Re: [dm-crypt] 1.6.2 - waiting for zero, luksFormat hung

[Milan Broz <gmazyland@gmail.com>]

On 11/22/2013 08:40 AM, shmick@riseup.net wrote:
> 
> 
> Milan Broz:
>> On 11/21/2013 03:55 PM, shmick@riseup.net wrote:
>>> ive also then tried with my distro default cryptsetup 1.4.1 but the same
>>> errors again in xts mode with AES or TWOFISH
>>>
>>> any ideas ???
>>
>> Did you modified udev device-mapper rules? If not, perhaps you should report
>> it to your distro bugzilla if it doesn't work with default package.
> 
> no, i didn't modify udev
> i've done some more testing and im very certain this an ascii character
> issue with cryptsetup 1.4.3


No. Why ascii characters problem?

"Waiting for zero" is situation where device mapper wait for semaphore
(used to synchronize block device creation) to reach zero, it is handled in udev rules
by generic device-mapper rules.

It has absolutely nothing to do with passwords or ciphers in cryptsetup. I am afraid
you are mixing several problems together here.
(You should get "No key available with this passphrase" error and not hang if it is
passphrase problem.)

BTW if you just need to open old container, do not compile own version and use
distro provided one (1.4 version is fully compatible here).

Anyway, adding cc to Peter who maybe know some problem in dm udev rules causing
this...

And if you cannot run even benchmark, you are still loading wrong libcryptsetup library,
(cryptsetup: relocation error: cryptsetup: symbol crypt_benchmark_kdf - this
function was added later, in cryptsetup 1.6).

Milan


> 
> i did test the following:
> 
> 1.
> will all the previous problems i was using 63 random ascii characters
> i triple checked all of these according to the wikipedia 95 character
> standards - they are all ok according to ascii tables
> 
> so i then tried with a basic password - it worked fine using identical
> luksFormat setings when creating and opening cryptsetup 1.4.3 from a live cd
> 
> however, i could not later open this same volume within linux mint 13
> amd-64 using cryptsetup 1.4.1 - same waiting for zero error
> 
> i don't know if different versions of mdadm affect this - i created RAID
> using latest mdadm 3.3 - the live cd has an earlier version, but it
> still opens and assembles the raid anyhow
> 
> 2.
> re-installed cryptsetup 1.6.2 in linux mint 13
> ./configure --with-libgcrypt-prefix=/usr/local LDFLAGS=-L/usr/local/lib
> 
> reboot
> 
> create volume
> cryptsetup --debug --hash sha512 --cipher twofish-xts-plain64
> --use-random --key-size 256 --iter-time 2000 luksFormat /dev/md0
> 
> try basic password with letters & numbers only
> 
> failed again, hung at 'waiting for zero'
> 
> same problem errors with failing to run benchmark
> 
> i don't understand how to get around this
> 
> desired outcomes:
> i want to use a long, secure passphrase, formatting with
> [twofish][aes]-xts-plain64 - to me it doesnt matter which cryptsetup
> version i use as long as it works
> 
> but so far i can't use 63 ascii charaters in any of the versions ive tried
> 
> ?
> 
>>
>> (The "waiting to zero" is in fact libdevmapper internal problem, cryptsetup is
>> just an user here.)
>> Usually it is caused byt removal of mandatory device-mapper udev rules.
> 
> what does this mean ?
> how would i fix this for myself and what would i change regarding a udev
> conf option ?
> 
>>
>> Cipher or dmcrypt configuration is perhaps not related to this.
>>
>> Milan
>>