[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <a9862b5b-61e5-4455-9b6f-d05050f11b64@efficios.com>
Date: Mon, 17 Feb 2025 15:23:13 -0500
From: Mathieu Desnoyers <mathieu.desnoyers@...icios.com>
To: Dmitry Vyukov <dvyukov@...gle.com>, peterz@...radead.org,
boqun.feng@...il.com, tglx@...utronix.de, mingo@...hat.com, bp@...en8.de,
dave.hansen@...ux.intel.com, hpa@...or.com, aruna.ramakrishna@...cle.com,
elver@...gle.com
Cc: "Paul E. McKenney" <paulmck@...nel.org>, x86@...nel.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH 4/4] selftests/rseq: Add test for rseq+pkeys
On 2025-02-17 06:07, Dmitry Vyukov wrote:
> Add a test that ensures that PKEY-protected struct rseq works
> and does not lead to process kills.
Rather than depend on the glibc rseq tunable to disable glibc
rseq support, you could simply setup a rseq_cs in a separate
page and get the kernel to hit _that_ page through the
rseq_cs pointer.
Thanks,
Mathieu
>
> Signed-off-by: Dmitry Vyukov <dvyukov@...gle.com>
> Cc: Mathieu Desnoyers <mathieu.desnoyers@...icios.com>
> Cc: Peter Zijlstra <peterz@...radead.org>
> Cc: "Paul E. McKenney" <paulmck@...nel.org>
> Cc: Boqun Feng <boqun.feng@...il.com>
> Cc: Thomas Gleixner <tglx@...utronix.de>
> Cc: Ingo Molnar <mingo@...hat.com>
> Cc: Borislav Petkov <bp@...en8.de>
> Cc: Dave Hansen <dave.hansen@...ux.intel.com>
> Cc: "H. Peter Anvin" <hpa@...or.com>
> Cc: Aruna Ramakrishna <aruna.ramakrishna@...cle.com>
> Cc: x86@...nel.org
> Cc: linux-kernel@...r.kernel.org
> ---
> tools/testing/selftests/rseq/Makefile | 2 +-
> tools/testing/selftests/rseq/pkey_test.c | 61 ++++++++++++++++++++++++
> 2 files changed, 62 insertions(+), 1 deletion(-)
>
> diff --git a/tools/testing/selftests/rseq/Makefile b/tools/testing/selftests/rseq/Makefile
> index 5a3432fceb586..9111d25fea3af 100644
> --- a/tools/testing/selftests/rseq/Makefile
> +++ b/tools/testing/selftests/rseq/Makefile
> @@ -16,7 +16,7 @@ OVERRIDE_TARGETS = 1
>
> TEST_GEN_PROGS = basic_test basic_percpu_ops_test basic_percpu_ops_mm_cid_test param_test \
> param_test_benchmark param_test_compare_twice param_test_mm_cid \
> - param_test_mm_cid_benchmark param_test_mm_cid_compare_twice
> + param_test_mm_cid_benchmark param_test_mm_cid_compare_twice pkey_test
>
> TEST_GEN_PROGS_EXTENDED = librseq.so
>
> diff --git a/tools/testing/selftests/rseq/pkey_test.c b/tools/testing/selftests/rseq/pkey_test.c
> new file mode 100644
> index 0000000000000..ba5c1f6e99ab5
> --- /dev/null
> +++ b/tools/testing/selftests/rseq/pkey_test.c
> @@ -0,0 +1,61 @@
> +// SPDX-License-Identifier: LGPL-2.1
> +/*
> + * Ensure that rseq works when rseq data is protected with PKEYs.
> + */
> +
> +#define _GNU_SOURCE
> +#include <err.h>
> +#include <errno.h>
> +#include <stdio.h>
> +#include <stdlib.h>
> +#include <string.h>
> +#include <sys/mman.h>
> +#include <sys/syscall.h>
> +#include <unistd.h>
> +
> +int main(int argc, char **argv)
> +{
> + void *rseq;
> + unsigned long page_size;
> + int pkey, i;
> +
> + pkey = pkey_alloc(0, 0);
> + if (pkey == -1) {
> + printf("[SKIP]\tKernel does not support PKEYs: %s\n",
> + strerror(errno));
> + return 0;
> + }
> +
> + /*
> + * Prevent glibc from registering own struct rseq.
> + * We need to know the rseq address to protect it, but also we need
> + * it to be placed on own page that does not contain other data
> + * (e.g. errno).
> + */
> + if (!getenv("RSEQ_TEST_REEXECED")) {
> + setenv("RSEQ_TEST_REEXECED", "1", 1);
> + setenv("GLIBC_TUNABLES", "glibc.pthread.rseq=0", 1);
> + if (execvpe(argv[0], argv, environ))
> + err(1, "execvpe failed");
> + }
> +
> + page_size = getpagesize();
> + rseq = mmap(NULL, page_size, PROT_READ | PROT_WRITE,
> + MAP_ANON | MAP_PRIVATE, -1, 0);
> + if (rseq == MAP_FAILED)
> + err(1, "mmap failed");
> + if (pkey_mprotect(rseq, page_size, PROT_READ | PROT_WRITE, pkey))
> + err(1, "pkey_mprotect failed");
> + if (syscall(__NR_rseq, rseq, 32, 0, 0))
> + err(1, "rseq failed");
> + if (pkey_set(pkey, PKEY_DISABLE_ACCESS))
> + err(1, "pkey_set failed");
> +
> + /*
> + * If the kernel misbehaves, context switches in the following loop
> + * will kill the process with SIGSEGV.
> + */
> + for (i = 0; i < 10; i++)
> + usleep(100);
> + return 0;
> +}
--
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com
Powered by blists - more mailing lists