Upgrade to 2.1.2: Stuck on 2.1

jimp

I have a faulty image in my hands now, hopefully I can track down a solution soon.

Current theory is that it was actually corrupt before the latest update and it just started to show it now, but once I have more time to experiment with the broken CF image I'll know for sure.

I'm certain we can come up with a fix, but it might be something scary like doing a DD of a good partition table to the start of the disk. Not something I'd generally recommend however in theory all NanoBSD images of the same size should have the same partition layout so it may be safe.

In the meantime I'd like to find out if everyone involved here had 4GB NanoBSD images, or both 2GB and 4GB, or even more.

me too

Both 4GB CF for me.

mjohnson

2 and 4GB for me.

RCS-Michael

Jim,

I have both 2G and 4G nanobsd images displaying this problem.

Michael

jimp

For those of you that have an issue, show me the output of:

fdisk -p /dev/ad0

And note if it's 2gb or 4gb.

If you have a working system of the same size to compare against, show the output from it also.

mkomar

Not working: 2GB

/dev/ad0

g c3875 h16 s63
p 1 0xa5 63 1902033
a 1
p 2 0xa5 1902159 1902033
p 3 0xa5 3804192 102816

Working: 4GB

/dev/ad0

g c7751 h16 s63
p 1 0xa5 63 3844449
p 2 0xa5 3844575 3844449
a 2
p 3 0xa5 7689024 102816

Not Working: 2GB

/dev/ad0

g c3875 h16 s63
p 1 0xa5 63 1902033
p 2 0xa5 1902159 1902033
a 2
p 3 0xa5 3804192 102816

Working: 2GB

/dev/ad0

g c3897 h16 s63
p 1 0xa5 63 1902033
a 1
p 2 0xa5 1902159 1902033
p 3 0xa5 3804192 102816

jimp

Those last two are interesting in that they're nearly identical and one works and the other doesn't. I expect some variation as we have, over time, slightly shrunk the NanoBSD slice sizes, but that is a bit curious.

The .img file I read from the CF with the "corrupt" table appears to be OK, despite the CF showing a damaged table. So I'm left to wonder if there may be some other CF-related factor at play.

The following commands could be dangerous so if you choose to attempt them, proceed with extreme caution. I tested these on my own ALIX with a good MBR and it survived, but there are no guarantees. You need only try one of these methods unless it doesn't help, then proceed to the next one.

Method #1: Rewrite the MBR+Partition table with dd

sysctl kern.geom.debugflags=16
dd if=/dev/ad0 of=/tmp/mbr_part_bkup.img bs=512 count=1
dd of=/dev/ad0 if=/tmp/mbr_part_bkup.img bs=512 count=1

Method #2: Have fdisk reset the partition table:

sysctl kern.geom.debugflags=16
fdisk -p /dev/ad0 > /tmp/fdisk_bkup.txt
fdisk -if /tmp/fdisk_bkup.txt /dev/ad0

Method #3: Take a "working" fdisk output and rewrite using it. I can't stress enough that you must make sure the partition boundaries line up, don't grab the fdisk output from a differently sized card:


# Get the "fdisk -p" output from a similar but working CF, save it in /tmp/fdisk_bkup.txt
sysctl kern.geom.debugflags=16
fdisk -if /tmp/fdisk_bkup.txt /dev/ad0

After any of those, chances are that no commands will work to reboot the unit, so either pull the power or run the following to force a panic+reboot:

sysctl debug.debugger_on_panic=0
sysctl debug.kdb.panic=1

After that has completed, try the upgrade once again.

Obviously that isn't something you'd want to try on a remote unit.

mkomar

All of mine are in remote locations and in production. Can't risk taking them down.

I'll be swapping them out with upgraded (Software) replacements in the next week or so.

Is there value in trying these fixes after that?

jimp

@mkomar:

All of mine are in remote locations and in production. Can't risk taking them down.

I'll be swapping them out with upgraded (Software) replacements in the next week or so.

Is there value in trying these fixes after that?

It would still help to know if any of the above methods would correct the faulty partition table, so that others can benefit from the knowledge.

mkomar

I should have a couple of those units on hand next week. I'll give it a shot and report back once I've done so.

mkomar

JimP - Should have my hands on one or two of the malfunctioning units in the next few days. I'd be happy to try the various fixes you have proposed, and/or if it would be of more value, I'd be happy to either get you serial access to one or both of them and/or get either CF card(s) and/or dd img dumps out to you.

Would any of the options work better than others as far as getting a 'known good' fix out there?

jimp

From my post a few entries up ( https://forum.pfsense.org/index.php?topic=75069.msg413219#msg413219 ) – I listed them in order of preference (and likely destructive potential!)

So try them in that order, method #1, then #2, then #3 only if both 1 and 2 fail.

mkomar

[2.1-RELEASE][root@pfsense]/root(1): sysctl kern.geom.debugflags=16
kern.geom.debugflags: 0 -> 16
[2.1-RELEASE][root@pfsense]/root(2): d if=/dev/ad0 of=/tmp/mbr_part_bkup.img bs=512 count=1
dd of=/dev/ad0 if=/tmp/mbr_part_bkup.img bs=512 count=1dd if=/dev/ad0 of=/tmp/mbr_part_bkup.img bs=512 count=1
1+0 records in
1+0 records out
512 bytes transferred in 0.000771 secs (664033 bytes/sec)
[2.1-RELEASE][root@pfsense]/root(3): dd of=/dev/ad0 if=/tmp/mbr_part_bkup.img bs=512 count=1
1+0 records in
GEOM_PART: integrity check failed (ad0, MBR)

512 bytes transferred in 0.024720 secs (20712 bGEOM: ad0s1: media size does not match label.
ytes/sec)
GEOM: ad0s2: media size does not match label.
[2.1-RELEASE][root@pfsense]/root(4):

pmiller

I also have the problem in this post and have been following it closely. I am particularly interested in an in-place fix since I don't have an extra CF card or reader to re-flash.

mkomar - in the third line of your post shouldn't there be 2 d's instead of one for the 'dd' command? Maybe a copy/paste error?

MattMeyer

Adding my experience to the thread. I am also having the same issue. I am using a SanDisk Extreme 4 GB CF. This is the output for fdisk -p /dev/ad0:

/dev/ad0

g c7751 h16 s63
p 1 0xa5 63 3861585
a 1
p 2 0xa5 3861711 3861585
p 3 0xa5 7723296 102816

i have tried recreating the MBR using both the dd method and the fdisk method. Both did not help with a successful upgrade. I do not have a working device that I can copy the MBR so option #3 is out.

Additionally, trying to change the boot slice does not "stick".

mkomar

@pmiller:

mkomar - in the third line of your post shouldn't there be 2 d's instead of one for the 'dd' command? Maybe a copy/paste error?

Must have been dropped somehow. If it wasn't 'dd' the we would see an error instead of the command output.

I've still got a 'broken' device standing by if someone is interested in finding a reliable on-line fix. In the mean time, I've just done a config backup/restore and replaced the production units.

gazoo

Since 2.1.3 came out, has this solved any of these problems for anyone? I was going to go for it, but I don't know if that's a good idea - wanted to see what others have seen for 2.1.3

I'm on nano alix 4g netgate

trunix

jimp, I've got the same problem on a 4gb CF. Output of fdisk -p /dev/ad0:

/dev/ad0

g c7745 h16 s63
p 1 0xa5 63 3854529
p 2 0xa5 3854655 3854529
a 2
p 3 0xa5 7709184 102816

I've tried method #1 and #2, but neither worked. The output of fdisk -if /tmp/fdisk_bkup.txt /dev/ad0 from method #2 is below in case it's notable. I didn't get any errors from method #1, the system just booted back into 2.1 on the same slice. The same thing happened after method #2. I'm also not able to switch the bootup slice for whatever reason.

fdisk: WARNING line 2: number of cylinders (7745) may be out-of-range
(must be within 1-1024 for normal BIOS operation, unless the entire disk
is dedicated to FreeBSD)
******* Working on device /dev/ad0 *******

This system and CF card have been in stable operation for awhile now and I've successfully installed all the updates from 2.0.1 to 2.1. I never got a chance to install 2.1.1, I've had similar problems attempting to install 2.1.2.

cmb

@gazoo:

Since 2.1.3 came out, has this solved any of these problems for anyone? I was going to go for it, but I don't know if that's a good idea - wanted to see what others have seen for 2.1.3

Nothing related to this in particular has changed. The worst that'll happen is it'll reboot back on the same version so there isn't any harm in trying. The vast majority can upgrade just fine, so there's no reason to not try, as it'll more than likely work for you.

For those who have an ALIX (or anything else with 256 MB RAM and nano), if you're running more than you reasonably should be on a box with 256 MB RAM, disabling some services (especially OpenVPN if you're running multiple instances) before upgrading will allow you to upgrade successfully in an unrelated circumstance to this thread, where it fails because you run out of RAM when trying to upgrade.

dchicks

I'm seeing the exact same reported issues that I've read in this thread…


gpart set -a active -i 1 ad0
gpart: table 'ad0' is corrupt: Operation not permitted

None of the prescribed fixes have done anything to help. I'm running a NetGate with M0n0wall on 256M RAM, of which only 54% is currently in use. I tried turning off a couple of services to see if it would help, but it is no use.

As with others, it looks like it's upgrading and says it will reboot when finished, but it appears to just log my out after a time, and no upgrade occurs. Do we have any clue why this isn't working? Do I need a new CF card?

Thanks,
Dave