+#!/usr/bin/python2
+
+# Copyright (c) 2003 Intel Corporation
+# All rights reserved.
+#
+# Copyright (c) 2004-2006 The Trustees of Princeton University
+# All rights reserved.
+
+
import string
+import re
+import os
+import UpdateBootStateWithPLC
from Exceptions import *
import utils
import compatibility
-from systeminfo import systeminfo
+import systeminfo
import BootAPI
+import notify_messages
+import ModelOptions
def Run( vars, log ):
"""
Load the kernel off of a node and boot to it.
This step assumes the disks are mounted on SYSIMG_PATH.
+ If successful, this function will not return. If it returns, no chain
+ booting has occurred.
Expect the following variables:
BOOT_CD_VERSION A tuple of the current bootcd version
SYSIMG_PATH the path where the system image will be mounted
(always starts with TEMP_PATH)
ROOT_MOUNTED the node root file system is mounted
-
+ NODE_SESSION the unique session val set when we requested
+ the current boot state
+ PLCONF_DIR The directory to store PL configuration files in
+
Sets the following variables:
ROOT_MOUNTED the node root file system is mounted
"""
SYSIMG_PATH= vars["SYSIMG_PATH"]
if SYSIMG_PATH == "":
raise ValueError, "SYSIMG_PATH"
-
+
+ PLCONF_DIR= vars["PLCONF_DIR"]
+ if PLCONF_DIR == "":
+ raise ValueError, "PLCONF_DIR"
+
+ # its ok if this is blank
+ NODE_SESSION= vars["NODE_SESSION"]
+
+ NODE_MODEL_OPTIONS= vars["NODE_MODEL_OPTIONS"]
+
+ PARTITIONS= vars["PARTITIONS"]
+ if PARTITIONS == None:
+ raise ValueError, "PARTITIONS"
+
except KeyError, var:
raise BootManagerException, "Missing variable in vars: %s\n" % var
except ValueError, var:
# simply creating an instance of this class and listing the system
# block devices will make them show up so vgscan can find the planetlab
# volume group
- systeminfo().get_block_device_list()
+ systeminfo.get_block_device_list(vars, log)
utils.sysexec( "vgscan", log )
utils.sysexec( "vgchange -ay planetlab", log )
utils.makedirs( SYSIMG_PATH )
- utils.sysexec( "mount /dev/planetlab/root %s" % SYSIMG_PATH, log )
- utils.sysexec( "mount /dev/planetlab/vservers %s/vservers" %
- SYSIMG_PATH, log )
+ cmd = "mount %s %s" % (PARTITIONS["root"],SYSIMG_PATH)
+ utils.sysexec( cmd, log )
+ cmd = "mount %s %s/vservers" % (PARTITIONS["vservers"],SYSIMG_PATH)
+ utils.sysexec( cmd, log )
+ cmd = "mount -t proc none %s/proc" % SYSIMG_PATH
+ utils.sysexec( cmd, log )
ROOT_MOUNTED= 1
vars['ROOT_MOUNTED']= 1
- node_update_cmd= "/usr/local/planetlab/bin/NodeUpdate.py start noreboot"
+ # write out the session value /etc/planetlab/session
+ try:
+ session_file_path= "%s/%s/session" % (SYSIMG_PATH,PLCONF_DIR)
+ session_file= file( session_file_path, "w" )
+ session_file.write( str(NODE_SESSION) )
+ session_file.close()
+ session_file= None
+ log.write( "Updated /etc/planetlab/session\n" )
+ except IOError, e:
+ log.write( "Unable to write out /etc/planetlab/session, continuing anyway\n" )
+
+ # update configuration files
+ log.write( "Updating configuration files.\n" )
+ if os.path.exists( SYSIMG_PATH + "/etc/init.d/conf_files" ):
+ cmd = "/etc/init.d/conf_files start --noscripts"
+ else:
+ cmd = "/usr/bin/PlanetLabConf.py noscripts"
+ utils.sysexec( "chroot %s %s" % (SYSIMG_PATH, cmd), log )
+ # update node packages
log.write( "Running node update.\n" )
- utils.sysexec( "chroot %s %s" % (SYSIMG_PATH,node_update_cmd), log )
+ cmd = "chroot %s /usr/bin/NodeUpdate.py start noreboot" \
+ % SYSIMG_PATH
+ utils.sysexec( cmd, log )
log.write( "Updating ssh public host key with PLC.\n" )
ssh_host_key= ""
ssh_host_key_file= None
except IOError, e:
pass
-
+
update_vals= {}
update_vals['ssh_host_key']= ssh_host_key
BootAPI.call_api_function( vars, "BootUpdateNode", (update_vals,) )
+ # get the kernel version
+ option = ''
+ if NODE_MODEL_OPTIONS & ModelOptions.SMP:
+ option = 'smp'
log.write( "Copying kernel and initrd for booting.\n" )
- utils.sysexec( "cp %s/boot/kernel-boot /tmp/kernel" % SYSIMG_PATH, log )
- utils.sysexec( "cp %s/boot/initrd-boot /tmp/initrd" % SYSIMG_PATH, log )
+ utils.sysexec( "cp %s/boot/kernel-boot%s /tmp/kernel" % (SYSIMG_PATH,option), log )
+ utils.sysexec( "cp %s/boot/initrd-boot%s /tmp/initrd" % (SYSIMG_PATH,option), log )
+
+ BootAPI.save(vars)
log.write( "Unmounting disks.\n" )
- utils.sysexec_noerr( "chroot %s umount /rcfs" % SYSIMG_PATH, log )
- utils.sysexec_noerr( "umount -r /dev/planetlab/vservers", log )
- utils.sysexec_noerr( "umount -r /dev/planetlab/root", log )
+ try:
+ # backwards compat, though, we should never hit this case post PL 3.2
+ os.stat("%s/rcfs/taskclass"%SYSIMG_PATH)
+ utils.sysexec_noerr( "chroot %s umount /rcfs" % SYSIMG_PATH, log )
+ except OSError, e:
+ pass
+
+ utils.sysexec_noerr( "umount %s/proc" % SYSIMG_PATH, log )
+ utils.sysexec_noerr( "umount -r %s/vservers" % SYSIMG_PATH, log )
+ utils.sysexec_noerr( "umount -r %s" % SYSIMG_PATH, log )
utils.sysexec_noerr( "vgchange -an", log )
ROOT_MOUNTED= 0
vars['ROOT_MOUNTED']= 0
- if BOOT_CD_VERSION[0] == 2:
- log.write( "Unloading modules and chaining booting to new kernel.\n" )
- else:
- log.write( "Chaining booting to new kernel.\n" )
+ log.write( "Unloading modules and chain booting to new kernel.\n" )
# further use of log after Upload will only output to screen
log.Upload()
- # on 2.x cds (2.4 kernel) for sure, we need to shutdown everything to
- # get kexec to work correctly
+ # regardless of whether kexec works or not, we need to stop trying to
+ # run anything
+ cancel_boot_flag= "/tmp/CANCEL_BOOT"
+ utils.sysexec( "touch %s" % cancel_boot_flag, log )
+
+ # on 2.x cds (2.4 kernel) for sure, we need to shutdown everything
+ # to get kexec to work correctly. Even on 3.x cds (2.6 kernel),
+ # there are a few buggy drivers that don't disable their hardware
+ # correctly unless they are first unloaded.
utils.sysexec_noerr( "ifconfig eth0 down", log )
for line in modules:
module= string.strip(line)
if module != "":
+ log.write( "Unloading %s\n" % module )
utils.sysexec_noerr( "modprobe -r %s" % module, log )
+
+ modules.close()
except IOError:
- log.write( "Couldn't load /tmp/loadedmodules to unload, continuing.\n" )
-
- utils.sysexec( "kexec --force --initrd=/tmp/initrd " \
- "--append=ramdisk_size=8192 /tmp/kernel" )
+ log.write( "Couldn't read /tmp/loadedmodules, continuing.\n" )
+
+ try:
+ modules= file("/proc/modules", "r")
+
+ # Get usage count for USB
+ usb_usage = 0
+ for line in modules:
+ try:
+ # Module Size UsageCount UsedBy State LoadAddress
+ parts= string.split(line)
+
+ if parts[0] == "usb_storage":
+ usb_usage += int(parts[2])
+ except IndexError, e:
+ log.write( "Couldn't parse /proc/modules, continuing.\n" )
+
+ modules.seek(0)
+
+ for line in modules:
+ try:
+ # Module Size UsageCount UsedBy State LoadAddress
+ parts= string.split(line)
+
+ # While we would like to remove all "unused" modules,
+ # you can't trust usage count, especially for things
+ # like network drivers or RAID array drivers. Just try
+ # and unload a few specific modules that we know cause
+ # problems during chain boot, such as USB host
+ # controller drivers (HCDs) (PL6577).
+ # if int(parts[2]) == 0:
+ if False and re.search('_hcd$', parts[0]):
+ if usb_usage > 0:
+ log.write( "NOT unloading %s since USB may be in use\n" % parts[0] )
+ else:
+ log.write( "Unloading %s\n" % parts[0] )
+ utils.sysexec_noerr( "modprobe -r %s" % parts[0], log )
+ except IndexError, e:
+ log.write( "Couldn't parse /proc/modules, continuing.\n" )
+ except IOError:
+ log.write( "Couldn't read /proc/modules, continuing.\n" )
+
+
+ kargs = "root=%s ramdisk_size=8192" % PARTITIONS["mapper-root"]
+ if NODE_MODEL_OPTIONS & ModelOptions.SMP:
+ kargs = kargs + " " + "acpi=off"
+ try:
+ kargsfb = open("/kargs.txt","r")
+ moreargs = kargsfb.readline()
+ kargsfb.close()
+ moreargs = moreargs.strip()
+ log.write( 'Parsed in "%s" kexec args from /kargs.txt\n' % moreargs )
+ kargs = kargs + " " + moreargs
+ except IOError:
+ # /kargs.txt does not exist, which is fine. Just kexec with default
+ # kargs, which is ramdisk_size=8192
+ pass
+
+ try:
+ utils.sysexec( 'kexec --force --initrd=/tmp/initrd ' \
+ '--append="%s" /tmp/kernel' % kargs)
+ except BootManagerException, e:
+ # if kexec fails, we've shut the machine down to a point where nothing
+ # can run usefully anymore (network down, all modules unloaded, file
+ # systems unmounted. write out the error, and cancel the boot process
+
+ log.write( "\n\n" )
+ log.write( "-------------------------------------------------------\n" )
+ log.write( "kexec failed with the following error. Please report\n" )
+ log.write( "this problem to support@planet-lab.org.\n\n" )
+ log.write( str(e) + "\n\n" )
+ log.write( "The boot process has been canceled.\n" )
+ log.write( "-------------------------------------------------------\n\n" )
return