To ensure the new process' parent is pid 1, so systemd won't
complain about supervising a process it does not own.
Fixes the following log spam on reloads:
Apr 25 10:50:54 deb-dev systemd[1]: proxmox-backup.service: Supervising process 1625 which is not our child. We'll most likely not notice when it exits.
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
Using O_CLOEXEC by default, and returning Fd handles to
ensure they get dropped on bail!() or panic!() if the RawFds
aren't used yet.
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
Now that we let hyper shutdown gracefully we need an owned
version of the listening socket to prevent it from closing
before running the reload preparations.
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
Reloadable resources are now 'Reloadable' instead of
'ReexecContinue'.
The struct handling the reload is a 'Reloader', not a
'ReexecStore'.
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
Proxy and daemon for now just want to handle reload via
`SIGHUP`, so provide a helper creating the signalfd stream
doing that - this is simply a filtered stream which passes
the remaining signals through, so it can be used exactly
like the signalfd stream could before to add more signals.
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
If the listening socket goes into some error state we'll get
std::io::Errors rather than higher level errors from the
native_tls::TlsAcceptor, those are usually fatal. (Ran into
this after performing a shutdown() on the file descriptor,
after which the future just endlessly loops in accept().)
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
This should have the same memory layoutas a simple `fn()`
with NULL function pointers mapping to None.
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
- make Client creation reusable
- add helper to create a CApiSocket for any Read + Write
streams
- add drop callback (required for the above)
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
- add comment about the boxed file descriptor (and others)
- reexport SigSet and Signal for convenience
- remove Result wrapper (only used once)
- enforce drop order of the PollEvented2 struct with respect
to the boxed fd
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
This provides a Stream<Item = siginfo> via nix' signalfd,
by wrapping it in tokio's PollEvented2 struct to allow
polling via tokio's event loop.
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
When using *blocking* I/O wait_for_* functions need to be
able to not read past their awaited packet.
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
This uses futures for everything which is mostly useful as a
test to see if the protocol crate's non-blocking I/O support
can handle it...
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
This is supposed to contain only the parts necessary to
communicate with the server via the proxmox backup protocol.
(Including the chunker, which is currently `include!()`d
from the main crate.
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
This is just an Async GET method, but semantically
different, so it gets its own router method.
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
We want to be able to iterate through digests of index
files, but without always having to distinguish between
dynamic and fixed types, so add a trait we can use as a
trait object.
Unfortunately the iterator needs to yield copies as
iterators cannot yield values with life times when
represented as trait objects (Box<dyn Iterator<Item = ?>>)
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
Single-method traits usually use the same name as their
method and aren't usually the 'noun' (which is usually an
implementation of them instead).
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
The protocol handler will receive chunk data plus a hash
pre-calculated by the client. It will verify the hash before
sending it up to the datastore in order to respond to the
client with an error on a mismatch, so there's no need to
recalculate the hash another time.
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
Error::from is already a function taking 1 parameter,
there's no need to wrap it with `|e| Error::from(e)`.
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
It now supports custom conversation methods, so instead of
new() we ask for a default authenticator taking a password
via with_password(). Since the password is now handled by
the now separate conversation handler, `set_credentials()`
is now called on the handler we get via `.get_handler()`.
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
Instead we just build the iterator using our
tools::fs::read_subdir iterator as follows:
Use the index range (0..0x10000) as an iterator, map() it
to yield a subdirectory iterator for each index, then
flatten it.
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
FileIterOps will provide operations on file entry iterators.
For now this provides a file type filter.
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
This wraps nix::dir::Entry with the parent and additionally
exposes the parent directory file descriptor since we need
it to perform openat().
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
Eg. if you have an x: Tied<Dir, Iterator> now you can simply
call x.next() as it'll dereference to the Iterator.
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
Tie two dependent values together, such as a nix::Dir with
its nix::dir::Iter<'a> where 'a is tied to the Dir's
lifetime, making it otherwise impossible to return them or
store them in a struct.
Alternatively we could try the 'rental' crate.
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
Wolfgang Bumiller