[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <08cc095f-bf31-c4e3-0658-f4dfec1d4958@infradead.org>
Date: Thu, 23 Jul 2020 07:22:53 -0700
From: Randy Dunlap <rdunlap@...radead.org>
To: Huang Guobin <huangguobin4@...wei.com>, haren@...ibm.com,
ddstreet@...e.org, herbert@...dor.apana.org.au
Cc: linux-kernel@...r.kernel.org
Subject: Re: [PATCH] lib: Verify array index is correct before using it
On 7/23/20 6:48 AM, Huang Guobin wrote:
> This code reads from the array before verifying that "c" is a valid
> index. Move test array offset code before use to fix it.
>
> Fixes: 2da572c959dd ("lib: add software 842 compression/decompression")
> Signed-off-by: Huang Guobin <huangguobin4@...wei.com>
Hi,
Why shouldn't these locations use array_index_nospec() from
#include <linux/nospec.h>
?
> ---
> lib/842/842_compress.c | 7 +++++--
> 1 file changed, 5 insertions(+), 2 deletions(-)
>
> diff --git a/lib/842/842_compress.c b/lib/842/842_compress.c
> index c02baa4168e1..10f9f8a0d05b 100644
> --- a/lib/842/842_compress.c
> +++ b/lib/842/842_compress.c
> @@ -222,12 +222,13 @@ static int add_bits(struct sw842_param *p, u64 d, u8 n)
> static int add_template(struct sw842_param *p, u8 c)
> {
> int ret, i, b = 0;
> - u8 *t = comp_ops[c];
> + u8 *t = NULL;
> bool inv = false;
>
> if (c >= OPS_MAX)
> return -EINVAL;
>
> + t = comp_ops[c];
> pr_debug("template %x\n", t[4]);
>
> ret = add_bits(p, t[4], OP_BITS);
> @@ -379,12 +380,14 @@ static int add_end_template(struct sw842_param *p)
>
> static bool check_template(struct sw842_param *p, u8 c)
> {
> - u8 *t = comp_ops[c];
> + u8 *t = NULL;
> int i, match, b = 0;
>
> if (c >= OPS_MAX)
> return false;
>
> + t = comp_ops[c];
> +
> for (i = 0; i < 4; i++) {
> if (t[i] & OP_ACTION_INDEX) {
> if (t[i] & OP_AMOUNT_2)
>
thanks.
--
~Randy
Powered by blists - more mailing lists