[dm-crypt] 4k block size in dm-crypt

Milan Broz gmazyland at gmail.com
Fri Dec 7 12:53:24 CET 2012


On 12/07/2012 04:58 AM, Vadim Finkelstein wrote:
> Regarding this: 
> http://www.saout.de/pipermail/dm-crypt/2011-April/001673.html
> 
> Was the issue of 4k block sizes ever solved? I have applied a 4k
> block size patch to my dm-crypt, but now seeing the same problem - if
> the data is read back with block size of 8192 (for example) it does
> not match what was previously written with block size of 4092.

What issue? The patch is not upstream and cannot be used wihout
additional changes.

If we add block-size parameter, we must reject all unaligned and
partial block reads to avoid this problem (and this part is still
missing, set proper io hinds for device and announce proper block size).

I would like to have such parameter one day in dmcrypt, but still
not sure what's the real use case (except workaround slow hw accelerators).

(And AES-NI and clones works quite well here with 512 block.)

Milan


More information about the dm-crypt mailing list