linux.git/arch/powerpc/kernel/vdso32, branch v5.12

powerpc/vdso32: Add missing _restgpr_31_x to fix build failure

2021-03-14T09:32:23+00:00

With some defconfig including CONFIG_CC_OPTIMIZE_FOR_SIZE,
(for instance mvme5100_defconfig and ps3_defconfig), gcc 5
generates a call to _restgpr_31_x.

Until recently it went unnoticed, but
commit 42ed6d56ade2 ("powerpc/vdso: Block R_PPC_REL24 relocations")
made it rise to the surface.

Provide that function (copied from lib/crtsavres.S) in
gettimeofday.S

Fixes: ab037dd87a2f ("powerpc/vdso: Switch VDSO to generic C implementation.")
Signed-off-by: Christophe Leroy 
Signed-off-by: Michael Ellerman 
Link: https://lore.kernel.org/r/a7aa198a88bcd33c6e35e99f70f86c7b7f2f9440.1615270757.git.christophe.leroy@csgroup.eu

powerpc/vdso: fix unnecessary rebuilds of vgettimeofday.o

2021-01-30T11:23:42+00:00

vgettimeofday.o is unnecessarily rebuilt. Adding it to 'targets' is not
enough to fix the issue. Kbuild is correctly rebuilding it because the
command line is changed.

PowerPC builds each vdso directory twice; first in vdso_prepare to
generate vdso{32,64}-offsets.h, second as part of the ordinary build
process to embed vdso{32,64}.so.dbg into the kernel.

The problem shows up when CONFIG_PPC_WERROR=y due to the following line
in arch/powerpc/Kbuild:

  subdir-ccflags-$(CONFIG_PPC_WERROR) := -Werror

In the preparation stage, Kbuild directly visits the vdso directories,
hence it does not inherit subdir-ccflags-y. In the second descend,
Kbuild adds -Werror, which results in the command line flipping
with/without -Werror.

It implies a potential danger; if a more critical flag that would impact
the resulted vdso, the offsets recorded in the headers might be different
from real offsets in the embedded vdso images.

Removing the unneeded second descend solves the problem.

Reported-by: Michael Ellerman 
Signed-off-by: Masahiro Yamada 
Signed-off-by: Michael Ellerman 
Link: https://lore.kernel.org/linuxppc-dev/87tuslxhry.fsf@mpe.ellerman.id.au/
Link: https://lore.kernel.org/r/20201223171142.707053-1-masahiroy@kernel.org

powerpc/vdso: Don't pass 64-bit ABI cflags to 32-bit VDSO

2020-12-21T11:06:26+00:00

When building the 32-bit VDSO, we are building 32-bit code as part of
a 64-bit kernel build. That requires us to tweak the cflags to trick
the compiler into building 32-bit code for us. The main way we do that
is by passing -m32, but there are other options that affect code
generation and ABI selection.

In particular when building vgettimeofday.c, we end up passing
-mcall-aixdesc because it's in KBUILD_CFLAGS, which causes the
compiler to generate function descriptors, and dot symbols, eg:

  $ nm arch/powerpc/kernel/vdso32/vgettimeofday.o
  000005d0 T .__c_kernel_clock_getres
  00000024 D __c_kernel_clock_getres
  ...

We get away with that at the moment because we also use the DOTSYM
macro, and that is also incorrectly prepending a '.' in 32-bit VDSO
code due to a separate bug.

But we shouldn't be generating function descriptors for this file,
there's no 32-bit ABI that includes function descriptors, so the
resulting object file is some frankenstein and it's surprising that it
even links.

So filter out all the ABI-related options we add to CFLAGS for 64-bit
builds, so that they're not used when building 32-bit code. With that
we only see regular text symbols:

  $ nm arch/powerpc/kernel/vdso32/vgettimeofday.o                                                                                                                                     michael@alpine1-p1
  000005d0 T __c_kernel_clock_getres
  00000000 T __c_kernel_clock_gettime
  00000200 T __c_kernel_clock_gettime64
  00000410 T __c_kernel_gettimeofday
  00000650 T __c_kernel_time

Fixes: ab037dd87a2f ("powerpc/vdso: Switch VDSO to generic C implementation.")
Signed-off-by: Michael Ellerman 
Link: https://lore.kernel.org/r/20201218111619.1206391-2-mpe@ellerman.id.au

powerpc/vdso: Block R_PPC_REL24 relocations

2020-12-21T11:06:25+00:00

Add R_PPC_REL24 relocations to the list of relocations we do NOT
support in the VDSO.

These are generated in some cases and we do not support relocating
them at runtime, so if they appear then the VDSO will not work at
runtime, therefore it's preferable to break the build if we see them.

Fixes: ab037dd87a2f ("powerpc/vdso: Switch VDSO to generic C implementation.")
Signed-off-by: Michael Ellerman 
Link: https://lore.kernel.org/r/20201218111619.1206391-1-mpe@ellerman.id.au

powerpc: Use common STABS_DEBUG and DWARF_DEBUG and ELF_DETAILS macro

2020-12-03T14:01:20+00:00

Use the common STABS_DEBUG and DWARF_DEBUG and ELF_DETAILS macro rule for
the linker script in an effort.

Signed-off-by: Youling Tang 
Signed-off-by: Michael Ellerman 
Link: https://lore.kernel.org/r/1606460857-2723-1-git-send-email-tangyouling@loongson.cn

powerpc/vdso: Remove VDSO32_LBASE and VDSO64_LBASE

2020-12-03T14:01:19+00:00

VDSO32_LBASE and VDSO64_LBASE are 0. Remove them to simplify code.

Signed-off-by: Christophe Leroy 
Signed-off-by: Michael Ellerman 
Link: https://lore.kernel.org/r/6c4d6570d886bbe1cc471e8ca01602e4b4d9beb5.1601197618.git.christophe.leroy@csgroup.eu

powerpc/vdso: Remove __kernel_datapage_offset

2020-12-03T14:01:18+00:00

__kernel_datapage_offset is not used anymore, remove it.

Signed-off-by: Christophe Leroy 
Signed-off-by: Michael Ellerman 
Link: https://lore.kernel.org/r/ddb5c746bec4e1a026d7c85243213a1876ef844f.1601197618.git.christophe.leroy@csgroup.eu

powerpc/vdso: Merge __kernel_sync_dicache_p5() into __kernel_sync_dicache()

2020-12-03T14:01:17+00:00

__kernel_sync_dicache_p5() is an alternative to
__kernel_sync_dicache() when cpu has CPU_FTR_COHERENT_ICACHE

Remove this alternative function and merge
__kernel_sync_dicache_p5() into __kernel_sync_dicache() using
standard CPU feature fixup.

Signed-off-by: Christophe Leroy 
Signed-off-by: Michael Ellerman 
Link: https://lore.kernel.org/r/4c7dcc6544882761b2b0249d7a8ec2c3a8088cb5.1601197618.git.christophe.leroy@csgroup.eu

powerpc/vdso: Use builtin symbols to locate fixup section

2020-12-03T14:01:17+00:00

Add builtin symbols to locate fixup section and use them
instead of locating sections through elf headers at runtime.

Signed-off-by: Christophe Leroy 
Signed-off-by: Michael Ellerman 
Link: https://lore.kernel.org/r/2954526981859ca1ccfcfc7a7c4263920e9ddfcb.1601197618.git.christophe.leroy@csgroup.eu

powerpc/vdso: Retrieve sigtramp offsets at buildtime

2020-12-03T14:01:17+00:00

This is copied from arm64.

Instead of using runtime generated signal trampoline offsets,
get offsets at buildtime.

If the said trampoline doesn't exist, build will fail. So no
need to check whether the trampoline exists or not in the VDSO.

Signed-off-by: Christophe Leroy 
Signed-off-by: Michael Ellerman 
Link: https://lore.kernel.org/r/f8bfd6812c3e3678b1cdb4d55a52f9eb022b40d3.1601197618.git.christophe.leroy@csgroup.eu