Hi Michael,

Sorry no logfiles as I just reloaded 2.21 from the nightly build. So all history in old partitions is lost. If the upgrade from 2.19 120 has been ok with other machines then my experience may have been a one off. I am not overly confident of the BIOS of this machine I have either.

As for 2.21 - running well CPU usage is as expected. Response is fast no issues. OpenVPN running for road warrior connections and no issues there. If any oddities come up I will report.

The backup issue is tricky. Even if you make it right before the upgrade it would still contain the conf for apache in this case with the php enabled. The upgrade would then remove the php reference but of course if for any reason you needed to restore then it will come back and web GUI would not restart. Because of your release a few upgrades ago I knew php was going so was relatively quick to spot the fix I needed to do. I had enabled the SSH server before web GUI was lost and could fix.

My lesson is backup before upgrade - and then again after upgrade labelling them with the release #. Maybe it is prudent to mention anything in release notes that might affect web GUI.

Regards, Paul

-----Original Message----- From: Michael Tremer michael.tremer@ipfire.org Sent: 28 June 2018 11:41 To: Peter Müller peter.mueller@link38.eu; Paul Titjen paul.titjen@ministc.com Cc: development@lists.ipfire.org Subject: Re: PFire 2.21

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA512

Hello,

On Wed, 2018-06-27 at 23:07 +0200, Peter Müller wrote:

Hello Paul,

this is the right place. Testing reports are always welcome. :-)

A crashing grub does not sound so good - especially if a firewall machine is physical unreachable and no remote KVM console or similar is available.

The 2.12 release was split up in two parts (one includes the new kernel, firmware files are in the second) and as far as I am aware, both should installed before rebooting.

Yes, you should install both in one go and then reboot. Pakfire should not really allow you to only install one.

...

This may be my machine as I had used a live parted in the past to make a properly useful size var partition.

I did not get this.

Changing the partition layout should not create any problems, because...

@Core developers: Are we changing partition layout on existing installations (enlarge /boot, drop /var)?

... once the system is installed we never change any partition layout and basically don't about it.

@Paul: Do you have any log files or so that can be useful to find out what went wrong here?

Yes, restoring too old backup files causes trouble. IPFire creates a new backup before running a core upgrade, but if this fails, it is hard to recover it. So better create a fresh backup before updating to 2.21. Should we include this in the release notes?

Pakfire automatically creates one. But people of course have to download that from the system.

Any observations after running on 2.21 (faster/slower/unstable/...)?

Best, - -Michael

Thanks, and best regards, Peter Müller

...

Dear all,

Hopefully I am in the right place to report on 2.21

Just in case not - a quick summary

Moved to testing from stable 120.

Pakfire upgraded first to 121 - no reboot suggested.

Pakfire upgraded next to 2.21 - This time reboot required.

Reboot failed with a broken grub. I could not fix this with the manual grub commands to get ipfire running at all so could not rewrite grub from a running system.

This may be my machine as I had used a live parted in the past to make a properly useful size var partition.

Bit the bullet and downloaded nightly build of 2.21 and reinstalled a clean ipfire.

So far so good.

Only one comment about restoring previous settings from a backup.

My backup was old and from a time when php was supported. After restoring the backup a surprise was in store the next time apache web interface was restarted. Apache failed as there is a reference in the conf files from backup to load php so extension.

Commenting out the line fixes it of course. Just thought I would mention that restoring backups from a version prior to the latest may have issues as features start to get expired.

By the way I really admire the stuff you do and a donation is en-route.

Regards,

Paul

Hello,

while testing some firewall stuff, I stumbled across bug #11777 (https://bugzilla.ipfire.org/show_bug.cgi?id=11777): In some cases, GeoIP country data in firewall rules and WebUI seem to differ. :-(

Since this makes debugging extremely hard and unreliable, could someone have a look at this please? Sorry for the noise, but this is a nasty one...

Thanks and best regards, Peter Müller

Hello,

just installed IPFire 2.21 - Core Update 122 on a testing machine.

Issues noticed during update: (a) Update to 122 was not installed automatically, but needs user interaction. (b) Machine rebooted properly and came up again without manual action required. (c) WebUI shortly displays "local recursor" for DNS status at the main page - DNSSEC status of nameservers, however, is green. These were displayed correctly again after ~ 2 minutes. (d) NRPE addon required reinstallation (probably due to some configuration changes). The service did not appear in the list at the WebUI; this needs some bugfixing. (e) charon displays connection errors "could not write to socket: operation not permitted" which disappeared after ~ 2 minutes and everything was properly established.

Summary: Reboot, basic functions WORKS Squid web proxy + URL filter WORKS IDS WORKS OpenVPN (N2N only) WORKS IPsec (N2N only) WORKS SSH WORKS QoS WORKS NRPE WORKS (after reinstallation, some bugs left)

CPU load (especially when it comes to HW interrupts) is a bit (but not significant) lower than it was while running C120. RAM consumption stays at the same level. Entropy is ~ 400 bits higher. Kernel reports two interesting log lines on boot:

19:02:35 kernel: alg: No test for seqiv(rfc4106(gcm(aes))) (seqiv(rfc4106-gcm-aesni))

18:57:49 kernel: xt_geoip: loading out-of-tree module taints kernel.

Just for the records. :-)

Systems seems to be safe against Spectre/Meltdown:

/sys/devices/system/cpu/vulnerabilities/meltdown: Mitigation: PTI /sys/devices/system/cpu/vulnerabilities/spec_store_bypass: Not affected /sys/devices/system/cpu/vulnerabilities/spectre_v1: Mitigation: __user pointer sanitization /sys/devices/system/cpu/vulnerabilities/spectre_v2: Mitigation: Full generic retpoline

In case any issues occur within the next time, I'll let you know. Excellent work so far!

Thanks, and best regards, Peter Müller

Hello Jörg,

with which kernel did you test this?

-Michael

On Sat, 2018-06-30 at 17:53 +0200, Jörg Thalau wrote:

Hello

I'm new here. I'm not sure if that's the right mailing list. Sometimes the ipfire hangs up. I researched the problem. The reason is the Intel network card driver. Could you implement a new version of Intel network card driver?

Thanks!

here the error message:

Jun 27 22:41:33 ipfire kernel: e1000e 0000:00:1f.6 green0: Detected Hardware Unit Hang: Jun 27 22:41:33 ipfire kernel: TDH <6a> Jun 27 22:41:33 ipfire kernel: TDT <7b> Jun 27 22:41:33 ipfire kernel: next_to_use <7b> Jun 27 22:41:33 ipfire kernel: next_to_clean <5f> Jun 27 22:41:33 ipfire kernel: buffer_info[next_to_clean]: Jun 27 22:41:33 ipfire kernel: time_stamp <117031b03> Jun 27 22:41:33 ipfire kernel: next_to_watch <6a> Jun 27 22:41:33 ipfire kernel: jiffies <117032408> Jun 27 22:41:33 ipfire kernel: next_to_watch.status <0> Jun 27 22:41:33 ipfire kernel: MAC Status <40080083> Jun 27 22:41:33 ipfire kernel: PHY Status <796d> Jun 27 22:41:33 ipfire kernel: PHY 1000BASE-T Status <3c00> Jun 27 22:41:33 ipfire kernel: PHY Extended Status <3000> Jun 27 22:41:33 ipfire kernel: PCI Status <10> Jun 27 22:41:35 ipfire kernel: e1000e 0000:00:1f.6 green0: Reset adapter unexpectedly Jun 27 22:41:37 ipfire ntpd[4079]: Deleting interface #10 green0, 192.168.144.1#123, interface stats: received=117, sent=117, dropped=0, active_time=1126210 secs Jun 27 22:41:39 ipfire kernel: e1000e: green0 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: Rx/Tx

...

Am 29.06.18 um 22:55 schrieb Peter Müller: Hello,

while testing some firewall stuff, I stumbled across bug #11777 (https://bugzilla.ipfire.org/show_bug.cgi?id=11777): In some cases, GeoIP country data in firewall rules and WebUI seem to differ. :-(

Since this makes debugging extremely hard and unreliable, could someone have a look at this please? Sorry for the noise, but this is a nasty one...

Thanks and best regards, Peter Müller

...

Hello,

just installed IPFire 2.21 - Core Update 122 on a testing machine.

Issues noticed during update: (a) Update to 122 was not installed automatically, but needs user interaction. (b) Machine rebooted properly and came up again without manual action required. (c) WebUI shortly displays "local recursor" for DNS status at the main page - DNSSEC status of nameservers, however, is green. These were displayed correctly again after ~ 2 minutes. (d) NRPE addon required reinstallation (probably due to some configuration changes). The service did not appear in the list at the WebUI; this needs some bugfixing. (e) charon displays connection errors "could not write to socket: operation not permitted" which disappeared after ~ 2 minutes and everything was properly established.

Summary: Reboot, basic functions WORKS Squid web proxy + URL filter WORKS IDS WORKS OpenVPN (N2N only) WORKS IPsec (N2N only) WORKS SSH WORKS QoS WORKS NRPE WORKS (after reinstallation, some bugs left)

CPU load (especially when it comes to HW interrupts) is a bit (but not significant) lower than it was while running C120. RAM consumption stays at the same level. Entropy is ~ 400 bits higher. Kernel reports two interesting log lines on boot:

19:02:35 kernel: alg: No test for seqiv(rfc4106(gcm(aes))) (seqiv(rfc4106-gcm-aesni))

18:57:49 kernel: xt_geoip: loading out-of-tree module taints kernel.

Just for the records. :-)

Systems seems to be safe against Spectre/Meltdown:

/sys/devices/system/cpu/vulnerabilities/meltdown: Mitigation: PTI /sys/devices/system/cpu/vulnerabilities/spec_store_bypass: Not affected /sys/devices/system/cpu/vulnerabilities/spectre_v1: Mitigation: __user pointer sanitization /sys/devices/system/cpu/vulnerabilities/spectre_v2: Mitigation: Full generic retpoline

In case any issues occur within the next time, I'll let you know. Excellent work so far!

Thanks, and best regards, Peter Müller

Hello,

could you install the latest testing release please?

https://planet.ipfire.org/post/ipfire-2-21-core-update-121-is-available-for-...

Best, -Michael

On Sat, 2018-06-30 at 18:05 +0200, Jörg Thalau wrote:

Hello Michael uname -r 3.14.79-ipfire IPFire 2.19 (x86_64) - Core Update 120 Should be last image version of ipfire. Jörg

Am 30.06.18 um 17:54 schrieb Michael Tremer:

...

Hello Jörg,

with which kernel did you test this?

-Michael

On Sat, 2018-06-30 at 17:53 +0200, Jörg Thalau wrote:

...

Hello

I'm new here. I'm not sure if that's the right mailing list. Sometimes the ipfire hangs up. I researched the problem. The reason is the Intel network card driver. Could you implement a new version of Intel network card driver?

Thanks!

here the error message:

Jun 27 22:41:33 ipfire kernel: e1000e 0000:00:1f.6 green0: Detected Hardware Unit Hang: Jun 27 22:41:33 ipfire kernel: TDH <6a> Jun 27 22:41:33 ipfire kernel: TDT <7b> Jun 27 22:41:33 ipfire kernel: next_to_use <7b> Jun 27 22:41:33 ipfire kernel: next_to_clean <5f> Jun 27 22:41:33 ipfire kernel: buffer_info[next_to_clean]: Jun 27 22:41:33 ipfire kernel: time_stamp <117031b03> Jun 27 22:41:33 ipfire kernel: next_to_watch <6a> Jun 27 22:41:33 ipfire kernel: jiffies <117032408> Jun 27 22:41:33 ipfire kernel: next_to_watch.status <0> Jun 27 22:41:33 ipfire kernel: MAC Status <40080083> Jun 27 22:41:33 ipfire kernel: PHY Status <796d> Jun 27 22:41:33 ipfire kernel: PHY 1000BASE-T Status <3c00> Jun 27 22:41:33 ipfire kernel: PHY Extended Status <3000> Jun 27 22:41:33 ipfire kernel: PCI Status <10> Jun 27 22:41:35 ipfire kernel: e1000e 0000:00:1f.6 green0: Reset adapter unexpectedly Jun 27 22:41:37 ipfire ntpd[4079]: Deleting interface #10 green0, 192.168.144.1#123, interface stats: received=117, sent=117, dropped=0, active_time=1126210 secs Jun 27 22:41:39 ipfire kernel: e1000e: green0 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: Rx/Tx

...

Am 29.06.18 um 22:55 schrieb Peter Müller: Hello,

while testing some firewall stuff, I stumbled across bug #11777 (https://bugzilla.ipfire.org/show_bug.cgi?id=11777): In some cases, GeoIP country data in firewall rules and WebUI seem to differ. :-(

Since this makes debugging extremely hard and unreliable, could someone have a look at this please? Sorry for the noise, but this is a nasty one...

Thanks and best regards, Peter Müller

...

Hello,

just installed IPFire 2.21 - Core Update 122 on a testing machine.

Issues noticed during update: (a) Update to 122 was not installed automatically, but needs user interaction. (b) Machine rebooted properly and came up again without manual action required. (c) WebUI shortly displays "local recursor" for DNS status at the main page - DNSSEC status of nameservers, however, is green. These were displayed correctly again after ~ 2 minutes. (d) NRPE addon required reinstallation (probably due to some configuration changes). The service did not appear in the list at the WebUI; this needs some bugfixing. (e) charon displays connection errors "could not write to socket: operation not permitted" which disappeared after ~ 2 minutes and everything was properly established.

Summary: Reboot, basic functions WORKS Squid web proxy + URL filter WORKS IDS WORKS OpenVPN (N2N only) WORKS IPsec (N2N only) WORKS SSH WORKS QoS WORKS NRPE WORKS (after reinstallation, some bugs left)

CPU load (especially when it comes to HW interrupts) is a bit (but not significant) lower than it was while running C120. RAM consumption stays at the same level. Entropy is ~ 400 bits higher. Kernel reports two interesting log lines on boot:

19:02:35 kernel: alg: No test for seqiv(rfc4106(gcm(aes))) (seqiv(rfc4106-gcm-aesni))

18:57:49 kernel: xt_geoip: loading out-of-tree module taints kernel.

Just for the records. :-)

Systems seems to be safe against Spectre/Meltdown:

/sys/devices/system/cpu/vulnerabilities/meltdown: Mitigation: PTI /sys/devices/system/cpu/vulnerabilities/spec_store_bypass: Not affected /sys/devices/system/cpu/vulnerabilities/spectre_v1: Mitigation: __user pointer sanitization /sys/devices/system/cpu/vulnerabilities/spectre_v2: Mitigation: Full generic retpoline

In case any issues occur within the next time, I'll let you know. Excellent work so far!

Thanks, and best regards, Peter Müller