Changeset 2937 in MondoRescue for branches/3.1/mindi/rootfs/sbin/start-netfs

Timestamp:

Jan 28, 2012, 1:51:28 AM (12 years ago)

Author:

Bruno Cornec

Message:

svn merge -r 2849:2935 ../3.0

• Adds 3 binaries called potentially by udev o support USB key mount at restore time (Victor Gattegno)
• mindi only sets the boot flag of a USB device if none is present already (should avoid flip/flop situation)
• Fix a compatibility bug in analyze-my-lvm whete exclude lists with '|' was incorrectly analyzed (' ' still expected)
• Precise that -J is not supported by mondorestore yet.
• Fix a compatibility bug in mindi where exclude lists with '|' was incorrectly analyzed (' ' still expected)
• Install an MBR on USB devices when possible
• Improve error message in case mondorestore isn't found
• Adds support for more virtio devices for KVM restore
• Improve mbr.bin handling
• Fix an issue in test of FindKernel for Xen and non Xen cases
  • Fix #492 by removing an extra useless fi
  • Move fvat nd fat modules to CDROM_MODS in order to have them in the initrd to support saving logs to a USB device e.g.
  • Fix an error when dmidecode reports comments before system name
  • Fix previous broken patch on comment management for bkphw
  • Improve msgs printed during kbd analysis
  • Fix a small remaining pb with new comment support
  • Try to again improve the mount/umount stuff in mondo -Adds mention to OEL, CentOS, ScientifLinux as RHEL clones
• Fix a panadoid_free done on the wrong variable (solve a memory leak)
• Solve #495 by skipping the potential (auto-read-only) when analyzin /proc/mdstat result
• Fix some markup in man pages
• adds the post-nuke script sample for Debian (Guillaume Pernot/Andrée Leidenfrost)
• Avoid a useless umount when changing of afio set, let insist_on_cd check and perform if needed
• Copy everything under .ssh of root account to allow for sshfs mount without passwd at restore time.
• Adds support for bnx2i driver
• Try to fix #496 by supporting the protocol in the output of the mount command (sshfs#...)
• Uses the -d option of the mount command to avoid leaving loop devices in use. Probably provoked by the usage of a more recent busybox, but anyway that option doesn't hurt, and does nothing for non-loop devices. Should fix the problems reported on the ML with NFS and more than 8 ISO images (8 being the default number of loop devices available)
• Attempt to fix #436 by adding some newtDrawForm where appropriate.
• Fix bugs in the kernel detection routine of mindi (some echo were done without capturing their output, confusing the return). This should also help the Xen detection issue
• Change the logging function to be nearer from 2.2.10, and start using some
• Initialize some global fields with an empty string, now that newt Draws more regularly progress forms
• Mindi progression still not good, but showing now 3% when calling mindi. Would need a specific handling for the progress bar to evolve more afterwards
• Add compatibility with the latest SSSTK + PSP content for HP ProLiant + branding homogeneous now. We only take conrep out of the SSSTK pending a package availability. Using hpacuscripting now. Review of parameters called following test made.
• Fix #501 by pushing BOOT_SIZE to 36864 for certain Ubuntu cases
• Fix some remaining inconsistencis in the ProLiant HW backup. ready for usage.
• Fix an issue on udevadm call (needs an intermediate variable)
• Fix a remaining ProLiant tool call
• Remove again some wrong variable initialization for Net in order to not have strange messages now we call Draw more often
• Adds support for hwaddr (MAC addresses) at restore time, and map to the correct NIC if it's found at restore time (based on code from Stefan Heijmans)
• Try to Fix #499 by changing init in order to never exit, and rebooting using sysrq in /proc (not portable). Maybe should move to using a real init, and a separated rcS script.
• Fixes the shell msg error by using setsid as per Busybox FAQ
• Fix #491 by correctly checking Xen kernel used for a dom0 env (apollo13)
• Improves logging in Die()
• Improve again ProLiant HW support, by adding references to the yet to come SSSTK rpm so that conrep is called at its future right place.
• Improve mindi support for kernel 3.x
• Fix #499 for good, by using a real init from busybox and moving our init script to etc/init.d/rcS. We also now call the reboot from busybox to exit, which indeed reset the system correctly (in a VM at least which wasn't the case before). Reauires a new mindi-busybox as well.
• Attempt to solve #503 by handling /lib64 separately for udev and only if not a link (to /lib) for Ubuntu 64
• Improve again previous tests for bug #503
• Improve -d option description for NFS
• Allow http download for ftp content
• New italian mirror
• Fix #505 by documenting environment variables
  • Fix a bug where on OBDR mode compression was turned off to create the OBDR content but not back on (report from the mailing list)
• solve http://trac.mondorescue.org/ticket/520 by Victor Gattegno
• Fix #524 as option -U needs a size also
• Removes useless gendf option
• Fix #511 now stabgrub is called also after conf file modifications
• Adds support for SUID binaries
• remove a useless xenkernelname variable from mindi
• Should fix #512 by clearing old scratchdir before changing to a new
• Simplify the function engaging SCSI for Smart Array controllers, and support as many controller as we have
• Add some binaries used on SLES 10 by udev
• Fix a syntax issue in newt-specific.c
• Fix a memory free bug where a used variable was freed to early
• Fix a memory free bug where a used variable was misused in asprintf/free
• Fix #498 by using an external program (hpsa_obdr_mode) which can set the mode of the tape to CD-ROM or Sequential at will.

That program will have to be downloaded from ​http://cciss.sf.net

• Adds support for the route command in mindi-busybox
• Force block size back to 32768 in OBDR mode after skipping the 2 tape marks, to read the additional rootfs data corerctly - Fix #498 for good it seems
• Adds support for a new pata_atiixp driver
• Fix #525 and improve printing when no /etc/raidtab, but still md devices
• Exclude /run similarly to /sys and /proc
• Fix a bug in mindi when having multiple swap partiions with same starting string (/dev/dm-1 and /dev/dm-16). We now add a space in the grep to avoid mischoice.
• Fix grub installation if we had to go in the hack part which wasn't using the correct variables to make decisions !
• Remove all mention of mondorescue.com from code
• Fix some website typo on the download page and precise that we now develop on Mageia
• Exclude sysfs devpts and tmpfs partitions found in RHEL 6.2 from analyze by mindi (made error messages)
• Force activation of first partition which should not be optional as everything is destroyed before
• Fix a bug in analyze-my-lvm where LVs were created in random order (the one of lvscan) instead of using the dm-xxx order, which creates issues at restore time between the mapping created then and the one we had at backup time which could be inconssistent

-mondoarchive now exists if the mt command is missing when needed

• Proper fix for analyze-my-lvm
• Fix #500 and #473 by adding support of old metadata format for mdadm if using grub < 0.9x or lilo on a md boot device (to be tested)
• Really never inject or eject when the -e option is passed or device is inappropriate (USB, Net, ...)
• Fix #577 by having factor back as a float (not a long ! it's a decimal value) and making the appropriate cast.
• Fix #578 by improving the regexp used for CD/DVD detection with -scanbus option
• Adds more verbosity when LVM creation fails.
• Fix #579 by checking with pipes on both sides of the string
• debugfs should be excluded from the device considered for fstab analysis

File:

• branches/3.1/mindi/rootfs/sbin/start-netfs (modified) (5 diffs)

branches/3.1/mindi/rootfs/sbin/start-netfs

@@ -10 +10 @@
 ipcount=3
 
-ipconf=""
-pre=""
-post=""
+# Get info from config file
 ipdev=""
+hwaddr=""
 ipaddress=""
 ipnetmask=""
@@ -19 +18 @@
 ipgateway=""
 proto=""
+ipconf=""
+pre=""
+post=""
 netfsmount=""
 imgname=""
@@ -26 +28 @@
 if [ -f $MINDI_CACHE/mondorestore.cfg ]; then
     ipdev=`grep netfs-dev $MINDI_CACHE/mondorestore.cfg 2> /dev/null | cut -d' ' -f2-`
+    hwaddr=`grep netfs-client-hwaddr $MINDI_CACHE/mondorestore.cfg 2> /dev/null | cut -d' ' -f2-`
     ipaddress=`grep netfs-client-ipaddr $MINDI_CACHE/mondorestore.cfg 2> /dev/null | cut -d' ' -f2-`
     ipnetmask=`grep netfs-client-netmask $MINDI_CACHE/mondorestore.cfg 2> /dev/null | cut -d' ' -f2-`
@@ -46 +49 @@
 for i in `cat $CMDLINE` ; do
     echo $i | grep -qi ping= && ipcount=`echo $i | cut -d= -f2`
+    echo $i | grep -qi hwaddr= && hwaddr=`echo $i | cut -d= -f2`
     echo $i | grep -qi ipconf= && ipconf=`echo $i | cut -d= -f2`
     echo $i | grep -qi netfsmount= && export netfsmount=`echo $i | cut -d= -f2`
@@ -93 +97 @@
             ipgateway=`echo $ipconf | cut -d: -f5`
         fi
+
+        # If same system, map to the right MAC address
+        hwaddr_found=`ifconfig $ipdev | /bin/grep HWaddr | awk '{print $NF}'`
+        if [ "$hwaddr" != "$hwaddr_found" ]; then
+            ipdev_new=`ifconfig -a | /bin/grep $hwaddr | awk '{print $1}'`
+            if [ "$ipdev_new" != "" ]; then
+                LogIt "Interface $ipdev changed to $ipdev_new (MAC: $hwaddr)"
+                ipdev=$ipdev_new
+            else
+                LogIt "NOTE: Interface $ipdev kept despite it doesn't match the $hwaddr MAC address"
+            fi
+        fi
+
         LogIt "Configuring $ipdev statically ($ipaddress/$ipnetmask)"
         ifconfig $ipdev $ipaddress netmask $ipnetmask broadcast $ipbroadcast
         route add default gw $ipgateway
+
     fi
+
+    # Leave time to the satck to wake up (reported by some users)
+    sleep 5
 
     # ping server helps waking interface up