| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
| |
This will make the register and query calls at the DHT wait for a JOIN
to complete. This avoids fails when calling register immediately after
creating a data transfer connection.
Signed-off-by: Dimitri Staessens <dimitri.staessens@ugent.be>
Signed-off-by: Sander Vrijders <sander.vrijders@ugent.be>
|
|
|
|
|
|
|
|
| |
Some responses for a query were missed because the lookup went into
LU_COMPLETE state while still having pending requests.
Signed-off-by: Dimitri Staessens <dimitri.staessens@ugent.be>
Signed-off-by: Sander Vrijders <sander.vrijders@ugent.be>
|
|
|
|
|
|
|
| |
Happy New Year, Ouroboros.
Signed-off-by: Dimitri Staessens <dimitri.staessens@ugent.be>
Signed-off-by: Sander Vrijders <sander.vrijders@ugent.be>
|
|
|
|
|
|
|
|
|
|
| |
The DHT will now wait for the DHT_RUNNING state before handling
request messages, ensuring that the parameters are set correctly
before they are read and sent at JOIN. Also fixes a re-introduced
assertion failure when a JOIN fails.
Signed-off-by: Dimitri Staessens <dimitri.staessens@ugent.be>
Signed-off-by: Sander Vrijders <sander.vrijders@ugent.be>
|
|
|
|
|
|
|
|
| |
ca2f47d had a missing bracket after a last-minute update to simplify
the logic.
Signed-off-by: Dimitri Staessens <dimitri.staessens@ugent.be>
Signed-off-by: Sander Vrijders <sander.vrijders@ugent.be>
|
|
|
|
|
|
|
|
|
|
| |
The DHT buckets were not updated in JOINING state, which sometimes
caused the DHT to miss its first peer update when joining and become
isolated. This also checks if a name is already registered to avoid
adding it multiple times.
Signed-off-by: Dimitri Staessens <dimitri.staessens@ugent.be>
Signed-off-by: Sander Vrijders <sander.vrijders@ugent.be>
|
|
|
|
|
|
|
| |
This fixes accessing buckets when a DHT join fails.
Signed-off-by: Dimitri Staessens <dimitri.staessens@ugent.be>
Signed-off-by: Sander Vrijders <sander.vrijders@ugent.be>
|
|
|
|
|
|
|
|
|
|
| |
The DHT join operation left an SDU in the rdrbuff for a couple of
seconds during retries, which may cause the system to block if there
is heavy traffic. This patch releases the sdb when the write fails and
creates a new packet for every retry.
Signed-off-by: Dimitri Staessens <dimitri.staessens@ugent.be>
Signed-off-by: Sander Vrijders <sander.vrijders@ugent.be>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
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.
|
|
|
|
|
|
| |
A pthread_cond_wait was using the pthread_cond_clock and thereby
timing out constantly, consuming 100% CPU. This changes it to use the
realtime clock.
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The lookup_update was waiting for the LU_INIT state to resolve under
dht->lock which prevented that update.
This PR fixes this bug, but leaves a (very rare) bug when the
lookup_destroy is called while the lookup_update is waiting for the
LU_INIT state to resolve. The solution also is a (harmless) lock
inversion, but this is also not the best.
Fixes #51
Fixes #52
|
|
|
|
|
|
|
|
|
|
|
| |
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.
|
|\
| |
| |
| | |
Some bugfixes
|
| | |
|
|\|
| |
| |
| | |
Some bugfixes
|
| | |
|
|/ |
|
|
|
|
| |
This avoids potential data races on rib.fuse.
|
|
|
|
|
|
|
| |
The DHT will now spawn a thread when receiving SDUs to avoid
starvation of sdu scheduler threads.
Also fixes some locking issues.
|
|\
| |
| |
| | |
ipcpd: Refuse query when DHT not in running state
|
| | |
|
|/ |
|
|
|
|
|
|
|
|
|
| |
The DHT will now enroll or sync when a data transfer connection is
added. This avoids the need to create a temporary data transfer
connection during enrollment (and speeds it up considerably).
The notifier system was modified to take an opaque pointer to the
object that registers as a parameter.
|
|
|
|
|
|
|
|
| |
This removes the RIB as a datastructure and CDAP as the protocol
between IPCPs. CDAP, the rib and related sources are deprecated. The
link-state protocol policy is udpated to use its own protocol based on
a simple broadcast strategy along a tree. The neighbors struct is
deprecated and moved to the library as a generic notifier component.
|
|
|
|
| |
This prevents a potential deadlock on the wrlock.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
The correct state was returned but not set internally in the
lookup_struct, causing lookup_wait to return a lookup in LU_PENDING if
the wait timedout.
This also unlocks the IRMd during register calls and fixes some
cleanup if lookup_wait gets interrupted.
Fixes #48
|
|
|
|
| |
This fixes several bugs as reported by the clang static analyzer.
|
|
|
|
|
| |
This fixes several memleaks as reported by valgrind. It also fixes
some calls to close() with -1.
|
|
|
|
|
|
|
|
|
|
| |
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.
|
| |
|
|\
| |
| |
| | |
lib: Fix instability in threadpool manager
|
| |
| |
| |
| |
| |
| | |
The DHT failed when receiving multiple responses in a short
timeframe. The lookup struct now keeps track of the number of
outstanding requests.
|
|/ |
|
|
|
|
| |
This fixes several assignments to the wrong enum type.
|
|
|
|
|
|
|
|
|
| |
The DHT lookup would stop if a node returned no new useful information
(i.e. the value for the key or unknown nodes closer to the key)
without waiting for other pending requests that could still return
useful information. Now it correctly tracks the number of outstanding
requests or returns if it doesn't get a response for KAD_T_RESP
seconds. This fixes multi-hop flow allocation over the normal.
|
| |
|
|
|
|
|
|
|
| |
Fixes a memory leak when the DHT is destroyed during a send_msg. Fixes
a deadlock when destroying the DHT with a pending lookup. Removes some
debug logs that clutter output due to expected redundant FIND_VALUE
responses.
|
|
|
|
|
|
|
| |
This will make the remote retry sending responses when the DT
component can't send the message. This is most useful at enrollment,
when the remote's routing table may need to wait a bit on the routing
component for a path to the new member,
|
|
|
|
|
| |
Not all returned contacts were processed when a FIND_NODE or
FIND_VALUE message was returned.
|
|
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.
|