Fix test-i2c leak, and mctp_pktbuf storage alignment

- Ensure mctp_pktbuf storage is correctly aligned
- Deallocate mctp and i2c instances to avoid failure with asan.

These previously succeeded in

Fix test-i2c leak, and mctp_pktbuf storage alignment

- Ensure mctp_pktbuf storage is correctly aligned
- Deallocate mctp and i2c instances to avoid failure with asan.

These previously succeeded in CI so are both fixed in this commit.

Fixes: e5b941d9d764 ("i2c: Add binding for MCTP over I2C transport")
Fixes: 4a09e1dc4883 ("core: Reuse buffers for tx, allow message pools")
Change-Id: I747bfff6faf3a5b0a982ae266bcef02ecbc4ee8a
Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>

show more ...

control: Add basic MCTP Control Protocol handler

This will respond to the 4 mandatory MCTP Control Protocol commands.
Applications can register supported types using mctp_control_add_type().

Change

control: Add basic MCTP Control Protocol handler

This will respond to the 4 mandatory MCTP Control Protocol commands.
Applications can register supported types using mctp_control_add_type().

Change-Id: Ia904bcbe118626adf9254ffa71dd8e17fbdfc9b7
Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>

show more ...

core: Add allocated tag expiry, mctp_set_now_op()

Allocated tags expire after 6 seconds, requiring a time source.
The default will use clock_gettime(), though implementations can set
MCTP_DEFAULT_CL

core: Add allocated tag expiry, mctp_set_now_op()

Allocated tags expire after 6 seconds, requiring a time source.
The default will use clock_gettime(), though implementations can set
MCTP_DEFAULT_CLOCK_GETTIME=0 and use mctp_set_now_op() for custom
behavior.

Change-Id: Iaacf500bcd458c987f4b79c698bac9ae5ad5a73e
Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>

show more ...

build: Generate libmctp-sizes.h

This can be used to allocate static storage for a struct mctp used with
heapless mctp_setup().

core.c internal data structures are moved to core-internal.h so that
s

build: Generate libmctp-sizes.h

This can be used to allocate static storage for a struct mctp used with
heapless mctp_setup().

core.c internal data structures are moved to core-internal.h so that
sizeof(struct mctp) can be compiled without linking other objects.

Change-Id: I72dcd46ef11d6f4b4f5ba1c9ae6c95e40dda40f8
Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>

show more ...

core: Add mctp_message_tx_request() function

This allocates a tag for messages sent with TO bit set.

Change-Id: Ia8403a06aa449e0218a30edf5ad69781c70d7c52
Signed-off-by: Matt Johnston <matt@codecons

core: Add mctp_message_tx_request() function

This allocates a tag for messages sent with TO bit set.

Change-Id: Ia8403a06aa449e0218a30edf5ad69781c70d7c52
Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>

show more ...

core: Allow running without heap allocations

Replace buses with a static array, and add mctp_setup/mctp_cleanup
to use an existing struct.

__mctp_alloc() and __mctp_free() no longer need to be func

core: Allow running without heap allocations

Replace buses with a static array, and add mctp_setup/mctp_cleanup
to use an existing struct.

__mctp_alloc() and __mctp_free() no longer need to be functional, though
are used as defaults if m_msg_alloc/m_msg_free are not provided.

Change-Id: Ib5826836efb59e416410b25f138cbaf5931278f3
Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>

show more ...

core: Reuse buffers for tx, allow message pools

Use new m_msg_alloc/m_msg_free operations for whole-message
MCTP buffers. m_realloc is no longer used, instead the maximum
sized buffer is allocated f

core: Reuse buffers for tx, allow message pools

Use new m_msg_alloc/m_msg_free operations for whole-message
MCTP buffers. m_realloc is no longer used, instead the maximum
sized buffer is allocated for each reassembly.
This allows applications to keep a pool of MCTP message buffers.

Don't create a queue of packets to transmit, instead reuse a single
binding-provided tx_storage buffer for each transmitted packet, which
can be static for bindings that have a known maximum packet size.

Asynchronous users/bindings can no longer rely on the core for queueing
TX packets, instead they should test mctp_is_tx_ready() prior to calling
mctp_message_tx(). The stack will return -EBUSY from mctp_message_tx()
if there is already a message pending to send.

Bindings must be updated to add the tx_storage member, and the core will
no longer free packets passed to mctp_bus_rx().

Change-Id: I2598bb91026ccef01b268c52b06c0f8e20bebb1e
Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>

show more ...

core: const input arg for mctp_pktbuf_push()

Change-Id: I321c81ba86bf4d4dcd8f242d25ce4c9c2487b736
Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>

pcap: Use SLL2 linktype for captures

This means we can use wireshark on our pcaps.

Signed-off-by: Rashmica Gupta <rashmica@linux.ibm.com>
Change-Id: I215af00e064d352e41d0a8eb87d5a30582998676

format: reformat with clang-format

Reformat everything with clang-format and remove .clang-ignore and
custom code formatter.

Signed-off-by: Patrick Williams <patrick@stwcx.xyz>
Change-Id: I93dfca67

format: reformat with clang-format

Reformat everything with clang-format and remove .clang-ignore and
custom code formatter.

Signed-off-by: Patrick Williams <patrick@stwcx.xyz>
Change-Id: I93dfca67948d3fbcc0f05ece4ae3e4dd9495ab57

show more ...

core: Define return value behaviours for binding Tx callbacks

Binding Tx callbacks must return 0 upon success or a negative error code
on failure. Some error codes invoke specific error handling beh

core: Define return value behaviours for binding Tx callbacks

Binding Tx callbacks must return 0 upon success or a negative error code
on failure. Some error codes invoke specific error handling behaviours.
If a binding Tx callback returns the following negative error codes:

1. EMSGSIZE: The packet whose transmission failed is dequeued from the
transmit queue and dropped, as it will never be successfully
transmitted

2. EBUSY: The packet whose transmission failed remains queued for a
subsequent attempt.

This prevents Tx queue stalls for bindings such as astlpc where
reinitialisation can renegotiate the Tx buffer size to a lower value
than the size of packets already in the Tx queue. Previously the
implementation in core failed to discard the packet from the binding Tx
queue if transmission of the head packet was not possible.

Signed-off-by: Andrew Jeffery <andrew@aj.id.au>
Change-Id: I7dde22bd7340f2c028adf5112d7d4ab78cac66a6

show more ...

core: Add TX/RX API that exposes message tag and tag owner

MCTP received packets can carry a message tag and tag owner bit
which is set by a remote MCTP endpoint. This can be used by the
remote MCTP

core: Add TX/RX API that exposes message tag and tag owner

MCTP received packets can carry a message tag and tag owner bit
which is set by a remote MCTP endpoint. This can be used by the
remote MCTP endpoint to track the responses. Thus, libmctp should
provide a mechanism for the upper layer MCTP applications to
respond with the same message tag.

This patchset extends TX and RX API with message tag and
tag owner bits.

Signed-off-by: Sumanth Bhat <sumanth.bhat@linux.intel.com>
Change-Id: I6d07eafa86c653abdd4313ab7cc77e5a93124477

show more ...

libmctp: Add mctp_unregister_bus()

Allow for safe cleanup of busses associated with bindings when the
binding has work to do in the tear-down path.

Signed-off-by: Andrew Jeffery <andrew@aj.id.au>
C

libmctp: Add mctp_unregister_bus()

Allow for safe cleanup of busses associated with bindings when the
binding has work to do in the tear-down path.

Signed-off-by: Andrew Jeffery <andrew@aj.id.au>
Change-Id: I7cc45e30d778554cfa33c49a47fe237b23fe5431

show more ...

core: Add infrastructure for packet captures

A callback and user data member is defined to abstract out the
implementation details of capturing the packets. The captured packets
are in DSP0236 form,

core: Add infrastructure for packet captures

A callback and user data member is defined to abstract out the
implementation details of capturing the packets. The captured packets
are in DSP0236 form, i.e. no binding metadata is present.

Signed-off-by: Andrew Jeffery <andrew@aj.id.au>
Change-Id: Id8234d6e8f7856c70a0ca3fe5d4cee38ec11bd8c

show more ...

astlpc: Introduce protocol v3 with integrity checks

v3 of the binding adds a CRC-32 value as a medium-specific trailer to
each packet passing over the binding interface.

The patch includes a naive

astlpc: Introduce protocol v3 with integrity checks

v3 of the binding adds a CRC-32 value as a medium-specific trailer to
each packet passing over the binding interface.

The patch includes a naive bit-shift implementation of CRC-32, we can
improve it later as necessary.

Signed-off-by: Andrew Jeffery <andrew@aj.id.au>
Change-Id: I93a95bccef30010d56e10e29b6d84554268ab7af

show more ...

libmctp: Split padding in pktbuf into header and trailer

The astlpc binding will shortly make use of a medium-specific trailer to
implement a CRC-32 integrity check over packet data. Allow for
mediu

libmctp: Split padding in pktbuf into header and trailer

The astlpc binding will shortly make use of a medium-specific trailer to
implement a CRC-32 integrity check over packet data. Allow for
medium-specific trailer data in the pktbuf in order to remove the need
for a bounce-buffer. The binding will trim the pktbuf to the payload
length before passing the instance up the stack.

Signed-off-by: Andrew Jeffery <andrew@aj.id.au>
Change-Id: Ib215a552ff218b007f7f119c5235fbc75d9ca6cb

show more ...

core: Make pkt_size and pkt_pad size_t

Fix bad signedness comparison:

```
core.c: In function ‘mctp_binding_set_tx_enabled’:
core.c:630:25: error: comparison of integer expressions of different sig

core: Make pkt_size and pkt_pad size_t

Fix bad signedness comparison:

```
core.c: In function ‘mctp_binding_set_tx_enabled’:
core.c:630:25: error: comparison of integer expressions of different signedness: ‘int’ and ‘long unsigned int’ [-Werror=sign-compare]
630 | if (binding->pkt_size < MCTP_PACKET_SIZE(MCTP_BTU)) {
| ^
cc1: all warnings being treated as errors
```

Signed-off-by: Andrew Jeffery <andrew@aj.id.au>
Change-Id: Idb91b16efa43526ea8e86fff666d0151f2987247

show more ...

core: Limit maximum size of an assembled MCTP message

If libmctp receives sequence of fragment MCTP packets and never receives
EOM packet, this will cause heap memory to grow without bounds. This
co

core: Limit maximum size of an assembled MCTP message

If libmctp receives sequence of fragment MCTP packets and never receives
EOM packet, this will cause heap memory to grow without bounds. This
commit puts an upper cap on maximum MCTP message size. This should
protect us from any malicious device trying to exploiting this.

Also, this prevents overwhelming of the device's resources.
Section 10.1.5 of DSP0236 (v1.3.1) allows configuration of
endpoints to protect its resources.

Signed-off-by: Sumanth Bhat <sumanth.bhat@linux.intel.com>
Change-Id: Id62cfab7c25b3e1ccf955f2e924844b58b4be154

show more ...

core: Return size_t for mctp_pktbuf_size()

Large packets are enabled by e.g. MTU negotiation in the astlpc binding.
Avoid truncating large packets in calls to the pktbuf APIs.

Signed-off-by: Andrew

core: Return size_t for mctp_pktbuf_size()

Large packets are enabled by e.g. MTU negotiation in the astlpc binding.
Avoid truncating large packets in calls to the pktbuf APIs.

Signed-off-by: Andrew Jeffery <andrew@aj.id.au>
Change-Id: I1244d70fe93ca7538ca09256110d648c495c6aa7

show more ...

astlpc: Introduce MTU negotiation

MTU negotiation is implemented in a backwards-compatible manner with
version 1 of the astlpc binding. Functionally, MTU negotation involves
proposing arrangements o

astlpc: Introduce MTU negotiation

MTU negotiation is implemented in a backwards-compatible manner with
version 1 of the astlpc binding. Functionally, MTU negotation involves
proposing arrangements of the Rx and Tx buffer layouts. It is assumed
that the MTU is a packet sized to fill the Tx buffer as described in the
control space.

For v1 of the binding the MTU is defined in terms of the MCTP_BTU
constant provided by libmctp.h. MCTP_BTU is used regardless of the
buffer sizes specified in the control space (which MUST describe buffers
supporting at least MCTP_BTU-sized packets).

For v2 of the binding the MTU is defined in terms of the appropriate
buffer's size field in the control space.

The sequence of events for negotiating the MTU under v2 is as follows:

1. The BMC initialises its binding, filling out the Rx and Tx buffer
properties with the largest configuration it supports.

2. The host initialises its binding, writing its maximum Rx buffer size
before sending `channel-init` to the BMC.

3. The BMC receives `channel-init`, negotiates protocol version 2 and
then validates the host's proposed buffer configuration. If the proposed
configuration is invalid (e.g. out-of-bounds values) the BMC terminates
channel initialisation leaving the channel-active bit clear and writing
the zero to the negotiated version field. If the proposal is valid, the
BMC calculates the buffer sizes according to the available constraints
and writes the chosen buffer configuration to the control region.

4. Assuming the version negotiation and buffer configuration are
successful, the BMC sets `channel-active` and notifies the host

5. The host reads `channel-active`, accepts the negotiation of v2 and
validates the buffer configuration. If the validation passes, then the
buffer configuration is the configuration used for the remainder of the
session. If validation fails then the host MUST NOT send MCTP packets
via the LPC binding until a valid buffer configuration can be
negotiated.

Change-Id: I89107593f220418d746c2d73771348ed8f7f3e87
Signed-off-by: Andrew Jeffery <andrew@aj.id.au>

show more ...

core: Support transport control commands

This change introduces a control message request handler for MCTP
bindings. If a handler is provided, transport control messages will be
forwarded to the han

core: Support transport control commands

This change introduces a control message request handler for MCTP
bindings. If a handler is provided, transport control messages will be
forwarded to the handler, otherwise they will be forwarded to the
default handler.

Change-Id: I62266d6bf2d512ec97759c0b8a3477c5e433d609
Signed-off-by: Wiktor Gołgowski <wiktor.golgowski@linux.intel.com>
[AJ: Split out general control message handler, formatting]
Signed-off-by: Andrew Jeffery <andrew@aj.id.au>

show more ...

libmctp: Add mctp_destroy()

Provide an abstraction to clean up memory to avoid false-positives with
leak sanitizers.

Signed-off-by: Andrew Jeffery <andrew@aj.id.au>
Change-Id: Id0f03ec61fcaf6eb0ed2

libmctp: Add mctp_destroy()

Provide an abstraction to clean up memory to avoid false-positives with
leak sanitizers.

Signed-off-by: Andrew Jeffery <andrew@aj.id.au>
Change-Id: Id0f03ec61fcaf6eb0ed27e8f735720f1ea7e5e8e

show more ...

libmctp: Clarify BTU-related symbols

DSP0236 defines a Baseline Transmission Unit (BTU), which dictates the
minimum supported packet payload size. The packet size is the BTU + the
packet header. The

libmctp: Clarify BTU-related symbols

DSP0236 defines a Baseline Transmission Unit (BTU), which dictates the
minimum supported packet payload size. The packet size is the BTU + the
packet header. The maximum (payload) transport unit size for a given
binding may be larger than the BTU.

Signed-off-by: Andrew Jeffery <andrew@aj.id.au>
Change-Id: I2c4013531efd572d95c5a61e4d59309490505a8f

show more ...

API: Add binding accessors to generic struct mctp_binding

Rather than have every binding include their own wrappers around
binding<->core functions, introduce an accessor to retrieve the struct
mctp

API: Add binding accessors to generic struct mctp_binding

Rather than have every binding include their own wrappers around
binding<->core functions, introduce an accessor to retrieve the struct
mctp_binding from each.

This means we no longer need the binding-specific registration
callbacks. However, we do now need a ->start callback, to allow bindings
to perform post-registration init.

Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
Change-Id: I6cee9e93f37520f85c155a0ca34017cc0675552c

show more ...

core,API: Add bridge support

This change introduces a facility to bridge messages between two
bindings.

This is implemented through a new mctp_bridge_busses() API, which
applies a new routing polic

core,API: Add bridge support

This change introduces a facility to bridge messages between two
bindings.

This is implemented through a new mctp_bridge_busses() API, which
applies a new routing policy, sending packets from one binding to the
other. This is in contrast to the current policy of dropping all
non-local packets.

To do this, the message context code needs to know both source and
destination EIDs, so add them to the mctp_msg_ctx_lookup() criteria.

Also, add a small test for bridge mode.

Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
Change-Id: If532613525ddbf81df249e26d0f23825996f7bda

show more ...