LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
From: Saleem Abdulrasool <abdulras@google.com>
To: Palmer Dabbelt <palmer@dabbelt.com>
Cc: linux-riscv@lists.infradead.org, linux-kernel@vger.kernel.org,
	nathan@kernel.org, Nick Desaulniers <ndesaulniers@google.com>,
	Bill Wendling <morbo@google.com>,
	clang-built-linux@googlegroups.com
Subject: Re: [PATCH] riscv: explicitly use symbol offsets for VDSO v2
Date: Wed, 4 Aug 2021 10:35:29 -0700	[thread overview]
Message-ID: <CAO8XFHs-s2SHajL8NpWsK_JcG0ZoyDr9qd9jGDxdQ6bU-cKqiw@mail.gmail.com> (raw)
In-Reply-To: <mhng-fc0ced4d-4e60-422b-8728-7e9bdd0276fb@palmerdabbelt-glaptop>

On Tue, Aug 3, 2021 at 8:55 PM Palmer Dabbelt <palmer@dabbelt.com> wrote:
>
> On Mon, 02 Aug 2021 14:00:17 PDT (-0700), abdulras@google.com wrote:
> > The current implementation of the `__rt_sigaction` reference computed an
> > absolute offset relative to the mapped base of the VDSO.  While this can
> > be handled in the medlow model, the medany model cannot handle this as
> > it is meant to be position independent.  The current implementation
> > relied on the BFD linker relaxing the PC-relative relocation into an
> > absolute relocation as it was a near-zero address allowing it to be
> > referenced relative to `zero`.
> >
> > We now extract the offsets and create a generated header allowing the
> > build with LLVM and lld to succeed as we no longer depend on the linker
> > rewriting address references near zero.  This change was largely
> > modelled after the ARM64 target which does something similar.
> >
> > Signed-off-by: Saleem Abdulrasool <abdulras@google.com>
> > ---
> >  arch/riscv/Makefile                        |  4 ++++
> >  arch/riscv/include/asm/vdso.h              | 11 ++--------
> >  arch/riscv/kernel/vdso/Makefile            | 25 ++++++++++------------
> >  arch/riscv/kernel/vdso/gen_vdso_offsets.sh |  5 +++++
> >  arch/riscv/kernel/vdso/so2s.sh             |  6 ------
> >  5 files changed, 22 insertions(+), 29 deletions(-)
> >  create mode 100755 arch/riscv/kernel/vdso/gen_vdso_offsets.sh
> >  delete mode 100755 arch/riscv/kernel/vdso/so2s.sh
> >
> > diff --git a/arch/riscv/Makefile b/arch/riscv/Makefile
> > index bc74afdbf31e..e026b2d0a5a4 100644
> > --- a/arch/riscv/Makefile
> > +++ b/arch/riscv/Makefile
> > @@ -108,6 +108,10 @@ PHONY += vdso_install
> >  vdso_install:
> >       $(Q)$(MAKE) $(build)=arch/riscv/kernel/vdso $@
> >
> > +prepare: vdso_prepare
> > +vdso_prepare: prepare0
> > +     $(Q)$(MAKE) $(build)=arch/riscv/kernel/vdso include/generated/vdso-offsets.h
> > +
> >  ifneq ($(CONFIG_XIP_KERNEL),y)
> >  ifeq ($(CONFIG_RISCV_M_MODE)$(CONFIG_SOC_CANAAN),yy)
> >  KBUILD_IMAGE := $(boot)/loader.bin
> > diff --git a/arch/riscv/include/asm/vdso.h b/arch/riscv/include/asm/vdso.h
> > index 1453a2f563bc..098795262b92 100644
> > --- a/arch/riscv/include/asm/vdso.h
> > +++ b/arch/riscv/include/asm/vdso.h
> > @@ -9,24 +9,17 @@
> >  #define _ASM_RISCV_VDSO_H
> >
> >  #include <linux/types.h>
> > +#include <generated/vdso-offsets.h>
> >
> >  #ifndef CONFIG_GENERIC_TIME_VSYSCALL
> >  struct vdso_data {
> >  };
> >  #endif
> >
> > -/*
> > - * The VDSO symbols are mapped into Linux so we can just use regular symbol
> > - * addressing to get their offsets in userspace.  The symbols are mapped at an
> > - * offset of 0, but since the linker must support setting weak undefined
> > - * symbols to the absolute address 0 it also happens to support other low
> > - * addresses even when the code model suggests those low addresses would not
> > - * otherwise be availiable.
> > - */
> >  #define VDSO_SYMBOL(base, name)                                                      \
> >  ({                                                                           \
> >       extern const char __vdso_##name[];                                      \
> > -     (void __user *)((unsigned long)(base) + __vdso_##name);                 \
> > +     (void __user *)((unsigned long)(base) + __vdso_##name##_offset);        \
> >  })
> >
> >  asmlinkage long sys_riscv_flush_icache(uintptr_t, uintptr_t, uintptr_t);
> > diff --git a/arch/riscv/kernel/vdso/Makefile b/arch/riscv/kernel/vdso/Makefile
> > index 24d936c147cd..f8cb9144a284 100644
> > --- a/arch/riscv/kernel/vdso/Makefile
> > +++ b/arch/riscv/kernel/vdso/Makefile
> > @@ -23,10 +23,10 @@ ifneq ($(c-gettimeofday-y),)
> >  endif
> >
> >  # Build rules
> > -targets := $(obj-vdso) vdso.so vdso.so.dbg vdso.lds vdso-syms.S
> > +targets := $(obj-vdso) vdso.so vdso.so.dbg vdso.lds
> >  obj-vdso := $(addprefix $(obj)/, $(obj-vdso))
> >
> > -obj-y += vdso.o vdso-syms.o
> > +obj-y += vdso.o
> >  CPPFLAGS_vdso.lds += -P -C -U$(ARCH)
> >
> >  # Disable -pg to prevent insert call site
> > @@ -43,20 +43,22 @@ $(obj)/vdso.o: $(obj)/vdso.so
> >  # link rule for the .so file, .lds has to be first
> >  $(obj)/vdso.so.dbg: $(obj)/vdso.lds $(obj-vdso) FORCE
> >       $(call if_changed,vdsold)
> > -LDFLAGS_vdso.so.dbg = -shared -s -soname=linux-vdso.so.1 \
> > +LDFLAGS_vdso.so.dbg = -shared -S -soname=linux-vdso.so.1 \
> >       --build-id=sha1 --hash-style=both --eh-frame-hdr
> >
> > -# We also create a special relocatable object that should mirror the symbol
> > -# table and layout of the linked DSO. With ld --just-symbols we can then
> > -# refer to these symbols in the kernel code rather than hand-coded addresses.
> > -$(obj)/vdso-syms.S: $(obj)/vdso.so FORCE
> > -     $(call if_changed,so2s)
> > -
> >  # strip rule for the .so file
> >  $(obj)/%.so: OBJCOPYFLAGS := -S
> >  $(obj)/%.so: $(obj)/%.so.dbg FORCE
> >       $(call if_changed,objcopy)
> >
> > +# Generate VDSO offsets using helper script
> > +gen-vdsosym := $(srctree)/$(src)/gen_vdso_offsets.sh
> > +quiet_cmd_vdsosym = VDSOSYM $@
> > +     cmd_vdsosym = $(NM) $< | $(gen-vdsosym) | LC_ALL=C sort > $@
> > +
> > +include/generated/vdso-offsets.h: $(obj)/vdso.so.dbg FORCE
> > +     $(call if_changed,vdsosym)
> > +
> >  # actual build commands
> >  # The DSO images are built using a special linker script
> >  # Make sure only to export the intended __vdso_xxx symbol offsets.
> > @@ -65,11 +67,6 @@ quiet_cmd_vdsold = VDSOLD  $@
> >                     $(OBJCOPY) $(patsubst %, -G __vdso_%, $(vdso-syms)) $@.tmp $@ && \
> >                     rm $@.tmp
> >
> > -# Extracts symbol offsets from the VDSO, converting them into an assembly file
> > -# that contains the same symbols at the same offsets.
> > -quiet_cmd_so2s = SO2S    $@
> > -      cmd_so2s = $(NM) -D $< | $(srctree)/$(src)/so2s.sh > $@
> > -
> >  # install commands for the unstripped file
> >  quiet_cmd_vdso_install = INSTALL $@
> >        cmd_vdso_install = cp $(obj)/$@.dbg $(MODLIB)/vdso/$@
> > diff --git a/arch/riscv/kernel/vdso/gen_vdso_offsets.sh b/arch/riscv/kernel/vdso/gen_vdso_offsets.sh
> > new file mode 100755
> > index 000000000000..c2e5613f3495
> > --- /dev/null
> > +++ b/arch/riscv/kernel/vdso/gen_vdso_offsets.sh
> > @@ -0,0 +1,5 @@
> > +#!/bin/sh
> > +# SPDX-License-Identifier: GPL-2.0
> > +
> > +LC_ALL=C
> > +sed -n -e 's/^[0]\+\(0[0-9a-fA-F]*\) . \(__vdso_[a-zA-Z0-9_]*\)$/\#define \2_offset\t0x\1/p'
> > diff --git a/arch/riscv/kernel/vdso/so2s.sh b/arch/riscv/kernel/vdso/so2s.sh
> > deleted file mode 100755
> > index e64cb6d9440e..000000000000
> > --- a/arch/riscv/kernel/vdso/so2s.sh
> > +++ /dev/null
> > @@ -1,6 +0,0 @@
> > -#!/bin/sh
> > -# SPDX-License-Identifier: GPL-2.0+
> > -# Copyright 2020 Palmer Dabbelt <palmerdabbelt@google.com>
> > -
> > -sed 's!\([0-9a-f]*\) T \([a-z0-9_]*\)\(@@LINUX_4.15\)*!.global \2\n.set \2,0x\1!' \
> > -| grep '^\.'
>
> This is giving me some unused variable warnings:
>
> In file included from arch/riscv/kernel/signal.c:16:
> arch/riscv/kernel/signal.c: In function ‘setup_rt_frame’:
> ./arch/riscv/include/asm/vdso.h:21:20: warning: unused variable ‘__vdso_rt_sigreturn’ [-Wunused-variable]
>    21 |  extern const char __vdso_##name[];     \
>       |                    ^~~~~~~
> arch/riscv/kernel/signal.c:197:28: note: in expansion of macro ‘VDSO_SYMBOL’
>   197 |  regs->ra = (unsigned long)VDSO_SYMBOL(
>       |                            ^~~~~~~~~~~

Thanks Palmer,

It seems that there was a difference in diagnostics between gcc and
clang.  The diagnostic indicated that I didn't scrub hard enough
during the cleanups.  I've sent out an updated version.  Thank you for
informing me of this missed cleanup!

      reply	other threads:[~2021-08-04 17:35 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-08-02 21:00 Saleem Abdulrasool
2021-08-02 23:14 ` kernel test robot
2021-08-04  3:54 ` Palmer Dabbelt
2021-08-04 17:35   ` Saleem Abdulrasool [this message]

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=CAO8XFHs-s2SHajL8NpWsK_JcG0ZoyDr9qd9jGDxdQ6bU-cKqiw@mail.gmail.com \
    --to=abdulras@google.com \
    --cc=clang-built-linux@googlegroups.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-riscv@lists.infradead.org \
    --cc=morbo@google.com \
    --cc=nathan@kernel.org \
    --cc=ndesaulniers@google.com \
    --cc=palmer@dabbelt.com \
    --subject='Re: [PATCH] riscv: explicitly use symbol offsets for VDSO v2' \
    /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

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).