Converted to draft, as the Hash for Shm is probably plain wrong if some fields are not part of the hash

The derived traits seem OK to me, but I agree that the custom impl for Shm seems a little suspect. For me, the primary purpose to have a type impl Hash is to be able to use it has a key in a HashMap. I guess it's not inconceivable that you'd want to do this. Did you have a specific usecase for implementing Hash on Shm ?

I'm working on a tool for memory diagnostics.
I want to store a list of physical pages used by each shared memory segment, so HashMap<Shm, Vec<Pfn>> makes a lot of sense.

I'll rework the PR to remove the custom impl stuff.

I'm not saying that procfs can't implement these, but could you put Shm in your own local wrapper type, and impl Hash and PartialEq according to the rules that make sense for your tool?

After some thinking, I believe that the custom Hash and PartialEq that I initially implemented is plain wrong.
I switched to using derive

Tell my if it's good for you

I rebased from master