On Mon, 12 Feb 2001, Scott Laird wrote:
>
> On 12 Feb 2001, H. Peter Anvin wrote:
> >
> > Just checked my own code, and SYSLINUX does indeed support 115200 (I
> > changed this to be a 32-bit register ages ago, apparently.) Still
> > doesn't answer the question "why"... all I think you do is increase
> > the risk for FIFO overrun and lost characters (flow control on a boot
> > loader console is vestigial at the best.)
>
> It's simple -- we want the kernel to have its serial console running at
> 115200, and we don't want to have to change speeds to talk to the
> bootloader.
Exactly.
Then HPA may ask: but why do you want to run the serial console at
115200?? The answer is simple: because we can (or more precisely, because
the HW can ;).
If the hardware is supposed to support 115.2Kbps, why can't / shouldn't
we use it?? Remember, this is not a modem connection, there is no
compression involved, both sides are running 115.2Kbps, so there should
NOT be a risk for FIFO overruns (unless you have buggy hardware). And in
this case, you can then decrease your baud rate. But at least you have the
_option_! :)
> Some boot processes, particularaly fsck, can be *REALLY*
> verbose on screwed up systems. I've seen systems take hours to run fsck,
> even on small filesystems, simply because they were blocking on a 9600 bps
> console.
This is true!!
Another one (not as critical as the fsck though): when compiling the
kernel, sometimes the kernel compilation is done, but the console output
isn't finished yet (I'm serious).
Later,
Ivan
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://vger.kernel.org/lkml/
