linux-kernel - Re: [PATCH] x86/elf: Fix core dump truncation on CPUs with no extended xfeatures

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

Message-ID: <aTfmLKlUjQN4e1Zw@gmail.com>
Date: Tue, 9 Dec 2025 10:04:44 +0100
From: Ingo Molnar <mingo@...nel.org>
To: yongxin.liu@...driver.com
Cc: x86@...nel.org, linux-kernel@...r.kernel.org, bp@...en8.de,
	tglx@...utronix.de, mingo@...hat.com, dave.hansen@...ux.intel.com,
	vigbalas@....com, stable@...r.kernel.org
Subject: Re: [PATCH] x86/elf: Fix core dump truncation on CPUs with no
 extended xfeatures

* yongxin.liu@...driver.com <yongxin.liu@...driver.com> wrote:

> From: Yongxin Liu <yongxin.liu@...driver.com>
>
> Zero can be a valid value of num_records. For example, on Intel Atom x6425RE,
> only x87 and SSE are supported (features 0, 1), and fpu_user_cfg.max_features
> is 3. The for_each_extended_xfeature() loop only iterates feature 2, which is
> not enabled, so num_records = 0. This is valid and should not cause core dump
> failure.
>
> The size check already validates consistency: if num_records = 0, then
> en.n_descsz = 0, so the check passes.
>
> Cc: stable@...r.kernel.org
> Fixes: ba386777a30b ("x86/elf: Add a new FPU buffer layout info to x86 core files")
> Signed-off-by: Yongxin Liu <yongxin.liu@...driver.com>
> ---
>  arch/x86/kernel/fpu/xstate.c | 2 --
>  1 file changed, 2 deletions(-)
>
> diff --git a/arch/x86/kernel/fpu/xstate.c b/arch/x86/kernel/fpu/xstate.c
> index 48113c5193aa..b1dd30eb21a8 100644
> --- a/arch/x86/kernel/fpu/xstate.c
> +++ b/arch/x86/kernel/fpu/xstate.c
> @@ -1984,8 +1984,6 @@ int elf_coredump_extra_notes_write(struct coredump_params *cprm)
>		return 1;
>
>	num_records = dump_xsave_layout_desc(cprm);
> -	if (!num_records)
> -		return 1;

The problem with your patch is that '0' is also used for other errors,
it's the all-around error flag for core dump helper functions such as
dump_emit():

                if (!dump_emit(cprm, &xc, sizeof(xc)))
                        return 0;

So please change dump_xsave_layout_desc() to use negatives
as genuine errors and otherwise returns num_records, and
change elf_coredump_extra_notes_write() to only abort
on genuine errors.

Thanks,

	Ingo