linux-kernel - Re: [PATCH RESEND 1/3] printk: convert byte-buffer to variable-length record buffer

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]

Message-ID: <CAPXgP13-XQdn0wMP=coVqS=Eko1vWkOrjOtfq4JQLHFZ3JNbgg@mail.gmail.com>
Date:	Wed, 9 May 2012 11:38:38 +0200
From:	Kay Sievers <kay@...y.org>
To:	Linus Torvalds <torvalds@...ux-foundation.org>
Cc:	Sasha Levin <levinsasha928@...il.com>,
	Greg Kroah-Hartmann <greg@...ah.com>,
	Ingo Molnar <mingo@...nel.org>, linux-kernel@...r.kernel.org
Subject: Re: [PATCH RESEND 1/3] printk: convert byte-buffer to variable-length
 record buffer

On Wed, May 9, 2012 at 5:52 AM, Linus Torvalds
<torvalds@...ux-foundation.org> wrote:
> On Tue, May 8, 2012 at 4:14 AM, Kay Sievers <kay@...y.org> wrote:
>>
>> Yeah, we need to make sure, we never merge the (always racy)
>> continuation printk() users with (non-racy) non-continuation users.
>> Therefore KERN_CONT is required to suppress the newline and to merge the
>> content with the earlier non-newline-terminated printk() line.
>
> Why?

The idea was: Prefixes are not used that often, but not using a prefix
should not expose the user to wrongly get appended to an earlier
non-terminated line of another thread.

The point was to limit the "risk" of wrong merges to users of
continuation, and not to users which send ordinary "atomic" lines.

> I really think this is just a bug in the new code.
>
> KERN_CONT should not be needed if the previous printk didn't have a final "\n".

It significantly limits wrong merges, especially for users which can
rightfully expect not to get a wrong merge.

> We made it easier to use printk for a reason a few months ago. The new
> rules are:
>
>  - If you have a KERN_<loglevel>, it *always* starts a new line, the
> obvious exception being KERN_CONT
>
>  - the loglevels *only* matter at the start of the printk - so if you
> have '\n' embedded in a single printk, that changes nothing
> what-so-ever. It's not line-based.

It is a different behaviour. "Innocent" users are not exposed to "risky" users.

>  - if you didn't have a '\n', and don't have a loglevel, KERN_CONT is implied.

But a lot of stuff which does not look for continuation, has no
prefix, hasn't it? I rather make the "I want to be appended" explicit,
instead giving the "I don't care about the log level" any meaning.

I think continuation is special, and ideally should not be much used.
We should not "optimize" for it, and not accept breakage introduced to
ordinary users that way.

> Quite frankly, those three rules (a) make sense and (b) make things easy.

It's true, it's much easier, but it's also much less reliable.

> Breaking them now is a bug. Please don't go adding ugly KERN_CONT when
> there really isn't any reason for it. Just fix the printk code you
> broke.

I can do this, I just don't think it's the right thing to do.

I surely would prefer reliability over rather weird heuristics for
special cases. Today, we should be able to trust at least
non-continuation printk users, which we can't, if we do expose them to
the very real problems of continuation.

Still not convinced? :)

Kay
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/