linux-kernel - Re:Re:[PATCH 6/6] irq: Potentially 'offset out of size' bug

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

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <87k0jjcpgn.ffs@tglx>
Date:   Tue, 14 Sep 2021 14:27:04 +0200
From:   Thomas Gleixner <tglx@...utronix.de>
To:     Jiasheng Jiang <jiasheng@...as.ac.cn>
Cc:     linux-kernel@...r.kernel.org
Subject: Re:Re:[PATCH 6/6] irq: Potentially 'offset out of size' bug

On Tue, Sep 14 2021 at 08:48, Jiasheng Jiang wrote:
>> On Fri, Sep 10 2021 at 03:26, Jiasheng Jiang wrote:
>> The find_next_bit() use nr_irqs as size, and using it without
>> any check might cause its returned value out of the sizei
>
> On Fri, Sep 10 2021 at 18:28, tglx wrote:
>> Why exactly is this a problem? The return value has to be checked at the
>> call site anyway.
>
> There is really a check at the call site, but the annotation of the 
> irq_get_next_irq() is 'Returns next irq number after offset or nr_irqs 
> if none is found', which tells the programmer should not check the
> return value of it. In case of a programmer write a new call for the
> irq_get_next_irq(), he may not check the return value because of the 
> annotation said.

The return value has always to be checked because nr_irqs is guaranteed
to be an invalid index.

> Therefore, it had better to add the check inside of irq_get_next_irq()
> to fit for the annotation.

Care to look what find_next_bit(..., size) does?

    * Returns the bit number for the next set bit
    * If no bits are set, returns @size.

So for:

    res = find_next_bit(addr, size, offset);

res is guaranteed to be:

    offset < res <= size

IOW. irq_get_next_irq() is doing exactly what the comment says.

So again, which problem are you trying to solve?

Thanks,

        tglx