Auto-reboots after 12-STABLE r352091

CyberCr33p

Active Member

Reaction score: 20
Messages: 243

Few hours ago I upgrade FreeBSD-12.0-STABLE (revision 351639) to FreeBSD-12.0-STABLE (revision 352091).

After running the new version for few hours 5 of my servers auto-reboot requiring fsck.

Logs show nothing.

Any idea if a commit between these 2 versions can cause it?

Commits are here:

 

SirDice

Administrator
Staff member
Administrator
Moderator

Reaction score: 8,228
Messages: 31,972

You may want to update regardless. Weird crashes like these don't happen often but -STABLE is a development version. So there's always an inherent risk of instability. Over the past 20 or so years I've been tracking -STABLE I've had this happen a couple of times.

I'll probably update some time this week too. I see it's been marked as 12.1-PRERELEASE now, so 12.1-RELEASE is really just around the corner :D
 
OP
OP
CyberCr33p

CyberCr33p

Active Member

Reaction score: 20
Messages: 243

Yes r352091 is marked as 12.1-PRERELEASE. I hope more people start using it so we can find what the issue is.
 

SirDice

Administrator
Staff member
Administrator
Moderator

Reaction score: 8,228
Messages: 31,972

There's no mention of a MFC (Merge From Current). And if I understand its context correctly it's for the RACK stack. Which isn't the default.
 
OP
OP
CyberCr33p

CyberCr33p

Active Member

Reaction score: 20
Messages: 243

But you are right SACK is not enabled by default. So it's should related to the "random restarts".
 
OP
OP
CyberCr33p

CyberCr33p

Active Member

Reaction score: 20
Messages: 243

I upgrade to FreeBSD-12.1-BETA1 and still have issues.
Here is the bug report:


The backtrace shows "tcp_sack_output".

sysctl net.inet.tcp.sack.enable
net.inet.tcp.sack.enable: 1

I disable it to see what happens.
 

SirDice

Administrator
Staff member
Administrator
Moderator

Reaction score: 8,228
Messages: 31,972

Upgraded everything last weekend, running r352571 now. Still no reboots.

Code:
dice@maelcum:~ % uname -a
FreeBSD maelcum.dicelan.home 12.1-STABLE FreeBSD 12.1-STABLE r352571 MAELCUM  amd64
 

SirDice

Administrator
Staff member
Administrator
Moderator

Reaction score: 8,228
Messages: 31,972

All of them are using an Intel PRO/1000 (I bought a whole stack of these a few years ago) with the em(4) driver.

Only one machine is using net/intel-em-kmod because I had some odd network issues on it (it loses connection at random intervals without any clear reason). Kind of annoying as it happens on my firewall/router that connects my whole network together. Interface just goes down for a second, then comes up. It's not the DHCP MTU issue as I'm using static addresses on it. Netflix or Youtube streams don't even notice so it's not interfering. It's just annoying to see in the logs:
Code:
Sep 22 18:12:12 maelcum miniupnpd[1559]: try_sendto(sock=7, len=467, dest=239.255.255.250:1900): sendto: No buffer space available
Sep 22 18:12:12 maelcum syslogd: sendto: No buffer space available
Sep 22 18:12:12 maelcum syslogd: sendto: No buffer space available
Sep 22 18:12:12 maelcum syslogd: sendto: No buffer space available
Sep 22 18:12:12 maelcum syslogd: sendto: No buffer space available
Sep 22 18:12:12 maelcum syslogd: sendto: No buffer space available
Sep 22 18:12:12 maelcum syslogd: sendto: No buffer space available
Sep 22 18:12:12 maelcum syslogd: sendto: No buffer space available
Sep 22 18:12:12 maelcum syslogd: sendto: No buffer space available
Sep 22 18:12:12 maelcum syslogd: sendto: No buffer space available
Sep 22 18:12:12 maelcum syslogd: sendto: No buffer space available
Sep 22 18:12:12 maelcum syslogd: sendto: No buffer space available
(I'm sending my syslogs to a remote machine, hence the syslog messages)
 
Top