2011-08-30

• ​http://fourdollars.blogspot.com/2011/02/4-debian-60.html

ITRI SSTC & DRBL

• 問題一：Ubuntu 10.04, 10.10, 11.04 都沒有 ​partimage 的 amd64 版本

  套件 partimage
      * dapper (admin): backup partitions into a compressed image file [universe]
        0.6.4-14: amd64 i386 powerpc
      * hardy (admin): backup partitions into a compressed image file [universe]
        0.6.7-1: i386
        0.6.4-17: amd64
      * lucid (admin): backup partitions into a compressed image file [universe]
        0.6.8-1: i386
      * maverick (admin): backup partitions into a compressed image file [universe]
        0.6.8-1: i386
      * natty (admin): backup partitions into a compressed image file [universe]
        0.6.8-1: i386
      * oneiric (admin): backup partitions into a compressed image file [universe]
        0.6.8-2build1: amd64 i386
  

• 問題二: 網卡不被支援 - Debian GNU/Linux 6.0.2 (squeeze) 不行，Ubuntu 要 10.10 以後版本的預設核心才行。

  00:19.0 Ethernet controller: Intel Corporation 82579LM Gigabit Network Connection (rev 04)
  00:19.0 0200: 8086:1502 (rev 04)
  

  • 2.6.32-5-amd64 的 e1000e 不支援這個 pciid，所以改裝 3.0.0-1-amd64

    jazz@eagle:/lib/modules/2.6.32-5-amd64$ cat modules.alias | grep 1502
    alias usb:v16CAp1502d*dc*dsc*dp*ic*isc*ip* aircable
    alias pci:v00001131d00007133sv00005169sd00001502bc*sc*i* saa7134
    jazz@eagle:/lib/modules/2.6.32-5-amd64$ cd ..
    jazz@eagle:/lib/modules$ cd 3.0.0-1-amd64/
    jazz@eagle:/lib/modules/3.0.0-1-amd64$ cat modules.alias | grep 1502
    alias pci:v00008086d00001502sv*sd*bc*sc*i* e1000e
    alias usb:v16CAp1502d*dc*dsc*dp*ic*isc*ip* aircable
    alias pci:v00001131d00007133sv00005169sd00001502bc*sc*i* saa7134
    

• 問題三：改裝 3.0.0-1-amd64 會有另外的問題，就是 initrd-pxe.3.0.0-1-amd64.img 無法正常載入 e1000e 的模組，會出現

  "XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX"
  "The driver of network card is NOT found!"
  "Is this kernel $kernel_ver too old so it does not support this network card ?"
  "Without network card driver, we can NOT go on!"
  "XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX"
  "Now enter shell to debug..."
  

• 但是在 shell 中不做任何動作，直接下 exit 之後，又可以正常網路開機。因此懷疑 3.0.0-1-amd64 的開機速度較快，或 modprobe 程序有誤。
• <解法> 目前暫時修改 /usr/lib/mkpxeinitrd-net/initrd-skel/bin/insert-modules 加入 sleep 1 的描述，則可以順利開機。

  --- bin/insert-modules.org	2011-08-30 21:40:18.000000000 +0800
  +++ bin/insert-modules	2011-08-30 21:17:16.000000000 +0800
  @@ -14,6 +14,7 @@
     esac
     echo -n "$module "
     modprobe $module $args 2>/dev/null
  +  sleep 1	
   done
   
   # Detect and load the network devices
  @@ -21,6 +22,7 @@
   [ -n "$NIC_MOD" ] && echo "The detected modules for hardware: $NIC_MOD. Try to load them..."
   for mod in $NIC_MOD; do
     modprobe $mod 2>/dev/null
  +  sleep 1
   done
   netdevices="$(get-nic-devs)"
   if [ -z "$netdevices" ]; then
  

• 看到一些奇怪的訊息，因此懷疑 scan_pci 這段產生的 ＄NIC_MOD 內容有誤～（待追蹤）

  modprobe Valid
  modprobe PCI
  modprobe device
  modprobe module
  modprobe is
  modprobe found!
  

• 問題四：出現 drm 錯誤訊息

  [drm:pch_irq_handler] *ERROR* PCH poison interrupt
  [drm:drm_edid_block_valid] *ERROR* EDID checksum is invalid
  

  • 經查：​Bug #795407 in linux (Ubuntu)與​Bug #712075 in linux (Ubuntu)，錯誤似乎與 DVI 接頭和 X-Windows 解析度偵測有關。目前看到解法是：

    echo 0 > /sys/module/drm_kms_helper/poll
    

• <備註> 應該要關閉 IPv6 支援 - [參考] ​IPv6 addrconf: prefix with wrong length 修改 /etc/sysctl.conf 加入

  net.ipv6.conf.all.accept_ra = 0
  net.ipv6.conf.all.autoconf = 0
  

• <備忘> 要記得裝 sysrqd 好預防重開之需求

  # apt-get install sysrqd
  # echo "mypassword" > /etc/sysrqd.secret && chmod 0600 /etc/sysrqd.secret
  # /opt/drbl/sbin/drbl-cp-host /etc/sysrqd.secret /etc/sysrqd.secret
  

  # /opt/drbl/sbin/drbl-client-service ganglia-monitor on
  # /opt/drbl/sbin/drbl-client-service sysrqd on