Re: [PATCH v3] proc/kcore: allow disabling /proc/kcore at boot
From: Christian Brauner
Date: Wed Nov 18 2026 - 08:11:24 EST
On Tue, Nov 17, 2026 at 02:20:38AM -0800, Chris Down wrote:
> /proc/kcore lets any process with CAP_SYS_RAWIO read all of kernel
> memory. That's a pretty large read primitive for anyone who obtains
> CAP_SYS_RAWIO (and it also gives away the KASLR layout).
>
> For this reason on some hosts we want to turn /proc/kcore off. In terms
> of what it can read, /proc/kcore isn't any more concerning than other
> ways to read kernel memory, like BPF with CAP_BPF and CAP_PERFMON. The
> distinction is that a process with CAP_SYS_RAWIO doesn't necessarily
> have CAP_BPF and CAP_PERFMON, because CAP_SYS_RAWIO may be needed for
> hardware access without necessarily wanting to grant general RAM access.
> CONFIG_STRICT_DEVMEM already makes that possible for /dev/mem, but
> /proc/kcore gives the same capability to do general purpose RAM reading
> anyway.
>
> However, on a kernel built with CONFIG_PROC_KCORE=y the only ways to
> block it today are lockdown=confidentiality (which also disables many
> other things we perhaps don't want to disable, like kprobes and
> bpf_probe_read_kernel()), or an LSM. That's annoying, because on the
> majority of hosts we do want it on for things like drgn/crash, so today
> we build two separate kernels.
>
> Let's add a "kcore=" boot parameter so that this can be disabled at
> boot, which lets us preserve those choices while using the same kernel
> image.
>
> Signed-off-by: Chris Down <chris@xxxxxxxxxxxxxx>
> ---
> v2 at https://lore.kernel.org/r/a8474f0e2d0ef5ea9d53aa6649c7e9c9ea6f15c2.1794800574.git.chris@xxxxxxxxxxxxxx
>
> v3: Make kcore_enabled __ro_after_init (Kees)
> v2: Give more justification in changelog (Omar)
>
> Documentation/admin-guide/kernel-parameters.txt | 6 ++++++
> fs/proc/kcore.c | 12 ++++++++++++
> 2 files changed, 18 insertions(+)
>
> diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
> index 176648941aec..205228294fc2 100644
> --- a/Documentation/admin-guide/kernel-parameters.txt
> +++ b/Documentation/admin-guide/kernel-parameters.txt
> @@ -2902,6 +2902,12 @@ Kernel parameters
> parameter KASAN will print report only for the first
> invalid access.
>
> + kcore= [KNL,EARLY]
> + Enable or disable /proc/kcore on kernels built with
> + CONFIG_PROC_KCORE=y.
> + Format: <bool>
> + Default: on
So I'm happy to take this on top of vfs-7.5.procfs but I'm really
unhappy about this proliferation of various hyper-specific opt-outs.
For example, we already have:
enum proc_mem_force {
PROC_MEM_FORCE_ALWAYS,
PROC_MEM_FORCE_PTRACE,
PROC_MEM_FORCE_NEVER
};
static enum proc_mem_force proc_mem_force_override __ro_after_init =
IS_ENABLED(CONFIG_PROC_MEM_NO_FORCE) ? PROC_MEM_FORCE_NEVER :
IS_ENABLED(CONFIG_PROC_MEM_FORCE_PTRACE) ? PROC_MEM_FORCE_PTRACE :
PROC_MEM_FORCE_ALWAYS;
to block FOLL_FORCE access on kernels which is another huge proc-based
attack surface. My thesis is that if one is worried about /proc/kcore
readout attacks they should be equally worried about these types of
attacks.
Iow, it would be nicer if someone took the time and came up with some
boot setting that is a list of memory protection settings instead of
spraying ever more kernel command line switches that nobody can
reasonably remember or use.