How to manually fsck an encrypted pupsave

[PaulBx1]

As of Puppy 4.1.2, "puppy pfix=fsck" does not work. This is a manual alternative:

Boot "puppy pfix=ram"

In a console window, mount the volume containing your pupsave. Then "cd" there.

modprobe cryptoloop
modprobe aes
losetup-FULL -e aes /dev/loop6 pup_save_crypta.2fs
e2fsck /dev/loop6

This procedure assumes heavy encryption and will need modification for light encryption (perhaps someone else can contribute that recipe). It also assumes your pupsave is named "pup_save_crypta.2fs", of course.

[vtpup]

I need assistance here for the commands needed to do an fsck on a light encrypted pup save -- anyone?

[MU]

I'm in hurry, so just a quick guess:

losetup-FULL -e xor /dev/loop6 pup_save_crypta.2fs
e2fsck /dev/loop6

http://noforum.de/man2html.php?page=losetup

Mark

[Pizzasgood]

According to the comments in the scripts in Puppy 4.2, that xor is a different one from the xor used for the encryption. According to the scripts it works like this:

Code: Select all

modprobe cryptoloop
losetup-FULL -E 1 /dev/loop6 pup_save_cryptx.2fs
e2fsck /dev/loop6 

I don't have one handy to test with though.

Also, if you have a bunch of filesystem images mounted, /dev/loop6 might already be in use. Running losetup-FULL -f will tell you the name of a free loopback device. And if you screw up while doing this, the loopback device you use might be left in-use. In that case, you can free it with losetup-FULL -d /dev/loop6

Just doing losetup-FULL /dev/loop6 will tell you information about the device.




REFERENCE: 683-687 of init from initrd.gz of Puppy 4.2.1:

Code: Select all

   case $PUPSAVEFILE in
    *cryptx*) #see /etc/rc.d/rc.shutdown.
     CRYPTO='-E 1' #v2.16final '-e xor' --bug, loads xor.ko which is something else.
     modprobe cryptoloop
     ;;

EDITED 2009.07.05: Changed 'losetup' to 'losetup-FULL' because it's more reliable across multiple versions of Puppy.

[vtpup]

Thanks Mu and pizzasgood-

I'm using 4.1.2, not 4.2 -- but that didn't make a difference, pizzasgood solution worked.

I did try Mu's suggetsion first but I got an error message after entering the password. Thanks Mu for giving it a shot that quickly!

Thanks pizzasgood once again for the massive amount of help you give people here.

To clarify for others the steps that worked for me:

In my case my savefile was on sda6/puppy412/ and was named pup_save_cryptx.2fs, change that if your setup is different (marked in blue below). So:

1.) Boot puppy CD w/ puppy pfix=ram option
2.) Mount sda6
3.) open terminal
4.) type and enter modprobe cryptoloop
5.) type and enter losetup -E 1 /dev/loop6 /mnt/sda6/puppy412/pup_save_cryptx.2fs
6.) answer the password prompt
7.) type and enter e2fsck /dev/loop6
8.) you may be asked if you want to correct specific problems -- I answered yes to all, however it's possible not to if you feel something shouldn't be changed.

Thanks once again Pizzasgood, Mu, and PaulBx1 for this. You guys are great!

Makes me wonder if there's a simple bugfix in 4.1.2's code so that you can do an fsck as you are supposed to at the bootup password prompt. Functionality is supposed to be: hit ENTER instead of password, then in answer to the red prompt, hit any other letter.

I'd really like to do a routine fsck every month or so using the normal bootup method.

[Pizzasgood]

I'll take a look later tonight and see if it's still a problem in 4.2.1. If it isn't, I can throw together a 4.1.2 frugal really quick and test with that. Can't be too hard to fix.

[heaths]

Hi thanks for your replies
simulationcredit

[Pizzasgood]

Well, in the init script from 4.2.1 I found this at lines 742-753:

Code: Select all

      losetup $CRYPTO /dev/loop1 /mnt/dev_save$PUPSAVEFILE
     fi
    fi

    echo "/dev/loop1 $CREATEPUPSAVE2FS ext2 defaults  1 1" >> /etc/fstab #v2.21
    #fsck_func loop1 ext2 $PUPSAVEFILE #v3.01. no, removed, takes ages, and won't mount afterward.
    mount -t ext2 -o noatime,rw /dev/loop1 $CREATEPUPSAVE2FS
    MNTSTAT=$?
    if [ "$MNTSTAT" = "0" ] ; then
     echo -n "...successfully mounted" >/dev/console
     break
    else 

The line we're interested in is 747:
#fsck_func loop1 ext2 $PUPSAVEFILE #v3.01. no, removed, takes ages, and won't mount afterward.

The fsck_func function automatically checks whether the pfix=fsck option has been given, and immediately returns unless it has. Otherwise it runs a scan. Except that this line was commented out, apparently because it wasn't working properly in 3.01. I'm not sure what that was about. I remember that there some issues with encryption back around that time, so maybe it was some 3.01 specific problem.

Anyways, I uncommented that line, rebuilt my initrd.gz file, and rebooted with pfix=fsck. As far as I can tell it worked. I got a message saying it was checking the file, and then after a short pause things continued as usual. After booting I looked in /initrd/tmp/ and found the file chret, which is where the fsck_func function stores the output from e2fsck. It listed a number of issues that it fixed automatically. Nothing seemed out of order.


I'm not going to post a bunch of initrd.gz files for the different versions of Puppy 4.x because I probably should have gone to bed half an hour ago. If anybody needs them and isn't willing to do it themselves, let me know and I can make them another night. But it's only a minor change to init - extract the initrd.gz file, open the 'init' script as text, and uncomment that line. It's line 747 in 4.2.1, but it will be at other places in other Puppies. Just paste the line into the find bar and it will come up. After uncommenting it (simply remove the #) rebuild and replace the initrd.gz file.


For those who need instructions on editing initrd.gz: How to edit initrd.gz (for Puppy 3.x and newer)

[vtpup]

Thanks for that Pizzasgood, I will try it later tonight.

I'm wondering if the reason for long delay, and/or reason it wouldn't mount is that the fsck process is throwing a Y/N question for the user, but possibly the user doesn't see or answer it? Maybe a subsequent silent abort leaves things mixed up. I don't know just throwing out a possibility since the code I used earlier definitely requires sitting by the keyboard. I had to hold down the Y key for autorepeat to get through all of the Y/N queries. Guess I had a lot of errors.

Is the fsck in initrd set to auto-fix errors, or does it still query?

Thanks again, and happy 4th of july (assuming you're in the states)!

[Pizzasgood]

This is the code for fsk_func in Puppy 4.2.1 (Lines 56-85 of init):

Code: Select all

#v3.01 fsck for ext2/3 prior to mounting...
fsck_func() { #passed params: partition filesystem [filename]
 [ "$PFSCK" != "yes" ] && return 0 #v404 default is not do any checking.
 #v3.93 restrict to ext2 only...
 #[ "$2" != "ext2" -a "$2" != "ext3" ] && return 0 #precaution.
 [ "$2" != "ext2" ] && return 0 #precaution.
 CHKDEV=$1
 #[ $3 ] && CHKDEV="`basename $3`"
 e2fsck -y /dev/$1 >/tmp/chkret &
 #sleep 1
 #if [ "`grep ': clean, ' /tmp/chkret`" = "" ];then
 # echo -e "\\033[1;35m" >/dev/console #35=purple
 # echo -n "Performing filesystem check on ${CHKDEV}, please wait" >/dev/console
 # echo -en "\\033[0;39m" >/dev/console
 #fi
 echo -en "\\033[1;35m" >/dev/console #35=purple
 MSGFSCK=" ${1} filesystem check, please wait..............."
 MSGFSCNT=1
 while [ "`pidof e2fsck`" != "" ];do
  if [ "${1}" = "loop1" ];then
   [ $MSGFSCNT -eq 1 ] && echo -n "filesystem check..." >/dev/console
  else
   [ $MSGFSCNT -lt 47 ] && MSGCHAR="`echo -n "$MSGFSCK" | cut -b $MSGFSCNT`"
   echo -n "$MSGCHAR" >/dev/console
  fi
  sleep 1
  MSGFSCNT=`expr $MSGFSCNT + 1`
 done
 echo -en "\\033[0;39m" >/dev/console
}

In particular, line 64:
e2fsck -y /dev/$1 >/tmp/chkret &
So it's assuming yes, at least in 4.2.1.


Yeah, I'm in TN right now, in a house sticking off the side of some mountain (Mom just moved here three weeks ago). Went to see Transformers yesterday, and played some Nethack. Then I forgot which day it was and thought the neighbors were shooting at something for a couple moments before I realized it was just fireworks.

[vtpup]

.... It's line 747 in 4.2.1, but it will be at other places in other Puppies. Just paste the line into the find bar and it will come up. After uncommenting it (simply remove the #) rebuild and replace the initrd.gz file.


For those who need instructions on editing initrd.gz: How to edit initrd.gz (for Puppy 3.x and newer)

It's Line #726 in Puppy 4.1.2, for anyone interested.

Man the 4th of july came and went, along with 3 more months. Sorry about the slow reply. On a mountain myself hanging off it in a cabin.

[stiginge]

Hi Paul BX, I've had trouble with my heavy encrypted savefile, documented here:

http://www.murga-linux.com/puppy/viewtopic.php?t=72236

I used the code you provided above but it doesn't seem to work due to some block issue, have you seen this before?

[watchdog]

In wary 5.3 with lightencrypted savefile:

Code: Select all

modprobe cryptoloop
losetup-FULL -E 1 /dev/loop2 warysave_cryptx-w53.2fs
e2fsck /dev/loop2
losetup-FULL -d /dev/loop2

worked for me.