5.10 kernel series fails to build on newer toolchain: FAILED unresolved symbol filp_close

5.10 kernel series fails to build on newer toolchain: FAILED unresolved symbol filp_close

[Philip Müller]

[-- Attachment #1: Type: text/plain, Size: 1411 bytes --]

For some odd reason 5.10 kernel series doesn't compile with a newer 
toolchain since 2025-02-09:

2025-02-09T17:32:07.7991299Z   GEN     .version
2025-02-09T17:32:07.8270062Z   CHK     include/generated/compile.h
2025-02-09T17:32:07.8540777Z   LD      vmlinux.o
2025-02-09T17:32:11.7210899Z   MODPOST vmlinux.symvers
2025-02-09T17:32:12.0869599Z   MODINFO modules.builtin.modinfo
2025-02-09T17:32:12.1403022Z   GEN     modules.builtin
2025-02-09T17:32:12.1475659Z   LD      .tmp_vmlinux.btf
2025-02-09T17:32:19.6117204Z   BTF     .btf.vmlinux.bin.o
2025-02-09T17:32:31.2916650Z   LD      .tmp_vmlinux.kallsyms1
2025-02-09T17:32:34.8731104Z   KSYMS   .tmp_vmlinux.kallsyms1.S
2025-02-09T17:32:35.4910608Z   AS      .tmp_vmlinux.kallsyms1.o
2025-02-09T17:32:35.9662538Z   LD      .tmp_vmlinux.kallsyms2
2025-02-09T17:32:39.2595984Z   KSYMS   .tmp_vmlinux.kallsyms2.S
2025-02-09T17:32:39.8802028Z   AS      .tmp_vmlinux.kallsyms2.o
2025-02-09T17:32:40.3659440Z   LD      vmlinux
2025-02-09T17:32:48.0031558Z   BTFIDS  vmlinux
2025-02-09T17:32:48.0143553Z FAILED unresolved symbol filp_close
2025-02-09T17:32:48.5019928Z make: *** [Makefile:1207: vmlinux] Error 255
2025-02-09T17:32:48.5061241Z ==> ERROR: A failure occurred in build().

5.10.234 built fine couple of days ago with the older one. There were 
slight changes made. 5.4 and 5.15 still compile.

Wonder what might be missing here ...

-- 
Best, Philip

[-- Attachment #2: toolchain.diff --]
[-- Type: text/x-patch, Size: 1989 bytes --]

--- old-t.txt	2025-03-14 08:15:56.205018752 +0700
+++ new-t.txt	2025-03-14 08:16:10.414018926 +0700
@@ -1,4 +1,4 @@
-Old toolchain:
+Newer toolchain:
 
 Packages (138) acl-2.3.2-1
 attr-2.5.2-1
@@ -6,16 +6,16 @@ audit-4.0.3-1
 autoconf-2.72-1
 automake-1.17-1
 bash-5.2.037-1
-binutils-2.43_1+r186+g61f8adadd6db-1
+binutils-2.44-1
 bison-3.8.2-8
 brotli-1.1.0-3
 bzip2-1.0.8-6
 ca-certificates-20240618-1
-ca-certificates-mozilla-3.107-1
+ca-certificates-mozilla-3.108-1
 ca-certificates-utils-20240618-1
 coreutils-9.6-2
 cryptsetup-2.7.5-2
-curl-8.11.1-3
+curl-8.12.0-3
 db5.3-5.3.28-5
 dbus-1.16.0-1
 dbus-broker-36-4
@@ -33,12 +33,12 @@ findutils-4.10.0-2
 flex-2.6.4-5
 gawk-5.3.1-1
 gc-8.2.8-2
-gcc-14.2.1+r730+gc061ad5a36ba-1
-gcc-libs-14.2.1+r730+gc061ad5a36ba-1
+gcc-14.2.1+r753+g1cd744a6828f-1
+gcc-libs-14.2.1+r753+g1cd744a6828f-1
 gdbm-1.24-1
 gettext-0.23.1-2
 glib2-2.82.4-2
-glibc-2.41+r2+g0a7c7a3e283a-1
+glibc-2.41+r6+gcf88351b685d-1
 gmp-6.3.0-2
 gnulib-l10n-20241231-1
 gnupg-2.4.7-1
@@ -48,7 +48,7 @@ grep-3.11-1
 groff-1.23.0-7
 guile-3.0.10-1
 gzip-1.13-4
-hwdata-0.391-1
+hwdata-0.392-1
 iana-etc-20241206-1
 icu-75.1-2
 jansson-2.14-4
@@ -61,7 +61,7 @@ libarchive-3.7.7-1
 libassuan-3.0.0-1
 libcap-2.71-1
 libcap-ng-0.8.5-3
-libelf-0.192-3
+libelf-0.192-4
 libevent-2.1.12-4
 libffi-3.4.6-1
 libgcrypt-1.11.0-2
@@ -80,10 +80,10 @@ libsasl-2.1.28-5
 libseccomp-2.5.5-4
 libsecret-0.21.6-1
 libssh2-1.11.1-1
-libsysprof-capture-47.2-2
-libtasn1-4.19.0-2
+libsysprof-capture-47.2-3
+libtasn1-4.20.0-1
 libtirpc-1.3.6-1
-libtool-2.5.4+r1+gbaa1fe41-2
+libtool-2.5.4+r1+gbaa1fe41-3
 libunistring-1.3-1
 libusb-1.0.27-1
 libverto-0.3.2-5
@@ -95,7 +95,7 @@ lz4-1:1.10.0-2
 m4-1.4.19-3
 make-4.4.1-2
 mpdecimal-4.0.0-2
-mpfr-4.2.1-5
+mpfr-4.2.1-6
 ncurses-6.5-3
 nettle-3.10.1-1
 npth-1.8-1
@@ -120,7 +120,7 @@ python-urllib3-2.3.0-1
 readline-8.2.013-1
 sed-4.9-3
 shadow-4.16.0-1
-sqlite-3.48.0-2
+sqlite-3.49.0-1
 sudo-1.9.16.p2-2
 systemd-257.2-2
 systemd-libs-257.2-2

[-- Attachment #3: build-logs.zip --]
[-- Type: application/zip, Size: 1356795 bytes --]

Re: 5.10 kernel series fails to build on newer toolchain: FAILED unresolved symbol filp_close

[Greg Kroah-Hartman]

On Fri, Mar 14, 2025 at 08:18:27AM +0700, Philip Müller wrote:
> For some odd reason 5.10 kernel series doesn't compile with a newer
> toolchain since 2025-02-09:
> 
> 2025-02-09T17:32:07.7991299Z   GEN     .version
> 2025-02-09T17:32:07.8270062Z   CHK     include/generated/compile.h
> 2025-02-09T17:32:07.8540777Z   LD      vmlinux.o
> 2025-02-09T17:32:11.7210899Z   MODPOST vmlinux.symvers
> 2025-02-09T17:32:12.0869599Z   MODINFO modules.builtin.modinfo
> 2025-02-09T17:32:12.1403022Z   GEN     modules.builtin
> 2025-02-09T17:32:12.1475659Z   LD      .tmp_vmlinux.btf
> 2025-02-09T17:32:19.6117204Z   BTF     .btf.vmlinux.bin.o
> 2025-02-09T17:32:31.2916650Z   LD      .tmp_vmlinux.kallsyms1
> 2025-02-09T17:32:34.8731104Z   KSYMS   .tmp_vmlinux.kallsyms1.S
> 2025-02-09T17:32:35.4910608Z   AS      .tmp_vmlinux.kallsyms1.o
> 2025-02-09T17:32:35.9662538Z   LD      .tmp_vmlinux.kallsyms2
> 2025-02-09T17:32:39.2595984Z   KSYMS   .tmp_vmlinux.kallsyms2.S
> 2025-02-09T17:32:39.8802028Z   AS      .tmp_vmlinux.kallsyms2.o
> 2025-02-09T17:32:40.3659440Z   LD      vmlinux
> 2025-02-09T17:32:48.0031558Z   BTFIDS  vmlinux
> 2025-02-09T17:32:48.0143553Z FAILED unresolved symbol filp_close
> 2025-02-09T17:32:48.5019928Z make: *** [Makefile:1207: vmlinux] Error 255
> 2025-02-09T17:32:48.5061241Z ==> ERROR: A failure occurred in build().
> 
> 5.10.234 built fine couple of days ago with the older one. There were slight
> changes made. 5.4 and 5.15 still compile.
> 
> Wonder what might be missing here ...

Can you bisect down to the offending commit?

And I think I saw kernelci hit this as well, but I don't have an answer
for it...

greg k-h

Re: 5.10 kernel series fails to build on newer toolchain: FAILED unresolved symbol filp_close

[Philip Müller]

On 14/3/25 12:39, Greg Kroah-Hartman wrote:
> Can you bisect down to the offending commit?
> 
> And I think I saw kernelci hit this as well, but I don't have an answer
> for it...

The same kernel compiles fine with the older toolchain. No changes were 
made to config nor patch-set when we tried to recompile the 5.10.234 
kernel with the newer toolchain. 5.10.235 fails similar to 5.10.234 on 
the same toolchain.

So maybe a commit is missing, which is present in either 5.4 or 5.15 series.

-- 
Best, Philip

Re: 5.10 kernel series fails to build on newer toolchain: FAILED unresolved symbol filp_close

[Laura Nao]

Hello,

On Fri, 14 Mar 2025 16:19:13 +0700, Philip Müller wrote:
> On 14/3/25 12:39, Greg Kroah-Hartman wrote:
>> Can you bisect down to the offending commit?
>> 
>> And I think I saw kernelci hit this as well, but I don't have an answer
>> for it...
>
> The same kernel compiles fine with the older toolchain. No changes were 
> made to config nor patch-set when we tried to recompile the 5.10.234 
> kernel with the newer toolchain. 5.10.235 fails similar to 5.10.234 on 
> the same toolchain.
>
> So maybe a commit is missing, which is present in either 5.4 or 5.15 series.

KernelCI is now reporting a pass on the stable-rc build (5.10.236-rc1),
though I was not able to spot exactly what fixed this.

Best,
Laura

Re: 5.10 kernel series fails to build on newer toolchain: FAILED unresolved symbol filp_close

[Greg KH]

On Thu, Mar 20, 2025 at 12:28:06PM +0100, Laura Nao wrote:
> Hello,
> 
> On Fri, 14 Mar 2025 16:19:13 +0700, Philip Müller wrote:
> > On 14/3/25 12:39, Greg Kroah-Hartman wrote:
> >> Can you bisect down to the offending commit?
> >> 
> >> And I think I saw kernelci hit this as well, but I don't have an answer
> >> for it...
> >
> > The same kernel compiles fine with the older toolchain. No changes were 
> > made to config nor patch-set when we tried to recompile the 5.10.234 
> > kernel with the newer toolchain. 5.10.235 fails similar to 5.10.234 on 
> > the same toolchain.
> >
> > So maybe a commit is missing, which is present in either 5.4 or 5.15 series.
> 
> KernelCI is now reporting a pass on the stable-rc build (5.10.236-rc1),
> though I was not able to spot exactly what fixed this.

That's good to know, thanks!

Re: 5.10 kernel series fails to build on newer toolchain: FAILED unresolved symbol filp_close

[Philip Müller]

Yes, I can confirm that with the current stable-queue patches on top of 
5.10.235 it compiles. I only had to not apply the following patch

ASoC: Intel: sof_sdw: Add support for Fatcat board with BT offload 
enabled in PTL platform

due to this compile error it created:

2025-03-28T08:55:34.9654713Z   CC      net/unix/sysctl_net_unix.o
2025-03-28T08:55:35.0056914Z sound/soc/intel/boards/sof_sdw.c:243:41: 
error: ‘SOC_SDW_PCH_DMIC’ undeclared here (not in a function); did you 
mean ‘SOF_SDW_PCH_DMIC’?
2025-03-28T08:55:35.0058317Z   243 |                 .driver_data = 
(void *)(SOC_SDW_PCH_DMIC |
2025-03-28T08:55:35.0059209Z       | 
     ^~~~~~~~~~~~~~~~
2025-03-28T08:55:35.0060110Z       | 
     SOF_SDW_PCH_DMIC
2025-03-28T08:55:35.0578564Z sound/soc/intel/boards/sof_sdw.c:244:41: 
error: implicit declaration of function ‘SOF_BT_OFFLOAD_SSP’ 
[-Werror=implicit-function-declaration]
2025-03-28T08:55:35.0580183Z   244 | 
     SOF_BT_OFFLOAD_SSP(2) |
2025-03-28T08:55:35.0581143Z       | 
     ^~~~~~~~~~~~~~~~~~
2025-03-28T08:55:35.1499457Z sound/soc/intel/boards/sof_sdw.c:245:41: 
error: ‘SOF_SSP_BT_OFFLOAD_PRESENT’ undeclared here (not in a function)
2025-03-28T08:55:35.1500818Z   245 | 
     SOF_SSP_BT_OFFLOAD_PRESENT),
2025-03-28T08:55:35.1501765Z       | 
     ^~~~~~~~~~~~~~~~~~~~~~~~~~
2025-03-28T08:55:35.2672363Z cc1: some warnings being treated as errors
2025-03-28T08:55:35.2740103Z make[4]: *** [scripts/Makefile.build:286: 
sound/soc/intel/boards/sof_sdw.o] Error 1
2025-03-28T08:55:35.2767794Z make[3]: *** [scripts/Makefile.build:503: 
sound/soc/intel/boards] Error 2
2025-03-28T08:55:35.2773462Z make[2]: *** [scripts/Makefile.build:503: 
sound/soc/intel] Error 2
2025-03-28T08:55:35.2801723Z make[1]: *** [scripts/Makefile.build:503: 
sound/soc] Error 2
2025-03-28T08:55:35.2802890Z make: *** [Makefile:1837: sound] Error 2

On 3/20/25 07:28, Laura Nao wrote:
> Hello,
> 
> On Fri, 14 Mar 2025 16:19:13 +0700, Philip Müller wrote:
>> On 14/3/25 12:39, Greg Kroah-Hartman wrote:
>>> Can you bisect down to the offending commit?
>>>
>>> And I think I saw kernelci hit this as well, but I don't have an answer
>>> for it...
>>
>> The same kernel compiles fine with the older toolchain. No changes were
>> made to config nor patch-set when we tried to recompile the 5.10.234
>> kernel with the newer toolchain. 5.10.235 fails similar to 5.10.234 on
>> the same toolchain.
>>
>> So maybe a commit is missing, which is present in either 5.4 or 5.15 series.
> 
> KernelCI is now reporting a pass on the stable-rc build (5.10.236-rc1),
> though I was not able to spot exactly what fixed this.
> 
> Best,
> Laura


-- 
Best, Philip

Re: 5.10 kernel series fails to build on newer toolchain: FAILED unresolved symbol filp_close

[Greg KH]

On Fri, Mar 28, 2025 at 07:06:10AM -0400, Philip Müller wrote:
> Yes, I can confirm that with the current stable-queue patches on top of
> 5.10.235 it compiles. I only had to not apply the following patch
> 
> ASoC: Intel: sof_sdw: Add support for Fatcat board with BT offload enabled
> in PTL platform

I've dropped this commit from the queue now, thanks.

greg k-h

Re: 5.10 kernel series fails to build on newer toolchain: FAILED unresolved symbol filp_close

[Philip Müller]

On 4/1/25 11:17, Greg KH wrote:
> On Fri, Mar 28, 2025 at 07:06:10AM -0400, Philip Müller wrote:
>> Yes, I can confirm that with the current stable-queue patches on top of
>> 5.10.235 it compiles. I only had to not apply the following patch
>>
>> ASoC: Intel: sof_sdw: Add support for Fatcat board with BT offload enabled
>> in PTL platform
> 
> I've dropped this commit from the queue now, thanks.
> 
> greg k-h

Somehow the issue came back with 5.10.238 ...

-- 
Best, Philip

Re: 5.10 kernel series fails to build on newer toolchain: FAILED unresolved symbol filp_close

[Greg KH]

On Thu, Jun 05, 2025 at 09:44:49AM +0200, Philip Müller wrote:
> On 4/1/25 11:17, Greg KH wrote:
> > On Fri, Mar 28, 2025 at 07:06:10AM -0400, Philip Müller wrote:
> > > Yes, I can confirm that with the current stable-queue patches on top of
> > > 5.10.235 it compiles. I only had to not apply the following patch
> > > 
> > > ASoC: Intel: sof_sdw: Add support for Fatcat board with BT offload enabled
> > > in PTL platform
> > 
> > I've dropped this commit from the queue now, thanks.
> > 
> > greg k-h
> 
> Somehow the issue came back with 5.10.238 ...

I have no context here, sorry...

Re: 5.10 kernel series fails to build on newer toolchain: FAILED unresolved symbol filp_close

[Philip Müller]

[-- Attachment #1: Type: text/plain, Size: 817 bytes --]

On 6/5/25 10:46, Greg KH wrote:
> I have no context here, sorry...

So basically, starting with GCC 15.1 the kernel series doesn't compile 
again and errors out with: FAILED unresolved symbol filp_close. I tested 
now v5.10.237 as well, which failed similar to v5.10.238.

There are some Debian reports out there:

https://linux.debian.bugs.dist.narkive.com/2JKeaFga/bug-1104662-failed-unresolved-symbol-filp-close-linux-kernel-5-10-237
https://www.mail-archive.com/debian-kernel@lists.debian.org/msg142397.html

And I also found this one:

https://lists-ec2.96boards.org/archives/list/linux-stable-mirror@lists.linaro.org/thread/7XFQI52N3KGUGFLPWCSJZW6DDFZCOXP4/

For GCC 14.1 I had to add the gnu 11 patch, which was discussed already. 
Also 5.4 and 5.15 still compile with the newer toolchain ...

-- 
Best, Philip

[-- Attachment #2: linux-5.10.236-std-gnu11.patch --]
[-- Type: text/x-patch, Size: 1183 bytes --]

Commit b3bee1e7c3f2b1b77182302c7b2131c804175870 x86/boot: Compile boot code with -std=gnu11 too
fixed a build failure when compiling with GCC 15. The same change is required for linux-5.10.236.

Signed-off-by: Chris Clayton <chris2553@googlemail.com>
Modified-by: Philip Mueller <philm@manjaro.org>
Link: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=b3bee1e7c3f2b1b77182302c7b2131c804175870


diff -rup linux-5.10.236.orig/arch/x86/Makefile linux-5.10.236/arch/x86/Makefile
--- linux-5.10.236.orig/arch/x86/Makefile	2025-04-10 13:37:44.000000000 +0100
+++ linux-5.10.236/arch/x86/Makefile	2025-04-26 19:37:38.294386968 +0100
@@ -31,7 +31,7 @@ endif
 CODE16GCC_CFLAGS := -m32 -Wa,$(srctree)/arch/x86/boot/code16gcc.h
 M16_CFLAGS	 := $(call cc-option, -m16, $(CODE16GCC_CFLAGS))
 
-REALMODE_CFLAGS	:= $(M16_CFLAGS) -g -Os -DDISABLE_BRANCH_PROFILING -D__DISABLE_EXPORTS \
+REALMODE_CFLAGS	:= -std=gnu11 $(M16_CFLAGS) -g -Os -DDISABLE_BRANCH_PROFILING -D__DISABLE_EXPORTS \
 		   -Wall -Wstrict-prototypes -march=i386 -mregparm=3 \
 		   -fno-strict-aliasing -fomit-frame-pointer -fno-pic \
 		   -mno-mmx -mno-sse $(call cc-option,-fcf-protection=none)

Re: 5.10 kernel series fails to build on newer toolchain: FAILED unresolved symbol filp_close

[Greg KH]

On Sun, Jun 08, 2025 at 08:27:35AM +0200, Philip Müller wrote:
> On 6/5/25 10:46, Greg KH wrote:
> > I have no context here, sorry...
> 
> So basically, starting with GCC 15.1 the kernel series doesn't compile again
> and errors out with: FAILED unresolved symbol filp_close. I tested now
> v5.10.237 as well, which failed similar to v5.10.238.
> 
> There are some Debian reports out there:
> 
> https://linux.debian.bugs.dist.narkive.com/2JKeaFga/bug-1104662-failed-unresolved-symbol-filp-close-linux-kernel-5-10-237
> https://www.mail-archive.com/debian-kernel@lists.debian.org/msg142397.html
> 
> And I also found this one:
> 
> https://lists-ec2.96boards.org/archives/list/linux-stable-mirror@lists.linaro.org/thread/7XFQI52N3KGUGFLPWCSJZW6DDFZCOXP4/
> 
> For GCC 14.1 I had to add the gnu 11 patch, which was discussed already.
> Also 5.4 and 5.15 still compile with the newer toolchain ...
> 
> -- 
> Best, Philip

> Commit b3bee1e7c3f2b1b77182302c7b2131c804175870 x86/boot: Compile boot code with -std=gnu11 too
> fixed a build failure when compiling with GCC 15. The same change is required for linux-5.10.236.
> 
> Signed-off-by: Chris Clayton <chris2553@googlemail.com>
> Modified-by: Philip Mueller <philm@manjaro.org>
> Link: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=b3bee1e7c3f2b1b77182302c7b2131c804175870
> 
> 
> diff -rup linux-5.10.236.orig/arch/x86/Makefile linux-5.10.236/arch/x86/Makefile
> --- linux-5.10.236.orig/arch/x86/Makefile	2025-04-10 13:37:44.000000000 +0100
> +++ linux-5.10.236/arch/x86/Makefile	2025-04-26 19:37:38.294386968 +0100
> @@ -31,7 +31,7 @@ endif
>  CODE16GCC_CFLAGS := -m32 -Wa,$(srctree)/arch/x86/boot/code16gcc.h
>  M16_CFLAGS	 := $(call cc-option, -m16, $(CODE16GCC_CFLAGS))
>  
> -REALMODE_CFLAGS	:= $(M16_CFLAGS) -g -Os -DDISABLE_BRANCH_PROFILING -D__DISABLE_EXPORTS \
> +REALMODE_CFLAGS	:= -std=gnu11 $(M16_CFLAGS) -g -Os -DDISABLE_BRANCH_PROFILING -D__DISABLE_EXPORTS \
>  		   -Wall -Wstrict-prototypes -march=i386 -mregparm=3 \
>  		   -fno-strict-aliasing -fomit-frame-pointer -fno-pic \
>  		   -mno-mmx -mno-sse $(call cc-option,-fcf-protection=none)

Can you resend this in a format we can apply it in?

Also for the newer kernels, this was only backported to 6.6.y, so
anything older than that should need this, right?

thanks,

greg k-h

Re: 5.10 kernel series fails to build on newer toolchain: FAILED unresolved symbol filp_close

[Philip Müller]

[-- Attachment #1: Type: text/plain, Size: 228 bytes --]

On 6/17/25 16:05, Greg KH wrote:
> Also for the newer kernels, this was only backported to 6.6.y, so
> anything older than that should need this, right?

Well, yes. The patches I applied on my end are attached.

-- 
Best, Philip

[-- Attachment #2: linux-5.4.292-std-gnu11.patch --]
[-- Type: text/x-patch, Size: 2219 bytes --]

Commit b3bee1e7c3f2b1b77182302c7b2131c804175870 x86/boot: Compile boot code with -std=gnu11 too
fixed a build failure when compiling with GCC 15. The same change is required for linux-5.4.292.

Signed-off-by: Chris Clayton <chris2553@googlemail.com>
Modified-by: Philip Mueller <philm@manjaro.org>
Link: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=b3bee1e7c3f2b1b77182302c7b2131c804175870


diff -rup linux-5.4.292.orig/arch/x86/Makefile linux-5.4.292/arch/x86/Makefile
--- linux-5.4.292.orig/arch/x86/Makefile	2025-04-10 13:37:44.000000000 +0100
+++ linux-5.4.292/arch/x86/Makefile	2025-04-26 19:37:38.294386968 +0100
@@ -31,7 +31,7 @@ endif
 CODE16GCC_CFLAGS := -m32 -Wa,$(srctree)/arch/x86/boot/code16gcc.h
 M16_CFLAGS	 := $(call cc-option, -m16, $(CODE16GCC_CFLAGS))
 
-REALMODE_CFLAGS	:= $(M16_CFLAGS) -g -Os -DDISABLE_BRANCH_PROFILING \
+REALMODE_CFLAGS	:= -std=gnu11 $(M16_CFLAGS) -g -Os -DDISABLE_BRANCH_PROFILING -D__DISABLE_EXPORTS \
 		   -Wall -Wstrict-prototypes -march=i386 -mregparm=3 \
 		   -fno-strict-aliasing -fomit-frame-pointer -fno-pic \
 		   -mno-mmx -mno-sse $(call cc-option,-fcf-protection=none)
--- linux-5.4.292.orig/drivers/firmware/efi/libstub/Makefile	2025-05-01 09:31:09.755059128 +0200
+++ linux-5.4.292/drivers/firmware/efi/libstub/Makefile	2025-05-01 16:48:22.029249917 +0200
@@ -7,7 +7,7 @@
 #
 cflags-$(CONFIG_X86_32)		:= -march=i386
 cflags-$(CONFIG_X86_64)		:= -mcmodel=small
-cflags-$(CONFIG_X86)		+= -m$(BITS) -D__KERNEL__ -O2 \
+cflags-$(CONFIG_X86)		+= -std=gnu11 -m$(BITS) -D__KERNEL__ -O2 \
 				   -fPIC -fno-strict-aliasing -mno-red-zone \
 				   -mno-mmx -mno-sse -fshort-wchar \
 				   -Wno-pointer-sign \
--- linux-5.4.292.orig/arch/x86/boot/compressed/Makefile	2025-05-01 09:31:09.755059128 +0200
+++ linux-5.4.292/arch/x86/boot/compressed/Makefile	2025-05-01 16:48:22.029249917 +0200
@@ -32,6 +32,7 @@ KBUILD_CFLAGS += -DDISABLE_BRANCH_PROFIL
 cflags-$(CONFIG_X86_32) := -march=i386
 cflags-$(CONFIG_X86_64) := -mcmodel=small
 KBUILD_CFLAGS += $(cflags-y)
+KBUILD_CFLAGS += -std=gnu11
 KBUILD_CFLAGS += -mno-mmx -mno-sse
 KBUILD_CFLAGS += $(call cc-option,-ffreestanding)
 KBUILD_CFLAGS += $(call cc-option,-fno-stack-protector)

[-- Attachment #3: linux-5.10.236-std-gnu11.patch --]
[-- Type: text/x-patch, Size: 1183 bytes --]

Commit b3bee1e7c3f2b1b77182302c7b2131c804175870 x86/boot: Compile boot code with -std=gnu11 too
fixed a build failure when compiling with GCC 15. The same change is required for linux-5.10.236.

Signed-off-by: Chris Clayton <chris2553@googlemail.com>
Modified-by: Philip Mueller <philm@manjaro.org>
Link: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=b3bee1e7c3f2b1b77182302c7b2131c804175870


diff -rup linux-5.10.236.orig/arch/x86/Makefile linux-5.10.236/arch/x86/Makefile
--- linux-5.10.236.orig/arch/x86/Makefile	2025-04-10 13:37:44.000000000 +0100
+++ linux-5.10.236/arch/x86/Makefile	2025-04-26 19:37:38.294386968 +0100
@@ -31,7 +31,7 @@ endif
 CODE16GCC_CFLAGS := -m32 -Wa,$(srctree)/arch/x86/boot/code16gcc.h
 M16_CFLAGS	 := $(call cc-option, -m16, $(CODE16GCC_CFLAGS))
 
-REALMODE_CFLAGS	:= $(M16_CFLAGS) -g -Os -DDISABLE_BRANCH_PROFILING -D__DISABLE_EXPORTS \
+REALMODE_CFLAGS	:= -std=gnu11 $(M16_CFLAGS) -g -Os -DDISABLE_BRANCH_PROFILING -D__DISABLE_EXPORTS \
 		   -Wall -Wstrict-prototypes -march=i386 -mregparm=3 \
 		   -fno-strict-aliasing -fomit-frame-pointer -fno-pic \
 		   -mno-mmx -mno-sse $(call cc-option,-fcf-protection=none)

[-- Attachment #4: linux-5.15.180-std-gnu11.patch --]
[-- Type: text/x-patch, Size: 1133 bytes --]

Commit b3bee1e7c3f2b1b77182302c7b2131c804175870 x86/boot: Compile boot code with -std=gnu11 too
fixed a build failure when compiling with GCC 15. The same change is required for linux-5.15.180.

Signed-off-by: Chris Clayton <chris2553@googlemail.com>
Link: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=b3bee1e7c3f2b1b77182302c7b2131c804175870


diff -rup linux-5.15.180.orig/arch/x86/Makefile linux-5.15.180/arch/x86/Makefile
--- linux-5.15.180.orig/arch/x86/Makefile	2025-04-10 13:37:44.000000000 +0100
+++ linux-5.15.180/arch/x86/Makefile	2025-04-26 19:37:38.294386968 +0100
@@ -43,7 +43,7 @@ endif
 
 # How to compile the 16-bit code.  Note we always compile for -march=i386;
 # that way we can complain to the user if the CPU is insufficient.
-REALMODE_CFLAGS	:= -m16 -g -Os -DDISABLE_BRANCH_PROFILING -D__DISABLE_EXPORTS \
+REALMODE_CFLAGS	:= -std=gnu11 -m16 -g -Os -DDISABLE_BRANCH_PROFILING -D__DISABLE_EXPORTS \
 		   -Wall -Wstrict-prototypes -march=i386 -mregparm=3 \
 		   -fno-strict-aliasing -fomit-frame-pointer -fno-pic \
 		   -mno-mmx -mno-sse $(call cc-option,-fcf-protection=none)

Re: 5.10 kernel series fails to build on newer toolchain: FAILED unresolved symbol filp_close

[Greg KH]

On Thu, Jun 19, 2025 at 10:34:52AM +0200, Philip Müller wrote:
> On 6/17/25 16:05, Greg KH wrote:
> > Also for the newer kernels, this was only backported to 6.6.y, so
> > anything older than that should need this, right?
> 
> Well, yes. The patches I applied on my end are attached.
> 
> -- 
> Best, Philip

> Commit b3bee1e7c3f2b1b77182302c7b2131c804175870 x86/boot: Compile boot code with -std=gnu11 too
> fixed a build failure when compiling with GCC 15. The same change is required for linux-5.4.292.

Something went wrong with the whitespace here :(

> 
> Signed-off-by: Chris Clayton <chris2553@googlemail.com>
> Modified-by: Philip Mueller <philm@manjaro.org>

You lost the original authorship and signed-off-by lines :(

> Link: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=b3bee1e7c3f2b1b77182302c7b2131c804175870

This isn't needed either.

Can you try again?

thanks,

greg k-h

Re: 5.10 kernel series fails to build on newer toolchain: FAILED unresolved symbol filp_close

[Philip Müller]

On 6/5/25 10:46, Greg KH wrote:
> I have no context here, sorry...

Seems with 5.10.239-rc1 it compiles again just fine ...

-- 
Best, Philip

Re: 5.10 kernel series fails to build on newer toolchain: FAILED unresolved symbol filp_close

[Alyssa Ross]

[-- Attachment #1: Type: text/plain, Size: 737 bytes --]

On Tue, Jun 24, 2025 at 09:05:31AM +0200, Philip Müller wrote:
> On 6/5/25 10:46, Greg KH wrote:
> > I have no context here, sorry...
>
> Seems with 5.10.239-rc1 it compiles again just fine ...

We've been seeing this issue[1] since 5.10.244 (specifically commit
b039655d31a1 ("genirq: Provide new interfaces for affinity hints")),
and still in 5.10.245.

Given that this has apparently come up before, and I don't see any
likely cause looking at that diff, I suppose it's probably some build
issue in 5.10 that can be triggered by innocent diffs, and so is liable
to keep fixing and breaking itself until somebody figure out the root
cause…

[1]: https://github.com/NixOS/nixpkgs/pull/448034#issuecomment-3364278085

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]

Re: 5.10 kernel series fails to build on newer toolchain: FAILED unresolved symbol filp_close

[David Nyström]

[-- Attachment #1: Type: text/plain, Size: 1793 bytes --]


On Sun, 26 Oct 2025, Alyssa Ross wrote:

> On Tue, Jun 24, 2025 at 09:05:31AM +0200, Philip Müller wrote:
>> On 6/5/25 10:46, Greg KH wrote:
>>> I have no context here, sorry...
>>
>> Seems with 5.10.239-rc1 it compiles again just fine ...
>
> We've been seeing this issue[1] since 5.10.244 (specifically commit
> b039655d31a1 ("genirq: Provide new interfaces for affinity hints")),
> and still in 5.10.245.
>
> Given that this has apparently come up before, and I don't see any
> likely cause looking at that diff, I suppose it's probably some build
> issue in 5.10 that can be triggered by innocent diffs, and so is liable
> to keep fixing and breaking itself until somebody figure out the root
> cause…
>
> [1]: https://github.com/NixOS/nixpkgs/pull/448034#issuecomment-3364278085
>

I have analyzed the issue, and it seems like never versions of GCC will 
agressively inline both in compile and linktime.
I'm not sure if this is fixed properly in master, I would assume its fixed 
by adding enough code into BTF exposed functions so that they not
optimized away anymore.

I can "fix" this locally via:
CONFIG_DEBUG_INFO_BTF=n

or

diff --git a/fs/Makefile b/fs/Makefile
index c660ce28f149..6fbaccbae1d9 100644
--- a/fs/Makefile
+++ b/fs/Makefile
@@ -6,6 +6,10 @@
  # Rewritten to use lists instead of if-statements.
  #

+# Prevent agressive link-time inlining of BTF symbols
+# Prevent GCC 15 IPA from removing filp_close symbol needed for BTF
+CFLAGS_open.o = -fno-inline -fno-ipa-sra
+

or by removing line below.

linux-stable$ git grep filp_close kernel/trace
kernel/trace/bpf_trace.c:BTF_ID(func, filp_close)

Last seems like the simplest option, but would cause regressions for BPF
programs I presume.
GCC 15+ does not seem to adhere to __noinline, at least not for this 
issue.