[klibc] [kvm-unit-tests PATCH v3 0/4] Update patch set

Bill Wendling morbo at google.com
Sat Jul 25 23:13:11 PDT 2020
Previous message (by thread): [klibc] [kvm-unit-tests PATCH v3 0/4] Update patch set
Next message (by thread): [klibc] [kvm-unit-tests PATCH v3 0/4] Update patch set
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
On Sat, Jul 25, 2020 at 4:10 PM Ben Hutchings <ben at decadent.org.uk> wrote:
> On Sat, 2020-07-25 at 15:16 -0700, Bill Wendling wrote:
> > On Sat, Jul 25, 2020 at 2:56 PM Ben Hutchings <ben at decadent.org.uk> wrote:
> > > On Sat, 2020-07-25 at 14:36 -0700, Bill Wendling wrote:
> > > [...]
> > > > I upstreamed a few patches to klibc to fix a few linker errors. Some
> > > > of these might help:
> > >
> > > I applied v3 of your patch series and never saw this later version.
> > >
> > > >   - cfc4e4c9591c73edc786aebbe83b371fec36fec8 - Conditionalized
> > > > non-supported flags.
> > > >   - 1b014f8401fc5b3dc803e2cae06ac5c803dc89cb -  Add "-fno-common"
> > > >   - 1e876662002eac9662357c7feaee12be04107d32 - Use proper libc
> > > >   - f07a8f75b8ebfea71254cd8829c0476b6b6683e1 - Adjust some flags for ld.lld
> > > >   - 7f0aa03493dc9ab8c1d2798f6586bfbc5f1de702 - Use "-ffreestanding"
> > > > (fixes using unlocked versions of calls)
> > > >
> > > > There are also a bunch of clang warnings Michael Davidson fixed in the
> > > > klibc tree.
> > >
> > > *Which* klibc tree?
> > >
> > > Ben.
> > >
> > The one here I think: https://git.kernel.org/pub/scm/libs/klibc/klibc.git
>
> That's where I'm pushing.  It has v3 of your patch series, and no other
> recent changes related to Clang or lld.
>
> So please send over any other fixes you have.
>
Hi Ben,

I'm sorry. I don't know what you want. The patches I see in
top-of-tree klibc look to be the ones I listed above. Though I don't
see the -ffreestanding and the -Ttext-segment changes

commit 3dbbaade8f836f18e1f739a61cc359d799d1127f
Author: Bill Wendling <morbo at google.com>
Date:   Mon Jun 4 16:55:01 2018 -0700

    klibc: add -ffreestanding to required flags

    Newer clang versions need the '-ffreestanding' flag because new
changes convert
    system calls to their unlocked versions, which don't exist in the kernel
    library.

    Signed-off-by: Bill Wendling <morbo at google.com>

diff --git a/gbuild/klibc/scripts/Kbuild.klibc
b/gbuild/klibc/scripts/Kbuild.klibc
index 4e9f5eb4846e..18f7291a1f61 100644
--- a/scripts/Kbuild.klibc
+++ b/scripts/Kbuild.klibc
@@ -67,7 +67,8 @@ include $(srctree)/scripts/Kbuild.include
 # ---------------------------------------------------------------------------

 KLIBCREQFLAGS     := $(call cc-option, -fno-stack-protector, ) \
-                     $(call cc-option, -fwrapv, )
+                     $(call cc-option, -fwrapv, ) \
+                     $(call cc-option, -ffreestanding, )
 KLIBCARCHREQFLAGS :=
 KLIBCOPTFLAGS     :=
 KLIBCWARNFLAGS    := -W -Wall -Wno-sign-compare -Wno-unused-parameter \


commit c3961aac9362f362c5aae8ea50d67c0a9766c1c7
Author: Stanislav Fomichev <sdf at google.com>
Date:   Thu Feb 27 09:51:29 2020 -0800

    klibc: support llvm's lld

    Clang that we have in google3 started emitting relocations that
    our outdated binutils ld doesn't support. In an effort to switch
    userspace to use llvm's ld, let's teach klibc some ld.lld quirks.

    Signed-off-by: Stanislav Fomichev <sdf at google.com>
    Signed-off-by: Bill Wendling <morbo at google.com>

diff --git a/Makefile b/Makefile
index a60161ccca5b..46d35fe77c3b 100644
--- a/Makefile
+++ b/Makefile
@@ -27,6 +27,9 @@ export NM := $(KLIBCROSS)nm
 export OBJCOPY  := $(KLIBCROSS)objcopy
 export OBJDUMP  := $(KLIBCROSS)objdump

+LLD := $(shell $(LD) --version 2>&1 | grep LLD)
+export IMAGE_BASE=$(if $(LLD),--image-base,-Ttext-segment)
+
 NOSTDINC_FLAGS := -nostdlib -nostdinc -isystem $(shell $(CC)
-print-file-name=include)

 ARCH           := $(shell uname -m | sed -e s/i.86/i386/ \
diff --git a/usr/klibc/arch/i386/MCONFIG b/usr/klibc/arch/i386/MCONFIG
index 0704669500f0..baa7e59854f8 100644
--- a/usr/klibc/arch/i386/MCONFIG
+++ b/usr/klibc/arch/i386/MCONFIG
@@ -30,4 +30,4 @@ KLIBCBITSIZE  = 32
 # This address needs to be reachable using normal inter-module
 # calls, and work on the memory models for this architecture
 # 96 MB - normal binaries start at 128 MB
-KLIBCSHAREDFLAGS = -Ttext-segment 0x06000000
+KLIBCSHAREDFLAGS = $(IMAGE_BASE) 0x06000000
diff --git a/usr/klibc/arch/mips64/MCONFIG b/usr/klibc/arch/mips64/MCONFIG
index 6a4b41b2f85c..3d7f20da08be 100644
--- a/usr/klibc/arch/mips64/MCONFIG
+++ b/usr/klibc/arch/mips64/MCONFIG
@@ -20,4 +20,4 @@ KLIBCBITSIZE  = 64
 # load address, use the very top of the 256M region (minus 2MB)
 #
 # Use -Ttext-segment so that the special .MIPS* sections are moved as well.
-KLIBCSHAREDFLAGS     = -Ttext-segment 0x12FE00000
+KLIBCSHAREDFLAGS     = $(IMAGE_BASE) 0x12FE00000
diff --git a/usr/klibc/arch/ppc/MCONFIG b/usr/klibc/arch/ppc/MCONFIG
index 8220f6b93fa4..ec859e8941d9 100644
--- a/usr/klibc/arch/ppc/MCONFIG
+++ b/usr/klibc/arch/ppc/MCONFIG
@@ -20,7 +20,7 @@ KLIBCBITSIZE       = 32
 # calls, and work on the memory models for this architecture
 # 256-16 MB - normal binaries start at 256 MB, and jumps are limited
 # to +/- 16 MB
-KLIBCSHAREDFLAGS     = -Ttext-segment 0x0f800000
+KLIBCSHAREDFLAGS     = $(IMAGE_BASE) 0x0f800000

 # The asm include files live in asm-powerpc
 KLIBCASMARCH = powerpc
diff --git a/usr/klibc/arch/ppc64/MCONFIG b/usr/klibc/arch/ppc64/MCONFIG
index d6e3a85c4b5f..6b553f9787bb 100644
--- a/usr/klibc/arch/ppc64/MCONFIG
+++ b/usr/klibc/arch/ppc64/MCONFIG
@@ -26,7 +26,7 @@ endif
 # calls, and work on the memory models for this architecture
 # 256-16 MB - normal binaries start at 256 MB, and jumps are limited
 # to +/- 16 MB
-KLIBCSHAREDFLAGS     = -Ttext-segment 0x0f000000
+KLIBCSHAREDFLAGS     = $(IMAGE_BASE) 0x0f000000

 # The asm include files live in asm-powerpc
 KLIBCASMARCH = powerpc
diff --git a/usr/klibc/arch/riscv64/MCONFIG b/usr/klibc/arch/riscv64/MCONFIG
index 61681509d89d..2ea257548417 100644
--- a/usr/klibc/arch/riscv64/MCONFIG
+++ b/usr/klibc/arch/riscv64/MCONFIG
@@ -17,4 +17,4 @@ endif
 KLIBCBITSIZE      = 64

 # Normal binaries start at 64 KB, so start the libary at 2 MB.
-KLIBCSHAREDFLAGS  =-Ttext-segment 0x00200000
+KLIBCSHAREDFLAGS  = $(IMAGE_BASE) 0x00200000
diff --git a/usr/klibc/arch/sparc64/MCONFIG b/usr/klibc/arch/sparc64/MCONFIG
index c83398559f84..767232fce80e 100644
--- a/usr/klibc/arch/sparc64/MCONFIG
+++ b/usr/klibc/arch/sparc64/MCONFIG
@@ -18,4 +18,4 @@ KLIBCLDFLAGS      = -m elf64_sparc
 # calls, and work on the memory models for this architecture
 # Normal binaries start at 1 MB; the linker wants 1 MB alignment,
 # and call instructions have a 30-bit signed offset, << 2.
-KLIBCSHAREDFLAGS = -Ttext-segment 0x80000000
+KLIBCSHAREDFLAGS = $(IMAGE_BASE) 0x80000000
diff --git a/usr/klibc/arch/x86_64/MCONFIG b/usr/klibc/arch/x86_64/MCONFIG
index a68d5692e04c..08a05f51863b 100644
--- a/usr/klibc/arch/x86_64/MCONFIG
+++ b/usr/klibc/arch/x86_64/MCONFIG
@@ -41,4 +41,4 @@ KLIBCLDFLAGS      = -m elf_x86_64
 # The old default was max-page-size=0x100000, but that also results
 # in a broken layout with binutils 2.30.  Since there's no
 # architectural page size betwen 4 KB and 2MB, set it to 4 KB.
-KLIBCSHAREDFLAGS     = -Ttext-segment 0x00200000 -z max-page-size=0x1000
+KLIBCSHAREDFLAGS     = $(IMAGE_BASE) 0x00200000 -z max-page-size=0x1000



bw
Previous message (by thread): [klibc] [kvm-unit-tests PATCH v3 0/4] Update patch set
Next message (by thread): [klibc] [kvm-unit-tests PATCH v3 0/4] Update patch set
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the klibc mailing list