[v3] regexdev: introduce regexdev subsystem
diff mbox series

Message ID 1580202029-37096-1-git-send-email-orika@mellanox.com
State Deferred
Delegated to: Thomas Monjalon
Headers show
Series
  • [v3] regexdev: introduce regexdev subsystem
Related show

Checks

Context Check Description
ci/Intel-compilation fail Compilation issues
ci/travis-robot success Travis build: passed
ci/iol-nxp-Performance success Performance Testing PASS
ci/iol-mellanox-Performance fail Performance Testing issues
ci/iol-testing fail Testing issues
ci/checkpatch success coding style OK

Commit Message

Ori Kam Jan. 28, 2020, 9 a.m. UTC
Even though there are some vendors which offer Regex HW offload, due to
lack of standard API, It is diffcult for DPDK consumer to use them
in a portable way.

This _RFC_ attempts to standardize the RegEx/DPI offload APIs for DPDK.

This RFC crafted based on SW Regex API frameworks such as libpcre and
hyperscan and a few of the RegEx HW IPs which I am aware of.

RegEx pattern matching applications:
* Next Generation Firewalls (NGFW)
* Deep Packet and Flow Inspection (DPI)
* Intrusion Prevention Systems (IPS)
* DDoS Mitigation
* Network Monitoring
* Data Loss Prevention (DLP)
* Smart NICs
* Grammar based content processing
* URL, spam and adware filtering
* Advanced auditing and policing of user/application security policies
* Financial data mining - parsing of streamed financial feeds
* Application recognition.
* Dmemory introspection.
* Natural Language Processing (NLP)
* Sentiment Analysis.
* Big data databse acceleration.
* Computational storage.

Request to review from HW and SW RegEx vendors and RegEx application
users to have portable DPDK API for RegEx.

The API schematics are based cryptodev, eventdev and ethdev existing
device API.

Signed-off-by: Jerin Jacob <jerinj@marvell.com>
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Signed-off-by: Ori Kam <orika@mellanox.com>
---
V3:
 * Change subject title.
V2:
 * Address ML comments.

RTE RegEx Device API
--------------------

Defines RTE RegEx Device APIs for RegEx operations and its provisioning.

The RegEx Device API is composed of two parts:

- The application-oriented RegEx API that includes functions to setup
  a RegEx device (configure it, setup its queue pairs and start it),
  update the rule database and so on.

- The driver-oriented RegEx API that exports a function allowing
  a RegEx poll Mode Driver (PMD) to simultaneously register itself as
  a RegEx device driver.

RegEx device components and definitions:

    +-----------------+
    |                 |
    |                 o---------+    rte_regex_[en|de]queue_burst()
    |   PCRE based    o------+  |               |
    |  RegEx pattern  |      |  |  +--------+   |
    | matching engine o------+--+--o        |   |    +------+
    |                 |      |  |  | queue  |<==o===>|Core 0|
    |                 o----+ |  |  | pair 0 |        |      |
    |                 |    | |  |  +--------+        +------+
    +-----------------+    | |  |
           ^               | |  |  +--------+
           |               | |  |  |        |        +------+
           |               | +--+--o queue  |<======>|Core 1|
       Rule|Database       |    |  | pair 1 |        |      |
    +------+----------+    |    |  +--------+        +------+
    |     Group 0     |    |    |
    | +-------------+ |    |    |  +--------+        +------+
    | | Rules 0..n  | |    |    |  |        |        |Core 2|
    | +-------------+ |    |    +--o queue  |<======>|      |
    |     Group 1     |    |       | pair 2 |        +------+
    | +-------------+ |    |       +--------+
    | | Rules 0..n  | |    |
    | +-------------+ |    |       +--------+
    |     Group 2     |    |       |        |        +------+
    | +-------------+ |    |       | queue  |<======>|Core n|
    | | Rules 0..n  | |    +-------o pair n |        |      |
    | +-------------+ |            +--------+        +------+
    |     Group n     |
    | +-------------+ |<-------rte_regex_rule_db_update()
    | | Rules 0..n  | |<-------rte_regex_rule_db_import()
    | +-------------+ |------->rte_regex_rule_db_export()
    +-----------------+

RegEx: A regular expression is a concise and flexible means for matching
strings of text, such as particular characters, words, or patterns of
characters. A common abbreviation for this is ?RegEx?.

RegEx device: A hardware or software-based implementation of RegEx
device API for PCRE based pattern matching syntax and semantics.

PCRE RegEx syntax and semantics specification:
http://regexkit.sourceforge.net/Documentation/pcre/pcrepattern.html

RegEx queue pair: Each RegEx device should have one or more queue pair
to
transmit a burst of pattern matching request and receive a burst of
receive the pattern matching response. The pattern matching
request/response
embedded in *rte_regex_ops* structure.

Rule: A pattern matching rule expressed in PCRE RegEx syntax along with
Match ID and Group ID to identify the rule upon the match.

Rule database: The RegEx device accepts regular expressions and converts
them
into a compiled rule database that can then be used to scan data.
Compilation allows the device to analyze the given pattern(s) and
pre-determine how to scan for these patterns in an optimized fashion
that
would be far too expensive to compute at run-time. A rule database
contains
a set of rules that compiled in device specific binary form.

Match ID or Rule ID: A unique identifier provided at the time of rule
creation for the application to identify the rule upon match.

Group ID: Group of rules can be grouped under one group ID to enable
rule isolation and effective pattern matching. A unique group identifier
provided at the time of rule creation for the application to identify
the
rule upon match.

Scan: A pattern matching request through *enqueue* API.

It may possible that a given RegEx device may not support all the
features
of PCRE. The application may probe unsupported features through
struct rte_regex_dev_info::pcre_unsup_flags

By default, all the functions of the RegEx Device API exported by a PMD
are lock-free functions which assume to not be invoked in parallel on
different logical cores to work on the same target object. For instance,
the dequeue function of a PMD cannot be invoked in parallel on two
logical
cores to operates on same RegEx queue pair. Of course, this function
can be invoked in parallel by different logical core on different queue
pair.
It is the responsibility of the upper level application to enforce this
rule.

In all functions of the RegEx API, the RegEx device is
designated by an integer >= 0 named the device identifier *dev_id*

At the RegEx driver level, RegEx devices are represented by a generic
data structure of type *rte_regex_dev*.

RegEx devices are dynamically registered during the PCI/SoC device
probing
phase performed at EAL initialization time.
When a RegEx device is being probed, a *rte_regex_dev* structure and
a new device identifier are allocated for that device. Then, the
regex_dev_init() function supplied by the RegEx driver matching the
probed
device is invoked to properly initialize the device.

The role of the device init function consists of resetting the hardware
or
software RegEx driver implementations.

If the device init operation is successful, the correspondence between
the device identifier assigned to the new device and its associated
*rte_regex_dev* structure is effectively registered.
Otherwise, both the *rte_regex_dev* structure and the device identifier
are
freed.

The functions exported by the application RegEx API to setup a device
designated by its device identifier must be invoked in the following
order:
    - rte_regex_dev_configure()
    - rte_regex_queue_pair_setup()
    - rte_regex_dev_start()

Then, the application can invoke, in any order, the functions
exported by the RegEx API to enqueue pattern matching job, dequeue
pattern
matching response, get the stats, update the rule database,
get/set device attributes and so on

If the application wants to change the configuration (i.e. call
rte_regex_dev_configure() or rte_regex_queue_pair_setup()), it must call
rte_regex_dev_stop() first to stop the device and then do the
reconfiguration
before calling rte_regex_dev_start() again. The enqueue and dequeue
functions should not be invoked when the device is stopped.

Finally, an application can close a RegEx device by invoking the
rte_regex_dev_close() function.

Each function of the application RegEx API invokes a specific function
of the PMD that controls the target device designated by its device
identifier.

For this purpose, all device-specific functions of a RegEx driver are
supplied through a set of pointers contained in a generic structure of
type
*regex_dev_ops*.
The address of the *regex_dev_ops* structure is stored in the
*rte_regex_dev*
structure by the device init function of the RegEx driver, which is
invoked during the PCI/SoC device probing phase, as explained earlier.

In other words, each function of the RegEx API simply retrieves the
*rte_regex_dev* structure associated with the device identifier and
performs an indirect invocation of the corresponding driver function
supplied in the *regex_dev_ops* structure of the *rte_regex_dev*
structure.

For performance reasons, the address of the fast-path functions of the
RegEx driver is not contained in the *regex_dev_ops* structure.
Instead, they are directly stored at the beginning of the
*rte_regex_dev*
structure to avoid an extra indirect memory access during their
invocation.

RTE RegEx device drivers do not use interrupts for enqueue or dequeue
operation. Instead, RegEx drivers export Poll-Mode enqueue and dequeue
functions to applications.

The *enqueue* operation submits a burst of RegEx pattern matching
request
to the RegEx device and the *dequeue* operation gets a burst of pattern
matching response for the ones submitted through *enqueue* operation.

Typical application utilisation of the RegEx device API will follow the
following programming flow.

- rte_regex_dev_configure()
- rte_regex_queue_pair_setup()
- rte_regex_rule_db_update() Needs to invoke if precompiled rule
  database not
  provided in rte_regex_dev_config::rule_db for
rte_regex_dev_configure()
  and/or application needs to update rule database.
- Create or reuse exiting mempool for *rte_regex_ops* objects.
- rte_regex_dev_start()
- rte_regex_enqueue_burst()
- rte_regex_dequeue_burst()

---
---
 config/common_base                           |    7 +
 doc/api/doxy-api-index.md                    |    1 +
 doc/api/doxy-api.conf.in                     |    1 +
 lib/Makefile                                 |    2 +
 lib/librte_regexdev/Makefile                 |   27 +
 lib/librte_regexdev/rte_regexdev.c           |    6 +
 lib/librte_regexdev/rte_regexdev.h           | 1411 ++++++++++++++++++++++++++
 lib/librte_regexdev/rte_regexdev_version.map |    3 +
 8 files changed, 1458 insertions(+)
 create mode 100644 lib/librte_regexdev/Makefile
 create mode 100644 lib/librte_regexdev/rte_regexdev.c
 create mode 100644 lib/librte_regexdev/rte_regexdev.h
 create mode 100644 lib/librte_regexdev/rte_regexdev_version.map

Comments

Jerin Jacob Feb. 22, 2020, 4:52 p.m. UTC | #1
> diff --git a/lib/librte_regexdev/rte_regexdev.h b/lib/librte_regexdev/rte_regexdev.h
> new file mode 100644
> index 0000000..c42128b
> --- /dev/null
> +++ b/lib/librte_regexdev/rte_regexdev.h
> @@ -0,0 +1,1411 @@
> +/* SPDX-License-Identifier: BSD-3-Clause
> + * Copyright(C) 2019 Marvell International Ltd.
> + * Copyright(C) 2020 Mellanox International Ltd.

There are a few comments from Xiang as well. So let's add Intel also
to the list.

> + */
> +
> +#ifndef _RTE_REGEXDEV_H_
> +#define _RTE_REGEXDEV_H_

> +
> +/**
> + * RegEx device information
> + */
> +struct rte_regex_dev_info {
> +       const char *driver_name; /**< RegEx driver name. */
> +       struct rte_device *dev; /**< Device information. */
> +       uint16_t max_matches;
> +       /**< Maximum matches per scan supported by this device. */
> +       uint16_t max_queue_pairs;
> +       /**< Maximum queue pairs supported by this device. */
> +       uint16_t max_payload_size;
> +       /**< Maximum payload size for a pattern match request or scan.
> +        * @see RTE_REGEX_DEV_CFG_CROSS_BUFFER_SCAN_F
> +        */
> +       uint32_t max_rules_per_group;
> +       /**< Maximum rules supported per group by this device.
> +        * This number can't be larger then 20 bits.

s/then/than

I think, we don't need to say this " This number can't be larger than 20 bits."
It may help SW drivers.



> +        */
> +       uint16_t max_groups;
> +       /**< Maximum group supported by this device.
> +        * This number can't be larger then 12 bits.
s/then/than
I think, we don't need to say this " This number can't be larger than 12 bits."
It may help SW drivers.

> +        */
> +       uint32_t regex_dev_capa;
> +       /**< RegEx device capabilities. @see RTE_REGEX_DEV_CAPA_* */
> +       uint64_t rule_flags;
> +       /**< Supported compiler rule flags.
> +        * @see RTE_REGEX_PCRE_RULE_*, struct rte_regex_rule::rule_flags
> +        */
> +       uint8_t max_scatter_gather;
> +       /**< The max supported number of buffers that can
> +        * be used in a single ops. The total size of all elements
> +        * must be less then max_payload_size.
> +        */
> +};
<snip>

> +int
> +rte_regex_rule_db_compile(uint8_t dev_id);
> +

I think your "rte_regex_rule_db_compile_activate() - compile and
activate the new rule set"
API name looks good. I am for rte_regex_rule_db_compile_activate().


> +/* Fast path APIs */
> +
> +/**
> + * The generic *rte_regex_match* structure to hold the RegEx match attributes.
> + * @see struct rte_regex_ops::matches
> + */
> +struct rte_regex_match {
> +       RTE_STD_C11
> +       union {
> +               uint64_t u64;
> +               struct {
> +                       uint32_t rule_id:20;
> +                       /**< Rule identifier to which the pattern matched.
> +                        * @see struct rte_regex_rule::rule_id
> +                        */
> +                       uint32_t group_id:12;
> +                       /**< Group identifier of the rule which the pattern
> +                        * matched. @see struct rte_regex_rule::group_id
> +                        */
> +                       uint16_t offset;

Since we have end_offset now, IMO, it is better to change this offset
to "start_offset".


> +                       /**< Starting Byte Position for matched rule. */
> +                       RTE_STD_C11
> +                       union {
> +                               uint16_t len;
> +                               /**< Length of match in bytes */
> +                               uint16_t end_offset;
> +                               /**< The end offset of the match. In case
> +                                * MATCH_AS_START configuration is disabled.
> +                                * @see RTE_REGEX_DEV_CFG_MATCH_AS_START
> +                                */

We have not concluded on this scheme. Have one field which has
different meaning will be difficult
for application. i.e fast path we need to have a check for this.

I think, Based on the majority of HW/SW implementation, we need to
either go with len or
end_offset. What Mellanox HW returns? len or end_offset?

or We can keep it as len or end_offset based on which drivers upstream first,
other drivers when it comes, we can see how to abstract it?

> +                       };
> +               };
> +       };
> +};

> +/**
> + * The generic *rte_regex_ops* structure to hold the RegEx attributes
> + * for enqueue and dequeue operation.
> + */
> +struct rte_regex_ops {
> +       /* W0 */
> +       uint16_t req_flags;
> +       /**< Request flags for the RegEx ops.
> +        * @see RTE_REGEX_OPS_REQ_*
> +        */
> +       uint16_t rsp_flags;
> +       /**< Response flags for the RegEx ops.
> +        * @see RTE_REGEX_OPS_RSP_*
> +        */
> +       uint16_t nb_actual_matches;
> +       /**< The total number of actual matches detected by the Regex device.*/
> +       uint16_t nb_matches;
> +       /**< The total number of matches returned by the RegEx device for this
> +        * scan. The size of *rte_regex_ops::matches* zero length array will be
> +        * this value.
> +        *
> +        * @see struct rte_regex_ops::matches, struct rte_regex_match
> +        */
> +
> +       /* W1 */
> +       uint16_t num_of_bufs;
> +       /**< The number of bufs that are part of this ops. The total size of
> +        * the length of all the buffer must be smaller then the max buffer
> +        * len.
> +        */
> +       uint16_t resv1;
> +       uint32_t resv2;

One of the point came up in our implementation is that.
HW can return an error due to various reasons.

One option could be to make nb_matches as zero? and update some flag?

What are your thoughts? updating the flag may be overkill.


> +
> +       /* W2 */
> +       struct rte_regex_iov *(*bufs)[];
> +       /**< Holds a pointer to the buffers list.*/

This memory gets submitted to HW so it can not be from the heap.
Cryptodev had a similar dilemma to use the container format for
multi-segment case, Finally they choose to with mbuf.

The following elements are in mbuf. Considering to avoid duplication and
avoid overhead most common usecase DPI(Assume if it is rte_regex_iov,
one need to copy all the elements from mbuf on fastpath).
I propose to have mbuf here instead of rte_regex_iov.

struct rte_regex_iov {
RTE_STD_C11
union {
uint64_t u64;
/**<  Allow 8-byte reserved on 32-bit system */
void *buf_addr;
/**< Virtual address of the pattern to be matched. */
};
rte_iova_t buf_iova;
/**< IOVA address of the pattern to be matched. */
uint16_t buf_size; /**< The buf size. */
};


> +
> +       /* W5 */
> +       RTE_STD_C11
> +       union {
> +               uint64_t cross_buf_id;
> +               /**< ID used by the RegEx device in order to handle cross
> +                * buffer detection.
> +                * This ID is given by the RegEx device on dequeue, and
> +                * the application must send it on the following enque.
> +                */
> +               void *cross_buf_ptr;
> +               /**< Pointer representation of *cross_buf_id* */

Could you have some example of how to use cross_buf_id?
Marvell HW does not support cross_buf_id, so we need to add this
feature as capability.



Regarding the rule attributes, We think, following needs to be added to control
the rule compilation behavior. If it not converging in first look, I
think, we can make
below as separate patch once we have basic things.


diff --git a/lib/librte_regexdev/rte_regexdev.h
b/lib/librte_regexdev/rte_regexdev.h
index 765da4aaa..1fa6a7135 100644
--- a/lib/librte_regexdev/rte_regexdev.h
+++ b/lib/librte_regexdev/rte_regexdev.h
@@ -767,6 +767,10 @@ enum rte_regex_rule_op {
 struct rte_regex_rule {
  enum rte_regex_rule_op op;
  /**< OP type of the rule either a OP_ADD or OP_DELETE */
+ uint32_t nb_prefix;
+ /**< Number of prefix entries */
+ char **prefixes;
+ /**< Rule prefix list */
  uint16_t group_id;
  /**< Group identifier to which the rule belongs to. */
  uint32_t rule_id;
@@ -784,6 +788,169 @@ struct rte_regex_rule {
  */
 };

+/** Enumerates RegEx rule prefix control type */
+enum rte_regex_prefix_control_entry_type
+{
+ RTE_REGEX_PREFIX_CONTROL_ENTRY_BLACKLIST = 0,
+ /**< The RegEx rule compiler will not use any prefixes that are
+ * specified. If there are no options but to use a blacklisted prefix,
+ * the rule will not be compiled.
+ */
+ RTE_REGEX_PREFIX_CONTROL_ENTRY_GRAYLIST = 1,
+ /**< The RegEx rule compiler will try not to use any prefixes that
+ * are specified.
+ * It will settle for these if there are no other options.
+ */
+ RTE_REGEX_PREFIX_CONTROL_ENTRY_WHITELIST = 2
+ /**< The RegEx rule compiler will try to use any prefixes that are
+ * specified. If it can't it will use other prefixes.
+ */
+
+};
+
+/** Structure to hold a RegEx rule prefix control entry */
+struct rte_regex_prefix_control_entry
+{
+        enum rte_regex_prefix_control_entry_type type;
+        /**< Indicates the type of the prefix selection control list entry */
+        char *value;
+        /**< The value associated with the prefix selection control list entry.
+         * This could be a string such as ABCD.
+         * Characters can also be represented in hex format
+         * such as \x00\x01\x02\x03.
+         */
+};
+
+/** Structure to hold a RegEx rule rule prefix control entry list */
+struct rte_regex_prefix_control
+{
+        uint32_t num;
+        /**< The number of prefix selection control list entries */
+        struct rte_regex_prefix_control_entry *prefixes;
+        /**< The list of prefix selection control list entries */
+};
+
+/** Enumerates RegEx rule compilation capabilities */
+
+#define RTE_REGEX_COMPILE_CAP_AUTO_RULE_ID_F (1ULL << 0)
+/**< RegEx device compilation supports auto rule id
+ * Assign an automatic incrementing ID to each rule.
+ */
+
+#define RTE_REGEX_COMPILE_CAP_CHECKSUM_F (1ULL << 1)
+/**< RegEx device compilation supports checksum calculation.
+ * When performing an incremental compile use the end checksum of the base rule
+ * db as the start checksum for the new rule db.
+ * Must be used in conjunction with an incremental compile.
+ */
+
+/* @warning might not be needed */
+#define RTE_REGEX_COMPILE_CAP_EM_F (1ULL << 2)
+/**< RegEx device compilation supports external memory
+ * Store rule db in external memory only.
+ */
+
+/* @warning might not be needed */
+#define RTE_REGEX_COMPILE_CAP_DISABLE_DISABLE_AIC_F (1ULL << 3)
+/**< RegEx device compilation supports disable of atomic incremental compile
+ * Disable the atomic incremental compile system.
+ * The incremental compilation will be performed but the output
+ * will not be atomized.
+ */
+
+#define RTE_REGEX_COMPILE_CAP_FORCE_F (1ULL << 4)
+/**< RegEx device compilation supports force compiliation
+ * Force the compilation to complete even if errors are found.
+ */
+
+#define RTE_REGEX_COMPILE_CAP_INCREMENTAL_F (1ULL << 5)
+/**< RegEx device compilation supports incremental compilation
+ * Perform an incremental compile using pre build rule db as the base.
+ */
+
+#define RTE_REGEX_COMPILE_CAP_NO_INC_COMPILE_PADDING_F (1ULL << 6)
+/**< RegEx device compilation supports padding removal
+ * Remove the padding that is normally introduced to assist with
+ * incremental compile. This will allow for a more compact rule db footprint
+ * but is not recommended if you intend to perform an incremental
+ * compile with the resulting rule db.
+ */
+
+#define RTE_REGEX_COMPILE_CAP_PCRE_PRE_8_36_F (1ULL << 7)
+/**< RegEx device compilation supports change of space class
+ * Set the space class to not include vertical tab (VT) as was used in
+ * PCRE before v8.36.
+ */
+
+#define RTE_REGEX_COMPILE_CAP_STRICT_QUANTIFIERS_F (1ULL << 8)
+/**< RegEx device compilation supports strict quantifiers
+ * To help with performance, by default the REGEX Compiler treats non-fixed
+ * bounded quantifiers as unbounded e.g. .{0,2048} will be the same as .*. This
+ * has the caveat of false positives being possible.
+ * This capability will ensure that the original construct is used in all
+ * cases meaning performance will be worse but no false positives will occur.
+ */
+
+#define RTE_REGEX_COMPILE_CAP_QUICK_REMOVE_RULES_F (1ULL << 9)
+/**< RegEx device compilation supports quick removal of rules
+ * Quick incremental compile to remove rules.
+ * Specify rule_ids_to_remove from rule DB.
+ * Must be used in conjunction with the incremental option.
+ */
+
+#define RTE_REGEX_COMPILE_CAP_QUICK_ADD_RULES_F (1ULL << 10)
+/**< RegEx device compilation supports quick addition of rules
+ * Quick incremental compile to add rules.
+ * Specify rule_ids_to_add to rule DB.
+ * Must be used in conjunction with the incremental option.
+ */
+
+/* @warning might not be needed */
+#define RTE_REGEX_COMPILE_CAP_UTF_8_F (1ULL << 11)
+/**< RegEx device compilation supports UTF-8 mode
+ * Switch on UTF-8 mode.
+ */
+
+#define RTE_REGEX_COMPILE_CAP_DISABLE_BIDIRECTIONAL_F (1ULL << 12)
+/**< RegEx device compilation supports disabling of bidirectional
+ * This disables the bidirectional rule compilation.
+ */
+
+#define RTE_REGEX_COMPILE_CAP_UTF_16_STRING_F (1ULL << 13)
+/**< RegEx device compilation supports UTF-16 strings efficiency
+ * Deal with UTF-16 strings more efficiently.
+ * The Compiler will assume RegEx constructs like the following are intended
+ * to match UTF-16 strings:A\x00?B\x00?C\x00?
+ * with this switch enabled the Compiler will interpret the
+ * above RegEx as:ABCD|A\x00B\x00C\x00
+ */
+
+#define RTE_REGEX_COMPILE_CAP_SWITCH_OFF_LOCALE_SUPPORT_F (1ULL << 14)
+/**< RegEx device compilation supports disabling locale
+ * Disable locale support.
+ */
+
+#define RTE_REGEX_COMPILE_CAP_SPLIT_ALTERNATIONS_F (1ULL << 15)
+/**< RegEx device compilation supports automatic alternation splitting
+ * Enable the automatic alternation splitting.
+ */
+
+/** Structure to hold a RegEx control list data sample */
+struct rte_regex_data_sample
+{
+        float threshold;
+        /**< A % threshold used to blacklist strings in the sample data.
+         * This % value represents the % of bytes in the sample data that
+         * represent the start of a 1 - 4-byte string.
+         * If the string occurs at > the % threshold number of bytes it will
+         * then be blacklisted.
+         */
+        size_t length;
+        /**< The number of bytes in the data sample to be analyzed */
+        char *data;
+        /**< The data sample to be analyzed */
+};
+
 /**
  * Update the rule database of a RegEx device.
  *
@@ -793,6 +960,12 @@ struct rte_regex_rule {
  *   which contain the regex rules attributes to be updated in rule database.
  * @param nb_rules
  *   The number of PCRE rules to update the rule database.
+ * @param prefixes
+ *   The prefix selection control list.
+ * @param compiler_options
+ *   The compiler options flags @see RTE_REGEX_COMPILE_CAP*.
+ * @param data_sample
+ *   The data sample for auto control list generation.
  *
  * @return
  *   The number of regex rules actually updated on the regex device's rule
@@ -811,7 +984,10 @@ struct rte_regex_rule {
  */
 uint16_t
 rte_regex_rule_db_update(uint8_t dev_id, const struct rte_regex_rule *rules,
- uint16_t nb_rules);
+ uint16_t nb_rules,
+ struct rte_regex_prefix_control *prefix_control,
+ uint32_t compiler_options
+ struct rte_regex_data_sample *data_sample);

 /**
  * Import a prebuilt rule database from a buffer to a RegEx device.
Ori Kam Feb. 23, 2020, 8:41 a.m. UTC | #2
Hi Jerin,

Thanks, for the review.
PSB

Ori Kam

> -----Original Message-----
> From: Jerin Jacob <jerinjacobk@gmail.com>
> Sent: Saturday, February 22, 2020 6:52 PM
> To: Ori Kam <orika@mellanox.com>
> Cc: Jerin Jacob <jerinj@marvell.com>; xiang.w.wang@intel.com; dpdk-dev
> <dev@dpdk.org>; Pavan Nikhilesh <pbhagavatula@marvell.com>; Shahaf
> Shuler <shahafs@mellanox.com>; Hemant Agrawal
> <hemant.agrawal@nxp.com>; Opher Reviv <opher@mellanox.com>; Alex
> Rosenbaum <alexr@mellanox.com>; dovrat@marvell.com; Prasun Kapoor
> <pkapoor@marvell.com>; Nipun Gupta <nipun.gupta@nxp.com>; Richardson,
> Bruce <bruce.richardson@intel.com>; yang.a.hong@intel.com;
> harry.chang@intel.com; gu.jian1@zte.com.cn; shanjiangh@chinatelecom.cn;
> zhangy.yun@chinatelecom.cn; lixingfu@huachentel.com; wushuai@inspur.com;
> yuyingxia@yxlink.com; fanchenggang@sunyainfo.com;
> davidfgao@tencent.com; liuzhong1@chinaunicom.cn;
> zhaoyong11@huawei.com; oc@yunify.com; jim@netgate.com;
> hongjun.ni@intel.com; j.bromhead@titan-ic.com; deri@ntop.org;
> fc@napatech.com; arthur.su@lionic.com; Thomas Monjalon
> <thomas@monjalon.net>
> Subject: Re: [dpdk-dev] [PATCH v3] regexdev: introduce regexdev subsystem
> 
> > diff --git a/lib/librte_regexdev/rte_regexdev.h
> b/lib/librte_regexdev/rte_regexdev.h
> > new file mode 100644
> > index 0000000..c42128b
> > --- /dev/null
> > +++ b/lib/librte_regexdev/rte_regexdev.h
> > @@ -0,0 +1,1411 @@
> > +/* SPDX-License-Identifier: BSD-3-Clause
> > + * Copyright(C) 2019 Marvell International Ltd.
> > + * Copyright(C) 2020 Mellanox International Ltd.
> 
> There are a few comments from Xiang as well. So let's add Intel also
> to the list.
> 

Sure no problem.

> > + */
> > +
> > +#ifndef _RTE_REGEXDEV_H_
> > +#define _RTE_REGEXDEV_H_
> 
> > +
> > +/**
> > + * RegEx device information
> > + */
> > +struct rte_regex_dev_info {
> > +       const char *driver_name; /**< RegEx driver name. */
> > +       struct rte_device *dev; /**< Device information. */
> > +       uint16_t max_matches;
> > +       /**< Maximum matches per scan supported by this device. */
> > +       uint16_t max_queue_pairs;
> > +       /**< Maximum queue pairs supported by this device. */
> > +       uint16_t max_payload_size;
> > +       /**< Maximum payload size for a pattern match request or scan.
> > +        * @see RTE_REGEX_DEV_CFG_CROSS_BUFFER_SCAN_F
> > +        */
> > +       uint32_t max_rules_per_group;
> > +       /**< Maximum rules supported per group by this device.
> > +        * This number can't be larger then 20 bits.
> 
> s/then/than
> 
> I think, we don't need to say this " This number can't be larger than 20 bits."
> It may help SW drivers.
> 

Agree I will remove the 20 bits part.

> 
> 
> > +        */
> > +       uint16_t max_groups;
> > +       /**< Maximum group supported by this device.
> > +        * This number can't be larger then 12 bits.
> s/then/than
> I think, we don't need to say this " This number can't be larger than 12 bits."
> It may help SW drivers.
>

Agree will remove the 12 bits part.
 
> > +        */
> > +       uint32_t regex_dev_capa;
> > +       /**< RegEx device capabilities. @see RTE_REGEX_DEV_CAPA_* */
> > +       uint64_t rule_flags;
> > +       /**< Supported compiler rule flags.
> > +        * @see RTE_REGEX_PCRE_RULE_*, struct rte_regex_rule::rule_flags
> > +        */
> > +       uint8_t max_scatter_gather;
> > +       /**< The max supported number of buffers that can
> > +        * be used in a single ops. The total size of all elements
> > +        * must be less then max_payload_size.
> > +        */
> > +};
> <snip>
> 
> > +int
> > +rte_regex_rule_db_compile(uint8_t dev_id);
> > +
> 
> I think your "rte_regex_rule_db_compile_activate() - compile and
> activate the new rule set"
> API name looks good. I am for rte_regex_rule_db_compile_activate().
> 

I like your name, will change to compile_activate.

> 
> > +/* Fast path APIs */
> > +
> > +/**
> > + * The generic *rte_regex_match* structure to hold the RegEx match
> attributes.
> > + * @see struct rte_regex_ops::matches
> > + */
> > +struct rte_regex_match {
> > +       RTE_STD_C11
> > +       union {
> > +               uint64_t u64;
> > +               struct {
> > +                       uint32_t rule_id:20;
> > +                       /**< Rule identifier to which the pattern matched.
> > +                        * @see struct rte_regex_rule::rule_id
> > +                        */
> > +                       uint32_t group_id:12;
> > +                       /**< Group identifier of the rule which the pattern
> > +                        * matched. @see struct rte_regex_rule::group_id
> > +                        */
> > +                       uint16_t offset;
> 
> Since we have end_offset now, IMO, it is better to change this offset
> to "start_offset".
> 

Agree, will change.

> 
> > +                       /**< Starting Byte Position for matched rule. */
> > +                       RTE_STD_C11
> > +                       union {
> > +                               uint16_t len;
> > +                               /**< Length of match in bytes */
> > +                               uint16_t end_offset;
> > +                               /**< The end offset of the match. In case
> > +                                * MATCH_AS_START configuration is disabled.
> > +                                * @see RTE_REGEX_DEV_CFG_MATCH_AS_START
> > +                                */
> 
> We have not concluded on this scheme. Have one field which has
> different meaning will be difficult
> for application. i.e fast path we need to have a check for this.
> 

This is the time to conclude 😊 . at least for the first version.
Why do we have one field with different meaning? 
The result can be ether len or end_offset.

> I think, Based on the majority of HW/SW implementation, we need to
> either go with len or
> end_offset. What Mellanox HW returns? len or end_offset?
>

From Mellanox perspective we prefer the len approach. We also think 
it is much more user oriented.
 
> or We can keep it as len or end_offset based on which drivers upstream first,
> other drivers when it comes, we can see how to abstract it?
> 

I can except that assuming we choose the start and len approach 😊

> > +                       };
> > +               };
> > +       };
> > +};
> 
> > +/**
> > + * The generic *rte_regex_ops* structure to hold the RegEx attributes
> > + * for enqueue and dequeue operation.
> > + */
> > +struct rte_regex_ops {
> > +       /* W0 */
> > +       uint16_t req_flags;
> > +       /**< Request flags for the RegEx ops.
> > +        * @see RTE_REGEX_OPS_REQ_*
> > +        */
> > +       uint16_t rsp_flags;
> > +       /**< Response flags for the RegEx ops.
> > +        * @see RTE_REGEX_OPS_RSP_*
> > +        */
> > +       uint16_t nb_actual_matches;
> > +       /**< The total number of actual matches detected by the Regex
> device.*/
> > +       uint16_t nb_matches;
> > +       /**< The total number of matches returned by the RegEx device for this
> > +        * scan. The size of *rte_regex_ops::matches* zero length array will be
> > +        * this value.
> > +        *
> > +        * @see struct rte_regex_ops::matches, struct rte_regex_match
> > +        */
> > +
> > +       /* W1 */
> > +       uint16_t num_of_bufs;
> > +       /**< The number of bufs that are part of this ops. The total size of
> > +        * the length of all the buffer must be smaller then the max buffer
> > +        * len.
> > +        */
> > +       uint16_t resv1;
> > +       uint32_t resv2;
> 
> One of the point came up in our implementation is that.
> HW can return an error due to various reasons.
> 
> One option could be to make nb_matches as zero? and update some flag?
> 
> What are your thoughts? updating the flag may be overkill.
>

I think we can return just zero matches for now.
 
> 
> > +
> > +       /* W2 */
> > +       struct rte_regex_iov *(*bufs)[];
> > +       /**< Holds a pointer to the buffers list.*/
> 
> This memory gets submitted to HW so it can not be from the heap.
> Cryptodev had a similar dilemma to use the container format for
> multi-segment case, Finally they choose to with mbuf.
> 
> The following elements are in mbuf. Considering to avoid duplication and
> avoid overhead most common usecase DPI(Assume if it is rte_regex_iov,
> one need to copy all the elements from mbuf on fastpath).
> I propose to have mbuf here instead of rte_regex_iov.
> 

The application only needs to set the data pointers. (no copy is required. )
I agree that there are advantages to the mbuf approach.
The main limitation for the mbufs approach is that the user will need to play with the offset
pointers and pointers to the next mbuf, in order to support cross buffer.
For example we have a packet and we want to add to the scan also the last part of the previous packet,
this means that the application must modify the data offset in the previous packet mbuf including 
changing the next pointer to point to the head of the new packet, and then return the values to the original position.

What do you think?
We can start with mbufs and see how it works, or start with the buffer and see how it works.



 
> struct rte_regex_iov {
> RTE_STD_C11
> union {
> uint64_t u64;
> /**<  Allow 8-byte reserved on 32-bit system */
> void *buf_addr;
> /**< Virtual address of the pattern to be matched. */
> };
> rte_iova_t buf_iova;
> /**< IOVA address of the pattern to be matched. */
> uint16_t buf_size; /**< The buf size. */
> };
> 
> 
> > +
> > +       /* W5 */
> > +       RTE_STD_C11
> > +       union {
> > +               uint64_t cross_buf_id;
> > +               /**< ID used by the RegEx device in order to handle cross
> > +                * buffer detection.
> > +                * This ID is given by the RegEx device on dequeue, and
> > +                * the application must send it on the following enque.
> > +                */
> > +               void *cross_buf_ptr;
> > +               /**< Pointer representation of *cross_buf_id* */
> 
> Could you have some example of how to use cross_buf_id?
> Marvell HW does not support cross_buf_id, so we need to add this
> feature as capability.
>

The idea is that this buffer will be used to keep some internal data for the engine.
For example the current state and what was found until now, and then reuse this
for the next buffer.
We can remove it for now if we agree that we can add it later.
 


One more thing, regarding the ops structure, I think it is better to split it to 2 different 
structures one enque and one for dequeue, since there are no real shared data and we will
be able to save memory, what do you think?

> 
> 
> Regarding the rule attributes, We think, following needs to be added to control
> the rule compilation behavior. If it not converging in first look, I
> think, we can make
> below as separate patch once we have basic things.
> 

Regarding the new code, we need also to add a function to get the capabilities for the compiler or
add a new field in the dev_info which will report the complier supported features.
I prefer a dedicated function.

> 
> diff --git a/lib/librte_regexdev/rte_regexdev.h
> b/lib/librte_regexdev/rte_regexdev.h
> index 765da4aaa..1fa6a7135 100644
> --- a/lib/librte_regexdev/rte_regexdev.h
> +++ b/lib/librte_regexdev/rte_regexdev.h
> @@ -767,6 +767,10 @@ enum rte_regex_rule_op {
>  struct rte_regex_rule {
>   enum rte_regex_rule_op op;
>   /**< OP type of the rule either a OP_ADD or OP_DELETE */
> + uint32_t nb_prefix;
> + /**< Number of prefix entries */
> + char **prefixes;
> + /**< Rule prefix list */
>   uint16_t group_id;
>   /**< Group identifier to which the rule belongs to. */
>   uint32_t rule_id;
> @@ -784,6 +788,169 @@ struct rte_regex_rule {
>   */
>  };
> 
> +/** Enumerates RegEx rule prefix control type */
> +enum rte_regex_prefix_control_entry_type
> +{
> + RTE_REGEX_PREFIX_CONTROL_ENTRY_BLACKLIST = 0,
> + /**< The RegEx rule compiler will not use any prefixes that are
> + * specified. If there are no options but to use a blacklisted prefix,
> + * the rule will not be compiled.
> + */
> + RTE_REGEX_PREFIX_CONTROL_ENTRY_GRAYLIST = 1,
> + /**< The RegEx rule compiler will try not to use any prefixes that
> + * are specified.
> + * It will settle for these if there are no other options.
> + */
> + RTE_REGEX_PREFIX_CONTROL_ENTRY_WHITELIST = 2
> + /**< The RegEx rule compiler will try to use any prefixes that are
> + * specified. If it can't it will use other prefixes.
> + */
> +
> +};
> +
> +/** Structure to hold a RegEx rule prefix control entry */
> +struct rte_regex_prefix_control_entry
> +{
> +        enum rte_regex_prefix_control_entry_type type;
> +        /**< Indicates the type of the prefix selection control list entry */
> +        char *value;
> +        /**< The value associated with the prefix selection control list entry.
> +         * This could be a string such as ABCD.
> +         * Characters can also be represented in hex format
> +         * such as \x00\x01\x02\x03.
> +         */
> +};
> +
> +/** Structure to hold a RegEx rule rule prefix control entry list */
> +struct rte_regex_prefix_control
> +{
> +        uint32_t num;
> +        /**< The number of prefix selection control list entries */
> +        struct rte_regex_prefix_control_entry *prefixes;
> +        /**< The list of prefix selection control list entries */
> +};
> +
> +/** Enumerates RegEx rule compilation capabilities */
> +
> +#define RTE_REGEX_COMPILE_CAP_AUTO_RULE_ID_F (1ULL << 0)
> +/**< RegEx device compilation supports auto rule id
> + * Assign an automatic incrementing ID to each rule.
> + */
> +
> +#define RTE_REGEX_COMPILE_CAP_CHECKSUM_F (1ULL << 1)
> +/**< RegEx device compilation supports checksum calculation.
> + * When performing an incremental compile use the end checksum of the
> base rule
> + * db as the start checksum for the new rule db.
> + * Must be used in conjunction with an incremental compile.
> + */
> +
> +/* @warning might not be needed */
> +#define RTE_REGEX_COMPILE_CAP_EM_F (1ULL << 2)
> +/**< RegEx device compilation supports external memory
> + * Store rule db in external memory only.
> + */
> +
> +/* @warning might not be needed */
> +#define RTE_REGEX_COMPILE_CAP_DISABLE_DISABLE_AIC_F (1ULL << 3)
> +/**< RegEx device compilation supports disable of atomic incremental
> compile
> + * Disable the atomic incremental compile system.
> + * The incremental compilation will be performed but the output
> + * will not be atomized.
> + */
> +
> +#define RTE_REGEX_COMPILE_CAP_FORCE_F (1ULL << 4)
> +/**< RegEx device compilation supports force compiliation
> + * Force the compilation to complete even if errors are found.
> + */
> +
> +#define RTE_REGEX_COMPILE_CAP_INCREMENTAL_F (1ULL << 5)
> +/**< RegEx device compilation supports incremental compilation
> + * Perform an incremental compile using pre build rule db as the base.
> + */
> +
> +#define RTE_REGEX_COMPILE_CAP_NO_INC_COMPILE_PADDING_F (1ULL <<
> 6)
> +/**< RegEx device compilation supports padding removal
> + * Remove the padding that is normally introduced to assist with
> + * incremental compile. This will allow for a more compact rule db footprint
> + * but is not recommended if you intend to perform an incremental
> + * compile with the resulting rule db.
> + */
> +
> +#define RTE_REGEX_COMPILE_CAP_PCRE_PRE_8_36_F (1ULL << 7)
> +/**< RegEx device compilation supports change of space class
> + * Set the space class to not include vertical tab (VT) as was used in
> + * PCRE before v8.36.
> + */
> +
> +#define RTE_REGEX_COMPILE_CAP_STRICT_QUANTIFIERS_F (1ULL << 8)
> +/**< RegEx device compilation supports strict quantifiers
> + * To help with performance, by default the REGEX Compiler treats non-fixed
> + * bounded quantifiers as unbounded e.g. .{0,2048} will be the same as .*. This
> + * has the caveat of false positives being possible.
> + * This capability will ensure that the original construct is used in all
> + * cases meaning performance will be worse but no false positives will occur.
> + */
> +
> +#define RTE_REGEX_COMPILE_CAP_QUICK_REMOVE_RULES_F (1ULL << 9)
> +/**< RegEx device compilation supports quick removal of rules
> + * Quick incremental compile to remove rules.
> + * Specify rule_ids_to_remove from rule DB.
> + * Must be used in conjunction with the incremental option.
> + */
> +
> +#define RTE_REGEX_COMPILE_CAP_QUICK_ADD_RULES_F (1ULL << 10)
> +/**< RegEx device compilation supports quick addition of rules
> + * Quick incremental compile to add rules.
> + * Specify rule_ids_to_add to rule DB.
> + * Must be used in conjunction with the incremental option.
> + */
> +
> +/* @warning might not be needed */
> +#define RTE_REGEX_COMPILE_CAP_UTF_8_F (1ULL << 11)
> +/**< RegEx device compilation supports UTF-8 mode
> + * Switch on UTF-8 mode.
> + */
> +
> +#define RTE_REGEX_COMPILE_CAP_DISABLE_BIDIRECTIONAL_F (1ULL << 12)
> +/**< RegEx device compilation supports disabling of bidirectional
> + * This disables the bidirectional rule compilation.
> + */
> +
> +#define RTE_REGEX_COMPILE_CAP_UTF_16_STRING_F (1ULL << 13)
> +/**< RegEx device compilation supports UTF-16 strings efficiency
> + * Deal with UTF-16 strings more efficiently.
> + * The Compiler will assume RegEx constructs like the following are intended
> + * to match UTF-16 strings:A\x00?B\x00?C\x00?
> + * with this switch enabled the Compiler will interpret the
> + * above RegEx as:ABCD|A\x00B\x00C\x00
> + */
> +
> +#define RTE_REGEX_COMPILE_CAP_SWITCH_OFF_LOCALE_SUPPORT_F (1ULL
> << 14)
> +/**< RegEx device compilation supports disabling locale
> + * Disable locale support.
> + */
> +
> +#define RTE_REGEX_COMPILE_CAP_SPLIT_ALTERNATIONS_F (1ULL << 15)
> +/**< RegEx device compilation supports automatic alternation splitting
> + * Enable the automatic alternation splitting.
> + */
> +
> +/** Structure to hold a RegEx control list data sample */
> +struct rte_regex_data_sample
> +{
> +        float threshold;
> +        /**< A % threshold used to blacklist strings in the sample data.
> +         * This % value represents the % of bytes in the sample data that
> +         * represent the start of a 1 - 4-byte string.
> +         * If the string occurs at > the % threshold number of bytes it will
> +         * then be blacklisted.
> +         */
> +        size_t length;
> +        /**< The number of bytes in the data sample to be analyzed */
> +        char *data;
> +        /**< The data sample to be analyzed */
> +};
> +
>  /**
>   * Update the rule database of a RegEx device.
>   *
> @@ -793,6 +960,12 @@ struct rte_regex_rule {
>   *   which contain the regex rules attributes to be updated in rule database.
>   * @param nb_rules
>   *   The number of PCRE rules to update the rule database.
> + * @param prefixes
> + *   The prefix selection control list.
> + * @param compiler_options
> + *   The compiler options flags @see RTE_REGEX_COMPILE_CAP*.
> + * @param data_sample
> + *   The data sample for auto control list generation.
>   *
>   * @return
>   *   The number of regex rules actually updated on the regex device's rule
> @@ -811,7 +984,10 @@ struct rte_regex_rule {
>   */
>  uint16_t
>  rte_regex_rule_db_update(uint8_t dev_id, const struct rte_regex_rule *rules,
> - uint16_t nb_rules);
> + uint16_t nb_rules,
> + struct rte_regex_prefix_control *prefix_control,
> + uint32_t compiler_options
> + struct rte_regex_data_sample *data_sample);
> 
>  /**
>   * Import a prebuilt rule database from a buffer to a RegEx device.
Jerin Jacob Feb. 23, 2020, 9:53 a.m. UTC | #3
On Sun, Feb 23, 2020 at 2:12 PM Ori Kam <orika@mellanox.com> wrote:
>
> Hi Jerin,
>
> Thanks, for the review.
> PSB


Hi Ori

Since we are finalizing the specification part, I thought of
enumerating the list of work needs to be
completed for a new subsystem in DPDK.

0) Finalize the first version of the spec. Hope v4 will do that.
1) Introduce common library code for based on the  specification
2) One HW based driver implementation
3) One SW reference driver: libpcre library provides complete PCRE
functionality.
4) app/test/test_regexdev.c like app/test/test_eventdev.c
5) Need a maintainer for maintaining the regex subsystem
6) The first version programming guide documentation
7) Add app/test-regexdev like app/test-eventdev
8) Add an examples/xxxxxx program

IMO The following items need to be completed to accept a subsystem in
dpdk(Need at least on HW and SW driver).

0) Finalize the first version of the spec. Hope v4 will do that.
1) Introduce common library code for based on the  specification
2) One HW based driver implementation
3) One SW reference driver: libpcre library provides complete PCRE
functionality.
4) app/test/test_regexdev.c like app/test/test_eventdev.c
5) Need a maintainer for maintaining the regex subsystem

We have item (3) so Marvell would like to work on item (3). Our HW
driver may ready by v20.05 or the worst case by 20.08.
Let us know what other items Mellanox or community would like to work
on. This is to avoid duplication of work
to get clarity on the next steps.

PSB


>
>
> Ori Kam
>
> > -----Original Message-----
> > From: Jerin Jacob <jerinjacobk@gmail.com>
> > Sent: Saturday, February 22, 2020 6:52 PM
> > To: Ori Kam <orika@mellanox.com>
> > Cc: Jerin Jacob <jerinj@marvell.com>; xiang.w.wang@intel.com; dpdk-dev
> > <dev@dpdk.org>; Pavan Nikhilesh <pbhagavatula@marvell.com>; Shahaf
> > Shuler <shahafs@mellanox.com>; Hemant Agrawal
> > <hemant.agrawal@nxp.com>; Opher Reviv <opher@mellanox.com>; Alex
> > Rosenbaum <alexr@mellanox.com>; dovrat@marvell.com; Prasun Kapoor
> > <pkapoor@marvell.com>; Nipun Gupta <nipun.gupta@nxp.com>; Richardson,
> > Bruce <bruce.richardson@intel.com>; yang.a.hong@intel.com;
> > harry.chang@intel.com; gu.jian1@zte.com.cn; shanjiangh@chinatelecom.cn;
> > zhangy.yun@chinatelecom.cn; lixingfu@huachentel.com; wushuai@inspur.com;
> > yuyingxia@yxlink.com; fanchenggang@sunyainfo.com;
> > davidfgao@tencent.com; liuzhong1@chinaunicom.cn;
> > zhaoyong11@huawei.com; oc@yunify.com; jim@netgate.com;
> > hongjun.ni@intel.com; j.bromhead@titan-ic.com; deri@ntop.org;
> > fc@napatech.com; arthur.su@lionic.com; Thomas Monjalon
> > <thomas@monjalon.net>
> > Subject: Re: [dpdk-dev] [PATCH v3] regexdev: introduce regexdev subsystem
> >
> > > diff --git a/lib/librte_regexdev/rte_regexdev.h
> > b/lib/librte_regexdev/rte_regexdev.h
> > > new file mode 100644
> > > index 0000000..c42128b
> > > --- /dev/null
> > > +++ b/lib/librte_regexdev/rte_regexdev.h
> > > @@ -0,0 +1,1411 @@
> > > +/* SPDX-License-Identifier: BSD-3-Clause
> > > + * Copyright(C) 2019 Marvell International Ltd.
> > > + * Copyright(C) 2020 Mellanox International Ltd.
> >
> > There are a few comments from Xiang as well. So let's add Intel also
> > to the list.
> >
>
> Sure no problem.

Thanks

>
> > > + */
> > > +
> > > +#ifndef _RTE_REGEXDEV_H_
> > > +#define _RTE_REGEXDEV_H_
> >
> > > +
> > > +/**
> > > + * RegEx device information
> > > + */
> > > +struct rte_regex_dev_info {
> > > +       const char *driver_name; /**< RegEx driver name. */
> > > +       struct rte_device *dev; /**< Device information. */
> > > +       uint16_t max_matches;
> > > +       /**< Maximum matches per scan supported by this device. */
> > > +       uint16_t max_queue_pairs;
> > > +       /**< Maximum queue pairs supported by this device. */
> > > +       uint16_t max_payload_size;
> > > +       /**< Maximum payload size for a pattern match request or scan.
> > > +        * @see RTE_REGEX_DEV_CFG_CROSS_BUFFER_SCAN_F
> > > +        */
> > > +       uint32_t max_rules_per_group;
> > > +       /**< Maximum rules supported per group by this device.
> > > +        * This number can't be larger then 20 bits.
> >
> > s/then/than
> >
> > I think, we don't need to say this " This number can't be larger than 20 bits."
> > It may help SW drivers.
> >
>
> Agree I will remove the 20 bits part.
>
> >
> >
> > > +        */
> > > +       uint16_t max_groups;
> > > +       /**< Maximum group supported by this device.
> > > +        * This number can't be larger then 12 bits.
> > s/then/than
> > I think, we don't need to say this " This number can't be larger than 12 bits."
> > It may help SW drivers.
> >
>
> Agree will remove the 12 bits part.
>
> > > +        */
> > > +       uint32_t regex_dev_capa;
> > > +       /**< RegEx device capabilities. @see RTE_REGEX_DEV_CAPA_* */
> > > +       uint64_t rule_flags;
> > > +       /**< Supported compiler rule flags.
> > > +        * @see RTE_REGEX_PCRE_RULE_*, struct rte_regex_rule::rule_flags
> > > +        */
> > > +       uint8_t max_scatter_gather;
> > > +       /**< The max supported number of buffers that can
> > > +        * be used in a single ops. The total size of all elements
> > > +        * must be less then max_payload_size.
> > > +        */
> > > +};
> > <snip>
> >
> > > +int
> > > +rte_regex_rule_db_compile(uint8_t dev_id);
> > > +
> >
> > I think your "rte_regex_rule_db_compile_activate() - compile and
> > activate the new rule set"
> > API name looks good. I am for rte_regex_rule_db_compile_activate().
> >
>
> I like your name, will change to compile_activate.

Ack.

>
> >
> > > +/* Fast path APIs */
> > > +
> > > +/**
> > > + * The generic *rte_regex_match* structure to hold the RegEx match
> > attributes.
> > > + * @see struct rte_regex_ops::matches
> > > + */
> > > +struct rte_regex_match {
> > > +       RTE_STD_C11
> > > +       union {
> > > +               uint64_t u64;
> > > +               struct {
> > > +                       uint32_t rule_id:20;
> > > +                       /**< Rule identifier to which the pattern matched.
> > > +                        * @see struct rte_regex_rule::rule_id
> > > +                        */
> > > +                       uint32_t group_id:12;
> > > +                       /**< Group identifier of the rule which the pattern
> > > +                        * matched. @see struct rte_regex_rule::group_id
> > > +                        */
> > > +                       uint16_t offset;
> >
> > Since we have end_offset now, IMO, it is better to change this offset
> > to "start_offset".
> >
>
> Agree, will change.
>
> >
> > > +                       /**< Starting Byte Position for matched rule. */
> > > +                       RTE_STD_C11
> > > +                       union {
> > > +                               uint16_t len;
> > > +                               /**< Length of match in bytes */
> > > +                               uint16_t end_offset;
> > > +                               /**< The end offset of the match. In case
> > > +                                * MATCH_AS_START configuration is disabled.
> > > +                                * @see RTE_REGEX_DEV_CFG_MATCH_AS_START
> > > +                                */
> >
> > We have not concluded on this scheme. Have one field which has
> > different meaning will be difficult
> > for application. i.e fast path we need to have a check for this.
> >
>
> This is the time to conclude . at least for the first version.
> Why do we have one field with different meaning?
> The result can be ether len or end_offset.
>
> > I think, Based on the majority of HW/SW implementation, we need to
> > either go with len or
> > end_offset. What Mellanox HW returns? len or end_offset?
> >
>
> From Mellanox perspective we prefer the len approach. We also think
> it is much more user oriented.
>
> > or We can keep it as len or end_offset based on which drivers upstream first,
> > other drivers when it comes, we can see how to abstract it?
> >
>
> I can except that assuming we choose the start and len approach

I think, we can have first version with "start and len" by removing
RTE_REGEX_DEV_CFG_MATCH_AS_START.
When can think, how to abstract new drivers when it upstream based on
the overhead.


>
> > > +                       };
> > > +               };
> > > +       };
> > > +};
> >
> > > +/**
> > > + * The generic *rte_regex_ops* structure to hold the RegEx attributes
> > > + * for enqueue and dequeue operation.
> > > + */
> > > +struct rte_regex_ops {
> > > +       /* W0 */
> > > +       uint16_t req_flags;
> > > +       /**< Request flags for the RegEx ops.
> > > +        * @see RTE_REGEX_OPS_REQ_*
> > > +        */
> > > +       uint16_t rsp_flags;
> > > +       /**< Response flags for the RegEx ops.
> > > +        * @see RTE_REGEX_OPS_RSP_*
> > > +        */
> > > +       uint16_t nb_actual_matches;
> > > +       /**< The total number of actual matches detected by the Regex
> > device.*/
> > > +       uint16_t nb_matches;
> > > +       /**< The total number of matches returned by the RegEx device for this
> > > +        * scan. The size of *rte_regex_ops::matches* zero length array will be
> > > +        * this value.
> > > +        *
> > > +        * @see struct rte_regex_ops::matches, struct rte_regex_match
> > > +        */
> > > +
> > > +       /* W1 */
> > > +       uint16_t num_of_bufs;
> > > +       /**< The number of bufs that are part of this ops. The total size of
> > > +        * the length of all the buffer must be smaller then the max buffer
> > > +        * len.
> > > +        */
> > > +       uint16_t resv1;
> > > +       uint32_t resv2;
> >
> > One of the point came up in our implementation is that.
> > HW can return an error due to various reasons.
> >
> > One option could be to make nb_matches as zero? and update some flag?
> >
> > What are your thoughts? updating the flag may be overkill.
> >
>
> I think we can return just zero matches for now.

Ack.

>
> >
> > > +
> > > +       /* W2 */
> > > +       struct rte_regex_iov *(*bufs)[];
> > > +       /**< Holds a pointer to the buffers list.*/
> >
> > This memory gets submitted to HW so it can not be from the heap.
> > Cryptodev had a similar dilemma to use the container format for
> > multi-segment case, Finally they choose to with mbuf.
> >
> > The following elements are in mbuf. Considering to avoid duplication and
> > avoid overhead most common usecase DPI(Assume if it is rte_regex_iov,
> > one need to copy all the elements from mbuf on fastpath).
> > I propose to have mbuf here instead of rte_regex_iov.
> >
>
> The application only needs to set the data pointers. (no copy is required. )
> I agree that there are advantages to the mbuf approach.
> The main limitation for the mbufs approach is that the user will need to play with the offset
> pointers and pointers to the next mbuf, in order to support cross buffer.
> For example we have a packet and we want to add to the scan also the last part of the previous packet,
> this means that the application must modify the data offset in the previous packet mbuf including
> changing the next pointer to point to the head of the new packet, and then return the values to the original position.
>
> What do you think?
> We can start with mbufs and see how it works, or start with the buffer and see how it works.

I think, we can start with mbuf to align with other subsystems. We
will see later the use case for struct rte_regex_iov.


>
>
>
>
> > struct rte_regex_iov {
> > RTE_STD_C11
> > union {
> > uint64_t u64;
> > /**<  Allow 8-byte reserved on 32-bit system */
> > void *buf_addr;
> > /**< Virtual address of the pattern to be matched. */
> > };
> > rte_iova_t buf_iova;
> > /**< IOVA address of the pattern to be matched. */
> > uint16_t buf_size; /**< The buf size. */
> > };
> >
> >
> > > +
> > > +       /* W5 */
> > > +       RTE_STD_C11
> > > +       union {
> > > +               uint64_t cross_buf_id;
> > > +               /**< ID used by the RegEx device in order to handle cross
> > > +                * buffer detection.
> > > +                * This ID is given by the RegEx device on dequeue, and
> > > +                * the application must send it on the following enque.
> > > +                */
> > > +               void *cross_buf_ptr;
> > > +               /**< Pointer representation of *cross_buf_id* */
> >
> > Could you have some example of how to use cross_buf_id?
> > Marvell HW does not support cross_buf_id, so we need to add this
> > feature as capability.
> >
>
> The idea is that this buffer will be used to keep some internal data for the engine.
> For example the current state and what was found until now, and then reuse this
> for the next buffer.
> We can remove it for now if we agree that we can add it later.

I think, adding it later would be better so that we can see how to
abstract it well.

>
>
>
> One more thing, regarding the ops structure, I think it is better to split it to 2 different
> structures one enque and one for dequeue, since there are no real shared data and we will
> be able to save memory, what do you think?

Ops are allocated from mempool so it will be overhead to manage both.
moreover, some
of the fields added in req can be used for resp as info. cryptodev
follows the similar concept,
I think, we can have symmetry with cryptodev wherever is possible to avoid
end-user to learn new API models.



>
> >
> >
> > Regarding the rule attributes, We think, following needs to be added to control
> > the rule compilation behavior. If it not converging in first look, I
> > think, we can make
> > below as separate patch once we have basic things.
> >
>
> Regarding the new code, we need also to add a function to get the capabilities for the compiler or
> add a new field in the dev_info which will report the complier supported features.

I agree. Lets remove this new code from the first version. We can add
it later with capability as
a new patch.

I assume you will send the v4 with these comments. I think, with v4 we
can start implementing common library code.
Ori Kam Feb. 23, 2020, 12:33 p.m. UTC | #4
Hi Jerin,

Best,
Ori
> -----Original Message-----
> From: Jerin Jacob <jerinjacobk@gmail.com>
> Sent: Sunday, February 23, 2020 11:54 AM
> To: Ori Kam <orika@mellanox.com>
> Cc: Jerin Jacob <jerinj@marvell.com>; xiang.w.wang@intel.com; dpdk-dev
> <dev@dpdk.org>; Pavan Nikhilesh <pbhagavatula@marvell.com>; Shahaf
> Shuler <shahafs@mellanox.com>; Hemant Agrawal
> <hemant.agrawal@nxp.com>; Opher Reviv <opher@mellanox.com>; Alex
> Rosenbaum <alexr@mellanox.com>; dovrat@marvell.com; Prasun Kapoor
> <pkapoor@marvell.com>; Nipun Gupta <nipun.gupta@nxp.com>; Richardson,
> Bruce <bruce.richardson@intel.com>; yang.a.hong@intel.com;
> harry.chang@intel.com; gu.jian1@zte.com.cn; shanjiangh@chinatelecom.cn;
> zhangy.yun@chinatelecom.cn; lixingfu@huachentel.com; wushuai@inspur.com;
> yuyingxia@yxlink.com; fanchenggang@sunyainfo.com;
> davidfgao@tencent.com; liuzhong1@chinaunicom.cn;
> zhaoyong11@huawei.com; oc@yunify.com; jim@netgate.com;
> hongjun.ni@intel.com; j.bromhead@titan-ic.com; deri@ntop.org;
> fc@napatech.com; arthur.su@lionic.com; Thomas Monjalon
> <thomas@monjalon.net>
> Subject: Re: [dpdk-dev] [PATCH v3] regexdev: introduce regexdev subsystem
> 
> On Sun, Feb 23, 2020 at 2:12 PM Ori Kam <orika@mellanox.com> wrote:
> >
> > Hi Jerin,
> >
> > Thanks, for the review.
> > PSB
> 
> 
> Hi Ori
> 
> Since we are finalizing the specification part, I thought of
> enumerating the list of work needs to be
> completed for a new subsystem in DPDK.
> 
> 0) Finalize the first version of the spec. Hope v4 will do that.
> 1) Introduce common library code for based on the  specification
> 2) One HW based driver implementation
> 3) One SW reference driver: libpcre library provides complete PCRE
> functionality.
> 4) app/test/test_regexdev.c like app/test/test_eventdev.c
> 5) Need a maintainer for maintaining the regex subsystem
> 6) The first version programming guide documentation
> 7) Add app/test-regexdev like app/test-eventdev
> 8) Add an examples/xxxxxx program
> 

> IMO The following items need to be completed to accept a subsystem in
> dpdk(Need at least on HW and SW driver).
> 
> 0) Finalize the first version of the spec. Hope v4 will do that.

I hope so to 😊

> 1) Introduce common library code for based on the  specification

I'm working on it. as soon as we agree on the API (this RFC will get acked ) I can work on this code.
I will send the entire code for ack when we decide if it will be part of 
20.05 or 20.08.

> 2) One HW based driver implementation

Just like you, our driver will be ready by 20.05 or 20.08

> 3) One SW reference driver: libpcre library provides complete PCRE
> functionality.

O.K. We are not working on this part.

> 4) app/test/test_regexdev.c like app/test/test_eventdev.c

We started to create a super basic app, after the API will be finalized and we will have HW
we can push it. (if you need it faster than feel free)

> 5) Need a maintainer for maintaining the regex subsystem
> 
We wish to maintain it if you agree.

> We have item (3) so Marvell would like to work on item (3). Our HW
> driver may ready by v20.05 or the worst case by 20.08.
> Let us know what other items Mellanox or community would like to work
> on. This is to avoid duplication of work
> to get clarity on the next steps.
> 

See my comments above.
From Mellanox the best date is 20.08 but we are trying to make it to 20.05, 
depended on HW.


> PSB
> 
> 
> >
> >
> > Ori Kam
> >
> > > -----Original Message-----
> > > From: Jerin Jacob <jerinjacobk@gmail.com>
> > > Sent: Saturday, February 22, 2020 6:52 PM
> > > To: Ori Kam <orika@mellanox.com>
> > > Cc: Jerin Jacob <jerinj@marvell.com>; xiang.w.wang@intel.com; dpdk-dev
> > > <dev@dpdk.org>; Pavan Nikhilesh <pbhagavatula@marvell.com>; Shahaf
> > > Shuler <shahafs@mellanox.com>; Hemant Agrawal
> > > <hemant.agrawal@nxp.com>; Opher Reviv <opher@mellanox.com>; Alex
> > > Rosenbaum <alexr@mellanox.com>; dovrat@marvell.com; Prasun Kapoor
> > > <pkapoor@marvell.com>; Nipun Gupta <nipun.gupta@nxp.com>;
> Richardson,
> > > Bruce <bruce.richardson@intel.com>; yang.a.hong@intel.com;
> > > harry.chang@intel.com; gu.jian1@zte.com.cn;
> shanjiangh@chinatelecom.cn;
> > > zhangy.yun@chinatelecom.cn; lixingfu@huachentel.com;
> wushuai@inspur.com;
> > > yuyingxia@yxlink.com; fanchenggang@sunyainfo.com;
> > > davidfgao@tencent.com; liuzhong1@chinaunicom.cn;
> > > zhaoyong11@huawei.com; oc@yunify.com; jim@netgate.com;
> > > hongjun.ni@intel.com; j.bromhead@titan-ic.com; deri@ntop.org;
> > > fc@napatech.com; arthur.su@lionic.com; Thomas Monjalon
> > > <thomas@monjalon.net>
> > > Subject: Re: [dpdk-dev] [PATCH v3] regexdev: introduce regexdev subsystem
> > >
> > > > diff --git a/lib/librte_regexdev/rte_regexdev.h
> > > b/lib/librte_regexdev/rte_regexdev.h
> > > > new file mode 100644
> > > > index 0000000..c42128b
> > > > --- /dev/null
> > > > +++ b/lib/librte_regexdev/rte_regexdev.h
> > > > @@ -0,0 +1,1411 @@
> > > > +/* SPDX-License-Identifier: BSD-3-Clause
> > > > + * Copyright(C) 2019 Marvell International Ltd.
> > > > + * Copyright(C) 2020 Mellanox International Ltd.
> > >
> > > There are a few comments from Xiang as well. So let's add Intel also
> > > to the list.
> > >
> >
> > Sure no problem.
> 
> Thanks
> 
> >
> > > > + */
> > > > +
> > > > +#ifndef _RTE_REGEXDEV_H_
> > > > +#define _RTE_REGEXDEV_H_
> > >
> > > > +
> > > > +/**
> > > > + * RegEx device information
> > > > + */
> > > > +struct rte_regex_dev_info {
> > > > +       const char *driver_name; /**< RegEx driver name. */
> > > > +       struct rte_device *dev; /**< Device information. */
> > > > +       uint16_t max_matches;
> > > > +       /**< Maximum matches per scan supported by this device. */
> > > > +       uint16_t max_queue_pairs;
> > > > +       /**< Maximum queue pairs supported by this device. */
> > > > +       uint16_t max_payload_size;
> > > > +       /**< Maximum payload size for a pattern match request or scan.
> > > > +        * @see RTE_REGEX_DEV_CFG_CROSS_BUFFER_SCAN_F
> > > > +        */
> > > > +       uint32_t max_rules_per_group;
> > > > +       /**< Maximum rules supported per group by this device.
> > > > +        * This number can't be larger then 20 bits.
> > >
> > > s/then/than
> > >
> > > I think, we don't need to say this " This number can't be larger than 20 bits."
> > > It may help SW drivers.
> > >
> >
> > Agree I will remove the 20 bits part.
> >
> > >
> > >
> > > > +        */
> > > > +       uint16_t max_groups;
> > > > +       /**< Maximum group supported by this device.
> > > > +        * This number can't be larger then 12 bits.
> > > s/then/than
> > > I think, we don't need to say this " This number can't be larger than 12 bits."
> > > It may help SW drivers.
> > >
> >
> > Agree will remove the 12 bits part.
> >
> > > > +        */
> > > > +       uint32_t regex_dev_capa;
> > > > +       /**< RegEx device capabilities. @see RTE_REGEX_DEV_CAPA_* */
> > > > +       uint64_t rule_flags;
> > > > +       /**< Supported compiler rule flags.
> > > > +        * @see RTE_REGEX_PCRE_RULE_*, struct
> rte_regex_rule::rule_flags
> > > > +        */
> > > > +       uint8_t max_scatter_gather;
> > > > +       /**< The max supported number of buffers that can
> > > > +        * be used in a single ops. The total size of all elements
> > > > +        * must be less then max_payload_size.
> > > > +        */
> > > > +};
> > > <snip>
> > >
> > > > +int
> > > > +rte_regex_rule_db_compile(uint8_t dev_id);
> > > > +
> > >
> > > I think your "rte_regex_rule_db_compile_activate() - compile and
> > > activate the new rule set"
> > > API name looks good. I am for rte_regex_rule_db_compile_activate().
> > >
> >
> > I like your name, will change to compile_activate.
> 
> Ack.
> 
> >
> > >
> > > > +/* Fast path APIs */
> > > > +
> > > > +/**
> > > > + * The generic *rte_regex_match* structure to hold the RegEx match
> > > attributes.
> > > > + * @see struct rte_regex_ops::matches
> > > > + */
> > > > +struct rte_regex_match {
> > > > +       RTE_STD_C11
> > > > +       union {
> > > > +               uint64_t u64;
> > > > +               struct {
> > > > +                       uint32_t rule_id:20;
> > > > +                       /**< Rule identifier to which the pattern matched.
> > > > +                        * @see struct rte_regex_rule::rule_id
> > > > +                        */
> > > > +                       uint32_t group_id:12;
> > > > +                       /**< Group identifier of the rule which the pattern
> > > > +                        * matched. @see struct rte_regex_rule::group_id
> > > > +                        */
> > > > +                       uint16_t offset;
> > >
> > > Since we have end_offset now, IMO, it is better to change this offset
> > > to "start_offset".
> > >
> >
> > Agree, will change.
> >
> > >
> > > > +                       /**< Starting Byte Position for matched rule. */
> > > > +                       RTE_STD_C11
> > > > +                       union {
> > > > +                               uint16_t len;
> > > > +                               /**< Length of match in bytes */
> > > > +                               uint16_t end_offset;
> > > > +                               /**< The end offset of the match. In case
> > > > +                                * MATCH_AS_START configuration is disabled.
> > > > +                                * @see RTE_REGEX_DEV_CFG_MATCH_AS_START
> > > > +                                */
> > >
> > > We have not concluded on this scheme. Have one field which has
> > > different meaning will be difficult
> > > for application. i.e fast path we need to have a check for this.
> > >
> >
> > This is the time to conclude . at least for the first version.
> > Why do we have one field with different meaning?
> > The result can be ether len or end_offset.
> >
> > > I think, Based on the majority of HW/SW implementation, we need to
> > > either go with len or
> > > end_offset. What Mellanox HW returns? len or end_offset?
> > >
> >
> > From Mellanox perspective we prefer the len approach. We also think
> > it is much more user oriented.
> >
> > > or We can keep it as len or end_offset based on which drivers upstream
> first,
> > > other drivers when it comes, we can see how to abstract it?
> > >
> >
> > I can except that assuming we choose the start and len approach
> 
> I think, we can have first version with "start and len" by removing
> RTE_REGEX_DEV_CFG_MATCH_AS_START.
> When can think, how to abstract new drivers when it upstream based on
> the overhead.
> 

Perfect

> 
> >
> > > > +                       };
> > > > +               };
> > > > +       };
> > > > +};
> > >
> > > > +/**
> > > > + * The generic *rte_regex_ops* structure to hold the RegEx attributes
> > > > + * for enqueue and dequeue operation.
> > > > + */
> > > > +struct rte_regex_ops {
> > > > +       /* W0 */
> > > > +       uint16_t req_flags;
> > > > +       /**< Request flags for the RegEx ops.
> > > > +        * @see RTE_REGEX_OPS_REQ_*
> > > > +        */
> > > > +       uint16_t rsp_flags;
> > > > +       /**< Response flags for the RegEx ops.
> > > > +        * @see RTE_REGEX_OPS_RSP_*
> > > > +        */
> > > > +       uint16_t nb_actual_matches;
> > > > +       /**< The total number of actual matches detected by the Regex
> > > device.*/
> > > > +       uint16_t nb_matches;
> > > > +       /**< The total number of matches returned by the RegEx device for
> this
> > > > +        * scan. The size of *rte_regex_ops::matches* zero length array will
> be
> > > > +        * this value.
> > > > +        *
> > > > +        * @see struct rte_regex_ops::matches, struct rte_regex_match
> > > > +        */
> > > > +
> > > > +       /* W1 */
> > > > +       uint16_t num_of_bufs;
> > > > +       /**< The number of bufs that are part of this ops. The total size of
> > > > +        * the length of all the buffer must be smaller then the max buffer
> > > > +        * len.
> > > > +        */
> > > > +       uint16_t resv1;
> > > > +       uint32_t resv2;
> > >
> > > One of the point came up in our implementation is that.
> > > HW can return an error due to various reasons.
> > >
> > > One option could be to make nb_matches as zero? and update some flag?
> > >
> > > What are your thoughts? updating the flag may be overkill.
> > >
> >
> > I think we can return just zero matches for now.
> 
> Ack.
> 
> >
> > >
> > > > +
> > > > +       /* W2 */
> > > > +       struct rte_regex_iov *(*bufs)[];
> > > > +       /**< Holds a pointer to the buffers list.*/
> > >
> > > This memory gets submitted to HW so it can not be from the heap.
> > > Cryptodev had a similar dilemma to use the container format for
> > > multi-segment case, Finally they choose to with mbuf.
> > >
> > > The following elements are in mbuf. Considering to avoid duplication and
> > > avoid overhead most common usecase DPI(Assume if it is rte_regex_iov,
> > > one need to copy all the elements from mbuf on fastpath).
> > > I propose to have mbuf here instead of rte_regex_iov.
> > >
> >
> > The application only needs to set the data pointers. (no copy is required. )
> > I agree that there are advantages to the mbuf approach.
> > The main limitation for the mbufs approach is that the user will need to play
> with the offset
> > pointers and pointers to the next mbuf, in order to support cross buffer.
> > For example we have a packet and we want to add to the scan also the last
> part of the previous packet,
> > this means that the application must modify the data offset in the previous
> packet mbuf including
> > changing the next pointer to point to the head of the new packet, and then
> return the values to the original position.
> >
> > What do you think?
> > We can start with mbufs and see how it works, or start with the buffer and
> see how it works.
> 
> I think, we can start with mbuf to align with other subsystems. We
> will see later the use case for struct rte_regex_iov.
>

Agree.
 
> 
> >
> >
> >
> >
> > > struct rte_regex_iov {
> > > RTE_STD_C11
> > > union {
> > > uint64_t u64;
> > > /**<  Allow 8-byte reserved on 32-bit system */
> > > void *buf_addr;
> > > /**< Virtual address of the pattern to be matched. */
> > > };
> > > rte_iova_t buf_iova;
> > > /**< IOVA address of the pattern to be matched. */
> > > uint16_t buf_size; /**< The buf size. */
> > > };
> > >
> > >
> > > > +
> > > > +       /* W5 */
> > > > +       RTE_STD_C11
> > > > +       union {
> > > > +               uint64_t cross_buf_id;
> > > > +               /**< ID used by the RegEx device in order to handle cross
> > > > +                * buffer detection.
> > > > +                * This ID is given by the RegEx device on dequeue, and
> > > > +                * the application must send it on the following enque.
> > > > +                */
> > > > +               void *cross_buf_ptr;
> > > > +               /**< Pointer representation of *cross_buf_id* */
> > >
> > > Could you have some example of how to use cross_buf_id?
> > > Marvell HW does not support cross_buf_id, so we need to add this
> > > feature as capability.
> > >
> >
> > The idea is that this buffer will be used to keep some internal data for the
> engine.
> > For example the current state and what was found until now, and then reuse
> this
> > for the next buffer.
> > We can remove it for now if we agree that we can add it later.
> 
> I think, adding it later would be better so that we can see how to
> abstract it well.
> 

Agree.

> >
> >
> >
> > One more thing, regarding the ops structure, I think it is better to split it to 2
> different
> > structures one enque and one for dequeue, since there are no real shared
> data and we will
> > be able to save memory, what do you think?
> 
> Ops are allocated from mempool so it will be overhead to manage both.
> moreover, some
> of the fields added in req can be used for resp as info. cryptodev
> follows the similar concept,
> I think, we can have symmetry with cryptodev wherever is possible to avoid
> end-user to learn new API models.

True that there will be overhead with 2 mempools (small one)
but lets assume 255 results. This means that the buffer should be 255 * sizeof(rte_regex_match) = 2K
also this will enable us to replace groupX with group[] which will allow even more groups.
In addition don't think that crypto is a good example.
The main difference is that in RegEx the output is different format then the input.


> 
> 
> 
> >
> > >
> > >
> > > Regarding the rule attributes, We think, following needs to be added to
> control
> > > the rule compilation behavior. If it not converging in first look, I
> > > think, we can make
> > > below as separate patch once we have basic things.
> > >
> >
> > Regarding the new code, we need also to add a function to get the
> capabilities for the compiler or
> > add a new field in the dev_info which will report the complier supported
> features.
> 
> I agree. Lets remove this new code from the first version. We can add
> it later with capability as
> a new patch.
> 

Agree

> I assume you will send the v4 with these comments. I think, with v4 we
> can start implementing common library code.

Just need to agree on the split (one more iteration 😊)
and I will start working on the common code.
Jerin Jacob Feb. 25, 2020, 5:57 a.m. UTC | #5
> > 4) app/test/test_regexdev.c like app/test/test_eventdev.c
>
> We started to create a super basic app, after the API will be finalized and we will have HW
> we can push it. (if you need it faster than feel free)

A simple Unit test case needs to be present for the APIs. On the
course of developing common code,
it can be developed to test the common code with dummy/skeleton driver.

>
> > 5) Need a maintainer for maintaining the regex subsystem
> >
> We wish to maintain it if you agree.

Yes. Please.

> > >
> > > One more thing, regarding the ops structure, I think it is better to split it to 2
> > different
> > > structures one enque and one for dequeue, since there are no real shared
> > data and we will
> > > be able to save memory, what do you think?
> >
> > Ops are allocated from mempool so it will be overhead to manage both.
> > moreover, some
> > of the fields added in req can be used for resp as info. cryptodev
> > follows the similar concept,
> > I think, we can have symmetry with cryptodev wherever is possible to avoid
> > end-user to learn new API models.
>
> True that there will be overhead with 2 mempools (small one)
> but lets assume 255 results. This means that the buffer should be 255 * sizeof(rte_regex_match) = 2K
> also this will enable us to replace groupX with group[] which will allow even more groups.
> In addition don't think that crypto is a good example.
> The main difference is that in RegEx the output is different format then the input.

# IMO, Some of the fields may be useful for a response as well. I
think application may be interested in following
req filed in the response.
a) buf_addr
b) scan_size
c) user_id (This would be main one)

# Having two mempools adds overhead per lcore L1 cache usage and extra
complexity to the application.

# IMO, From a performance perspective, one mempool is good due to less
stress on the cache and it is costly to
add new mempool for HW mempool implementations.

# I think, group[] use case we can add it when it required by
introducing "matches_start_offset" field, which will
tell the req, where is the end of group[] and where "matches" start
with single mempool scheme also.

# I think, one of the other use case for "matches_start_offset" that,
It may possible to put vendor-specific
opaque data. It will be filled by driver on response. The application
can reference the matches as

struct rte_regex_match *matches = RTE_PTR_ADD(ops, ops->matches_start_offset);

>
> > I assume you will send the v4 with these comments. I think, with v4 we
> > can start implementing common library code.
>
> Just need to agree on the split (one more iteration )
> and I will start working on the common code.

Ack.
Ori Kam Feb. 25, 2020, 7:48 a.m. UTC | #6
> -----Original Message-----
> From: Jerin Jacob <jerinjacobk@gmail.com>
> Sent: Tuesday, February 25, 2020 7:57 AM
> To: Ori Kam <orika@mellanox.com>
> Cc: Jerin Jacob <jerinj@marvell.com>; xiang.w.wang@intel.com; dpdk-dev
> <dev@dpdk.org>; Pavan Nikhilesh <pbhagavatula@marvell.com>; Shahaf
> Shuler <shahafs@mellanox.com>; Hemant Agrawal
> <hemant.agrawal@nxp.com>; Opher Reviv <opher@mellanox.com>; Alex
> Rosenbaum <alexr@mellanox.com>; dovrat@marvell.com; Prasun Kapoor
> <pkapoor@marvell.com>; Nipun Gupta <nipun.gupta@nxp.com>; Richardson,
> Bruce <bruce.richardson@intel.com>; yang.a.hong@intel.com;
> harry.chang@intel.com; gu.jian1@zte.com.cn; shanjiangh@chinatelecom.cn;
> zhangy.yun@chinatelecom.cn; lixingfu@huachentel.com; wushuai@inspur.com;
> yuyingxia@yxlink.com; fanchenggang@sunyainfo.com;
> davidfgao@tencent.com; liuzhong1@chinaunicom.cn;
> zhaoyong11@huawei.com; oc@yunify.com; jim@netgate.com;
> hongjun.ni@intel.com; j.bromhead@titan-ic.com; deri@ntop.org;
> fc@napatech.com; arthur.su@lionic.com; Thomas Monjalon
> <thomas@monjalon.net>
> Subject: Re: [dpdk-dev] [PATCH v3] regexdev: introduce regexdev subsystem
> 
> > > 4) app/test/test_regexdev.c like app/test/test_eventdev.c
> >
> > We started to create a super basic app, after the API will be finalized and we
> will have HW
> > we can push it. (if you need it faster than feel free)
> 
> A simple Unit test case needs to be present for the APIs. On the
> course of developing common code,
> it can be developed to test the common code with dummy/skeleton driver.
> 

Agree this is what we are currently have.

> >
> > > 5) Need a maintainer for maintaining the regex subsystem
> > >
> > We wish to maintain it if you agree.
> 
> Yes. Please.
> 

Great.

> > > >
> > > > One more thing, regarding the ops structure, I think it is better to split it
> to 2
> > > different
> > > > structures one enque and one for dequeue, since there are no real shared
> > > data and we will
> > > > be able to save memory, what do you think?
> > >
> > > Ops are allocated from mempool so it will be overhead to manage both.
> > > moreover, some
> > > of the fields added in req can be used for resp as info. cryptodev
> > > follows the similar concept,
> > > I think, we can have symmetry with cryptodev wherever is possible to avoid
> > > end-user to learn new API models.
> >
> > True that there will be overhead with 2 mempools (small one)
> > but lets assume 255 results. This means that the buffer should be 255 *
> sizeof(rte_regex_match) = 2K
> > also this will enable us to replace groupX with group[] which will allow even
> more groups.
> > In addition don't think that crypto is a good example.
> > The main difference is that in RegEx the output is different format then the
> input.
> 
> # IMO, Some of the fields may be useful for a response as well. I
> think application may be interested in following
> req filed in the response.
> a) buf_addr

I don't see how this can be used in the response. since if working in out of order result.
you don’t know which result will be returned. 
I also think it is error prone to use the same op for the enqueue and dequeue.

> b) scan_size

Please see above.

> c) user_id (This would be main one)

Agree

> 
> # Having two mempools adds overhead per lcore L1 cache usage and extra
> complexity to the application.
> 
> # IMO, From a performance perspective, one mempool is good due to less
> stress on the cache and it is costly to
> add new mempool for HW mempool implementations.
> 
> # I think, group[] use case we can add it when it required by
> introducing "matches_start_offset" field, which will
> tell the req, where is the end of group[] and where "matches" start
> with single mempool scheme also.
> 
> # I think, one of the other use case for "matches_start_offset" that,
> It may possible to put vendor-specific
> opaque data. It will be filled by driver on response. The application
> can reference the matches as
> 
> struct rte_regex_match *matches = RTE_PTR_ADD(ops, ops-
> 
>matches_start_offset);
> 

O.K for now we will keep  it as is, and we will see what will be in the future.

> >
> > > I assume you will send the v4 with these comments. I think, with v4 we
> > > can start implementing common library code.
> >
> > Just need to agree on the split (one more iteration )
> > and I will start working on the common code.
> 
> Ack.

Great,
I'm starting to work on V4 with all comments so the RFC will be acked and then will start 
coding the rest of the common code.

Patch
diff mbox series

diff --git a/config/common_base b/config/common_base
index f9a68f3..4810849 100644
--- a/config/common_base
+++ b/config/common_base
@@ -806,6 +806,12 @@  CONFIG_RTE_LIBRTE_PMD_OCTEONTX2_DMA_RAWDEV=y
 CONFIG_RTE_LIBRTE_PMD_NTB_RAWDEV=y
 
 #
+# Compile regex device support
+#
+CONFIG_RTE_LIBRTE_REGEXDEV=y
+CONFIG_RTE_LIBRTE_REGEXDEV_DEBUG=n
+
+#
 # Compile librte_ring
 #
 CONFIG_RTE_LIBRTE_RING=y
@@ -1098,3 +1104,4 @@  CONFIG_RTE_APP_CRYPTO_PERF=y
 # Compile the eventdev application
 #
 CONFIG_RTE_APP_EVENTDEV=y
+
diff --git a/doc/api/doxy-api-index.md b/doc/api/doxy-api-index.md
index dff496b..787f7c2 100644
--- a/doc/api/doxy-api-index.md
+++ b/doc/api/doxy-api-index.md
@@ -26,6 +26,7 @@  The public API headers are grouped by topics:
   [event_timer_adapter]    (@ref rte_event_timer_adapter.h),
   [event_crypto_adapter]   (@ref rte_event_crypto_adapter.h),
   [rawdev]             (@ref rte_rawdev.h),
+  [regexdev]           (@ref rte_regexdev.h),
   [metrics]            (@ref rte_metrics.h),
   [bitrate]            (@ref rte_bitrate.h),
   [latency]            (@ref rte_latencystats.h),
diff --git a/doc/api/doxy-api.conf.in b/doc/api/doxy-api.conf.in
index 1c4392e..56c08eb 100644
--- a/doc/api/doxy-api.conf.in
+++ b/doc/api/doxy-api.conf.in
@@ -58,6 +58,7 @@  INPUT                   = @TOPDIR@/doc/api/doxy-api-index.md \
                           @TOPDIR@/lib/librte_rcu \
                           @TOPDIR@/lib/librte_reorder \
                           @TOPDIR@/lib/librte_rib \
+                          @TOPDIR@/lib/librte_regexdev \
                           @TOPDIR@/lib/librte_ring \
                           @TOPDIR@/lib/librte_sched \
                           @TOPDIR@/lib/librte_security \
diff --git a/lib/Makefile b/lib/Makefile
index 46b91ae..82ff950 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -44,6 +44,8 @@  DEPDIRS-librte_eventdev := librte_eal librte_ring librte_ethdev librte_hash \
                            librte_mempool librte_timer librte_cryptodev
 DIRS-$(CONFIG_RTE_LIBRTE_RAWDEV) += librte_rawdev
 DEPDIRS-librte_rawdev := librte_eal librte_ethdev
+DIRS-$(CONFIG_RTE_LIBRTE_REGEXDEV) += librte_regexdev
+DEPDIRS-librte_regexdev := librte_eal
 DIRS-$(CONFIG_RTE_LIBRTE_VHOST) += librte_vhost
 DEPDIRS-librte_vhost := librte_eal librte_mempool librte_mbuf librte_ethdev \
 			librte_net librte_hash librte_cryptodev
diff --git a/lib/librte_regexdev/Makefile b/lib/librte_regexdev/Makefile
new file mode 100644
index 0000000..90b4029
--- /dev/null
+++ b/lib/librte_regexdev/Makefile
@@ -0,0 +1,27 @@ 
+# SPDX-License-Identifier: BSD-3-Clause
+# Copyright(C) 2019 Marvell International Ltd.
+# Copyright(C) 2020 Mellanox International Ltd.
+#
+
+include $(RTE_SDK)/mk/rte.vars.mk
+
+# library name
+LIB = librte_regexdev.a
+
+EXPORT_MAP := rte_regex_version.map
+
+# library version
+LIBABIVER := 1
+
+# build flags
+CFLAGS += -O3
+CFLAGS += $(WERROR_FLAGS)
+
+# library source files
+# all source are stored in SRCS-y
+SRCS-$(CONFIG_RTE_LIBRTE_REGEXDEV) := rte_regexdev.c
+
+# export include files
+SYMLINK-$(CONFIG_RTE_LIBRTE_REGEXDEV)-include += rte_regexdev.h
+
+include $(RTE_SDK)/mk/rte.lib.mk
diff --git a/lib/librte_regexdev/rte_regexdev.c b/lib/librte_regexdev/rte_regexdev.c
new file mode 100644
index 0000000..b901877
--- /dev/null
+++ b/lib/librte_regexdev/rte_regexdev.c
@@ -0,0 +1,6 @@ 
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(C) 2019 Marvell International Ltd.
+ * Copyright(C) 2020 Mellanox International Ltd.
+ */
+
+#include <rte_regexdev.h>
diff --git a/lib/librte_regexdev/rte_regexdev.h b/lib/librte_regexdev/rte_regexdev.h
new file mode 100644
index 0000000..c42128b
--- /dev/null
+++ b/lib/librte_regexdev/rte_regexdev.h
@@ -0,0 +1,1411 @@ 
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(C) 2019 Marvell International Ltd.
+ * Copyright(C) 2020 Mellanox International Ltd.
+ */
+
+#ifndef _RTE_REGEXDEV_H_
+#define _RTE_REGEXDEV_H_
+
+/**
+ * @file
+ *
+ * RTE RegEx Device API
+ *
+ * Defines RTE RegEx Device APIs for RegEx operations and its provisioning.
+ *
+ * The RegEx Device API is composed of two parts:
+ *
+ * - The application-oriented RegEx API that includes functions to setup
+ *   a RegEx device (configure it, setup its queue pairs and start it),
+ *   update the rule database and so on.
+ *
+ * - The driver-oriented RegEx API that exports a function allowing
+ *   a RegEx poll Mode Driver (PMD) to simultaneously register itself as
+ *   a RegEx device driver.
+ *
+ * RegEx device components and definitions:
+ *
+ *     +-----------------+
+ *     |                 |
+ *     |                 o---------+    rte_regex_[en|de]queue_burst()
+ *     |   PCRE based    o------+  |               |
+ *     |  RegEx pattern  |      |  |  +--------+   |
+ *     | matching engine o------+--+--o        |   |    +------+
+ *     |                 |      |  |  | queue  |<==o===>|Core 0|
+ *     |                 o----+ |  |  | pair 0 |        |      |
+ *     |                 |    | |  |  +--------+        +------+
+ *     +-----------------+    | |  |
+ *            ^               | |  |  +--------+
+ *            |               | |  |  |        |        +------+
+ *            |               | +--+--o queue  |<======>|Core 1|
+ *        Rule|Database       |    |  | pair 1 |        |      |
+ *     +------+----------+    |    |  +--------+        +------+
+ *     |     Group 0     |    |    |
+ *     | +-------------+ |    |    |  +--------+        +------+
+ *     | | Rules 0..n  | |    |    |  |        |        |Core 2|
+ *     | +-------------+ |    |    +--o queue  |<======>|      |
+ *     |     Group 1     |    |       | pair 2 |        +------+
+ *     | +-------------+ |    |       +--------+
+ *     | | Rules 0..n  | |    |
+ *     | +-------------+ |    |       +--------+
+ *     |     Group 2     |    |       |        |        +------+
+ *     | +-------------+ |    |       | queue  |<======>|Core n|
+ *     | | Rules 0..n  | |    +-------o pair n |        |      |
+ *     | +-------------+ |            +--------+        +------+
+ *     |     Group n     |
+ *     | +-------------+ |<-------rte_regex_rule_db_update()
+ *     | | Rules 0..n  | |<-------rte_regex_rule_db_import()
+ *     | +-------------+ |------->rte_regex_rule_db_export()
+ *     +-----------------+
+ *
+ * RegEx: A regular expression is a concise and flexible means for matching
+ * strings of text, such as particular characters, words, or patterns of
+ * characters. A common abbreviation for this is “RegEx”.
+ *
+ * RegEx device: A hardware or software-based implementation of RegEx
+ * device API for PCRE based pattern matching syntax and semantics.
+ *
+ * PCRE RegEx syntax and semantics specification:
+ * http://regexkit.sourceforge.net/Documentation/pcre/pcrepattern.html
+ *
+ * RegEx queue pair: Each RegEx device should have one or more queue pair to
+ * transmit a burst of pattern matching request and receive a burst of
+ * receive the pattern matching response. The pattern matching request/response
+ * embedded in *rte_regex_ops* structure.
+ *
+ * Rule: A pattern matching rule expressed in PCRE RegEx syntax along with
+ * Match ID and Group ID to identify the rule upon the match.
+ *
+ * Rule database: The RegEx device accepts regular expressions and converts them
+ * into a compiled rule database that can then be used to scan data.
+ * Compilation allows the device to analyze the given pattern(s) and
+ * pre-determine how to scan for these patterns in an optimized fashion that
+ * would be far too expensive to compute at run-time. A rule database contains
+ * a set of rules that compiled in device specific binary form.
+ *
+ * Match ID or Rule ID: A unique identifier provided at the time of rule
+ * creation for the application to identify the rule upon match.
+ *
+ * Group ID: Group of rules can be grouped under one group ID to enable
+ * rule isolation and effective pattern matching. A unique group identifier
+ * provided at the time of rule creation for the application to identify the
+ * rule upon match.
+ *
+ * Scan: A pattern matching request through *enqueue* API.
+ *
+ * It may possible that a given RegEx device may not support all the features
+ * of PCRE. The application may probe unsupported features through
+ * struct rte_regex_dev_info::pcre_unsup_flags
+ *
+ * By default, all the functions of the RegEx Device API exported by a PMD
+ * are lock-free functions which assume to not be invoked in parallel on
+ * different logical cores to work on the same target object. For instance,
+ * the dequeue function of a PMD cannot be invoked in parallel on two logical
+ * cores to operates on same RegEx queue pair. Of course, this function
+ * can be invoked in parallel by different logical core on different queue pair.
+ * It is the responsibility of the upper level application to enforce this rule.
+ *
+ * In all functions of the RegEx API, the RegEx device is
+ * designated by an integer >= 0 named the device identifier *dev_id*
+ *
+ * At the RegEx driver level, RegEx devices are represented by a generic
+ * data structure of type *rte_regex_dev*.
+ *
+ * RegEx devices are dynamically registered during the PCI/SoC device probing
+ * phase performed at EAL initialization time.
+ * When a RegEx device is being probed, a *rte_regex_dev* structure and
+ * a new device identifier are allocated for that device. Then, the
+ * regex_dev_init() function supplied by the RegEx driver matching the probed
+ * device is invoked to properly initialize the device.
+ *
+ * The role of the device init function consists of resetting the hardware or
+ * software RegEx driver implementations.
+ *
+ * If the device init operation is successful, the correspondence between
+ * the device identifier assigned to the new device and its associated
+ * *rte_regex_dev* structure is effectively registered.
+ * Otherwise, both the *rte_regex_dev* structure and the device identifier are
+ * freed.
+ *
+ * The functions exported by the application RegEx API to setup a device
+ * designated by its device identifier must be invoked in the following order:
+ *     - rte_regex_dev_configure()
+ *     - rte_regex_queue_pair_setup()
+ *     - rte_regex_dev_start()
+ *
+ * Then, the application can invoke, in any order, the functions
+ * exported by the RegEx API to enqueue pattern matching job, dequeue pattern
+ * matching response, get the stats, update the rule database,
+ * get/set device attributes and so on
+ *
+ * If the application wants to change the configuration (i.e. call
+ * rte_regex_dev_configure() or rte_regex_queue_pair_setup()), it must call
+ * rte_regex_dev_stop() first to stop the device and then do the reconfiguration
+ * before calling rte_regex_dev_start() again. The enqueue and dequeue
+ * functions should not be invoked when the device is stopped.
+ *
+ * Finally, an application can close a RegEx device by invoking the
+ * rte_regex_dev_close() function.
+ *
+ * Each function of the application RegEx API invokes a specific function
+ * of the PMD that controls the target device designated by its device
+ * identifier.
+ *
+ * For this purpose, all device-specific functions of a RegEx driver are
+ * supplied through a set of pointers contained in a generic structure of type
+ * *regex_dev_ops*.
+ * The address of the *regex_dev_ops* structure is stored in the *rte_regex_dev*
+ * structure by the device init function of the RegEx driver, which is
+ * invoked during the PCI/SoC device probing phase, as explained earlier.
+ *
+ * In other words, each function of the RegEx API simply retrieves the
+ * *rte_regex_dev* structure associated with the device identifier and
+ * performs an indirect invocation of the corresponding driver function
+ * supplied in the *regex_dev_ops* structure of the *rte_regex_dev* structure.
+ *
+ * For performance reasons, the address of the fast-path functions of the
+ * RegEx driver is not contained in the *regex_dev_ops* structure.
+ * Instead, they are directly stored at the beginning of the *rte_regex_dev*
+ * structure to avoid an extra indirect memory access during their invocation.
+ *
+ * RTE RegEx device drivers do not use interrupts for enqueue or dequeue
+ * operation. Instead, RegEx drivers export Poll-Mode enqueue and dequeue
+ * functions to applications.
+ *
+ * The *enqueue* operation submits a burst of RegEx pattern matching request
+ * to the RegEx device and the *dequeue* operation gets a burst of pattern
+ * matching response for the ones submitted through *enqueue* operation.
+ *
+ * Typical application utilisation of the RegEx device API will follow the
+ * following programming flow.
+ *
+ * - rte_regex_dev_configure()
+ * - rte_regex_queue_pair_setup()
+ * - rte_regex_rule_db_update() Needs to invoke if precompiled rule database not
+ *   provided in rte_regex_dev_config::rule_db for rte_regex_dev_configure()
+ *   and/or application needs to update rule database.
+ * - Create or reuse exiting mempool for *rte_regex_ops* objects.
+ * - rte_regex_dev_start()
+ * - rte_regex_enqueue_burst()
+ * - rte_regex_dequeue_burst()
+ *
+ */
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <rte_common.h>
+#include <rte_config.h>
+#include <rte_dev.h>
+#include <rte_errno.h>
+#include <rte_memory.h>
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice.
+ *
+ * Get the total number of RegEx devices that have been successfully
+ * initialised.
+ *
+ * @return
+ *   The total number of usable RegEx devices.
+ */
+__rte_experimental
+uint8_t
+rte_regex_dev_count(void);
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice.
+ *
+ * Get the device identifier for the named RegEx device.
+ *
+ * @param name
+ *   RegEx device name to select the RegEx device identifier.
+ *
+ * @return
+ *   Returns RegEx device identifier on success.
+ *   - <0: Failure to find named RegEx device.
+ */
+__rte_experimental
+int
+rte_regex_dev_get_dev_id(const char *name);
+
+/* Enumerates RegEx device capabilities */
+#define RTE_REGEX_DEV_CAPA_RUNTIME_COMPILATION_F (1ULL << 0)
+/**< RegEx device does support compiling the rules at runtime unlike
+ * loading only the pre-built rule database using
+ * struct rte_regex_dev_config::rule_db in rte_regex_dev_configure()
+ * @see struct rte_regex_dev_config::rule_db, rte_regex_dev_configure()
+ * @see struct rte_regex_dev_info::regex_dev_capa
+ */
+
+#define RTE_REGEX_DEV_CAPA_SUPP_PCRE_START_ANCHOR_F (1ULL << 1)
+/**< RegEx device support PCRE Anchor to start of match flag.
+ * Example RegEx is '/\Gfoo\d/'. Here '\G' asserts position at the end of the
+ * previous match or the start of the string for the first match.
+ * This position will change each time the RegEx is applied to the subject
+ * string. If the RegEx is applied to 'foo1foo2Zfoo3' the first two matches will
+ * be successful for 'foo1foo2' and fail for 'Zfoo3'.
+ * @see struct rte_regex_dev_info::regex_dev_capa
+ */
+
+#define RTE_REGEX_DEV_CAPA_SUPP_PCRE_ATOMIC_GROUPING_F (1ULL << 2)
+/**< RegEx device support PCRE Atomic grouping.
+ * Atomic groups are represented by '(?>)'. An atomic group is a group that,
+ * when the RegEx engine exits from it, automatically throws away all
+ * backtracking positions remembered by any tokens inside the group.
+ * Example RegEx is 'a(?>bc|b)c' if the given patterns are 'abc' and 'abcc' then
+ * 'a(bc|b)c' matches both where as 'a(?>bc|b)c' matches only abcc because
+ * atomic groups don't allow backtracing back to 'b'.
+ * @see struct rte_regex_dev_info::regex_dev_capa
+ */
+
+#define RTE_REGEX_DEV_SUPP_PCRE_BACKTRACKING_CTRL_F (1ULL << 3)
+/**< RegEx device support PCRE backtracking control verbs.
+ * Some examples of backtracing verbs are (*COMMIT), (*ACCEPT), (*FAIL),
+ * (*SKIP), (*PRUNE).
+ * @see struct rte_regex_dev_info::regex_dev_capa
+ */
+
+#define RTE_REGEX_DEV_SUPP_PCRE_CALLOUTS_F (1ULL << 4)
+/**< RegEx device support PCRE callouts.
+ * PCRE supports calling external function in between matches by using '(?C)'.
+ * Example RegEx 'ABC(?C)D' if a given patter is 'ABCD' then the RegEx engine
+ * will parse ABC perform a userdefined callout and return a successful match at
+ * D.
+ * @see struct rte_regex_dev_info::regex_dev_capa
+ */
+
+#define RTE_REGEX_DEV_SUPP_PCRE_BACKREFERENCE_F (1ULL << 5)
+/**< RegEx device support PCRE backreference.
+ * Example RegEx is '(\2ABC|(GHI))+' \2 matches the same text as most recently
+ * matched by the 2nd capturing group i.e. 'GHI'.
+ * @see struct rte_regex_dev_info::regex_dev_capa
+ */
+
+#define RTE_REGEX_DEV_SUPP_PCRE_GREEDY_F (1ULL << 6)
+/**< RegEx device support PCRE Greedy mode.
+ * For example if the RegEx is 'AB\d*?' then '*?' represents zero or unlimited
+ * matches. In greedy mode the pattern 'AB12345' will be matched completely
+ * where as the ungreedy mode 'AB' will be returned as the match.
+ * @see struct rte_regex_dev_info::regex_dev_capa
+ */
+
+#define RTE_REGEX_DEV_SUPP_PCRE_LOOKAROUND_ASRT_F (1ULL << 7)
+/**< RegEx device support PCRE Lookaround assertions
+ * (Zero-width assertions). Example RegEx is '[a-z]+\d+(?=!{3,})' if
+ * the given pattern is 'dwad1234!' the RegEx engine doesn't report any matches
+ * because the assert '(?=!{3,})' fails. The pattern 'dwad123!!!' would return a
+ * successful match.
+ * @see struct rte_regex_dev_info::regex_dev_capa
+ */
+
+#define RTE_REGEX_DEV_SUPP_PCRE_MATCH_POINT_RST_F (1ULL << 8)
+/**< RegEx device doesn't support PCRE match point reset directive.
+ * Example RegEx is '[a-z]+\K\d+' if the pattern is 'dwad123'
+ * then even though the entire pattern matches only '123'
+ * is reported as a match.
+ * @see struct rte_regex_dev_info::regex_dev_capa
+ */
+
+#define RTE_REGEX_DEV_SUPP_NEWLINE_CONVENTIONS_F (1ULL << 9)
+/**< RegEx support PCRE newline convention.
+ * Newline conventions are represented as follows:
+ * (*CR)        carriage return
+ * (*LF)        linefeed
+ * (*CRLF)      carriage return, followed by linefeed
+ * (*ANYCRLF)   any of the three above
+ * (*ANY)       all Unicode newline sequences
+ * @see struct rte_regex_dev_info::regex_dev_capa
+ */
+
+#define RTE_REGEX_DEV_SUPP_PCRE_NEWLINE_SEQ_F (1ULL << 10)
+/**< RegEx device support PCRE newline sequence.
+ * The escape sequence '\R' will match any newline sequence.
+ * It is equivalent to: '(?>\r\n|\n|\x0b|\f|\r|\x85)'.
+ * @see struct rte_regex_dev_info::regex_dev_capa
+ */
+
+#define RTE_REGEX_DEV_SUPP_PCRE_POSSESSIVE_QUALIFIERS_F (1ULL << 11)
+/**< RegEx device support PCRE possessive qualifiers.
+ * Example RegEx possessive qualifiers '*+', '++', '?+', '{m,n}+'.
+ * Possessive quantifier repeats the token as many times as possible and it does
+ * not give up matches as the engine backtracks. With a possessive quantifier,
+ * the deal is all or nothing.
+ * @see struct rte_regex_dev_info::regex_dev_capa
+ */
+
+#define RTE_REGEX_DEV_SUPP_PCRE_SUBROUTINE_REFERENCES_F (1ULL << 12)
+/**< RegEx device support PCRE Subroutine references.
+ * PCRE Subroutine references allow for sub patterns to be assessed
+ * as part of the RegEx. Example RegEx is '(foo|fuzz)\g<1>+bar' matches the
+ * pattern 'foofoofuzzfoofuzzbar'.
+ * @see struct rte_regex_dev_info::regex_dev_capa
+ */
+
+#define RTE_REGEX_DEV_SUPP_PCRE_UTF_8_F (1ULL << 13)
+/**< RegEx device support UTF-8 character encoding.
+ * @see struct rte_regex_dev_info::pcre_unsup_flags
+ */
+
+#define RTE_REGEX_DEV_SUPP_PCRE_UTF_16_F (1ULL << 14)
+/**< RegEx device support UTF-16 character encoding.
+ * @see struct rte_regex_dev_info::regex_dev_capa
+ */
+
+#define RTE_REGEX_DEV_SUPP_PCRE_UTF_32_F (1ULL << 15)
+/**< RegEx device support UTF-32 character encoding.
+ * @see struct rte_regex_dev_info::regex_dev_capa
+ */
+
+#define RTE_REGEX_DEV_SUPP_PCRE_WORD_BOUNDARY_F (1ULL << 16)
+/**< RegEx device support word boundaries.
+ * The meta character '\b' represents word boundary anchor.
+ * @see struct rte_regex_dev_info::regex_dev_capa
+ */
+
+#define RTE_REGEX_DEV_SUPP_PCRE_FORWARD_REFERENCES_F (1ULL << 17)
+/**< RegEx device support Forward references.
+ * Forward references allow you to use a back reference to a group that appears
+ * later in the RegEx. Example RegEx is '(\3ABC|(DEF|(GHI)))+' matches the
+ * following string 'GHIGHIABCDEF'.
+ * @see struct rte_regex_dev_info::regex_dev_capa
+ */
+
+#define RTE_REGEX_DEV_SUPP_MATCH_AS_START (1ULL << 18)
+/**< RegEx device support match as start.
+ * Match as start means that the match result holds the starting position of
+ * match and the length of the match.
+ * @see struct rte_regex_dev_info::regex_dev_capa
+ */
+
+/* Enumerates PCRE rule flags */
+#define RTE_REGEX_PCRE_RULE_ALLOW_EMPTY_F (1ULL << 0)
+/**< When this flag is set, the pattern that can match against an empty string,
+ * such as '.*' are allowed.
+ * @see struct rte_regex_dev_info::rule_flags, struct rte_regex_rule::rule_flags
+ */
+
+#define RTE_REGEX_PCRE_RULE_ANCHORED_F (1ULL << 1)
+/**< When this flag is set, the pattern is forced to be "anchored", that is, it
+ * is constrained to match only at the first matching point in the string that
+ * is being searched. Similar to '^' and represented by \A.
+ * @see struct rte_regex_dev_info::rule_flags, struct rte_regex_rule::rule_flags
+ */
+
+#define RTE_REGEX_PCRE_RULE_CASELESS_F (1ULL << 2)
+/**< When this flag is set, letters in the pattern match both upper and lower
+ * case letters in the subject.
+ * @see struct rte_regex_dev_info::rule_flags, struct rte_regex_rule::rule_flags
+ */
+
+#define RTE_REGEX_PCRE_RULE_DOTALL_F (1ULL << 3)
+/**< When this flag is set, a dot metacharacter in the pattern matches any
+ * character, including one that indicates a newline.
+ * @see struct rte_regex_dev_info::rule_flags, struct rte_regex_rule::rule_flags
+ */
+
+#define RTE_REGEX_PCRE_RULE_DUPNAMES_F (1ULL << 4)
+/**< When this flag is set, names used to identify capture groups need not be
+ * unique.
+ * @see struct rte_regex_dev_info::rule_flags, struct rte_regex_rule::rule_flags
+ */
+
+#define RTE_REGEX_PCRE_RULE_EXTENDED_F (1ULL << 5)
+/**< When this flag is set, most white space characters in the pattern are
+ * totally ignored except when escaped or inside a character class.
+ * @see struct rte_regex_dev_info::rule_flags, struct rte_regex_rule::rule_flags
+ */
+
+#define RTE_REGEX_PCRE_RULE_MATCH_UNSET_BACKREF_F (1ULL << 6)
+/**< When this flag is set, a backreference to an unset capture group matches an
+ * empty string.
+ * @see struct rte_regex_dev_info::rule_flags, struct rte_regex_rule::rule_flags
+ */
+
+#define RTE_REGEX_PCRE_RULE_MULTILINE_F (1ULL << 7)
+/**< When this flag  is set, the '^' and '$' constructs match immediately
+ * following or immediately before internal newlines in the subject string,
+ * respectively, as well as at the very start and end.
+ * @see struct rte_regex_dev_info::rule_flags, struct rte_regex_rule::rule_flags
+ */
+
+#define RTE_REGEX_PCRE_RULE_NO_AUTO_CAPTURE_F (1ULL << 8)
+/**< When this Flag is set, it disables the use of numbered capturing
+ * parentheses in the pattern. References to capture groups (backreferences or
+ * recursion/subroutine calls) may only refer to named groups, though the
+ * reference can be by name or by number.
+ * @see struct rte_regex_dev_info::rule_flags, struct rte_regex_rule::rule_flags
+ */
+
+#define RTE_REGEX_PCRE_RULE_UCP_F (1ULL << 9)
+/**< By default, only ASCII characters are recognized, When this flag is set,
+ * Unicode properties are used instead to classify characters.
+ * @see struct rte_regex_dev_info::rule_flags, struct rte_regex_rule::rule_flags
+ */
+
+#define RTE_REGEX_PCRE_RULE_UNGREEDY_F (1ULL << 10)
+/**< When this flag is set, the "greediness" of the quantifiers is inverted
+ * so that they are not greedy by default, but become greedy if followed by
+ * '?'.
+ * @see struct rte_regex_dev_info::rule_flags, struct rte_regex_rule::rule_flags
+ */
+
+#define RTE_REGEX_PCRE_RULE_UTF_F (1ULL << 11)
+/**< When this flag is set, RegEx engine has to regard both the pattern and the
+ * subject strings that are subsequently processed as strings of UTF characters
+ * instead of single-code-unit strings.
+ * @see struct rte_regex_dev_info::rule_flags, struct rte_regex_rule::rule_flags
+ */
+
+#define RTE_REGEX_PCRE_RULE_NEVER_BACKSLASH_C_F (1ULL << 12)
+/**< This Flag locks out the use of '\C' in the pattern that is being compiled.
+ * This escape matches one data unit, even in UTF mode which can cause
+ * unpredictable behavior in UTF-8 or UTF-16 modes, because it may leave the
+ * current matching point in the middle of a multi-code-unit character.
+ * @see struct rte_regex_dev_info::rule_flags, struct rte_regex_rule::rule_flags
+ */
+
+
+/**
+ * RegEx device information
+ */
+struct rte_regex_dev_info {
+	const char *driver_name; /**< RegEx driver name. */
+	struct rte_device *dev;	/**< Device information. */
+	uint16_t max_matches;
+	/**< Maximum matches per scan supported by this device. */
+	uint16_t max_queue_pairs;
+	/**< Maximum queue pairs supported by this device. */
+	uint16_t max_payload_size;
+	/**< Maximum payload size for a pattern match request or scan.
+	 * @see RTE_REGEX_DEV_CFG_CROSS_BUFFER_SCAN_F
+	 */
+	uint32_t max_rules_per_group;
+	/**< Maximum rules supported per group by this device.
+	 * This number can't be larger then 20 bits.
+	 */
+	uint16_t max_groups;
+	/**< Maximum group supported by this device.
+	 * This number can't be larger then 12 bits.
+	 */
+	uint32_t regex_dev_capa;
+	/**< RegEx device capabilities. @see RTE_REGEX_DEV_CAPA_* */
+	uint64_t rule_flags;
+	/**< Supported compiler rule flags.
+	 * @see RTE_REGEX_PCRE_RULE_*, struct rte_regex_rule::rule_flags
+	 */
+	uint8_t max_scatter_gather;
+	/**< The max supported number of buffers that can
+	 * be used in a single ops. The total size of all elements
+	 * must be less then max_payload_size.
+	 */
+};
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice.
+ *
+ * Retrieve the contextual information of a RegEx device.
+ *
+ * @param dev_id
+ *   The identifier of the device.
+ *
+ * @param[out] dev_info
+ *   A pointer to a structure of type *rte_regex_dev_info* to be filled with the
+ *   contextual information of the device.
+ *
+ * @return
+ *   - 0: Success, driver updates the contextual information of the RegEx device
+ *   - <0: Error code returned by the driver info get function.
+ *
+ */
+__rte_experimental
+int
+rte_regex_dev_info_get(uint8_t dev_id, struct rte_regex_dev_info *dev_info);
+
+/* Enumerates RegEx device configuration flags */
+#define RTE_REGEX_DEV_CFG_CROSS_BUFFER_SCAN_F (1ULL << 0)
+/**< Cross buffer scan refers to the ability to be able to detect
+ * matches that occur across buffer boundaries, where the buffers are related
+ * to each other in some way. Enable this flag when to scan payload size
+ * greater struct struct rte_regex_dev_info::max_payload_size and/or
+ * matches can present across scan buffer boundaries.
+ *
+ * @see struct rte_regex_dev_info::max_payload_size
+ * @see struct rte_regex_dev_config::dev_cfg_flags, rte_regex_dev_configure()
+ * @see RTE_REGEX_OPS_RSP_PMI_SOJ_F
+ * @see RTE_REGEX_OPS_RSP_PMI_EOJ_F
+ */
+
+#define RTE_REGEX_DEV_CFG_MATCH_AS_START (1ULL << 1)
+/**< Match as start is the ability to return the result as starting offset and
+ * len. When this flag is set, the result for each match will hold the starting
+ * offset of the match in offset, and the length of the match in len.
+ * If this flag is not set, then the match result will only hold the end of
+ * the match offset in the end_offset. While the offset will be zero.
+ *
+ * @see RTE_REGEX_DEV_SUPP_MATCH_AS_START
+ */
+
+/** RegEx device configuration structure */
+struct rte_regex_dev_config {
+	uint16_t nb_max_matches;
+	/**< Maximum matches per scan configured on this device.
+	 * This value cannot exceed the *max_matches*
+	 * which previously provided in rte_regex_dev_info_get().
+	 * The value 0 is allowed, in which case, value 1 used.
+	 * @see struct rte_regex_dev_info::max_matches
+	 */
+	uint16_t nb_queue_pairs;
+	/**< Number of RegEx queue pairs to configure on this device.
+	 * This value cannot exceed the *max_queue_pairs* which previously
+	 * provided in rte_regex_dev_info_get().
+	 * @see struct rte_regex_dev_info::max_queue_pairs
+	 */
+	uint32_t nb_rules_per_group;
+	/**< Number of rules per group to configure on this device.
+	 * This value cannot exceed the *max_rules_per_group*
+	 * which previously provided in rte_regex_dev_info_get().
+	 * The value 0 is allowed, in which case,
+	 * struct rte_regex_dev_info::max_rules_per_group used.
+	 * @see struct rte_regex_dev_info::max_rules_per_group
+	 */
+	uint16_t nb_groups;
+	/**< Number of groups to configure on this device.
+	 * This value cannot exceed the *max_groups*
+	 * which previously provided in rte_regex_dev_info_get().
+	 * @see struct rte_regex_dev_info::max_groups
+	 */
+	const char *rule_db;
+	/**< Import initial set of prebuilt rule database on this device.
+	 * The value NULL is allowed, in which case, the device will not
+	 * be configured prebuilt rule database. Application may use
+	 * rte_regex_rule_db_update() or rte_regex_rule_db_import() API
+	 * to update or import rule database after the
+	 * rte_regex_dev_configure().
+	 * @see rte_regex_rule_db_update(), rte_regex_rule_db_import()
+	 */
+	uint32_t rule_db_len;
+	/**< Length of *rule_db* buffer. */
+	uint32_t dev_cfg_flags;
+	/**< RegEx device configuration flags, See RTE_REGEX_DEV_CFG_*  */
+};
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice.
+ *
+ * Configure a RegEx device.
+ *
+ * This function must be invoked first before any other function in the
+ * API. This function can also be re-invoked when a device is in the
+ * stopped state.
+ *
+ * The caller may use rte_regex_dev_info_get() to get the capability of each
+ * resources available for this regex device.
+ *
+ * @param dev_id
+ *   The identifier of the device to configure.
+ * @param cfg
+ *   The RegEx device configuration structure.
+ *
+ * @return
+ *   - 0: Success, device configured. Otherwise negative errno is returned.
+ */
+__rte_experimental
+int
+rte_regex_dev_configure(uint8_t dev_id, const struct rte_regex_dev_config *cfg);
+
+/* Enumerates RegEx queue pair configuration flags */
+#define RTE_REGEX_QUEUE_PAIR_CFG_OOS_F (1ULL << 0)
+/**< Out of order scan, If not set, a scan must retire after previously issued
+ * in-order scans to this queue pair. If set, this scan can be retired as soon
+ * as device returns completion. Application should not set out of order scan
+ * flag if it needs to maintain the ingress order of scan request.
+ *
+ * @see struct rte_regex_qp_conf::qp_conf_flags, rte_regex_queue_pair_setup()
+ */
+
+struct rte_regex_ops;
+typedef void (*regexdev_stop_flush_t)(uint8_t dev_id, uint16_t qp_id,
+				      struct rte_regex_ops *op);
+/**< Callback function called during rte_regex_dev_stop(), invoked once per
+ * flushed RegEx op.
+ */
+
+/** RegEx queue pair configuration structure */
+struct rte_regex_qp_conf {
+	uint32_t qp_conf_flags;
+	/**< Queue pair config flags, See RTE_REGEX_QUEUE_PAIR_CFG_* */
+	uint16_t nb_desc;
+	/**< The number of descriptors to allocate for this queue pair. */
+	regexdev_stop_flush_t cb;
+	/**< Callback function called during rte_regex_dev_stop(), invoked
+	 * once per flushed regex op. Value NULL is allowed, in which case
+	 * callback will not be invoked. This function can be used to properly
+	 * dispose of outstanding regex ops from response queue,
+	 * for example ops containing memory pointers.
+	 * @see rte_regex_dev_stop()
+	 */
+};
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice.
+ *
+ * Allocate and set up a RegEx queue pair for a RegEx device.
+ *
+ * @param dev_id
+ *   The identifier of the device.
+ * @param queue_pair_id
+ *   The index of the RegEx queue pair to setup. The value must be in the range
+ *   [0, nb_queue_pairs - 1] previously supplied to rte_regex_dev_configure().
+ * @param qp_conf
+ *   The pointer to the configuration data to be used for the RegEx queue pair.
+ *   NULL value is allowed, in which case default configuration	used.
+ *
+ * @return
+ *   0 on success. Otherwise negative errno is returned.
+ */
+__rte_experimental
+int
+rte_regex_queue_pair_setup(uint8_t dev_id, uint16_t queue_pair_id,
+			   const struct rte_regex_qp_conf *qp_conf);
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice.
+ *
+ * Start a RegEx device.
+ *
+ * The device start step is the last one and consists of setting the RegEx
+ * queues to start accepting the pattern matching scan requests.
+ *
+ * On success, all basic functions exported by the API (RegEx enqueue,
+ * RegEx dequeue and so on) can be invoked.
+ *
+ * @param dev_id
+ *   RegEx device identifier.
+ *
+ * @return
+ *   0 on success. Otherwise negative errno is returned.
+ */
+__rte_experimental
+int
+rte_regex_dev_start(uint8_t dev_id);
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice.
+ *
+ * Stop a RegEx device.
+ *
+ * Stop a RegEx device. The device can be restarted with a call to
+ * rte_regex_dev_start().
+ *
+ * This function causes all queued response regex ops to be drained in the
+ * response queue. While draining ops out of the device,
+ * struct rte_regex_qp_conf::cb will be invoked for each ops.
+ *
+ * @param dev_id
+ *   RegEx device identifier.
+ *
+ * @see struct rte_regex_qp_conf::cb, rte_regex_queue_pair_setup()
+ */
+__rte_experimental
+void
+rte_regex_dev_stop(uint8_t dev_id);
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice.
+ *
+ * Close a RegEx device. The device cannot be restarted!
+ *
+ * @param dev_id
+ *   RegEx device identifier
+ *
+ * @return
+ *   0 on success. Otherwise negative errno is returned.
+ */
+__rte_experimental
+int
+rte_regex_dev_close(uint8_t dev_id);
+
+/* Device get/set attributes */
+
+/** Enumerates RegEx device attribute identifier */
+enum rte_regex_dev_attr_id {
+	RTE_REGEX_DEV_ATTR_SOCKET_ID,
+	/**< The NUMA socket id to which the device is connected or
+	 * a default of zero if the socket could not be determined.
+	 * datatype: *int*
+	 * operation: *get*
+	 */
+	RTE_REGEX_DEV_ATTR_MAX_MATCHES,
+	/**< Maximum number of matches per scan.
+	 * datatype: *uint8_t*
+	 * operation: *get* and *set*
+	 *
+	 * @see RTE_REGEX_OPS_RSP_MAX_MATCH_F
+	 */
+	RTE_REGEX_DEV_ATTR_MAX_SCAN_TIMEOUT,
+	/**< Upper bound scan time in ns.
+	 * datatype: *uint16_t*
+	 * operation: *get* and *set*
+	 *
+	 * @see RTE_REGEX_OPS_RSP_MAX_SCAN_TIMEOUT_F
+	 */
+	RTE_REGEX_DEV_ATTR_MAX_PREFIX,
+	/**< Maximum number of prefix detected per scan.
+	 * This would be useful for denial of service detection.
+	 * datatype: *uint16_t*
+	 * operation: *get* and *set*
+	 *
+	 * @see RTE_REGEX_OPS_RSP_MAX_PREFIX_F
+	 */
+};
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice.
+ *
+ * Get an attribute from a RegEx device.
+ *
+ * @param dev_id
+ *   RegEx device identifier.
+ * @param attr_id
+ *   The attribute ID to retrieve.
+ * @param attr_value
+ *   A pointer that will be filled in with the attribute
+ *   value if successful.
+ *
+ * @return
+ *   - 0: Successfully retrieved attribute value.
+ *   - -EINVAL: Invalid device or  *attr_id* provided, or *attr_value* is NULL.
+ *   - -ENOTSUP: if the device doesn't support specific *attr_id*.
+ */
+__rte_experimental
+int
+rte_regex_dev_attr_get(uint8_t dev_id, enum rte_regex_dev_attr_id attr_id,
+		       void *attr_value);
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice.
+ *
+ * Set an attribute to a RegEx device.
+ *
+ * @param dev_id
+ *   RegEx device identifier.
+ * @param attr_id
+ *   The attribute ID to retrieve.
+ * @param attr_value
+ *   Pointer that will be filled in with the attribute value
+ *   by the application.
+ *
+ * @return
+ *   - 0: Successfully applied the attribute value.
+ *   - -EINVAL: Invalid device or  *attr_id* provided, or *attr_value* is NULL.
+ *   - -ENOTSUP: if the device doesn't support specific *attr_id*.
+ */
+__rte_experimental
+int
+rte_regex_dev_attr_set(uint8_t dev_id, enum rte_regex_dev_attr_id attr_id,
+		       const void *attr_value);
+
+/* Rule related APIs */
+/** Enumerates RegEx rule operation. */
+enum rte_regex_rule_op {
+	RTE_REGEX_RULE_OP_ADD,
+	/**< Add RegEx rule to rule database. */
+	RTE_REGEX_RULE_OP_REMOVE
+	/**< Remove RegEx rule from rule database. */
+};
+
+/** Structure to hold a RegEx rule attributes. */
+struct rte_regex_rule {
+	enum rte_regex_rule_op op;
+	/**< OP type of the rule either a OP_ADD or OP_DELETE. */
+	uint16_t group_id;
+	/**< Group identifier to which the rule belongs to. */
+	uint32_t rule_id;
+	/**< Rule identifier which is returned on successful match. */
+	const char *pcre_rule;
+	/**< Buffer to hold the PCRE rule. */
+	uint16_t pcre_rule_len;
+	/**< Length of the PCRE rule. */
+	uint64_t rule_flags;
+	/* PCRE rule flags. Supported device specific PCRE rules enumerated
+	 * in struct rte_regex_dev_info::rule_flags. For successful rule
+	 * database update, application needs to provide only supported
+	 * rule flags.
+	 * @See RTE_REGEX_PCRE_RULE_*, struct rte_regex_dev_info::rule_flags
+	 */
+};
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice.
+ *
+ * Update the local rule set.
+ * This functions only modify the rule set in memory.
+ *
+ * @param dev_id.
+ *   RegEx device identifier.
+ * @param rules.
+ *   Points to an array of *nb_rules* objects of type *rte_regex_rule* structure
+ *   which contain the regex rules attributes to be updated in rule database.
+ * @param nb_rules.
+ *   The number of PCRE rules to update the rule database.
+ *
+ * @return
+ *   The number of regex rules actually updated on the regex device's rule
+ *   database. The return value can be less than the value of the *nb_rules*
+ *   parameter when the regex devices fails to update the rule database or
+ *   if invalid parameters are specified in a *rte_regex_rule*.
+ *   If the return value is less than *nb_rules*, the remaining PCRE rules
+ *   at the end of *rules* are not consumed and the caller has to take
+ *   care of them and rte_errno is set accordingly.
+ *   Possible errno values include:
+ *   - -EINVAL:  Invalid device ID or rules is NULL
+ *   - -ENOTSUP: The last processed rule is not supported on this device.
+ *   - -ENOSPC: No space available in rule database.
+ *
+ * @see rte_regex_rule_db_import(), rte_regex_rule_db_export(),
+ *   rte_regex_rule_db_compile()
+ */
+__rte_experimental
+int
+rte_regex_rule_db_update(uint8_t dev_id, const struct rte_regex_rule *rules,
+			 uint32_t nb_rules);
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice.
+ *
+ * Compile local rule set and burn the complied result to the
+ * RegEx deive.
+ *
+ * @param dev_id.
+ *   RegEx device identifier.
+ *
+ * @return
+ *   0 on success, otherwise negative errno.
+ *
+ * @see rte_regex_rule_db_import(), rte_regex_rule_db_export(,
+ *   rte_regex_rule_db_update()
+ */
+__rte_experimental
+int
+rte_regex_rule_db_compile(uint8_t dev_id);
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice.
+ *
+ * Import a prebuilt rule database from a buffer to a RegEx device.
+ *
+ * @param dev_id
+ *   RegEx device identifier.
+ * @param rule_db
+ *   Points to prebuilt rule database.
+ * @param rule_db_len
+ *   Length of the rule database.
+ *
+ * @return
+ *   - 0: Successfully updated the prebuilt rule database.
+ *   - -EINVAL:  Invalid device ID or rule_db is NULL
+ *   - -ENOTSUP: Rule database import is not supported on this device.
+ *   - -ENOSPC: No space available in rule database.
+ *
+ * @see rte_regex_rule_db_update(), rte_regex_rule_db_export()
+ */
+__rte_experimental
+int
+rte_regex_rule_db_import(uint8_t dev_id, const char *rule_db,
+			 uint32_t rule_db_len);
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice.
+ *
+ * Export the prebuilt rule database from a RegEx device to the buffer.
+ *
+ * @param dev_id
+ *   RegEx device identifier.
+ * @param[out] rule_db
+ *   Block of memory to insert the rule database. Must be at least size in
+ *   capacity. If set to NULL, function returns required capacity.
+ *
+ * @return
+ *   - 0: Successfully exported the prebuilt rule database.
+ *   - size: If rule_db set to NULL then required capacity for *rule_db*
+ *   - -EINVAL:  Invalid device ID
+ *   - -ENOTSUP: Rule database export is not supported on this device.
+ *
+ * @see rte_regex_rule_db_update(), rte_regex_rule_db_import()
+ */
+__rte_experimental
+int
+rte_regex_rule_db_export(uint8_t dev_id, char *rule_db);
+
+/* Extended statistics */
+/** Maximum name length for extended statistics counters */
+#define RTE_REGEX_DEV_XSTATS_NAME_SIZE 64
+
+/**
+ * A name-key lookup element for extended statistics.
+ *
+ * This structure is used to map between names and ID numbers
+ * for extended RegEx device statistics.
+ */
+struct rte_regex_dev_xstats_map {
+	uint16_t id;
+	/**< xstat identifier */
+	char name[RTE_REGEX_DEV_XSTATS_NAME_SIZE];
+	/**< xstat name */
+};
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice.
+ *
+ * Retrieve names of extended statistics of a regex device.
+ *
+ * @param dev_id
+ *   The identifier of the regex device.
+ * @param[out] xstats_map
+ *   Block of memory to insert id and names into. Must be at least size in
+ *   capacity. If set to NULL, function returns required capacity.
+ * @return
+ *   - Positive value on success:
+ *        -The return value is the number of entries filled in the stats map.
+ *        -If xstats_map set to NULL then required capacity for xstats_map.
+ *   - Negative value on error:
+ *      -ENODEV for invalid *dev_id*
+ *      -ENOTSUP if the device doesn't support this function.
+ */
+__rte_experimental
+int
+rte_regex_dev_xstats_names_get(uint8_t dev_id,
+			       struct rte_regex_dev_xstats_map *xstats_map);
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice.
+ *
+ * Retrieve extended statistics of an regex device.
+ *
+ * @param dev_id
+ *   The identifier of the device.
+ * @param ids
+ *   The id numbers of the stats to get. The ids can be got from the stat
+ *   position in the stat list from rte_regex_dev_xstats_names_get(), or
+ *   by using rte_regex_dev_xstats_by_name_get().
+ * @param values
+ *   The values for each stats request by ID.
+ * @param n
+ *   The number of stats requested.
+ * @return
+ *   - Positive value: number of stat entries filled into the values array
+ *   - Negative value on error:
+ *      -ENODEV for invalid *dev_id*
+ *      -ENOTSUP if the device doesn't support this function.
+ */
+__rte_experimental
+int
+rte_regex_dev_xstats_get(uint8_t dev_id, const uint16_t ids[],
+			 uint64_t values[], uint16_t n);
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice.
+ *
+ * Retrieve the value of a single stat by requesting it by name.
+ *
+ * @param dev_id
+ *   The identifier of the device.
+ * @param name
+ *   The stat name to retrieve.
+ * @param id
+ *   If non-NULL, the numerical id of the stat will be returned, so that further
+ *   requests for the stat can be got using rte_regex_dev_xstats_get, which will
+ *   be faster as it doesn't need to scan a list of names for the stat.
+ * @param[out] value.
+ *   Must be non-NULL, retrieved xstat value will be stored in this address.
+ *
+ * @return
+ *   - 0: Successfully retrieved xstat value.
+ *   - -EINVAL: invalid parameters
+ *   - -ENOTSUP: if not supported.
+ */
+__rte_experimental
+int
+rte_regex_dev_xstats_by_name_get(uint8_t dev_id, const char *name,
+				 uint16_t *id, uint64_t *value);
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice.
+ *
+ * Reset the values of the xstats of the selected component in the device.
+ *
+ * @param dev_id
+ *   The identifier of the device.
+ * @param ids
+ *   Selects specific statistics to be reset. When NULL, all statistics will be
+ *   reset. If non-NULL, must point to array of at least *nb_ids* size.
+ * @param nb_ids
+ *   The number of ids available from the *ids* array. Ignored when ids is NULL.
+ *
+ * @return
+ *   - 0: Successfully reset the statistics to zero.
+ *   - -EINVAL: invalid parameters.
+ *   - -ENOTSUP: if not supported.
+ */
+__rte_experimental
+int
+rte_regex_dev_xstats_reset(uint8_t dev_id, const uint16_t ids[],
+			   uint16_t nb_ids);
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice.
+ *
+ * Trigger the RegEx device self test.
+ *
+ * @param dev_id
+ *   The identifier of the device.
+ * @return
+ *   - 0: Selftest successful.
+ *   - -ENOTSUP if the device doesn't support selftest.
+ *   - other values < 0 on failure.
+ */
+__rte_experimental
+int
+rte_regex_dev_selftest(uint8_t dev_id);
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice.
+ *
+ * Dump internal information about *dev_id* to the FILE* provided in *f*.
+ *
+ * @param dev_id
+ *   The identifier of the device.
+ * @param f
+ *   A pointer to a file for output.
+ *
+ * @return
+ *   0 on success, negative errno on failure.
+ */
+__rte_experimental
+int
+rte_regex_dev_dump(uint8_t dev_id, FILE *f);
+
+/* Fast path APIs */
+
+/**
+ * The generic *rte_regex_match* structure to hold the RegEx match attributes.
+ * @see struct rte_regex_ops::matches
+ */
+struct rte_regex_match {
+	RTE_STD_C11
+	union {
+		uint64_t u64;
+		struct {
+			uint32_t rule_id:20;
+			/**< Rule identifier to which the pattern matched.
+			 * @see struct rte_regex_rule::rule_id
+			 */
+			uint32_t group_id:12;
+			/**< Group identifier of the rule which the pattern
+			 * matched. @see struct rte_regex_rule::group_id
+			 */
+			uint16_t offset;
+			/**< Starting Byte Position for matched rule. */
+			RTE_STD_C11
+			union {
+				uint16_t len;
+				/**< Length of match in bytes */
+				uint16_t end_offset;
+				/**< The end offset of the match. In case
+				 * MATCH_AS_START configuration is disabled.
+				 * @see RTE_REGEX_DEV_CFG_MATCH_AS_START
+				 */
+			};
+		};
+	};
+};
+
+/* Enumerates RegEx request flags. */
+#define RTE_REGEX_OPS_REQ_GROUP_ID1_VALID_F (1 << 0)
+/**< Set when struct rte_regex_rule::group_id1 valid */
+
+#define RTE_REGEX_OPS_REQ_GROUP_ID2_VALID_F (1 << 1)
+/**< Set when struct rte_regex_rule::group_id2 valid */
+
+#define RTE_REGEX_OPS_REQ_GROUP_ID3_VALID_F (1 << 2)
+/**< Set when struct rte_regex_rule::group_id3 valid */
+
+#define RTE_REGEX_OPS_REQ_STOP_ON_MATCH_F (1 << 4)
+/**< The RegEx engine will stop scanning and return the first match. */
+
+#define RTE_REGEX_OPS_REQ_MATCH_HIGH_PRIORITY_F (1 << 5)
+/**< In High Priority mode a maximum of one match will be returned per scan to
+ * reduce the post-processing required by the application. The match with the
+ * lowest Rule id, lowest start pointer and lowest match length will be
+ * returned.
+ *
+ * @see struct rte_regex_ops::nb_actual_matches
+ * @see struct rte_regex_ops::nb_matches
+ */
+
+
+/* Enumerates RegEx response flags. */
+#define RTE_REGEX_OPS_RSP_PMI_SOJ_F (1 << 0)
+/**< Indicates that the RegEx device has encountered a partial match at the
+ * start of scan in the given buffer.
+ *
+ * @see RTE_REGEX_DEV_CFG_CROSS_BUFFER_SCAN_F
+ */
+
+#define RTE_REGEX_OPS_RSP_PMI_EOJ_F (1 << 1)
+/**< Indicates that the RegEx device has encountered a partial match at the
+ * end of scan in the given buffer.
+ *
+ * @see RTE_REGEX_DEV_CFG_CROSS_BUFFER_SCAN_F
+ */
+
+#define RTE_REGEX_OPS_RSP_MAX_SCAN_TIMEOUT_F (1 << 2)
+/**< Indicates that the RegEx device has exceeded the max timeout while
+ * scanning the given buffer.
+ *
+ * @see RTE_REGEX_DEV_ATTR_MAX_SCAN_TIMEOUT
+ */
+
+#define RTE_REGEX_OPS_RSP_MAX_MATCH_F (1 << 3)
+/**< Indicates that the RegEx device has exceeded the max matches while
+ * scanning the given buffer.
+ *
+ * @see RTE_REGEX_DEV_ATTR_MAX_MATCHES
+ */
+
+#define RTE_REGEX_OPS_RSP_MAX_PREFIX_F (1 << 4)
+/**< Indicates that the RegEx device has reached the max allowed prefix length
+ * while scanning the given buffer.
+ *
+ * @see RTE_REGEX_DEV_ATTR_MAX_PREFIX
+ */
+
+/** Struct to hold scatter gather elements in ops. */
+struct rte_regex_iov {
+	RTE_STD_C11
+	union {
+		uint64_t u64;
+		/**<  Allow 8-byte reserved on 32-bit system */
+		void *buf_addr;
+		/**< Virtual address of the pattern to be matched. */
+	};
+	rte_iova_t buf_iova;
+	/**< IOVA address of the pattern to be matched. */
+	uint16_t buf_size; /**< The buf size. */
+};
+
+/**
+ * The generic *rte_regex_ops* structure to hold the RegEx attributes
+ * for enqueue and dequeue operation.
+ */
+struct rte_regex_ops {
+	/* W0 */
+	uint16_t req_flags;
+	/**< Request flags for the RegEx ops.
+	 * @see RTE_REGEX_OPS_REQ_*
+	 */
+	uint16_t rsp_flags;
+	/**< Response flags for the RegEx ops.
+	 * @see RTE_REGEX_OPS_RSP_*
+	 */
+	uint16_t nb_actual_matches;
+	/**< The total number of actual matches detected by the Regex device.*/
+	uint16_t nb_matches;
+	/**< The total number of matches returned by the RegEx device for this
+	 * scan. The size of *rte_regex_ops::matches* zero length array will be
+	 * this value.
+	 *
+	 * @see struct rte_regex_ops::matches, struct rte_regex_match
+	 */
+
+	/* W1 */
+	uint16_t num_of_bufs;
+	/**< The number of bufs that are part of this ops. The total size of
+	 * the length of all the buffer must be smaller then the max buffer
+	 * len.
+	 */
+	uint16_t resv1;
+	uint32_t resv2;
+
+	/* W2 */
+	struct rte_regex_iov *(*bufs)[];
+	/**< Holds a pointer to the buffers list.*/
+
+	/* W3 */
+	uint16_t group_id0;
+	/**< First group_id to match the rule against. Minimum one group id
+	 * must be provided by application.
+	 * When RTE_REGEX_OPS_REQ_GROUP_ID1_VALID_F set then group_id1
+	 * is valid, respectively similar flags for group_id2 and group_id3.
+	 * Upon the match, struct rte_regex_match::group_id shall be updated
+	 * with matching group ID by the device. Group ID scheme provides
+	 * rule isolation and effective pattern matching.
+	 */
+	uint16_t group_id1;
+	/**< Second group_id to match the rule against.
+	 *
+	 * @see RTE_REGEX_OPS_REQ_GROUP_ID1_VALID_F
+	 */
+	uint16_t group_id2;
+	/**< Third group_id to match the rule against.
+	 *
+	 * @see RTE_REGEX_OPS_REQ_GROUP_ID2_VALID_F
+	 */
+	uint16_t group_id3;
+	/**< Forth group_id to match the rule against.
+	 *
+	 * @see RTE_REGEX_OPS_REQ_GROUP_ID3_VALID_F
+	 */
+
+	/* W4 */
+	RTE_STD_C11
+	union {
+		uint64_t user_id;
+		/**< Application specific opaque value. An application may use
+		 * this field to hold application specific value to share
+		 * between dequeue and enqueue operation.
+		 * Implementation should not modify this field.
+		 */
+		void *user_ptr;
+		/**< Pointer representation of *user_id* */
+	};
+
+	/* W5 */
+	RTE_STD_C11
+	union {
+		uint64_t cross_buf_id;
+		/**< ID used by the RegEx device in order to handle cross
+		 * buffer detection.
+		 * This ID is given by the RegEx device on dequeue, and
+		 * the application must send it on the following enque.
+		 */
+		void *cross_buf_ptr;
+		/**< Pointer representation of *cross_buf_id* */
+	};
+
+	/* W6 */
+	struct rte_regex_match matches[];
+	/**< Zero length array to hold the match tuples.
+	 * The struct rte_regex_ops::nb_matches value holds the number of
+	 * elements in this array.
+	 *
+	 * @see struct rte_regex_ops::nb_matches
+	 */
+};
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice.
+ *
+ * Enqueue a burst of scan request on a RegEx device.
+ *
+ * The rte_regex_enqueue_burst() function is invoked to place
+ * regex operations on the queue *qp_id* of the device designated by
+ * its *dev_id*.
+ *
+ * The *nb_ops* parameter is the number of operations to process which are
+ * supplied in the *ops* array of *rte_regex_op* structures.
+ *
+ * The rte_regex_enqueue_burst() function returns the number of
+ * operations it actually enqueued for processing. A return value equal to
+ * *nb_ops* means that all packets have been enqueued.
+ *
+ * @param dev_id
+ *   The identifier of the device.
+ * @param qp_id
+ *   The index of the queue pair which packets are to be enqueued for
+ *   processing. The value must be in the range [0, nb_queue_pairs - 1]
+ *   previously supplied to rte_regex_dev_configure().
+ * @param ops
+ *   The address of an array of *nb_ops* pointers to *rte_regex_op* structures
+ *   which contain the regex operations to be processed.
+ * @param nb_ops
+ *   The number of operations to process.
+ *
+ * @return
+ *   The number of operations actually enqueued on the regex device. The return
+ *   value can be less than the value of the *nb_ops* parameter when the
+ *   regex devices queue is full or if invalid parameters are specified in
+ *   a *rte_regex_op*. If the return value is less than *nb_ops*, the remaining
+ *   ops at the end of *ops* are not consumed and the caller has to take care
+ *   of them.
+ */
+__rte_experimental
+uint16_t
+rte_regex_enqueue_burst(uint8_t dev_id, uint16_t qp_id,
+			struct rte_regex_ops **ops, uint16_t nb_ops);
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice.
+ *
+ * Dequeue a burst of scan response from a queue on the RegEx device.
+ * The dequeued operation are stored in *rte_regex_op* structures
+ * whose pointers are supplied in the *ops* array.
+ *
+ * The rte_regex_dequeue_burst() function returns the number of ops
+ * actually dequeued, which is the number of *rte_regex_op* data structures
+ * effectively supplied into the *ops* array.
+ *
+ * A return value equal to *nb_ops* indicates that the queue contained
+ * at least *nb_ops* operations, and this is likely to signify that other
+ * processed operations remain in the devices output queue. Applications
+ * implementing a "retrieve as many processed operations as possible" policy
+ * can check this specific case and keep invoking the
+ * rte_regex_dequeue_burst() function until a value less than
+ * *nb_ops* is returned.
+ *
+ * The rte_regex_dequeue_burst() function does not provide any error
+ * notification to avoid the corresponding overhead.
+ *
+ * @param dev_id
+ *   The RegEx device identifier
+ * @param qp_id
+ *   The index of the queue pair from which to retrieve processed packets.
+ *   The value must be in the range [0, nb_queue_pairs - 1] previously
+ *   supplied to rte_regex_dev_configure().
+ * @param ops
+ *   The address of an array of pointers to *rte_regex_op* structures that must
+ *   be large enough to store *nb_ops* pointers in it.
+ * @param nb_ops
+ *   The maximum number of operations to dequeue.
+ *
+ * @return
+ *   The number of operations actually dequeued, which is the number
+ *   of pointers to *rte_regex_op* structures effectively supplied to the
+ *   *ops* array. If the return value is less than *nb_ops*, the remaining
+ *   ops at the end of *ops* are not consumed and the caller has to take care
+ *   of them.
+ */
+__rte_experimental
+uint16_t
+rte_regex_dequeue_burst(uint8_t dev_id, uint16_t qp_id,
+			struct rte_regex_ops **ops, uint16_t nb_ops);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _RTE_REGEXDEV_H_ */
diff --git a/lib/librte_regexdev/rte_regexdev_version.map b/lib/librte_regexdev/rte_regexdev_version.map
new file mode 100644
index 0000000..aabae5e
--- /dev/null
+++ b/lib/librte_regexdev/rte_regexdev_version.map
@@ -0,0 +1,3 @@ 
+EXPERIMENTAL {
+	global:
+};