3 # Copyright (c) 2003 Intel Corporation
6 # Copyright (c) 2004-2006 The Trustees of Princeton University
17 import notify_messages
20 from Exceptions import BootManagerException
22 import UpdateNodeConfiguration
23 import StopRunlevelAgent
28 Load the kernel off of a node and boot to it.
29 This step assumes the disks are mounted on SYSIMG_PATH.
30 If successful, this function will not return. If it returns, no chain
33 Expect the following variables:
34 SYSIMG_PATH the path where the system image will be mounted
35 (always starts with TEMP_PATH)
36 ROOT_MOUNTED the node root file system is mounted
37 NODE_SESSION the unique session val set when we requested
38 the current boot state
39 PLCONF_DIR The directory to store PL configuration files in
41 Sets the following variables:
42 ROOT_MOUNTED the node root file system is mounted
45 log.write("\n\nStep: Chain booting node.\n")
47 # make sure we have the variables we need
49 SYSIMG_PATH = vars["SYSIMG_PATH"]
51 raise ValueError("SYSIMG_PATH")
53 PLCONF_DIR = vars["PLCONF_DIR"]
55 raise ValueError("PLCONF_DIR")
57 # its ok if this is blank
58 NODE_SESSION = vars["NODE_SESSION"]
60 NODE_MODEL_OPTIONS = vars["NODE_MODEL_OPTIONS"]
62 PARTITIONS = vars["PARTITIONS"]
63 if PARTITIONS == None:
64 raise ValueError("PARTITIONS")
66 except KeyError as var:
67 raise BootManagerException("Missing variable in vars: {}\n".format(var))
68 except ValueError as var:
69 raise BootManagerException("Variable in vars, shouldn't be: {}\n".format(var))
72 if vars.has_key('ROOT_MOUNTED'):
73 ROOT_MOUNTED = vars['ROOT_MOUNTED']
76 log.write("Mounting node partitions\n")
78 # simply creating an instance of this class and listing the system
79 # block devices will make them show up so vgscan can find the planetlab
81 systeminfo.get_block_devices_dict(vars, log)
83 utils.sysexec("vgscan", log)
84 utils.sysexec("vgchange -ay planetlab", log)
86 utils.makedirs(SYSIMG_PATH)
88 cmd = "mount {} {}".format(PARTITIONS["root"], SYSIMG_PATH)
89 utils.sysexec(cmd, log)
90 cmd = "mount -t proc none {}/proc".format(SYSIMG_PATH)
91 utils.sysexec(cmd, log)
92 cmd = "mount {} {}/vservers".format(PARTITIONS["vservers"], SYSIMG_PATH)
93 utils.sysexec(cmd, log)
96 vars['ROOT_MOUNTED'] = 1
98 utils.display_disks_status(PARTITIONS, "In ChainBootNode", log)
100 # write out the session value /etc/planetlab/session
102 session_file_path = "{}/{}/session".format(SYSIMG_PATH, PLCONF_DIR)
103 session_file = file(session_file_path, "w")
104 session_file.write(str(NODE_SESSION))
107 log.write("Updated /etc/planetlab/session\n")
109 log.write("Unable to write out /etc/planetlab/session, continuing anyway\n")
111 # update configuration files
112 log.write("Updating configuration files.\n")
113 # avoid using conf_files initscript as we're moving to systemd on some platforms
115 if (vars['ONE_PARTITION'] != '1'):
117 cmd = "/usr/bin/env python /usr/share/NodeManager/conf_files.py --noscripts"
118 utils.sysexec_chroot(SYSIMG_PATH, cmd, log)
120 log.write("conf_files failed with \n {}".format(e))
122 # update node packages
123 #log.write("Running node update.\n")
124 #if os.path.exists(SYSIMG_PATH + "/usr/bin/NodeUpdate.py"):
125 # cmd = "/usr/bin/NodeUpdate.py start noreboot"
127 # for backwards compatibility
128 # cmd = "/usr/local/planetlab/bin/NodeUpdate.py start noreboot"
129 #utils.sysexec_chroot(SYSIMG_PATH, cmd, log)
130 log.write("NodeUpdate skipped\n")
132 # Re-generate initrd right before kexec call
133 # this is not required anymore on recent depls.
134 if vars['virt'] == 'vs':
135 MakeInitrd.Run(vars, log)
137 # the following step should be done by NM
138 UpdateNodeConfiguration.Run(vars, log)
140 log.write("Updating ssh public host key with PLC.\n")
143 ssh_host_key_file = file("{}/etc/ssh/ssh_host_rsa_key.pub".format(SYSIMG_PATH), "r")
144 ssh_host_key = ssh_host_key_file.read().strip()
145 ssh_host_key_file.close()
146 ssh_host_key_file = None
151 update_vals['ssh_rsa_key'] = ssh_host_key
152 BootAPI.call_api_function(vars, "BootUpdateNode", (update_vals,))
155 # get the kernel version
157 if NODE_MODEL_OPTIONS & ModelOptions.SMP:
160 log.write("Copying kernel and initrd for booting.\n")
161 if vars['virt'] == 'vs':
162 utils.sysexec("cp {}/boot/kernel-boot{} /tmp/kernel".format(SYSIMG_PATH, option), log)
163 utils.sysexec("cp {}/boot/initrd-boot{} /tmp/initrd".format(SYSIMG_PATH, option), log)
165 # Use chroot to call rpm, b/c the bootimage&nodeimage rpm-versions may not work together
167 kversion = os.popen("chroot {} rpm -qa kernel | tail -1 | cut -c 8-"\
168 .format(SYSIMG_PATH)).read().rstrip()
169 major_version = int(kversion[0]) # Check if the string looks like a kernel version
171 # Try a different method for non-rpm-based distributions
172 kversion = os.popen("ls -lrt {}/lib/modules | tail -1 | awk '{{print $9;}}'"\
173 .format(SYSIMG_PATH)).read().rstrip()
175 # kernel and initrd come in various locations depending on the distro
177 kernel_candidates = []
178 kernel_candidates.append("/boot/vmlinuz-{}*".format(kversion))
179 # on f23 starting dec. 2015 - what a twisted naming scheme
180 kernel_candidates.append("/boot/*/{}*/linux".format(kversion))
182 initrd_candidates = []
183 # f16/18: expect initramfs image here
184 initrd_candidates.append ("/boot/initramfs-{}.img".format(kversion))
185 # f20: uses a uid of some kind, e.g. /boot/543f88c129de443baaa65800cf3927ce/<kversion>/initrd
186 initrd_candidates.append ("/boot/*/{}/initrd".format(kversion))
188 initrd_candidates.append ("/boot/initrd.img-{}".format(kversion))
190 def install_from_sysimg_to_tmp (candidates, name):
193 for pattern in candidates:
194 matches = glob.glob(SYSIMG_PATH+pattern)
195 log.write("locating {}: found {} matches in {}\n".format(name, len(matches), pattern))
200 raise Exception("Unable to locate {} for kexec'ing".format(name))
201 utils.sysexec("cp {} /tmp/{}".format(found, name))
203 install_from_sysimg_to_tmp(kernel_candidates, 'kernel')
204 install_from_sysimg_to_tmp(initrd_candidates, 'initrd')
208 log.write("Unmounting disks.\n")
210 if (vars['ONE_PARTITION'] != '1'):
211 utils.sysexec("umount {}/vservers".format(SYSIMG_PATH), log)
212 utils.sysexec("umount {}/proc".format(SYSIMG_PATH), log)
213 utils.sysexec_noerr("umount {}/dev".format(SYSIMG_PATH), log)
214 utils.sysexec_noerr("umount {}/sys".format(SYSIMG_PATH), log)
215 utils.sysexec("umount {}".format(SYSIMG_PATH), log)
216 utils.sysexec("vgchange -an", log)
219 vars['ROOT_MOUNTED'] = 0
221 # Change runlevel to 'boot' prior to kexec.
222 StopRunlevelAgent.Run(vars, log)
224 log.write("Unloading modules and chain booting to new kernel.\n")
226 # further use of log after Upload will only output to screen
227 log.Upload("/root/.bash_eternal_history")
229 # regardless of whether kexec works or not, we need to stop trying to
231 cancel_boot_flag = "/tmp/CANCEL_BOOT"
232 utils.sysexec("touch {}".format(cancel_boot_flag), log)
234 # on 2.x cds (2.4 kernel) for sure, we need to shutdown everything
235 # to get kexec to work correctly. Even on 3.x cds (2.6 kernel),
236 # there are a few buggy drivers that don't disable their hardware
237 # correctly unless they are first unloaded.
239 utils.sysexec_noerr("ifconfig eth0 down", log)
241 utils.sysexec_noerr("killall dhclient", log)
243 if vars['virt'] == 'vs':
244 utils.sysexec_noerr("umount -a -r -t ext2,ext3", log)
246 utils.sysexec_noerr("umount -a -r -t ext2,ext3,btrfs", log)
247 utils.sysexec_noerr("modprobe -r lvm-mod", log)
249 # modules that should not get unloaded
250 # unloading cpqphp causes a kernel panic
251 blacklist = [ "floppy", "cpqphp", "i82875p_edac", "mptspi"]
253 modules = file("/tmp/loadedmodules","r")
256 module = string.strip(line)
257 if module in blacklist :
258 log.write("Skipping unload of kernel module '{}'.\n".format(module))
260 log.write("Unloading {}\n".format(module))
261 utils.sysexec_noerr("modprobe -r {}".format(module), log)
262 if "e1000" in module:
263 log.write("Unloading e1000 driver; sleeping 4 seconds...\n")
268 log.write("Couldn't read /tmp/loadedmodules, continuing.\n")
271 modules = file("/proc/modules", "r")
273 # Get usage count for USB
277 # Module Size UsageCount UsedBy State LoadAddress
278 parts = string.split(line)
280 if parts[0] == "usb_storage":
281 usb_usage += int(parts[2])
282 except IndexError as e:
283 log.write("Couldn't parse /proc/modules, continuing.\n")
289 # Module Size UsageCount UsedBy State LoadAddress
290 parts = string.split(line)
292 # While we would like to remove all "unused" modules,
293 # you can't trust usage count, especially for things
294 # like network drivers or RAID array drivers. Just try
295 # and unload a few specific modules that we know cause
296 # problems during chain boot, such as USB host
297 # controller drivers (HCDs) (PL6577).
298 # if int(parts[2]) == 0:
299 if False and re.search('_hcd$', parts[0]):
301 log.write("NOT unloading {} since USB may be in use\n".format(parts[0]))
303 log.write("Unloading {}\n".format(parts[0]))
304 utils.sysexec_noerr("modprobe -r {}".format(parts[0]), log)
305 except IndexError as e:
306 log.write("Couldn't parse /proc/modules, continuing.\n")
308 log.write("Couldn't read /proc/modules, continuing.\n")
311 kargs = "root={} ramdisk_size=8192".format(PARTITIONS["root"])
312 if NODE_MODEL_OPTIONS & ModelOptions.SMP:
313 kargs = kargs + " " + "acpi=off"
315 kargsfb = open("/kargs.txt","r")
316 moreargs = kargsfb.readline()
318 moreargs = moreargs.strip()
319 log.write('Parsed in "{}" kexec args from /kargs.txt\n'.format(moreargs))
320 kargs = kargs + " " + moreargs
322 # /kargs.txt does not exist, which is fine. Just kexec with default
323 # kargs, which is ramdisk_size=8192
326 utils.sysexec_noerr('hwclock --systohc --utc ', log)
327 # utils.breakpoint("Before kexec");
329 utils.sysexec('kexec --force --initrd=/tmp/initrd --append="{}" /tmp/kernel'.format(kargs), log)
330 except BootManagerException as e:
331 # if kexec fails, we've shut the machine down to a point where nothing
332 # can run usefully anymore (network down, all modules unloaded, file
333 # systems unmounted. write out the error, and cancel the boot process
336 log.write("-------------------------------------------------------\n")
337 log.write("kexec failed with the following error. Please report\n")
338 log.write("this problem to support@planet-lab.org.\n\n")
339 log.write(str(e) + "\n\n")
340 log.write("The boot process has been canceled.\n")
341 log.write("-------------------------------------------------------\n\n")