summaryrefslogtreecommitdiff
path: root/src/ipcpd
Commit message (Collapse)AuthorAgeFilesLines
* ipcpd: Mitigate race in wait_req_arr and wait_respDimitri Staessens2024-02-261-6/+15
| | | | | | | | | | | | | | | | | | | | The wait_flow_req_arr and wait_flow_resp functions are there to ensure the responses of the IRMd to flow allocation requests arrive in the correct order. These functions use a mutex: alloc_lock. After these functions return, the IPCP will switch to it's own (usually read-write) lock. In the local IPCP, this leaves room for a race where the state of the flow is accessed in alloc_resp before it is updated in wait_flow_req_arr. This race is only visible in the local IPCP, as the other IPCP have to send information between these calls, but it is theoretically possible when using any IPCP for local IPC. In the ipcpd-local, it happens ~0.01% to ~0.03% of flow allocations. This mitigates the problem in the ipcpd-local by adding a 1ms wait to the flow allocation if this race is detected. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* irmd: Fix flow deallocation timeoutDimitri Staessens2024-02-261-3/+9
| | | | | | | | The timeout was set to a value calculated as abstime for a cond_wait instead of a timeout, causing flows to linger in the IPCP. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Add build option to use pid as addressDimitri Staessens2024-02-233-1/+8
| | | | | | | | | This adds a build option IPCP_DEBUG_LOCAL that will use the pid as the DT name (address) in the unicast IPCP, which is handy for integration testing and debugging. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Clean up connection when cancelling connectDimitri Staessens2024-02-231-1/+8
| | | | | | | | The connection element was not free'd on shutdown during a connect operation. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* lib: Revise app flow allocationDimitri Staessens2024-02-238-179/+169
| | | | | | | | | | | | | This revises the application flow allocator to use the flow_info struct/message between the components. Revises the messaging to move the use protocol buffers to its own source (serdes-irm). Adds a timeout to the IRMd flow allocator to make sure flow allocations don't hang forever (this was previously taken care of by the sanitize thread). Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* build: Fix missing newlines at end of fileDimitri Staessens2024-02-191-1/+1
| | | | | | | Some files had a newline at the end, others didn't. Now they all do. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* irmd: Revise IRMd internalsDimitri Staessens2024-02-1911-71/+68
| | | | | | | | | | | | | | | | | | | | | | | | | | | This is a full revision of the IRMd internal implementation. The registry is now a proper subcomponent managing its own internal lock (a single mutex). Some tests are added for the registry and its data structures. Some macros for tests are added in <ouroboros/test.h>. Flow allocation is now more symmetric between the client side (alloc) and server size (accept). Each will create a flow in pending state (ALLOC_PENDING/ACCEPT_PENDING) that is potentially fulfilled by an IPCP using respond_alloc and respond_accept primitives. Deallocation is split in flow_dealloc (application side) and ipcp_flow_dealloc (IPCP side) to get the flow in DEALLOC_PENDING and DEALLOCATED state. Cleanup of failed flow allocation is now properly handled instead of relying on the sanitizer thread. The new sanitizer only needs to monitor crashed processes. On shutdown, the IRMd will now detect hanging processes and SIGKILL them and clean up their fuse mountpoints if needed. A lot of other things have been cleaned up and shuffled around a bit. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Fix hash conversion in ipcpd-udpDimitri Staessens2024-01-311-1/+3
| | | | | | | | The bootstrap function was not returning the correct hash. Bug introduced in 99545fa2. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Fix memleak in rib_readdirDimitri Staessens2024-01-311-2/+2
| | | | | | | On fail_dup, the last element (0) was not free'd. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* include: Use common definition between lib and IRMdDimitri Staessens2024-01-316-17/+9
| | | | | | | | Some definitions/enums were different between the library and IRMd (flow_state, ipcp_state). This moves them to common ground. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* build: Update licenses to 2024Dimitri Staessens2024-01-1371-71/+71
| | | | | | | Slow but steady. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* include: Rename layer.layer_name to layer.nameDimitri Staessens2023-11-295-7/+7
| | | | | | | | The layer_info had a member layer_name which is a bit redundant. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Fix eth management packets blocking rdrbuff0.20.3Dimitri Staessens2023-11-262-20/+19
| | | | | | | | | | The ipcpd-eth-* reserve a packet buffer slot for the N+1 data packets whenever receiving a frame. For management frames, that slot is not needed and it was not released, thus blocking the rdrbuff. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* lib: Fix reading FUSE output on RaspbianDimitri Staessens2023-11-261-2/+6
| | | | | | | | | | | | | | | | | The previous patch only fixed listing the contents on the fuse filesystem. Apparently files with st_blocks = 0 as seen as empty on Raspbian, and the FUSE read() function isn't invoked for such files. Setting st_blocks to 1 fixes that, but st_blksize is ignored for fuse. So, on raspbian the filesize is now a huge number, but at least reading the fuse filesystem works. Corrected the filesystem attributes for the IPCP output for systems that don't rely on st_blocks to assess filesize. Also set the file mode to 0644 as these are not executables. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* lib: Fix FUSE output on RaspbianDimitri Staessens2023-11-201-1/+1
| | | | | | | | | | | | For some reason, 'ls' on raspbian invoked the fuse readdir() in a loop where the first call had fuse_file_info * info set to NULL and subsequent calls had info->nonseekable set to 1. Since we don't check the value the info struct, this caused an infinite loop when trying to list the contents of the fuse filesystem subdirectories of /tmp/ouroboros/. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Fix log line in ipcpd-udpDimitri Staessens2023-11-081-1/+2
| | | | | | | The dst was printed as a string instead of using hash-formatting. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Improve logging in ipcpd-udpDimitri Staessens2023-10-251-9/+14
| | | | | | | Printed some more info on a few errors using strerror. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Broadcast destruction of DHT requests0.20.0Dimitri Staessens2023-09-201-1/+1
| | | | | | | | When a request is cancelled/destroyed, all blocking threads should exit. Noticed some hangs on the DHT, this seems to fix/reduce it. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Fix dereference of unknown pointer typeDimitri Staessens2023-09-201-2/+2
| | | | | | | | The event handler was dereferencing the pointer to a connection, but the pointer type is not known yet. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Register event handler after scheduler startDimitri Staessens2023-09-201-10/+15
| | | | | | | | | | | The event handler was registered before the scheduler was started. Which could in theory cause addition of fds to an uninitialized scheduler. The event handler is now registered after the scheduler is created as part of dt_start. Likewise it now unregisters as part of dt_stop. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Revise loggingDimitri Staessens2023-08-3013-245/+344
| | | | | | | | This revises the logging in the IPCPs to be a more consistent and reduce duplicate messages in nested functions. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* lib: Move protobuf definitions to pb/ directoryDimitri Staessens2023-08-305-41/+39
| | | | | | | | | | This moves the protobuf definition in the library to a pb/ directory. Also renames the protobuf files and does a quick review of the #define guards in the include library to specify _LIB_ for internal/non-public library headers. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Move alloc race mitigation to common sourceDimitri Staessens2023-08-306-235/+100
| | | | | | | | | | All flow allocator code was duplicating the mitigation for a race where the IRMd response for the flow allocation with a new flow fd was arriving before the response to the flow_req_arr. This is now moved to the ipcp common source. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Don't release sdb before writing packetDimitri Staessens2023-08-301-20/+25
| | | | | | | | | The flow allocator fa_alloc_resp would release the packet buffer (sdb) before writing if the response was a failure. Also sets the IPCP allocation timeout in nanoseconds as per the comment. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Set IPCP states in common sourcesDimitri Staessens2023-08-306-25/+12
| | | | | | | | The state of the IPCP was set and checked in the main files, but it's more convenient to do it in the common source. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Remove unused rwlockDimitri Staessens2023-08-301-2/+1
| | | | | Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Simplify including common enrollment codeDimitri Staessens2023-08-305-31/+6
| | | | | | | | | | | | | | | | | | | | | | | | | | | The common enrollment code between the unicast and broadcast IPCP included the same C source from a source file in the unicast and broadcast directories. Now it's handled by defining COMMON_SOURCES for the unicast and broadcast IPCP. For now, only the enrollment component is common. Two things are needed for the connection manager: * Routing to be fixed to use a supporting broadcast layer. Then management flows can be removed from the unicast IPCP (as they are in fact data transfer flows of the broadcast layer that supports the routing dissemination traffic). * DHT to enroll as part of IPCP enrollment. Now it enrolls when a data transfer flow is established to be more akin to how Kademlia nodes join a Kademlia DHT, but this should be revised to adhere more to O7s enrollment concepts. After that, the connection manager code will also be completely shared between the unicast and broadcast layer and the connmgr will also be common code. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Add IDs to enrollmentDimitri Staessens2023-08-234-171/+200
| | | | | | | | The enrollment messages now have a 64-bit ID to easier track enrollments in the logs in larger scale tests. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* include: Revise printing hashesDimitri Staessens2023-08-236-17/+22
| | | | | | | | | | | The code was a bit convoluted to print hashes as hex strings. Renamed to HASH_FMT32 and HASH_VAL32 to make clear we are printing the first 32 bits only, and added options to print 64 up to 512 bits as well. This doesn't depend on endianness anymore. Adds a small test for the hash (printing) functions. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Fix leaked filedescriptors in ipcpd-udpDimitri Staessens2023-07-311-2/+10
| | | | | | | Found by GCC static analyzer. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Use consistent function namingDimitri Staessens2023-03-243-70/+70
| | | | | | | | | | For instance ipcp_udp_* vs eth_ipcp_*. Now all functions are <type>_ipcp_*. Als cleans up some minor things. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* icpdp: Fix layer information for ipcpd-udpDimitri Staessens2023-03-243-5/+5
| | | | | | | | | | | | | The refactors removed the need to set the hash algorithm for the ipcpd-udp and the ipcpd-broadcast. However, the algorithm was not set at bootstrap, so the ipcpd-udp was trying to use an SHA3-256 instead of an MD5, causing flow allocation over the UDP to fail. The ipcpd-broadcast used the default, so there was no problem. Fixed by setting the correct algorithm for these ipcpds at bootstrap. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Fix condvar initialization in ipcpd-udpDimitri Staessens2023-03-241-2/+13
| | | | | | | The clock was not explicitly initialized in the ipcpd-udp. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Fix long log line in enrollmentDimitri Staessens2023-03-181-1/+2
| | | | | Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* lib: Remove semicolon after protobuf messagesDimitri Staessens2023-03-181-2/+2
| | | | | | | | Doesn't seem to be needed, this makes it uniform in all protobuf files. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* irmd: Add configuration file supportDimitri Staessens2023-03-087-225/+109
| | | | | | | | | | | | | | | | | | | | | This adds initial support for configuration files using the C99 TOML parser (to be installed separately from https://github.com/cktan/tomlc99). The default location for the IRMd configuration file is /etc/ouroboros/irmd.conf. This is configurable at build time. An example file will be installed in the configuration directory with the name irmd.conf.example. Config file support can be disabled using the DISABLE_CONFIGFILE build option. There were some refactors and changes to the configuration messages and protobuf files. This works towards consolidation of protobuf C as an option for more generic handling of serialization/deserialization of various messages. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Fix address conversion in ipcpd-udpDimitri Staessens2023-03-011-2/+2
| | | | | | | Bug introduced in 269f25d3. The wrong pointer was passed to inet_ntop. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* lib: Refactor hash internalsDimitri Staessens2023-02-231-20/+2
| | | | | | | | The internal hash enum now matches the public one w.r.t. directory hash policies. This removes some unnecessary conversion. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: refactor IPCP configurationDimitri Staessens2023-02-1310-406/+423
| | | | | | | | | | The ipcp configuration struct now has internal structures for the different IPCPs and for IPCP components of the unicast IPCP. Split the very long IPCP main loop into individual handler functions. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* build: Update copyright to 2023Dimitri Staessens2023-02-1372-72/+72
| | | | | | | 2022 was a rather slow year... Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* lib: Fix prototypes missing 'void'Dimitri Staessens2023-02-132-6/+6
| | | | | | | Found by Clang version 15. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Refactor main functionsDimitri Staessens2022-07-137-92/+78
| | | | | | | | | | | | | | | | | | | The structure of main functions of the IPCPs was a bit strange with a ipcp_shutdown() call that combined waiting for a terminating signal with stopping the internal threads. This is now revised into a symmetrical design of ipcp_start(), which now includes the create response towards the IRMd. ipcp_sigwait(), which waits for a shutdown signal ipcp_stop() that then stops the internal threads. Now the main() functions of the IPCPs will make sense without checking what that ipcp_shutdown() functions actually does. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Fix deadlock on exit in udp, local, ethDimitri Staessens2022-05-223-6/+6
| | | | | | | | | | | The IPCP flow_set was getting destroyed with the IPCP main loop still running, causing potential deadlocks. Reported-by: Thijs Paelman Confirmed-by: Dimitri Staessens <[email protected]> Tested-by: Dimitri Staessens <[email protected]> Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* build: Update copyright to 2022Dimitri Staessens2022-04-0372-72/+72
| | | | | | | Growing pains. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* lib: Add np1_flow_read and np1_flow_write callsDimitri Staessens2022-03-307-19/+39
| | | | | | | | | | | | | | | | Reading/writing to (N + 1)-flows from the IPCP was using a raw QoS flow to bypass some functions in the ipcp_flow_read call. But this call was broken for keepalive packets. Fixing the ipcp_flow_read call for (N - 1) flows causes the IPCPs to drop 0-byte keepalive packets coming from (N + 1) client flows. >From now on, there is a dedicated call for (N + 1) reads/writes from the IPCPs that's more efficient and cleaner. The (N + 1) flow internal QoS is now also defaulted to a qos_np1 qosspec, instead of tampering with the qosspec requested by the (N + 1) client. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* lib: Refactor reading packet from rbuffDimitri Staessens2022-03-305-10/+10
| | | | | | | | | | | | Reading packets from the rbuff and checking their validity (non-zero size, pass crc check, pass decryption) is now extracted into a function. Also adds a function to get the length of an sdu_du_buff instead of subtracting the tail and head pointers. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Fix memcpy with NULL in piggyback APIDimitri Staessens2022-03-083-6/+12
| | | | | | | | If there is no piggyback data, memcpy was passed a NULL pointer in memcpy(buf, NULL, 0) calls, which is undefined behaviour. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipcpd: Refactor kad_req_createDimitri Staessens2022-03-081-20/+25
| | | | | | | A small refactor of the kad_req_create function's cleanup code. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* ipdpd: Pass MPL to application at flow_allocationDimitri Staessens2022-03-0811-12/+43
| | | | | | | | | | | | The maximum packet lifetime (MPL) is a property of the flow that needs to be passed to the reliable transmission protocol (FRCP) for its correct operation. Previously, the value of MPL was set fixed as one of the (fixed) Delta-t parameters. This patch makes the MPL a property of the layer, and it can now be set per layer-type at build time. This is a step towards a proper MPL estimator in the flow allocator. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
* irmd, ipcp: Remove socket option in acceptloopDimitri Staessens2022-03-031-6/+1
| | | | | | | | | | | | | We cancel the thread, so the SO_RCVTIMEO is not needed anymore (it dated from when we checked the state every so often. The address sanitizer is complaining about the the cleanup handlers in the acceptloops after the thread gets cancelled in the read(). I've tried to resolve it, but no avail. Pretty convinced it's a false-positive, so ASan will ignore these functions for now. Signed-off-by: Dimitri Staessens <[email protected]> Signed-off-by: Sander Vrijders <[email protected]>
generated by cgit v1.2.3 (git 2.39.1) at 2024-12-01 09:35:04 +0000