This adds support for netdevs implementing the new API that provides
`netdev_driver_t::confirm_send()`. This allows implementing netdevs
in an event based non-blocking fashion, making live of driver
developers a bit easier. In addition, `gnrc_tx_sync` will now throttle
users of `sock_udp_send()` so that they can only send datagrams as
fast as the network stack and hardware is able to send out.
Finally, this lays the groundwork to fetch TX statistics (such as
TX timestamps, reception of layer 2 ACKs/NACKs, etc.) from the network
devices.
Instead of retrieving a pointer with NETOPT_STATS, retrieve the current
data. This avoids data corruptions when reading from one thread (e.g.
the thread running the shell (ifconfig command)) while another thread
is updating it (e.g. the netif thread).
The issue affects all boards, as users typically expect the count of
TX packets and the number of TX bytes to refer to the same state. For
16 bit and 8 bit platforms even a single netstat entry can read back
corrupted.
This fixes the issue by just copying the whole netstat_t struct over
without requiring explicit locking on the user side. A multi-threaded
network stack still needs to synchronize the thread responding to
netopt_get with the thread writing to the netstat_t structure, but that
is an implementation detail no relevant to the user of the API.
Using `container_of()` to translate from `netif_t *` to the containing
`gnrc_netif_t *` adds a bit of plausibility: It will fail when `gnrc_netif_t`
stops to contain `netif_t` as member. Additionally, this silences false
positives of -Wcast-align
Initialize addr with 0 first.
Otherwise if prefix is < 64 there will be random bits of stack memory
in the bits of the address that are neither touched by ipv6_addr_set_aiid()
nor ipv6_addr_init_prefix().
Make gnrc_netif_create() block until the interface is created and
registered.
This avoids a race condition where after calling gnrc_netif_init_devs()
not all interfaces are available yet when iterating through the list
of interfaces with gnrc_netif_iter().
A lot of things break if `GNRC_NETIF_FLAGS_HAS_L2ADDR` is not set.
In order to handle router advertisements and auto-configureation,
generate a faux l2 address based on the netdev ID.
If we call `netif_register()` before we can be sure that the interface
can be configured, a 'zombie' interface remains in the list, causing
all kinds of trouble down the line.
Only call `netif_register()` if `init()` was successful.
The new `gnrc_tx_sync` module allows users of the GNRC network stack to
synchronize with the actual transmission of outgoing packets. This is directly
integrated into gnrc_sock. Hence, if `gnrc_tx_sync` is used, calls to e.g.
sock_udp_send() will block until the network stack has processed the message.
Use cases:
1. Prevent packet drop when sending at high rate
- If the application is sending faster than the stack can handle, the
message queues will overflow and outgoing packets are lost
2. Passing auxiliary data about the transmission back the stack
- When e.g. the number of required retransmissions, the transmission time
stamp, etc. should be made available to a user of an UDP sock, a
synchronization mechanism is needed
3. Simpler error reporting without footguns
- The current approach of using `core/msg` for passing up error messages is
difficult to use if other message come in. Currently, gnrc_sock is
busy-waiting and fetching messages from the message queue until the number
of expected status reports is received. It will enqueue all
non-status-report messages again at the end of the queue. This has
multiple issues:
- Busy waiting is especially in lower power scenarios with time slotted
MAC protocols harmful, as the CPU will remain active and consume
power even though the it could sleep until the TX slot is reached
- The status reports from the network stack are send to the user thread
blocking. If the message queue of the user thread is full, the network
stack would block until the user stack can fetch the messages. If
another higher priority thread would start sending a message, it
would busy wait for its status reports to completely come in. Hence,
the first thread doesn't get CPU time to fetch messages and unblock
the network stack. As a result, the system would lock up completely.
- Just adding the error/status code to the gnrc_tx_sync_t would preallocate
and reserve memory for the error reporting. That way gnrc_sock does not
need to search through the message queue for status reports and the
network stack does not need to block for the user thread fetching it.
The driver uses the netdev interface. Due to the limited
capabilities of the transceiver (32 byte FIFO and no source address in the layer2 frame),
it relies on 6LowPAN compression and adds the source address to the frame for that.
The gnrc_netif_ipv6_addr_add_internal() function unconditionally
acquires the global netif rmutex lock but doesn't release this lock on
this specific path (i.e. if gnrc_netif_ipv6_group_join_internal failed).
This can cause a deadlock as no other thread will afterwards be capable
of acquiring the netif lock.
xtimer.h must not be included, when the xtimer module is not use. Otherwise
compilation on the waspmote-pro with https://github.com/RIOT-OS/RIOT/pull/14799
will not longer work. gnrc_netif_pktq includes xtimer.h and uses xtimer, but
gnrc_netif includes gnrc_netif_pktq.h regardless of whether gnrc_netif_pktq
is used. This makes sure that gnrc_netif_pktq.h is only included when actually
used.
Replace direct accesses to sched_active_thread and sched_active_pid with
the helper functions thread_getpid() and thread_get_active(). This serves
two purposes:
1. It makes accidental writes to those variable from outside core less likely.
2. Casting off the volatile qualifier is now well contained to those two
functions
These functions are independent from GNRC and can be used by stack-agnistoc
code.
Avoid pulling in a GNRC dependency by moving those two helper functions to
`netif`.
The old function names are kept as `static inline` wrapper functions to avoid
breaking API users.
