The manpage for recvmsg says -1 will be returned on error, Zero indicates an
"orderly shutdown", presumably only in case of stream sockets.
Further, UNIX Network Programming, Vol 1 says ".. a return value of 0 from
recvfrom is acceptable for a datagram protocol"
Such packets have been observed in the wild, aimed at PTP's multicast
address and port, possibly related to malformed management queries.
Patch to properly check return from recvmesg and not trigger the fault
codepath. Instead, such packets are treated as "Bad Message" the same as
non-zero but still too-short UDP payloads.
Signed-off-by: David Mirabito <davidjm@arista.com>
Add global option for the hardware time stamp setting.
The function could:
Normally set the filters as the PTP daemon require.
Check that the filters are proper but do not change them.
Full, set the RX filter to all and the TX filter as the PTP daemon require.
[ RC: added missing extern keyword and fixed indentation. ]
Signed-off-by: Erez Geva <erez.geva.ext@siemens.com>
Signed-off-by: Erez Geva <ErezGeva2@gmail.com>
When compiling with "-O1 -D_FORTIFY_SOURCE=1", gcc throws a false positive
warning. The code is not performance sensitive, so this patch silences
the warning by initializing the offending local variable explicitly.
Signed-off-by: Richard Cochran <richardcochran@gmail.com>
The 1588 standard defines one step operation for both Sync and
PDelay_Resp messages. Up until now, hardware with P2P one step has
been rare, and kernel support was lacking. This patch adds support of
the mode in anticipation of new kernel and hardware developments.
Signed-off-by: Richard Cochran <richardcochran@gmail.com>
Convert a software timestamp to the internal tmv_t representation at
the earliest possible opportunity, to match the behaviour for hardware
timestamps.
Signed-off-by: Michael Brown <mbrown@fensystems.co.uk>
Convert a hardware timestamp to the internal tmv_t representation at
the earliest possible opportunity. This allows us to:
- eliminate multiple redundant calls to timespec_to_tmv()
- use tmv_add() instead of open-coded manipulation of a struct
timespec in ts_add()
- use tmv_to_Timestamp() instead of open-coded manipulation of a
struct timespec and struct Timestamp in ts_to_Timestamp()
- use tmv_is_zero() instead of open-coded manipulation of a struct
timespec in msg_sots_valid()
Signed-off-by: Michael Brown <mbrown@fensystems.co.uk>
When a network interface doesn't support the most specific PTP filter
(e.g. it can timestamp either all or no received packets), it is
expected that the driver will report a different filter than was
requested. Turn the warnings into debug messages to avoid confusing
the users.
Signed-off-by: Miroslav Lichvar <mlichvar@redhat.com>
The current implementation of ptp4l always assumes 6 octets MAC
address, which is correct for Ethernet interfaces but not for IPoIB
interfaces (that have 20 octets MAC), therefore running ptp4l over
IPoIB interface does not function correctly.
In Infiniband, every interface has three identifiers:
GUID, GID, and LID.
The GUID is similar in concept to a MAC address. From RFC4392:
The EUI-64 portion of a GID is referred to as the Global Unique
Identifier (GUID) and is the only persistent identifier of a port.
Therefore, to support IPoIB interfaces, the GUID of the port should
be used instead of the MAC.
This patch checks the interface type before creating the clock identity,
for Infiniband ports, it retrieves the GUID of the port using sysfs
and use it to create the clock identity.
sysfs method was chosen since the GUID is the 6 lsb bytes of
the 20 byte device address, and SIOCGIFHWADDR ioctl call returns
the 14 msb bytes of the device address, so it is not possible to
get the GUID using SIOCGIFHWADDR ioctl call.
[ RC: fixed trivial coding style error, space after switch keyword. ]
Signed-off-by: Feras Daoud <ferasda@mellanox.com>
Reviewed-by: Alex Vesker <valex@mellanox.com>
The clock module will want to know the interface indexes, in order to
implement link monitoring. However, the clock does not open any sockets
directly. This helper function lets us keep the clock module free of
socket level code.
Signed-off-by: Richard Cochran <richardcochran@gmail.com>
In the last years there are several media streaming standards
evolving that are relying on PTP. These standards make requirements
about the DSCP priority of PTP messages. This patch introduces two
new configuration options 'dscp_event' and 'dscp_general' to address
that issue and to be able to set the DSCP priority separately for
PTP event messages and PTP general messages.
Signed-off-by: Henry Jesuiter <henry.jesuiter@alcnetworx.de>
Signed-off-by: Richard Cochran <richardcochran@gmail.com>
The raw Ethernet transport code invented its own way of storing the MAC
address into our "struct address" data structure. However, this private
format is incompatible with the sockaddr_ll returned from the networking
stack. This patch converts the code to use the proper format.
Signed-off-by: Richard Cochran <richardcochran@gmail.com>
The current implementation fetches a transmit time stamp by polling on the
socket with pollfd.events set to zero, and then checking if POLLERR has
been returned by the kernel in pollfd.revents. This has the unfortunate
side effect of sleeping in poll() for the entire time out duration,
regardless of when the error queue becomes readable.
Linux kernel version 3.10 introduced a new socket option that allows
polling for transmit time stamps explicitly, waking the process as soon as
a time stamp becomes available. This patch enables the socket option,
falling back to the old behavior if necessary.
Suggested-by: Joe Schaack <jschaack@xes-inc.com>
Signed-off-by: Richard Cochran <richardcochran@gmail.com>
Copy the address directly to struct sockaddr_in or sockaddr_in6 instead
of sockaddr as Coverity doesn't seem to understand the union and reports
a buffer overflow.
Signed-off-by: Miroslav Lichvar <mlichvar@redhat.com>
This was broken in commit e804e6, ifa_addr is a pointer to sockaddr,
it shouldn't be referenced for the memcpy call.
Acked-by: Jiri Benc <jbenc@redhat.com>
This modifies all transports to use a new common address type, struct
address. This address is stored in a ptp_message for all received messages.
For sending, the "default" address is used with the default sending
functions, transport_send and transport_peer. The default address depends on
the transport; it's supposed to be the multicast address assigned by the
transport specification.
Later, a new transport_sendto function will be implemented that sends to the
address contained in the passed ptp_message.
Signed-off-by: Jiri Benc <jbenc@redhat.com>
This patch cleans up the error message for the Tx timestamp polling, and adds a
line which indicates where the bug might have originated. This should help a
user more easily diagnose what went wrong.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Because of packet reordering that can occur in the network, in the
hardware, or in the networking stack, a follow up message can appear
to arrive in the application before the matching sync message. As this
is a normal occurrence, and the sequenceID message field ensures
proper matching, the ptp4l program accepts out of order packets.
This patch adds an additional check using the software time stamps
from the networking stack to verify that the sync message did arrive
first. This check is only useful if the sequence IDs generated by
the master might possibly be incorrect.
Signed-off-by: Richard Cochran <richardcochran@gmail.com>
This patch modifies sk_receive in order to use poll() on POLLERR instead of the
tryagain loop as this resolves issues with drivers who do not return timestamps
quickly enough. It also resolves the issue of wasting time repeating every
microsecond. It lets the kernel sleep our application until the data or timeout arrives.
This change also replaces the old tx_timestamp_retries config value with
tx_timestamp_timeout specified in milliseconds (the smallest length of time poll
accepts). This does have the side effect of increasing the minimum delay before
missing a timestamp by up to 1ms, but the poll should return sooner in the
normal case where a packet timestamp was not dropped.
This change vastly improves some devices and cleans the code up by simplifying a
race condition window due to drivers returning tx timestamp on the error queue.
[ RC - removed the unused 'try_again' variable. ]
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
The Linux kernel supports a hardware time stamping mode that allows
sending a one step sync message. This commit adds support for this mode
by expanding the time stamp type enumeration. In order to enable this
mode, the configuration must specify both hardware time stamping and set
the twoStepFlag to false.
We still do not support the one step peer delay request mechanism since
there is neither kernel nor hardware support for it at this time.
Signed-off-by: Richard Cochran <richardcochran@gmail.com>
this patch changes sk_interface_phc to sk_get_ts_info, by allowing the function
to store all the data returned by Ethtool's get_ts_info IOCTL in a struct. A new
struct "sk_ts_info" contains the same data as well as a field for specifying the
structure as valid (in order to support old kernels without the IOCTL). The
valid field should be set only when the IOCTL successfully populates the fields.
A follow-on patch will add new functionality possible because of these
changes. This patch only updates the programs which use the call to perform the
minimum they already do, using the new interface.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
It seems with some cards the current default of 2 is too small, increase
the number to prevent users from having to investigate the EAGAIN error
messages.
Signed-off-by: Miroslav Lichvar <mlichvar@redhat.com>
Now that the code automatically falls back to transport-specific time
stamping, this option is no longer needed.
Signed-off-by: Richard Cochran <richardcochran@gmail.com>
Acked-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Jiri Benc <jbenc@redhat.com>
Start with the most general HWTSTAMP option. If that fails, fall back
to the option that best fits the interface's transport.
Signed-off-by: Richard Cochran <richardcochran@gmail.com>
Acked-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Jiri Benc <jbenc@redhat.com>
If the kernel supports the ioctl, but the driver does not (like igb in
kernel version 3.5), then ptp4l will incorrectly choose the system clock.
Signed-off-by: Richard Cochran <richardcochran@gmail.com>
Some of the time stamping hardware out there only recognizes layer 2
packets, and these do not work without changing the receive filter in
the SIOCSHWTSTAMP request.
Signed-off-by: Richard Cochran <richardcochran@gmail.com>
If the new ethtool operation is supported, then use it to verify that the PHC
selected by the user is correct. If the user doesn't specify a PHC and ethtool
is supported then automatically select the PHC device.
If the user specifies a PHC device, and the ethtool operation is suppported,
automatically confirm that the PHC device requested is correct. This check is
performed for all ports, in order to verify that a boundary clock setup is
valid.
The check for PHC device validity is not done in the transport because the
only thing necessary for performing the check is the port name. Handled this
in the port_open code instead.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
In some error cases, no message is logged. Now we always complain loudly
when an error occurs.
Signed-off-by: Richard Cochran <richardcochran@gmail.com>
David Mirabito via Linuxptp-devel
Miroslav Lichvar
Anders Selhammer
Erez Geva
Richard Cochran
Michael Brown
Feras Daoud
Jesuiter, Henry (ALC NetworX GmbH)
Jiri Benc
Stephan Gatzka
Jacob Keller
Jiri Bohac
Geoff Salmon