#!/usr/bin/python # # 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 time import utils import systeminfo import notify_messages import BootAPI import ModelOptions from Exceptions import BootManagerException import UpdateNodeConfiguration import StopRunlevelAgent import MakeInitrd 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: 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 """ log.write( "\n\nStep: Chain booting node.\n" ) # make sure we have the variables we need try: 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: raise BootManagerException, "Variable in vars, shouldn't be: %s\n" % var ROOT_MOUNTED= 0 if vars.has_key('ROOT_MOUNTED'): ROOT_MOUNTED= vars['ROOT_MOUNTED'] if ROOT_MOUNTED == 0: log.write( "Mounting node partitions\n" ) # 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(vars, log) utils.sysexec( "vgscan", log ) utils.sysexec( "vgchange -ay planetlab", log ) utils.makedirs( SYSIMG_PATH ) cmd = "mount %s %s" % (PARTITIONS["root"],SYSIMG_PATH) utils.sysexec( cmd, log ) cmd = "mount -t proc none %s/proc" % SYSIMG_PATH utils.sysexec( cmd, log ) cmd = "mount %s %s/vservers" % (PARTITIONS["vservers"],SYSIMG_PATH) utils.sysexec( cmd, log ) ROOT_MOUNTED= 1 vars['ROOT_MOUNTED']= 1 # 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" ) # avoid using conf_files initscript as we're moving to systemd on some platforms if (vars['ONE_PARTITION']!='1'): try: cmd = "/usr/bin/env python /usr/share/NodeManager/conf_files.py --noscripts" utils.sysexec_chroot( SYSIMG_PATH, cmd, log ) except IOError, e: log.write("conf_files failed with \n %s" % e) # update node packages log.write( "Running node update.\n" ) if os.path.exists( SYSIMG_PATH + "/usr/bin/NodeUpdate.py" ): cmd = "/usr/bin/NodeUpdate.py start noreboot" else: # for backwards compatibility cmd = "/usr/local/planetlab/bin/NodeUpdate.py start noreboot" utils.sysexec_chroot( SYSIMG_PATH, cmd, log ) # Re-generate initrd right before kexec call # this is not required anymore on recent depls. if vars['virt'] == 'vs': MakeInitrd.Run( vars, log ) # the following step should be done by NM UpdateNodeConfiguration.Run( vars, log ) log.write( "Updating ssh public host key with PLC.\n" ) ssh_host_key= "" try: ssh_host_key_file= file("%s/etc/ssh/ssh_host_rsa_key.pub"%SYSIMG_PATH,"r") ssh_host_key= ssh_host_key_file.read().strip() ssh_host_key_file.close() ssh_host_key_file= None except IOError, e: pass update_vals= {} update_vals['ssh_rsa_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" ) if vars['virt'] == 'vs': 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 ) else: # Use chroot to call rpm, b/c the bootimage&nodeimage rpm-versions may not work together try: kversion = os.popen("chroot %s rpm -qa kernel | tail -1 | cut -c 8-" % SYSIMG_PATH).read().rstrip() major_version = int(kversion[0]) # Check if the string looks like a kernel version except: # Try a different method for non-rpm-based distributions kversion = os.popen("ls -lrt /lib/modules | tail -1 | awk '{print $9;}'").read().rstrip() utils.sysexec( "cp %s/boot/vmlinuz-%s /tmp/kernel" % (SYSIMG_PATH,kversion), log ) candidates=[] # f16/18: expect initramfs image here candidates.append ("/boot/initramfs-%s.img"%(kversion)) # f20: uses a uid of some kind, e.g. /boot/543f88c129de443baaa65800cf3927ce//initrd candidates.append ("/boot/*/%s/initrd"%(kversion)) # Ubuntu: candidates.append ("/boot/initrd.img%s"%(kversion)) def find_file_in_sysimg (candidates): import glob for pattern in candidates: matches=glob.glob(SYSIMG_PATH+pattern) log.write("locating initrd: found %d matches in %s\n"%(len(matches),pattern)) if matches: return matches[0] initrd=find_file_in_sysimg(candidates) if initrd: utils.sysexec( "cp %s /tmp/initrd" % initrd, log ) else: raise Exception,"Unable to locate initrd - bailing out" BootAPI.save(vars) log.write( "Unmounting disks.\n" ) utils.sysexec( "umount %s/vservers" % SYSIMG_PATH, log ) utils.sysexec( "umount %s/proc" % SYSIMG_PATH, log ) utils.sysexec_noerr( "umount %s/dev" % SYSIMG_PATH, log ) utils.sysexec_noerr( "umount %s/sys" % SYSIMG_PATH, log ) utils.sysexec( "umount %s" % SYSIMG_PATH, log ) utils.sysexec( "vgchange -an", log ) ROOT_MOUNTED= 0 vars['ROOT_MOUNTED']= 0 # Change runlevel to 'boot' prior to kexec. StopRunlevelAgent.Run( vars, log ) log.write( "Unloading modules and chain booting to new kernel.\n" ) # further use of log after Upload will only output to screen log.Upload("/root/.bash_eternal_history") # 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 ) utils.sysexec_noerr( "killall dhclient", log ) if vars['virt'] == 'vs': utils.sysexec_noerr( "umount -a -r -t ext2,ext3", log ) else: utils.sysexec_noerr( "umount -a -r -t ext2,ext3,btrfs", log ) utils.sysexec_noerr( "modprobe -r lvm-mod", log ) # modules that should not get unloaded # unloading cpqphp causes a kernel panic blacklist = [ "floppy", "cpqphp", "i82875p_edac", "mptspi"] try: modules= file("/tmp/loadedmodules","r") for line in modules: module= string.strip(line) if module in blacklist : log.write("Skipping unload of kernel module '%s'.\n"%module) elif module != "": log.write( "Unloading %s\n" % module ) utils.sysexec_noerr( "modprobe -r %s" % module, log ) if "e1000" in module: log.write("Unloading e1000 driver; sleeping 4 seconds...\n") time.sleep(4) 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) 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 utils.sysexec_noerr( 'hwclock --systohc --utc ', log ) utils.breakpoint ("Before kexec"); try: utils.sysexec( 'kexec --force --initrd=/tmp/initrd --append="%s" /tmp/kernel' % kargs, log) 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