linux-kernel - Re: [GIT PULL] MFD fixes for v5.2

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

Date:   Mon, 17 Jun 2019 10:33:20 -0700
From:   Linus Torvalds <torvalds@...ux-foundation.org>
To:     Lee Jones <lee.jones@...aro.org>,
        Dan Carpenter <dan.carpenter@...cle.com>
Cc:     Linux List Kernel Mailing <linux-kernel@...r.kernel.org>
Subject: Re: [GIT PULL] MFD fixes for v5.2

On Mon, Jun 17, 2019 at 3:01 AM Lee Jones <lee.jones@...aro.org> wrote:
>
> Enjoy!

No.

This is still entirely wrong.

You can't just randomly cast an "u32 *" to "unsigned long *".

It wasn't correct when you did it the other way in regmap_read(), but
it's also not correct when you now for it this way for
for_each_set_bit().

You can do

    u32 regmap_bits;
    unsigned long bits;

and then

    ret = regmap_read(stmfx->map, STMFX_REG_IRQ_PENDING, &regmap_bits);
    ...
    bits = regmap_bits;
    for_each_set_bit(n, &bits, STMFX_REG_IRQ_SRC_MAX) ..

but casting pointers at either point is *completely* wrong.

Yes, yes, it happens to work on little-endian, but on a 64-bit
big-endian machine, the low 32 bits of the "unsigned int" will have
absolutely _zero_ overlap with the low 32 bits of the "unsigned long"
in memory.

When you moved the cast to for_each_set_bit(), it only moves the
access of the bogus bits to another place instead.

So that patch doesn't fix anything at all, it only moves the same error around.

                      Linus