| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This moves Resource Information Base (RIB) initialization into the
ipcp_init() function, so all IPCPs initialize a RIB. The RIB not shows
some common IPCP information, such as the IPCP name, IPCP state and
the layer name if the IPCP is part of a layer.
The initialization of the hash algorithm and layer name was moved out
of the common ipcp source because IPCPs may only know this information
after enrollment. Some IPCPs were not even storing this information.
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The UDP layer will now use a single (configurable) UDP port, default
3435. This makes it easer to allocate flows as a client from behind a
NAT firewall without having to configure port forwarding rules. So
basically, from now on Ouroboros traffic is transported over a
bidirectional <src><port>:<dst><port> UDP tunnel. The reason for not
using/allowing different client/server ports is that it would require
reading from different sockets using select() or something similar,
but since we need the EID anyway (mgmt packets arrive on the same
server UDP port), there's not a lot of benefit in doing it. Now the
operation is similar to the ipcpd-eth, with the port somewhat
functioning as a "layer name", where in UDP, the Ethertype functions
as a "layer name".
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
|
| |
The ugent email addresses are shut down, updated to Ouroboros mail
addresses.
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
| |
Happy New Year, Ouroboros!
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This adds congestion avoidance policies to the unicast IPCP. The
default policy is a multi-bit explicit congestion avoidance algorithm
based on data-center TCP congestion avoidance (DCTCP) to relay
information about the maximum queue depth that packets experienced to
the receiver. There's also a "nop" policy to disable congestion
avoidance for testing and benchmarking purposes.
The (initial) API for congestion avoidance policies is:
void * (* ctx_create)(void);
void (* ctx_destroy)(void * ctx);
These calls create / and or destroy a context for congestion control
for a specific flow. Thread-safety of the context is the
responsability of the flow allocator (operations on the ctx should be
performed under a lock).
ca_wnd_t (* ctx_update_snd)(void * ctx,
size_t len);
This is the sender call to update the context, and should be called
for every packet that is sent on the flow. The len parameter in this
API is the packet length, which allows calculating the bandwidth. It
returns an opaque union type that is used for the call to check/wait
if the congestion window is open or closed (and allowing to release
locks before waiting).
bool (* ctx_update_rcv)(void * ctx,
size_t len,
uint8_t ecn,
uint16_t * ece);
This is the call to update the flow congestion context on the receiver
side. It should be called for every received packet. It gets the ecn
value from the packet and its length, and returns the ECE (explicit
congestion experienced) value to be sent to the sender in case of
congestion. The boolean returned signals whether or not a congestion
update needs to be sent.
void (* ctx_update_ece)(void * ctx,
uint16_t ece);
This is the call for the sending side top update the context when it
receives an ECE update from the receiver.
void (* wnd_wait)(ca_wnd_t wnd);
This is a (blocking) call that waits for the congestion window to
clear. It should be stateless (to avoid waiting under locks). This may
change later on if passing the context is needed for different algorithms.
uint8_t (* calc_ecn)(int fd,
size_t len);
This is the call that intermediate IPCPs(routers) should use to update
the ECN field on passing packets.
The multi-bit ECN policy bases the value for the ECN field on the
depth of the rbuff queue packets will be sent on. I created another
call to grab the queue depth as fccntl is write-locking the
application. We can further optimize this to avoid most locking on the
rbuff.
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
The initial implementation for the ECDHE key exchange was doing the
key exchange after a flow was established. The public keys are now
sent allowg on the flow allocation messages, so that an encrypted
tunnel can be created within 1 RTT. The flow allocation steps had to
be extended to pass the opaque data ('piggybacking').
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
| |
The Packet Forwarding Function (PFF) was user-configurable using the
irm tool. However, this isn't really wanted since the PFF is dictated
by the routing algorithm. This moves the responsability for selecting
the correct PFF from the network admin to the unicast IPCP
implementation. Each routing policy now has to specify which PFF it
will use.
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
| |
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
|
| |
This completes the renaming of the normal IPCP to the unicast IPCP in
the sources, to get everything consistent with the documentation.
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
|
|
| |
The default case in the configuration parsing prints an erroneous
"Unknown IPCP type" message when a local IPCP is bootstrapped. The
IPCP_LOCAL type is now handled separately (no action needed).
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
The UDP IPCP now uses a fixed server UDP port (default 3435) for all
communications. This allows passing firewalls more easily since only a
single port needs to be opened. The client port can be fixed as well
if needed (default random). It uses an internal eid, so the MTU of the
UDP layer is reduced by 4 bytes, similar to the Ethernet IPCPs.
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
| |
Updates the copyright notice in all sources to 2019.
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
| |
This adds a new flow_join operaiton for broadcast, which is a much
safer solution than overloading destination name semantics. The
internal API now also has a different IPCP_FLOW_JOIN operation. The
IRMd doesn't need to query broadcasts IPCPs for the name, it can just
check if an IPCP with the layer name exists. The broadcast IPCP
doesn't need to implement the query proxy call anymore.
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
|
|
|
| |
This adds a broadcast IPCP that allows us to easily create multicast
applications. The broadcast IPCP accepts flows for "<layer_name>.mc".
A tool, obc (Ouroboros broadcast), is added that sends and reads a
message to a broadcast layer.
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
|
|
| |
The ipcp connect command can now set a specific qos cube for data
transfer flows. For management flows, the tool ignores this and
defaults to raw until data flows are stable enough.
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
|
|
| |
The signals are now handled in the main thread instead of an
asynchronous signal handler. The acceptloop is now correctly cancelled
and the associated timeouts are removed.
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
| |
Renames port_id to flow_id according to updated nomenclature.
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
| |
The flow allocator now passes the full qos specification to the
endpoint, instead of just a cube. This is a more flexible
architecture, as it makes QoS cubes internal to the layers.
Adds endianness transforms for the flow allocator protocol in the
normal IPCP.
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
|
| |
This removes the _DEFAULT_SOURCE definition in the endian header as it
should not be there. This avoids double and conflicting definitions.
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
|
| |
The cpu was a size_t but printed as an unsigned long, giving
compilation errors on 32 bit machines.
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
|
|
|
| |
This makes the buffer size used by the UNIX sockets configurable. In
case of a lot of IPCPs in the system it might become too small with
the default value, resulting in irm command failures. The user can now
easily configure it with an adequate value.
Signed-off-by: Sander Vrijders <[email protected]>
Signed-off-by: Dimitri Staessens <[email protected]>
|
|
|
|
|
|
|
|
| |
The signal handler is completely embedded in the source file. There
was no more need to call it from elsewhere.
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
|
|
| |
This adds a function that locks a thread to a random core. This
greatly improves performance on multi-cpu systems. There is no
portable way to do this, this only implements it for GNU/Linux.
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The reg/unreg API is simplified to registering and unregistering a
single name with a single IPCP. The functionality associated with
registering names was moved from the IRMd to the irm tool. The
function to list IPCPs was simplified to return all IPCPs in the
system with their basic properties needed for management.
The above changes led to some needed changes in the irm tool and the
management functions that were depending on the previous behaviour of
list_ipcps.
Command line functionality to list IPCPs in the system is also added
to the irm tool.
Some older code was refactored.
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This adds an IPC Process that uses DIX Ethernet with an Ethertype that
is configurable at bootstrap. This allows parallel DIX layers over the
same Ethernet network with different Ethertypes (and one LLC
layer). It allows jumbo frames in the future, and should avoid the
problems we have with some routers not handling LLC traffic very
well. The destination endpoint ID is sent as a 16 bit integer, so the
maximum payload is 1498 bytes in standard Ethernet, and 8998 bytes
when Jumbo frames are used.
The implementation is very similar to the Ethernet LLC IPCP, so it is
implemented using preprocessor macros in the single source instead of
duplicating code.
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
This makes the TTL non-optional and allows the maximum (initial) value
of the TTL to be specified at bootstrap (the default is set to
60). The fd in the DT PCI is now called EID (Endpoint ID). The names
"dif" and "ae" have been replaced by "layer" and "component"
respectively in all sources.
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
| |
Happy New Year, Ouroboros.
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This changes the terminology to use layer instead of DIF and deprecate
the word "shim" for the IPCPs that attach to Ethernet LLC and UDP .The
terminology has not yet been changed in the variable names etc.
This reflects the design choices in Ouroboros to make IPCPs pure
resource allocators instead of also providing an "IPC service". The
Ouroboros IPCPs that attach to Ethernet and UDP implement the
allocator and are thus not really shims.
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
|
|
|
| |
This makes the hashing algorithm configurable once more for the local
and the shim-eth-llc, since their scope is so small that it is up to
the network administrator to select a correct algorithm for the whole
network.
Signed-off-by: Sander Vrijders <[email protected]>
Signed-off-by: Dimitri Staessens <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
This refactors ouroboros to use "program" instead of "application
process" and "process" instead of "application process instance" to
align with current naming in current Operating Systems courses instead
of the ISO nomenclature adopted by RINA. This change permeates through
the entire implementation. Also contains some minor other refactors.
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
|
|
| |
The commands were read under a mutex. Since the cmd struct was just
malloc'd, this is not needed. Also fixes closing the socket if the
malloc fails.
Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This commit deprecates ouroboros_init and ouroboros_fini and adds them
as a constructor or destructor, causing these function to be run
automatically when a program that links to the library calls and exits
main(). For this to fully work, the library had to be split so that we
can avoid the irmd calling these functions (the IRMd has to create the
shm structures on which these calls depend).
The library is split in 3 parts: libouroboros-dev, libouroboros-irm
and libouroboros-common. The latter is linked to the other two so that
including libouroboros-dev or libouroboros-irm will also link
libouroboros-common.
|
|
|
|
|
|
|
|
|
| |
This makes the threadpool use pthread_cancel instead of setting an
exit flag that threadpool managed threads check periodically. This
drastically reduces CPU consumption in the irmd when running a lot of
applications. It requires cancellation handlers in the ipcp and irmd
to be implemented to ensure safe cancellation during operation and
shutdown.
|
|
|
|
|
|
|
|
|
|
|
| |
This adds a threadpool manager to the DHT. This was needed because the
detached thread could cause a data race on shutdown.
The threadpool manager is revised to allow multiple instances in a
single program.
The irmd and ipcp now store commands in a buffer (list) instead of a
single buffer before passing it to handler threads.
|
|
|
|
|
| |
This turns the PDU Forwarding Function of the IPCP into a policy. For
now only the simple PFF policy is available.
|
|
|
|
|
|
|
|
|
|
|
|
| |
This enables user-written tools to instruct IPCPs to establish and
tear down connections (a.k.a. adjacencies) between its internal
components (Management and Data Transfer).
For more info, do "irm ipcp connect" or "irm ipcp disconnect" on the
command line.
This commit exposes a deletion bug in the RIB where FSO's fail to
unpack/parse. This will be fixed when the RIB is deprecated.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The graph adjacency manager has been deprecated in favor of providing
an external interface into the connectivity manager so that
adjacencies can be controlled from the command line, user scripts or
user applications.
The gam and its associated policies were removed from the normal IPCP
and the IRM configuration tools. The "/members" part of the RIB was
deprecated. Removal of the gam means that initial connectivity based
on changes in the RIB can't be provided, so some changes were
required throughout the normal IPCP.
The enrollment procedure was revised to establish its own
connectivity. First, it gets boot information from a peer by
establishing a connection to the remote enrollment component and
downloading the IPCP configuratoin. This is now done using its own
protocol buffers message in anticipation of deprecation of the RIB and
CDAP for communication within a DIF.
After the boot information is downloaded, it establishes a data
transfer flow for enrolling the directory (DHT). After the DHT has
enrolled, it signals the peer to that enrollment is done, and the data
transfer connection is torn down.
Signaling connections is done via the nbs struct, which is now passed
to the connmgr, which enables control of the connectivity graph from
external sources.
|
|
|
|
|
| |
This fixes several memleaks as reported by valgrind. It also fixes
some calls to close() with -1.
|
|\
| |
| |
| | |
build: Revise the build system
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This revises the build system to have configuration per system
component. System settings can now be set using cmake.
The standard compliance defines were removed from configuration header
and are set in the sources where needed. Also some small code
refactors, such as moving the data for shims out of the ipcp structure
to the respective shims were performed.
|
|/
|
|
|
|
| |
Fixes a data race in the IRMd when IPCPs bail immediately after being
created (e.g. due to an error). Fixes a race between the acceptor
threads and the threads managed by threadpool manager upon shutdown.
|
|
|
|
|
| |
This adds an extra condvar so the mainloops can signal the acceptor
thread without waking up all mainloops.
|
|
|
|
|
|
|
|
|
|
|
| |
The IRMd and IPCPs will now call accept on their command sockets from
a single thread that will dispatch work to the other threads.
This solves a problem on OS X and FreeBSD where accept() doesn't time
out when setting SO_RCVTIMEO on the socket. Calling kqueue or select()
on that socket to wait for events before calling accept() didn't solve
it since select() or kqueue() might wake up multiple threads, with the
non-working threads again blocked on the accept() on shutdown.
|
| |
|
|
|
|
| |
This fixes several assignments to the wrong enum type.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This adds support for libgcrypt. If at least version 1.7.0 of
libgcrypt is present, it may be used for secure random number
generation and is used for hashing in the irmd/ipcp.
The hash definitions are moved to the internal hash.h header, and
defined independently of the hashes that are defined as part of the
directory policy for the normal IPCP. The translation is moved from
the IRMd to ipcpd/ipcp.h. The bootstrap call from the IRMd expects the
IPCP to return the correct hash algorithm with a dif_info struct,
which is in line with the behavior of the enroll call.
This also improves how some platform checks in the build system are
handled.
|
|
|
|
|
| |
This refactors dev.c to have a common function called init_flow that
is called whenever a flow is allocated in dev.c
|
|
|
|
|
|
|
|
| |
This adds a threadpool manager component in the library that is used
in the IRMd and IPCPs. The threadpool manager now doesn't detach
threads but does a join when they exit. This solves a data race in the
previous implementation where some threads were not completely finished
upon release of some resources.
|
|\
| |
| |
| | |
ipcpd: Add DHT as directory in normal IPCP
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This implements a Distributed Hash Table (DHT) based on the Kademlia
protocol, with default parameters set as used in the BitTorrent
Mainline DHT. This initial implementation is almost feature complete,
except for some things to be done after a testing period: caching and
stale peer bumping, and setting the expiration timeout via the IRM
tool.
|