add Kernel patch to allow netpoll while veth is in use

[pve-kernel-3.10.0.git] / README

KERNEL SOURCE:
==============

We use the RHEL7 kernel sources, available from:

ftp://ftp.redhat.com/redhat/rhel/


Additional/Updated Modules:
---------------------------

- include latest ixgbe driver from intel/sourceforge

- include latest igb driver from intel/sourceforge

- include latest e1000e driver from intel/sourceforge

# this driver does not compile with RHEL7 kernel
#- include latest Broadcom bnx2 drivers
#
#  * original file linux-7.8.56.zip contains
#    netxtreme2-7.8.56.tar.gz (added to repository)

- include latest Adaptec aacraid driver

  * http://download.adaptec.com/raid/aac/linux/aacraid-linux-src-1.2.1-40300.tgz

- include latest ARECA RAID drivers

  * http://www.areca.com.tw/support/s_linux/linux.htm

FIRMWARE:
=========

We create our own firmware package, which includes the firmware for
all proxmox-ve kernels. So far this include

pve-kernel-2.6.18
pve-kernel-2.6.24
pve-kernel-2.6.32
pve-kernel-2.6.35
pve-kernel-3.10.0

We use 'find-firmware.pl' to extract lists of required firmeware
files.  The script 'assemble-firmware.pl' is used to read those lists
and copy the files from various source directory into a target
directory.

We do not include firmeware for some wireless HW when there is a
separate debian package for that, for example:

zd1211-firmware
atmel-firmware
bluez-firmware 


PATCHES:
--------

 bridge-patch.diff: Avoid bridge problems with changing MAC
  see also: http://forum.openvz.org/index.php?t=msg&th=5291

  Behaviour after 2.6.27 has changed slighly - after setting mac address
  of bridge device, then address won't change. So we could omit
  that patch, requiring to set hwaddress in /etc/network/interfaces.


Additional information
----------------------

We use the default configuration provided by RHEL, and apply
the following modification:

see config-3.10.0.diff

- enable common CONFIG_BLK_DEV_XXX to avoid hardware detection
  problems (udev, undate-initramfs hav serious problems without that)

  	 CONFIG_BLK_DEV_SD=y
  	 CONFIG_BLK_DEV_SR=y
  	 CONFIG_BLK_DEV_DM=y

- compile NBD and RBD modules
	 CONFIG_BLK_DEV_NBD=m
	 CONFIG_BLK_DEV_RBD=m

- set LOOP_MIN_COUNT to 8 (debian defaults)
	 CONFIG_BLK_DEV_LOOP_MIN_COUNT=8

- disable module signatures (CONFIG_MODULE_SIG)
 
- enable IBM JFS file system 

  This is disabled in RHEL kernel for no real reason, so we enable
  it as requested by users (bug #64)

- enable apple HFS and HFSPLUS

  This is disabled in RHEL kernel for no real reason, so we enable
  it as requested by users

- enable CONFIG_BCACHE=m (requested by user)

- enable CONFIG_BRIDGE=y

  Else we get warnings on boot, that
  net.bridge.bridge-nf-call-iptables is an unknown key

- disable CONFIG_DEFAULT_SECURITY_SELINUX

  Use same SELINUX config as debian kernels (CONFIG_DEFAULT_SECURITY_DAC=y)

- set CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE=y

  because if not set, it can give some dynamic memory or cpu frequencies 
  change, and vms can crash (mainly windows guest).

  see http://forum.proxmox.com/threads/18238-Windows-7-x64-VMs-crashing-randomly-during-process-termination?p=93273#post93273

- set CONFIG_CC_STACKPROTECTOR_REGULAR=y

  gcc on debian wheezy does not supportCONFIG_CC_STACKPROTECTOR_STRONG
 
# Note: enable now for testing
#- disable CONFIG_BRIDGE_IGMP_SNOOPING 
#
# To avoid problems with multicasting. Various users reported
#  multipath problems, which can be avoided by turning off IGMP
# snooping.

# deadline is default anyways now
#- use 'deadline' as default scheduler
#
#  This is the suggested setting for KVM. We also measure bad fsync
#  performance with ext4 and cfq.

Testing final kernel with kvm
-----------------------------

kvm -kernel data/boot/vmlinuz-3.10.0-2-pve -initrd initrd.img-3.10.0-2-pve -append "vga=791 video=vesafb:ywrap,mtrr" /dev/zero