Current-Users archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: [5.0_RC2] long delay in /boot

On Feb 22, 2009, at 11:38 , Jared D. McNeill wrote:

On Sun, 22 Feb 2009, Andrew Doran wrote:
On Sun, Feb 22, 2009 at 10:38:29AM -0500, Jared D. McNeill wrote:

Joerg Sonnenberger wrote:
On Sat, Feb 21, 2009 at 03:55:50PM -0500, Andreas Wrede wrote:
On a NetBSD/amd64 5.0_RC2 system, the boot process hangs for 4 min 52
sec (+- 1 sec) after printing the loaded segment sizes and before
printing the Copyright line. This is happening in /boot as booting /
usr/pkg/mdec/memtestplus is also delayed by the same 4 min 52 sec.

Try to build a kernel without pckbd. We had some issues on cetrain HP
systems with the built-in keyboard emulation being extremely slow,
resulting in similar symptoms.

How is that going to help /boot?

Presumably this is the same problem that affects some Soekris boxes? Or are
they from another vendor?

Must be a different vendor, Soekris doesn't have any amd64 products. ISTR that the Soekris problem was that it took ~5 minutes for boot2 to show up, instead of ~5 minutes for the kernel to start.

Yes and on my NET5501 the delay between mbr boot and /boot is still 2 min 10 sec, with mbr and bootxx_ffsv1 built from 5.0_RC2 sources. After that, /boot loads the kernel and runs it with no delay.

The box with the /boot-to-kernel delay is a ASUS P5Q with a Core 2 Quad CPU.

How can I turn on DEBUG for sys/arch/i386/stand/lib/exec.c, which I think where I would start looking?

On Sun, Feb 22, 2009 at 12:38:13PM -0500, Steven M. Bellovin wrote:
On Sun, 22 Feb 2009 12:19:53 -0500 (EST)
"Jared D. McNeill" <> wrote:

Which points strongly to the boot loader, since memtestplus doesn't use
any part of our kernel.

But it might initialise the pckbc the same way NetBSD does in early
console setup.

In case it matters, I tested both with and without a keyboard attached to pckbc - no difference.


Home | Main Index | Thread Index | Old Index