Re: [Ksummit-discuss] [TECH TOPIC] asynchronous printk

[David Woodhouse <dwmw2@infradead.org>]

[-- Attachment #1: Type: text/plain, Size: 1309 bytes --]

On Thu, 2016-07-21 at 04:19 -0700, Josh Triplett wrote:
> 
> I've run into various cases where "console=ttyS0,115200" didn't produce
> any serial output at all, but "earlyprintk=ttyS0,115200,keep" did.

Note that those are very different things. Because 'ttyS0' for
earlyprintk really means '16x50 UART at 0x3f8'. While the later one
actually depends on it being discovered by the various options we have
for finding them, some of which are only available quite late.

> Perhaps "earlyprintk" expects less from the serial port than "console"
> does, but that suggests that the differences involve more than just how
> early the driver can work.

There *are* differences, yes. But they are almost entirely gratuitous.
Fundamentally, the only interesting thing is how early the driver can
shovel bits out the port, and thus how early it can be registered.

In the case of serial ports, we can register the console early and just
need to sure it keeps working when the driver is *later* fully
initialised.

It's slightly more fun in cases like the bootx console which actually
does stop working when the graphics hardware is properly set up, but
that's not really hard to cope with in the generic code either; it
doesn't need to be *such* a horrid mess of special handling.

-- 
dwmw2

[-- Attachment #2: smime.p7s --]
[-- Type: application/x-pkcs7-signature, Size: 5760 bytes --]