Re: r326734: serial console fails: boot stuck at: Consoles: internal video/keyboard

From: O. Hartmann <o.hartmann_at_walstatt.org>
Date: Sun, 10 Dec 2017 02:35:03 +0100
Am Sat, 9 Dec 2017 14:30:38 -0700
Warner Losh <imp_at_bsdimp.com> schrieb:

> On Sat, Dec 9, 2017 at 1:47 PM, O. Hartmann <o.hartmann_at_walstatt.org> wrote:
> 
> > Am Sat, 9 Dec 2017 13:00:12 -0700
> > Warner Losh <imp_at_bsdimp.com> schrieb:
> >  
> > > On Sat, Dec 9, 2017 at 12:13 PM, Michael Tuexen <tuexen_at_freebsd.org>  
> > wrote:  
> > >  
> > > > > On 9. Dec 2017, at 19:41, O. Hartmann <ohartmann_at_walstatt.org>  
> > wrote:  
> > > > >
> > > > > Running a PCengines APU2C4 with FreeBSD Current, r326734, fails to  
> > boot,  
> > > > it gets stuck at  
> > > > > the very first messages on the serial console prompting
> > > > >
> > > > > Consoles: internal video/keyboard
> > > > >
> > > > > and then nothing more.  
> > > > I don't think it is a general problem with r326734. I'm running on the
> > > > same hardware
> > > > that revision (with some modifications to the TCP I'm testing):  
> > > > > uname -a  
> > > > FreeBSD nf3.testbed 12.0-CURRENT FreeBSD 12.0-CURRENT #31 r326734M: Sat
> > > > Dec  9 20:03:59 CET 2017     tuexen_at_nf3.testbed:/usr/home/
> > > > tuexen/head/sys/amd64/compile/TCP  amd64
> > > >  
> > >
> > > Hi Michael,
> > >
> > > What's your boot setup? BIOS vs UEFI? UFS vs ZFS? GELI? GPT vs MBR?  
> > Plants  
> > > vs Zombies? And did you update your boot blocks or not?
> > >
> > > Warner  
> >
> > All right, I got your point here, my bad, my apaologies.
> >
> > I boot from a NanoBSD-prepared image residing on a SD card. The partiton
> > scheme is GPT,
> > non UEFI (the SeaBIOS, latest version available for the APU2C2 available)
> > doesn't support
> > UEFI. NanoBSD scripts (legacy.sh) has been modified to fit our requests.
> >
> > The modifications target at most the creation of a GPT partition scheme
> > layout,
> > installing boot/pmbr as bootcode and boot/gptboot on a freebsd-boot
> > partition as well as
> > boot/boot1.efifat onto a dedicated efi-type partition. The UEFI partition
> > is usually
> > created before(!) the freebsd-boot partition, but in this specific
> > version, there is NO
> > EFI partition, only the gptboot containing partition. The binaries of the
> > above mentioned
> > bootcode/efi code images are taken from the newly build world. I do not
> > use boot0sio, the
> > portion where it is installed is excluded by some "if GPT; then" clause.
> >
> > The last known-good version I reported in working is the image I backuped
> > last time. I
> > lost the USB 2.0 flash device today containing the last version, that was
> > not far from
> > r326218, the version that corrupted FreeBSD, I guess it was r326184, but I
> > really do not
> > know and I havn't a backup of that image.
> >
> > The buildworld process is maintained by a bunch of WITHOUT_ statements in
> > a file driving
> > NanoBSD. This just for the record just in case FreeBSD build system changed
> > something significantly.
> >  
> 
> Should, but if you could share what those all are, I can tell you if any
> might cause an issue that would help us root cause.
> 
> 
> > Building a world takes 90 minutes or more, so bi-secting the problem would
> > be a pain in
> > the arse (that said, I imply that world and kernel need to be in sync).
> >
> > I would appreciate hints or tipps where to look after or how to increase
> > verbosity
> > especially at the first boot stage.
> >  
> 
> OK. So you are booting with gptboot off an SD card on the PC Engines system
> via legacy methods for a UFS root partition. GELI is not involved, correct?

I do not use GELI. GEOM_GELI is in ther kernel compiled in. No modules are allowed to be
loaded.

The disk image created and mounted via md0 (mdconfig -a -t vndo -f filename.img) looks
like:

=>     40  5175704  md0  GPT  (2.5G)
       40     1024    2  boot0  (512K)
     1064  2059464    3  dsks1a  (1.0G)
  2060528  2063643    4  dsks2a  (1.0G)
  4124171  1048576    5  dsks3  (512M)
  5172747     2997       - free -  (1.5M)

This is the image type without UEFI partition, legacy boot only via gptboot.

> 
> Sadly, there's no easy way to increase the verbosity of the early boot
> since the boot blocks have to live in such a constrained environment
> there's a strong bias against including debug code.
> 
> But what we know is encouraging. We know that gptboot properly loaded, and
> that it found /boot/loader, loaded it, and then we had the issue.  This
> helps us constrain things somewhat in testing. For the moment, we'll assume
> that gptboot is good, and concentrate on when /boot/loader went south.
> Before we do that, can I get any loader.conf and /boot.config /boot/config
> files you have?

boot.config/loader.conf are non existent.

[loader.conf.local]
boot_serial="YES"
comconsole_speed="115200"
console="comconsole"

autoboot_delay="5"

verbose_loading="YES"
loader_logo="none"
beastie_disable="NO"

kern.geom.label.gptid.enable=0
hw.physmem=1073741824
# Da mehr als 1 igb NIC an Bord! Siehe man igb(4)
kern.ipc.nmbclusters=757350
hw.intrbalance=1
hw.em.max_interrupt_rate=16000
net.fibs=6
net.add_addr_allfibs=0

> 
> src/stand is easily buildable standalone these days, so my thought is
> trying to build it at r326550 and putting that onto into your image.
> src/stand stakes 1 minute to build on my box, so 5 on yours max :). You can
> even build it standalone in a separate tree if that would help keep
> whatever world you're using to do the builds purer. If we can't get
> satisfaction doing that, then we'll drop back to plan B: if you give me all
> the WITH/WITHOUT you are using, I could send you a tarball with about two
> dozen /boot/loaders to try which should make bisecting not a total pain in
> the arse. I'd likely toss in gptboot as well, since it's small. Let me know
> :)

Well, that is a kind offer, thanks. Tomorrow morning, I'll try r326550. For now, I spent
too much  time trying to rid of this.

Kind regards,

Oliver

> 
> Since this is very early in /boot/loader's purview, I don't think
> kernel/world matter at all (apart from the fact that loader is built as
> part of world).
> 
> The PCengines APU with the most recent SeaBIOS isn't capable of booting
> > FreeBSD from USB
> > 3.0 devices. Even from USB 2.0 flashdrives working images fail at the boot
> > loader with
> > "failed with error 19".  
> 
> 
> OK. I assume this is a different issue. :)
> 
> Warner
> 
> 
> > >
> > >
> > >  
> > > > Best regards
> > > > Michael  
> > > > >
> > > > > FreeBSD CURRENT FreeBSD 12.0-CURRENT #52 r324234: Tue Oct  3 11:00:53  
> > > > CEST 2017 amd64  
> > > > > works fine.
> > > > >
> > > > > What the heck has changed?
> > > > >
> > > > > Kind regards,
> > > > >
> > > > > oh
> > > > >
> > > > > --
> > > > > O. Hartmann
> > > > >
> > > > > Ich widerspreche der Nutzung oder Übermittlung meiner Daten für
> > > > > Werbezwecke oder für die Markt- oder Meinungsforschung (§ 28 Abs. 4  
> > > > BDSG).
> > > >
> > > > _______________________________________________
> > > > freebsd-current_at_freebsd.org mailing list
> > > > https://lists.freebsd.org/mailman/listinfo/freebsd-current
> > > > To unsubscribe, send any mail to "freebsd-current-unsubscribe_at_  
> > freebsd.org"  
> > > >  
> > > _______________________________________________
> > > freebsd-current_at_freebsd.org mailing list
> > > https://lists.freebsd.org/mailman/listinfo/freebsd-current
> > > To unsubscribe, send any mail to "freebsd-current-unsubscribe_at_  
> > freebsd.org"
> >
> >
> >
> > --
> > O. Hartmann
> >
> > Ich widerspreche der Nutzung oder Übermittlung meiner Daten für
> > Werbezwecke oder für die Markt- oder Meinungsforschung (§ 28 Abs. 4 BDSG).
> >  



-- 
O. Hartmann

Ich widerspreche der Nutzung oder Übermittlung meiner Daten für
Werbezwecke oder für die Markt- oder Meinungsforschung (§ 28 Abs. 4 BDSG).

Received on Sun Dec 10 2017 - 00:35:15 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:41:14 UTC