X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=source%2Fsteps%2FChainBootNode.py;h=5dec7395ef5ba8373f673cdf0d3f0f6d56338f01;hb=04f40909223edfa75aad8d7baa45925aee7ae812;hp=e2d2e8944e892cb6b03bf1b24967d6e13d01c73e;hpb=bb5fbc8ba39ef808bff8f2ec5c2eda7c11b5304e;p=bootmanager.git diff --git a/source/steps/ChainBootNode.py b/source/steps/ChainBootNode.py index e2d2e89..5dec739 100644 --- a/source/steps/ChainBootNode.py +++ b/source/steps/ChainBootNode.py @@ -1,17 +1,27 @@ +#!/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 InstallWriteConfig -import UpdateBootStateWithPLC -from Exceptions import * import utils -import compatibility -from systeminfo import systeminfo -import BootAPI +import systeminfo import notify_messages - +import BootAPI import ModelOptions +from Exceptions import BootManagerException + +import UpdateNodeConfiguration +import StopRunlevelAgent +import MakeInitrd def Run( vars, log ): """ @@ -21,7 +31,6 @@ def Run( vars, log ): 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 @@ -37,10 +46,6 @@ def Run( vars, log ): # make sure we have the variables we need try: - BOOT_CD_VERSION= vars["BOOT_CD_VERSION"] - if BOOT_CD_VERSION == "": - raise ValueError, "BOOT_CD_VERSION" - SYSIMG_PATH= vars["SYSIMG_PATH"] if SYSIMG_PATH == "": raise ValueError, "SYSIMG_PATH" @@ -54,45 +59,81 @@ def Run( vars, log ): 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 'ROOT_MOUNTED' in vars.keys(): + if vars.has_key('ROOT_MOUNTED'): ROOT_MOUNTED= vars['ROOT_MOUNTED'] if ROOT_MOUNTED == 0: log.write( "Mounting node partitions\n" ) - # old cds need extra utilities to run lvm - if BOOT_CD_VERSION[0] == 2: - compatibility.setup_lvm_2x_cd( vars, log ) - # 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 ) - utils.sysexec( "mount -t proc none %s/proc" % SYSIMG_PATH, log ) + 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 - 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" ) - log.write( "Running node update.\n" ) - utils.sysexec( "chroot %s %s" % (SYSIMG_PATH,node_update_cmd), log ) + # 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= "" @@ -104,26 +145,10 @@ def Run( vars, log ): 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 + update_vals['ssh_rsa_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 = '' @@ -131,42 +156,58 @@ def Run( vars, log ): option = 'smp' log.write( "Copying kernel and initrd for booting.\n" ) - 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 ) + 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" ) - 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 /dev/planetlab/vservers", log ) - utils.sysexec_noerr( "umount -r /dev/planetlab/root", log ) - utils.sysexec_noerr( "vgchange -an", log ) + 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 - # 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 + # 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() + log.Upload("/root/.bash_eternal_history") # regardless of whether kexec works or not, we need to stop trying to # run anything @@ -180,22 +221,30 @@ def Run( vars, log ): utils.sysexec_noerr( "ifconfig eth0 down", log ) - if BOOT_CD_VERSION[0] == 2: - utils.sysexec_noerr( "killall dhcpcd", log ) - elif BOOT_CD_VERSION[0] == 3: - utils.sysexec_noerr( "killall dhclient", log ) + utils.sysexec_noerr( "killall dhclient", log ) - utils.sysexec_noerr( "umount -a -r -t ext2,ext3", 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 != "": + 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: @@ -230,7 +279,7 @@ def Run( vars, log ): # problems during chain boot, such as USB host # controller drivers (HCDs) (PL6577). # if int(parts[2]) == 0: - if re.search('_hcd$', parts[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: @@ -242,7 +291,7 @@ def Run( vars, log ): log.write( "Couldn't read /proc/modules, continuing.\n" ) - kargs = "root=/dev/mapper/planetlab-root ramdisk_size=8192" + kargs = "root=%s ramdisk_size=8192" % PARTITIONS["mapper-root"] if NODE_MODEL_OPTIONS & ModelOptions.SMP: kargs = kargs + " " + "acpi=off" try: @@ -257,9 +306,10 @@ def Run( vars, log ): # 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) + 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