, ArmVirtpkg ARM: enable strict memory protection Ard Biesheuvel
2017-03-02 10:36 ` [PATCH v2 1/4] ArmPkg/CpuDxe ARM: avoid splitting page table sections unnecessarily Ard Biesheuvel
2017-03-06 14:12 ` Leif Lindholm
2017-03-06 14:55 ` Ard Biesheuvel
2017-03-06 16:24 ` Leif Lindholm
2017-03-02 10:36 ` [PATCH v2 2/4] ArmPkg/CpuDxe ARM: avoid unnecessary cache/TLB maintenance Ard Biesheuvel
2017-03-06 14:27 ` Leif Lindholm
2017-03-06 15:06 ` Ard Biesheuvel
2017-03-02 10:36 ` [PATCH v2 3/4] ArmPkg/CpuDxe ARM: honour RO/XP attributes in SetMemoryAttributes() Ard Biesheuvel
2017-03-06 14:48 ` Leif Lindholm
2017-03-06 15:11 ` Ard Biesheuvel
2017-03-06 16:40 ` Leif Lindholm [this message]
2017-03-02 10:36 ` [PATCH v2 4/4] ArmVirtPkg: enable PE/COFF image and memory protection for ARM platforms 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=20170306164047.GC16034@bivouac.eciton.net \
--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