linux-kernel - Re: Linux 6.15-rc1

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: <Z/Qg0jQ/xCn67jOL@lstrano-desk.jf.intel.com>
Date: Mon, 7 Apr 2025 12:00:34 -0700
From: Matthew Brost <matthew.brost@...el.com>
To: Linus Torvalds <torvalds@...ux-foundation.org>
CC: Guenter Roeck <linux@...ck-us.net>, Miri Korenblit
	<miriam.rachel.korenblit@...el.com>, Johannes Berg <johannes.berg@...el.com>,
	Oak Zeng <oak.zeng@...el.com>, Linux Kernel Mailing List
	<linux-kernel@...r.kernel.org>, the arch/x86 maintainers <x86@...nel.org>
Subject: Re: Linux 6.15-rc1

On Mon, Apr 07, 2025 at 08:55:01AM -0700, Linus Torvalds wrote:
> On Mon, 7 Apr 2025 at 08:14, Guenter Roeck <linux@...ck-us.net> wrote:
> >
> > Build failures:
> >
> > In file included from drivers/net/wireless/intel/iwlwifi/mld/tests/utils.c:18:
> > drivers/net/wireless/intel/iwlwifi/mld/tests/../iface.h:169:5: warning: "CONFIG_PM_SLEEP" is not defined, evaluates to 0 [-Wundef]
> >   169 | #if CONFIG_PM_SLEEP
> > [...]
> > Triggered by commit d1e879ec600f9 ("wifi: iwlwifi: add iwlmld sub-driver").
> > The code was not build tested with CONFIG_PM disabled.
> 
> Ok, should be trivial to fix, let's get the guilty parties involved..
> 
> > Building x86_64:allyesconfig ... failed
> > --------------
> > Error log:
> > x86_64-linux-ld: kernel image bigger than KERNEL_IMAGE_SIZE
> >
> > No idea what to do here. Increase KERNEL_IMAGE_SIZE ? Disable some options ?
> 
> Hmm. It's a fairly random #define, and I suspect we could just
> increase it for COMPILE_TEST.
> 
> allyesconfig is kind of useless, but it does occasionally find the
> "same name used in different modules" case and other random things,
> and disabling options seems overly complicated for this odd case.
> 
> Adding x86 people to see if there are preferences.
> 
> > Building xtensa:allmodconfig ... failed
> >
> > ERROR: modpost: "__divdi3" [drivers/gpu/drm/xe/xe.ko] undefined!
> > ERROR: modpost: "__udivdi3" [drivers/gpu/drm/xe/xe.ko] undefined!
> >
> > Commit 9c44fd5f6e8a ("drm/xe: Add migrate layer functions for SVM support")
> > didn't watch out for 64-bit divide operations.
> 
> Hmm. Why didn't that show up on x86-32? Or maybe it did and you're
> just quoting the xtensa case.
> 

We have x86-32 in our CI and this was not caught.

> Adding Oak and Matthew.
> 
> I assume it's the DIV_ROUND_UP(entries, 0x1ff) thing.
> 

This is the problem. It was fixed here [1].

[1] https://patchwork.freedesktop.org/series/146689/

The patch is showing up in drm-tip too but must have not been included
in the 6.15 PR:

c9092257506a drm/xe: avoid plain 64-bit division

> Considering that that pte_update_cmd_size() returns an explicitly
> 32-bit number with
> 
>         num_dword += entries * 2;
>         return num_dword;
> 

Yea this a little odd. Can clean up in follow up patch.

Matt

> it's kind of silly for 'entries' to be calculated in 64 bits. It does
> look like it should just use div_u64().
> 
> > Building xtensa:de212:kc705-nommu:nommu_kc705_defconfig ... failed
> > ------------
> > Error log:
> > kernel/trace/trace.o:(.init.literal+0x250): undefined reference to `get_vm_area'
> 
> I see that Steven already saw this one.
> 
>              Linus