biosnoop doen't play nice with device-mapper/dmcrypt

My system uses dmcrypt to encrypt the underlying block-device. Running biosnoop reports all block I/O requests as originating from `dmcrypt_writer`, a kernel thread, masking the originating pid behind the requests.

There's a nice block diagram in [wp:Device_mapper](https://en.wikipedia.org/wiki/Device_mapper). I guess the "Optional" bit in the middle muddies the water. I also wonder if this is an issue specific to dmcrypt or whether other dm targets also share this problem (notably LVM). Both fairly common.