** Description changed:

+ 1) modern udevd does not support WAIT_FOR_SYSFS stanza
+ 2) this introduces a sysfs race where devices are attempted to be configured, 
when not yet ready to be online
+ 
+ since sysconfig, s390-tools has gained a modern toolset for activating
+ s390x devices. The combination of tools is: lszdev & chzdev. One to ls
+ devices on z, and another to change devices on z.
+ 
+ the chzdev tool is flexible and can serialize/deserialize current state
+ of devices, into a persistent state by generating appropriate race free
+ udev rules.
+ 
+ realistically sysconfig is currently sub-par solution which is not
+ supportable for the duration of the LTS cycles. Whilst the chzdev
+ tooling has IBM upstream support and is compatible with udev we
+ currently ship and overall is more user-friendly and has better admin
+ user experience, and script-ability.
+ 
+ I'm requesting FFe to:
+ - drop hwup/hwdown support
+ - migrate hwup/hwdown configuration to chzdev udev rules
+ - change installer to generate chzdev udev rules, instead of hwup/hwdown 
configuration
+ 
+ Subject to testing:
+ - full installer testing on all supported target platforms
+ - upgrade testing of all supported target platforms
+ 
+ Failing to land this for xenial, I shall seek to land this as an SRU.
+ 
+ Regards,
+ 
+ Dimitri.
+ 
+ Original bug report below.
+ 
  -- Problem Description --
  Ubuntu 16.04 kernel 4.4.0-13 is installed on LPAR without cio_ignore in 
kernel command line, so a lot of devices are visible:
  
  root@s83lp41:~# ls -1U /sys/bus/ccw/drivers/qeth | wc -l
  154
  (50 OSA triplets + 4 non-devices)
  
  With so many devices the system may boot (initial or reboot) with only
  some or no interfaces brought up. A network is configured only on some
  of the triplets; the issue is well reproduced with two interfaces.
  
  lsqeth shows only active interfaces
  Inactive cannot be brought up with ifup:
  
  root@s83lp41:~# ifup encb080
  Cannot find device "encb080"
  Failed to bring up encb080.
  
  However, they become available after manually issuing hwup:
  
  root@s83lp41:~# hwup -A -D /devices/css0/0.0.1c67/0.0.b080 ccw 0.0.b080
  Configuring device 0.0.b080: ok. (portname OSAPORT) (portno 0) ok.
  
  After hwup interfaces are configured automatically and are listed in
  lsqeth.
  
  root@s83lp41:~# ifup encb080
  ifup: waiting for lock on /run/network/ifstate.encb080
  ifup: interface encb080 already configured

** Changed in: sysconfig (Ubuntu)
   Importance: Undecided => High

** Changed in: s390-sysconfig-writer (Ubuntu)
   Importance: Undecided => High

** Changed in: s390-sysconfig-writer (Ubuntu)
       Status: New => Confirmed

** Changed in: sysconfig (Ubuntu)
       Status: New => Confirmed

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1560545

Title:
  [FFe] hwup/down is racy, migrate to chzdev (previously: Some or all
  OSA devices may be unavailable after boot if too many are visible)

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/s390-sysconfig-writer/+bug/1560545/+subscriptions

-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to