Re: [PATCH 1/2] mips: vdso: fix 'jalr t9' crash in vdso code

["Victor Kamensky (kamensky)" <kamensky@cisco.com>]

Hi Paul,

Appreciate you looking into this.

Please see inline. Look for 'kamensky>'. Please forgive
my awful corporate mail agent deficiencies with inline replies.
Please let me know if you cannot read my inline replies,
I will re-post this message from my personal account.

________________________________________
From: linux-mips-owner@vger.kernel.org <linux-mips-owner@vger.kernel.org> on behalf of Paul Burton <paulburton@kernel.org>
Sent: Monday, February 10, 2020 11:33 AM
To: Victor Kamensky (kamensky)
Cc: linux-mips@vger.kernel.org; Ralf Baechle; James Hogan; Vincenzo Frascino; bruce.ashfield@gmail.com; richard.purdie@linuxfoundation.org
Subject: Re: [PATCH 1/2] mips: vdso: fix 'jalr t9' crash in vdso code

Hi Victor,

On Mon, Feb 10, 2020 at 11:14:24AM -0800, Paul Burton wrote:
> Which kernel version did you find this issue with? Could you check
> whether you have commit bbcc5672b006 ("MIPS: Avoid VDSO ABI breakage due
> to global register variable")? That was merged for v5.5-rc5.

Ah, I just spotted your detailed cover letter - my bad :)

Please try moving from v5.4 to v5.4.18 or higher, or just cherry-picking
the commit I mentioned.

kamensky> Double commit that you mentioned is present in 5.4.15 (since
kamensky> v5.4.9) that I use. It is 5b004a238460113276319536534928c58d95e599

[kamensky@kamensky-p53s linux]$ git tag --contains 5b004a238460113276319536534928c58d95e599
v5.4.10
v5.4.11
v5.4.12
v5.4.13
v5.4.14
v5.4.15
v5.4.16
v5.4.17
v5.4.18
v5.4.9

kamensky> You can check your tree: disassemble arch/mips/vdso/vgettimeofday-n32.o
kamensky> you would see 'jalr t9' calls, then disassemble arsm/mips/vdso/vdos-n32.so
kamensky> and check that those 'jalr t9' places got converted into 'bal' instructions,
kamensky> even though t9 set up instructions are there. This conversion happens by
kamensky> ld during relocation processing if code was compiled with options I propose
kamensky> to add explicitly. Walk through gcc and binutils places that do that are in my
kamensky> cover letter write up.

kamensky> You can reproduce the issue by doing opposite of what I propose:
kamensky> disable -mrelax-pic-calls and -mexplicit-relocs defaults in your
kamensky> toolchain by either using negate variant of option, ie
kamensky> -mno-relax-pic-calls or -mno-explicit-relocs or both. Something
kamensky> like in diff below. I've tried it on mips-poky-linux-gcc toolchain
kamensky> that got correct defaults and worked for us before, after adding
kamensky> no variants it shows the same 'jalr t9' crash as with other yocto
kamensky> mips toolchains.

diff --git a/arch/mips/vdso/Makefile b/arch/mips/vdso/Makefile
index 996a934ece7d..2131c3082580 100644
--- a/arch/mips/vdso/Makefile
+++ b/arch/mips/vdso/Makefile
@@ -31,7 +31,9 @@ cflags-vdso := $(ccflags-vdso) \
        -O3 -g -fPIC -fno-strict-aliasing -fno-common -fno-builtin -G 0 \
        -fno-stack-protector -fno-jump-tables -DDISABLE_BRANCH_PROFILING \
        $(call cc-option, -fno-asynchronous-unwind-tables) \
-       $(call cc-option, -fno-stack-protector)
+       $(call cc-option, -fno-stack-protector) \
+       $(call cc-option, -mno-relax-pic-calls) \
+       $(call cc-option, -mno-explicit-relocs)

Thanks,
Victor

Thanks,
    Paul