[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20251030143400.09fc0a89@pumpkin>
Date: Thu, 30 Oct 2025 14:34:00 +0000
From: David Laight <david.laight.linux@...il.com>
To: Christian Kujau <lists@...dbynature.de>
Cc: linux-kernel@...r.kernel.org, "Liam R. Howlett"
 <Liam.Howlett@...cle.com>, Arnd Bergmann <arnd@...db.de>
Subject: Re: maple_tree.c:3738:1: error: the frame size of 1040 bytes is
 larger than 1024 bytes
On Thu, 30 Oct 2025 14:35:12 +0100 (CET)
Christian Kujau <lists@...dbynature.de> wrote:
> Hi,
> 
> while cross-compiling the latest mainline checkout for x86_64, this 
> happened:
> 
> ------------------
> $ make -j24 O=${DIR} ARCH=x86 CROSS_COMPILE=x86_64-linux-gnu- \
>             CC="x86_64-linux-gnu-gcc" bzImage
> [...]
>   CC      lib/maple_tree.o
> lib/maple_tree.c: In function ‘mas_wr_bnode’:
> lib/maple_tree.c:3738:1: error: the frame size of 1040 bytes is larger than 1024 bytes [-Werror=frame-larger-than=]
>  3738 | }
>       | ^
> cc1: all warnings being treated as errors
> ------------------
> 
> This is basically "allnoconfig" with CONFIG_WERROR and some debugging 
> options enabled. I found 44081c77e8a4 ("maple_tree: reduce stack usage 
> with gcc-9 and earlier") where a similar issue has been reported before, 
> but this was 2023. I'm trying to build this on a Debian/13 arm64 system:
> 
> $ x86_64-linux-gnu-gcc --version | head -1
> x86_64-linux-gnu-gcc (Debian 14.2.0-19) 14.2.0
> 
> Git bisect pointed to:
> 
> ------------------
>  commit 9b05890a25d9197e39fcf5b2298f0b911c323306
>  Author: Liam R. Howlett <Liam.Howlett@...cle.com>
>  Date:   Wed Sep 3 15:00:01 2025 +0200
> 
>     maple_tree: Prefilled sheaf conversion and testing
> ------------------
> 
> This looks related, but I was unable to revert only this commit. I can try 
> harder, but maybe someone has an idea how to tackle this.
Does marking mas_wr_bnode() 'noinline' help?
Some functions are marked noinline for KASAN builds, the comment suggests
mas_wr_bnode() is one of them - but it isn't marked at all.
I'd try an unconditional noinline first.
	David
> 
> Thanks,
> Christian.
Powered by blists - more mailing lists
 
