* See what it looks like to replace "master key" with "root key". There are two places that would require more challenging code changes: the storage path `core/master`, and its contents (the JSON-serialized EncodedKeyringtructure.)
* Restore accidentally deleted line
* Add changelog
* Update root->recovery
* Fix test
Co-authored-by: Nick Cabatoff <ncabatoff@hashicorp.com>
* Make sure we sanitize the rotation config on each clone
* Add regression test for missing rotation config
* use Equals
* simplify
Co-authored-by: Scott G. Miller <smiller@hashicorp.com>
* Add NIST guidance on rotating keys used for AES-GCM encryption
* Capture more places barrier encryption is used
* spacing issue
* Probabilistically track an estimated encryption count by key term
* Un-reorder imports
* wip
* get rid of sampling
Shamir seals now come in two varieties: legacy and new-style. Legacy
Shamir is automatically converted to new-style when a rekey operation
is performed. All new Vault initializations using Shamir are new-style.
New-style Shamir writes an encrypted master key to storage, just like
AutoUnseal. The stored master key is encrypted using the shared key that
is split via Shamir's algorithm. Thus when unsealing, we take the key
fragments given, combine them into a Key-Encryption-Key, and use that
to decrypt the master key on disk. Then the master key is used to read
the keyring that decrypts the barrier.
This lets other parts of Vault that can't depend on the vault package
take advantage of the subview functionality.
This also allows getting rid of BarrierStorage and vault.Entry, two
totally redundant abstractions.
The result will still pass gofmtcheck and won't trigger additional
changes if someone isn't using goimports, but it will avoid the
piecemeal imports changes we've been seeing.
If we have a panic defer functions are run but unlocks aren't. Since we
can't really trust plugins and storage, this backs out the changes for
those parts of the request path.
* Remove a lot of deferred functions in the request path.
There is an interesting benchmark at https://www.reddit.com/r/golang/comments/3h21nk/simple_micro_benchmark_to_measure_the_overhead_of/
It shows that defer actually adds quite a lot of overhead -- maybe 100ns
per call but we defer a *lot* of functions in the request path. So this
removes some of the ones in request handling, ha, barrier, router, and
physical cache.
One meta-note: nearly every metrics function is in a defer which means
every metrics call we add could add a non-trivial amount of time, e.g.
for every 10 extra metrics statements we add 1ms to a request. I don't
know how to solve this right now without doing what I did in some of
these cases and putting that call into a simple function call that then
goes before each return.
* Simplify barrier defer cleanup
