KVM ARM Archive on lore.kernel.org
 help / color / Atom feed
From: Alexandru Elisei <alexandru.elisei@arm.com>
To: Andrew Jones <drjones@redhat.com>
Cc: pbonzini@redhat.com, kvmarm@lists.cs.columbia.edu, kvm@vger.kernel.org
Subject: Re: [kvm-unit-tests PATCH] arm64: Compile with -mno-outline-atomics for GCC >= 10
Date: Sat, 18 Jul 2020 14:50:19 +0100
Message-ID: <202d475d-95df-2350-a8e9-9264144993ac@arm.com> (raw)
In-Reply-To: <20200718091145.zheu46pfjwsntr3a@kamzik.brq.redhat.com>

Hi,

On 7/18/20 10:11 AM, Andrew Jones wrote:
> On Fri, Jul 17, 2020 at 05:47:27PM +0100, Alexandru Elisei wrote:
>> GCC 10.1.0 introduced the -m{,no-}outline-atomics flags which, according to
>> man 1 gcc:
>>
>> "Enable or disable calls to out-of-line helpers to implement atomic
>> operations.  These helpers will, at runtime, determine if the LSE
>> instructions from ARMv8.1-A can be used; if not, they will use the
>> load/store-exclusive instructions that are present in the base ARMv8.0 ISA.
>> [..] This option is on by default."
>>
>> Unfortunately the option causes the following error at compile time:
>>
>> aarch64-linux-gnu-ld -nostdlib -pie -n -o arm/spinlock-test.elf -T /path/to/kvm-unit-tests/arm/flat.lds \
>> 	arm/spinlock-test.o arm/cstart64.o lib/libcflat.a lib/libfdt/libfdt.a /usr/lib/gcc/aarch64-linux-gnu/10.1.0/libgcc.a lib/arm/libeabi.a arm/spinlock-test.aux.o
>> aarch64-linux-gnu-ld: /usr/lib/gcc/aarch64-linux-gnu/10.1.0/libgcc.a(lse-init.o): in function `init_have_lse_atomics':
>> lse-init.c:(.text.startup+0xc): undefined reference to `__getauxval'
>>
>> This is happening because we are linking against our own libcflat which
>> doesn't implement the function __getauxval().
>>
>> Disable the use of the out-of-line functions by compiling with
>> -mno-outline-atomics if we detect a GCC version greater than 10.
>>
>> Signed-off-by: Alexandru Elisei <alexandru.elisei@arm.com>
>> ---
>>
>> Tested with gcc versions 10.1.0 and 5.4.0 (cross-compilation), 9.3.0
>> (native).
>>
>> I've been able to suss out the reason for the build failure from this
>> rejected gcc patch [1].
>>
>> [1] https://patches.openembedded.org/patch/172460/
>>
>>  arm/Makefile.arm64 | 6 ++++++
>>  1 file changed, 6 insertions(+)
>>
>> diff --git a/arm/Makefile.arm64 b/arm/Makefile.arm64
>> index dfd0c56fe8fb..3223cb966789 100644
>> --- a/arm/Makefile.arm64
>> +++ b/arm/Makefile.arm64
>> @@ -9,6 +9,12 @@ ldarch = elf64-littleaarch64
>>  arch_LDFLAGS = -pie -n
>>  CFLAGS += -mstrict-align
>>  
>> +# The -mno-outline-atomics flag is only valid for GCC versions 10 and greater.
>> +GCC_MAJOR_VERSION=$(shell $(CC) -dumpversion 2> /dev/null | cut -f1 -d.)
>> +ifeq ($(shell expr "$(GCC_MAJOR_VERSION)" ">=" "10"), 1)
>> +CFLAGS += -mno-outline-atomics
>> +endif
> How about this patch instead?
>
> diff --git a/Makefile b/Makefile
> index 3ff2f91600f6..0e21a49096ba 100644
> --- a/Makefile
> +++ b/Makefile
> @@ -17,6 +17,11 @@ DESTDIR := $(PREFIX)/share/kvm-unit-tests/
>  
>  .PHONY: arch_clean clean distclean cscope
>  
> +# cc-option
> +# Usage: OP_CFLAGS+=$(call cc-option, -falign-functions=0, -malign-functions=0)
> +cc-option = $(shell if $(CC) -Werror $(1) -S -o /dev/null -xc /dev/null \
> +              > /dev/null 2>&1; then echo "$(1)"; else echo "$(2)"; fi ;)
> +
>  #make sure env CFLAGS variable is not used
>  CFLAGS =
>  
> @@ -43,12 +48,6 @@ OBJDIRS += $(LIBFDT_objdir)
>  #include architecture specific make rules
>  include $(SRCDIR)/$(TEST_DIR)/Makefile
>  
> -# cc-option
> -# Usage: OP_CFLAGS+=$(call cc-option, -falign-functions=0, -malign-functions=0)
> -
> -cc-option = $(shell if $(CC) -Werror $(1) -S -o /dev/null -xc /dev/null \
> -              > /dev/null 2>&1; then echo "$(1)"; else echo "$(2)"; fi ;)
> -
>  COMMON_CFLAGS += -g $(autodepend-flags) -fno-strict-aliasing -fno-common
>  COMMON_CFLAGS += -Wall -Wwrite-strings -Wempty-body -Wuninitialized
>  COMMON_CFLAGS += -Wignored-qualifiers -Werror
> diff --git a/arm/Makefile.arm64 b/arm/Makefile.arm64
> index dfd0c56fe8fb..dbc7524d3070 100644
> --- a/arm/Makefile.arm64
> +++ b/arm/Makefile.arm64
> @@ -9,6 +9,9 @@ ldarch = elf64-littleaarch64
>  arch_LDFLAGS = -pie -n
>  CFLAGS += -mstrict-align
>  
> +mno_outline_atomics := $(call cc-option, -mno-outline-atomics, "")
> +CFLAGS += $(mno_outline_atomics)
> +
>  define arch_elf_check =
>  	$(if $(shell ! $(OBJDUMP) -R $(1) >&/dev/null && echo "nok"),
>  		$(error $(shell $(OBJDUMP) -R $(1) 2>&1)))
>
>
> Thanks,
> drew

Looks much better than my version. Do you want me to spin a v2 or do you want to
send it as a separate patch? If that's the case, I tested the same way I did my
patch (gcc 10.1.0 and 5.4.0 for cross-compiling, 9.3.0 native):

Tested-by: Alexandru Elisei <alexandru.elisei@arm.com>

Thanks,
Alex
_______________________________________________
kvmarm mailing list
kvmarm@lists.cs.columbia.edu
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm

  reply index

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-07-17 16:47 Alexandru Elisei
2020-07-18  9:11 ` Andrew Jones
2020-07-18 13:50   ` Alexandru Elisei [this message]
2020-07-27 12:21     ` Alexandru Elisei
2020-07-27 12:30       ` Andrew Jones
2020-07-27 12:39         ` Alexandru Elisei
2020-07-27 17:25         ` Paolo Bonzini

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-all 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=202d475d-95df-2350-a8e9-9264144993ac@arm.com \
    --to=alexandru.elisei@arm.com \
    --cc=drjones@redhat.com \
    --cc=kvm@vger.kernel.org \
    --cc=kvmarm@lists.cs.columbia.edu \
    --cc=pbonzini@redhat.com \
    /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

KVM ARM Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/kvmarm/0 kvmarm/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 kvmarm kvmarm/ https://lore.kernel.org/kvmarm \
		kvmarm@lists.cs.columbia.edu
	public-inbox-index kvmarm

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/edu.columbia.cs.lists.kvmarm


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git