linux-kernel - Re: linux-next: Tree for Aug 21 (riscv-64 asm)

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <87r0nu19i3.fsf@all.your.base.are.belong.to.us>
Date:   Wed, 23 Aug 2023 09:52:04 +0200
From:   Björn Töpel <bjorn@...nel.org>
To:     Randy Dunlap <rdunlap@...radead.org>,
        Stephen Rothwell <sfr@...b.auug.org.au>,
        Linux Next Mailing List <linux-next@...r.kernel.org>
Cc:     Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        linux-riscv <linux-riscv@...ts.infradead.org>
Subject: Re: linux-next: Tree for Aug 21 (riscv-64 asm)

Björn Töpel <bjorn@...nel.org> writes:

> Björn Töpel <bjorn@...nel.org> writes:
>
>> Randy Dunlap <rdunlap@...radead.org> writes:
>>
>>> Hi,
>>>
>>> On 8/22/23 01:11, Björn Töpel wrote:
>>>> Randy Dunlap <rdunlap@...radead.org> writes:
>>>> 
>>>>> On 8/20/23 23:46, Stephen Rothwell wrote:
>>>>>> Hi all,
>>>>>>
>>>>>> Changes since 20230818:
>>>>>>
>>>>>
>>>>> ../arch/riscv/kernel/traps.c: In function 'do_irq':
>>>>> ../arch/riscv/kernel/traps.c:384:1: error: s0 cannot be used in 'asm' here
>>>>>   384 | }
>>>>>       | ^
>>>>> ../arch/riscv/kernel/irq.c: In function 'do_softirq_own_stack':
>>>>> ../arch/riscv/kernel/irq.c:94:1: error: s0 cannot be used in 'asm' here
>>>>>    94 | }
>>>>>       | ^
>>>>>
>>>>>
>>>>> 2 out of 10 risc-v 64-bit builds failed with these errors.
>>>>>
>>>>> Is this a toolchain problem or something else?
>>>> 
>>>> Hmm, do you have a link to config/toolchain/log, or similar?
>>>
>>> The full randconfig file is attached.
>>>
>>> The toolchain is Arnd's build of gcc 13.1.0 from:
>>>   https://mirrors.edge.kernel.org/pub/tools/crosstool/
>>> (x86_64 host)
>>>
>>> Hm, I see that the latest/current there is gcc 13.2.0.
>>> I'll upgrade later today.
>>
>> Thanks Randy! I can reproduce with GCC 12 on my machine.
>>
>> The config has
>>  | # CONFIG_FRAME_POINTER is not set
>> but for some reason "-fno-omit-frame-pointer" is passed to GCC anyway,
>> which is why GCC complains about s0 (FP on RISC-V) in asm.
>>
>>
>> I'll dig a bit more.
>
> There are multiple issues:
>  * riscv enables -fno-omit-frame-pointer unconditionally for PERF_EVENTS
>  * CONFIG_FUNCTION_TRACER w/o dynamic ftrace enables -pg, which pulls in
>    -fno-omit-frame-pointer on riscv.

Maybe something like (too crude?):

diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
index 10e7a7ad175a..cd8075c092c3 100644
--- a/arch/riscv/Kconfig
+++ b/arch/riscv/Kconfig
@@ -62,6 +62,7 @@ config RISCV
        select COMMON_CLK
        select CPU_PM if CPU_IDLE || HIBERNATION
        select EDAC_SUPPORT
+       select FRAME_POINTER if PERF_EVENTS || (FUNCTION_TRACER && !DYNAMIC_FTRACE)
        select GENERIC_ARCH_TOPOLOGY
        select GENERIC_ATOMIC64 if !64BIT
        select GENERIC_CLOCKEVENTS_BROADCAST if SMP
diff --git a/arch/riscv/Makefile b/arch/riscv/Makefile
index 6ec6d52a4180..1329e060c548 100644
--- a/arch/riscv/Makefile
+++ b/arch/riscv/Makefile
@@ -87,9 +87,6 @@ endif
 ifeq ($(CONFIG_CMODEL_MEDANY),y)
        KBUILD_CFLAGS += -mcmodel=medany
 endif
-ifeq ($(CONFIG_PERF_EVENTS),y)
-        KBUILD_CFLAGS += -fno-omit-frame-pointer
-endif
 
 # Avoid generating .eh_frame sections.
 KBUILD_CFLAGS += -fno-asynchronous-unwind-tables -fno-unwind-tables