linux-kernel - Re: [PATCH] epoll: fix dereferenced before check pt

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: <20171222000207.GY21978@ZenIV.linux.org.uk>
Date:   Fri, 22 Dec 2017 00:02:07 +0000
From:   Al Viro <viro@...IV.linux.org.uk>
To:     Vasyl Gomonovych <gomonovych@...il.com>
Cc:     linux-fsdevel@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] epoll: fix dereferenced before check pt

On Thu, Dec 21, 2017 at 04:54:17PM +0100, Vasyl Gomonovych wrote:
> This patch fixes the warning reported by smatch:
> 
> fs/eventpoll.c:889 ep_item_poll() warn: variable dereferenced before check 'pt'

... and of course, that is all the analysis one might want - the tool has
spoken.  Except that the only thing proven is that the tool is either misused
or just plain crappy.

Callers of ep_item_poll() are easily found - it's static, so they all are in
the same file.  All four of them -
                if (ep_item_poll(epi, &pt, depth)) {
        revents = ep_item_poll(epi, &epq.pt, 1);
        if (ep_item_poll(epi, &pt, 1)) {
                revents = ep_item_poll(epi, &pt, 1);

Figuring out why an address of a local variable or of a field thereof can never
be NULL is left as an exercise for an introductory course on C.

"Fixing warnings" is bloody wrong - warning (from any tool) is a heuristic
pointing to a place that might be worth looking into.  In cases when you've
found a real problem with help from some tool, sure, go ahead and credit that
tool in the commit message.  Along with the description of the problem
you've found.