On Mon, 16 Mar 2026 17:27:28 +0000 Anatoly Burakov <[email protected]> wrote:
> Most rte_flow parsers in DPDK suffer from huge implementation complexity > because > even though 99% of what people use rte_flow parsers for is parsing protocol > graphs, no parser is written explicitly as a graph. This patchset attempts to > suggest a viable model to build rte_flow parsers as graphs, by offering a > lightweight header only library to build rte_flow parsering graphs without too > much boilerplate and complexity. > > Most of the patchset is about Intel drivers, but they are meant as > reimplementations as well as examples for the rest of the community to assess > how to build parsers using this new infrastructure. I expect the first two > patches will be of most interest to non-Intel reviewers, as they deal with > building two reusable parser architecture pieces. > > The first piece is a new flow graph helper in ethdev. Its purpose is > deliberately narrow: it targets the protocol-graph part of rte_flow pattern > parsing, where drivers walk packet headers and validate legal item sequences > and > parameters. That does not cover all possible rte_flow features, especially > more > exotic flow items, but it does cover a large and widely shared part of what > existing drivers need to do. Or, to put it in other words, the only flow items > this infrastructure *doesn't* cover is things that do not lend themselves well > to be parsed as a graph of protocol headers (e.g. conntrack items). Everything > else should be covered or cover-able. > > The second piece is a reusable flow engine framework for Intel Ethernet > drivers. > This is kept Intel-local for now because I do not feel it is generic enough to > be presented as an ethdev-wide engine model. Even so, the intent is to > establish > a cleaner parser architecture with a defined interaction model, explicit > memory > ownership rules, and engine definitions that do not block > secondary-process-safe > usage. It is my hope that we could also promote something like this into > ethdev > proper and remove the necessity for drivers to build so much boilerplate > around > rte_flow parsing (and more often than not doing it in a way that is more > complex > than it needs to be). > > Most of the rest of the series is parser reimplementation, but that is mainly > the vehicle for demonstrating and validating those two pieces. ixgbe and i40e > are wired into the new common parsing path, and their existing parsers are > migrated incrementally to the graph-based model. Besides reducing ad hoc > parser > code, this also makes validation more explicit and more consistent with the > actual install path. In a few places that means invalid inputs that were > previously ignored, deferred, or interpreted loosely are now rejected earlier > and more strictly, without any increase in code complexity (in fact, with > marked > *decrease* of it!). > > Series depends on previously submitted patchsets: > > - IAVF global buffer fix [1] > - Common attr parsing stuff [2] > > [1] https://patches.dpdk.org/project/dpdk/list/?series=37585&state=* > [2] https://patches.dpdk.org/project/dpdk/list/?series=37663&state=* > > Anatoly Burakov (21): > ethdev: add flow graph API > net/intel/common: add flow engines infrastructure > net/intel/common: add utility functions > net/ixgbe: add support for common flow parsing > net/ixgbe: reimplement ethertype parser > net/ixgbe: reimplement syn parser > net/ixgbe: reimplement L2 tunnel parser > net/ixgbe: reimplement ntuple parser > net/ixgbe: reimplement security parser > net/ixgbe: reimplement FDIR parser > net/ixgbe: reimplement hash parser > net/i40e: add support for common flow parsing > net/i40e: reimplement ethertype parser > net/i40e: reimplement FDIR parser > net/i40e: reimplement tunnel QinQ parser > net/i40e: reimplement VXLAN parser > net/i40e: reimplement NVGRE parser > net/i40e: reimplement MPLS parser > net/i40e: reimplement gtp parser > net/i40e: reimplement L4 cloud parser > net/i40e: reimplement hash parser > > drivers/net/intel/common/flow_engine.h | 1003 ++++ > drivers/net/intel/common/flow_util.h | 165 + > drivers/net/intel/i40e/i40e_ethdev.c | 56 +- > drivers/net/intel/i40e/i40e_ethdev.h | 49 +- > drivers/net/intel/i40e/i40e_fdir.c | 47 - > drivers/net/intel/i40e/i40e_flow.c | 4092 +---------------- > drivers/net/intel/i40e/i40e_flow.h | 44 + > drivers/net/intel/i40e/i40e_flow_ethertype.c | 258 ++ > drivers/net/intel/i40e/i40e_flow_fdir.c | 1806 ++++++++ > drivers/net/intel/i40e/i40e_flow_hash.c | 1289 ++++++ > drivers/net/intel/i40e/i40e_flow_tunnel.c | 1510 ++++++ > drivers/net/intel/i40e/i40e_hash.c | 980 +--- > drivers/net/intel/i40e/i40e_hash.h | 8 +- > drivers/net/intel/i40e/meson.build | 4 + > drivers/net/intel/ixgbe/ixgbe_ethdev.c | 40 +- > drivers/net/intel/ixgbe/ixgbe_ethdev.h | 13 +- > drivers/net/intel/ixgbe/ixgbe_fdir.c | 13 +- > drivers/net/intel/ixgbe/ixgbe_flow.c | 3130 +------------ > drivers/net/intel/ixgbe/ixgbe_flow.h | 38 + > .../net/intel/ixgbe/ixgbe_flow_ethertype.c | 240 + > drivers/net/intel/ixgbe/ixgbe_flow_fdir.c | 1510 ++++++ > drivers/net/intel/ixgbe/ixgbe_flow_hash.c | 182 + > drivers/net/intel/ixgbe/ixgbe_flow_l2tun.c | 228 + > drivers/net/intel/ixgbe/ixgbe_flow_ntuple.c | 483 ++ > drivers/net/intel/ixgbe/ixgbe_flow_security.c | 297 ++ > drivers/net/intel/ixgbe/ixgbe_flow_syn.c | 280 ++ > drivers/net/intel/ixgbe/meson.build | 7 + > lib/ethdev/meson.build | 1 + > lib/ethdev/rte_flow_graph.h | 414 ++ > 29 files changed, 9867 insertions(+), 8320 deletions(-) > create mode 100644 drivers/net/intel/common/flow_engine.h > create mode 100644 drivers/net/intel/common/flow_util.h > create mode 100644 drivers/net/intel/i40e/i40e_flow.h > create mode 100644 drivers/net/intel/i40e/i40e_flow_ethertype.c > create mode 100644 drivers/net/intel/i40e/i40e_flow_fdir.c > create mode 100644 drivers/net/intel/i40e/i40e_flow_hash.c > create mode 100644 drivers/net/intel/i40e/i40e_flow_tunnel.c > create mode 100644 drivers/net/intel/ixgbe/ixgbe_flow.h > create mode 100644 drivers/net/intel/ixgbe/ixgbe_flow_ethertype.c > create mode 100644 drivers/net/intel/ixgbe/ixgbe_flow_fdir.c > create mode 100644 drivers/net/intel/ixgbe/ixgbe_flow_hash.c > create mode 100644 drivers/net/intel/ixgbe/ixgbe_flow_l2tun.c > create mode 100644 drivers/net/intel/ixgbe/ixgbe_flow_ntuple.c > create mode 100644 drivers/net/intel/ixgbe/ixgbe_flow_security.c > create mode 100644 drivers/net/intel/ixgbe/ixgbe_flow_syn.c > create mode 100644 lib/ethdev/rte_flow_graph.h > Detailed AI review found some issues here. Review: [RFC PATCH v1 00/21] Intel flow parser rework (flow graph + engine framework) No Reviewed-by: a functional bug remains (see patch 02). Patch 02/21 - net/intel/common: add flow engines (flow_engine.h) Error: created flows are never tracked; destroy corrupts memory, flush/reset leak. ci_flow_create() allocates, parses, installs, and returns a flow but never links it into engine_conf->flows. The series has TAILQ_INIT + 3x TAILQ_REMOVE + 2x FOREACH over that list, but no TAILQ_INSERT into it anywhere. Once an engine is registered: - ci_flow_destroy() runs TAILQ_REMOVE on a node never inserted. ci_flow_alloc() zeroes the ci_flow header, so node.tqe_prev == NULL, and TAILQ_REMOVE's "*(elm)->field.tqe_prev = ..." is a NULL-pointer write -> crash/corruption on first rule destroy. - ci_flow_flush() and ci_flow_engine_conf_reset() iterate an always-empty list, so created rules are never uninstalled from HW and are leaked (incl. at close). Fix: on the success path in ci_flow_create(), before goto unlock: TAILQ_INSERT_TAIL(&engine_conf->flows, flow, node); Info: ci_flow_engine_conf_reset() calls ci_flow_free(engine, flow) without the "if (engine == NULL) continue;" guard that ci_flow_flush() has; ci_flow_free() dereferences engine->ops. Unreachable today (list always empty) but inconsistent. Patch 01/21 - ethdev: add flow graph API (rte_flow_graph.h) Warning: __rte_internal is applied to six static inline functions. That macro is a symbol-visibility attribute for exported (non-inline) symbols; on static inline it is meaningless and, without ALLOW_INTERNAL_API, expands to __attribute__((error("Symbol is not public ABI"))). Verified against current main: a TU including the header and calling rte_flow_graph_parse() fails at -O0 without ALLOW_INTERNAL_API ("Symbol is not public ABI", x8); clean with it. In-tree drivers set the flag so CI passes, but the header ships in driver_sdk_headers and the section(".text.internal") placement is still wrong. Also applied inconsistently (the two constraint helpers omit it). Recommend dropping __rte_internal from the inlines. Warning: doc references nonexistent macro. struct rte_flow_graph_edge comment and the next field comment say "terminated by RTE_FLOW_EDGE_END"; the defined sentinel is RTE_FLOW_NODE_EDGE_END. Info: rte_flow_graph_parse() visits node[0] with item == NULL. If a driver gives node[0] constraints or a validate/process callback, __flow_graph_node_is_expected() dereferences item->spec (NULL) or passes NULL to the callback. Default empty graph is safe; the "node[0] is a bare start node" contract is implicit - document/guard. Design (RFC-level): both the graph (per-node validate/process fn-pointers) and ci_flow_engine_ops (13 callbacks) are callback-table frameworks. The header aims to avoid a "driver-within-a-driver framework" - worth confirming the callback surface is minimal (e.g. optional ctx_validate, separate flow_alloc/flow_free) before the non-RFC spin.
