[dm-devel] [PATCH] a deadlock bug in the kernel-side device mapper code

[malahal at us.ibm.com]

Mikulas Patocka [mpatocka at redhat.com] wrote:
> This patch introduces a new lock, _name_read_lock, that is placed around
> regions that modify pointer to the hash with dm_set_mdptr or that modify
> hc->name or hc->uuid. When this lock is taken, the caller can safely read
> the name and uuid.
> 
> This lock has much smaller span than _hash_lock and thus lock recursion
> can't happen.

This lock actually guarantees that hash cell, name (and uuid) don't
disappear when someone holds this lock. How about naming it to something
like _hashcell_name_uuid_lock()???

Thanks, Malahal.