The device PM subsystem _depends_ on device, not vice-versa. Devices
only hold a reference to struct pm_device now, and initialize this
reference with the value provided in Z_DEVICE_DEFINE. This requirement
can be solved with a forward struct declaration, meaning there is no
need to include device PM headers.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
Replace unpacked in6_addr structures with raw buffers in packed icmpv6
structs, to prevent compiler warnings about unaligned access.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Replace unpacked in6_addr structures with raw buffers in net_ipv6_hdr
struct, to prevent compiler warnings about unaligned access.
Remove __packed parameter from `struct net_6lo_context` since the
structure isn't really serialized.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Replace unpacked in_addr structures with raw buffers in net_ipv4_hdr
struct, to prevent compiler warnings about unaligned access.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Replace unpacked in_addr/in6_addr structures with raw buffers in
net_arp_hdr struct, to prevent compiler warnings about unaligned
access.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Make sure that in_addr/in6_addr structure size match the respective
binary IP address size with BUILD_ASSERT.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Since drivers implement a callback based on action and not the state,
we should be using the API based on the action instead of the one based
on the state.
Signed-off-by: Flavio Ceolin <flavio.ceolin@intel.com>
Update the macro prototype to explicitly require the length of the
desired user data. Update all in-tree usage of this macro.
Signed-off-by: Jordan Yates <jordan.yates@data61.csiro.au>
Update the macro prototype to explicitly require the length of the
desired user data. Update all in-tree usage of this macro.
Signed-off-by: Jordan Yates <jordan.yates@data61.csiro.au>
The print specifier for `atomic_t` should be updated
to `%ld`, `%lu`, or `%lx` to account for the type
change of `atomic_t` to `long`.
Signed-off-by: Christopher Friedt <chrisfriedt@gmail.com>
TCP2 is no longer needed as it is the unique implementation since the
legacy one has been removed.
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
Move the structure definition into the relevant header. Rename the
access relevantly as well. It's easier to read without mss_option being
used in various places (struct and access).
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
Too long lines and indentations mostly. Let's just keep it concistent
over the file.
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
As the already existing macro K_MEM_SLAB_DEFINE results in
two variable definitions, the preceding static modifier leads to
a seemingly working solution, though linkage conflicts will occur
when the same memory slab name is used across multiple modules.
The new K_MEM_SLAB_DEFINE_STATIC macro duplicates the functionality of
K_MEM_SLAB_DEFINE with the difference that the static keywords are
internally prepended before both variable definitions.
The implementation has been tested on my Zephyr project (the build
issue faded out). The documentation has been updated altogether
with all incorrect occurences of static K_MEM_SLAB_DEFINE.
Signed-off-by: Pavel Hübner <pavel.hubner@hardwario.com>
Reason why the prority was at its lowest is unknown, but now that it may
be used to send local packets (which used to be sent right away),
it seems to affect TCP scheduling in loopback mode. Raising the prority
so it matches how it was previously (i.e. sent right away) should fix
things. (Note however that this issue was not broadly present, only
sockets.tls test seemed to be affected.)
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
Closing a connection, thus calling net_context_put() will not close a
TCP connection properly, and will leak tcp connection memory.
This is because: net_context_put calls net_context_unref which calls
net_tcp_unref which leads to unref tcp connection and thus sets
ctx->tcp to NULL. Back to net_context_put, that one finally calls
net_tcp_put: but that bails out directly since ctx->tcp is NULL.
Fixing it by inverting net_tcp_put() and net_context_unref() calls
within net_context_put().
Fixes#38598
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
On any target, running a TCP server and a net shell can show the issue:
net tcp connect local_ip port
will fail. Usally it ends up by consumming all tcp connection memory.
This is because in tcp_in(), state changes will most of the time lead to
sending SYN/ACK/etc... packets under the same thread, which will run all
through net_send_data(), back to tcp_in(). Thus a forever loop on SYN ->
SYN|ACK -> SYN -> SYN|ACK until tcp connection cannot be allocated
anymore.
Fixing it by scheduling any local packet to be sent on the queue.
Fixes#38576
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
ipv6_prefix_find() wrongly tests if the unicast address is in use
instead of the prefix. This has two implications:
- The function can return an expired prefix to net_if_ipv6_prefix_add(),
which will do nothing more to enable it (since it assumes that it is
already enabled). As a result, the prefix will not be used by the
rest of the stack due to prefix->is_used being false.
- ipv6_prefix_find() loops using a bound of NET_IF_MAX_IPV6_PREFIX, but
the size of the unicast[] array is defined by NET_IF_MAX_IPV6_ADDR.
This could lead to an out-of-bound access if NET_IF_MAX_IPV6_ADDR is
smaller than NET_IF_MAX_IPV6_PREFIX.
Signed-off-by: Florian Vaussard <florian.vaussard@gmail.com>
With the introduction of `EXPERIMENTAL` and `WARN_EXPERIMENTAL` in
Zephyr all subsys/net and drivers/ethernet/Kconfig.e1000 settings
having `[EXPERIMENTAL]` in their prompt has has been updated to include
`select EXPERIMENTAL` so that developers can enable warnings when
experimental features are enabled.
The following settings has EXPERIMENTAL removed as they are considered
mature:
- NET_OFFLOAD
- NET_PROMISCUOUS_MODE
Signed-off-by: Torsten Rasmussen <Torsten.Rasmussen@nordicsemi.no>
Add __printf_like modifier to validate strings used by shell.
Fixing warnings triggered by this change.
Signed-off-by: Krzysztof Chruscinski <krzysztof.chruscinski@nordicsemi.no>
If data for `context_sendto()` was provided in a form of
`struct msghdr` (for instance via `sendmsg()`), it was not verified that
the provided data would actually fit into allocated net_pkt. In result,
and error could be returned in case the provided data was larger than
net_pkt allows.
Fix this, by verifying the remaining buffer length when iterating over
`struct msghdr`. Once the buffer is filled up, break the loop. In
result, functions like `sendmsg()` will return the actual length of data
sent instead of an error.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
As per #38352, we would like to start building out PTP (IEEE 1588)
support for superset of gPTP functionality in Zephyr. This is the first
step to abstract away some key interfaces from NET_GPTP umbrella to
NET_L2_PTP.
Signed-off-by: Alex Sergeev <asergeev@carbonrobotics.com>
RFC 2460 Sec. 5 requires that a ICMPv6 Time Exceeded message is sent
upon reassembly timeout, if we received the first fragment (i.e. the one
with a Fragment Offset of zero).
Implement this requirement.
Signed-off-by: Florian Vaussard <florian.vaussard@gmail.com>
The purpose of shift_packets() is to make room to insert one fragment in
the list. This is not what it does currently, potentially leading to
-ENOMEM even if there is enough free room.
To see the current behaviour, let's assume that we receive 3 fragments
in reverse order:
- Frag3(offset = 0x40, M=0)
- Frag2(offset = 0x20, M=1)
- Frag1(offset = 0x00, M=1)
After receiving Frag3 and Frag2, pkt[] will look like:
.-------.-------.-------.
| Frag2 | Frag3 | NULL |
| 0x20 | 0x40 | |
'-------'-------'-------'
pkt[0] pkt[1] pkt[2]
When receiving Frag1, shift_packets(pos = 0) is called to make some room
at position 0. It will iterate up to i = 2 where there is a free
element. The current algorithm will try to shift pkt[0] to pkt[2], which
is indeed impossible but also unnecessary. It is only required to shift
pkt[0] and pkt[1] by one element in order to free pkt[0] to insert
Frag1.
Update the algorithm in order to shift the memory only by one element.
As a result, the ENOMEM test is only simpler: as long as we encounter
one free element, we are guaranteed that we can shift by one element.
Also assign a NULL value to the newly freed element since memmove() only
copy bytes.
Signed-off-by: Florian Vaussard <florian.vaussard@gmail.com>
Currently net_ipv6_handle_fragment_hdr() performs 2 distinct tests: it
checks the M-bit of the most recent fragment to decide if we can proceed
with the reassembly. Then it performs some sanity checks which can lead
to dropping the whole packet if not successful.
The test on the M-bit assumes that fragments arrive in order. But this
will fail if packets arrive out-of-order, since the last fragment can
arrive before some other fragments. In that case, we proceed with the
reassembly but it will fail because not all the fragments have been
received.
We need a more complete check before proceeding with the reassembly:
- We received the first fragment (offset = 0)
- All intermediate fragments are contiguous
- The More bit of the last fragment is 0
Since these conditions can also detect a malformed fragmented packet, we
can replace the existing sanity check that is performed before
reassembly. As a bonus, we can now detect and rejected overlapping
fragments, since this can have some security issues (see RFC 5722).
Signed-off-by: Florian Vaussard <florian.vaussard@gmail.com>
Currently we only store the fragment offset. But in some cases it might
be necessary to also inspect the M-bit (More Fragment) of all received
fragments.
Modify the semantics of the field to store all the flags, rename the
setter to account for this change, and add a getter for the M-bit.
Signed-off-by: Florian Vaussard <florian.vaussard@gmail.com>
The special handling of the 1st fragment in unnecessary, since it will
be correctly handled even without it. Moreover it causes some corner
cases, like a single packet with a fragment header (M=0), to be
incorrectly handled since the reassembly code is skipped.
Remove the special handling of the 1st fragment to fix these problems.
Signed-off-by: Florian Vaussard <florian.vaussard@gmail.com>
Currently the requirement of the length being a multiple of 8 is not
tested for the first fragment, since the first fragment takes a
different path due to the goto.
Move the test earlier in the process, so that it is performed on all
fragments, including the first one.
Signed-off-by: Florian Vaussard <florian.vaussard@gmail.com>
If we have less fragments than what can be stored in the reassembly
array, some loops will blindly dereference NULL pointers.
Add checks for NULL pointers when necessary and exit the loop.
Signed-off-by: Florian Vaussard <florian.vaussard@gmail.com>
Currently the stack is limited to a maximum of 2 incoming fragments per
packet. While this can be enough in most cases, it might not be enough
in other cases.
Make this value configurable at build time.
Signed-off-by: Florian Vaussard <florian.vaussard@gmail.com>
Currently prev_hdr_offset always equals 6, which is the offset of
the nexthdr field in the IPv6 header. This value is used to overwrite it
when removing an IPv6 Fragment header, so it will work as long as there
is no other Extension header between the IPv6 header and the Fragment
header.
However this does not work in the other cases: the nexthdr field of the
IPv6 header will be overwritten instead of the nexthdr field of the last
Extension header before the Fragment, leading to unwanted results.
Update prev_hdr_offset so that it always point to the nexthdr field of
the previous header, either the IPv6 header or an Extension header.
Signed-off-by: Florian Vaussard <florian.vaussard@gmail.com>
The current validation code waits to process the header before rejecting
it, while some checks can be already enforced when reading the nexthdr
field of the previous header.
The main problem is a wrong pointer field in the resulting ICMPv6 error
message: the pointer should have the offset of the invalid nexthdr
field, while currently it will the offset the invalid header.
To solve that problem, reorganize the loop in two parts: the first
switch validates nexthdr, while the second switch processes the current
header. This allows to reject invalid nexthdr earlier.
The check for duplicated headers is also generalized, so that we can
catch other kind of headers (like the Fragment header).
Signed-off-by: Florian Vaussard <florian.vaussard@gmail.com>
By definition, NET_IPV6_NEXTHDR_NONE is void. So we must stop processing
before trying to read any data, since we will start reading values that
are outside the Extension Header (likely the payload, if any).
Signed-off-by: Florian Vaussard <florian.vaussard@gmail.com>
When an unknown option is encountered, an ICMPv6 error message must be
sent in some cases. The message contains a pointer field, which must be
the offset to the unknown option. Currently the offset is computed from
the beginning of the option list, while it should be computed with
respect to the beginning of the IPv6 header.
Record the offset when reading the option type and pass it later to
ipv6_drop_on_unknown_option() to correctly set the pointer field. Also
rename the argument in ipv6_drop_on_unknown_option() to make the
purpose more clear.
Signed-off-by: Florian Vaussard <florian.vaussard@gmail.com>
Currently PADN data are not skipped, which results in the stack to think
that the next header starts in the middle of the padding. We have to
skip the bytes before going on.
Also clarify the PAD1 does not have any length field.
Signed-off-by: Florian Vaussard <florian.vaussard@gmail.com>
The current names are confusing. Indeed "nexthdr" if the type of the
header currently processed, while "next_nexthdr" is the nexthdr field of
the current header.
Rename them to improve readability and make it less error-prone.
Signed-off-by: Florian Vaussard <florian.vaussard@gmail.com>
ICMPv6 error messages are not sent (on native_posix) because the first
net_pkt_write() returns an error.
pkt has just been allocated using net_pkt_alloc_with_buffer(). Trying to
write an empty packet in overwrite mode will result in an error. There
is no need to be in overwrite mode, since we want to write the LL
src/dst addresses at the beginning.
Signed-off-by: Florian Vaussard <florian.vaussard@gmail.com>
net_packet_socket_input() was changed to hardcode the return of
NET_CONTINUE and that caused a segmentation fault/crash in
net_core/process_data(), in cases when pkt was unreferred and
NET_OK was returned from net_conn_input()
This happened with socket combo of: AF_PACKET+SOCK_RAW+IPPROTO_RAW.
Signed-off-by: Jani Hirsimäki <jani.hirsimaki@nordicsemi.no>
The nbr_lock var actually depends on CONFIG_NET_IPV6_NBR_CACHE
(not CONFIG_NET_IPV6_ND), so move its initialization call.
Signed-off-by: Stancu Florin <niflostancu@gmail.com>
