import string
import re
+import os
+import InstallWriteConfig
+import UpdateBootStateWithPLC
from Exceptions import *
import utils
import compatibility
from systeminfo import systeminfo
import BootAPI
+import notify_messages
+
+from GetAndUpdateNodeDetails import SMP_OPT
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"]
+
except KeyError, var:
raise BootManagerException, "Missing variable in vars: %s\n" % var
except ValueError, var:
ssh_host_key_file= None
except IOError, e:
pass
-
+
+ # 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_vals= {}
update_vals['ssh_host_key']= ssh_host_key
BootAPI.call_api_function( vars, "BootUpdateNode", (update_vals,) )
+ # rewrite modprobe.conf in case there were any module changes
+ # from a new kernel installed.
+ log.write( "Rewriting /etc/modprobe.conf\n" )
+ (network_count,storage_count)= \
+ InstallWriteConfig.write_modprobeconf_file( vars, log )
+
+ # get the kernel version
+ option = ''
+ if NODE_MODEL_OPTIONS & SMP_OPT:
+ 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 )
log.write( "Unmounting disks.\n" )
- utils.sysexec_noerr( "chroot %s umount /rcfs" % SYSIMG_PATH, 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 -r /dev/planetlab/vservers", log )
utils.sysexec_noerr( "umount -r /dev/planetlab/root", log )
utils.sysexec_noerr( "vgchange -an", log )
ROOT_MOUNTED= 0
vars['ROOT_MOUNTED']= 0
- log.write( "Unloading modules and chaining booting to new kernel.\n" )
+ # before we do the real kexec, check to see if we had any
+ # network drivers written to modprobe.conf. if not, return -1,
+ # which will cause this node to be switched to a debug state.
+ if network_count == 0:
+ log.write( "\nIt appears we don't have any network drivers. Aborting.\n" )
+
+ vars['BOOT_STATE']= 'dbg'
+ vars['STATE_CHANGE_NOTIFY']= 1
+ vars['STATE_CHANGE_NOTIFY_MESSAGE']= \
+ notify_messages.MSG_NO_DETECTED_NETWORK
+ UpdateBootStateWithPLC.Run( vars, log )
+
+ return
+
+ log.write( "Unloading modules and chain booting to new kernel.\n" )
# further use of log after Upload will only output to screen
log.Upload()
utils.sysexec_noerr( "modprobe -r %s" % module, log )
modules.close()
+ except IOError:
+ 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)
- # You can't trust usage count, especially for things
+ 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 re.search('_hcd$', parts[0]):
- log.write( "Unloading %s\n" % parts[0] )
- utils.sysexec_noerr( "modprobe -r %s" % parts[0], log )
+ 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 load /tmp/loadedmodules to unload, continuing.\n" )
+ log.write( "Couldn't read /proc/modules, continuing.\n" )
+
+
+ kargs = "ramdisk_size=8192"
+ 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=ramdisk_size=8192 /tmp/kernel" )
+ 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