Re: [PATCH v1 6/6] x86/boot: Support nocfg parameter for earlyprintk

[Ingo Molnar]

* Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx> wrote:

> On Tue, 2018-01-16 at 04:12 +0100, Ingo Molnar wrote:
> > * Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx> wrote:
> > 
> > > @@ -133,12 +135,16 @@ static void parse_earlyprintk(void)
> > >  		if (arg[pos] == ',')
> > >  			pos++;
> > >  
> > > -		baud = simple_strtoull(arg + pos, &e, 0);
> > > -		if (baud == 0 || arg + pos == e)
> > > -			baud = DEFAULT_BAUD;
> > > +		if (strncmp(arg + pos, "nocfg", 5)) {
> > > +			baud = simple_strtoull(arg + pos, &e, 0);
> > > +			if (baud == 0 || arg + pos == e)
> > > +				baud = DEFAULT_BAUD;
> > > +		} else {
> > > +			configure = false;
> > > +		}
> > >  	}
> > >  
> > > -	early_serial_init(port, baud);
> > > +	early_serial_init(port, baud, configure);
> > >  }
> > >  
> > >  #define BASE_BAUD (1843200/16)
> > > @@ -162,6 +168,7 @@ static void parse_console_uart8250(void)
> > >  	char optstr[64], *options;
> > >  	int baud = DEFAULT_BAUD;
> > >  	unsigned long port = 0;
> > > +	bool configure = true;
> > >  
> > >  	/*
> > >  	 * console=uart8250,io,0x3f8,115200n8
> > > @@ -179,12 +186,16 @@ static void parse_console_uart8250(void)
> > >  	else
> > >  		return;
> > >  
> > > -	if (options && (options[0] == ','))
> > > -		baud = simple_strtoull(options + 1, &options, 0);
> > > -	else
> > > +	if (options[0] == ',') {
> > > +		if (strncmp(options + 1, "nocfg", 5))
> > > +			baud = simple_strtoull(options + 1,
> > > &options, 0);
> > > +		else
> > > +			configure = false;
> > > +	} else {
> > >  		baud = probe_baud(port);
> > 
> > These code patters seem very similar - could a common function be
> > factored out, to 
> > simplify future changes (such as the one done here)?
> 
> Need to think about. Moreoever, arch/x86/kernel/early_print.c contains
> even more duplication, though I understand why it's split to different
> folders.
> 
> And on top of that we have earlycon (which indeed would be more
> preferable solution). Perhaps, instead of playing with earlyprintk at
> boot stage we might parse earlycon option that more flexible?
> 
> P.S. In any choice at least patch 1 (and maybe patch 2) would be needed.

I'm fine with your current approach - and earlyprintk is preferred by many kernel 
developers. Was just wondering how hard it would be to create a common parser - 
and whether that's desirable at all (it might not be).

Thanks,

	Ingo