Обсуждение: Wrong buffer limits check

Поиск
Список
Период
Сортировка

Wrong buffer limits check

От
Mikhail Gribkov
Дата:
Hi hackers,

I have tried to analyse Postgres code with Svace static analyzer [1] and found something I think is a real bug.

In pgp-decrypt.c, in prefix_init function the following check:
if (len > sizeof(tmpbuf))

seem to be erroneous and should really look this way:
if (len > PGP_MAX_BLOCK)

Otherwise the below checks in this line could lead to buffer overflows:
if (buf[len - 2] != buf[len] || buf[len - 1] != buf[len + 1])

This is because buf will point to tmpbuf, while tmpbuf have a size of PGP_MAX_BLOCK + 2.

What do you think? The proposed patch towarts the current master branch is attached.

[1] - https://svace.pages.ispras.ru/svace-website/en/

--
 best regards,
    Mikhail A. Gribkov

e-mail: youzhick@gmail.com
http://www.strava.com/athletes/5085772
phone: +7(916)604-71-12
Telegram: @youzhick

Вложения

Re: Wrong buffer limits check

От
Daniel Gustafsson
Дата:
> On 29 Jan 2024, at 14:37, Mikhail Gribkov <youzhick@gmail.com> wrote:

> I have tried to analyse Postgres code with Svace static analyzer [1] and found something I think is a real bug.
>
> In pgp-decrypt.c, in prefix_init function the following check:
> if (len > sizeof(tmpbuf))
>
> seem to be erroneous and should really look this way:
> if (len > PGP_MAX_BLOCK)

Studying the code I think you're right, we should be ensuring that the cipher
block size isn't exceeding PGP_MAX_BLOCK.  In practice it seems night
impossible to hit given the ciphers in cipher_list, but we should still fix it.
Unsurprisingly this seems to have been there forever (since July 2005) so needs
to be backpatched to all supported branches for the sake of consistency

--
Daniel Gustafsson