Linux PARISC architecture development
 help / color / mirror / Atom feed
From: Vineet Gupta <vgupta@kernel.org>
To: "Arnd Bergmann" <arnd@arndb.de>,
	"Thomas Weißschuh" <thomas.weissschuh@linutronix.de>,
	"Dinh Nguyen" <dinguyen@kernel.org>,
	"schuster.simon@siemens-energy.com"
	<schuster.simon@siemens-energy.com>,
	"Ley Foon Tan" <lftan@altera.com>
Cc: linux-kernel@vger.kernel.org, kernel test robot <lkp@intel.com>,
	linux-snps-arc@lists.infradead.org, linux-parisc@vger.kernel.org,
	"linux-openrisc@vger.kernel.org" <linux-openrisc@vger.kernel.org>
Subject: Re: [PATCH] nios2: ksyms: Declare __mulsi3 export
Date: Sat, 31 Oct 2026 16:57:11 -0700	[thread overview]
Message-ID: <3175caa0-9f49-465d-959e-6b6c9bbb4f01@kernel.org> (raw)
In-Reply-To: <47eb6b48-d664-440f-84b6-ba09f34d5e08@app.fastmail.com>

On 10/19/26 04:16, Arnd Bergmann wrote:
> On Mon, Oct 19, 2026, at 11:49, Thomas Weißschuh wrote:
>> GCC may emit libcalls to __mulsi3(). As modpost is not aware that this
>> function is provided through the linked in libgcc, it fails:
>>
>> ERROR: modpost: kernel/locking/test-ww_mutex.ko: symbol '__mulsi3' undefined!
>> ERROR: modpost: kernel/time/test_udelay.ko: symbol '__mulsi3' undefined!
>> ERROR: modpost: kernel/resource_kunit.ko: symbol '__mulsi3' undefined!
>> ERROR: modpost: fs/fuse/virtiofs.ko: symbol '__mulsi3' undefined!
>> etc.
>>
>> Make sure modpost can see the symbol from libgcc.
>>
>> Reported-by: kernel test robot <lkp@intel.com>
>> Closes:
>> https://lore.kernel.org/oe-kbuild-all/202610171933.J4bFOfL8-lkp@intel.com/
>> Fixes: 2fc8483fdcde ("nios2: Build infrastructure")
>> Signed-off-by: Thomas Weißschuh <thomas.weissschuh@linutronix.de>
> Acked-by: Arnd Bergmann <arnd@arndb.de>
>
> The patch looks correct to me, but it would also be nice to not use
> libgcc at all. What I see is that only arc, nios2, openrisc and
> parisc still link against the toolchain-provided libgcc, everything
> else uses the kernel's own runtime.
>
> It's probably not even that hard to do right, here is what I found
> is imported from libgcc.

And seems like ARC is the biggest offender, what a shame.

> * Generic library helper exists, can be enabled through Kconfig:
>
> === __lshrdi3 ===
> arch/arc/kernel/arcksyms.c
> arch/nios2/kernel/nios2_ksyms.c
> arch/openrisc/kernel/or32_ksyms.c
> arch/parisc/kernel/parisc_ksyms.c
> -> lib/lshrdi3.c
> === __ashldi3 ===
> arch/arc/kernel/arcksyms.c
> arch/nios2/kernel/nios2_ksyms.c
> arch/openrisc/kernel/or32_ksyms.c
> arch/parisc/kernel/parisc_ksyms.c
> -> lib/ashldi3.c
> === __ashrdi3 ===
> arch/arc/kernel/arcksyms.c
> arch/nios2/kernel/nios2_ksyms.c
> arch/openrisc/kernel/or32_ksyms.c
> arch/parisc/kernel/parisc_ksyms.c
> -> lib/ashrdi3.c
> === __cmpdi2 ===
> arch/arc/kernel/arcksyms.c
> -> lib/cmpdi2.c
> === __muldi3 ===
> arch/arc/kernel/arcksyms.c
> arch/nios2/kernel/nios2_ksyms.c
> arch/openrisc/kernel/or32_ksyms.c
> arch/parisc/kernel/parisc_ksyms.c
> -> lib/muldi3.c
> === __ucmpdi2 ===
> arch/arc/kernel/arcksyms.c
> arch/nios2/kernel/nios2_ksyms.c
> arch/openrisc/kernel/or32_ksyms.c
> arch/parisc/kernel/parisc_ksyms.c
> -> lib/ucmpdi2.c

It seems Palmer added those back in 2018 which ARC was still developed 
actively, I totally missed them.
I now have a patch to switch ARC to generic helpers for those.

> * 64-bit division probably should not be provided on arc or
>    nios2 (and sparc32, which has a private copy).
>    parisc64 should get these from libgcc/libgcc2.c.
>
> === __divdi3 ===
> arch/parisc/kernel/parisc_ksyms.c
> === __udivmoddi4 ===
> arch/nios2/kernel/nios2_ksyms.c
> === __udivdi3 ===
> arch/arc/kernel/arcksyms.c

It seems in some of the old binaries I have, dw mmc driver was at some 
point calling 64/64 divide which eventually got fixed.
But agree as of today there are no callers of 64-bit div/mod. I will get 
rid of ?div*di* from ARC.


> arch/parisc/kernel/parisc_ksyms.c
> === __umoddi3 ===
> arch/nios2/kernel/nios2_ksyms.c
> arch/parisc/kernel/parisc_ksyms.c
> === __moddi3 ===
> arch/nios2/kernel/nios2_ksyms.c
> arch/parisc/kernel/parisc_ksyms.c
>
> * floating-point math should not be provided by kernel:
>
> === __divdf3 ===
> arch/arc/kernel/arcksyms.c
> === __divsf3 ===
> arch/arc/kernel/arcksyms.c
> === __fixunsdfsi ===
> arch/arc/kernel/arcksyms.c
> === __floatunsidf ===
> arch/arc/kernel/arcksyms.c
> === __floatunsisf ===
> arch/arc/kernel/arcksyms.c
> === __muldf3 ===
> arch/arc/kernel/arcksyms.c

And a second patch gets rid of those. FWIW back in the day some ARCAngel 
board's HDD driver needed those. Crazy times.

> * no generic implementation in kernel, need to copy from
>    libgcc/lib1funcs.S code:
>
> === __divsi3 ===
> arch/arc/kernel/arcksyms.c
> arch/nios2/kernel/nios2_ksyms.c
> arch/openrisc/kernel/or32_ksyms.c
> -> arch/arm/lib/lib1funcs.S
>     arch/m68k/lib/divsi3.S
>     arch/microblaze/lib/divsi3.S
>     arch/xtensa/lib/divsi3.S
> === __modsi3 ===
> arch/arc/kernel/arcksyms.c
> arch/nios2/kernel/nios2_ksyms.c
> arch/openrisc/kernel/or32_ksyms.c
> -> arch/m68k/lib/modsi3.S
>     arch/microblaze/lib/modsi3.S
>     arch/xtensa/lib/modsi3.S
> === __udivsi3 ===
> arch/arc/kernel/arcksyms.c
> arch/nios2/kernel/nios2_ksyms.c
> arch/openrisc/kernel/or32_ksyms.c
> -> arch/arm/lib/lib1funcs.S
>     arch/m68k/lib/udivsi3.S
>     arch/microblaze/lib/udivsi3.S
>     arch/sh/lib/udivsi3.S
>     arch/xtensa/lib/udivsi3.S
> === __umodsi3 ===
> arch/arc/kernel/arcksyms.c
> arch/nios2/kernel/nios2_ksyms.c
> arch/openrisc/kernel/or32_ksyms.c
> -> arch/arm/lib/lib1funcs.S
>     arch/m68k/lib/umodsi3.S
>     arch/microblaze/lib/umodsi3.S
>     arch/xtensa/lib/umodsi3.S

ARCv2 has hardware divide so these are not needed, but ARC700 doesn't 
and there's a ton of calls for those.

┌───────────┬───────┬───────────┐
│  routine  │ calls │ functions │
├───────────┼───────┼───────────┤
│ __udivsi3 │ 415   │ 238       │
├───────────┼───────┼───────────┤
│ __umodsi3 │ 203   │ 131       │
├───────────┼───────┼───────────┤
│ __divsi3  │ 77    │ 57        │
├───────────┼───────┼───────────┤
│ __modsi3  │ 41    │ 33        │
└───────────┴───────┴───────────┘

I can take a stab at providing the generic libgcc implementation in 
lib/* but I only expect the arches using libgcc version to switch.

Thx,
-Vineet

  parent reply	other threads:[~2026-10-31 23:57 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <20261019-nios2-mulsi3-v1-1-4fd75b1bacef@linutronix.de>
2026-10-19 11:16 ` [PATCH] nios2: ksyms: Declare __mulsi3 export Arnd Bergmann
2026-10-19 14:34   ` Dinh Nguyen
2026-10-20  9:40     ` David Laight
2026-10-31 23:57   ` Vineet Gupta [this message]
2026-11-01  0:26     ` Vineet Gupta
2026-11-01 11:19       ` Arnd Bergmann

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=3175caa0-9f49-465d-959e-6b6c9bbb4f01@kernel.org \
    --to=vgupta@kernel.org \
    --cc=arnd@arndb.de \
    --cc=dinguyen@kernel.org \
    --cc=lftan@altera.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-openrisc@vger.kernel.org \
    --cc=linux-parisc@vger.kernel.org \
    --cc=linux-snps-arc@lists.infradead.org \
    --cc=lkp@intel.com \
    --cc=schuster.simon@siemens-energy.com \
    --cc=thomas.weissschuh@linutronix.de \
    /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
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox