On Tue, Feb 27, 2018 at 05:19:59PM +0000, Ard Biesheuvel wrote:
> On 27 February 2018 at 13:47, Ard Biesheuvel <[email protected]> 
> wrote:
> > The Designware PCIe IP in the SynQuacer SoC needs a little help to
> > appear sane to the OS. Not only does it lack a true root port, and
> > therefore does not perform any filtering whatsoever of type 0 config
> > TLPs that are not intended for the link peer, it also has trouble
> > issuing 64-bit wide MMIO accesses, which are often used on MMIO BARs
> > with memory semantics (e.g., frame buffers).
> >
> > So let's create a stage 2 mapping covering the entire physical address
> > space, and remap some ECAM regions and demote write combine attributes
> > to device/strongly ordered. This is not a water tight fix, but it does
> > work around the issues in the majority of cases.
> >
> > (Note that the ECAM remapping can also be addressed in the SMMU mapping
> > of the PCIe IP exposed to the CPU, but this is currently under
> > development, and it does not hurt to have it in two places)
> >
> > Contributed-under: TianoCore Contribution Agreement 1.1
> > Signed-off-by: Ard Biesheuvel <[email protected]>
> > ---
> > This is a followup to/replacement for '[RFC PATCH edk2-non-osi]
> > Platform/DeveloperBox: add prebuilt binary containing stage 2 page tables'
> >
> >  Platform/Socionext/DeveloperBox/DeveloperBox.dsc          |  1 +
> >  Platform/Socionext/DeveloperBox/DeveloperBox.fdf          |  5 +-
> >  Silicon/Socionext/SynQuacer/Stage2Tables/GNUmakefile      | 23 +++++
> >  Silicon/Socionext/SynQuacer/Stage2Tables/Stage2Tables.S   | 88 
> > ++++++++++++++++++++
> >  Silicon/Socionext/SynQuacer/Stage2Tables/Stage2Tables.inf | 32 +++++++
> >  5 files changed, 148 insertions(+), 1 deletion(-)
> >
> ...
> > diff --git a/Silicon/Socionext/SynQuacer/Stage2Tables/Stage2Tables.inf 
> > b/Silicon/Socionext/SynQuacer/Stage2Tables/Stage2Tables.inf
> > new file mode 100644
> > index 000000000000..9bec659af444
> > --- /dev/null
> > +++ b/Silicon/Socionext/SynQuacer/Stage2Tables/Stage2Tables.inf
> > @@ -0,0 +1,32 @@
> > +## @file
> > +#
> > +#  Copyright (c) 2018, Linaro, Ltd. All rights reserved.<BR>
> > +#
> > +#  This program and the accompanying materials
> > +#  are licensed and made available under the terms and conditions of the 
> > BSD License
> > +#  which accompanies this distribution.  The full text of the license may 
> > be found at
> > +#  http://opensource.org/licenses/bsd-license.php
> > +#
> > +#  THE PROGRAM IS DISTRIBUTED UNDER THE BSD LICENSE ON AN "AS IS" BASIS,
> > +#  WITHOUT WARRANTIES OR REPRESENTATIONS OF ANY KIND, EITHER EXPRESS OR 
> > IMPLIED.
> > +#
> > +##
> > +
> > +[Defines]
> > +  INF_VERSION                    = 0x0001001A
> > +  BASE_NAME                      = Stage2Tables
> > +  FILE_GUID                      = e11cbca5-8f82-41a7-8441-02b48acc13a7
> > +  MODULE_TYPE                    = USER_DEFINED
> > +  VERSION_STRING                 = 1.0
> > +  CUSTOM_MAKEFILE                = GCC|GNUmakefile
> > +
> > +[Sources]
> > +  Stage2Tables.S
> > +
> > +[Packages]
> > +  MdePkg/MdePkg.dec
> > +  Silicon/Socionext/SynQuacer/SynQuacer.dec
> > +
> > +[BuildOptions]
> > +  *_*_*_OBJCOPY_PATH  == objcopy
> > +  *_*_*_OBJCOPY_FLAGS == -I elf64-little -O binary -j .rodata
> 
> I can improve this and fix Clang at the same time by adding
> 
>   *_*_*_ASM_FLAGS == -nostdlib
> -Wl,-e,0x81f8000,--section-start=.rodata=0x81f8000
>   *_CLANG35_*_ASM_FLAGS = -no-integrated-as
>   *_CLANG38_*_ASM_FLAGS = -no-integrated-as
> 
> (and remove the linker arguments from the Makefile)

I can confirm this resolves the CLANG issue.

Could we do this with a CLANG_ALL_ASM_FLAGS, rather than listing each
new toolchain profile as they get added?

/
    Leif
_______________________________________________
edk2-devel mailing list
[email protected]
https://lists.01.org/mailman/listinfo/edk2-devel

Reply via email to