Re: [PATCH 3/3] [v4] lib/vsprintf: no_hash_pointers prints all addresses as unhashed

[Vlastimil Babka <vbabka@suse.cz>]

On 3/2/21 2:29 PM, Petr Mladek wrote:
> On Tue 2021-03-02 13:51:35, Geert Uytterhoeven wrote:
>> > > > +
>> > > > +       pr_warn("**********************************************************\n");
>> > > > +       pr_warn("**   NOTICE NOTICE NOTICE NOTICE NOTICE NOTICE NOTICE   **\n");
>> > > > +       pr_warn("**                                                      **\n");
>> > > > +       pr_warn("** This system shows unhashed kernel memory addresses   **\n");
>> > > > +       pr_warn("** via the console, logs, and other interfaces. This    **\n");
>> > > > +       pr_warn("** might reduce the security of your system.            **\n");
>> > > > +       pr_warn("**                                                      **\n");
>> > > > +       pr_warn("** If you see this message and you are not debugging    **\n");
>> > > > +       pr_warn("** the kernel, report this immediately to your system   **\n");
>> > > > +       pr_warn("** administrator!                                       **\n");
>> > > > +       pr_warn("**                                                      **\n");
>> > > > +       pr_warn("**   NOTICE NOTICE NOTICE NOTICE NOTICE NOTICE NOTICE   **\n");
>> > > > +       pr_warn("**********************************************************\n");
>> > > > +
>> > > > +       return 0;
>> > > > +}
>> > > > +early_param("no_hash_pointers", no_hash_pointers_enable);
>> > >
>> > > While bloat-o-meter is not smart enough to notice the real size impact,
>> > > this does add more than 500 bytes of string data to the kernel.
>> > > Do we really need such a large message?
>> > > Perhaps the whole no_hash_pointers machinery should be protected by
>> > > "#ifdef CONFIG_DEBUG_KERNEL"?
> 
> This was the deal. The configure option is a no-go, see below and also
> https://lore.kernel.org/r/CAHk-=wgaK4cz=K-JB4p-KPXBV73m9bja2w1W1Lr3iu8+NEPk7A@mail.gmail.com

I think it's a no-go only when enabling such option equals to "no_hash_pointers"
being always passed. What Geert suggests is that you need both
CONFIG_DEBUG_KERNEL *and* no_hash_pointers and that's different.

>> > We recently stumbled across this, and it appears an increasing number
>> > of production kernels enable CONFIG_DEBUG_KERNEL [1], so it likely
>> > isn't the solution (we tried to use CONFIG_DEBUG_KERNEL in similar
>> 
>> I guess the people who do care about kernel size do know to disable
>> CONFIG_DEBUG_KERNEL, so it would help them.
>> The everything-but-the-kitchen-sink distro people don't care about kernel
>> size anyway.
> 
> The problem with the configure option is not about size. The problem is
> that there would be many kernels in the wild with this option enabled.
> People would install them without knowing that they are less secure.

Same as above.

> Distros would need to provide both kernels. Well, they already do.
> But it might be worse. Some distros might even want to enable it
> by default.
> 
> Also many bugs might be debugged without this option. Some bugs
> are hard to reproduce and might be visible only on production
> systems. It should be possible to enable this only when really
> needed and the user must be aware of the risk.

So this is basically a kernel tinyfication issue, right? Is that still pursued
today? Are there better config options suitable for this than CONFIG_DEBUG_KERNEL?

>> > Would placing the strings into an __initconst array help?
>> 
>> That would indeed help to reduce run-time memory consumption.
> 
> Sure. We could do this. Do you want to send a patch, please?
> 
>> It would not solve the raw kernel size increase.
> 
> I see. Well, the compression should be pretty efficient
> for a text (with many spaces).
> 
> Best Regards,
> Petr
>