Re: [RFC v2] kernel/printk.c - handling more than one CON_BOOT

From: Andrew Morton
Date: Wed Jul 01 2009 - 17:03:50 EST


On Wed, 1 Jul 2009 16:50:19 -0400
Robin Getz <rgetz@xxxxxxxxxxxxxxxxxxxx> wrote:

> On Wed 1 Jul 2009 15:44, Andrew Morton pondered:
> > On Tue, 30 Jun 2009 22:51:52 -0400
> > Robin Getz <rgetz@xxxxxxxxxxxxxxxxxxxx> wrote:
> >
> > > From: Robin Getz <rgetz@xxxxxxxxxxxxxxxxxxxx>
> > >
> > > Today, register_console() assumes the following usage:
> > > - The first console to register with a flag set to CON_BOOT is the
> > > one and only bootconsole.
> > > - If another register_console() is called with an additional
> > > CON_BOOT, it is silently rejected.
> > > - As soon as a console without the CON_BOOT set calls registers
> > > the bootconsole is automatically unregistered.
> > > - Once there is a "real" console - register_console() will silently
> > > reject any consoles with it's CON_BOOT flag set.
> >
> > hm. I never knew all that. Thanks for taking the time to explain all
> > this - it helps.
>
> Is it also worthwhile to mention that this function (when used with
> early_printk) can get called _very_ early in the boot process, (before
> setup_arch() completes)

OK.

> > I think it would be useful if we had a description of the new design,
> > similar to your description of the "Today" behaviour above. Perhaps
> > as a comment over register_console()?
>
> Can do - do you want it as v3 or on-top of this one?

I'm easy either way. Ingo might have applied to one of his trees, in
which case a replacement patch might cause him problems, dunno.

> > > In many systems (alpha, blackfin, microblaze, mips, powerpc, sh, &
> > > x86), there are early_printk implementations, which use the CON_BOOT
> > > which come out serial ports, vga, usb, & memory buffers. In many
> > > embedded systems, it would be nice to have two - in case the primary
> > > fails, you always have access to a backup memory buffer - but this
> > > requires at least two CON_BOOT consoles.
> > >
> > > This changeset allows multiple boot consoles, and changes the
> > functionality
> > > to, be mostly the same as the above.
> > > - Any number CON_BOOT consoles of can be registered
> > > - A "real" console will unregister all the CON_BOOT consoles
> > > - Once a "real" console is registered, no more CON_BOOT consoles
> > > can be added (still silently rejected)
> >
> > Is the "silent" rejection desirable? Perhaps that's a
> > programming/configuration error which the developer should be informed
> > of?
>
> That is what it is today. I would actually prefer not silent - as it
> would have meant 10 min that I would have saved myself trying to figure out
> what was going on... :)
>
> But when I was figuring out how things worked, and had a few too many printks
> in register_console - I was getting BUG: "recent printk recursion!" - so I
> just left it silent (rather than figuring this out).

OK - I was just wondering. Whatever you think is best..

> >
> > > diff -puN
> > kernel/printk.c~kernel-printkc-handling-more-than-one-con_boot
> > kernel/printk.c
> > > --- a/kernel/printk.c~kernel-printkc-handling-more-than-one-con_boot
> > > +++ a/kernel/printk.c
> > > @@ -37,6 +37,12 @@
> > > #include <asm/uaccess.h>
> > >
> > > /*
> > > + * for_each_console() allows you to iterate on each console
> > > + */
> > > +#define for_each_console(con) \
> > > + for (con = console_drivers; con != NULL; con = con->next)
> >
> > hum. Fair enough.
>
> Was there an issue with this? (Or you just want me to split that up into a
> separate patch?)

Just a general allergy to fancypants macros, especially those which
hide control flow. But this one is straightforward enough.

Formally one should parenthesize `con' here, but if that change is ever
actually necessary then we have other problems, and the macro was
probably a bad idea anyway.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/