June 2017 - git-lng-odp - lists.linaro.org

[COMMIT] branch master updated. v1.15.0.0-10-g497c339f

by git＠git-us.linaro.org

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "". The branch, master has been updated via 497c339f94e35d5fadeff539cebfbf44d698dc7e (commit) via fdd3df3547c78d81d1244eac3fa65512e9023237 (commit) via e5b211403233dda67507f18fd91f9826022fd287 (commit) via d232db102fc77e6cb2fd9d62ba387c22f68c042a (commit) from e7e92031e08fe525e421ab30223b7398881ed42e (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- commit 497c339f94e35d5fadeff539cebfbf44d698dc7e Author: Yi He <yi.he(a)linaro.org> Date: Fri Jun 23 11:26:09 2017 +0800 travis: enable all tests when make distcheck Make distcheck with all test configurations to confirm test programs installation in odp distribution package. Signed-off-by: Yi He <yi.he(a)linaro.org> Reviewed-by: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/.travis.yml b/.travis.yml index 67a2df64..2cc8caf7 100644 --- a/.travis.yml +++ b/.travis.yml @@ -137,7 +137,7 @@ jobs: compiler: gcc script: - ./bootstrap - - ./configure + - ./configure --prefix=$HOME/odp-install --enable-test-cpp --enable-test-vald --enable-test-helper --enable-test-perf --enable-user-guides --enable-test-perf-proc --enable-test-example - sudo LD_LIBRARY_PATH="/usr/local/lib:$LD_LIBRARY_PATH" make distcheck - stage: test env: TEST=doxygen commit fdd3df3547c78d81d1244eac3fa65512e9023237 Author: Yi He <yi.he(a)linaro.org> Date: Fri Jun 23 11:21:50 2017 +0800 Makefile.am: update distcheck configuration Update AM_DISTCHECK_CONFIGURE_FLAGS to enable all tests. Signed-off-by: Yi He <yi.he(a)linaro.org> Reviewed-by: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/Makefile.am b/Makefile.am index daa49788..fff1276a 100644 --- a/Makefile.am +++ b/Makefile.am @@ -1,8 +1,11 @@ ACLOCAL_AMFLAGS=-I m4 AUTOMAKE_OPTIONS = foreign -AM_DISTCHECK_CONFIGURE_FLAGS = --enable-test-cpp \ - --enable-test-perf \ - --enable-test-vald \ +AM_DISTCHECK_CONFIGURE_FLAGS = --enable-test-cpp \ + --enable-test-example \ + --enable-test-helper \ + --enable-test-perf \ + --enable-test-perf-proc \ + --enable-test-vald \ --with-testdir #@with_platform@ works alone in subdir but not as part of a path??? commit e5b211403233dda67507f18fd91f9826022fd287 Author: Yi He <yi.he(a)linaro.org> Date: Mon Jun 19 17:13:50 2017 +0800 test: time: fix distcheck failure When configured odp with --enable-test-vald, make distcheck will fail by not being able to find the test script time.sh in the distribution. This patch installs time test_SCRIPTS in distribution build. Signed-off-by: Yi He <yi.he(a)linaro.org> Reviewed-by: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/test/common_plat/validation/api/time/Makefile.am b/test/common_plat/validation/api/time/Makefile.am index 82682dfc..ca9c0562 100644 --- a/test/common_plat/validation/api/time/Makefile.am +++ b/test/common_plat/validation/api/time/Makefile.am @@ -16,3 +16,4 @@ time_main_LDADD = libtesttime.la $(LIBCUNIT_COMMON) $(LIBODP) EXTRA_DIST = time_test.h $(TESTSCRIPTS) dist_check_SCRIPTS = $(TESTSCRIPTS) +test_SCRIPTS = $(dist_check_SCRIPTS) commit d232db102fc77e6cb2fd9d62ba387c22f68c042a Author: Yi He <yi.he(a)linaro.org> Date: Tue May 23 15:08:59 2017 +0800 test: traffic_mngr: fix distcheck failure When configured odp with --enable-test-vald, make distcheck will fail by not being able to find the traffic_mngr test binary and script in the distribution. This patch installs traffic_mngr test_PROGRAMS and test_SCRIPTS in distribution build. Signed-off-by: Yi He <yi.he(a)linaro.org> Reviewed-by: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/test/common_plat/validation/api/traffic_mngr/Makefile.am b/test/common_plat/validation/api/traffic_mngr/Makefile.am index a012c1b3..44d50a91 100644 --- a/test/common_plat/validation/api/traffic_mngr/Makefile.am +++ b/test/common_plat/validation/api/traffic_mngr/Makefile.am @@ -10,9 +10,10 @@ TESTS = $(TESTSCRIPTS) noinst_LTLIBRARIES = libtesttraffic_mngr.la libtesttraffic_mngr_la_SOURCES = traffic_mngr.c -bin_PROGRAMS = traffic_mngr_main$(EXEEXT) +test_PROGRAMS = traffic_mngr_main$(EXEEXT) dist_traffic_mngr_main_SOURCES = traffic_mngr_main.c traffic_mngr_main_LDADD = libtesttraffic_mngr.la -lm $(LIBCUNIT_COMMON) $(LIBODP) EXTRA_DIST = traffic_mngr.h $(TESTSCRIPTS) dist_check_SCRIPTS = $(TESTSCRIPTS) +test_SCRIPTS = $(dist_check_SCRIPTS) ----------------------------------------------------------------------- Summary of changes: .travis.yml | 2 +- Makefile.am | 9 ++++++--- test/common_plat/validation/api/time/Makefile.am | 1 + test/common_plat/validation/api/traffic_mngr/Makefile.am | 3 ++- 4 files changed, 10 insertions(+), 5 deletions(-) hooks/post-receive --

6 years, 10 months

1
0
0 0

[COMMIT] branch api-next updated. v1.15.0.0-371-ge3c2e06c

by git＠git-us.linaro.org

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "". The branch, api-next has been updated via e3c2e06c4420f41d7053e615ac1cc6c1dc315407 (commit) via e7e92031e08fe525e421ab30223b7398881ed42e (commit) from f9dcbcc3f45fcd6d825c087fc55c631d6b30fa8c (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- commit e3c2e06c4420f41d7053e615ac1cc6c1dc315407 Merge: f9dcbcc3 e7e92031 Author: Maxim Uvarov <maxim.uvarov(a)linaro.org> Date: Fri Jun 23 12:36:45 2017 +0300 Merge branch 'master' into api-next Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --cc platform/linux-generic/Makefile.am index e6ff0033,989a65b6..2293f7f0 --- a/platform/linux-generic/Makefile.am +++ b/platform/linux-generic/Makefile.am @@@ -249,19 -229,24 +261,29 @@@ __LIB__libodp_linux_la_SOURCES = odp_timer_wheel.c \ odp_traffic_mngr.c \ odp_version.c \ - odp_weak.c + odp_weak.c \ + drv_atomic.c \ + drv_barrier.c \ + drv_driver.c \ + drv_shm.c \ - drv_spinlock.c \ - arch/@ARCH_DIR@/odp_cpu_arch.c \ - arch/@ARCH_DIR@/odp_sysinfo_parse.c - - __LIB__libodp_linux_la_LIBADD = $(ATOMIC_LIBS) ++ drv_spinlock.c + if ARCH_IS_ARM + __LIB__libodp_linux_la_SOURCES += arch/arm/odp_cpu_arch.c \ + arch/arm/odp_sysinfo_parse.c + endif + if ARCH_IS_MIPS64 + __LIB__libodp_linux_la_SOURCES += arch/mips64/odp_cpu_arch.c \ + arch/mips64/odp_sysinfo_parse.c + endif + if ARCH_IS_POWERPC + __LIB__libodp_linux_la_SOURCES += arch/powerpc/odp_cpu_arch.c \ + arch/powerpc/odp_sysinfo_parse.c + endif if ARCH_IS_X86 - __LIB__libodp_linux_la_SOURCES += arch/@ARCH_DIR@/cpu_flags.c + __LIB__libodp_linux_la_SOURCES += arch/x86/cpu_flags.c \ + arch/x86/odp_cpu_arch.c \ + arch/x86/odp_sysinfo_parse.c endif if HAVE_PCAP ----------------------------------------------------------------------- Summary of changes: configure.ac | 3 +++ platform/linux-generic/Makefile.am | 40 ++++++++++++++++++++++++++++++-------- 2 files changed, 35 insertions(+), 8 deletions(-) hooks/post-receive --

6 years, 10 months

1
0
0 0

[COMMIT] branch master updated. v1.15.0.0-6-ge7e92031

by git＠git-us.linaro.org

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "". The branch, master has been updated via e7e92031e08fe525e421ab30223b7398881ed42e (commit) from 09a7800c4c4a093fb962e362952f9cf562d2fc98 (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- ----------------------------------------------------------------------- Summary of changes: configure.ac | 3 +++ platform/linux-generic/Makefile.am | 40 ++++++++++++++++++++++++++++++-------- 2 files changed, 35 insertions(+), 8 deletions(-) hooks/post-receive --

6 years, 10 months

1
0
0 0

[COMMIT] branch api-next updated. v1.15.0.0-369-gf9dcbcc3

by git＠git-us.linaro.org

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "". The branch, api-next has been updated via f9dcbcc3f45fcd6d825c087fc55c631d6b30fa8c (commit) via 845743bd80da73f2a296f24090dfe058d3be7dec (commit) via bbd08a6c0c31c12275089c3134c9e7e58eb0fafa (commit) via 0468dc01b3ae1517d57cf960b511ab066a9d5f05 (commit) via 274a0b016d8c2eed9f065186ad5458c5edb94dad (commit) via 0ce5d628d817003ca0f96beddece3f618f4593bb (commit) via 65a46f66bfcb6b2034d4b0563facc36cbac1e286 (commit) via 441fb964f22b2baaa97641e6f8dda88d32d153f6 (commit) via d5ff31f14e87c815d3a0e0461206aa2687152dd5 (commit) via 09a7800c4c4a093fb962e362952f9cf562d2fc98 (commit) via ddf2ac38b1288283f7b3d7c9e16e36f392685866 (commit) via f4337956ee3be1e783e2389eb5a5ceecb4b0b0ca (commit) via cd1f187d33d65cb54781b5a3c0864229b19c11c3 (commit) via 0d24adacfe17c8e0e1348f19cd8b75b64cb13ccf (commit) from 4bcd15e2bfb839ac0ef6419aa9e7b720958638f7 (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- commit f9dcbcc3f45fcd6d825c087fc55c631d6b30fa8c Author: Bill Fischofer <bill.fischofer(a)linaro.org> Date: Tue Jun 13 07:55:15 2017 -0500 doc: userguide: add odp_init_global() documentation for unused features Update User Guide startup section to include current parameters for odp_init_global() and odp_init_local() as well as optimization hints for unused features. Signed-off-by: Bill Fischofer <bill.fischofer(a)linaro.org> Reviewed-by: Petri Savolainen <petri.savolainen(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/doc/users-guide/users-guide.adoc b/doc/users-guide/users-guide.adoc index ead8da5e..d78e0777 100755 --- a/doc/users-guide/users-guide.adoc +++ b/doc/users-guide/users-guide.adoc @@ -530,20 +530,44 @@ calling the terminate functions should only be done when the application is sure it has closed the ingress and subsequently drained all queues, etc. === Startup -The first API that must be called by an ODP application is 'odp_init_global()'. +The first API that must be called by an ODP application is `odp_init_global()`: +[source,c] +----- +int odp_init_global(odp_instance_t *instance, + const odp_init_t *param, + const odp_platform_init_t *platform_param); +----- This takes two pointers. The first, `odp_init_t`, contains ODP initialization data that is platform independent and portable, while the second, `odp_platform_init_t`, is passed unparsed to the implementation to be used for platform specific data that is not yet, or may never be -suitable for the ODP API. +suitable for the ODP API. Each of these parameters is optional and may be +specified as NULL to accept the implementation-defined default initialization +values. -Calling odp_init_global() establishes the ODP API framework and MUST be +Calling `odp_init_global()` establishes the ODP API framework and MUST be called before any other ODP API may be called. Note that it is only called -once per application. Following global initialization, each thread in turn +once per application. A successful call to `odp_init_global()` returns rc = 0 +and sets the `instance` variable supplied as input to the call to an handle +representing this unique ODP instance. + +The `odp_init_t` parameter is used to specify various customizations to the +ODP environment being established by this call. For example, the caller can +specify the maximum number of worker threads it will use, the thread masks +associated with these threads, as well as whether the default logging or +abort functions are to be overridden with an application-supplied handler. + +The application may also provide optimization hints to the ODP implementation +if it knows that it will never use specific ODP feature sets, such as the +packet classifier or traffic manager. Implementations may use such hints to +provide optimized behavior to applications that are known not to need these +features. + +Following global initialization, each thread in turn calls 'odp_init_local()'. This establishes the local ODP thread context for that thread and MUST be called before other ODP APIs may be -called by that thread. The sole argument to this call is the _thread type_, -which is either `ODP_THREAD_WORKER` or `ODP_THREAD_CONTROL`. +called by that thread. The sole argument to this call is the `instance` +variable returned by `odp_init_global()`. === Shutdown Shutdown is the logical reverse of the initialization procedure, with commit 845743bd80da73f2a296f24090dfe058d3be7dec Author: Bill Fischofer <bill.fischofer(a)linaro.org> Date: Tue Jun 13 07:55:14 2017 -0500 validation: init: use odp_init_param_init() in init tests Provide test coverage for odp_init_param_init() API. Signed-off-by: Bill Fischofer <bill.fischofer(a)linaro.org> Reviewed-by: Petri Savolainen <petri.savolainen(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/test/common_plat/validation/api/init/init.c b/test/common_plat/validation/api/init/init.c index 61055fad..6f9556d7 100644 --- a/test/common_plat/validation/api/init/init.c +++ b/test/common_plat/validation/api/init/init.c @@ -24,10 +24,10 @@ static int odp_init_log(odp_log_level_t level, const char *fmt, ...); void init_test_odp_init_global_replace_abort(void) { int status; - struct odp_init_t init_data; + odp_init_t init_data; odp_instance_t instance; - memset(&init_data, 0, sizeof(init_data)); + odp_init_param_init(&init_data); init_data.abort_fn = &odp_init_abort; status = odp_init_global(&instance, &init_data, NULL); @@ -77,10 +77,10 @@ int init_main_abort(int argc, char *argv[]) void init_test_odp_init_global_replace_log(void) { int status; - struct odp_init_t init_data; + odp_init_t init_data; odp_instance_t instance; - memset(&init_data, 0, sizeof(init_data)); + odp_init_param_init(&init_data); init_data.log_fn = &odp_init_log; replacement_logging_used = 0; commit bbd08a6c0c31c12275089c3134c9e7e58eb0fafa Author: Bill Fischofer <bill.fischofer(a)linaro.org> Date: Tue Jun 13 07:55:13 2017 -0500 linux-generic: init: implement odp_init_param_init() Signed-off-by: Bill Fischofer <bill.fischofer(a)linaro.org> Reviewed-by: Petri Savolainen <petri.savolainen(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/platform/linux-generic/odp_init.c b/platform/linux-generic/odp_init.c index 31bd7c73..62a1fbc2 100644 --- a/platform/linux-generic/odp_init.c +++ b/platform/linux-generic/odp_init.c @@ -147,6 +147,11 @@ static int read_configfile(void) return 0; } +void odp_init_param_init(odp_init_t *param) +{ + memset(param, 0, sizeof(odp_init_t)); +} + int odp_init_global(odp_instance_t *instance, const odp_init_t *params, const odp_platform_init_t *platform_params ODP_UNUSED) commit 0468dc01b3ae1517d57cf960b511ab066a9d5f05 Author: Bill Fischofer <bill.fischofer(a)linaro.org> Date: Tue Jun 13 07:55:12 2017 -0500 api: init: add support for unused features Add the not_used field to odp_init_t to permit applications to specify that they will not use various ODP features. This may allow implementations to provide optimized behavior. Also add the odp_init_param_init() API to initialize odp_init_t to default values. Signed-off-by: Bill Fischofer <bill.fischofer(a)linaro.org> Reviewed-by: Petri Savolainen <petri.savolainen(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/include/odp/api/spec/init.h b/include/odp/api/spec/init.h index 154cdf8f..e8ec4113 100644 --- a/include/odp/api/spec/init.h +++ b/include/odp/api/spec/init.h @@ -29,6 +29,7 @@ extern "C" { #include <odp/api/std_types.h> #include <odp/api/hints.h> +#include <odp/api/feature.h> #include <odp/api/thread.h> #include <odp/api/cpumask.h> @@ -153,9 +154,23 @@ typedef struct odp_init_t { odp_log_func_t log_fn; /** Replacement for the default abort fn */ odp_abort_func_t abort_fn; + /** Unused features. These are hints to the ODP implementation that + * the application will not use any APIs associated with these + * features. Implementations may use this information to provide + * optimized behavior. Results are undefined if applications assert + * that a feature will not be used and it is used anyway. + */ + odp_feature_t not_used; } odp_init_t; /** + * Initialize the odp_init_t to default values for all fields + * + * @param[out] param Address of the odp_init_t to be initialized + */ +void odp_init_param_init(odp_init_t *param); + +/** * @typedef odp_platform_init_t * ODP platform initialization data * commit 274a0b016d8c2eed9f065186ad5458c5edb94dad Author: Bill Fischofer <bill.fischofer(a)linaro.org> Date: Tue Jun 13 07:55:11 2017 -0500 api: feature: add odp feature bits Add new odp_feature_t bits that permit other APIs/components to refer to various ODP features. Signed-off-by: Bill Fischofer <bill.fischofer(a)linaro.org> Reviewed-by: Petri Savolainen <petri.savolainen(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/include/odp/api/spec/feature.h b/include/odp/api/spec/feature.h new file mode 100644 index 00000000..0cfc141d --- /dev/null +++ b/include/odp/api/spec/feature.h @@ -0,0 +1,69 @@ +/* Copyright (c) 2017, Linaro Limited + * All rights reserved. + * + * SPDX-License-Identifier: BSD-3-Clause + */ + +/** + * @file + * + * ODP features. + * Define various ODP feature sets that can be referenced by other + * components. + */ + +#ifndef ODP_API_FEATURE_H_ +#define ODP_API_FEATURE_H_ +#include <odp/visibility_begin.h> + +#ifdef __cplusplus +extern "C" { +#endif + +#include <odp/api/std_types.h> + +/** @defgroup odp_features ODP_FEATURE + * ODP feature definitions + * @{ + */ + +/** Definition of ODP features */ +typedef union odp_feature_t { + /** All features */ + uint32_t all_feat; + + /** Individual feature bits */ + struct { + /** Classifier APIs, e.g., odp_cls_xxx(), odp_cos_xxx() */ + uint32_t cls:1; + + /** Crypto APIs, e.g., odp_crypto_xxx() */ + uint32_t crypto:1; + + /** IPsec APIs, e.g., odp_ipsec_xxx() */ + uint32_t ipsec:1; + + /** Scheduler APIs, e.g., odp_schedule_xxx() */ + uint32_t schedule:1; + + /** Time APIs are, e.g., odp_time_xxx() */ + uint32_t time:1; + + /** Timer APIs, e.g., odp_timer_xxx(), odp_timeout_xxx() */ + uint32_t timer:1; + + /** Traffic Manager APIs, e.g., odp_tm_xxx() */ + uint32_t tm:1; + } feat; +} odp_feature_t; + +/** + * @} + */ + +#ifdef __cplusplus +} +#endif + +#include <odp/visibility_end.h> +#endif diff --git a/include/odp_api.h b/include/odp_api.h index 8146e024..86232ee1 100644 --- a/include/odp_api.h +++ b/include/odp_api.h @@ -32,6 +32,7 @@ extern "C" { #include <odp/api/barrier.h> #include <odp/api/spinlock.h> #include <odp/api/atomic.h> +#include <odp/api/feature.h> #include <odp/api/init.h> #include <odp/api/system_info.h> #include <odp/api/thread.h> diff --git a/platform/Makefile.inc b/platform/Makefile.inc index 5343bfe1..24f06914 100644 --- a/platform/Makefile.inc +++ b/platform/Makefile.inc @@ -32,6 +32,7 @@ odpapispecinclude_HEADERS = \ $(top_srcdir)/include/odp/api/spec/deprecated.h \ $(top_srcdir)/include/odp/api/spec/errno.h \ $(top_srcdir)/include/odp/api/spec/event.h \ + $(top_srcdir)/include/odp/api/spec/feature.h \ $(top_srcdir)/include/odp/api/spec/support.h \ $(top_srcdir)/include/odp/api/spec/hash.h \ $(top_srcdir)/include/odp/api/spec/hints.h \ diff --git a/platform/linux-generic/Makefile.am b/platform/linux-generic/Makefile.am index ecb581c1..e6ff0033 100644 --- a/platform/linux-generic/Makefile.am +++ b/platform/linux-generic/Makefile.am @@ -38,7 +38,7 @@ odpapiinclude_HEADERS = \ $(srcdir)/include/odp/api/deprecated.h \ $(srcdir)/include/odp/api/errno.h \ $(srcdir)/include/odp/api/event.h \ - $(srcdir)/include/odp/api/support.h \ + $(srcdir)/include/odp/api/feature.h \ $(srcdir)/include/odp/api/hash.h \ $(srcdir)/include/odp/api/hints.h \ $(srcdir)/include/odp/api/init.h \ @@ -59,6 +59,7 @@ odpapiinclude_HEADERS = \ $(srcdir)/include/odp/api/spinlock_recursive.h \ $(srcdir)/include/odp/api/std_clib.h \ $(srcdir)/include/odp/api/std_types.h \ + $(srcdir)/include/odp/api/support.h \ $(srcdir)/include/odp/api/sync.h \ $(srcdir)/include/odp/api/system_info.h \ $(srcdir)/include/odp/api/thread.h \ diff --git a/platform/linux-generic/include/odp/api/feature.h b/platform/linux-generic/include/odp/api/feature.h new file mode 100644 index 00000000..55a86a83 --- /dev/null +++ b/platform/linux-generic/include/odp/api/feature.h @@ -0,0 +1,34 @@ +/* Copyright (c) 2017, Linaro Limited + * All rights reserved. + * + * SPDX-License-Identifier: BSD-3-Clause + */ + +/** + * @file + * + * ODP features. + */ + +#ifndef ODP_PLAT_FEATURE_H_ +#define ODP_PLAT_FEATURE_H_ + +#ifdef __cplusplus +extern "C" { +#endif + +/** @ingroup odp_feature + * @{ + */ + +/** + * @} + */ + +#include <odp/api/spec/feature.h> + +#ifdef __cplusplus +} +#endif + +#endif commit 0ce5d628d817003ca0f96beddece3f618f4593bb Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Fri Jun 16 13:49:40 2017 +0300 api: ipsec: disable event is the last event Disable event is guaranteed to be the last event for the SA, so that application can use it for SA destroy synchronization. Signed-off-by: Petri Savolainen <petri.savolainen(a)linaro.org> Reviewed-by: Bill Fischofer <bill.fischofer(a)linaro.org> Reviewed-by: Balasubramanian Manoharan <bala.manoharan(a)linaro.org> Reviewed-by: Nikhil Agarwal <Nikhil.agarwal(a)linaro.org> Reviewed-by: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/include/odp/api/spec/ipsec.h b/include/odp/api/spec/ipsec.h index 372ef462..e602e4b8 100644 --- a/include/odp/api/spec/ipsec.h +++ b/include/odp/api/spec/ipsec.h @@ -818,7 +818,10 @@ odp_ipsec_sa_t odp_ipsec_sa_create(const odp_ipsec_sa_param_t *param); * * When in synchronous operation mode, the call will return when it's possible * to destroy the SA. In asynchronous mode, the same is indicated by an - * ODP_EVENT_IPSEC_STATUS event sent to the queue specified for the SA. + * ODP_EVENT_IPSEC_STATUS event sent to the queue specified for the SA. The + * status event is guaranteed to be the last event for the SA, i.e. all + * in-progress operations have completed and resulting events (including status + * events) have been enqueued before it. * * @param sa IPSEC SA to be disabled * commit 65a46f66bfcb6b2034d4b0563facc36cbac1e286 Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Fri Jun 16 13:49:39 2017 +0300 api: ipsec: change IPSEC result to packet Input and output of IPSEC operations are packets. Parameter and result structures are cleaner when packet arrays are direct parameters to functions. Also API is more flexible for application and API pipelining when output is packets with additional metadata. Application or API pipeline stages which do not care about IPSEC results may work on basic packet metadata. IPSEC result event type changes from ODP_EVENT_IPSEC_RESULT to ODP_EVENT_PACKET. Event subtype (ODP_EVENT_PACKET_IPSEC) can be used to identify packets with IPSEC metadata. Signed-off-by: Petri Savolainen <petri.savolainen(a)linaro.org> Reviewed-by: Bill Fischofer <bill.fischofer(a)linaro.org> Reviewed-by: Balasubramanian Manoharan <bala.manoharan(a)linaro.org> Reviewed-by: Nikhil Agarwal <Nikhil.agarwal(a)linaro.org> Reviewed-by: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/include/odp/api/spec/ipsec.h b/include/odp/api/spec/ipsec.h index 65f0b066..372ef462 100644 --- a/include/odp/api/spec/ipsec.h +++ b/include/odp/api/spec/ipsec.h @@ -552,16 +552,18 @@ typedef enum odp_ipsec_frag_mode_t { /** * Packet lookup mode + * + * Lookup mode controls how an SA participates in SA lookup offload. + * Inbound operations perform SA lookup if application does not provide a SA as + * a parameter. In inline mode, a lookup miss directs the packet back to normal + * packet input interface processing. SA lookup failure status (error.sa_lookup) + * is reported through odp_ipsec_packet_result_t. */ typedef enum odp_ipsec_lookup_mode_t { - /** Inbound SA lookup is disabled. */ + /** Inbound SA lookup is disabled for the SA. */ ODP_IPSEC_LOOKUP_DISABLED = 0, - /** Inbound SA lookup is enabled. Lookup matches only SPI value. - * In inline mode, a lookup miss directs the packet back to normal - * packet input interface processing. In other modes, the SA lookup - * failure status (error.sa_lookup) is reported through - * odp_ipsec_packet_result_t. */ + /** Inbound SA lookup is enabled. Lookup matches only SPI value. */ ODP_IPSEC_LOOKUP_SPI, /** Inbound SA lookup is enabled. Lookup matches both SPI value and @@ -572,13 +574,13 @@ typedef enum odp_ipsec_lookup_mode_t { } odp_ipsec_lookup_mode_t; /** - * Result event pipeline configuration + * IPSEC pipeline configuration */ typedef enum odp_ipsec_pipeline_t { - /** Do not pipeline */ + /** Do not pipeline. Send all resulting events to the application. */ ODP_IPSEC_PIPELINE_NONE = 0, - /** Send IPSEC result events to the classifier. + /** Send resulting packets to the classifier * * IPSEC capability 'pipeline_cls' determines if pipelined * classification is supported. */ @@ -659,9 +661,9 @@ typedef struct odp_ipsec_sa_param_t { */ uint32_t mtu; - /** Select pipelined destination for IPSEC result events + /** Select pipelined destination for resulting events * - * Asynchronous and inline modes generate result events. Select where + * Asynchronous and inline modes generate events. Select where * those events are sent. Inbound SAs may choose to use pipelined * classification. The default value is ODP_IPSEC_PIPELINE_NONE. */ @@ -670,18 +672,19 @@ typedef struct odp_ipsec_sa_param_t { /** Destination queue for IPSEC events * * Operations in asynchronous or inline mode enqueue resulting events - * into this queue. + * into this queue. The default queue ('default_queue') is used when + * SA is not known. */ odp_queue_t dest_queue; - /** Classifier destination CoS for IPSEC result events + /** Classifier destination CoS for resulting packets * - * Result events for successfully decapsulated packets are sent to - * classification through this CoS. Other result events are sent to - * 'dest_queue'. This field is considered only when 'pipeline' is + * Successfully decapsulated packets are sent to classification + * through this CoS. Other resulting events are sent to 'dest_queue'. + * This field is considered only when 'pipeline' is * ODP_IPSEC_PIPELINE_CLS. The CoS must not be shared between any pktio * interface default CoS. The maximum number of different CoS supported - * is defined by IPsec capability max_cls_cos. + * is defined by IPSEC capability max_cls_cos. */ odp_cos_t dest_cos; @@ -850,17 +853,6 @@ int odp_ipsec_sa_destroy(odp_ipsec_sa_t sa); */ uint64_t odp_ipsec_sa_to_u64(odp_ipsec_sa_t sa); -/** - * IPSEC operation level options - * - * These may be used to override some SA level options - */ -typedef struct odp_ipsec_op_opt_t { - /** Fragmentation mode */ - odp_ipsec_frag_mode_t mode; - -} odp_ipsec_op_opt_t; - /** IPSEC operation status has no errors */ #define ODP_IPSEC_OK 0 @@ -870,7 +862,8 @@ typedef struct odp_ipsec_op_status_t { union { /** Error flags */ struct { - /** Protocol error. Not a valid ESP or AH packet. */ + /** Protocol error. Not a valid ESP or AH packet, + * packet data length error, etc. */ uint32_t proto : 1; /** SA lookup failed */ @@ -934,41 +927,70 @@ typedef struct odp_ipsec_op_status_t { } odp_ipsec_op_status_t; /** - * IPSEC operation input parameters + * IPSEC outbound operation options + * + * These may be used to override some SA level options */ -typedef struct odp_ipsec_op_param_t { - /** Number of packets to be processed */ - int num_pkt; +typedef struct odp_ipsec_out_opt_t { + /** Fragmentation mode */ + odp_ipsec_frag_mode_t mode; + +} odp_ipsec_out_opt_t; +/** + * IPSEC outbound operation parameters + */ +typedef struct odp_ipsec_out_param_t { /** Number of SAs * + * Outbound IPSEC operation needs SA from application. Use either + * single SA for all packets, or a SA per packet. + * * Valid values are: - * * 0: No SAs (default) - * * 1: Single SA for all packets - * * num_pkt: SA per packet + * - 1: Single SA for all packets + * - N: A SA per packet. N must match the number of packets. */ int num_sa; - /** Number of operation options + /** Number of outbound operation options * * Valid values are: - * * 0: No options (default) - * * 1: Single option for all packets - * * num_pkt: An option per packet + * - 0: No options + * - 1: Single option for all packets + * - N: An option per packet. N must match the number of packets. */ int num_opt; - /** Pointer to an array of packets + /** Pointer to an array of IPSEC SAs */ + odp_ipsec_sa_t *sa; + + /** Pointer to an array of outbound operation options + * + * May be NULL when num_opt is zero. + */ + odp_ipsec_out_opt_t *opt; + +} odp_ipsec_out_param_t; + +/** + * IPSEC inbound operation parameters + */ +typedef struct odp_ipsec_in_param_t { + /** Number of SAs * - * Each packet must have a valid value for these metadata: - * * L3 offset: Offset to the first byte of the (outmost) IP header - * * L4 offset: For inbound direction, when udp_encap is enabled - - * offset to the first byte of the encapsulating UDP - * header + * Inbound IPSEC operation processes a packet using the SA provided by + * the application. If the application does not provide an SA, the + * operation searches for the SA by matching the input packet with all + * inbound SAs according to the lookup mode (odp_ipsec_lookup_mode_t) + * configured in each SA. When passing SAs, use either single SA for + * all packets, or a SA per packet. * - * @see odp_packet_l3_offset(), odp_packet_l4_offset() + * Valid values are: + * - 0: No SAs. SA lookup is done for all packets. + * - 1: Single SA for all packets + * - N: A SA per packet. N must match the number of packets. */ - odp_packet_t *pkt; + int num_sa; /** Pointer to an array of IPSEC SAs * @@ -976,18 +998,12 @@ typedef struct odp_ipsec_op_param_t { */ odp_ipsec_sa_t *sa; - /** Pointer to an array of operation options - * - * May be NULL when num_opt is zero. - */ - odp_ipsec_op_opt_t *opt; - -} odp_ipsec_op_param_t; +} odp_ipsec_in_param_t; /** * Outbound inline IPSEC operation parameters */ -typedef struct odp_ipsec_inline_op_param_t { +typedef struct odp_ipsec_out_inline_param_t { /** Packet output interface for inline output operation * * Outbound inline IPSEC operation uses this packet IO interface to @@ -1011,7 +1027,7 @@ typedef struct odp_ipsec_inline_op_param_t { uint32_t len; } outer_hdr; -} odp_ipsec_inline_op_param_t; +} odp_ipsec_out_inline_param_t; /** * IPSEC operation result for a packet @@ -1020,16 +1036,6 @@ typedef struct odp_ipsec_packet_result_t { /** IPSEC operation status */ odp_ipsec_op_status_t status; - /** Number of output packets created from the corresponding input packet - * - * Without fragmentation offload this is always one. However, if the - * input packet was fragmented during the operation this is larger than - * one for the first returned fragment and zero for the rest of the - * fragments. All the fragments (of the same source packet) are stored - * consecutively in the 'pkt' array. - */ - int num_out; - /** IPSEC SA that was used to create the packet * * Operation updates this SA handle value, when SA look up is performed @@ -1040,7 +1046,8 @@ typedef struct odp_ipsec_packet_result_t { odp_ipsec_sa_t sa; /** Packet outer header status before inbound inline processing. - * This is valid only when status.flag.inline_mode is set. + * This is valid only when outer headers are retained + * (see odp_ipsec_inbound_config_t) and status.flag.inline_mode is set. */ struct { /** Points to the first byte of retained outer headers. These @@ -1048,7 +1055,7 @@ typedef struct odp_ipsec_packet_result_t { * implementation specific memory space. Since the memory space * may overlap with e.g. packet head/tailroom, the content * becomes invalid if packet data storage is modified in - * anyway. The memory space may not be sharable to other + * any way. The memory space may not be shareable to other * threads. */ uint8_t *ptr; @@ -1059,51 +1066,6 @@ typedef struct odp_ipsec_packet_result_t { } odp_ipsec_packet_result_t; /** - * IPSEC operation results - */ -typedef struct odp_ipsec_op_result_t { - /** Number of packets - * - * Application sets this to the maximum number of packets the operation - * may output (number of elements in 'pkt' and 'res' arrays). - * The operation updates it with the actual number of packets - * outputted. - */ - int num_pkt; - - /** Pointer to an array of packets - * - * Operation outputs packets into this array. The array must have - * at least 'num_pkt' elements. - * - * Each successfully transformed packet has a valid value for these - * metadata regardless of the inner packet parse configuration. - * (odp_ipsec_inbound_config_t): - * * L3 offset: Offset to the first byte of the (outmost) IP header - * * pktio: For inbound inline IPSEC processed packets, original - * packet input interface - * - * Other metadata for parse results and error checks depend on - * configuration (selected parse and error check levels). - */ - odp_packet_t *pkt; - - /** Pointer to an array of per packet operation results - * - * Operation outputs results for each outputted packet into this array. - * The array must have at least 'num_pkt' elements. The results include - * operation status and packet form information for each outputted - * packet. - * - * For example, some packets may not have been transformed due to - * an error, but the original packet is returned with appropriate - * packet result information instead. - */ - odp_ipsec_packet_result_t *res; - -} odp_ipsec_op_result_t; - -/** * IPSEC status ID */ typedef enum odp_ipsec_status_id_t { @@ -1136,20 +1098,33 @@ typedef struct odp_ipsec_status_t { * * This operation does inbound IPSEC processing in synchronous mode * (ODP_IPSEC_OP_MODE_SYNC). A successful operation returns the number of - * packets consumed and outputs a new packet handle as well as an operation - * result for each outputted packet. The operation does not modify packets that - * it does not consume. It cannot consume all input packets if 'output.num_pkt' - * is smaller than 'input.num_pkt'. + * packets consumed and outputs a new packet handle for each outputted packet. + * Outputted packets contain IPSEC result metadata (odp_ipsec_packet_result_t), + * which should be checked for transformation errors, etc. Outputted packets + * with error status have not been transformed but the original packet is + * returned. The operation does not modify packets that it does not consume. + * It cannot consume all input packets if 'num_out' is smaller than 'num_in'. * * Packet context pointer and user area content are copied from input to output * packets. Output packets are allocated from the same pool(s) as input packets. * - * When 'input.num_sa' is zero, this operation performs SA look up for each + * When 'param.num_sa' is zero, this operation performs SA look up for each * packet. Otherwise, application must provide the SA(s) as part of operation - * input parameters (odp_ipsec_op_param_t). The operation outputs used SA(s) as - * part of per packet operation results (odp_ipsec_packet_result_t), or an error + * input parameters (odp_ipsec_in_param_t). The operation outputs used SA(s) as + * part of per packet results (odp_ipsec_packet_result_t), or an error * status if a SA was not found. * + * Each input packet must have a valid value for these metadata (other metadata + * is ignored): + * - L3 offset: Offset to the first byte of the (outmost) IP header + * - L4 offset: When udp_encap is enabled, offset to the first byte of the + * encapsulating UDP header + * + * Additionally, implementation checks input IP packet length (odp_packet_len() + * minus odp_packet_l3_offset()) against protocol headers and reports an error + * (status.error.proto) if packet data length is less than protocol headers + * indicate. + * * Packets are processed in the input order. Packet order is maintained from * input 'pkt' array to output 'pkt' array. Packet order is not guaranteed * between calling threads. @@ -1162,35 +1137,61 @@ typedef struct odp_ipsec_status_t { * restored. The amount and content of packet data before the IP header is * undefined. * - * @param input Operation input parameters - * @param[out] output Operation results - * - * @return Number of input packets consumed (0 ... input.num_pkt) + * Each successfully transformed packet has a valid value for these metadata + * regardless of the inner packet parse configuration + * (odp_ipsec_inbound_config_t): + * - L3 offset: Offset to the first byte of the (outmost) IP header + * - pktio: For inline IPSEC processed packets, original packet input + * interface + * + * Other metadata for parse results and error checks depend on configuration + * (selected parse and error check levels). + * + * @param pkt_in Packets to be processed + * @param num_in Number of packets to be processed + * @param[out] pkt_out Packet handle array for resulting packets + * @param[in, out] num_out Number of resulting packets. Application sets this + * to 'pkt_out' array size. A successful operation sets + * this to the number of outputted packets + * (1 ... num_out). + * @param param Inbound operation parameters + * + * @return Number of input packets consumed (0 ... num_in) * @retval <0 On failure * - * @see odp_packet_user_ptr(), odp_packet_user_area() + * @see odp_packet_user_ptr(), odp_packet_user_area(), odp_packet_l3_offset(), + * odp_packet_l4_offset() */ -int odp_ipsec_in(const odp_ipsec_op_param_t *input, - odp_ipsec_op_result_t *output); +int odp_ipsec_in(const odp_packet_t pkt_in[], int num_in, + odp_packet_t pkt_out[], int *num_out, + const odp_ipsec_in_param_t *param); /** * Outbound synchronous IPSEC operation * * This operation does outbound IPSEC processing in synchronous mode * (ODP_IPSEC_OP_MODE_SYNC). A successful operation returns the number of - * packets consumed and outputs a new packet handle as well as an operation - * result for each outputted packet. The operation does not modify packets that - * it does not consume. It cannot consume all input packets if 'output.num_pkt' - * is smaller than 'input.num_pkt'. + * packets consumed and outputs a new packet handle for each outputted packet. + * Outputted packets contain IPSEC result metadata (odp_ipsec_packet_result_t), + * which should be checked for transformation errors, etc. Outputted packets + * with error status have not been transformed but the original packet is + * returned. The operation does not modify packets that it does not consume. + * It cannot consume all input packets if 'num_out' is smaller than 'num_in'. * * Packet context pointer and user area content are copied from input to output * packets. Output packets are allocated from the same pool(s) as input packets. * * When outbound IP fragmentation offload is enabled, the number of outputted - * packets (and corresponding per packet results) may be greater than - * the number of input packets. In that case, application may examine 'num_out' - * of each packet result (odp_ipsec_packet_result_t) to find out which - * fragments are originated from which input packet. + * packets may be greater than the number of input packets. + * + * Each input packet must have a valid value for these metadata (other metadata + * is ignored): + * - L3 offset: Offset to the first byte of the (outmost) IP header + * - L4 offset: Offset to the L4 header if L4 checksum offload is requested + * + * Additionally, input IP packet length (odp_packet_len() minus + * odp_packet_l3_offset()) must match values in protocol headers. Otherwise + * results are undefined. * * Packets are processed in the input order. Packet order is maintained from * input 'pkt' array to output 'pkt' array. Packet order is not guaranteed @@ -1201,31 +1202,40 @@ int odp_ipsec_in(const odp_ipsec_op_param_t *input, * with IPSEC, etc headers constructed according to the standards. The amount * and content of packet data before the IP header is undefined. * - * @param input Operation input parameters - * @param[out] output Operation results + * Each successfully transformed packet has a valid value for these metadata: + * - L3 offset: Offset to the first byte of the (outmost) IP header + * + * @param pkt_in Packets to be processed + * @param num_in Number of packets to be processed + * @param[out] pkt_out Packet handle array for resulting packets + * @param[in, out] num_out Number of resulting packets. Application sets this + * to 'pkt_out' array size. A successful operation sets + * this to the number of outputted packets + * (1 ... num_out). + * @param param Outbound operation parameters * - * @return Number of input packets consumed (0 ... input.num_pkt) + * @return Number of input packets consumed (0 ... num_in) * @retval <0 On failure * - * @see odp_packet_user_ptr(), odp_packet_user_area() + * @see odp_packet_user_ptr(), odp_packet_user_area(), odp_packet_l3_offset() */ -int odp_ipsec_out(const odp_ipsec_op_param_t *input, - odp_ipsec_op_result_t *output); +int odp_ipsec_out(const odp_packet_t pkt_in[], int num_in, + odp_packet_t pkt_out[], int *num_out, + const odp_ipsec_out_param_t *param); /** * Inbound asynchronous IPSEC operation * * This operation does inbound IPSEC processing in asynchronous mode. It - * processes packets otherwise identically to odp_ipsec_in(), but outputs all - * results through one or more ODP_EVENT_IPSEC_RESULT events with the following - * ordering considerations. + * processes packets otherwise identically to odp_ipsec_in(), but outputs + * resulting packets as ODP_EVENT_PACKET events (with ODP_EVENT_PACKET_IPSEC + * subtype). The following ordering considerations apply to the events. * * Asynchronous mode maintains packet order per SA when application calls the * operation within an ordered or atomic scheduler context of the same queue. - * Resulting events for the same SA are enqueued in order and packet handles - * (for the same SA) are stored in order within an event. Packet order per SA at - * a destination queue is the same as if application would have enqueued packets - * there with odp_queue_enq_multi(). + * Resulting events for the same SA are enqueued in order. Packet order per SA + * at a destination queue is the same as if application would have enqueued + * packets there with odp_queue_enq_multi(). * * Packet order is also maintained when application otherwise guarantees * (e.g. using locks) that the operation is not called simultaneously from @@ -1239,29 +1249,31 @@ int odp_ipsec_out(const odp_ipsec_op_param_t *input, * may be processed simultaneously in both modes (initiated by this function * and inline operation). * - * @param input Operation input parameters + * @param pkt Packets to be processed + * @param num Number of packets to be processed + * @param param Inbound operation parameters * - * @return Number of input packets consumed (0 ... input.num_pkt) + * @return Number of input packets consumed (0 ... num) * @retval <0 On failure * * @see odp_ipsec_in(), odp_ipsec_result() */ -int odp_ipsec_in_enq(const odp_ipsec_op_param_t *input); +int odp_ipsec_in_enq(const odp_packet_t pkt[], int num, + const odp_ipsec_in_param_t *param); /** * Outbound asynchronous IPSEC operation * * This operation does outbound IPSEC processing in asynchronous mode. It - * processes packets otherwise identically to odp_ipsec_out(), but outputs all - * results through one or more ODP_EVENT_IPSEC_RESULT events with the following - * ordering considerations. + * processes packets otherwise identically to odp_ipsec_out(), but outputs + * resulting packets as ODP_EVENT_PACKET events (with ODP_EVENT_PACKET_IPSEC + * subtype). The following ordering considerations apply to the events. * * Asynchronous mode maintains packet order per SA when application calls the * operation within an ordered or atomic scheduler context of the same queue. - * Resulting events for the same SA are enqueued in order and packet handles - * (for the same SA) are stored in order within an event. Packet order per SA at - * a destination queue is the same as if application would have enqueued packets - * there with odp_queue_enq_multi(). + * Resulting events for the same SA are enqueued in order. Packet order per SA + * at a destination queue is the same as if application would have enqueued + * packets there with odp_queue_enq_multi(). * * Packet order is also maintained when application otherwise guarantees * (e.g. using locks) that the operation is not called simultaneously from @@ -1273,14 +1285,17 @@ int odp_ipsec_in_enq(const odp_ipsec_op_param_t *input); * The function may be used also in inline processing mode, e.g. for IPSEC * packets for which inline processing is not possible. * - * @param input Operation input parameters + * @param pkt Packets to be processed + * @param num Number of packets to be processed + * @param param Outbound operation parameters * - * @return Number of input packets consumed (0 ... input.num_pkt) + * @return Number of input packets consumed (0 ... num) * @retval <0 On failure * * @see odp_ipsec_out(), odp_ipsec_result() */ -int odp_ipsec_out_enq(const odp_ipsec_op_param_t *input); +int odp_ipsec_out_enq(const odp_packet_t pkt[], int num, + const odp_ipsec_out_param_t *param); /** * Outbound inline IPSEC operation @@ -1288,42 +1303,75 @@ int odp_ipsec_out_enq(const odp_ipsec_op_param_t *input); * This operation does outbound inline IPSEC processing for the packets. It's * otherwise identical to odp_ipsec_out_enq(), but outputs all successfully * transformed packets to the specified output interface, instead of generating - * result events for those. + * events for those. * * Inline operation parameters are defined per packet. The array of parameters - * must have 'op_param.num_pkt' elements and is pointed to by 'inline_param'. + * must have 'num' elements and is pointed to by 'inline_param'. * - * @param op_param Operation parameters - * @param inline_param Outbound inline operation specific parameters + * @param pkt Packets to be processed + * @param num Number of packets to be processed + * @param param Outbound operation parameters + * @param inline_param Outbound inline operation specific parameters * - * @return Number of packets consumed (0 ... op_param.num_pkt) + * @return Number of packets consumed (0 ... num) * @retval <0 On failure * * @see odp_ipsec_out_enq() */ -int odp_ipsec_out_inline(const odp_ipsec_op_param_t *op_param, - const odp_ipsec_inline_op_param_t *inline_param); +int odp_ipsec_out_inline(const odp_packet_t pkt[], int num, + const odp_ipsec_out_param_t *param, + const odp_ipsec_out_inline_param_t *inline_param); + +/** + * Convert IPSEC processed packet event to packet handle + * + * Get packet handle to an IPSEC processed packet event. Event subtype must be + * ODP_EVENT_IPSEC_PACKET. IPSEC operation results can be examined with + * odp_ipsec_result(). + * + * @param ev Event handle + * + * @return Packet handle + * + * @see odp_event_subtype(), odp_ipsec_result() + */ +odp_packet_t odp_ipsec_packet_from_event(odp_event_t ev); /** - * Get IPSEC results from an ODP_EVENT_IPSEC_RESULT event + * Convert IPSEC processed packet handle to event + * + * The packet handle must be an output of an IPSEC operation. * - * Copies IPSEC operation results from an event. The event must be of - * type ODP_EVENT_IPSEC_RESULT. It must be freed before the application passes - * any resulting packet handles to other ODP calls. + * @param pkt Packet handle from IPSEC operation * - * @param[out] result Pointer to operation result for output. Maybe NULL, if - * application is interested only on the number of - * packets. - * @param event An ODP_EVENT_IPSEC_RESULT event + * @return Event handle + */ +odp_event_t odp_ipsec_packet_to_event(odp_packet_t pkt); + +/** + * Get IPSEC operation results from an IPSEC processed packet * - * @return Number of packets in the event. If this is larger than - * 'result.num_pkt', all packets did not fit into result struct and - * application must call the function again with a larger result struct. + * Successful IPSEC operations of all types (SYNC, ASYNC and INLINE) produce + * packets which contain IPSEC result metadata. This function copies the + * operation results from an IPSEC processed packet. Event subtype of this kind + * of packet is ODP_EVENT_PACKET_IPSEC. Results are undefined if a non-IPSEC + * processed packet is passed as input. + * + * Some packet API operations output a new packet handle + * (e.g. odp_packet_concat()). IPSEC metadata remain valid as long as the packet + * handle is not changed from the original (output of e.g. odp_ipsec_in() or + * odp_ipsec_packet_from_event() call) IPSEC processed packet handle. + * + * @param[out] result Pointer to operation result for output + * @param packet An IPSEC processed packet (ODP_EVENT_PACKET_IPSEC) + * + * @retval 0 On success * @retval <0 On failure * - * @see odp_ipsec_in_enq(), odp_ipsec_out_enq() + * @see odp_ipsec_in(), odp_ipsec_in_enq(), odp_ipsec_out(), + * odp_ipsec_out_enq(), odp_ipsec_packet_from_event() */ -int odp_ipsec_result(odp_ipsec_op_result_t *result, odp_event_t event); +int odp_ipsec_result(odp_ipsec_packet_result_t *result, odp_packet_t packet); /** * Get IPSEC status information from an ODP_EVENT_IPSEC_STATUS event diff --git a/platform/linux-generic/odp_ipsec.c b/platform/linux-generic/odp_ipsec.c index 10918dfb..c7eeb4ec 100644 --- a/platform/linux-generic/odp_ipsec.c +++ b/platform/linux-generic/odp_ipsec.c @@ -73,51 +73,68 @@ int odp_ipsec_sa_destroy(odp_ipsec_sa_t sa) return -1; } -int odp_ipsec_in(const odp_ipsec_op_param_t *input, - odp_ipsec_op_result_t *output) -{ - (void)input; - (void)output; +int odp_ipsec_in(const odp_packet_t pkt_in[], int num_in, + odp_packet_t pkt_out[], int *num_out, + const odp_ipsec_in_param_t *param) +{ + (void)pkt_in; + (void)num_in; + (void)pkt_out; + (void)num_out; + (void)param; return -1; } -int odp_ipsec_out(const odp_ipsec_op_param_t *input, - odp_ipsec_op_result_t *output) +int odp_ipsec_out(const odp_packet_t pkt_in[], int num_in, + odp_packet_t pkt_out[], int *num_out, + const odp_ipsec_out_param_t *param) { - (void)input; - (void)output; + (void)pkt_in; + (void)num_in; + (void)pkt_out; + (void)num_out; + (void)param; return -1; } -int odp_ipsec_in_enq(const odp_ipsec_op_param_t *input) +int odp_ipsec_in_enq(const odp_packet_t pkt[], int num, + const odp_ipsec_in_param_t *param) { - (void)input; + (void)pkt; + (void)num; + (void)param; return -1; } -int odp_ipsec_out_enq(const odp_ipsec_op_param_t *input) +int odp_ipsec_out_enq(const odp_packet_t pkt[], int num, + const odp_ipsec_out_param_t *param) { - (void)input; + (void)pkt; + (void)num; + (void)param; return -1; } -int odp_ipsec_out_inline(const odp_ipsec_op_param_t *op_param, - const odp_ipsec_inline_op_param_t *inline_param) +int odp_ipsec_out_inline(const odp_packet_t pkt[], int num, + const odp_ipsec_out_param_t *param, + const odp_ipsec_out_inline_param_t *inline_param) { - (void)op_param; + (void)pkt; + (void)num; + (void)param; (void)inline_param; return -1; } -int odp_ipsec_result(odp_ipsec_op_result_t *result, odp_event_t event) +int odp_ipsec_result(odp_ipsec_packet_result_t *result, odp_packet_t packet) { (void)result; - (void)event; + (void)packet; return -1; } @@ -145,6 +162,20 @@ void *odp_ipsec_sa_context(odp_ipsec_sa_t sa) return NULL; } +odp_packet_t odp_ipsec_packet_from_event(odp_event_t ev) +{ + (void)ev; + + return ODP_PACKET_INVALID; +} + +odp_event_t odp_ipsec_packet_to_event(odp_packet_t pkt) +{ + (void)pkt; + + return ODP_EVENT_INVALID; +} + uint64_t odp_ipsec_sa_to_u64(odp_ipsec_sa_t sa) { return _odp_pri(sa); commit 441fb964f22b2baaa97641e6f8dda88d32d153f6 Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Fri Jun 16 13:49:38 2017 +0300 api: event: add subtype to expand event type Event subtype gives more detailed information about the event. Two subtypes (basic and IPSEC packet) are introduced initially. Later on, other packet producing APIs (crypto, comp, etc) may also produce packet events with additional subtypes. Signed-off-by: Petri Savolainen <petri.savolainen(a)linaro.org> Reviewed-by: Bill Fischofer <bill.fischofer(a)linaro.org> Reviewed-by: Balasubramanian Manoharan <bala.manoharan(a)linaro.org> Reviewed-by: Nikhil Agarwal <Nikhil.agarwal(a)linaro.org> Reviewed-by: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/include/odp/api/spec/event.h b/include/odp/api/spec/event.h index f22efce5..2ad3ce84 100644 --- a/include/odp/api/spec/event.h +++ b/include/odp/api/spec/event.h @@ -37,21 +37,91 @@ extern "C" { /** * @typedef odp_event_type_t - * ODP event types: - * ODP_EVENT_BUFFER, ODP_EVENT_PACKET, ODP_EVENT_TIMEOUT, - * ODP_EVENT_CRYPTO_COMPL, ODP_EVENT_IPSEC_RESULT, ODP_EVENT_IPSEC_STATUS + * Event type + * + * Event type specifies purpose and general format of an event. It can be + * checked with odp_event_type() or odp_event_types(). Each event type has + * functions (e.g. odp_buffer_from_event()) to convert between the generic event + * handle (odp_event_t) and the type specific handle (e.g. odp_buffer_t). + * Results are undefined, if conversion function of a wrong event type is used. + * Application cannot change event type by chaining conversion functions. + * + * List of event types: + * - ODP_EVENT_BUFFER + * - Buffer event (odp_buffer_t) for simple data storage and message passing + * - ODP_EVENT_PACKET + * - Packet event (odp_packet_t) containing packet data and plenty of + * packet processing related metadata + * - ODP_EVENT_TIMEOUT + * - Timeout event (odp_timeout_t) from a timer + * - ODP_EVENT_CRYPTO_COMPL + * - Crypto completion event (odp_crypto_compl_t) + * - ODP_EVENT_IPSEC_STATUS + * - IPSEC status update event (odp_ipsec_status_t) */ /** - * Get event type + * @typedef odp_event_subtype_t + * Event subtype * - * @param event Event handle + * Event subtype expands event type specification by providing more detailed + * purpose and format of an event. It can be checked with odp_event_subtype() or + * odp_event_types(). Each event subtype may define specific functions + * (e.g. odp_ipsec_packet_from_event()) to convert between the generic event + * handle (odp_event_t) and event type specific handle (e.g. odp_packet_t). When + * subtype is known, these subtype specific functions should be preferred over + * the event type general function (e.g. odp_packet_from_event()). Results are + * undefined, if conversion function of a wrong event subtype is used. + * Application cannot change event subtype by chaining conversion functions. + * + * List of event subtypes: + * - ODP_EVENT_PACKET_BASIC + * - Packet event (odp_packet_t) with basic packet metadata + * - ODP_EVENT_PACKET_IPSEC + * - Packet event (odp_packet_t) generated as a result of an IPsec + * operation. It contains IPSEC specific metadata in addition to the basic + * packet metadata. + * - ODP_EVENT_NO_SUBTYPE + * - An event type does not have any subtypes defined + */ + +/** + * Event type of an event + * + * Event type specifies purpose and general format of an event. + * + * @param event Event handle * * @return Event type */ odp_event_type_t odp_event_type(odp_event_t event); /** + * Event subtype of an event + * + * Event subtype expands event type specification by providing more detailed + * purpose and format of an event. + * + * @param event Event handle + * + * @return Event subtype + */ +odp_event_subtype_t odp_event_subtype(odp_event_t event); + +/** + * Event type and subtype of an event + * + * Returns event type and outputs event subtype. + * + * @param event Event handle + * @param[out] subtype Pointer to event subtype for output + * + * @return Event type + */ +odp_event_type_t odp_event_types(odp_event_t event, + odp_event_subtype_t *subtype); + +/** * Get printable value for an odp_event_t * * @param hdl odp_event_t handle to be printed diff --git a/include/odp/arch/default/api/abi/event.h b/include/odp/arch/default/api/abi/event.h index 87220d63..ab3c0f75 100644 --- a/include/odp/arch/default/api/abi/event.h +++ b/include/odp/arch/default/api/abi/event.h @@ -29,10 +29,15 @@ typedef enum odp_event_type_t { ODP_EVENT_PACKET = 2, ODP_EVENT_TIMEOUT = 3, ODP_EVENT_CRYPTO_COMPL = 4, - ODP_EVENT_IPSEC_RESULT = 5, - ODP_EVENT_IPSEC_STATUS = 6 + ODP_EVENT_IPSEC_STATUS = 5 } odp_event_type_t; +typedef enum odp_event_subtype_t { + ODP_EVENT_NO_SUBTYPE = 0, + ODP_EVENT_PACKET_BASIC = 1, + ODP_EVENT_PACKET_IPSEC = 2 +} odp_event_subtype_t; + /** * @} */ diff --git a/platform/linux-generic/include/odp/api/plat/event_types.h b/platform/linux-generic/include/odp/api/plat/event_types.h index 0f517834..5b3a07e3 100644 --- a/platform/linux-generic/include/odp/api/plat/event_types.h +++ b/platform/linux-generic/include/odp/api/plat/event_types.h @@ -39,9 +39,15 @@ typedef enum odp_event_type_t { ODP_EVENT_PACKET = 2, ODP_EVENT_TIMEOUT = 3, ODP_EVENT_CRYPTO_COMPL = 4, - ODP_EVENT_IPSEC_RESULT = 5 + ODP_EVENT_IPSEC_STATUS = 5 } odp_event_type_t; +typedef enum odp_event_subtype_t { + ODP_EVENT_NO_SUBTYPE = 0, + ODP_EVENT_PACKET_BASIC = 1, + ODP_EVENT_PACKET_IPSEC = 2 +} odp_event_subtype_t; + /** * @} */ commit d5ff31f14e87c815d3a0e0461206aa2687152dd5 Merge: 4bcd15e2 09a7800c Author: Maxim Uvarov <maxim.uvarov(a)linaro.org> Date: Thu Jun 22 18:41:49 2017 +0300 Merge branch 'master' into api-next diff --cc platform/linux-generic/odp_packet_io.c index 8fb5b5ee,e5436a90..4dd28549 --- a/platform/linux-generic/odp_packet_io.c +++ b/platform/linux-generic/odp_packet_io.c @@@ -564,15 -563,14 +563,14 @@@ static inline int pktin_recv_buf(odp_pk for (i = 0; i < pkts; i++) { pkt = packets[i]; pkt_hdr = odp_packet_hdr(pkt); - buf = _odp_packet_to_buffer(pkt); - buf_hdr = buf_hdl_to_hdr(buf); + buf_hdr = packet_to_buf_hdr(pkt); if (pkt_hdr->p.input_flags.dst_queue) { - queue_entry_t *dst_queue; + queue_t dst_queue; int ret; - dst_queue = queue_to_qentry(pkt_hdr->dst_queue); - ret = queue_enq(dst_queue, buf_hdr); + dst_queue = queue_fn->from_ext(pkt_hdr->dst_queue); + ret = queue_fn->enq(dst_queue, buf_hdr); if (ret < 0) odp_packet_free(pkt); continue; @@@ -582,9 -580,9 +580,9 @@@ return num_rx; } -int pktout_enqueue(queue_entry_t *qentry, odp_buffer_hdr_t *buf_hdr) +int pktout_enqueue(queue_t qentry, odp_buffer_hdr_t *buf_hdr) { - odp_packet_t pkt = _odp_packet_from_buffer(buf_hdr->handle.handle); + odp_packet_t pkt = packet_from_buf_hdr(buf_hdr); int len = 1; int nbr; @@@ -611,9 -612,9 +609,9 @@@ int pktout_enq_multi(queue_t qentry, od return nbr; for (i = 0; i < num; ++i) - pkt_tbl[i] = _odp_packet_from_buffer(buf_hdr[i]->handle.handle); + pkt_tbl[i] = packet_from_buf_hdr(buf_hdr[i]); - nbr = odp_pktout_send(qentry->s.pktout, pkt_tbl, num); + nbr = odp_pktout_send(queue_fn->get_pktout(qentry), pkt_tbl, num); return nbr; } ----------------------------------------------------------------------- Summary of changes: .travis.yml | 1 + doc/application-api-guide/examples.dox | 5 + doc/users-guide/users-guide.adoc | 36 +- example/Makefile.am | 1 + example/ipfragreass/.gitignore | 3 + example/ipfragreass/Makefile.am | 23 + example/ipfragreass/odp_ipfragreass.c | 376 ++++++++++ example/ipfragreass/odp_ipfragreass_atomics.h | 55 ++ example/ipfragreass/odp_ipfragreass_atomics_arm.h | 120 ++++ example/ipfragreass/odp_ipfragreass_fragment.c | 99 +++ example/ipfragreass/odp_ipfragreass_fragment.h | 28 + example/ipfragreass/odp_ipfragreass_helpers.c | 124 ++++ example/ipfragreass/odp_ipfragreass_helpers.h | 79 +++ example/ipfragreass/odp_ipfragreass_ip.h | 251 +++++++ example/ipfragreass/odp_ipfragreass_reassemble.c | 771 +++++++++++++++++++++ example/ipfragreass/odp_ipfragreass_reassemble.h | 211 ++++++ example/m4/configure.m4 | 1 + include/odp/api/spec/event.h | 80 ++- include/odp/api/spec/feature.h | 69 ++ include/odp/api/spec/init.h | 15 + include/odp/api/spec/ipsec.h | 417 ++++++----- include/odp/arch/default/api/abi/event.h | 9 +- include/odp_api.h | 1 + platform/Makefile.inc | 1 + platform/linux-generic/Makefile.am | 3 +- platform/linux-generic/arch/arm/odp_cpu_arch.c | 38 +- .../include/odp/api/{support.h => feature.h} | 10 +- .../include/odp/api/plat/event_types.h | 8 +- .../linux-generic/include/odp_packet_internal.h | 10 + platform/linux-generic/odp_init.c | 5 + platform/linux-generic/odp_ipsec.c | 67 +- platform/linux-generic/odp_packet_io.c | 8 +- platform/linux-generic/pktio/loop.c | 4 +- test/common_plat/validation/api/init/init.c | 8 +- 34 files changed, 2704 insertions(+), 233 deletions(-) create mode 100644 example/ipfragreass/.gitignore create mode 100644 example/ipfragreass/Makefile.am create mode 100644 example/ipfragreass/odp_ipfragreass.c create mode 100644 example/ipfragreass/odp_ipfragreass_atomics.h create mode 100644 example/ipfragreass/odp_ipfragreass_atomics_arm.h create mode 100644 example/ipfragreass/odp_ipfragreass_fragment.c create mode 100644 example/ipfragreass/odp_ipfragreass_fragment.h create mode 100644 example/ipfragreass/odp_ipfragreass_helpers.c create mode 100644 example/ipfragreass/odp_ipfragreass_helpers.h create mode 100644 example/ipfragreass/odp_ipfragreass_ip.h create mode 100644 example/ipfragreass/odp_ipfragreass_reassemble.c create mode 100644 example/ipfragreass/odp_ipfragreass_reassemble.h create mode 100644 include/odp/api/spec/feature.h copy platform/linux-generic/include/odp/api/{support.h => feature.h} (59%) hooks/post-receive --

6 years, 10 months

1
0
0 0

[COMMIT] branch monarch_lts updated. v1.11.0.0_monarch-6-g0fcec770

by git＠git-us.linaro.org

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "". The branch, monarch_lts has been updated via 0fcec77081fd1df2f862419b42014bbf700d9c32 (commit) from 164fc322a6bb346d6c1f545ac69501b774cc56c2 (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- commit 0fcec77081fd1df2f862419b42014bbf700d9c32 Author: Matias Elo <matias.elo(a)nokia.com> Date: Tue Jun 20 10:13:20 2017 +0300 linux-gen: socket: remove limits for maximum RX/TX burst size backport of: 373f97f8 Remove unnecessary limits for maximum RX/TX burst size. Fixes: https://bugs.linaro.org/show_bug.cgi?id=3039 Signed-off-by: Matias Elo <matias.elo(a)nokia.com> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/platform/linux-generic/include/odp_packet_socket.h b/platform/linux-generic/include/odp_packet_socket.h index ccff69a5..8c1012a8 100644 --- a/platform/linux-generic/include/odp_packet_socket.h +++ b/platform/linux-generic/include/odp_packet_socket.h @@ -27,11 +27,6 @@ * Packet socket config: */ -/** Max receive (Rx) burst size*/ -#define ODP_PACKET_SOCKET_MAX_BURST_RX 32 -/** Max transmit (Tx) burst size*/ -#define ODP_PACKET_SOCKET_MAX_BURST_TX 32 - /* * This makes sure that building for kernels older than 3.1 works * and a fanout requests fails (for invalid packet socket option) @@ -46,8 +41,6 @@ typedef struct { int sockfd; /**< socket descriptor */ odp_pool_t pool; /**< pool to alloc packets from */ unsigned char if_mac[ETH_ALEN]; /**< IF eth mac addr */ - uint8_t *cache_ptr[ODP_PACKET_SOCKET_MAX_BURST_RX]; - odp_shm_t shm; } pkt_sock_t; /** packet mmap ring */ diff --git a/platform/linux-generic/pktio/socket.c b/platform/linux-generic/pktio/socket.c index 0309b0d0..624104de 100644 --- a/platform/linux-generic/pktio/socket.c +++ b/platform/linux-generic/pktio/socket.c @@ -458,8 +458,6 @@ static int sock_close(pktio_entry_t *pktio_entry) return -1; } - odp_shm_free(pkt_sock->shm); - return 0; } @@ -471,13 +469,11 @@ static int sock_setup_pkt(pktio_entry_t *pktio_entry, const char *netdev, { int sockfd; int err; - int i; unsigned int if_idx; struct ifreq ethreq; struct sockaddr_ll sa_ll; char shm_name[ODP_SHM_NAME_LEN]; pkt_sock_t *pkt_sock = &pktio_entry->s.pkt_sock; - uint8_t *addr; odp_pktio_stats_t cur_stats; /* Init pktio entry */ @@ -491,18 +487,6 @@ static int sock_setup_pkt(pktio_entry_t *pktio_entry, const char *netdev, snprintf(shm_name, ODP_SHM_NAME_LEN, "%s-%s", "pktio", netdev); shm_name[ODP_SHM_NAME_LEN - 1] = '\0'; - pkt_sock->shm = odp_shm_reserve(shm_name, PACKET_JUMBO_LEN, - PACKET_JUMBO_LEN * - ODP_PACKET_SOCKET_MAX_BURST_RX, 0); - if (pkt_sock->shm == ODP_SHM_INVALID) - return -1; - - addr = odp_shm_addr(pkt_sock->shm); - for (i = 0; i < ODP_PACKET_SOCKET_MAX_BURST_RX; i++) { - pkt_sock->cache_ptr[i] = addr; - addr += PACKET_JUMBO_LEN; - } - sockfd = socket(AF_PACKET, SOCK_RAW, htons(ETH_P_ALL)); if (sockfd == -1) { __odp_errno = errno; @@ -616,15 +600,11 @@ static int sock_mmsg_recv(pktio_entry_t *pktio_entry, int index ODP_UNUSED, odp_time_t *ts = NULL; const int sockfd = pkt_sock->sockfd; int msgvec_len; - struct mmsghdr msgvec[ODP_PACKET_SOCKET_MAX_BURST_RX]; + struct mmsghdr msgvec[len]; int nb_rx = 0; int recv_msgs; - uint8_t **recv_cache; int i; - if (odp_unlikely(len > ODP_PACKET_SOCKET_MAX_BURST_RX)) - return -1; - odp_ticketlock_lock(&pktio_entry->s.rxl); if (pktio_entry->s.config.pktin.bit.ts_all || @@ -632,10 +612,10 @@ static int sock_mmsg_recv(pktio_entry_t *pktio_entry, int index ODP_UNUSED, ts = &ts_val; memset(msgvec, 0, sizeof(msgvec)); - recv_cache = pkt_sock->cache_ptr; if (pktio_cls_enabled(pktio_entry)) { - struct iovec iovecs[ODP_PACKET_SOCKET_MAX_BURST_RX]; + struct iovec iovecs[len]; + uint8_t recv_cache[len][PACKET_JUMBO_LEN]; for (i = 0; i < (int)len; i++) { msgvec[i].msg_hdr.msg_iovlen = 1; @@ -686,8 +666,7 @@ static int sock_mmsg_recv(pktio_entry_t *pktio_entry, int index ODP_UNUSED, pkt_table[nb_rx++] = pkt; } } else { - struct iovec iovecs[ODP_PACKET_SOCKET_MAX_BURST_RX] - [ODP_BUFFER_MAX_SEG]; + struct iovec iovecs[len][ODP_BUFFER_MAX_SEG]; for (i = 0; i < (int)len; i++) { pkt_table[i] = packet_alloc(pkt_sock->pool, @@ -770,15 +749,12 @@ static int sock_mmsg_send(pktio_entry_t *pktio_entry, int index ODP_UNUSED, const odp_packet_t pkt_table[], int len) { pkt_sock_t *pkt_sock = &pktio_entry->s.pkt_sock; - struct mmsghdr msgvec[ODP_PACKET_SOCKET_MAX_BURST_TX]; - struct iovec iovecs[ODP_PACKET_SOCKET_MAX_BURST_TX][ODP_BUFFER_MAX_SEG]; + struct mmsghdr msgvec[len]; + struct iovec iovecs[len][ODP_BUFFER_MAX_SEG]; int ret; int sockfd; int n, i; - if (odp_unlikely(len > ODP_PACKET_SOCKET_MAX_BURST_TX)) - return -1; - odp_ticketlock_lock(&pktio_entry->s.txl); sockfd = pkt_sock->sockfd; ----------------------------------------------------------------------- Summary of changes: platform/linux-generic/include/odp_packet_socket.h | 7 ----- platform/linux-generic/pktio/socket.c | 36 ++++------------------ 2 files changed, 6 insertions(+), 37 deletions(-) hooks/post-receive --

6 years, 10 months

1
0
0 0

[COMMIT] branch master updated. v1.15.0.0-5-g09a7800c

by git＠git-us.linaro.org

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "". The branch, master has been updated via 09a7800c4c4a093fb962e362952f9cf562d2fc98 (commit) from ddf2ac38b1288283f7b3d7c9e16e36f392685866 (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- commit 09a7800c4c4a093fb962e362952f9cf562d2fc98 Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Thu Jun 22 12:17:13 2017 +0300 linux-gen: time: fix ARM compile for GCC 4.8 Use __aarch64__ instead of __ARM_ARCH, since it's backwards compatible between GCC versions. Fixes bug https://bugs.linaro.org/show_bug.cgi?id=3066 Signed-off-by: Petri Savolainen <petri.savolainen(a)linaro.org> Reviewed-by: Brian Brooks <brian.brooks(a)arm.com> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/platform/linux-generic/arch/arm/odp_cpu_arch.c b/platform/linux-generic/arch/arm/odp_cpu_arch.c index 91d439d9..28ee508b 100644 --- a/platform/linux-generic/arch/arm/odp_cpu_arch.c +++ b/platform/linux-generic/arch/arm/odp_cpu_arch.c @@ -68,7 +68,7 @@ int cpu_has_global_time(void) uint64_t cpu_global_time(void) { -#if __ARM_ARCH == 8 +#ifdef __aarch64__ uint64_t cntvct; /* @@ -88,7 +88,7 @@ uint64_t cpu_global_time(void) uint64_t cpu_global_time_freq(void) { -#if __ARM_ARCH == 8 +#ifdef __aarch64__ uint64_t cntfrq; __asm__ volatile("mrs %0, cntfrq_el0" : "=r"(cntfrq) : : ); ----------------------------------------------------------------------- Summary of changes: platform/linux-generic/arch/arm/odp_cpu_arch.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) hooks/post-receive --

6 years, 10 months

1
0
0 0

[COMMIT] branch master updated. v1.15.0.0-4-gddf2ac38

by git＠git-us.linaro.org

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "". The branch, master has been updated via ddf2ac38b1288283f7b3d7c9e16e36f392685866 (commit) from f4337956ee3be1e783e2389eb5a5ceecb4b0b0ca (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- commit ddf2ac38b1288283f7b3d7c9e16e36f392685866 Author: Maxim Uvarov <maxim.uvarov(a)linaro.org> Date: Thu Jun 22 11:42:05 2017 +0300 travis: back to stable image until issue is resolved after update tap pktio hangs. Back to stable image untill we solve failure. Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/.travis.yml b/.travis.yml index 50b3b9c2..67a2df64 100644 --- a/.travis.yml +++ b/.travis.yml @@ -10,6 +10,7 @@ language: c sudo: required dist: trusty +group: deprecated-2017Q2 env: global: # COVERITY_SCAN_TOKEN ----------------------------------------------------------------------- Summary of changes: .travis.yml | 1 + 1 file changed, 1 insertion(+) hooks/post-receive --

6 years, 10 months

1
0
0 0

[COMMIT] branch api-next updated. v1.15.0.0-355-g4bcd15e2

by git＠git-us.linaro.org

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "". The branch, api-next has been updated via 4bcd15e2bfb839ac0ef6419aa9e7b720958638f7 (commit) from 58a8ec90b57eae43be6461d1ae407e70a87c65d2 (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- commit 4bcd15e2bfb839ac0ef6419aa9e7b720958638f7 Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Mon Jun 12 14:11:33 2017 +0300 linux-gen: queue: clean up after modular interface Clean up function and parameter naming after modular interface patch. Queue_t type is referred as "queue internal": queue_int or q_int. Term "handle" is reserved for API level handles (e.g. odp_queue_t, odp_pktio_t, etc) through out linux-gen implementation. Signed-off-by: Petri Savolainen <petri.savolainen(a)linaro.org> Reviewed-and-tested-by: Bill Fischofer <bill.fischofer(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/platform/linux-generic/include/odp_queue_if.h b/platform/linux-generic/include/odp_queue_if.h index 4359435f..168d0e9e 100644 --- a/platform/linux-generic/include/odp_queue_if.h +++ b/platform/linux-generic/include/odp_queue_if.h @@ -53,24 +53,24 @@ typedef int (*queue_term_global_fn_t)(void); typedef int (*queue_init_local_fn_t)(void); typedef int (*queue_term_local_fn_t)(void); typedef queue_t (*queue_from_ext_fn_t)(odp_queue_t handle); -typedef odp_queue_t (*queue_to_ext_fn_t)(queue_t handle); -typedef int (*queue_enq_fn_t)(queue_t handle, odp_buffer_hdr_t *); -typedef int (*queue_enq_multi_fn_t)(queue_t handle, odp_buffer_hdr_t **, int); -typedef odp_buffer_hdr_t *(*queue_deq_fn_t)(queue_t handle); -typedef int (*queue_deq_multi_fn_t)(queue_t handle, odp_buffer_hdr_t **, int); -typedef odp_pktout_queue_t (*queue_get_pktout_fn_t)(queue_t handle); -typedef void (*queue_set_pktout_fn_t)(queue_t handle, odp_pktio_t pktio, +typedef odp_queue_t (*queue_to_ext_fn_t)(queue_t q_int); +typedef int (*queue_enq_fn_t)(queue_t q_int, odp_buffer_hdr_t *); +typedef int (*queue_enq_multi_fn_t)(queue_t q_int, odp_buffer_hdr_t **, int); +typedef odp_buffer_hdr_t *(*queue_deq_fn_t)(queue_t q_int); +typedef int (*queue_deq_multi_fn_t)(queue_t q_int, odp_buffer_hdr_t **, int); +typedef odp_pktout_queue_t (*queue_get_pktout_fn_t)(queue_t q_int); +typedef void (*queue_set_pktout_fn_t)(queue_t q_int, odp_pktio_t pktio, int index); -typedef odp_pktin_queue_t (*queue_get_pktin_fn_t)(queue_t handle); -typedef void (*queue_set_pktin_fn_t)(queue_t handle, odp_pktio_t pktio, +typedef odp_pktin_queue_t (*queue_get_pktin_fn_t)(queue_t q_int); +typedef void (*queue_set_pktin_fn_t)(queue_t q_int, odp_pktio_t pktio, int index); -typedef void (*queue_set_enq_fn_t)(queue_t handle, queue_enq_fn_t func); -typedef void (*queue_set_enq_multi_fn_t)(queue_t handle, +typedef void (*queue_set_enq_fn_t)(queue_t q_int, queue_enq_fn_t func); +typedef void (*queue_set_enq_multi_fn_t)(queue_t q_int, queue_enq_multi_fn_t func); -typedef void (*queue_set_deq_fn_t)(queue_t handle, queue_deq_fn_t func); -typedef void (*queue_set_deq_multi_fn_t)(queue_t handle, +typedef void (*queue_set_deq_fn_t)(queue_t q_int, queue_deq_fn_t func); +typedef void (*queue_set_deq_multi_fn_t)(queue_t q_int, queue_deq_multi_fn_t func); -typedef void (*queue_set_type_fn_t)(queue_t handle, odp_queue_type_t type); +typedef void (*queue_set_type_fn_t)(queue_t q_int, odp_queue_type_t type); /* Queue functions towards other internal components */ typedef struct { diff --git a/platform/linux-generic/include/odp_queue_internal.h b/platform/linux-generic/include/odp_queue_internal.h index 0c31ce8a..d79abd23 100644 --- a/platform/linux-generic/include/odp_queue_internal.h +++ b/platform/linux-generic/include/odp_queue_internal.h @@ -78,9 +78,9 @@ static inline uint32_t queue_to_id(odp_queue_t handle) return _odp_typeval(handle) - 1; } -static inline queue_entry_t *qentry_from_int(queue_t handle) +static inline queue_entry_t *qentry_from_int(queue_t q_int) { - return (queue_entry_t *)(void *)(handle); + return (queue_entry_t *)(void *)(q_int); } static inline queue_t qentry_to_int(queue_entry_t *qentry) diff --git a/platform/linux-generic/include/odp_schedule_if.h b/platform/linux-generic/include/odp_schedule_if.h index 9adacef7..5d10cd37 100644 --- a/platform/linux-generic/include/odp_schedule_if.h +++ b/platform/linux-generic/include/odp_schedule_if.h @@ -26,7 +26,7 @@ typedef int (*schedule_init_queue_fn_t)(uint32_t queue_index, typedef void (*schedule_destroy_queue_fn_t)(uint32_t queue_index); typedef int (*schedule_sched_queue_fn_t)(uint32_t queue_index); typedef int (*schedule_unsched_queue_fn_t)(uint32_t queue_index); -typedef int (*schedule_ord_enq_multi_fn_t)(queue_t handle, +typedef int (*schedule_ord_enq_multi_fn_t)(queue_t q_int, void *buf_hdr[], int num, int *ret); typedef int (*schedule_init_global_fn_t)(void); typedef int (*schedule_term_global_fn_t)(void); diff --git a/platform/linux-generic/odp_queue.c b/platform/linux-generic/odp_queue.c index 3e18f578..19945584 100644 --- a/platform/linux-generic/odp_queue.c +++ b/platform/linux-generic/odp_queue.c @@ -35,20 +35,22 @@ #include <string.h> #include <inttypes.h> +static int queue_init(queue_entry_t *queue, const char *name, + const odp_queue_param_t *param); + typedef struct queue_table_t { queue_entry_t queue[ODP_CONFIG_QUEUES]; } queue_table_t; static queue_table_t *queue_tbl; -static queue_t queue_from_ext(odp_queue_t handle); -static int _queue_enq(queue_t handle, odp_buffer_hdr_t *buf_hdr); -static odp_buffer_hdr_t *_queue_deq(queue_t handle); +static inline queue_entry_t *handle_to_qentry(odp_queue_t handle) +{ + uint32_t queue_id; -static int _queue_enq_multi(queue_t handle, odp_buffer_hdr_t *buf_hdr[], - int num); -static int _queue_deq_multi(queue_t handle, odp_buffer_hdr_t *buf_hdr[], - int num); + queue_id = queue_to_id(handle); + return get_qentry(queue_id); +} static inline odp_queue_t queue_from_id(uint32_t queue_id) { @@ -70,50 +72,6 @@ queue_entry_t *get_qentry(uint32_t queue_id) return &queue_tbl->queue[queue_id]; } -static int queue_init(queue_entry_t *queue, const char *name, - const odp_queue_param_t *param) -{ - if (name == NULL) { - queue->s.name[0] = 0; - } else { - strncpy(queue->s.name, name, ODP_QUEUE_NAME_LEN - 1); - queue->s.name[ODP_QUEUE_NAME_LEN - 1] = 0; - } - memcpy(&queue->s.param, param, sizeof(odp_queue_param_t)); - if (queue->s.param.sched.lock_count > sched_fn->max_ordered_locks()) - return -1; - - if (param->type == ODP_QUEUE_TYPE_SCHED) { - queue->s.param.deq_mode = ODP_QUEUE_OP_DISABLED; - - if (param->sched.sync == ODP_SCHED_SYNC_ORDERED) { - unsigned i; - - odp_atomic_init_u64(&queue->s.ordered.ctx, 0); - odp_atomic_init_u64(&queue->s.ordered.next_ctx, 0); - - for (i = 0; i < queue->s.param.sched.lock_count; i++) - odp_atomic_init_u64(&queue->s.ordered.lock[i], - 0); - } - } - queue->s.type = queue->s.param.type; - - queue->s.enqueue = _queue_enq; - queue->s.dequeue = _queue_deq; - queue->s.enqueue_multi = _queue_enq_multi; - queue->s.dequeue_multi = _queue_deq_multi; - - queue->s.pktin = PKTIN_INVALID; - queue->s.pktout = PKTOUT_INVALID; - - queue->s.head = NULL; - queue->s.tail = NULL; - - return 0; -} - - static int queue_init_global(void) { uint32_t i; @@ -204,27 +162,27 @@ static int queue_capability(odp_queue_capability_t *capa) static odp_queue_type_t queue_type(odp_queue_t handle) { - return qentry_from_int(queue_from_ext(handle))->s.type; + return handle_to_qentry(handle)->s.type; } static odp_schedule_sync_t queue_sched_type(odp_queue_t handle) { - return qentry_from_int(queue_from_ext(handle))->s.param.sched.sync; + return handle_to_qentry(handle)->s.param.sched.sync; } static odp_schedule_prio_t queue_sched_prio(odp_queue_t handle) { - return qentry_from_int(queue_from_ext(handle))->s.param.sched.prio; + return handle_to_qentry(handle)->s.param.sched.prio; } static odp_schedule_group_t queue_sched_group(odp_queue_t handle) { - return qentry_from_int(queue_from_ext(handle))->s.param.sched.group; + return handle_to_qentry(handle)->s.param.sched.group; } static int queue_lock_count(odp_queue_t handle) { - queue_entry_t *queue = qentry_from_int(queue_from_ext(handle)); + queue_entry_t *queue = handle_to_qentry(handle); return queue->s.param.sched.sync == ODP_SCHED_SYNC_ORDERED ? (int)queue->s.param.sched.lock_count : -1; @@ -299,7 +257,7 @@ void sched_cb_queue_destroy_finalize(uint32_t queue_index) static int queue_destroy(odp_queue_t handle) { queue_entry_t *queue; - queue = qentry_from_int(queue_from_ext(handle)); + queue = handle_to_qentry(handle); if (handle == ODP_QUEUE_INVALID) return -1; @@ -352,14 +310,14 @@ static int queue_context_set(odp_queue_t handle, void *context, uint32_t len ODP_UNUSED) { odp_mb_full(); - qentry_from_int(queue_from_ext(handle))->s.param.context = context; + handle_to_qentry(handle)->s.param.context = context; odp_mb_full(); return 0; } static void *queue_context(odp_queue_t handle) { - return qentry_from_int(queue_from_ext(handle))->s.param.context; + return handle_to_qentry(handle)->s.param.context; } static odp_queue_t queue_lookup(const char *name) @@ -385,7 +343,7 @@ static odp_queue_t queue_lookup(const char *name) return ODP_QUEUE_INVALID; } -static inline int enq_multi(queue_t handle, odp_buffer_hdr_t *buf_hdr[], +static inline int enq_multi(queue_t q_int, odp_buffer_hdr_t *buf_hdr[], int num) { int sched = 0; @@ -393,8 +351,8 @@ static inline int enq_multi(queue_t handle, odp_buffer_hdr_t *buf_hdr[], queue_entry_t *queue; odp_buffer_hdr_t *hdr, *tail, *next_hdr; - queue = qentry_from_int(handle); - if (sched_fn->ord_enq_multi(handle, (void **)buf_hdr, num, &ret)) + queue = qentry_from_int(q_int); + if (sched_fn->ord_enq_multi(q_int, (void **)buf_hdr, num, &ret)) return ret; /* Optimize the common case of single enqueue */ @@ -462,17 +420,17 @@ static inline int enq_multi(queue_t handle, odp_buffer_hdr_t *buf_hdr[], return num; /* All events enqueued */ } -static int _queue_enq_multi(queue_t handle, odp_buffer_hdr_t *buf_hdr[], - int num) +static int queue_int_enq_multi(queue_t q_int, odp_buffer_hdr_t *buf_hdr[], + int num) { - return enq_multi(handle, buf_hdr, num); + return enq_multi(q_int, buf_hdr, num); } -static int _queue_enq(queue_t handle, odp_buffer_hdr_t *buf_hdr) +static int queue_int_enq(queue_t q_int, odp_buffer_hdr_t *buf_hdr) { int ret; - ret = enq_multi(handle, &buf_hdr, 1); + ret = enq_multi(q_int, &buf_hdr, 1); if (ret == 1) return 0; @@ -489,7 +447,7 @@ static int queue_enq_multi(odp_queue_t handle, const odp_event_t ev[], int num) if (num > QUEUE_MULTI_MAX) num = QUEUE_MULTI_MAX; - queue = qentry_from_int(queue_from_ext(handle)); + queue = handle_to_qentry(handle); for (i = 0; i < num; i++) buf_hdr[i] = buf_hdl_to_hdr(odp_buffer_from_event(ev[i])); @@ -503,13 +461,13 @@ static int queue_enq(odp_queue_t handle, odp_event_t ev) odp_buffer_hdr_t *buf_hdr; queue_entry_t *queue; - queue = qentry_from_int(queue_from_ext(handle)); + queue = handle_to_qentry(handle); buf_hdr = buf_hdl_to_hdr(odp_buffer_from_event(ev)); return queue->s.enqueue(qentry_to_int(queue), buf_hdr); } -static inline int deq_multi(queue_t handle, odp_buffer_hdr_t *buf_hdr[], +static inline int deq_multi(queue_t q_int, odp_buffer_hdr_t *buf_hdr[], int num) { odp_buffer_hdr_t *hdr, *next; @@ -517,7 +475,7 @@ static inline int deq_multi(queue_t handle, odp_buffer_hdr_t *buf_hdr[], queue_entry_t *queue; int updated = 0; - queue = qentry_from_int(handle); + queue = qentry_from_int(q_int); LOCK(&queue->s.lock); if (odp_unlikely(queue->s.status < QUEUE_STATUS_READY)) { /* Bad queue, or queue has been destroyed. @@ -583,18 +541,18 @@ static inline int deq_multi(queue_t handle, odp_buffer_hdr_t *buf_hdr[], return i; } -static int _queue_deq_multi(queue_t handle, odp_buffer_hdr_t *buf_hdr[], - int num) +static int queue_int_deq_multi(queue_t q_int, odp_buffer_hdr_t *buf_hdr[], + int num) { - return deq_multi(handle, buf_hdr, num); + return deq_multi(q_int, buf_hdr, num); } -static odp_buffer_hdr_t *_queue_deq(queue_t handle) +static odp_buffer_hdr_t *queue_int_deq(queue_t q_int) { odp_buffer_hdr_t *buf_hdr = NULL; int ret; - ret = deq_multi(handle, &buf_hdr, 1); + ret = deq_multi(q_int, &buf_hdr, 1); if (ret == 1) return buf_hdr; @@ -611,7 +569,7 @@ static int queue_deq_multi(odp_queue_t handle, odp_event_t events[], int num) if (num > QUEUE_MULTI_MAX) num = QUEUE_MULTI_MAX; - queue = qentry_from_int(queue_from_ext(handle)); + queue = handle_to_qentry(handle); ret = queue->s.dequeue_multi(qentry_to_int(queue), buf_hdr, num); @@ -627,7 +585,7 @@ static odp_event_t queue_deq(odp_queue_t handle) queue_entry_t *queue; odp_buffer_hdr_t *buf_hdr; - queue = qentry_from_int(queue_from_ext(handle)); + queue = handle_to_qentry(handle); buf_hdr = queue->s.dequeue(qentry_to_int(queue)); if (buf_hdr) @@ -636,6 +594,49 @@ static odp_event_t queue_deq(odp_queue_t handle) return ODP_EVENT_INVALID; } +static int queue_init(queue_entry_t *queue, const char *name, + const odp_queue_param_t *param) +{ + if (name == NULL) { + queue->s.name[0] = 0; + } else { + strncpy(queue->s.name, name, ODP_QUEUE_NAME_LEN - 1); + queue->s.name[ODP_QUEUE_NAME_LEN - 1] = 0; + } + memcpy(&queue->s.param, param, sizeof(odp_queue_param_t)); + if (queue->s.param.sched.lock_count > sched_fn->max_ordered_locks()) + return -1; + + if (param->type == ODP_QUEUE_TYPE_SCHED) { + queue->s.param.deq_mode = ODP_QUEUE_OP_DISABLED; + + if (param->sched.sync == ODP_SCHED_SYNC_ORDERED) { + unsigned i; + + odp_atomic_init_u64(&queue->s.ordered.ctx, 0); + odp_atomic_init_u64(&queue->s.ordered.next_ctx, 0); + + for (i = 0; i < queue->s.param.sched.lock_count; i++) + odp_atomic_init_u64(&queue->s.ordered.lock[i], + 0); + } + } + queue->s.type = queue->s.param.type; + + queue->s.enqueue = queue_int_enq; + queue->s.dequeue = queue_int_deq; + queue->s.enqueue_multi = queue_int_enq_multi; + queue->s.dequeue_multi = queue_int_deq_multi; + + queue->s.pktin = PKTIN_INVALID; + queue->s.pktout = PKTOUT_INVALID; + + queue->s.head = NULL; + queue->s.tail = NULL; + + return 0; +} + void queue_lock(queue_entry_t *queue) { LOCK(&queue->s.lock); @@ -776,64 +777,65 @@ static uint64_t queue_to_u64(odp_queue_t hdl) return _odp_pri(hdl); } -static odp_pktout_queue_t queue_get_pktout(queue_t handle) +static odp_pktout_queue_t queue_get_pktout(queue_t q_int) { - return qentry_from_int(handle)->s.pktout; + return qentry_from_int(q_int)->s.pktout; } -static void queue_set_pktout(queue_t handle, odp_pktio_t pktio, int index) +static void queue_set_pktout(queue_t q_int, odp_pktio_t pktio, int index) { - qentry_from_int(handle)->s.pktout.pktio = pktio; - qentry_from_int(handle)->s.pktout.index = index; + queue_entry_t *qentry = qentry_from_int(q_int); + + qentry->s.pktout.pktio = pktio; + qentry->s.pktout.index = index; } -static odp_pktin_queue_t queue_get_pktin(queue_t handle) +static odp_pktin_queue_t queue_get_pktin(queue_t q_int) { - return qentry_from_int(handle)->s.pktin; + return qentry_from_int(q_int)->s.pktin; } -static void queue_set_pktin(queue_t handle, odp_pktio_t pktio, int index) +static void queue_set_pktin(queue_t q_int, odp_pktio_t pktio, int index) { - qentry_from_int(handle)->s.pktin.pktio = pktio; - qentry_from_int(handle)->s.pktin.index = index; + queue_entry_t *qentry = qentry_from_int(q_int); + + qentry->s.pktin.pktio = pktio; + qentry->s.pktin.index = index; } -static void queue_set_enq_func(queue_t handle, queue_enq_fn_t func) +static void queue_set_enq_func(queue_t q_int, queue_enq_fn_t func) { - qentry_from_int(handle)->s.enqueue = func; + qentry_from_int(q_int)->s.enqueue = func; } -static void queue_set_enq_multi_func(queue_t handle, queue_enq_multi_fn_t func) +static void queue_set_enq_multi_func(queue_t q_int, queue_enq_multi_fn_t func) { - qentry_from_int(handle)->s.enqueue_multi = func; + qentry_from_int(q_int)->s.enqueue_multi = func; } -static void queue_set_deq_func(queue_t handle, queue_deq_fn_t func) +static void queue_set_deq_func(queue_t q_int, queue_deq_fn_t func) { - qentry_from_int(handle)->s.dequeue = func; + qentry_from_int(q_int)->s.dequeue = func; } -static void queue_set_deq_multi_func(queue_t handle, queue_deq_multi_fn_t func) +static void queue_set_deq_multi_func(queue_t q_int, queue_deq_multi_fn_t func) { - qentry_from_int(handle)->s.dequeue_multi = func; + qentry_from_int(q_int)->s.dequeue_multi = func; } -static void queue_set_type(queue_t handle, odp_queue_type_t type) +static void queue_set_type(queue_t q_int, odp_queue_type_t type) { - qentry_from_int(handle)->s.type = type; + qentry_from_int(q_int)->s.type = type; } static queue_t queue_from_ext(odp_queue_t handle) { - uint32_t queue_id; - - queue_id = queue_to_id(handle); - return qentry_to_int(get_qentry(queue_id)); + return qentry_to_int(handle_to_qentry(handle)); } -static odp_queue_t queue_to_ext(queue_t handle) +static odp_queue_t queue_to_ext(queue_t q_int) { - return qentry_from_int(handle)->s.handle; + return qentry_from_int(q_int)->s.handle; } /* API functions */ @@ -866,10 +868,10 @@ queue_fn_t queue_default_fn = { .term_local = queue_term_local, .from_ext = queue_from_ext, .to_ext = queue_to_ext, - .enq = _queue_enq, - .enq_multi = _queue_enq_multi, - .deq = _queue_deq, - .deq_multi = _queue_deq_multi, + .enq = queue_int_enq, + .enq_multi = queue_int_enq_multi, + .deq = queue_int_deq, + .deq_multi = queue_int_deq_multi, .get_pktout = queue_get_pktout, .set_pktout = queue_set_pktout, .get_pktin = queue_get_pktin, diff --git a/platform/linux-generic/odp_schedule.c b/platform/linux-generic/odp_schedule.c index 011d4dc4..92634235 100644 --- a/platform/linux-generic/odp_schedule.c +++ b/platform/linux-generic/odp_schedule.c @@ -712,12 +712,12 @@ static inline int copy_events(odp_event_t out_ev[], unsigned int max) return i; } -static int schedule_ord_enq_multi(queue_t handle, void *buf_hdr[], +static int schedule_ord_enq_multi(queue_t q_int, void *buf_hdr[], int num, int *ret) { int i; uint32_t stash_num = sched_local.ordered.stash_num; - queue_entry_t *dst_queue = qentry_from_int(handle); + queue_entry_t *dst_queue = qentry_from_int(q_int); queue_entry_t *src_queue = sched_local.ordered.src_queue; if (!sched_local.ordered.src_queue || sched_local.ordered.in_order) diff --git a/platform/linux-generic/odp_schedule_iquery.c b/platform/linux-generic/odp_schedule_iquery.c index bdf1a460..79086843 100644 --- a/platform/linux-generic/odp_schedule_iquery.c +++ b/platform/linux-generic/odp_schedule_iquery.c @@ -1159,12 +1159,12 @@ static inline void schedule_release_context(void) schedule_release_atomic(); } -static int schedule_ord_enq_multi(queue_t handle, void *buf_hdr[], +static int schedule_ord_enq_multi(queue_t q_int, void *buf_hdr[], int num, int *ret) { int i; uint32_t stash_num = thread_local.ordered.stash_num; - queue_entry_t *dst_queue = qentry_from_int(handle); + queue_entry_t *dst_queue = qentry_from_int(q_int); queue_entry_t *src_queue = thread_local.ordered.src_queue; if (!thread_local.ordered.src_queue || thread_local.ordered.in_order) diff --git a/platform/linux-generic/odp_schedule_sp.c b/platform/linux-generic/odp_schedule_sp.c index 91d70e3a..66715c13 100644 --- a/platform/linux-generic/odp_schedule_sp.c +++ b/platform/linux-generic/odp_schedule_sp.c @@ -414,10 +414,10 @@ static int unsched_queue(uint32_t qi ODP_UNUSED) return 0; } -static int ord_enq_multi(queue_t handle, void *buf_hdr[], int num, +static int ord_enq_multi(queue_t q_int, void *buf_hdr[], int num, int *ret) { - (void)handle; + (void)q_int; (void)buf_hdr; (void)num; (void)ret; ----------------------------------------------------------------------- Summary of changes: platform/linux-generic/include/odp_queue_if.h | 28 +-- .../linux-generic/include/odp_queue_internal.h | 4 +- platform/linux-generic/include/odp_schedule_if.h | 2 +- platform/linux-generic/odp_queue.c | 218 +++++++++++---------- platform/linux-generic/odp_schedule.c | 4 +- platform/linux-generic/odp_schedule_iquery.c | 4 +- platform/linux-generic/odp_schedule_sp.c | 4 +- 7 files changed, 133 insertions(+), 131 deletions(-) hooks/post-receive --

6 years, 10 months

1
0
0 0

[COMMIT] branch api-next updated. v1.15.0.0-354-g58a8ec90

by git＠git-us.linaro.org

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "". The branch, api-next has been updated via 58a8ec90b57eae43be6461d1ae407e70a87c65d2 (commit) via ebbd8301f396e3f2f106ed94ce5c3970ec8bacce (commit) from a5ef33a6f2575cd011cb05c3fb1b06d1c017f879 (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- commit 58a8ec90b57eae43be6461d1ae407e70a87c65d2 Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Fri Jun 2 12:38:20 2017 +0300 api: ipsec: add max number of cos capability Added capability for maximum number of different CoS in CLS pipelining. Signed-off-by: Petri Savolainen <petri.savolainen(a)linaro.org> Reviewed-by: Bill Fischofer <bill.fischofer(a)linaro.org> Reviewed-by: Balasubramanian Manoharan <bala.manoharan(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/include/odp/api/spec/ipsec.h b/include/odp/api/spec/ipsec.h index bbb28a31..65f0b066 100644 --- a/include/odp/api/spec/ipsec.h +++ b/include/odp/api/spec/ipsec.h @@ -262,6 +262,12 @@ typedef struct odp_ipsec_capability_t { */ odp_support_t pipeline_cls; + /** Maximum number of different destination CoSes in classification + * pipelining. The same CoS may be used for many SAs. This is equal or + * less than 'max_cos' capability in classifier API. + */ + uint32_t max_cls_cos; + /** Soft expiry limit in seconds support * * 0: Limit is not supported @@ -674,7 +680,8 @@ typedef struct odp_ipsec_sa_param_t { * classification through this CoS. Other result events are sent to * 'dest_queue'. This field is considered only when 'pipeline' is * ODP_IPSEC_PIPELINE_CLS. The CoS must not be shared between any pktio - * interface default CoS. + * interface default CoS. The maximum number of different CoS supported + * is defined by IPsec capability max_cls_cos. */ odp_cos_t dest_cos; commit ebbd8301f396e3f2f106ed94ce5c3970ec8bacce Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Fri Jun 2 12:38:19 2017 +0300 api: ipsec: refine packet order specification State more explicitly that async ipsec operations maintain packet order the same way as odp_queue_enq_multi(). Signed-off-by: Petri Savolainen <petri.savolainen(a)linaro.org> Reviewed-by: Bill Fischofer <bill.fischofer(a)linaro.org> Reviewed-by: Balasubramanian Manoharan <bala.manoharan(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/include/odp/api/spec/ipsec.h b/include/odp/api/spec/ipsec.h index 15f5e8be..bbb28a31 100644 --- a/include/odp/api/spec/ipsec.h +++ b/include/odp/api/spec/ipsec.h @@ -1213,13 +1213,19 @@ int odp_ipsec_out(const odp_ipsec_op_param_t *input, * results through one or more ODP_EVENT_IPSEC_RESULT events with the following * ordering considerations. * - * Asynchronous mode maintains (operation input) packet order per SA when - * application calls the operation within an ordered or atomic scheduler context - * of the same queue. Packet order is also maintained when application - * otherwise guarantees (e.g. using locks) that the operation is not called - * simultaneously from multiple threads for the same SA(s). Resulting - * events for the same SA are enqueued in order, and packet handles (for the - * same SA) are stored in order within an event. + * Asynchronous mode maintains packet order per SA when application calls the + * operation within an ordered or atomic scheduler context of the same queue. + * Resulting events for the same SA are enqueued in order and packet handles + * (for the same SA) are stored in order within an event. Packet order per SA at + * a destination queue is the same as if application would have enqueued packets + * there with odp_queue_enq_multi(). + * + * Packet order is also maintained when application otherwise guarantees + * (e.g. using locks) that the operation is not called simultaneously from + * multiple threads for the same SA(s). + * + * Logically, packet processing (e.g. sequence number check) happens in the + * output order as defined above. * * The function may be used also in inline processing mode, e.g. for IPSEC * packets for which inline processing is not possible. Packets for the same SA @@ -1243,13 +1249,19 @@ int odp_ipsec_in_enq(const odp_ipsec_op_param_t *input); * results through one or more ODP_EVENT_IPSEC_RESULT events with the following * ordering considerations. * - * Asynchronous mode maintains (operation input) packet order per SA when - * application calls the operation within an ordered or atomic scheduler context - * of the same queue. Packet order is also maintained when application - * otherwise guarantees (e.g. using locks) that the operation is not called - * simultaneously from multiple threads for the same SA(s). Resulting - * events for the same SA are enqueued in order, and packet handles (for the - * same SA) are stored in order within an event. + * Asynchronous mode maintains packet order per SA when application calls the + * operation within an ordered or atomic scheduler context of the same queue. + * Resulting events for the same SA are enqueued in order and packet handles + * (for the same SA) are stored in order within an event. Packet order per SA at + * a destination queue is the same as if application would have enqueued packets + * there with odp_queue_enq_multi(). + * + * Packet order is also maintained when application otherwise guarantees + * (e.g. using locks) that the operation is not called simultaneously from + * multiple threads for the same SA(s). + * + * Logically, packet processing (e.g. sequence number assignment) happens in the + * output order as defined above. * * The function may be used also in inline processing mode, e.g. for IPSEC * packets for which inline processing is not possible. ----------------------------------------------------------------------- Summary of changes: include/odp/api/spec/ipsec.h | 49 ++++++++++++++++++++++++++++++-------------- 1 file changed, 34 insertions(+), 15 deletions(-) hooks/post-receive --

6 years, 10 months

1
0
0 0

[COMMIT] branch master updated. v1.15.0.0-3-gf4337956

by git＠git-us.linaro.org

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "". The branch, master has been updated via f4337956ee3be1e783e2389eb5a5ceecb4b0b0ca (commit) via cd1f187d33d65cb54781b5a3c0864229b19c11c3 (commit) from 0d24adacfe17c8e0e1348f19cd8b75b64cb13ccf (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- commit f4337956ee3be1e783e2389eb5a5ceecb4b0b0ca Author: Joe Savage <joe.savage(a)arm.com> Date: Thu Apr 13 15:46:15 2017 +0100 example: add IPv4 fragmentation/reassembly example Add an example application implementing lock-free IPv4 fragmentation and reassembly functionality using ODP's packet "concat" and "split". Signed-off-by: Joe Savage <joe.savage(a)arm.com> Reviewed-and-tested-by: Bill Fischofer <bill.fischofer(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/doc/application-api-guide/examples.dox b/doc/application-api-guide/examples.dox index 60d4058f..80fe4670 100644 --- a/doc/application-api-guide/examples.dox +++ b/doc/application-api-guide/examples.dox @@ -28,3 +28,8 @@ * @example odp_timer_test.c * ODP timer example application */ + + /** + * @example odp_ipfragreass.c + * ODP IPv4 lock-free fragmentation and reassembly example application + */ diff --git a/example/Makefile.am b/example/Makefile.am index dfc07b67..9503a1ba 100644 --- a/example/Makefile.am +++ b/example/Makefile.am @@ -2,6 +2,7 @@ SUBDIRS = classifier \ generator \ hello \ ipsec \ + ipfragreass \ l2fwd_simple \ l3fwd \ packet \ diff --git a/example/ipfragreass/.gitignore b/example/ipfragreass/.gitignore new file mode 100644 index 00000000..d25d7584 --- /dev/null +++ b/example/ipfragreass/.gitignore @@ -0,0 +1,3 @@ +odp_ipfragreass +*.log +*.trs diff --git a/example/ipfragreass/Makefile.am b/example/ipfragreass/Makefile.am new file mode 100644 index 00000000..f805a69b --- /dev/null +++ b/example/ipfragreass/Makefile.am @@ -0,0 +1,23 @@ +include $(top_srcdir)/example/Makefile.inc + +bin_PROGRAMS = odp_ipfragreass$(EXEEXT) +odp_ipfragreass_LDFLAGS = $(AM_LDFLAGS) -static +odp_ipfragreass_CFLAGS = $(AM_CFLAGS) -I${top_srcdir}/example + +noinst_HEADERS = \ + $(top_srcdir)/example/ipfragreass/odp_ipfragreass_atomics.h \ + $(top_srcdir)/example/ipfragreass/odp_ipfragreass_atomics_arm.h \ + $(top_srcdir)/example/ipfragreass/odp_ipfragreass_fragment.h \ + $(top_srcdir)/example/ipfragreass/odp_ipfragreass_helpers.h \ + $(top_srcdir)/example/ipfragreass/odp_ipfragreass_ip.h \ + $(top_srcdir)/example/ipfragreass/odp_ipfragreass_reassemble.h \ + $(top_srcdir)/example/example_debug.h + +dist_odp_ipfragreass_SOURCES = odp_ipfragreass.c \ + odp_ipfragreass_fragment.c \ + odp_ipfragreass_helpers.c \ + odp_ipfragreass_reassemble.c + +if test_example +TESTS = odp_ipfragreass +endif diff --git a/example/ipfragreass/odp_ipfragreass.c b/example/ipfragreass/odp_ipfragreass.c new file mode 100644 index 00000000..6ad0db9f --- /dev/null +++ b/example/ipfragreass/odp_ipfragreass.c @@ -0,0 +1,376 @@ +/* Copyright (c) 2017, Linaro Limited + * All rights reserved. + * + * SPDX-License-Identifier: BSD-3-Clause + */ + +/** + * @file + * + * @example odp_ipfragreass.c ODP IPv4 lock-free fragmentation and reassembly + */ + +#include <stdio.h> +#include <stdlib.h> +#include <time.h> +#include <assert.h> + +#include <odp/helper/odph_api.h> +#include <example_debug.h> + +#include "odp_ipfragreass_fragment.h" +#include "odp_ipfragreass_reassemble.h" +#include "odp_ipfragreass_helpers.h" + +#define NUM_PACKETS 200 /**< Number of packets to fragment/reassemble */ +#define MAX_WORKERS 32 /**< Maximum number of worker threads */ +#define FRAGLISTS 16384 /**< Hash map size for reassembly */ + +#define MIN_MF_FRAG_SIZE 576 /**< Minimum fragment size */ +#define MAX_PKT_LEN 8192 /**< Maximum packet size */ +#define MAX_FRAGS_PER_PKT 6 /**< Maximum number of fragments per packet */ + +/** + * Derived parameters for packet storage (inc. pool configuration) + */ +#define POOL_MIN_SEG_LEN IP_HDR_LEN_MIN +#define POOL_UAREA_SIZE sizeof(struct packet) +#define MAX_FRAGS (MAX_FRAGS_PER_PKT * NUM_PACKETS) + +/** Output queue for fragmentation, input queue for reassembly */ +static odp_queue_t fragments; + +/** Output queue for reassembly, input queue for validation */ +static odp_queue_t reassembled_pkts; + +/** Number of packets reassembled thus far */ +static odp_atomic_u32_t packets_reassembled; + +/** Number of fragments processed per thread in reassembly (for printing) */ +static struct { + uint32_t frags; +} ODP_ALIGNED_CACHE thread_stats[MAX_WORKERS]; + +/** Shared hash map structure for reassembly */ +static union fraglist *fraglists; + +/** Barrier for synchronising reassembly worker threads */ +static odp_barrier_t barrier; + +/** + * Initialise the base structures required for execution of this application + * + * @param[out] instance ODP instance handle to initialise + * @param[out] fragment_pool Output for fragment pool creation + * @param[out] shm Output for reassembly shared memory creation + * @param[out] cpumask Output for worker threads CPU mask + * @param[out] num_workers Output for number of worker threads + */ +static void init(odp_instance_t *instance, odp_pool_t *fragment_pool, + odp_shm_t *shm, odp_cpumask_t *cpumask, int *num_workers) +{ + unsigned int seed = time(NULL); + int i; + odp_pool_param_t pool_params; + odp_queue_param_t frag_queue_params; + odp_queue_param_t reass_queue_params; + char cpumask_str[ODP_CPUMASK_STR_SIZE]; + + srand(seed); + printf("= Seed: %d\n", seed); + printf("= MTU: %d\n", MTU); + + /* ODP initialisation */ + if (odp_init_global(instance, NULL, NULL)) { + fprintf(stderr, "ERROR: ODP global init failed.\n"); + exit(1); + } + if (odp_init_local(*instance, ODP_THREAD_CONTROL)) { + fprintf(stderr, "ERROR: ODP local init failed.\n"); + exit(1); + } + + /* Create a pool for packet storage */ + odp_pool_param_init(&pool_params); + pool_params.pkt.seg_len = POOL_MIN_SEG_LEN; + pool_params.pkt.len = MAX_PKT_LEN; + pool_params.pkt.num = 2 * MAX_FRAGS + MAX_WORKERS; + pool_params.pkt.uarea_size = POOL_UAREA_SIZE; + pool_params.type = ODP_POOL_PACKET; + *fragment_pool = odp_pool_create("packet pool", &pool_params); + if (*fragment_pool == ODP_POOL_INVALID) { + fprintf(stderr, "ERROR: packet pool create failed.\n"); + exit(1); + } + + /* Reserve (and initialise) shared memory for reassembly fraglists */ + *shm = odp_shm_reserve("fraglists", FRAGLISTS * sizeof(union fraglist), + ODP_CACHE_LINE_SIZE, 0); + if (*shm == ODP_SHM_INVALID) { + fprintf(stderr, "ERROR: odp_shm_reserve\n"); + exit(1); + } + fraglists = odp_shm_addr(*shm); + if (fraglists == NULL) { + fprintf(stderr, "ERROR: odp_shm_addr\n"); + exit(1); + } + for (i = 0; i < FRAGLISTS; ++i) + init_fraglist(&fraglists[i]); + + /* Create a queue for holding fragments */ + odp_queue_param_init(&frag_queue_params); + frag_queue_params.type = ODP_QUEUE_TYPE_PLAIN; + frag_queue_params.enq_mode = ODP_QUEUE_OP_MT_UNSAFE; + fragments = odp_queue_create("fragments", &frag_queue_params); + if (fragments == ODP_QUEUE_INVALID) { + fprintf(stderr, "ERROR: odp_queue_create\n"); + exit(1); + } + + /* Create a queue for holding reassembled packets */ + odp_queue_param_init(&reass_queue_params); + reass_queue_params.type = ODP_QUEUE_TYPE_PLAIN; + reass_queue_params.deq_mode = ODP_QUEUE_OP_MT_UNSAFE; + reassembled_pkts = odp_queue_create("reassembled packets", + &reass_queue_params); + if (reassembled_pkts == ODP_QUEUE_INVALID) { + fprintf(stderr, "ERROR: odp_queue_create\n"); + exit(1); + } + + /* Set up worker threads */ + *num_workers = odp_cpumask_default_worker(cpumask, *num_workers); + odp_barrier_init(&barrier, *num_workers + 1); + odp_cpumask_to_str(cpumask, cpumask_str, sizeof(cpumask_str)); + printf("= Workers: %d\n", *num_workers); + printf("= CPU Mask: %s (first CPU: %d)\n\n", cpumask_str, + odp_cpumask_first(cpumask)); +} + +/** + * Reassembly worker thread function + * + * Repeatedly dequeues input fragments, validating them and then passing them + * to the reassembly procedure "reassemble_ipv4_packets". When a packet has + * been reassembled, it is added to the output queue, and when NUM_PACKETS + * packets have been completed the function returns. Thread 0 additionally + * executes the garbage collection procedure to clean up stale fragments. + * + * @param arg The thread number of this worker (masquerading as a pointer) + * + * @return Always returns zero + */ +static int run_worker(void *arg EXAMPLE_UNUSED) +{ + int threadno = odp_thread_id() - 1; + int iterations = 0; + odp_event_t ev; + + odp_barrier_wait(&barrier); + while (odp_atomic_load_u32(&packets_reassembled) < NUM_PACKETS) { + odp_packet_t pkt; + odp_time_t timestamp; + odph_ipv4hdr_t *hdr; + struct packet *fragment; + int reassembled; + + ev = odp_queue_deq(fragments); + if (ev == ODP_EVENT_INVALID) + break; + assert(odp_event_type(ev) == ODP_EVENT_PACKET); + ++thread_stats[threadno].frags; + + pkt = odp_packet_from_event(ev); + hdr = odp_packet_data(pkt); + fragment = odp_packet_user_area(pkt); + timestamp = odp_time_global(); + assert(fragment != NULL); + assert(odp_packet_len(pkt) == ipv4hdr_payload_len(*hdr) + + ipv4hdr_ihl(*hdr)); + assert(!ipv4hdr_more_fragments(*hdr) || + (odp_packet_len(pkt) >= MIN_MF_FRAG_SIZE && + (odp_packet_len(pkt) + - ipv4hdr_ihl(*hdr)) % 8 == 0)); + assert(odp_packet_len(pkt) <= MAX_PKT_LEN); + fragment->handle = pkt; + fragment->prev = NULL; + fragment->arrival.t = odp_time_to_ns(timestamp) / TS_RES_NS; + + reassembled = reassemble_ipv4_packets(fraglists, FRAGLISTS, + fragment, 1, + reassembled_pkts); + if (reassembled > 0) + odp_atomic_add_u32(&packets_reassembled, reassembled); + + /* + * Perform garbage collection of stale fragments every 50 + * iterations. (In real applications, use a timer!) + */ + if (threadno == 0 && iterations++ > 50) { + iterations = 0; + garbage_collect_fraglists(fraglists, FRAGLISTS, + reassembled_pkts, 0); + } + } + + while ((ev = odp_queue_deq(fragments)) != ODP_EVENT_INVALID) { + assert(odp_event_type(ev) == ODP_EVENT_PACKET); + odp_packet_free(odp_packet_from_event(ev)); + } + + return 0; +} + +/** + * ODP fragmentation and reassembly example main function + */ +int main(void) +{ + odp_instance_t instance; + odp_pool_t fragment_pool; + odp_shm_t shm; + odp_cpumask_t cpumask; + odph_odpthread_t threads[MAX_WORKERS] = {}; + odph_odpthread_params_t thread_params; + odp_packet_t dequeued_pkts[NUM_PACKETS]; + odp_event_t ev; + odp_u16be_t ip_id = 0; + odp_packet_t orig_pkts[NUM_PACKETS]; + odp_packet_t fragment_buffer[MAX_FRAGS]; + int total_fragments = 0; + int i; + int num_workers = MAX_WORKERS; + int reassembled; + + init(&instance, &fragment_pool, &shm, &cpumask, &num_workers); + + /* Packet generation & fragmentation */ + printf("\n= Fragmenting %d packets...\n", NUM_PACKETS); + for (i = 0; i < NUM_PACKETS; ++i) { + odp_packet_t packet; + int num_fragments; + + packet = pack_udp_ipv4_packet(fragment_pool, ip_id++, + MAX_PKT_LEN, + MTU + IP_HDR_LEN_MAX + 1); + if (packet == ODP_PACKET_INVALID) { + fprintf(stderr, "ERROR: pack_udp_ipv4_packet\n"); + return 1; + } + + orig_pkts[i] = odp_packet_copy(packet, fragment_pool); + if (orig_pkts[i] == ODP_PACKET_INVALID) { + fprintf(stderr, "ERROR: odp_packet_copy\n"); + return 1; + } + + if (fragment_ipv4_packet(packet, + &fragment_buffer[total_fragments], + &num_fragments)) { + fprintf(stderr, "ERROR: fragment_ipv4_packet\n"); + return 1; + } + + total_fragments += num_fragments; + } + + /* Shuffle the fragments around so they aren't necessarily in order */ + printf("\n= Shuffling %d fragments...\n", total_fragments); + shuffle(fragment_buffer, total_fragments); + + /* Insert the fragments into a queue for consumption */ + for (i = 0; i < total_fragments; ++i) { + ev = odp_packet_to_event(fragment_buffer[i]); + + if (odp_queue_enq(fragments, ev) < 0) { + fprintf(stderr, "ERROR: odp_queue_enq\n"); + return 1; + } + } + + /* Spawn the worker threads for reassembly */ + memset(&thread_params, 0, sizeof(thread_params)); + thread_params.start = run_worker; + thread_params.arg = 0; + thread_params.thr_type = ODP_THREAD_WORKER; + thread_params.instance = instance; + odph_odpthreads_create(threads, &cpumask, &thread_params); + + /* Go! */ + printf("\n= Starting reassembly...\n"); + odp_barrier_wait(&barrier); + + /* Wait for all threads to complete and output statistics */ + odph_odpthreads_join(threads); + for (i = 0; i < num_workers; ++i) + printf("=== Thread %02d processed %3d fragments\n", i, + thread_stats[i].frags); + + /* Dequeue the reassembled packets */ + for (reassembled = 0; (ev = odp_queue_deq(reassembled_pkts)) != + ODP_EVENT_INVALID; ++reassembled) { + assert(reassembled < NUM_PACKETS); + assert(odp_event_type(ev) == ODP_EVENT_PACKET); + dequeued_pkts[reassembled] = odp_packet_from_event(ev); + } + + /* Check reassembled packets against the originals */ + printf("\n= Checking reassembled packets...\n"); + for (i = 0; i < reassembled; ++i) { + int j = -1; + int k; + odp_packet_t packet = dequeued_pkts[i]; + uint32_t len = odp_packet_len(packet); + odph_ipv4hdr_t hdr; + odph_ipv4hdr_t reassembled_hdr; + + reassembled_hdr = *(odph_ipv4hdr_t *)odp_packet_data(packet); + for (k = 0; k < reassembled; ++k) { + hdr = *(odph_ipv4hdr_t *)odp_packet_data(orig_pkts[k]); + if (hdr.src_addr == reassembled_hdr.src_addr && + hdr.dst_addr == reassembled_hdr.dst_addr && + hdr.id == reassembled_hdr.id && + hdr.proto == reassembled_hdr.proto) { + assert(j < 0); + j = k; + } + } + assert(j >= 0); + + assert(odp_packet_is_valid(packet)); + assert(len == odp_packet_len(orig_pkts[j])); + assert(!packet_memcmp(orig_pkts[j], packet, 0, 0, len)); + } + printf("=== Successfully reassembled %d of %d packets\n", reassembled, + NUM_PACKETS); + assert(reassembled == NUM_PACKETS); + printf("\n= Complete!\n"); + + /* Free packets */ + for (i = 0; i < reassembled; ++i) + odp_packet_free(dequeued_pkts[i]); + for (i = 0; i < NUM_PACKETS; ++i) + odp_packet_free(orig_pkts[i]); + garbage_collect_fraglists(fraglists, FRAGLISTS, reassembled_pkts, 1); + + /* ODP cleanup and termination */ + assert(!odp_queue_destroy(fragments)); + assert(!odp_queue_destroy(reassembled_pkts)); + assert(!odp_shm_free(shm)); + if (odp_pool_destroy(fragment_pool)) { + fprintf(stderr, + "ERROR: fragment_pool destruction failed\n"); + return 1; + } + if (odp_term_local()) { + fprintf(stderr, "ERROR: odp_term_local\n"); + return 1; + } + if (odp_term_global(instance)) { + fprintf(stderr, "ERROR: odp_term_global\n"); + return 1; + } + + return 0; +} diff --git a/example/ipfragreass/odp_ipfragreass_atomics.h b/example/ipfragreass/odp_ipfragreass_atomics.h new file mode 100644 index 00000000..691978e5 --- /dev/null +++ b/example/ipfragreass/odp_ipfragreass_atomics.h @@ -0,0 +1,55 @@ +/* Copyright (c) 2017, Linaro Limited + * All rights reserved. + * + * SPDX-License-Identifier: BSD-3-Clause + */ + +#ifndef ODP_FRAGREASS_PP_ATOMICS_H_ +#define ODP_FRAGREASS_PP_ATOMICS_H_ + +#include <example_debug.h> + +#if __SIZEOF_POINTER__ == 4 +/** + * A wrapper function to perform a 64-bit "strong" atomic compare and swap + * (CAS) operation. This generic version of the function is used if an + * architecture-specific (e.g. lockless) variant is not specified. + * + * @param var The location at which the CAS should take place + * @param exp A pointer to the expected value + * @param neu A pointer to the new value to be written on success + * @param mo_success The memory order on success + * @param mo_failure The memory order on failure + * + * @return Whether the operation succeeded + */ +static inline bool atomic_strong_cas_dblptr(uint64_t *var, uint64_t *exp, + uint64_t neu, int mo_success, + int mo_failure) +{ + return __atomic_compare_exchange_n(var, exp, neu, 0, mo_success, + mo_failure); +} +#elif __SIZEOF_POINTER__ == 8 +/** + * A wrapper function to perform a 128-bit "strong" atomic compare and swap + * (CAS) operation. This generic version of the function is used if an + * architecture-specific (e.g. lockless) variant is not specified. + * + * @param var The location at which the CAS should take place + * @param exp A pointer to the expected value + * @param neu A pointer to the new value to be written on success + * @param mo_success The memory order on success + * @param mo_failure The memory order on failure + * + * @return Whether the operation succeeded + */ +static inline bool atomic_strong_cas_dblptr(__int128 *var, __int128 *exp, + __int128 neu, int mo_success, + int mo_failure) +{ + return __atomic_compare_exchange_n(var, exp, neu, 0, mo_success, + mo_failure); +} +#endif +#endif diff --git a/example/ipfragreass/odp_ipfragreass_atomics_arm.h b/example/ipfragreass/odp_ipfragreass_atomics_arm.h new file mode 100644 index 00000000..99c37a77 --- /dev/null +++ b/example/ipfragreass/odp_ipfragreass_atomics_arm.h @@ -0,0 +1,120 @@ +/* Copyright (c) 2017, Linaro Limited + * All rights reserved. + * + * SPDX-License-Identifier: BSD-3-Clause + */ + +#ifndef ODP_FRAGREASS_PP_ATOMICS_ARM_H_ +#define ODP_FRAGREASS_PP_ATOMICS_ARM_H_ + +#include <example_debug.h> + +#if __SIZEOF_POINTER__ == 8 && defined(__aarch64__) +static inline __int128 lld(__int128 *var, int mo) +{ + __int128 old; + + if (mo == __ATOMIC_ACQUIRE) + __asm__ volatile("ldaxp %0, %H0, [%1]" : "=&r" (old) + : "r" (var) : "memory"); + else /* mo == __ATOMIC_RELAXED */ + __asm__ volatile("ldxp %0, %H0, [%1]" : "=&r" (old) + : "r" (var) : ); + return old; +} + +static inline uint32_t scd(__int128 *var, __int128 neu, int mo) +{ + uint32_t ret; + + if (mo == __ATOMIC_RELEASE) + __asm__ volatile("stlxp %w0, %1, %H1, [%2]" : "=&r" (ret) + : "r" (neu), "r" (var) : "memory"); + else /* mo == __ATOMIC_RELAXED */ + __asm__ volatile("stxp %w0, %1, %H1, [%2]" : "=&r" (ret) + : "r" (neu), "r" (var) : ); + return ret; +} + +static inline bool atomic_strong_cas_dblptr(__int128 *var, __int128 *exp, + __int128 neu, int mo_success, + int mo_failure EXAMPLE_UNUSED) +{ + register __int128 old; + register __int128 expected = *exp; + int ll_mo, sc_mo; + + ll_mo = (mo_success != __ATOMIC_RELAXED && + mo_success != __ATOMIC_RELEASE) ? __ATOMIC_ACQUIRE + : __ATOMIC_RELAXED; + sc_mo = (mo_success == __ATOMIC_RELEASE || + mo_success == __ATOMIC_ACQ_REL || + mo_success == __ATOMIC_SEQ_CST) ? __ATOMIC_RELEASE + : __ATOMIC_RELAXED; + + /* + * To prevent spurious failures and ensure atomicity, we must write some + * value back -- whether it's the value we wanted to write, or the value + * that is currently there. Repeat until we perform a successful write. + */ + do { + old = lld(var, ll_mo); + } while (scd(var, old == expected ? neu : old, sc_mo)); + + *exp = old; + return (old == expected); +} +#elif __SIZEOF_POINTER__ == 4 && defined(__ARM_ARCH) && __ARM_ARCH == 7 +static inline uint64_t lld(uint64_t *var, int mo) +{ + uint64_t old; + + __asm__ volatile("ldrexd %0, %H0, [%1]" : "=&r" (old) : "r" (var) : ); + if (mo == __ATOMIC_ACQUIRE) + __asm__ volatile("dmb ish" ::: "memory"); + return old; +} + +static inline uint32_t scd(uint64_t *var, uint64_t neu, int mo) +{ + uint32_t ret; + + if (mo == __ATOMIC_RELEASE) + __asm__ volatile("dmb ish" ::: "memory"); + __asm__ volatile("strexd %0, %1, %H1, [%2]" : "=&r" (ret) + : "r" (neu), "r" (var) : ); + return ret; +} + +static inline bool atomic_strong_cas_dblptr(uint64_t *var, uint64_t *exp, + uint64_t neu, int mo_success, + int mo_failure EXAMPLE_UNUSED) +{ + register uint64_t old; + register uint64_t expected = *exp; + int ll_mo, sc_mo; + + ll_mo = (mo_success != __ATOMIC_RELAXED && + mo_success != __ATOMIC_RELEASE) ? __ATOMIC_ACQUIRE + : __ATOMIC_RELAXED; + sc_mo = (mo_success == __ATOMIC_RELEASE || + mo_success == __ATOMIC_ACQ_REL || + mo_success == __ATOMIC_SEQ_CST) ? __ATOMIC_RELEASE + : __ATOMIC_RELAXED; + + /* + * To prevent spurious failures and ensure atomicity, we must write some + * value back -- whether it's the value we wanted to write, or the value + * that is currently there. Repeat until we perform a successful write. + */ + do { + old = lld(var, ll_mo); + } while (scd(var, old == expected ? neu : old, sc_mo)); + + *exp = old; + return (old == expected); +} +#else +#include "odp_ipfragreass_atomics.h" +#endif +#endif diff --git a/example/ipfragreass/odp_ipfragreass_fragment.c b/example/ipfragreass/odp_ipfragreass_fragment.c new file mode 100644 index 00000000..b2fe6ad2 --- /dev/null +++ b/example/ipfragreass/odp_ipfragreass_fragment.c @@ -0,0 +1,99 @@ +/* Copyright (c) 2017, Linaro Limited + * All rights reserved. + * + * SPDX-License-Identifier: BSD-3-Clause + */ + +#include <stdio.h> +#include <assert.h> + +#include "odp_ipfragreass_ip.h" +#include "odp_ipfragreass_fragment.h" + +int fragment_ipv4_packet(odp_packet_t orig_packet, odp_packet_t *out, + int *out_len) +{ + int fragments = 0; + odp_bool_t first_fragment = 1; + odph_ipv4hdr_t *orig_header = odp_packet_data(orig_packet); + uint32_t header_len = ipv4hdr_ihl(*orig_header); + uint32_t orig_len = odp_packet_len(orig_packet); + uint32_t bytes_remaining = orig_len - header_len; + uint32_t frag_offset = 0; + odp_packet_t frag = orig_packet; + odp_packet_t rest = ODP_PACKET_INVALID; + + if (bytes_remaining <= MTU) + return -1; + + /* + * The main fragmentation loop (continue until all bytes from the + * original payload have been assigned to a fragment) + */ + while (bytes_remaining) { + odph_ipv4hdr_t *header; + uint32_t frag_len; + + frag_len = (bytes_remaining > MTU ? (MTU + header_len) + : (bytes_remaining + + header_len)); + bytes_remaining -= (frag_len - header_len); + if (bytes_remaining) { + uint32_t split_point; + int split_success; + + split_point = first_fragment ? (frag_len) + : (frag_len - header_len); + split_success = odp_packet_split(&frag, split_point, + &rest); + if (split_success < 0) { + fprintf(stderr, + "ERROR: odp_packet_split\n"); + return -1; + } else if (first_fragment && split_success > 0) { + orig_header = odp_packet_data(frag); + } + } + + /* Add a header to the packet if necessary */ + if (first_fragment) { + first_fragment = 0; + } else { + if (odp_packet_extend_head(&frag, header_len, NULL, + NULL) < 0) { + fprintf(stderr, + "ERROR: odp_packet_extend_head\n"); + return -1; + } + if (odp_packet_copy_from_mem(frag, 0, header_len, + orig_header)) { + fprintf(stderr, + "ERROR: odp_packet_copy_from_mem\n"); + return -1; + } + } + + /* Update the header */ + header = odp_packet_data(frag); + ipv4hdr_set_more_fragments(header, bytes_remaining != 0); + header->tot_len = odp_cpu_to_be_16(frag_len); + ipv4hdr_set_fragment_offset_oct(header, (frag_offset / 8)); + assert((bytes_remaining == 0) || + (odp_packet_len(frag) == (MTU + header_len))); + assert((bytes_remaining != 0) || + (ipv4hdr_reass_payload_len(*header) + header_len + == orig_len)); + odp_packet_l3_offset_set(frag, 0); + header->chksum = 0; + odph_ipv4_csum_update(frag); + + out[fragments++] = frag; + frag = rest; + frag_offset += (frag_len - header_len); + } + + if (out_len) + *out_len = fragments; + + return 0; +} diff --git a/example/ipfragreass/odp_ipfragreass_fragment.h b/example/ipfragreass/odp_ipfragreass_fragment.h new file mode 100644 index 00000000..f497fad7 --- /dev/null +++ b/example/ipfragreass/odp_ipfragreass_fragment.h @@ -0,0 +1,28 @@ +/* Copyright (c) 2017, Linaro Limited + * All rights reserved. + * + * SPDX-License-Identifier: BSD-3-Clause + */ + +#ifndef ODP_FRAGREASS_PP_FRAG_H_ +#define ODP_FRAGREASS_PP_FRAG_H_ + +#include <odp/helper/ip.h> + +#define MTU 1480 /**< IPv4 payload MTU */ + +ODP_STATIC_ASSERT(!(MTU % 8), "ODPFRAG_MTU__SIZE_ERROR"); + +/** + * Break apart a larger-than-MTU packet into smaller IPv4 fragments + * + * @param orig_packet The larger-than-MTU packet to fragment + * @param[out] out The packet buffer to write fragments out to + * @param[out] out_len The number of fragments produced + * + * @return 0 on success, -1 otherwise + */ +int fragment_ipv4_packet(odp_packet_t orig_packet, odp_packet_t *out, + int *out_len); + +#endif diff --git a/example/ipfragreass/odp_ipfragreass_helpers.c b/example/ipfragreass/odp_ipfragreass_helpers.c new file mode 100644 index 00000000..e8d89b97 --- /dev/null +++ b/example/ipfragreass/odp_ipfragreass_helpers.c @@ -0,0 +1,124 @@ +/* Copyright (c) 2017, Linaro Limited + * All rights reserved. + * + * SPDX-License-Identifier: BSD-3-Clause + */ + +#include <stdio.h> +#include <stdlib.h> + +#include <assert.h> +#include <arpa/inet.h> + +#include "odp_ipfragreass_helpers.h" +#include "odp_ipfragreass_ip.h" + +#define IP_SRC_ADDR 0xC0A80001 /* 192.168.0.1 */ +#define IP_DST_ADDR 0xC0A80002 /* 192.168.0.2 */ + +static void set_random_payload(odp_packet_t packet, uint32_t payload_offset, + uint32_t size) +{ + uint32_t i, j; + uint32_t bytes_remaining; + unsigned char *buffer; + unsigned char seed = rand() % (UINT8_MAX + 1); + + /* + * Set the payload to a run of consecutive numbers from a random seed. + * Not completely random as some structure is good for debugging. + */ + for (i = 0; i < size; ++i) { + buffer = odp_packet_offset(packet, payload_offset + i, + &bytes_remaining, NULL); + for (j = 0; i < size && j < bytes_remaining; ++j, ++i) + buffer[j] = (unsigned char)(seed + i); + } +} + +odp_packet_t pack_udp_ipv4_packet(odp_pool_t pool, odp_u16be_t ip_id, + uint32_t max_size, uint32_t min_size) +{ + odp_packet_t result; + unsigned char *buffer; + odph_ipv4hdr_t *header; + uint32_t size; + uint32_t header_len; + uint32_t header_len_words; + + size = (rand() % (max_size - min_size)) + min_size; + result = odp_packet_alloc(pool, size); + if (result == ODP_PACKET_INVALID) + return ODP_PACKET_INVALID; + odp_packet_ts_set(result, odp_time_global()); + + /* Set the IPv4 header */ + header_len_words = (rand() % (IP_IHL_MAX - IP_IHL_MIN + 1)); + header_len_words += IP_IHL_MIN; + header_len = WORDS_TO_BYTES(header_len_words); + assert(header_len >= IP_HDR_LEN_MIN && header_len <= IP_HDR_LEN_MAX); + odp_packet_l3_offset_set(result, 0); + odp_packet_has_ipv4_set(result, 1); + buffer = odp_packet_data(result); + header = (odph_ipv4hdr_t *)buffer; + ipv4hdr_init(header); + header->id = odp_cpu_to_be_16(ip_id); + header->proto = ODPH_IPPROTO_UDP; + header->src_addr = odp_cpu_to_be_32(IP_SRC_ADDR); + header->dst_addr = odp_cpu_to_be_32(IP_DST_ADDR); + ipv4hdr_set_ihl_words(header, header_len_words); + ipv4hdr_set_payload_len(header, size - header_len); + header->chksum = 0; + odph_ipv4_csum_update(result); + + /* Set the payload */ + size -= header_len; + set_random_payload(result, header_len, size); + + return result; +} + +void shuffle(odp_packet_t *packets, int num_packets) +{ + int i; + + if (num_packets <= 1) + return; + + for (i = 0; i < num_packets; ++i) { + odp_packet_t tmp = packets[i]; + int j = rand() % num_packets; + + packets[i] = packets[j]; + packets[j] = tmp; + } +} + +int packet_memcmp(odp_packet_t a, odp_packet_t b, uint32_t offset_a, + uint32_t offset_b, uint32_t length) +{ + uint32_t i = 0; + void *data_a, *data_b; + + while (i < length) { + int status; + uint32_t bytes_remaining_a; + uint32_t bytes_remaining_b; + uint32_t bytes_remaining; + + data_a = odp_packet_offset(a, offset_a + i, &bytes_remaining_a, + NULL); + data_b = odp_packet_offset(b, offset_b + i, &bytes_remaining_b, + NULL); + bytes_remaining = min(bytes_remaining_a, bytes_remaining_b); + bytes_remaining = min(bytes_remaining, length - i); + assert(bytes_remaining != 0 && data_a && data_b); + + status = memcmp(data_a, data_b, bytes_remaining); + if (status != 0) + return status; + i += bytes_remaining; + } + + return 0; +} diff --git a/example/ipfragreass/odp_ipfragreass_helpers.h b/example/ipfragreass/odp_ipfragreass_helpers.h new file mode 100644 index 00000000..ffb3bfb7 --- /dev/null +++ b/example/ipfragreass/odp_ipfragreass_helpers.h @@ -0,0 +1,79 @@ +/* Copyright (c) 2017, Linaro Limited + * All rights reserved. + * + * SPDX-License-Identifier: BSD-3-Clause + */ + +#ifndef ODP_FRAGREASS_PP_HELPERS_H_ +#define ODP_FRAGREASS_PP_HELPERS_H_ + +#include <odp/helper/ip.h> + +#if defined(__ARM_ARCH) +#include "odp_ipfragreass_atomics_arm.h" +#else +#include "odp_ipfragreass_atomics.h" +#endif + +/** + * Generate a random IPv4 UDP packet from the specified parameters + * + * @param pool The pool used to allocate the packet + * @param ip_id The IP ID of the packet to be generated + * @param max_size The maximum size of the generated packet + * @param min_size The minimum size of the generated packet + * + * @return A handle to the generated packet + */ +odp_packet_t pack_udp_ipv4_packet(odp_pool_t pool, odp_u16be_t ip_id, + uint32_t max_size, uint32_t min_size); + +/** + * Roughly perform a random shuffle on an array of packets + * + * @param packets A pointer to the packets to shuffle + * @param num_packets The number of packets in the array + */ +void shuffle(odp_packet_t *packets, int num_packets); + +/** + * Compare the contents of two packets + * + * @param a The first packet to compare + * @param b The second packet to compare + * @param offset_a The offset in the first packet to begin comparing at + * @param offset_b The offset in the second packet to begin comparing at + * @param length The number of bytes to compare + * + * @return Returns the same values as memcmp (0 if both packets are equal) + */ +int packet_memcmp(odp_packet_t a, odp_packet_t b, uint32_t offset_a, + uint32_t offset_b, uint32_t length); + +/** + * Get the smallest of two uint32_t values + * + * @param a The first value + * @param b The second value + * + * @return The smallest of the two input values + */ +static inline uint32_t min(uint32_t a, uint32_t b) +{ + return a < b ? a : b; +} + +/** + * Get the largest of two uint32_t values + * + * @param a The first value + * @param b The second value + * + * @return The largest of the two input values + */ +static inline uint32_t max(uint32_t a, uint32_t b) +{ + return a > b ? a : b; +} + +#endif diff --git a/example/ipfragreass/odp_ipfragreass_ip.h b/example/ipfragreass/odp_ipfragreass_ip.h new file mode 100644 index 00000000..e7281e5a --- /dev/null +++ b/example/ipfragreass/odp_ipfragreass_ip.h @@ -0,0 +1,251 @@ +/* Copyright (c) 2017, Linaro Limited + * All rights reserved. + * + * SPDX-License-Identifier: BSD-3-Clause + */ + +#ifndef ODP_FRAGREASS_PP_IP_H_ +#define ODP_FRAGREASS_PP_IP_H_ + +#include <odp/helper/ip.h> + +#define IP_HDR_LEN_MIN ODPH_IPV4HDR_LEN +#define IP_HDR_LEN_MAX 60 +#define IP_IHL_MIN ODPH_IPV4HDR_IHL_MIN +#define IP_IHL_MAX 15 +#define IP_OCTET_MAX ((1U << 14U) - 1U) /* 65535 bytes, 8192 octs, 14 bits */ +#define IP_FRAGOFF_MAX 8192 /* 13 bits */ + +#define IP_FRAG_RESV 0x8000 /**< "Reserved" fragment flag (must be zero) */ +#define IP_FRAG_DONT 0x4000 /**< "Don't Fragment" (DF) fragment flag */ +#define IP_FRAG_MORE 0x2000 /**< "More Fragments" (MF) fragment flag */ +#define IP_FRAG_MASK 0x1fff /**< "Fragment Offset" mask */ + +#define BYTES_TO_OCTS(bytes) (((bytes) + 7U) / 8U) +#define OCTS_TO_BYTES(bytes) ((bytes) * 8) +#define WORDS_TO_BYTES(words) ((words) * 4) + +/** + * Initialise an IPv4 header structure + * + * @param h A pointer to the header structure to initialise + */ +static inline void ipv4hdr_init(odph_ipv4hdr_t *h) +{ + h->ver_ihl = 0x45; /* IHL of 5, Version of 4 [0x45 = 0100 0101] */ + h->tos = 0; + h->tot_len = odp_cpu_to_be_16(IP_HDR_LEN_MIN); + h->frag_offset = 0; + h->ttl = UINT8_MAX; + h->chksum = 0; +} + +/** + * Get the Internet Header Length (IHL) of an IPv4 header in words + * + * @param h The header to get the IHL of + * + * @return The IHL of "h" in words + */ +static inline uint8_t ipv4hdr_ihl_words(odph_ipv4hdr_t h) +{ + return ODPH_IPV4HDR_IHL(h.ver_ihl); +} + +/** + * Get the Internet Header Length (IHL) of an IPv4 header in bytes + * + * @param h The header to get the IHL of + * + * @return The IHL of "h" in bytes + */ +static inline uint8_t ipv4hdr_ihl(odph_ipv4hdr_t h) +{ + return WORDS_TO_BYTES(ipv4hdr_ihl_words(h)); +} + +/** + * Set the Internet Header Length (IHL) of an IPv4 header in words + * + * @param h A pointer to the header to set the IHL of + * @param ihl The new IHL in words + */ +static inline void ipv4hdr_set_ihl_words(odph_ipv4hdr_t *h, uint8_t ihl) +{ + h->ver_ihl = (ODPH_IPV4HDR_VER(h->ver_ihl) << 4) + | ODPH_IPV4HDR_IHL(ihl); +} + +/** + * Check whether the "Don't Fragment" (DF) flag is set in an IPv4 header + * + * @param h The header to check the DF flag of + * + * @return Whether the DF flag is set + */ +static inline bool ipv4hdr_dont_fragment(odph_ipv4hdr_t h) +{ + return (h.frag_offset & odp_cpu_to_be_16(IP_FRAG_DONT)); +} + +/** + * Set the "Don't Fragment" (DF) flag of an IPv4 header + * + * @param h A pointer to the header to set the DF of + * @param df Whether the DF flag should be set or unset + */ +static inline void ipv4hdr_set_dont_fragment(odph_ipv4hdr_t *h, bool df) +{ + if (df) + h->frag_offset |= odp_cpu_to_be_16(IP_FRAG_DONT); + else + h->frag_offset &= ~odp_cpu_to_be_16(IP_FRAG_DONT); +} + +/** + * Check whether the "More Fragments" (MF) flag is set in an IPv4 header + * + * @param h The header to check the MF flag of + * + * @return Whether the MF flag is set + */ +static inline bool ipv4hdr_more_fragments(odph_ipv4hdr_t h) +{ + return (h.frag_offset & odp_cpu_to_be_16(IP_FRAG_MORE)); +} + +/** + * Set the "More Fragments" (MF) flag of an IPv4 header + * + * @param h A pointer to the header to set the MF of + * @param mf Whether the MF flag should be set or unset + */ +static inline void ipv4hdr_set_more_fragments(odph_ipv4hdr_t *h, bool mf) +{ + if (mf) + h->frag_offset |= odp_cpu_to_be_16(IP_FRAG_MORE); + else + h->frag_offset &= ~odp_cpu_to_be_16(IP_FRAG_MORE); +} + +/** + * Check whether an IPv4 header represents an IPv4 fragment or not + * + * @param h The header to check + * + * @return Whether "h" is the header of an IPv4 fragment or not + */ +static inline bool ipv4hdr_is_fragment(odph_ipv4hdr_t h) +{ + return (h.frag_offset & odp_cpu_to_be_16(IP_FRAG_MORE | IP_FRAG_MASK)); +} + +/** + * Get the fragment offset of an IPv4 header in octets + * + * @param h The header to get the fragment offset of + * + * @return The fragment offset of "h" in octets + */ +static inline uint16_t ipv4hdr_fragment_offset_oct(odph_ipv4hdr_t h) +{ + return (odp_be_to_cpu_16(h.frag_offset) & IP_FRAG_MASK); +} + +/** + * Set the fragment offset of an IPv4 header in octets + * + * @param h A pointer to the header to set the fragment offset of + * @param offset The new fragment offset in octets + */ +static inline void ipv4hdr_set_fragment_offset_oct(odph_ipv4hdr_t *h, + uint16_t offset) +{ + h->frag_offset = odp_cpu_to_be_16((odp_be_to_cpu_16(h->frag_offset) + & ~IP_FRAG_MASK) | offset); +} + +/** + * Get the payload length of an IPv4 header in bytes + * + * @param h The header to get the payload length of + * + * @return The payload length of "h" in bytes + */ +static inline uint16_t ipv4hdr_payload_len(odph_ipv4hdr_t h) +{ + uint32_t packet_len = odp_be_to_cpu_16(h.tot_len); + uint32_t header_size = ipv4hdr_ihl(h); + + return (uint16_t)(packet_len >= header_size ? (packet_len - header_size) + : 0); +} + +/** + * Get the payload length of an IPv4 header in octets + * + * @param h The header to get the payload length of + * + * @return The payload length of "h" in octets + */ +static inline uint16_t ipv4hdr_payload_len_oct(odph_ipv4hdr_t h) +{ + return BYTES_TO_OCTS(ipv4hdr_payload_len(h)); +} + +/** + * Set the payload length of an IPv4 header in bytes + * + * @param h A pointer to the header to set the payload length of + * @param len The new payload length in bytes + */ +static inline void ipv4hdr_set_payload_len(odph_ipv4hdr_t *h, uint16_t len) +{ + h->tot_len = odp_cpu_to_be_16(len + ipv4hdr_ihl(*h)); +} + +/** + * Get the upper bound length in bytes of the reassembled payload that this + * fragment contributes to from an IPv4 header + * + * If this is called with the final fragment in a series, the exact length of + * the reassembled payload is computed. Otherwise, UINT16_MAX is returned. + * + * @param h The header to get the reassembled payload length from + * + * @return The upper bound for the reassembled payload length of "h" in bytes + */ +static inline uint16_t ipv4hdr_reass_payload_len(odph_ipv4hdr_t h) +{ + uint16_t fragment_offset_oct; + + if (ipv4hdr_more_fragments(h)) + return UINT16_MAX; + + fragment_offset_oct = ipv4hdr_fragment_offset_oct(h); + return OCTS_TO_BYTES(fragment_offset_oct) + ipv4hdr_payload_len(h); +} + +/** + * Get the upper bound length in octets of the reassembled payload that this + * fragment contributes to from an IPv4 header + * + * If this is called with the final fragment in a series, the exact length of + * the reassembled payload is computed. Otherwise, IP_OCTET_MAX is returned. + * + * @param h The header to estimate the reassembled payload length from + * + * @return The estimate for the reassembled payload length of "h" in octets + */ +static inline uint16_t ipv4hdr_reass_payload_len_oct(odph_ipv4hdr_t h) +{ + uint16_t fragment_offset_oct; + + if (ipv4hdr_more_fragments(h)) + return IP_OCTET_MAX; + + fragment_offset_oct = ipv4hdr_fragment_offset_oct(h); + return fragment_offset_oct + ipv4hdr_payload_len_oct(h); +} + +#endif diff --git a/example/ipfragreass/odp_ipfragreass_reassemble.c b/example/ipfragreass/odp_ipfragreass_reassemble.c new file mode 100644 index 00000000..fd1618f5 --- /dev/null +++ b/example/ipfragreass/odp_ipfragreass_reassemble.c @@ -0,0 +1,771 @@ +/* Copyright (c) 2017, Linaro Limited + * All rights reserved. + * + * SPDX-License-Identifier: BSD-3-Clause + */ + +#include <stdio.h> +#include <stdlib.h> +#include <assert.h> + +#include "odp_ipfragreass_reassemble.h" +#include "odp_ipfragreass_helpers.h" + +#define ROT(x, k) (((x) << (k)) | ((x) >> (32 - (k)))) + +/** + * Check whether two packets have the same flow (src/dst/id/proto) + * + * @param current The first packet to compare + * @param frag The second packet to compare + * + * @return true if the flows match, false otherwise + */ +static inline odp_bool_t equal_flow(struct packet *current, struct packet *frag) +{ + odph_ipv4hdr_t *curr_h = odp_packet_data(current->handle); + odph_ipv4hdr_t *frag_h = odp_packet_data(frag->handle); + + return (memcmp(&curr_h->src_addr, &frag_h->src_addr, + sizeof(curr_h->src_addr) + sizeof(curr_h->dst_addr)) + == 0 && + curr_h->id == frag_h->id && + curr_h->proto == frag_h->proto); +} + +/** + * Check whether one packet has a "later" flow than another + * + * @param current The first packet to compare + * @param frag The second packet to compare + * + * @return true if the first packet's flow is later, false otherwise + */ +static inline odp_bool_t later_flow(struct packet *current, struct packet *frag) +{ + odph_ipv4hdr_t *curr_h = odp_packet_data(current->handle); + odph_ipv4hdr_t *frag_h = odp_packet_data(frag->handle); + + return (memcmp(&curr_h->src_addr, &frag_h->src_addr, + sizeof(curr_h->src_addr) + sizeof(curr_h->dst_addr)) + > 0 || + curr_h->id > frag_h->id || + curr_h->proto > frag_h->proto); +} + +/** + * Find the earliest of two fraglist timestamps, considering a "now" timestamp + * + * @param a The first timestamp + * @param b The second timestamp + * @param now A timestamp indication of the time "now" + * + * @return The earliest of the first and second timestamps + */ +static inline struct flts earliest(struct flts a, struct flts b, + struct flts now) +{ + struct flts result; + struct flts elapsed_a; + struct flts elapsed_b; + + now.t += TS_NOW_TOLERANCE; + elapsed_a.t = now.t - a.t; + elapsed_b.t = now.t - b.t; + result.t = now.t - max(elapsed_a.t, elapsed_b.t); + return result; +} + +/** + * Hash the flow information within an IPv4 header + * + * @param hdr A pointer to the header to hash + * + * @return A hash of the src/dst/id/proto information in the header + */ +static inline uint32_t hash(odph_ipv4hdr_t *hdr) +{ + uint32_t a = hdr->src_addr; + uint32_t b = hdr->dst_addr; + uint32_t c = hdr->id << 16 | hdr->proto; + + /* A degenerate 3x32-bit Jenkins hash */ + c ^= b; + c -= ROT(b, 14); + a ^= c; + a -= ROT(c, 11); + b ^= a; + b -= ROT(a, 25); + c ^= b; + c -= ROT(b, 16); + a ^= c; + a -= ROT(c, 4); + b ^= a; + b -= ROT(a, 14); + c ^= b; + c -= ROT(b, 24); + return c; +} + +/** + * Check whether one fragment is "later" than another, considering a timestamp + * of "now" + * + * This definition of "later" relies firstly on flow, then on the endpoint of + * the fragment, then on the fragment offset, then on arrival time. + * + * @param a The first fragment to compare + * @param b The second fragment to compare + * @param now A timestamp indication of the time "now" + * + * @return true if the first fragment is "later", false otherwise + */ +static inline odp_bool_t later_fragment(struct packet *a, struct packet *b, + struct flts now) +{ + odph_ipv4hdr_t hdr_a = *(odph_ipv4hdr_t *)odp_packet_data(a->handle); + odph_ipv4hdr_t hdr_b = *(odph_ipv4hdr_t *)odp_packet_data(b->handle); + uint32_t offset_a = ipv4hdr_fragment_offset_oct(hdr_a); + uint32_t offset_b = ipv4hdr_fragment_offset_oct(hdr_b); + uint32_t payload_len_a = ipv4hdr_payload_len(hdr_a); + uint32_t payload_len_b = ipv4hdr_payload_len(hdr_b); + uint32_t endpoint_a = OCTS_TO_BYTES(offset_a) + payload_len_a; + uint32_t endpoint_b = OCTS_TO_BYTES(offset_b) + payload_len_b; + + if (later_flow(a, b)) { + return 1; + } else if (equal_flow(a, b)) { + if (endpoint_a > endpoint_b) { + return 1; + } else if (endpoint_a == endpoint_b) { + if (offset_a < offset_b) { + return 1; + } else if (offset_a == offset_b) { + return b->arrival.t == earliest(a->arrival, + b->arrival, + now).t; + } + } + } + + return 0; +} + +/** + * Attempt to extract a whole packet from a list of sorted fragments + * + * If a complete packet is formed, its tail pointer is returned, and the tail + * pointer for the remaining packets is written out to remaining_packets. + * + * @param[in] tail The tail of the list of fragments to parse + * @param[out] remaining_packets The pointer to any remaining packets + * @param[out] made_changes Whether any changes were made to the fragments + * + * @return The tail pointer of any reassembled packet, or NULL otherwise + */ +static struct packet *extract_complete_packet(struct packet *tail, + struct packet **remaining_packets, + int *made_changes) +{ + /* + * Iterate through the flows in this fragment list (until a packet is + * reassembled). + */ + while (tail) { + /* + * Work backwards through the fragments in a single flow, + * attempting to glue together a whole packet. Upon finding a + * discontinuity, break out of the loop for this flow and try + * the next (if there is one). + */ + struct packet *current = tail; + odph_ipv4hdr_t tail_hdr; + uint16_t final_frag_offset; + uint16_t expected_frag_offset; + + tail_hdr = *(odph_ipv4hdr_t *)odp_packet_data(tail->handle); + final_frag_offset = ipv4hdr_fragment_offset_oct(tail_hdr); + expected_frag_offset = final_frag_offset; + while (current) { + odph_ipv4hdr_t curr_hdr; + uint16_t curr_offset_oct; + odp_bool_t final_fragment; + struct packet *prev; + void *tmp; + odph_ipv4hdr_t prev_hdr; + uint16_t prev_off_oct; + uint16_t prev_oct; + + tmp = odp_packet_data(current->handle); + curr_hdr = *(odph_ipv4hdr_t *)tmp; + curr_offset_oct = ipv4hdr_fragment_offset_oct(curr_hdr); + final_fragment = (curr_offset_oct == final_frag_offset); + + /* + * If the final fragment in the chain has "More + * Fragments" set, it's not the final packet of the + * datagram as a whole. + */ + if (final_fragment && ipv4hdr_more_fragments(curr_hdr)) + break; + + /* + * If this is the first fragment in a chain, we may have + * completed the reassembly of a whole packet. + */ + prev = prev_packet(*current); + if (prev == NULL || !equal_flow(current, prev)) { + if (curr_offset_oct) + break; + + /* + * Extract the complete packet from the list of + * remaining packets + */ + if (remaining_packets) + *remaining_packets = prev; + set_prev_packet(current, NULL); + if (made_changes) + *made_changes = 1; + return tail; + } + + /* + * Fragments should be consistent with those previously + * processed + */ + tmp = odp_packet_data(prev->handle); + prev_hdr = *(odph_ipv4hdr_t *)tmp; + prev_off_oct = ipv4hdr_fragment_offset_oct(prev_hdr); + prev_oct = BYTES_TO_OCTS(ipv4hdr_payload_len(prev_hdr)); + if (curr_offset_oct != prev_off_oct + prev_oct) { + if (prev_off_oct + prev_oct < curr_offset_oct) { + /* + * If there's no overlap, this is just a + * regular discontinuity + */ + break; + } + + /* + * Fragment duplication or overlap has occurred! + * We don't handle such occurrences in this + * simple example application. + */ + assert(0); + break; + } + + expected_frag_offset -= prev_oct; + current = prev; + } + + /* + * Since we haven't had any luck finding a whole packet within + * this flow, let's try to look at other flows in this fraglist + * (if there are any others). + */ + if (!current) { + tail = NULL; + break; + } + + while (prev_packet(*current) && + equal_flow(current, prev_packet(*current))) { + current = prev_packet(*current); + } + tail = prev_packet(*current); + remaining_packets = prev_packet_ptr(current); + } + + return NULL; +} + +/* + * Glue together a list of fragments sorted by fragment offset, writing the + * result to an output queue + * + * @param tail The tail pointer to the list of fragments to reassemble + * @param out The output queue to write the result to + * + * @return 0 on success, -1 otherwise + */ +static int send_packet(struct packet *tail, odp_queue_t out) +{ + struct packet result = *tail; + struct packet *current = prev_packet(result); + odph_ipv4hdr_t *header; + uint32_t length; + + /* + * Reassemble the complete packet (working backwards from the last + * fragment) + */ + while (current && equal_flow(current, &result)) { + struct packet new_result = *current; + int concat_success; + + current = prev_packet(new_result); + header = odp_packet_data(result.handle); + odp_packet_pull_head(result.handle, ipv4hdr_ihl(*header)); + concat_success = odp_packet_concat(&new_result.handle, + result.handle); + if (concat_success < 0) { + fprintf(stderr, "ERROR: odp_packet_concat\n"); + return -1; + } + result = new_result; + } + + /* Fix the header */ + header = odp_packet_data(result.handle); + length = odp_packet_len(result.handle); + assert(length >= IP_HDR_LEN_MIN || length <= UINT16_MAX); + header->tot_len = odp_cpu_to_be_16(length); + ipv4hdr_set_more_fragments(header, 0); + ipv4hdr_set_fragment_offset_oct(header, 0); + header->chksum = 0; + odph_ipv4_csum_update(result.handle); + + assert(odp_queue_enq(out, odp_packet_to_event(result.handle)) >= 0); + return 0; +} + +/** + * Sort a fraglist using the "later_fragment" function + * + * @param fl A pointer to the fraglist to sort + * @param now A timestamp indication of the time "now" + */ +static void sort_fraglist(union fraglist *fl, struct flts now) +{ + struct packet *to_insert = fl->tail; + + fl->tail = NULL; + while (to_insert) { + struct packet *to_insert_next = prev_packet(*to_insert); + + if (fl->tail == NULL || + later_fragment(to_insert, fl->tail, now)) { + set_prev_packet(to_insert, fl->tail); + fl->tail = to_insert; + } else { + struct packet *current = fl->tail; + + while (prev_packet(*current) && + later_fragment(prev_packet(*current), to_insert, + now)) { + current = prev_packet(*current); + } + set_prev_packet(to_insert, prev_packet(*current)); + set_prev_packet(current, to_insert); + } + to_insert = to_insert_next; + } +} + +/** + * Add a thread local fraglist to a shared fraglist + * + * @param fl Pointer to the shared fraglist to add "frags" to + * @param frags The thread local fraglist to add to "fl" + * @param frags_head Pointer to the head fragment of "frags" + * @param now A timestamp indication of the time "now" + * @param out The queue to which reassembled packets should be written + * @param dont_assemble Whether reassembly should be attempted by default + * + * @return The number of packets reassembled and sent to the output queue + */ +static int add_fraglist_to_fraglist(union fraglist *fl, union fraglist frags, + struct packet *frags_head, struct flts now, + odp_queue_t out, odp_bool_t dont_assemble) +{ + int reassembled = 0; + + /* + * We may need to recursively call this function a number of times, + * keeping count of the total number of packets reassembled. Sadly, + * tail call optimisation doesn't seem to be working very well, so + * we're using good ol' fashioned GOTOs instead. + */ +redo:; + union fraglist oldfl; + union fraglist newfl; + union fraglist nullfl; + struct flts oldfl_earliest; + struct flts frags_earliest; + + __atomic_load(&fl->half[0], &oldfl.half[0], __ATOMIC_RELAXED); + __atomic_load(&fl->half[1], &oldfl.half[1], __ATOMIC_RELAXED); + + /* + * If we're updating a non-empty fraglist, we should always attempt + * reassembly! + */ + if (oldfl.tail != NULL) + dont_assemble = 0; + + /* Insert the new fragment(s) to the tail of the fraglist */ + set_prev_packet(frags_head, oldfl.tail); + newfl.tail = frags.tail; + + /* + * Update the fraglist variables (accumulating the length of the + * received pieces into "part_len", and updating the perceived 'true' + * length of the whole packet along with the timestamp of the earliest + * fragment in this list). + */ + oldfl_earliest.t = oldfl.earliest; + frags_earliest.t = frags.earliest; + newfl.part_len = min(IP_OCTET_MAX, oldfl.part_len + frags.part_len); + newfl.whole_len = min(oldfl.whole_len, frags.whole_len); + newfl.earliest = (oldfl.tail == NULL ? frags.earliest + : earliest(oldfl_earliest, + frags_earliest, + now).t); + + /* + * Check if it looks like we have all the fragments for a whole packet + * yet. If not, just write out our changes and move on. + */ + if (newfl.part_len < newfl.whole_len || dont_assemble) { + if (!atomic_strong_cas_dblptr(&fl->raw, &oldfl.raw, newfl.raw, + __ATOMIC_RELEASE, + __ATOMIC_RELAXED)) { + /* Failed to add this fragment? Try again. */ + set_prev_packet(frags_head, NULL); + goto redo; + } + return reassembled; + } + + /* + * It /appears/ that we have all the fragments for a packet. Things are + * not always as they appear, however, particularly in the case of a + * hash map collision where part_len and whole_len may be incorrect + * (and, hence, must be verified). + * + * Take exclusive ownership over this fraglist while we attempt + * reassembly. If we're truly done with it, then this releases the slot, + * otherwise we'll update the slot with our changes later. + */ + init_fraglist(&nullfl); + if (!atomic_strong_cas_dblptr(&fl->raw, &oldfl.raw, nullfl.raw, + __ATOMIC_ACQUIRE, __ATOMIC_RELAXED)) { + /* Failed to take this fraglist? Try again. */ + set_prev_packet(frags_head, NULL); + goto redo; + } + + /* + * Find any complete packets within the fraglist, cut them out of the + * list, and send them to the output queue. Note that there may be + * several complete packets, as we may have added multiple new fragments + * into the fraglist. + */ + struct packet *remaining_packets; + struct packet *complete_datagram; + int fraglist_changed = 0; + int call_changed_fraglist = 0; + union fraglist update; + + sort_fraglist(&newfl, now); + remaining_packets = newfl.tail; + dont_assemble = 1; + while ((complete_datagram = + extract_complete_packet(remaining_packets, &remaining_packets, + &call_changed_fraglist)) || + call_changed_fraglist) { + fraglist_changed = 1; + if (complete_datagram) { + assert(!send_packet(complete_datagram, out)); + ++reassembled; + dont_assemble = 0; + } + call_changed_fraglist = 0; + } + + /* No remaining fragments in this fraglist? We're done. */ + if (!remaining_packets) + return reassembled; + + /* + * If there are still fragments in this fraglist, we have changes to + * write back. + * + * Note that we may have to reassemble more packets, as adding the + * fragments this thread has exclusive access to into the shared + * fraglist may entail new packets being completed. Thus, we have to + * repeat this whole add_fraglist_to_fraglist process with the remaining + * fragments. + */ + update.tail = remaining_packets; + update.part_len = newfl.part_len; + update.whole_len = newfl.whole_len; + update.earliest = newfl.earliest; + + /* + * We've cut fragments from the fragment list chain, and so should + * recalculate the part_len, whole_len, and earliest variables before + * writing out our changes. + */ + if (fraglist_changed) { + struct packet *current = remaining_packets; + + update.earliest = now.t; + while (current) { + odph_ipv4hdr_t *h; + uint16_t part_oct; + uint16_t whole_oct; + struct flts update_earliest; + + update_earliest.t = update.earliest; + h = odp_packet_data(current->handle); + part_oct = ipv4hdr_payload_len_oct(*h); + whole_oct = ipv4hdr_reass_payload_len_oct(*h); + update.part_len = min(IP_OCTET_MAX, + update.part_len + part_oct); + update.whole_len = min(update.whole_len, whole_oct); + update.earliest = earliest(update_earliest, + current->arrival, now).t; + frags_head = current; + current = prev_packet(*current); + } + frags = update; + goto redo; + } + + frags = update; + frags_head = frags.tail; + while (prev_packet(*frags_head)) + frags_head = prev_packet(*frags_head); + goto redo; +} + +/** + * Add a single fragment to a shared fraglist + * + * @param fl Pointer to the shared fraglist to add "frag" to + * @param frag Pointer to the fragment to add to "fl" + * @param frag_payload_len The payload length of "frag" in octets + * @param frag_reass_payload_len The estimated reassembled payload length of + * "frag" in octets + * @param out The queue to which reassembled packets should be written + * + * @return The number of packets reassembled and sent to the output + */ +static int add_frag_to_fraglist(union fraglist *fl, struct packet *frag, + uint16_t frag_payload_len, + uint16_t frag_reass_payload_len, + odp_queue_t out) +{ + union fraglist frags; + + frags.tail = frag; + frags.part_len = frag_payload_len; + frags.whole_len = frag_reass_payload_len; + frags.earliest = frag->arrival.t; + + return add_fraglist_to_fraglist(fl, frags, frags.tail, frag->arrival, + out, 0); +} + +/** + * Remove the stale flows from a shared fraglist this thread has exclusive + * access over + * + * @param fl Pointer to the shared fraglist to clean stale flows from + * @param oldfl The value of the fraglist before we took exclusive access + * @param time_now A timestamp indication of the time "now" + * @param out The queue to which reassembled packets should be written + * @param force Whether all flows in the fraglist should be considered stale + */ +static void remove_stale_flows(union fraglist *fl, union fraglist oldfl, + struct flts timestamp_now, odp_queue_t out, + odp_bool_t force) +{ + union fraglist newfl = oldfl; + struct packet *current; + struct packet *current_tail; + struct packet *remaining_frags_head; + struct flts flow_earliest; + + /* + * Sort the fraglist so we can step through its fragments flow-by-flow + */ + sort_fraglist(&newfl, timestamp_now); + + /* Remove stale flows from the fraglist */ + current = newfl.tail; + current_tail = newfl.tail; + remaining_frags_head = NULL; + flow_earliest = current->arrival; + newfl.earliest = timestamp_now.t; + while (current) { + struct packet *prev = prev_packet(*current); + + /* + * If this is the first fragment in a chain, we can make the + * decision on whether this flow should be kept or discarded + */ + if (prev == NULL || !equal_flow(current, prev)) { + struct flts elapsed; + uint64_t elapsed_ns; + + elapsed.t = timestamp_now.t - flow_earliest.t; + elapsed_ns = (elapsed.t * TS_RES_NS); + if ((elapsed_ns >= FLOW_TIMEOUT_NS && + elapsed.t + TS_NOW_TOLERANCE >= + TS_NOW_TOLERANCE) || force) { + struct packet *to_free = current_tail; + + while (to_free != prev) { + struct packet *next; + + next = prev_packet(*to_free); + odp_packet_free(to_free->handle); + to_free = next; + } + + if (remaining_frags_head) + set_prev_packet(remaining_frags_head, + prev); + else + newfl.tail = prev; + } else { + odph_ipv4hdr_t *h; + uint16_t part_oct; + uint16_t whole_oct; + struct flts newfl_earliest; + + newfl_earliest.t = newfl.earliest; + remaining_frags_head = current; + h = odp_packet_data(current->handle); + part_oct = ipv4hdr_payload_len_oct(*h); + whole_oct = ipv4hdr_reass_payload_len_oct(*h); + newfl.part_len = min(IP_OCTET_MAX, + newfl.part_len + part_oct); + newfl.whole_len = min(newfl.whole_len, + whole_oct); + newfl.earliest = earliest(newfl_earliest, + flow_earliest, + timestamp_now).t; + } + + current_tail = prev; + flow_earliest.t = EARLIEST_MAX; + } else { + flow_earliest = earliest(flow_earliest, + current->arrival, + timestamp_now); + } + + current = prev; + } + + /* + * If there are any remaining fragments, write them back into the + * fraglist + */ + if (remaining_frags_head) + add_fraglist_to_fraglist(fl, newfl, remaining_frags_head, + timestamp_now, out, 0); +} + +/** + * Clean up any stale flows within a fraglist + * + * @param fl Pointer to the shared fraglist to clean stale flows from + * @param out The queue to which reassembled packets should be written + * @param force Whether all flows in the fraglist should be considered stale + */ +static void garbage_collect_fraglist(union fraglist *fl, odp_queue_t out, + odp_bool_t force) +{ + uint64_t time_now; + struct flts timestamp_now; + struct flts elapsed; + uint64_t elapsed_ns; + union fraglist oldfl; + odp_bool_t success = 1; + + do { + time_now = odp_time_to_ns(odp_time_global()); + timestamp_now.t = time_now / TS_RES_NS; + __atomic_load(&fl->half[0], &oldfl.half[0], __ATOMIC_RELAXED); + __atomic_load(&fl->half[1], &oldfl.half[1], __ATOMIC_RELAXED); + elapsed.t = timestamp_now.t - oldfl.earliest; + + if (oldfl.tail == NULL || + elapsed.t + TS_NOW_TOLERANCE < TS_NOW_TOLERANCE) + return; + + elapsed_ns = (elapsed.t * TS_RES_NS); + assert(force || elapsed_ns <= 86400000000000); + if (elapsed_ns >= FLOW_TIMEOUT_NS || force) { + union fraglist nullfl; + + init_fraglist(&nullfl); + success = atomic_strong_cas_dblptr(&fl->raw, &oldfl.raw, + nullfl.raw, + __ATOMIC_ACQUIRE, + __ATOMIC_RELAXED); + + if (success) + remove_stale_flows(fl, oldfl, timestamp_now, + out, force); + } + } while (!success); +} + +int reassemble_ipv4_packets(union fraglist *fraglists, int num_fraglists, + struct packet *fragments, int num_fragments, + odp_queue_t out) +{ + int i; + int packets_reassembled = 0; + + for (i = 0; i < num_fragments; ++i) { + struct packet frag; + odph_ipv4hdr_t *hdr; + uint16_t frag_payload_len; + uint16_t frag_reass_payload_len; + uint32_t key; + union fraglist *fl; + int status; + + frag = fragments[i]; + hdr = odp_packet_data(frag.handle); + frag_payload_len = ipv4hdr_payload_len_oct(*hdr); + frag_reass_payload_len = ipv4hdr_reass_payload_len_oct(*hdr); + + /* + * Find the appropriate hash map bucket for fragments in this + * flow. In the case of collisions, fragments for multiple flows + * are simply stored in the same list. + */ + key = hash(hdr); + fl = &fraglists[key % num_fraglists]; + + status = add_frag_to_fraglist(fl, &fragments[i], + frag_payload_len, + frag_reass_payload_len, out); + if (status < 0) { + fprintf(stderr, + "ERROR: failed to add fragment to fraglist\n"); + return -1; + } + packets_reassembled += status; + } + + return packets_reassembled; +} + +void garbage_collect_fraglists(union fraglist *fraglists, int num_fraglists, + odp_queue_t out, odp_bool_t destroy_all) +{ + int i; + + for (i = 0; i < num_fraglists; ++i) + garbage_collect_fraglist(&fraglists[i], out, destroy_all); +} diff --git a/example/ipfragreass/odp_ipfragreass_reassemble.h b/example/ipfragreass/odp_ipfragreass_reassemble.h new file mode 100644 index 00000000..ffc4bfb5 --- /dev/null +++ b/example/ipfragreass/odp_ipfragreass_reassemble.h @@ -0,0 +1,211 @@ +/* Copyright (c) 2017, Linaro Limited + * All rights reserved. + * + * SPDX-License-Identifier: BSD-3-Clause + */ + +#ifndef ODP_FRAGREASS_PP_REASSEMBLE_H_ +#define ODP_FRAGREASS_PP_REASSEMBLE_H_ + +#include <odp/helper/ip.h> + +#include "odp_ipfragreass_ip.h" +#include "odp_ipfragreass_helpers.h" + +ODP_STATIC_ASSERT((__SIZEOF_POINTER__ == 4 || __SIZEOF_POINTER__ == 8), + "ODPREASS_PTR__SIZE_ERROR"); + +/** + * The time in nanoseconds after reception of the earliest fragment that a + * flow of traffic is considered to be stale + */ +#define FLOW_TIMEOUT_NS 15000000000ULL + +/** Convert nanoseconds into a unit for packet.arrival */ +#if __SIZEOF_POINTER__ == 4 +#define TS_RES_NS ((uint64_t)5000000000) /**< ns -> 5s */ +#elif __SIZEOF_POINTER__ == 8 +#define TS_RES_NS ((uint64_t)1000000) /**< ns -> 1ms */ +#endif + +/** + * The maximum value of the packet.arrival field. + */ +#if __SIZEOF_POINTER__ == 4 +#define EARLIEST_MAX 15 +#elif __SIZEOF_POINTER__ == 8 +#define EARLIEST_MAX UINT32_MAX +#endif + +/** + * The time in packet.arrival ticks that indications of the time "now" are + * permitted to be off by. + */ +#if __SIZEOF_POINTER__ == 4 +#define TS_NOW_TOLERANCE 1 +#elif __SIZEOF_POINTER__ == 8 +#define TS_NOW_TOLERANCE 5000 +#endif + +/** + * The timestamp format used for fragments. Sadly, this has to be a structure + * as we may need a bit field. + */ +struct flts { +#if __SIZEOF_POINTER__ == 4 + uint8_t t:4; +#elif __SIZEOF_POINTER__ == 8 + uint32_t t; +#endif +}; + +/** + * Metadata for reassembly, to be stored alongside each fragment + */ +struct packet { + odp_packet_t handle; /**< The ODP packet handle for this fragment */ + struct packet *prev; /**< Pointer to the fragment "before" this one */ + struct flts arrival; /**< Arrival timestamp for this fragment */ +}; + +/** + * A list of IP fragments associated with one or more traffic flows, along with + * some related data + * + * This is used as an atomically-updated hash map bucket in reassembly, and is + * assumed to be packed with no padding. + */ +union fraglist { + struct { + /** + * The timestamp of the earliest arriving fragment in this + * fraglist + */ +#if __SIZEOF_POINTER__ == 4 + uint32_t earliest:4; +#elif __SIZEOF_POINTER__ == 8 + uint32_t earliest; +#endif + + /** + * The sum of the payloads of the fragments in this list + * + * That is, the size of reassembling all the fragments in this + * list into one big packet (minus the header). + */ + uint32_t part_len:14; + + /** + * The smallest reassembled payload length upper bound from + * all fragments in this list + * + * This is the threshold over which, given the right + * circumstances, "part_len" might indicate that we are able + * to reassemble a packet. + */ + uint32_t whole_len:14; + + /** + * The tail of a "reverse" linked list of fragments + * + * Each fragment element has a "prev" pointer to the element + * before it in the list. When used in a multi-threaded + * environment, new elements should be inserted atomically by + * modifying this tail pointer. + */ + struct packet *tail; + }; + +#if __SIZEOF_POINTER__ == 4 + struct { + uint32_t half[2]; + }; + uint64_t raw; +#elif __SIZEOF_POINTER__ == 8 + struct { + uint64_t half[2]; + }; + __int128 raw; +#endif +}; + +/** + * Initialise a fraglist structure + * + * @param fl The fraglist to initialise + */ +static inline void init_fraglist(union fraglist *fl) +{ + fl->earliest = EARLIEST_MAX; + fl->part_len = 0; + fl->whole_len = IP_OCTET_MAX; + fl->tail = NULL; +} + +/** + * Get the packet "before" a packet in a linked list + * + * @param packet The packet from which the previous should be located + * + * @return A pointer to the packet before the input packet (can be NULL) + */ +static inline struct packet *prev_packet(struct packet packet) +{ + return packet.prev; +} + +/** + * Get the address of the pointer to the packet "before" a packet in a linked + * list + * + * @param packet The packet for which the previous pointer should be located + * + * @return A pointer to the "prev" packet pointer of the input packet + */ +static inline struct packet **prev_packet_ptr(struct packet *packet) +{ + return &packet->prev; +} + +/** + * Set the packet "before" a packet in a linked list + * + * @param packet The packet to set the previous packet from + * @param prev The packet to set as being before "packet" + */ +static inline void set_prev_packet(struct packet *packet, struct packet *prev) +{ + packet->prev = prev; +} + +/** + * Attempt packet reassembly with the aid of a number of new fragments + * + * Add "num_fragments" fragments to a fraglist hash map (with "num_fraglist" + * entries), attempting reassembly and writing any successfully reassembled + * packets out to the "out" queue. + * + * @param fraglists The hash map structure to add the fragments to + * @param num_fraglists The number of entries in the hash map + * @param fragments Pointer to the fragments to add + * @param num_fragments The number of fragments to add + * @param out The queue to which reassembled packets should be written + * + * @return The number of packets successfully reassembled and written to "out" + */ +int reassemble_ipv4_packets(union fraglist *fraglists, int num_fraglists, + struct packet *fragments, int num_fragments, + odp_queue_t out); + +/** + * Clean up any stale flows within a fraglist hash map + * + * @param fraglists The hash map structure to clean flows from + * @param num_fraglists The number of entries in the hash map + * @param out The queue to which reassembled packets should be written + * @param destroy_all Whether all encountered flows should be cleaned up + */ +void garbage_collect_fraglists(union fraglist *fraglists, int num_fraglists, + odp_queue_t out, odp_bool_t destroy_all); + +#endif diff --git a/example/m4/configure.m4 b/example/m4/configure.m4 index 620db04f..18218d05 100644 --- a/example/m4/configure.m4 +++ b/example/m4/configure.m4 @@ -14,6 +14,7 @@ AC_CONFIG_FILES([example/classifier/Makefile example/generator/Makefile example/hello/Makefile example/ipsec/Makefile + example/ipfragreass/Makefile example/l2fwd_simple/Makefile example/l3fwd/Makefile example/packet/Makefile commit cd1f187d33d65cb54781b5a3c0864229b19c11c3 Author: Joyce Kong <joyce.kong(a)arm.com> Date: Wed Jun 7 15:19:08 2017 +0800 linux-gen: improve conversion between buf_hdr_t and packet_t Signed-off-by: Joyce Kong <joyce.kong(a)arm.com> Reviewed-and-tested-by: Bill Fischofer <bill.fischofer(a)linaro.org> Reviewed-by: Brian Brooks <brian.brooks(a)arm.com> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/platform/linux-generic/include/odp_packet_internal.h b/platform/linux-generic/include/odp_packet_internal.h index a480a748..cdd139bb 100644 --- a/platform/linux-generic/include/odp_packet_internal.h +++ b/platform/linux-generic/include/odp_packet_internal.h @@ -148,6 +148,16 @@ static inline odp_packet_t packet_handle(odp_packet_hdr_t *pkt_hdr) return (odp_packet_t)pkt_hdr; } +static inline odp_buffer_hdr_t *packet_to_buf_hdr(odp_packet_t pkt) +{ + return &odp_packet_hdr(pkt)->buf_hdr; +} + +static inline odp_packet_t packet_from_buf_hdr(odp_buffer_hdr_t *buf_hdr) +{ + return (odp_packet_t)(odp_packet_hdr_t *)buf_hdr; +} + static inline void copy_packet_parser_metadata(odp_packet_hdr_t *src_hdr, odp_packet_hdr_t *dst_hdr) { diff --git a/platform/linux-generic/odp_packet_io.c b/platform/linux-generic/odp_packet_io.c index 877978ba..e5436a90 100644 --- a/platform/linux-generic/odp_packet_io.c +++ b/platform/linux-generic/odp_packet_io.c @@ -554,7 +554,6 @@ static inline int pktin_recv_buf(odp_pktin_queue_t queue, odp_packet_t packets[num]; odp_packet_hdr_t *pkt_hdr; odp_buffer_hdr_t *buf_hdr; - odp_buffer_t buf; int i; int pkts; int num_rx = 0; @@ -564,8 +563,7 @@ static inline int pktin_recv_buf(odp_pktin_queue_t queue, for (i = 0; i < pkts; i++) { pkt = packets[i]; pkt_hdr = odp_packet_hdr(pkt); - buf = _odp_packet_to_buffer(pkt); - buf_hdr = buf_hdl_to_hdr(buf); + buf_hdr = packet_to_buf_hdr(pkt); if (pkt_hdr->p.input_flags.dst_queue) { queue_entry_t *dst_queue; @@ -584,7 +582,7 @@ static inline int pktin_recv_buf(odp_pktin_queue_t queue, int pktout_enqueue(queue_entry_t *qentry, odp_buffer_hdr_t *buf_hdr) { - odp_packet_t pkt = _odp_packet_from_buffer(buf_hdr->handle.handle); + odp_packet_t pkt = packet_from_buf_hdr(buf_hdr); int len = 1; int nbr; @@ -614,7 +612,7 @@ int pktout_enq_multi(queue_entry_t *qentry, odp_buffer_hdr_t *buf_hdr[], return nbr; for (i = 0; i < num; ++i) - pkt_tbl[i] = _odp_packet_from_buffer(buf_hdr[i]->handle.handle); + pkt_tbl[i] = packet_from_buf_hdr(buf_hdr[i]); nbr = odp_pktout_send(qentry->s.pktout, pkt_tbl, num); return nbr; diff --git a/platform/linux-generic/pktio/loop.c b/platform/linux-generic/pktio/loop.c index e9ad22ba..cc636eba 100644 --- a/platform/linux-generic/pktio/loop.c +++ b/platform/linux-generic/pktio/loop.c @@ -81,7 +81,7 @@ static int loopback_recv(pktio_entry_t *pktio_entry, int index ODP_UNUSED, for (i = 0; i < nbr; i++) { uint32_t pkt_len; - pkt = _odp_packet_from_buffer(odp_hdr_to_buf(hdr_tbl[i])); + pkt = packet_from_buf_hdr(hdr_tbl[i]); pkt_len = odp_packet_len(pkt); @@ -163,7 +163,7 @@ static int loopback_send(pktio_entry_t *pktio_entry, int index ODP_UNUSED, len = QUEUE_MULTI_MAX; for (i = 0; i < len; ++i) { - hdr_tbl[i] = buf_hdl_to_hdr(_odp_packet_to_buffer(pkt_tbl[i])); + hdr_tbl[i] = packet_to_buf_hdr(pkt_tbl[i]); bytes += odp_packet_len(pkt_tbl[i]); } ----------------------------------------------------------------------- Summary of changes: doc/application-api-guide/examples.dox | 5 + example/Makefile.am | 1 + example/ipfragreass/.gitignore | 3 + example/ipfragreass/Makefile.am | 23 + example/ipfragreass/odp_ipfragreass.c | 376 ++++++++++ example/ipfragreass/odp_ipfragreass_atomics.h | 55 ++ example/ipfragreass/odp_ipfragreass_atomics_arm.h | 120 ++++ example/ipfragreass/odp_ipfragreass_fragment.c | 99 +++ example/ipfragreass/odp_ipfragreass_fragment.h | 28 + example/ipfragreass/odp_ipfragreass_helpers.c | 124 ++++ example/ipfragreass/odp_ipfragreass_helpers.h | 79 +++ example/ipfragreass/odp_ipfragreass_ip.h | 251 +++++++ example/ipfragreass/odp_ipfragreass_reassemble.c | 771 +++++++++++++++++++++ example/ipfragreass/odp_ipfragreass_reassemble.h | 211 ++++++ example/m4/configure.m4 | 1 + .../linux-generic/include/odp_packet_internal.h | 10 + platform/linux-generic/odp_packet_io.c | 8 +- platform/linux-generic/pktio/loop.c | 4 +- 18 files changed, 2162 insertions(+), 7 deletions(-) create mode 100644 example/ipfragreass/.gitignore create mode 100644 example/ipfragreass/Makefile.am create mode 100644 example/ipfragreass/odp_ipfragreass.c create mode 100644 example/ipfragreass/odp_ipfragreass_atomics.h create mode 100644 example/ipfragreass/odp_ipfragreass_atomics_arm.h create mode 100644 example/ipfragreass/odp_ipfragreass_fragment.c create mode 100644 example/ipfragreass/odp_ipfragreass_fragment.h create mode 100644 example/ipfragreass/odp_ipfragreass_helpers.c create mode 100644 example/ipfragreass/odp_ipfragreass_helpers.h create mode 100644 example/ipfragreass/odp_ipfragreass_ip.h create mode 100644 example/ipfragreass/odp_ipfragreass_reassemble.c create mode 100644 example/ipfragreass/odp_ipfragreass_reassemble.h hooks/post-receive --

6 years, 10 months

1
0
0 0

2024

2023

2022

2021

2020

2019

2018

2017

2016

git-lng-odp June 2017