Merge patch series "coredump: allow for flexible coredump handling"

Christian Brauner <brauner@kernel.org> says:

In addition to the extensive selftests I've already written a
(non-production ready) simple Rust coredump server for this in
userspace:

https://github.com/brauner/dumdum.git

Extend the coredump socket to allow the coredump server to tell the
kernel how to process individual coredumps. This allows for fine-grained
coredump management. Userspace can decide to just let the kernel write
out the coredump, or generate the coredump itself, or just reject it.

When the crashing task connects to the coredump socket the kernel will
send a struct coredump_req to the coredump server. The kernel will set
the size member of struct coredump_req allowing the coredump server how
much data can be read.

The coredump server uses MSG_PEEK to peek the size of struct
coredump_req. If the kernel uses a newer struct coredump_req the
coredump server just reads the size it knows and discard any remaining
bytes in the buffer. If the kernel uses an older struct coredump_req
the coredump server just reads the size the kernel knows.

The returned struct coredump_req will inform the coredump server what
features the kernel supports. The coredump_req->mask member is set to
the currently know features.

The coredump server may only use features whose bits were raised by the
kernel in coredump_req->mask.

In response to a coredump_req from the kernel the coredump server sends
a struct coredump_ack to the kernel. The kernel informs the coredump
server what version of struct coredump_ack it supports by setting struct
coredump_req->size_ack to the size it knows about. The coredump server
may only send as many bytes as coredump_req->size_ack indicates (a
smaller size is fine of course). The coredump server must set
coredump_ack->size accordingly.

The coredump server sets the features it wants to use in struct
coredump_ack->mask. Only bits returned in struct coredump_req->mask may
be used.

In case an invalid struct coredump_ack is sent to the kernel a non-zero
u32 integer is sent indicating the reason for the failure. If it was
successful a zero u32 integer is sent.

In the initial version the following features are supported in
coredump_{req,ack}->mask:

* COREDUMP_KERNEL
  The kernel will write the coredump data to the socket.

* COREDUMP_USERSPACE
  The kernel will not write coredump data but will indicate to the
  parent that a coredump has been generated. This is used when userspace
  generates its own coredumps.

* COREDUMP_REJECT
  The kernel will skip generating a coredump for this task.

* COREDUMP_WAIT
  The kernel will prevent the task from exiting until the coredump
  server has shutdown the socket connection.

The flexible coredump socket can be enabled by using the "@@" prefix
instead of the single "@" prefix for the regular coredump socket:

  @@/run/systemd/coredump.socket

will enable flexible coredump handling. Current kernels already enforce
that "@" must be followed by "/" and will reject anything else. So
extending this is backward and forward compatible.

Link: https://lore.kernel.org/20250603-work-coredump-socket-protocol-v2-0-05a5f0c18ecc@kernel.org
* patches from https://lore.kernel.org/20250603-work-coredump-socket-protocol-v2-0-05a5f0c18ecc@kernel.org:
  selftests/coredump: add coredump server selftests
  tools: add coredump.h header
  selftests/coredump: cleanup coredump tests
  selftests/coredump: fix build
  coredump: allow for flexible coredump handling

Signed-off-by: Christian Brauner <brauner@kernel.org>