[socket] Modify the error reported when writing a very large datagram packet

When attempting to write a datagram packet larger than the socket's
entire capacity, we previously returned ZX_ERR_SHOULD_WAIT. This is
misleading: the wait can never be satisfied, as socket capacity cannot
grow. This CL changes it to report ZX_ERR_OUT_OF_BOUNDS instead.

ZX-2553 #done

Test: modified /boot/test/core/socket accordingly, and added `k ut
mbuf` tests

Change-Id: I40ac9e8ec0c1e228ee19219499e951e7e6870a8a

[docs][syscalls] Add TODOs about documenting system call rights

ZX-2399 #comment

Test: no functional change
Change-Id: I1f285c677e2444fcf1d720d7ecbd9f490c303962

[docs][syscalls] Clarify ZX_ERR_NO_MEMORY as eventually going away

Not touched:
vmo_op_range.md
vmo_set_size.md
vmo_write.md
The current text is ok for these.

ZX-2303 #comment patch

Tested: Just doc changes, read before/after.

Change-Id: I2c6e1039e2b205e9cc377b9888755fa7ac676c78

[docs] Improve zx_socket_write docs

Clarify that short writes are still considered a success and that
callers should wait when they receive ZX_ERR_SHOULD_WAIT.

Change-Id: I322db7a5ebef3925fe95fa5c9f7568e7d569ffd6

[syscalls] Update parameter types and names

This CL updates the parameter types and names for a number of syscalls
to match the Zircon System API rubric. There's more work to do to make
all the syscalls conform to the rubric, but this change is a start.

Change-Id: I218ac5e7e0cbd80a8c69fef7891ad40b78b6b407

[zx] Magenta -> Zircon

The Great Renaming is here!

Change-Id: I3229bdeb2a3d0e40fb4db6fec8ca7d971fbffb94

[kernel][socket] control plane

Implement an out-of-band socket control plane to be used for
configuring a socket connection. Typical POSIX functions (like
bind and listen) can be implemented by sending an out-of-band
message to the other side of the socket using MX_SOCKET_CONTROL.

Change-Id: Ic1a8d1a2fbf8cc9b1e815bb1abd6fdb0ccf3271a

Allow shutting down sockets for read, write or both

This replaces half-closing sockets with calls to shut down a socket
endpoint for read, write, or both as a separate operation from closing.
Shutting an endpoint down for read disables all further reads from that
endpoint (past data already buffered) and all further writes into the
peer endpoint. Shutting an endpoint down for writing disables all
further writing into that endpoint and all reader from the peer
endpoint (past data already buffered).

Shutting a socket endpoint down no longer raises the PEER_CLOSED
signal on the peer endpoint. Separate signals are raised when reading and
writing are disabled and the PEER_CLOSED signal is only raised when the peer
is actually closed.

In mxio, this fully implements the posix shutdown(2) calls for local sockets
created with the socketpair(2) call. Remote sockets support shutdown(SHUT_WR)
using their existing mechanism. Once remote sockets are migrated to use an
mx_socket directly, they can re-use this machinery to fully support partial
shutdowns.

This is a reland of 2f2e0a0f8f1cefdd9505d5ba18d4c8aeaacb4188 which was reverted
in a8fa45f268598f75478464f568d543ad3e58a0cb, with the following changes:

mxio/remotesocket.c updated to use the new shutdown bits for reading
and writing

The READ_DISABLED signal is now only raised when reading is disabled in
that direction on the socket and the socket is drained in that
direction.

New tests added for blocking recv() and send() and fixes in the mxio
bindings to behave correctly for these cases.

Change-Id: I455855d29a6a79d2e98ad47b37222965332b9b98

Revert "Allow shutting down sockets for read, write or both"

This reverts commit 2f2e0a0f8f1cefdd9505d5ba18d4c8aeaacb4188.

NET-129 #done

Change-Id: Iac5291251a8b3f28e77169aa8e59677dd3568f4c

Allow shutting down sockets for read, write or both

This replaces half-closing sockets with calls to shut down a socket
endpoint for read, write, or both as a separate operation from closing.
Shutting an endpoint down for read disables all further reads from that
endpoint (past data already buffered) and all further writes into the
peer endpoint. Shutting an endpoint down for writing disables all
further writing into that endpoint and all reader from the peer
endpoint (past data already buffered).

Shutting a socket endpoint down no longer raises the PEER_CLOSED
signal on the peer endpoint. Separate signals are raised when reading and
writing are disabled and the PEER_CLOSED signal is only raised when the peer
is actually closed.

In mxio, this fully implements the posix shutdown(2) calls for local sockets
created with the socketpair(2) call. Remote sockets support shutdown(SHUT_WR)
using their existing mechanism. Once remote sockets are migrated to use an
mx_socket directly, they can re-use this machinery to fully support partial
shutdowns.

Change-Id: If73a3550a97a87785be32c182ca4fb44abb98fc5

[syscalls][socket] Update docs and test for mx_socket_write() error conditions

Fix the docs: (MX_)ERR_TOO_BIG does not exist as an error code.

Furthermore, mx_socket_write() doesn't seem to have a code path for
the exact condition of "*size* is larger than the largest allowable
size for socket messages".

The actual error behaviour is already covered in the docs under
"DESCRIPTION".

Change the test to be stricter: Test the status code, and test whether
'written' was or wasn't filled out.

Change-Id: I257695477ac673b24cb108728f96a2e207942733

[docs][errors] Rename all the errors in docs/ to MX_OK and MX_ERR_*

Change-Id: I62451456a83145760c6454a619d4c699ce8e9017

[syscalls][socket] add MX_SOCKET_DATAGRAM support

Change-Id: Ifa9a2e7f5d40b2d04f7a6467d451df2116376d47

[magenta][errors] Rename ERR_REMOTE_CLOSED to ERR_PEER_CLOSED

This error, at least at the Magenta syscall level, is really about
peers. This new name aligns with the MX_*_PEER_CLOSED signals. And
REMOTE implies a distance or asymmetry that isn't necessarily true.

This leaves a compatibility alias for the old name in place until the
rest of Fuchsia can be cleaned up.

Change-Id: Icf95003b2813270d865c6876b62e21d9f5b8c0f6

[syscalls] Uniformally use "options" rather than "flags" as a syscall parameter name

With the exception of the VMAR syscalls which really do want to talk
about map_flags.

Part of MG-559

Change-Id: I51673de824b4072990f01724a3bf522a9a84f4f4

[docs][syscalls] Fix all the manpages

Change-Id: Ie9ba25060f3331de8d9c178caf6332fb48fc80f5

[socket] Allow actual to be NULL in mx_socket_read.

Users should be allowed to pass NULL for actual, in the same way as
mx_socket_write.

Change-Id: I9678be6efe77a4e44949e86d05fcba2eba7d192c

Remove mx_size_t and mx_ssize_t types

mx_ssize_t is no longer used. mx_size_t is superfluous and
can be replaced by the standard C type size_t.

Change-Id: I2137af6f166099d2e8aaca4fab63ff101f2a43ac

[docs] Fix typo in socket write manpage

Change-Id: Icb0dfa056fa3ea4c83359402f70a6e2245d9e5ca

[docs] Add socket man pages

Change-Id: Icb7debc576fc46777f38a7e47170a19eed854062