zuluCrypt vs truecrypt ( shameless plug :-) )

[muungwana]

ok,so the problem is not reproducible.

I have a hunch it was caused by heavy prolonged external disk I/O while the CPU was under a sufficient amount of stress.

More people are updating to 3.2.x kernel and hence if there is an underlying problem are these two reports are its symptoms,then more people will start reporting it and hopefully a predictable pattern will be found.

[Bald Brick]

muungwana'

You've got a convert. After playing with zuluCrypt for three days I've switched from TrueCrypt to zuluCrypt. And I think the app should be in the repository; if you don't feel like requesting it yourself I can do it.

[muungwana]

i very much appreciate your appreciation of the little value i have added to the FOSS world.

I will also appreciate if you could make a package request, people usually do not like people making package requests for their own work.

Please dont hesitate to comment if you hit any bug,have any suggestion or criticism.

[gseaman]

No need to make a request. I'll package it.

Galen

[muungwana]

ok,so the problem is not reproducible.

I have a hunch it was caused by heavy prolonged external disk I/O while the CPU was under a sufficient amount of stress.

More people are updating to 3.2.x kernel and hence if there is an underlying problem are these two reports are its symptoms,then more people will start reporting it and hopefully a predictable pattern will be found.

I updated to kernel 3.2.x and the system locks shut somewhere around 17% every time it tries to write random data to the file during encrypted container file creation process.

I found a discussion thread below talking about this. Will move away from using cryptsetup to generate random data.
http://www.saout.de/pipermail/dm-crypt/2010-August/001111.html

[muungwana]

@Bald Brick

can you do the following test on the computer that locked up earlier? i am trying to pin point the lockup with the cryptsetup people.

1. what kernel were you running when you got the lock up?
2. what are your computer spec?

1. log in as root.
2. create a 5GB file with the following command:

dd if=/dev/zero of=zzz bs=1000 count=5M

3. create a cryptsetup mapper with below command,enter random password when asked

cryptsetup --cipher=aes-cbc-essiv:sha256  --hash=ripemd160  --key-size=256 create zzz zzz

4. write random data to the file through the mapper

dd if=/dev/zero of=/dev/mapper/zzz bs=1000 count=5M

did the last command return without the computer hanging?

it it succeeded, them run this command to remove the mapper

cryptsetup remove zzz

[Bald Brick]

@Bald Brick

can you do the following test on the computer that locked up earlier? i am trying to pin point the lockup with the cryptsetup people.

1. what kernel were you running when you got the lock up?

kernel.h-3.2.18-pclos2.bfs

2. what are your computer spec?

My specs are in my signature, apart from the specs of the removable drive that I was trying to write to when the hang happened. It's an old 300GB Buffalo HD-PEU2. The computer itself is an Acer Aspire M5201.

1. log in as root.
2. create a 5GB file with the following command:

dd if=/dev/zero of=zzz bs=1000 count=5M

3. create a cryptsetup mapper with below command,enter random password when asked

cryptsetup --cipher=aes-cbc-essiv:sha256  --hash=ripemd160  --key-size=256 create zzz zzz

4. write random data to the file through the mapper

dd if=/dev/zero of=/dev/mapper/zzz bs=1000 count=5M

did the last command return without the computer hanging?

it it succeeded, them run this command to remove the mapper

cryptsetup remove zzz

No problems this time. I actually logged in as root (didn't just su to root) and this is the output:

[root@localhost ~]# dd if=/dev/zero of=zzz bs=1000 count=5M
5242880+0 records in
5242880+0 records out
5242880000 bytes (5.2 GB) copied, 100.427 s, 52.2 MB/s
[root@localhost ~]# cryptsetup --cipher=aes-cbc-essiv:sha256  --hash=ripemd160  --key-size=256 create zzz zzz
Enter passphrase:
[root@localhost ~]# dd if=/dev/zero of=/dev/mapper/zzz bs=1000 count=5M
5242880+0 records in
5242880+0 records out
5242880000 bytes (5.2 GB) copied, 241.8 s, 21.7 MB/s
[root@localhost ~]# cryptsetup remove zzz
[root@localhost ~]#

This, of course, created zzz in my /root folder so I repeated the process from the HD-PEU2 partition (this time after just su'ing to root):

[erik  /media/HD-PEU2]$ su -
Password:
[root@localhost ~]# cd /media/HD-PEU2
[root@localhost HD-PEU2]# dd if=/dev/zero of=zzz bs=1000 count=5M
5242880+0 records in
5242880+0 records out
5242880000 bytes (5.2 GB) copied, 145.739 s, 36.0 MB/s
[root@localhost HD-PEU2]# cryptsetup --cipher=aes-cbc-essiv:sha256  --hash=ripemd160  --key-size=256 create zzz zzz
Enter passphrase:
[root@localhost HD-PEU2]# dd if=/dev/zero of=/dev/mapper/zzz bs=1000 count=5M
5242880+0 records in
5242880+0 records out
5242880000 bytes (5.2 GB) copied, 336.143 s, 15.6 MB/s
[root@localhost HD-PEU2]# cryptsetup remove zzz
[root@localhost HD-PEU2]#

As you can see the copying speed was a bit slower, but otherwise the result was the same. I'm beginning to think that my lockup was just something that happened by chance.

[muungwana]

on my system, i have 1 GB ram and the speed get lower as the written file get bigger until it locks up when the file is big enough( 1GB ).

you have 7.8GB RAM and 5GB file worked, what about with 10GB file, a file bigger than your RAM?

you got a lock up when you try to create a 10GB file,  can you repeat the experiment on last time, changing "count=10M" above. That will create a 10GB filea file bigger than your RAM,the same size as in your first attempt.

also what file system does that partition use?

[Bald Brick]

on my system, i have 1 GB ram and the speed get lower as the written file get bigger until it locks up when the file is big enough( 1GB ).

you have 7.8GB RAM and 5GB file worked, what about with 10GB file, a file bigger than your RAM?

Yes, but after my lockup I managed to create a 30GB volume on the same drive, and a bit later a 20GB volume on my 32 bit install. (I wanted to check zuluCrypt there too.)

you got a lock up when you try to create a 10GB file,  can you repeat the experiment on last time, changing "count=10M" above. That will create a 10GB filea file bigger than your RAM,the same size as in your first attempt.

Everything worked as it should:

[root@localhost ~]# cd /media/HD-PEU2/
[root@localhost HD-PEU2]# dd if=/dev/zero of=zzz bs=1000 count=10M
10485760+0 records in
10485760+0 records out
10485760000 bytes (10 GB) copied, 337.936 s, 31.0 MB/s
[root@localhost HD-PEU2]# cryptsetup --cipher=aes-cbc-essiv:sha256  --hash=ripemd160  --key-size=256 create zzz zzz
Enter passphrase:
[root@localhost HD-PEU2]# dd if=/dev/zero of=/dev/mapper/zzz bs=1000 count=10M
10485760+0 records in
10485760+0 records out
10485760000 bytes (10 GB) copied, 801.819 s, 13.1 MB/s
[root@localhost HD-PEU2]# cryptsetup remove zzz
[root@localhost HD-PEU2]#

This time I monitored memory usage throughout most of the process. It never went over 1.1GiB (1.18GB). So swap wasn't used at all.

also what file system does that partition use?

I could have sworn that it was ext4, but I checked and it's actually xfs.

[Bald Brick]

At first I suspected that my lockup happened because of overheating, but even if the container-creation process seems to gobble up all the CPU cycles it can, that is not likely to result in overheating in the few minutes the process takes. And the few times that some parts of the system really have become too hot (after hours of the CPU running at close to 100%) that hasn't led to the computer locking up: it has just spontaneously started a shut-down sequence.

[muungwana]

ok, thanks for your extensive tests and explanations, very much appreciated.