[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <61528e45-b2a7-5ef9-09dd-9cdc63c4599e@omp.ru>
Date: Mon, 12 Aug 2024 12:41:03 +0300
From: Sergey Shtylyov <s.shtylyov@....ru>
To: Roman Smirnov <r.smirnov@....ru>, "David S. Miller" <davem@...emloft.net>,
Eric Dumazet <edumazet@...gle.com>, Jakub Kicinski <kuba@...nel.org>, Paolo
Abeni <pabeni@...hat.com>, Breno Leitao <leitao@...ian.org>
CC: <netdev@...r.kernel.org>, <linux-kernel@...r.kernel.org>, Karina Yankevich
<k.yankevich@....ru>, <lvc-project@...uxtesting.org>
Subject: Re: [PATCH] drivers: net: bsd_comp: fix integer overflow in
bsd_decompress()
Hello!
Failed to notice an issue with the subject in the internal review;
I think it should look like "net: ppp: bsd_comp: fix integer overflow
in bsd_decompress()"...
On 8/12/24 11:43 AM, Roman Smirnov wrote:
> The result of a bit shift has type int.
So far, so good... :-)
> If ibuf is greater than or
*ibuf maybe? :-)
> equal to 128, a sign switch will occur.
I wonder whether you had looked at the .lsy file before writing
that...
Actually, movzvl (%rdi),%eax is used when reading *buf, so no
sign extension occurs at this point... it occurs when casting the
result of shift to *unsignjed long*
> After that, the higher 32 bits in accm will be set to 1.
Only if we have 1 in the bit 31 after shl %cl,%eax...
> Cast the result of the expression to unsigned long.
I strongly suspect we should re-declare accm as *unsigned int*
instead...
> Found by Linux Verification Center (linuxtesting.org) with Svace.
>
> Signed-off-by: Roman Smirnov <r.smirnov@....ru>
> ---
> drivers/net/ppp/bsd_comp.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/net/ppp/bsd_comp.c b/drivers/net/ppp/bsd_comp.c
> index 55954594e157..078fe8c9bee8 100644
> --- a/drivers/net/ppp/bsd_comp.c
> +++ b/drivers/net/ppp/bsd_comp.c
> @@ -918,7 +918,7 @@ static int bsd_decompress (void *state, unsigned char *ibuf, int isize,
> */
>
> bitno -= 8;
> - accm |= *ibuf++ << bitno;
> + accm |= (unsigned long)(*ibuf++) << bitno;
> if (tgtbitno < bitno)
> {
> continue;
>
MBR, Sergey
Powered by blists - more mailing lists