| 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
| ||
|
Date: Wed, 18 Jul 2018 23:28:13 +0800
From: Coly Li <colyli@...e.de>
To: Noah Massey <noah.massey@...il.com>
Cc: linux-kernel@...r.kernel.org, linux-bcache@...r.kernel.org,
linux-block@...r.kernel.org, gregkh@...uxfoundation.org,
torvalds@...ux-foundation.org, tglx@...utronix.de,
kstewart@...uxfoundation.org
Subject: Re: [PATCH v3 3/3] lib/test_crc: Add test cases for crc calculation
On 2018/7/18 2:51 AM, Noah Massey wrote:
> On Tue, Jul 17, 2018 at 10:56 AM Coly Li <colyli@...e.de> wrote:
>>
>> This patch adds a kernel module to test the consistency of multiple crc
>> calculation in Linux kernel. It is enabled with CONFIG_TEST_CRC enabled.
>>
>> The test results are printed into kernel message, which look like,
>>
>> test_crc: crc64: PASSED (0x4e6b1ff972fa8c55, expected 0x4e6b1ff972fa8c55)
>> test_crc: crc64_bch: PASSED (0x0e4f1391d7a4a62e, expected 0x0e4f1391d7a4a62e)
>> test_crc: crc64_update: FAILED (0x03d4d0d85685d9a1, expected 0x3d4d0d85685d9a1f)
>>
>> kernel 0day system has framework to check kernel message, then the above
>> result can be handled by 0day system. If crc calculation inconsistency
>> happens, it can be detected quite soon.
>>
>> lib/test_crc.c is a testing frame work for many crc consistency
>> testings. For now, there are only test caes for 3 crc routines,
>> - crc64()
>> - crc64_bch()
>> - crc64_update()
>>
>> Changelog:
>> v3: Add test cases passed/failed statistic
>> More fixes for review comments of v2
>> v2: Fixes for review comments of v1
>> v1: Initial version.
>>
>> Signed-off-by: Coly Li <colyli@...e.de>
>> Reviewed-by: Hannes Reinecke <hare@...e.de>
>> Cc: Greg Kroah-Hartman <gregkh@...uxfoundation.org>
>> Cc: Linus Torvalds <torvalds@...ux-foundation.org>
>> Cc: Thomas Gleixner <tglx@...utronix.de>
>> Cc: Kate Stewart <kstewart@...uxfoundation.org>
>> ---
>> lib/Kconfig.debug | 10 ++++
>> lib/Makefile | 1 +
>> lib/test_crc.c | 138 ++++++++++++++++++++++++++++++++++++++++++++++
>> 3 files changed, 149 insertions(+)
>> create mode 100644 lib/test_crc.c
>>
>> diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
>> index 8838d1158d19..a9c1de0c2a7d 100644
>> --- a/lib/Kconfig.debug
>> +++ b/lib/Kconfig.debug
>> @@ -1911,6 +1911,16 @@ config TEST_SYSCTL
>>
>> If unsure, say N.
>>
>> +config TEST_CRC
>> + tristate "CRC calculation test driver"
>> + depends on CRC64
>> + help
>> + This builds the "test_crc" module. This driver enables to test the
>> + CRC calculation consistency to make sure new modification does not
>> + break existing checksum calculation.
>> +
>> + if unsure, say N.
>> +
>> config TEST_UDELAY
>> tristate "udelay test driver"
>> default n
>> diff --git a/lib/Makefile b/lib/Makefile
>> index 40c215181687..224d047d026a 100644
>> --- a/lib/Makefile
>> +++ b/lib/Makefile
>> @@ -49,6 +49,7 @@ obj-$(CONFIG_FIND_BIT_BENCHMARK) += find_bit_benchmark.o
>> obj-$(CONFIG_TEST_BPF) += test_bpf.o
>> obj-$(CONFIG_TEST_FIRMWARE) += test_firmware.o
>> obj-$(CONFIG_TEST_SYSCTL) += test_sysctl.o
>> +obj-$(CONFIG_TEST_CRC) += test_crc.o
>> obj-$(CONFIG_TEST_HASH) += test_hash.o test_siphash.o
>> obj-$(CONFIG_TEST_KASAN) += test_kasan.o
>> CFLAGS_test_kasan.o += -fno-builtin
>> diff --git a/lib/test_crc.c b/lib/test_crc.c
>> new file mode 100644
>> index 000000000000..441bf835fbd3
>> --- /dev/null
>> +++ b/lib/test_crc.c
>> @@ -0,0 +1,138 @@
>> +// SPDX-License-Identifier: GPL-2.0
>> +/*
>> + * CRC test driver
>> + *
>> + * Copyright (C) 2018 Coly Li <colyli@...e.de>
>> + *
>> + * This module provides an simple framework to check the consistency of
>> + * Linux kernel CRC calculation routines in lib/crc*.c. This driver
>> + * requires CONFIG_CRC* items to be enabled if the associated routines are
>> + * tested here. The test results will be printed to kernel message
>> + * when this test driver is loaded.
>> + *
>> + * Current test routines are,
>> + * - crc64()
>> + * - crc64_bch()
>> + * - crc64_update()
>> + *
>> + */
>> +
>> +#include <linux/async.h>
>> +#include <linux/delay.h>
>> +#include <linux/fs.h>
>> +#include <linux/list.h>
>> +#include <linux/module.h>
>> +#include <linux/printk.h>
>> +#include <linux/miscdevice.h>
>> +#include <linux/slab.h>
>> +#include <linux/uaccess.h>
>> +#include <linux/vmalloc.h>
>> +#include <linux/crc64.h>
>> +
>> +struct crc_test_record {
>> + char *name;
>> + u64 data[4];
>> + u64 initval;
>> + u64 expval;
>> + int (*handler)(struct crc_test_record *rec);
>> +};
>> +
>> +static int chk_and_msg(const char *name, u64 crc, u64 expval)
>> +{
>> + int ret = 0;
>> +
>> + if (crc == expval) {
>> + pr_info("test_crc: %s: PASSED:(0x%016llx, expected 0x%016llx)\n",
>> + name, crc, expval);
>
> I don't think we should have specific kernel output for passed tests.
> If a new test is added which follows this pattern, the 0-day will fail
> because the kernel output would change. Along the lines of "silence is
> golden", if no test hit the error output, we're good.
>
>> + } else {
>> + pr_err("test_crc: %s: FAILED:(0x%016llx, expected 0x%016llx)\n",
>> + name, crc, expval);
>> + ret = -EINVAL;
>> + }
>> +
>> + return ret;
>> +}
>> +
>> +/* Add your crc test cases here */
>> +static int test_crc64(struct crc_test_record *rec)
>> +{
>> + u64 crc;
>> +
>> + crc = crc64(rec->data, sizeof(rec->data));
>> + return chk_and_msg(rec->name, crc, rec->expval);
>> +}
>> +
>> +static int test_crc64_bch(struct crc_test_record *rec)
>> +{
>> + u64 crc;
>> +
>> + crc = crc64_bch(rec->data, sizeof(rec->data));
>> + return chk_and_msg(rec->name, crc, rec->expval);
>> +}
>> +
>> +static int test_crc64_update(struct crc_test_record *rec)
>> +{
>> + u64 crc = rec->initval;
>> +
>> + crc = crc64_update(crc, rec->data, sizeof(rec->data));
>> + return chk_and_msg(rec->name, crc, rec->expval);
>> +}
>> +
>> +/*
>> + * Set up your crc test initial data here.
>> + * Do not change the existing items, they are hard coded with
>> + * pre-calculated values.
>> + */
>> +static struct crc_test_record test_data[] = {
>> + { .name = "crc64",
>> + .data = { 0x42F0E1EBA9EA3693, 0x85E1C3D753D46D26,
>> + 0xC711223CFA3E5BB5, 0x493366450E42ECDF },
>> + .initval = 0,
>> + .expval = 0xe2b9911e7b997201,
>> + .handler = test_crc64,
>> + },
>> + { .name = "crc64_bch",
>> + .data = { 0x42F0E1EBA9EA3693, 0x85E1C3D753D46D26,
>> + 0xC711223CFA3E5BB5, 0x493366450E42ECDF },
>> + .initval = 0,
>> + .expval = 0xd2753a20fd862892,
>> + .handler = test_crc64_bch,
>> + },
>> + { .name = "crc64_update",
>> + .data = { 0x42F0E1EBA9EA3693, 0x85E1C3D753D46D26,
>> + 0xC711223CFA3E5BB5, 0x493366450E42ECDF },
>> + .initval = 0x61C8864680B583EB,
>> + .expval = 0xb2c863673f4292bf,
>> + .handler = test_crc64_update,
>> + },
>> + {}
>> +};
>> +
>> +static int __init test_crc_init(void)
>> +{
>> + int i;
>> + int v, err = 0;
>> +
>> + pr_info("Kernel CRC consitency testing:\n");
>> + for (i = 0; test_data[i].name; i++) {
>> + v = test_data[i].handler(&test_data[i]);
>> + if (v < 0)
>> + err++;
>> + }
>> +
>> + if (err == 0)
>> + pr_info("test_crc: all %d tests passed\n", i);
>
> Similar to previous comment: we should not report the number of passed
> tests, since adding a test would invalidate previous golden output.
> Also, consider the situation where some tests are conditionally
> executed depending on kconfig.
Sure, I fix this in v4 series. Thanks.
Coly Li
[snipped]
Powered by blists - more mailing lists