| Message ID | 20161123142949.GD23528@citrix.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Hi Wei, On 23/11/16 14:29, Wei Liu wrote: > On Wed, Nov 23, 2016 at 04:12:28PM +0200, Andrii Anisov wrote: >>> But before you write any code, let me try to understand the real issue >>> first: you're trying to cross-build ARM tools on x86, but x86 >>> iasl doesn't support ARM ACPI definition(s), right? >> Well, as I stated here [1], I'm pretty far from ACPI and understanding >> of what's going wrong with the compilation. But I have a strong >> feeling that this option should be optional. >> >> [1] https://lists.xenproject.org/archives/html/xen-devel/2016-11/msg01903.html > > What is the exact rune you got when compiling mk_dsdt ? > > If everything it set up properly you should be using a cross-build gcc > which should have the correct architecture? > > Can you try this (untested) patch? > > > diff --git a/tools/libacpi/Makefile b/tools/libacpi/Makefile > index ccc32c9..2b9d5b8 100644 > --- a/tools/libacpi/Makefile > +++ b/tools/libacpi/Makefile > @@ -44,7 +44,7 @@ $(H_SRC): $(ACPI_BUILD_DIR)/%.h: %.asl iasl > rm -f $(addprefix $(ACPI_BUILD_DIR)/, $*.aml $*.hex) > > $(MK_DSDT): mk_dsdt.c > - $(HOSTCC) $(HOSTCFLAGS) $(CFLAGS_xeninclude) -D__XEN_TOOLS__ -o $@ mk_dsdt.c > + $(CC) $(CFLAGS) $(CFLAGS_xeninclude) -D__XEN_TOOLS__ -o $@ mk_dsdt.c That will not work, CC will build an ARM binary. However mk_dsdt is used to generate the static ACPI table integrated in the libxl, so mk_dsdt needs to be built with HOSTCC. I think the code should use CONFIG_ARM_64/CONFIG_ARM_32 rather than __*__ one. Regards,
On Wed, Nov 23, 2016 at 02:34:04PM +0000, Julien Grall wrote: > Hi Wei, > > On 23/11/16 14:29, Wei Liu wrote: > >On Wed, Nov 23, 2016 at 04:12:28PM +0200, Andrii Anisov wrote: > >>>But before you write any code, let me try to understand the real issue > >>>first: you're trying to cross-build ARM tools on x86, but x86 > >>>iasl doesn't support ARM ACPI definition(s), right? > >>Well, as I stated here [1], I'm pretty far from ACPI and understanding > >>of what's going wrong with the compilation. But I have a strong > >>feeling that this option should be optional. > >> > >>[1] https://lists.xenproject.org/archives/html/xen-devel/2016-11/msg01903.html > > > >What is the exact rune you got when compiling mk_dsdt ? > > > >If everything it set up properly you should be using a cross-build gcc > >which should have the correct architecture? > > > >Can you try this (untested) patch? > > > > > >diff --git a/tools/libacpi/Makefile b/tools/libacpi/Makefile > >index ccc32c9..2b9d5b8 100644 > >--- a/tools/libacpi/Makefile > >+++ b/tools/libacpi/Makefile > >@@ -44,7 +44,7 @@ $(H_SRC): $(ACPI_BUILD_DIR)/%.h: %.asl iasl > > rm -f $(addprefix $(ACPI_BUILD_DIR)/, $*.aml $*.hex) > > > > $(MK_DSDT): mk_dsdt.c > >- $(HOSTCC) $(HOSTCFLAGS) $(CFLAGS_xeninclude) -D__XEN_TOOLS__ -o $@ mk_dsdt.c > >+ $(CC) $(CFLAGS) $(CFLAGS_xeninclude) -D__XEN_TOOLS__ -o $@ mk_dsdt.c > > That will not work, CC will build an ARM binary. However mk_dsdt is used to > generate the static ACPI table integrated in the libxl, so mk_dsdt needs to > be built with HOSTCC. Yes, you're right. > > I think the code should use CONFIG_ARM_64/CONFIG_ARM_32 rather than __*__ > one. > That might work -- provided they are defined even in a cross-build. > Regards, > > -- > Julien Grall
diff --git a/tools/libacpi/Makefile b/tools/libacpi/Makefile index ccc32c9..2b9d5b8 100644 --- a/tools/libacpi/Makefile +++ b/tools/libacpi/Makefile @@ -44,7 +44,7 @@ $(H_SRC): $(ACPI_BUILD_DIR)/%.h: %.asl iasl rm -f $(addprefix $(ACPI_BUILD_DIR)/, $*.aml $*.hex) $(MK_DSDT): mk_dsdt.c - $(HOSTCC) $(HOSTCFLAGS) $(CFLAGS_xeninclude) -D__XEN_TOOLS__ -o $@ mk_dsdt.c + $(CC) $(CFLAGS) $(CFLAGS_xeninclude) -D__XEN_TOOLS__ -o $@ mk_dsdt.c $(ACPI_BUILD_DIR)/dsdt_anycpu_qemu_xen.asl: dsdt.asl dsdt_acpi_info.asl $(MK_DSDT) # Remove last bracket