git-lng-odp

git-lng-odp@lists.linaro.org

2 participants
1206 discussions

[COMMIT] branch master updated. v1.19.0.2_tigermoth-218-g45cd09238

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 45cd0923809b2b4b972c9e9843ca7323183380c7 (commit) via 8de012860878e3cb70aab7b6151223b201a8d9a1 (commit) via 52d54fae53f5e65f232cb21b116e51ea36b4eb0b (commit) from eab91f7101cdf47c59b14b5a511a82c701e66ff4 (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 45cd0923809b2b4b972c9e9843ca7323183380c7 Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Wed Oct 24 15:54:53 2018 +0300 linux-gen: sched: increase max spread weight Increase max spread weight. The default value is kept the same. Signed-off-by: Petri Savolainen <petri.savolainen(a)linaro.org> Reviewed-by: Bill Fischofer <bill.fischofer(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/platform/linux-generic/odp_schedule_basic.c b/platform/linux-generic/odp_schedule_basic.c index 456704a08..7208200aa 100644 --- a/platform/linux-generic/odp_schedule_basic.c +++ b/platform/linux-generic/odp_schedule_basic.c @@ -58,7 +58,7 @@ ODP_STATIC_ASSERT((ODP_SCHED_PRIO_NORMAL > 0) && /* A thread polls a non preferred sched queue every this many polls * of the prefer queue. */ -#define MAX_PREFER_WEIGHT 63 +#define MAX_PREFER_WEIGHT 127 #define MIN_PREFER_WEIGHT 1 #define MAX_PREFER_RATIO (MAX_PREFER_WEIGHT + 1) commit 8de012860878e3cb70aab7b6151223b201a8d9a1 Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Wed Oct 24 15:45:38 2018 +0300 linux-gen: sched: use spread weight from config file Use the new config file option instead of fixed prefer ratio. Signed-off-by: Petri Savolainen <petri.savolainen(a)linaro.org> Reviewed-by: Bill Fischofer <bill.fischofer(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/platform/linux-generic/odp_schedule_basic.c b/platform/linux-generic/odp_schedule_basic.c index 214d42c8b..456704a08 100644 --- a/platform/linux-generic/odp_schedule_basic.c +++ b/platform/linux-generic/odp_schedule_basic.c @@ -333,7 +333,7 @@ static inline uint8_t prio_spread_index(uint32_t index) static void sched_local_init(void) { int i; - uint8_t spread; + uint8_t spread, prefer_ratio; uint8_t num_spread = sched->config.num_spread; uint8_t offset = 1; @@ -344,11 +344,12 @@ static void sched_local_init(void) sched_local.stash.queue = ODP_QUEUE_INVALID; spread = prio_spread_index(sched_local.thr); + prefer_ratio = sched->config.prefer_ratio; for (i = 0; i < SPREAD_TBL_SIZE; i++) { sched_local.spread_tbl[i] = spread; - if (num_spread > 1 && (i % MAX_PREFER_RATIO) == 0) { + if (num_spread > 1 && (i % prefer_ratio) == 0) { sched_local.spread_tbl[i] = prio_spread_index(spread + offset); offset++; @@ -362,6 +363,7 @@ static int schedule_init_global(void) { odp_shm_t shm; int i, j, grp; + int prefer_ratio; ODP_DBG("Schedule init ... "); @@ -382,8 +384,10 @@ static int schedule_init_global(void) return -1; } + prefer_ratio = sched->config.prefer_ratio; + /* When num_spread == 1, only spread_tbl[0] is used. */ - sched->max_spread = (sched->config.num_spread - 1) * MAX_PREFER_RATIO; + sched->max_spread = (sched->config.num_spread - 1) * prefer_ratio; sched->shm = shm; odp_spinlock_init(&sched->mask_lock); commit 52d54fae53f5e65f232cb21b116e51ea36b4eb0b Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Wed Oct 24 15:20:21 2018 +0300 linux-gen: sched: add spread weight config file option Add new config file option to control scheduler internal queue preference ratio. Signed-off-by: Petri Savolainen <petri.savolainen(a)linaro.org> Reviewed-by: Bill Fischofer <bill.fischofer(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/config/odp-linux-generic.conf b/config/odp-linux-generic.conf index af651d7f6..2417d23f2 100644 --- a/config/odp-linux-generic.conf +++ b/config/odp-linux-generic.conf @@ -16,7 +16,7 @@ # Mandatory fields odp_implementation = "linux-generic" -config_file_version = "0.1.1" +config_file_version = "0.1.2" # Shared memory options shm: { @@ -80,13 +80,27 @@ queue_basic: { } sched_basic: { - # Priority level spread. Each priority level is spread into multiple - # scheduler internal queues. A higher spread value typically improves - # parallelism and thus is better for high thread counts, but causes - # uneven service level for low thread counts. Typically, optimal - # value is the number of threads using the scheduler. + # Priority level spread + # + # Each priority level is spread into multiple scheduler internal queues. + # This value defines the number of those queues. Minimum value is 1. + # Each thread prefers one of the queues over other queues. A higher + # spread value typically improves parallelism and thus is better for + # high thread counts, but causes uneven service level for low thread + # counts. Typically, optimal value is the number of threads using + # the scheduler. prio_spread = 4 + # Weight of the preferred scheduler internal queue + # + # Each thread prefers one of the internal queues over other queues. + # This value controls how many times the preferred queue is polled + # between a poll to another internal queue. Minimum value is 1. A higher + # value typically improves parallelism as threads work mostly on their + # preferred queues, but causes uneven service level for low thread + # counts as non-preferred queues are served less often + prio_spread_weight = 63 + # Burst size configuration per priority. The first array element # represents the highest queue priority. The scheduler tries to get # burst_size_default[prio] events from a queue and stashes those that diff --git a/platform/linux-generic/odp_schedule_basic.c b/platform/linux-generic/odp_schedule_basic.c index 58396293b..214d42c8b 100644 --- a/platform/linux-generic/odp_schedule_basic.c +++ b/platform/linux-generic/odp_schedule_basic.c @@ -58,10 +58,12 @@ ODP_STATIC_ASSERT((ODP_SCHED_PRIO_NORMAL > 0) && /* A thread polls a non preferred sched queue every this many polls * of the prefer queue. */ -#define PREFER_RATIO 64 +#define MAX_PREFER_WEIGHT 63 +#define MIN_PREFER_WEIGHT 1 +#define MAX_PREFER_RATIO (MAX_PREFER_WEIGHT + 1) /* Spread weight table */ -#define SPREAD_TBL_SIZE ((MAX_SPREAD - 1) * PREFER_RATIO) +#define SPREAD_TBL_SIZE ((MAX_SPREAD - 1) * MAX_PREFER_RATIO) /* Maximum number of packet IO interfaces */ #define NUM_PKTIO ODP_CONFIG_PKTIO_ENTRIES @@ -182,6 +184,7 @@ typedef struct { uint8_t burst_default[NUM_PRIO]; uint8_t burst_max[NUM_PRIO]; uint8_t num_spread; + uint8_t prefer_ratio; } config; uint16_t max_spread; @@ -256,13 +259,29 @@ static int read_config_file(sched_global_t *sched) } if (val > MAX_SPREAD || val < MIN_SPREAD) { - ODP_ERR("Bad value %s = %u\n", str, val); + ODP_ERR("Bad value %s = %u [min: %u, max: %u]\n", str, val, + MIN_SPREAD, MAX_SPREAD); return -1; } sched->config.num_spread = val; ODP_PRINT(" %s: %i\n", str, val); + str = "sched_basic.prio_spread_weight"; + if (!_odp_libconfig_lookup_int(str, &val)) { + ODP_ERR("Config option '%s' not found.\n", str); + return -1; + } + + if (val > MAX_PREFER_WEIGHT || val < MIN_PREFER_WEIGHT) { + ODP_ERR("Bad value %s = %u [min: %u, max: %u]\n", str, val, + MIN_PREFER_WEIGHT, MAX_PREFER_WEIGHT); + return -1; + } + + sched->config.prefer_ratio = val + 1; + ODP_PRINT(" %s: %i\n", str, val); + str = "sched_basic.burst_size_default"; if (_odp_libconfig_lookup_array(str, burst_val, NUM_PRIO) != NUM_PRIO) { @@ -329,7 +348,7 @@ static void sched_local_init(void) for (i = 0; i < SPREAD_TBL_SIZE; i++) { sched_local.spread_tbl[i] = spread; - if (num_spread > 1 && (i % PREFER_RATIO) == 0) { + if (num_spread > 1 && (i % MAX_PREFER_RATIO) == 0) { sched_local.spread_tbl[i] = prio_spread_index(spread + offset); offset++; @@ -364,7 +383,7 @@ static int schedule_init_global(void) } /* When num_spread == 1, only spread_tbl[0] is used. */ - sched->max_spread = (sched->config.num_spread - 1) * PREFER_RATIO; + sched->max_spread = (sched->config.num_spread - 1) * MAX_PREFER_RATIO; sched->shm = shm; odp_spinlock_init(&sched->mask_lock); diff --git a/platform/linux-generic/test/process-mode.conf b/platform/linux-generic/test/process-mode.conf index d80df25c4..7a06544ab 100644 --- a/platform/linux-generic/test/process-mode.conf +++ b/platform/linux-generic/test/process-mode.conf @@ -1,6 +1,6 @@ # Mandatory fields odp_implementation = "linux-generic" -config_file_version = "0.1.1" +config_file_version = "0.1.2" # Shared memory options shm: { ----------------------------------------------------------------------- Summary of changes: config/odp-linux-generic.conf | 26 +++++++++++++++----- platform/linux-generic/odp_schedule_basic.c | 35 ++++++++++++++++++++++----- platform/linux-generic/test/process-mode.conf | 2 +- 3 files changed, 50 insertions(+), 13 deletions(-) hooks/post-receive --

7 years, 3 months

[COMMIT] branch master updated. v1.19.0.2_tigermoth-215-geab91f710

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 eab91f7101cdf47c59b14b5a511a82c701e66ff4 (commit) from cc91c95216ca80d17c638c07659873ecf24f25e6 (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 eab91f7101cdf47c59b14b5a511a82c701e66ff4 Author: Maxim Uvarov <maxim.uvarov(a)linaro.org> Date: Sun Oct 21 10:19:41 2018 +0300 remove scripts/build-pktio-dpdk Script was introduced to quick build odp with dpdk pktio support for dev or testing propose. Now it's more easy to run docker container with same parameters as CI does. Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> Signed-off-by: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov(a)linaro.org> Reviewed-by: Bill Fischofer <bill.fischofer(a)linaro.org> Reviewed-and-tested-by: Matias Elo <matias.elo(a)nokia.com> diff --git a/DEPENDENCIES b/DEPENDENCIES index 6b345b9c3..abe6d7411 100644 --- a/DEPENDENCIES +++ b/DEPENDENCIES @@ -166,23 +166,42 @@ Prerequisites for building the OpenDataPlane (ODP) API Use DPDK for ODP packet I/O. + Note: only packet I/O is accelerated with DPDK. Use + https://github.com/Linaro/odp-dpdk.git + for fully accelerated odp dpdk platform. + +3.4.1 DPDK pktio requirements + DPDK pktio adds a depency to NUMA library. # Debian/Ubuntu $ sudo apt-get install libnuma-dev # CentOS/RedHat/Fedora $ sudo yum install numactl-devel - Note: only packet I/O is accelerated with DPDK. Use - https://github.com/Linaro/odp-dpdk.git - for fully accelerated odp dpdk platform. +3.4.2 Native DPDK install + # Debian/Ubuntu starting from 18.04 + $ sudo apt-get install dpdk-dev + +3.4.2 Built DPDK from src + git clone --branch=17.11 http://dpdk.org/git/dpdk-stable dpdk + + #Make and edit DPDK configuration + TARGET="x86_64-native-linuxapp-gcc" + make config T=${TARGET} O=${TARGET} + pushd ${TARGET} -3.4.1 Building DPDK and ODP with DPDK pktio support + #To use I/O without DPDK supported NIC's enable pcap pmd: + sed -ri 's,(CONFIG_RTE_LIBRTE_PMD_PCAP=).*,\1y,' .config + popd - DPDK packet I/O has been tested to work with DPDK v17.11. + #Build DPDK + make build O=${TARGET} EXTRA_CFLAGS="-fPIC" + make install O=${TARGET} DESTDIR=${TARGET} - Follow steps in ./scripts/build-pktio-dpdk + #compile ODP + ./configure --with-dpdk-path=`pwd`/dpdk/${TARGET}/usr/local -3.4.2 Setup system +3.4.3 Setup system # Load DPDK modules $ sudo /sbin/modprobe uio @@ -194,13 +213,13 @@ Prerequisites for building the OpenDataPlane (ODP) API 512 x 2MB hugepages. All this can be done with the DPDK setup script (<dpdk-dir>/tools/dpdk-setup.sh). -3.4.3 Running ODP with DPDK pktio +3.4.4 Running ODP with DPDK pktio ODP applications will try use DPDK for packet I/O by default. If some other I/O type is desired instead, DPDK I/O can be disabled by setting the environment variable ODP_PKTIO_DISABLE_DPDK. - DPDK interfaces are accessed using indices. For example, two first DPDK + DPDK interfaces are accessed using indexes. For example, two first DPDK interfaces can be used with the odp_l2fwd example as follows: $ cd <odp_dir> $ sudo ./test/performance/odp_l2fwd -i 0,1 -c 2 -m 0 diff --git a/scripts/build-pktio-dpdk b/scripts/build-pktio-dpdk deleted file mode 100755 index b0c0a4d0e..000000000 --- a/scripts/build-pktio-dpdk +++ /dev/null @@ -1,40 +0,0 @@ -#!/bin/bash - -TARGET=${TARGET:-"x86_64-native-linuxapp-gcc"} - -export ROOT_DIR=$(readlink -e $(dirname $0) | sed 's|/scripts||') -pushd ${ROOT_DIR} - -echo '#include "pcap.h"' | cpp -H -o /dev/null 2>&1 -if [ "$?" != "0" ]; then - echo "Error: pcap is not installed. You may need to install libpcap-dev" -fi - -echo '#include "numa.h"' | cpp -H -o /dev/null 2>&1 -if [ "$?" != "0" ]; then - echo "Error: NUMA library is not installed. You need to install libnuma-dev" - exit 1 -fi - -git -c advice.detachedHead=false clone -q --depth=1 --single-branch --branch=17.11 http://dpdk.org/git/dpdk-stable dpdk -pushd dpdk -git log --oneline --decorate - -#Make and edit DPDK configuration -make config T=${TARGET} O=${TARGET} -pushd ${TARGET} -#To use I/O without DPDK supported NIC's enable pcap pmd: -sed -ri 's,(CONFIG_RTE_LIBRTE_PMD_PCAP=).*,\1y,' .config -popd - -#Build DPDK -make build O=${TARGET} EXTRA_CFLAGS="-fPIC" -make install O=${TARGET} DESTDIR=${TARGET} -popd - -#Build ODP -./bootstrap; -./configure --enable-test-vald --enable-test-perf --enable-test-cpp \ - --enable-debug --enable-debug-print \ - --with-dpdk-path=`pwd`/dpdk/${TARGET}/usr/local -make ----------------------------------------------------------------------- Summary of changes: DEPENDENCIES | 37 ++++++++++++++++++++++++++++--------- scripts/build-pktio-dpdk | 40 ---------------------------------------- 2 files changed, 28 insertions(+), 49 deletions(-) delete mode 100755 scripts/build-pktio-dpdk hooks/post-receive --

7 years, 3 months

[COMMIT] branch master updated. v1.19.0.2_tigermoth-214-gcc91c9521

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 cc91c95216ca80d17c638c07659873ecf24f25e6 (commit) via 4aa40267ae25e6ee021c8ce067bae649db2d0d8d (commit) via 83e9a6c3896d2d8a949a7bece565bae7dfd1d967 (commit) from 8dc48817b06a5a303cf2b778eda68cc3627dd625 (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 cc91c95216ca80d17c638c07659873ecf24f25e6 Author: Matias Elo <matias.elo(a)nokia.com> Date: Wed Oct 17 15:34:53 2018 +0300 linux-gen: timer: run inline timers during queue dequeue operations Process inline timers (if enabled) during odp_queue_deq() and odp_queue_deq_multi() calls. This fixes inline timers not working if odp_schedule() or odp_schedule_multi() was never called. Signed-off-by: Matias Elo <matias.elo(a)nokia.com> Reviewed-by: Bill Fischofer <bill.fischofer(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/platform/linux-generic/odp_queue_basic.c b/platform/linux-generic/odp_queue_basic.c index 73a5536d8..4e23fb53b 100644 --- a/platform/linux-generic/odp_queue_basic.c +++ b/platform/linux-generic/odp_queue_basic.c @@ -16,6 +16,7 @@ #include <odp_pool_internal.h> #include <odp_init_internal.h> #include <odp_shm_internal.h> +#include <odp_timer_internal.h> #include <odp/api/shared_memory.h> #include <odp/api/schedule.h> #include <odp_schedule_if.h> @@ -28,6 +29,7 @@ #include <odp/api/traffic_mngr.h> #include <odp_libconfig_internal.h> #include <odp/api/plat/queue_inline_types.h> +#include <odp_global_data.h> #define NUM_INTERNAL_QUEUES 64 @@ -976,6 +978,10 @@ static int queue_api_deq_multi(odp_queue_t handle, odp_event_t ev[], int num) { queue_entry_t *queue = qentry_from_handle(handle); + if (odp_global_rw->inline_timers && + odp_atomic_load_u64(&queue->s.num_timers)) + timer_run(); + if (num > QUEUE_MULTI_MAX) num = QUEUE_MULTI_MAX; @@ -987,6 +993,10 @@ static odp_event_t queue_api_deq(odp_queue_t handle) { queue_entry_t *queue = qentry_from_handle(handle); + if (odp_global_rw->inline_timers && + odp_atomic_load_u64(&queue->s.num_timers)) + timer_run(); + return (odp_event_t)queue->s.dequeue(handle); } diff --git a/platform/linux-generic/odp_queue_scalable.c b/platform/linux-generic/odp_queue_scalable.c index 6ec7f14b7..3acfc0841 100644 --- a/platform/linux-generic/odp_queue_scalable.c +++ b/platform/linux-generic/odp_queue_scalable.c @@ -24,9 +24,11 @@ #include <odp_pool_internal.h> #include <odp_queue_scalable_internal.h> #include <odp_schedule_if.h> +#include <odp_timer_internal.h> #include <odp_ishm_internal.h> #include <odp_ishmpool_internal.h> #include <odp/api/plat/queue_inline_types.h> +#include <odp_global_data.h> #include <string.h> #include <inttypes.h> @@ -845,6 +847,11 @@ static int queue_deq_multi(odp_queue_t handle, odp_event_t ev[], int num) num = QUEUE_MULTI_MAX; queue = qentry_from_ext(handle); + + if (odp_global_rw->inline_timers && + odp_atomic_load_u64(&queue->s.num_timers)) + timer_run(); + return queue->s.dequeue_multi(handle, (odp_buffer_hdr_t **)ev, num); } @@ -853,6 +860,11 @@ static odp_event_t queue_deq(odp_queue_t handle) queue_entry_t *queue; queue = qentry_from_ext(handle); + + if (odp_global_rw->inline_timers && + odp_atomic_load_u64(&queue->s.num_timers)) + timer_run(); + return (odp_event_t)queue->s.dequeue(handle); } diff --git a/platform/linux-generic/odp_timer.c b/platform/linux-generic/odp_timer.c index 507226580..a22617559 100644 --- a/platform/linux-generic/odp_timer.c +++ b/platform/linux-generic/odp_timer.c @@ -55,6 +55,7 @@ #include <odp/api/time.h> #include <odp/api/plat/time_inlines.h> #include <odp/api/timer.h> +#include <odp_queue_if.h> #include <odp_timer_internal.h> #include <odp/api/plat/queue_inlines.h> #include <odp_global_data.h> @@ -442,6 +443,8 @@ static inline odp_timer_t timer_alloc(timer_pool_t *tp, tp->num_alloc, _ODP_MEMMODEL_RLS); hdl = tp_idx_to_handle(tp, idx); + /* Add timer to queue */ + queue_fn->timer_add(queue); } else { __odp_errno = ENFILE; /* Reusing file table overflow */ hdl = ODP_TIMER_INVALID; @@ -461,6 +464,9 @@ static inline odp_buffer_t timer_free(timer_pool_t *tp, uint32_t idx) * grab any timeout buffer */ odp_buffer_t old_buf = timer_set_unused(tp, idx); + /* Remove timer from queue */ + queue_fn->timer_rem(tim->queue); + /* Destroy timer */ timer_fini(tim, &tp->tick_buf[idx]); commit 4aa40267ae25e6ee021c8ce067bae649db2d0d8d Author: Matias Elo <matias.elo(a)nokia.com> Date: Wed Oct 17 13:29:42 2018 +0300 linux-gen: timer: reduce inline timer overhead Reduce inline timer overhead by not polling timers until at least one timer pool has been created. Signed-off-by: Matias Elo <matias.elo(a)nokia.com> Reviewed-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_global_data.h b/platform/linux-generic/include/odp_global_data.h index f1be1502c..a12d12973 100644 --- a/platform/linux-generic/include/odp_global_data.h +++ b/platform/linux-generic/include/odp_global_data.h @@ -40,6 +40,7 @@ typedef struct { /* Read-only global data. Members should not be modified after global init * to enable process more support. */ struct odp_global_data_ro_t { + odp_init_t init_param; /* directory for odp mmaped files */ char *shm_dir; /* overload default with env */ @@ -69,6 +70,7 @@ struct odp_global_data_rw_t { pthread_t inotify_thread; int inotify_pcapng_is_running; odp_bool_t dpdk_initialized; + odp_bool_t inline_timers; }; extern struct odp_global_data_ro_t odp_global_ro; diff --git a/platform/linux-generic/include/odp_timer_internal.h b/platform/linux-generic/include/odp_timer_internal.h index 8dda9e6ff..02ba92e0d 100644 --- a/platform/linux-generic/include/odp_timer_internal.h +++ b/platform/linux-generic/include/odp_timer_internal.h @@ -19,6 +19,7 @@ #include <odp_buffer_internal.h> #include <odp_pool_internal.h> #include <odp/api/timer.h> +#include <odp_global_data.h> /* Minimum number of scheduling rounds between checking timer pools. */ #define CONFIG_TIMER_RUN_RATELIMIT_ROUNDS 1 @@ -38,22 +39,12 @@ typedef struct { odp_timer_t timer; } odp_timeout_hdr_t; -/* - * Whether to run timer pool processing 'inline' (on worker cores) or in - * background threads (thread-per-timerpool). - * - * If the application will use both scheduler and timer this flag is set - * to true, otherwise false. This application conveys this information via - * the 'not_used' bits in odp_init_t which are passed to odp_global_init(). - */ -extern odp_bool_t inline_timers; - unsigned _timer_run(void); /* Static inline wrapper to minimize modification of schedulers. */ static inline unsigned timer_run(void) { - return inline_timers ? _timer_run() : 0; + return odp_global_rw->inline_timers ? _timer_run() : 0; } #endif diff --git a/platform/linux-generic/odp_init.c b/platform/linux-generic/odp_init.c index dca06641a..0f7e0df28 100644 --- a/platform/linux-generic/odp_init.c +++ b/platform/linux-generic/odp_init.c @@ -268,6 +268,8 @@ int odp_init_global(odp_instance_t *instance, odp_global_ro.abort_fn = odp_override_abort; if (params != NULL) { + odp_global_ro.init_param = *params; + if (params->log_fn != NULL) odp_global_ro.log_fn = params->log_fn; if (params->abort_fn != NULL) diff --git a/platform/linux-generic/odp_timer.c b/platform/linux-generic/odp_timer.c index eeb863530..507226580 100644 --- a/platform/linux-generic/odp_timer.c +++ b/platform/linux-generic/odp_timer.c @@ -68,8 +68,6 @@ * for checking the freshness of received timeouts */ #define TMO_INACTIVE ((uint64_t)0x8000000000000000) -odp_bool_t inline_timers = false; - /****************************************************************************** * Mutual exclusion in the absence of CAS16 *****************************************************************************/ @@ -333,8 +331,26 @@ static odp_timer_pool_t timer_pool_new(const char *name, odp_spinlock_init(&tp->lock); odp_ticketlock_lock(&timer_global->lock); timer_global->timer_pool[tp_idx] = tp; + + if (timer_global->num_timer_pools == 1) { + odp_bool_t inline_tim; + + /* + * Whether to run timer pool processing 'inline' (on worker + * cores) or in background threads (thread-per-timerpool). + * + * If the application will use scheduler this flag is set to + * true, otherwise false. This application conveys this + * information via the 'not_used' bits in odp_init_t which are + * passed to odp_global_init(). + */ + inline_tim = !odp_global_ro.init_param.not_used.feat.schedule; + + odp_global_rw->inline_timers = inline_tim; + } + odp_ticketlock_unlock(&timer_global->lock); - if (!inline_timers) { + if (!odp_global_rw->inline_timers) { if (tp->param.clk_src == ODP_CLOCK_CPU) itimer_init(tp); } @@ -368,7 +384,7 @@ static void odp_timer_pool_del(timer_pool_t *tp) odp_spinlock_lock(&tp->lock); - if (!inline_timers) { + if (!odp_global_rw->inline_timers) { /* Stop POSIX itimer signals */ if (tp->param.clk_src == ODP_CLOCK_CPU) itimer_fini(tp); @@ -390,6 +406,11 @@ static void odp_timer_pool_del(timer_pool_t *tp) timer_global->timer_pool[tp->tp_idx] = NULL; timer_global->timer_pool_used[tp->tp_idx] = 0; timer_global->num_timer_pools--; + + /* Disable inline timer polling */ + if (timer_global->num_timer_pools == 0) + odp_global_rw->inline_timers = false; + odp_ticketlock_unlock(&timer_global->lock); rc = odp_shm_free(shm); @@ -1093,6 +1114,11 @@ int odp_timer_capability(odp_timer_clk_src_t clk_src, odp_timer_pool_t odp_timer_pool_create(const char *name, const odp_timer_pool_param_t *param) { + if (odp_global_ro.init_param.not_used.feat.timer) { + ODP_ERR("Trying to use disabled ODP feature.\n"); + return ODP_TIMER_POOL_INVALID; + } + if (param->res_ns < timer_global->highest_res_ns) { __odp_errno = EINVAL; return ODP_TIMER_POOL_INVALID; @@ -1295,6 +1321,7 @@ void odp_timeout_free(odp_timeout_t tmo) int odp_timer_init_global(const odp_init_t *params) { odp_shm_t shm; + odp_bool_t inline_timers = false; shm = odp_shm_reserve("_odp_timer", sizeof(timer_global_t), ODP_CACHE_LINE_SIZE, 0); commit 83e9a6c3896d2d8a949a7bece565bae7dfd1d967 Author: Matias Elo <matias.elo(a)nokia.com> Date: Tue Oct 2 14:13:35 2018 +0300 linux-gen: queue: add internal interface for adding/removing inline timers Add internal queue interface functions queue_timer_add_fn_t and queue_timer_rem_fn_t for adding/removing inline timers. Signed-off-by: Matias Elo <matias.elo(a)nokia.com> Reviewed-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_basic_internal.h b/platform/linux-generic/include/odp_queue_basic_internal.h index 41ca424c7..c8ed978ce 100644 --- a/platform/linux-generic/include/odp_queue_basic_internal.h +++ b/platform/linux-generic/include/odp_queue_basic_internal.h @@ -54,6 +54,7 @@ struct queue_entry_s { ring_spsc_t ring_spsc; }; + odp_atomic_u64_t num_timers; int status; queue_deq_multi_fn_t orig_dequeue_multi; diff --git a/platform/linux-generic/include/odp_queue_if.h b/platform/linux-generic/include/odp_queue_if.h index 6ec7e24c2..5fe28dac1 100644 --- a/platform/linux-generic/include/odp_queue_if.h +++ b/platform/linux-generic/include/odp_queue_if.h @@ -39,6 +39,8 @@ typedef void (*queue_set_enq_deq_fn_t)(odp_queue_t queue, queue_enq_multi_fn_t enq_multi, queue_deq_fn_t deq, queue_deq_multi_fn_t deq_multi); +typedef void (*queue_timer_add_fn_t)(odp_queue_t queue); +typedef void (*queue_timer_rem_fn_t)(odp_queue_t queue); /* Queue functions towards other internal components */ typedef struct { @@ -51,6 +53,8 @@ typedef struct { queue_get_pktin_fn_t get_pktin; queue_set_pktin_fn_t set_pktin; queue_set_enq_deq_fn_t set_enq_deq_fn; + queue_timer_add_fn_t timer_add; + queue_timer_rem_fn_t timer_rem; /* Original queue dequeue multi function (before override). May be used * by an overriding dequeue function. */ diff --git a/platform/linux-generic/include/odp_queue_scalable_internal.h b/platform/linux-generic/include/odp_queue_scalable_internal.h index 05932a36a..3c582076c 100644 --- a/platform/linux-generic/include/odp_queue_scalable_internal.h +++ b/platform/linux-generic/include/odp_queue_scalable_internal.h @@ -35,6 +35,7 @@ struct queue_entry_s { sched_elem_t sched_elem; odp_ticketlock_t ODP_ALIGNED_CACHE lock; + odp_atomic_u64_t num_timers; int status; queue_enq_fn_t ODP_ALIGNED_CACHE enqueue; diff --git a/platform/linux-generic/odp_queue_basic.c b/platform/linux-generic/odp_queue_basic.c index ccd97da1e..73a5536d8 100644 --- a/platform/linux-generic/odp_queue_basic.c +++ b/platform/linux-generic/odp_queue_basic.c @@ -806,6 +806,7 @@ static int queue_init(queue_entry_t *queue, const char *name, queue->s.param.deq_mode = ODP_QUEUE_OP_DISABLED; queue->s.type = queue_type; + odp_atomic_init_u64(&queue->s.num_timers, 0); queue->s.pktin = PKTIN_INVALID; queue->s.pktout = PKTOUT_INVALID; @@ -949,6 +950,20 @@ static int queue_api_enq_multi(odp_queue_t handle, (odp_buffer_hdr_t **)(uintptr_t)ev, num); } +static void queue_timer_add(odp_queue_t handle) +{ + queue_entry_t *queue = qentry_from_handle(handle); + + odp_atomic_inc_u64(&queue->s.num_timers); +} + +static void queue_timer_rem(odp_queue_t handle) +{ + queue_entry_t *queue = qentry_from_handle(handle); + + odp_atomic_dec_u64(&queue->s.num_timers); +} + static int queue_api_enq(odp_queue_t handle, odp_event_t ev) { queue_entry_t *queue = qentry_from_handle(handle); @@ -1007,5 +1022,7 @@ queue_fn_t queue_basic_fn = { .get_pktin = queue_get_pktin, .set_pktin = queue_set_pktin, .set_enq_deq_fn = queue_set_enq_deq_func, - .orig_deq_multi = queue_orig_multi + .orig_deq_multi = queue_orig_multi, + .timer_add = queue_timer_add, + .timer_rem = queue_timer_rem }; diff --git a/platform/linux-generic/odp_queue_scalable.c b/platform/linux-generic/odp_queue_scalable.c index b7ff2195f..6ec7f14b7 100644 --- a/platform/linux-generic/odp_queue_scalable.c +++ b/platform/linux-generic/odp_queue_scalable.c @@ -124,6 +124,8 @@ static int queue_init(queue_entry_t *queue, const char *name, ring[ring_idx] = NULL; queue->s.type = queue->s.param.type; + odp_atomic_init_u64(&queue->s.num_timers, 0); + queue->s.enqueue = _queue_enq; queue->s.dequeue = _queue_deq; queue->s.enqueue_multi = _queue_enq_multi; @@ -958,6 +960,20 @@ static int queue_orig_multi(odp_queue_t handle, buf_hdr, num); } +static void queue_timer_add(odp_queue_t handle) +{ + queue_entry_t *queue = qentry_from_ext(handle); + + odp_atomic_inc_u64(&queue->s.num_timers); +} + +static void queue_timer_rem(odp_queue_t handle) +{ + queue_entry_t *queue = qentry_from_ext(handle); + + odp_atomic_dec_u64(&queue->s.num_timers); +} + /* API functions */ _odp_queue_api_fn_t queue_scalable_api = { .queue_create = queue_create, @@ -990,5 +1006,7 @@ queue_fn_t queue_scalable_fn = { .get_pktin = queue_get_pktin, .set_pktin = queue_set_pktin, .set_enq_deq_fn = queue_set_enq_deq_func, - .orig_deq_multi = queue_orig_multi + .orig_deq_multi = queue_orig_multi, + .timer_add = queue_timer_add, + .timer_rem = queue_timer_rem }; ----------------------------------------------------------------------- Summary of changes: platform/linux-generic/include/odp_global_data.h | 2 ++ .../include/odp_queue_basic_internal.h | 1 + platform/linux-generic/include/odp_queue_if.h | 4 +++ .../include/odp_queue_scalable_internal.h | 1 + .../linux-generic/include/odp_timer_internal.h | 13 ++----- platform/linux-generic/odp_init.c | 2 ++ platform/linux-generic/odp_queue_basic.c | 29 ++++++++++++++- platform/linux-generic/odp_queue_scalable.c | 32 ++++++++++++++++- platform/linux-generic/odp_timer.c | 41 +++++++++++++++++++--- 9 files changed, 108 insertions(+), 17 deletions(-) hooks/post-receive --

7 years, 3 months

[COMMIT] branch master updated. v1.19.0.2_tigermoth-211-g8dc48817b

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 8dc48817b06a5a303cf2b778eda68cc3627dd625 (commit) via f21e5e8137becb8bda722ac8b9dd27d476af77f8 (commit) via f46ecd688b8c0de5f273ad92478f473eea167112 (commit) from 0e53dabacd38643eb17834b15b5ae8e459700a4f (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 8dc48817b06a5a303cf2b778eda68cc3627dd625 Author: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov(a)linaro.org> Date: Mon Oct 22 17:01:14 2018 +0300 m4: update ax_prog_doxygen.m4 to latest version Signed-off-by: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov(a)linaro.org> Reviewed-by: Bill Fischofer <bill.fischofer(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/m4/ax_prog_doxygen.m4 b/m4/ax_prog_doxygen.m4 index 426ba0dfb..ed1dc83b4 100644 --- a/m4/ax_prog_doxygen.m4 +++ b/m4/ax_prog_doxygen.m4 @@ -1,5 +1,5 @@ # =========================================================================== -# http://www.gnu.org/software/autoconf-archive/ax_prog_doxygen.html +# https://www.gnu.org/software/autoconf-archive/ax_prog_doxygen.html # =========================================================================== # # SYNOPSIS @@ -21,7 +21,7 @@ # The DX_*_FEATURE macros control the default setting for the given # Doxygen feature. Supported features are 'DOXYGEN' itself, 'DOT' for # generating graphics, 'HTML' for plain HTML, 'CHM' for compressed HTML -# help (for MS users), 'CHI' for generating a seperate .chi file by the +# help (for MS users), 'CHI' for generating a separate .chi file by the # .chm file, and 'MAN', 'RTF', 'XML', 'PDF' and 'PS' for the appropriate # output formats. The environment variable DOXYGEN_PAPER_SIZE may be # specified to override the default 'a4wide' paper size. @@ -97,7 +97,7 @@ # and this notice are preserved. This file is offered as-is, without any # warranty. -#serial 18 +#serial 24 ## ----------## ## Defaults. ## @@ -164,7 +164,7 @@ AC_DEFUN([DX_TEST_FEATURE], [test "$DX_FLAG_$1" = 1]) AC_DEFUN([DX_CHECK_DEPEND], [ test "$DX_FLAG_$1" = "$2" \ || AC_MSG_ERROR([doxygen-DX_CURRENT_FEATURE ifelse([$2], 1, - requires, contradicts) doxygen-DX_CURRENT_FEATURE]) + requires, contradicts) doxygen-$1]) ]) # DX_CLEAR_DEPEND(FEATURE, REQUIRED_FEATURE, REQUIRED_STATE) @@ -265,14 +265,10 @@ m4_define([DX_loop], m4_dquote(m4_if(m4_eval(3 < m4_count($@)), 1, [m4_for([DX_i], 4, m4_count($@), 2, [, m4_eval(DX_i[/2])])], [])))dnl -# Environment variables used inside Doxyfile: +# Environment variables used inside doxygen.cfg: DX_ENV_APPEND(SRCDIR, $srcdir) -DX_ENV_APPEND(BUILDDIR, $builddir) -DX_ENV_APPEND(VERSION, $VERSION) -DX_ENV_APPEND(WITH_PLATFORM, $with_platform) DX_ENV_APPEND(PROJECT, $DX_PROJECT) DX_ENV_APPEND(VERSION, $PACKAGE_VERSION) -DX_ENV_APPEND(WITH_ARCH, $ARCH_DIR) # Doxygen itself: DX_ARG_ABLE(doc, [generate any doxygen documentation], @@ -325,8 +321,8 @@ DX_ARG_ABLE(chm, [generate doxygen compressed HTML help documentation], DX_ENV_APPEND(GENERATE_HTMLHELP, YES)], [DX_ENV_APPEND(GENERATE_HTMLHELP, NO)]) -# Seperate CHI file generation. -DX_ARG_ABLE(chi, [generate doxygen seperate compressed HTML help index file], +# Separate CHI file generation. +DX_ARG_ABLE(chi, [generate doxygen separate compressed HTML help index file], [DX_CHECK_DEPEND(chm, 1)], [DX_CLEAR_DEPEND(chm, 1)], [], @@ -382,94 +378,82 @@ a4wide|a4|letter|legal|executive) esac # Rules: -if test $DX_FLAG_html -eq 1; then - DX_SNIPPET_html="## ------------------------------- ## +AS_IF([[test $DX_FLAG_html -eq 1]], +[[DX_SNIPPET_html="## ------------------------------- ## ## Rules specific for HTML output. ## ## ------------------------------- ## -DX_CLEAN_HTML = \$(DX_DOCDIR)/html[]dnl +DX_CLEAN_HTML = \$(DX_DOCDIR)/html]dnl m4_foreach([DX_i], [m4_shift(DX_loop)], [[\\ - \$(DX_DOCDIR]DX_i[)/html]]) + \$(DX_DOCDIR]DX_i[)/html]])[ -" -else - DX_SNIPPET_html="" -fi -if test $DX_FLAG_chi -eq 1; then - DX_SNIPPET_chi=" -DX_CLEAN_CHI = \$(DX_DOCDIR)/\$(PACKAGE).chi[]dnl +"]], +[[DX_SNIPPET_html=""]]) +AS_IF([[test $DX_FLAG_chi -eq 1]], +[[DX_SNIPPET_chi=" +DX_CLEAN_CHI = \$(DX_DOCDIR)/\$(PACKAGE).chi]dnl m4_foreach([DX_i], [m4_shift(DX_loop)], [[\\ - \$(DX_DOCDIR]DX_i[)/\$(PACKAGE).chi]])" -else - DX_SNIPPET_chi="" -fi -if test $DX_FLAG_chm -eq 1; then - DX_SNIPPET_chm="## ------------------------------ ## + \$(DX_DOCDIR]DX_i[)/\$(PACKAGE).chi]])["]], +[[DX_SNIPPET_chi=""]]) +AS_IF([[test $DX_FLAG_chm -eq 1]], +[[DX_SNIPPET_chm="## ------------------------------ ## ## Rules specific for CHM output. ## ## ------------------------------ ## -DX_CLEAN_CHM = \$(DX_DOCDIR)/chm[]dnl +DX_CLEAN_CHM = \$(DX_DOCDIR)/chm]dnl m4_foreach([DX_i], [m4_shift(DX_loop)], [[\\ - \$(DX_DOCDIR]DX_i[)/chm]])\ + \$(DX_DOCDIR]DX_i[)/chm]])[\ ${DX_SNIPPET_chi} -" -else - DX_SNIPPET_chm="" -fi -if test $DX_FLAG_man -eq 1; then - DX_SNIPPET_man="## ------------------------------ ## +"]], +[[DX_SNIPPET_chm=""]]) +AS_IF([[test $DX_FLAG_man -eq 1]], +[[DX_SNIPPET_man="## ------------------------------ ## ## Rules specific for MAN output. ## ## ------------------------------ ## -DX_CLEAN_MAN = \$(DX_DOCDIR)/man[]dnl +DX_CLEAN_MAN = \$(DX_DOCDIR)/man]dnl m4_foreach([DX_i], [m4_shift(DX_loop)], [[\\ - \$(DX_DOCDIR]DX_i[)/man]]) + \$(DX_DOCDIR]DX_i[)/man]])[ -" -else - DX_SNIPPET_man="" -fi -if test $DX_FLAG_rtf -eq 1; then - DX_SNIPPET_rtf="## ------------------------------ ## +"]], +[[DX_SNIPPET_man=""]]) +AS_IF([[test $DX_FLAG_rtf -eq 1]], +[[DX_SNIPPET_rtf="## ------------------------------ ## ## Rules specific for RTF output. ## ## ------------------------------ ## -DX_CLEAN_RTF = \$(DX_DOCDIR)/rtf[]dnl +DX_CLEAN_RTF = \$(DX_DOCDIR)/rtf]dnl m4_foreach([DX_i], [m4_shift(DX_loop)], [[\\ - \$(DX_DOCDIR]DX_i[)/rtf]]) + \$(DX_DOCDIR]DX_i[)/rtf]])[ -" -else - DX_SNIPPET_rtf="" -fi -if test $DX_FLAG_xml -eq 1; then - DX_SNIPPET_xml="## ------------------------------ ## +"]], +[[DX_SNIPPET_rtf=""]]) +AS_IF([[test $DX_FLAG_xml -eq 1]], +[[DX_SNIPPET_xml="## ------------------------------ ## ## Rules specific for XML output. ## ## ------------------------------ ## -DX_CLEAN_XML = \$(DX_DOCDIR)/xml[]dnl +DX_CLEAN_XML = \$(DX_DOCDIR)/xml]dnl m4_foreach([DX_i], [m4_shift(DX_loop)], [[\\ - \$(DX_DOCDIR]DX_i[)/xml]]) + \$(DX_DOCDIR]DX_i[)/xml]])[ -" -else - DX_SNIPPET_xml="" -fi -if test $DX_FLAG_ps -eq 1; then - DX_SNIPPET_ps="## ----------------------------- ## +"]], +[[DX_SNIPPET_xml=""]]) +AS_IF([[test $DX_FLAG_ps -eq 1]], +[[DX_SNIPPET_ps="## ----------------------------- ## ## Rules specific for PS output. ## ## ----------------------------- ## -DX_CLEAN_PS = \$(DX_DOCDIR)/\$(PACKAGE).ps[]dnl +DX_CLEAN_PS = \$(DX_DOCDIR)/\$(PACKAGE).ps]dnl m4_foreach([DX_i], [m4_shift(DX_loop)], [[\\ - \$(DX_DOCDIR]DX_i[)/\$(PACKAGE).ps]]) + \$(DX_DOCDIR]DX_i[)/\$(PACKAGE).ps]])[ DX_PS_GOAL = doxygen-ps doxygen-ps: \$(DX_CLEAN_PS) -m4_foreach([DX_i], [DX_loop], +]m4_foreach([DX_i], [DX_loop], [[\$(DX_DOCDIR]DX_i[)/\$(PACKAGE).ps: \$(DX_DOCDIR]DX_i[)/\$(PACKAGE).tag \$(DX_V_LATEX)cd \$(DX_DOCDIR]DX_i[)/latex; \\ rm -f *.aux *.toc *.idx *.ind *.ilg *.log *.out; \\ @@ -485,25 +469,22 @@ m4_foreach([DX_i], [DX_loop], done; \\ \$(DX_DVIPS) -o ../\$(PACKAGE).ps refman.dvi -]])dnl -" -else - DX_SNIPPET_ps="" -fi -if test $DX_FLAG_pdf -eq 1; then - DX_SNIPPET_pdf="## ------------------------------ ## +]])["]], +[[DX_SNIPPET_ps=""]]) +AS_IF([[test $DX_FLAG_pdf -eq 1]], +[[DX_SNIPPET_pdf="## ------------------------------ ## ## Rules specific for PDF output. ## ## ------------------------------ ## -DX_CLEAN_PDF = \$(DX_DOCDIR)/\$(PACKAGE).pdf[]dnl +DX_CLEAN_PDF = \$(DX_DOCDIR)/\$(PACKAGE).pdf]dnl m4_foreach([DX_i], [m4_shift(DX_loop)], [[\\ - \$(DX_DOCDIR]DX_i[)/\$(PACKAGE).pdf]]) + \$(DX_DOCDIR]DX_i[)/\$(PACKAGE).pdf]])[ DX_PDF_GOAL = doxygen-pdf doxygen-pdf: \$(DX_CLEAN_PDF) -m4_foreach([DX_i], [DX_loop], +]m4_foreach([DX_i], [DX_loop], [[\$(DX_DOCDIR]DX_i[)/\$(PACKAGE).pdf: \$(DX_DOCDIR]DX_i[)/\$(PACKAGE).tag \$(DX_V_LATEX)cd \$(DX_DOCDIR]DX_i[)/latex; \\ rm -f *.aux *.toc *.idx *.ind *.ilg *.log *.out; \\ @@ -519,31 +500,26 @@ m4_foreach([DX_i], [DX_loop], done; \\ mv refman.pdf ../\$(PACKAGE).pdf -]])dnl -" -else - DX_SNIPPET_pdf="" -fi -if test $DX_FLAG_ps -eq 1 -o $DX_FLAG_pdf -eq 1; then - DX_SNIPPET_latex="## ------------------------------------------------- ## +]])["]], +[[DX_SNIPPET_pdf=""]]) +AS_IF([[test $DX_FLAG_ps -eq 1 -o $DX_FLAG_pdf -eq 1]], +[[DX_SNIPPET_latex="## ------------------------------------------------- ## ## Rules specific for LaTeX (shared for PS and PDF). ## ## ------------------------------------------------- ## DX_V_LATEX = \$(_DX_v_LATEX_\$(V)) _DX_v_LATEX_ = \$(_DX_v_LATEX_\$(AM_DEFAULT_VERBOSITY)) -_DX_v_LATEX_0 = @echo \" LATEX \" \$[]][[]@; +_DX_v_LATEX_0 = @echo \" LATEX \" \$][@; -DX_CLEAN_LATEX = \$(DX_DOCDIR)/latex[]dnl +DX_CLEAN_LATEX = \$(DX_DOCDIR)/latex]dnl m4_foreach([DX_i], [m4_shift(DX_loop)], [[\\ - \$(DX_DOCDIR]DX_i[)/latex]]) + \$(DX_DOCDIR]DX_i[)/latex]])[ -" -else - DX_SNIPPET_latex="" -fi +"]], +[[DX_SNIPPET_latex=""]]) -if test $DX_FLAG_doc -eq 1; then - DX_SNIPPET_doc="## --------------------------------- ## +AS_IF([[test $DX_FLAG_doc -eq 1]], +[[DX_SNIPPET_doc="## --------------------------------- ## ## Format-independent Doxygen rules. ## ## --------------------------------- ## @@ -563,23 +539,24 @@ _DX_v_DXGEN_0 = @echo \" DXGEN \" \$<; .INTERMEDIATE: doxygen-run \$(DX_PS_GOAL) \$(DX_PDF_GOAL) -doxygen-run:[]m4_foreach([DX_i], [DX_loop], - [[ \$(DX_DOCDIR]DX_i[)/\$(PACKAGE).tag]]) +doxygen-run:]m4_foreach([DX_i], [DX_loop], + [[ \$(DX_DOCDIR]DX_i[)/\$(PACKAGE).tag]])[ doxygen-doc: doxygen-run \$(DX_PS_GOAL) \$(DX_PDF_GOAL) -m4_foreach([DX_i], [DX_loop], +]m4_foreach([DX_i], [DX_loop], [[\$(DX_DOCDIR]DX_i[)/\$(PACKAGE).tag: \$(DX_CONFIG]DX_i[) \$(pkginclude_HEADERS) \$(A""M_V_at)rm -rf \$(DX_DOCDIR]DX_i[) \$(DX_V_DXGEN)\$(DX_ENV) DOCDIR=\$(DX_DOCDIR]DX_i[) \$(DX_DOXYGEN) \$(DX_CONFIG]DX_i[) \$(A""M_V_at)echo Timestamp >\$][@ ]])dnl -DX_CLEANFILES = \\ +[DX_CLEANFILES = \\] m4_foreach([DX_i], [DX_loop], -[[ \$(DX_DOCDIR]DX_i[)/\$(PACKAGE).tag \\ +[[ \$(DX_DOCDIR]DX_i[)/doxygen_sqlite3.db \\ + \$(DX_DOCDIR]DX_i[)/\$(PACKAGE).tag \\ ]])dnl - -r \\ +[ -r \\ \$(DX_CLEAN_HTML) \\ \$(DX_CLEAN_CHM) \\ \$(DX_CLEAN_CHI) \\ @@ -588,19 +565,17 @@ m4_foreach([DX_i], [DX_loop], \$(DX_CLEAN_XML) \\ \$(DX_CLEAN_PS) \\ \$(DX_CLEAN_PDF) \\ - \$(DX_CLEAN_LATEX)" -else - DX_SNIPPET_doc="" -fi + \$(DX_CLEAN_LATEX)"]], +[[DX_SNIPPET_doc=""]]) AC_SUBST([DX_RULES], ["${DX_SNIPPET_doc}"])dnl AM_SUBST_NOTMAKE([DX_RULES]) #For debugging: -echo DX_FLAG_doc=$DX_FLAG_doc +#echo DX_FLAG_doc=$DX_FLAG_doc #echo DX_FLAG_dot=$DX_FLAG_dot #echo DX_FLAG_man=$DX_FLAG_man -echo DX_FLAG_html=$DX_FLAG_html +#echo DX_FLAG_html=$DX_FLAG_html #echo DX_FLAG_chm=$DX_FLAG_chm #echo DX_FLAG_chi=$DX_FLAG_chi #echo DX_FLAG_rtf=$DX_FLAG_rtf commit f21e5e8137becb8bda722ac8b9dd27d476af77f8 Author: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov(a)linaro.org> Date: Mon Oct 22 17:00:39 2018 +0300 doc: pass generated include files to Doxygen Pass generated files to doxygen to let it know about ODP_DEPRECATED()/ODP_VERSION_API_xxx Signed-off-by: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov(a)linaro.org> Reviewed-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/Doxyfile b/doc/application-api-guide/Doxyfile index 00c18df2e..b101153d1 100644 --- a/doc/application-api-guide/Doxyfile +++ b/doc/application-api-guide/Doxyfile @@ -4,6 +4,7 @@ PROJECT_NAME = "API Reference Manual" PROJECT_NUMBER = $(VERSION) PROJECT_LOGO = $(SRCDIR)/doc/images/ODP-Logo-HQ.svg INPUT = $(SRCDIR)/doc/application-api-guide \ + include \ $(SRCDIR)/include EXAMPLE_PATH = $(SRCDIR)/example $(SRCDIR) WARNINGS = NO diff --git a/doc/platform-api-guide/Doxyfile b/doc/platform-api-guide/Doxyfile index d716b4a3c..02e325cc6 100644 --- a/doc/platform-api-guide/Doxyfile +++ b/doc/platform-api-guide/Doxyfile @@ -5,8 +5,8 @@ PROJECT_NUMBER = $(VERSION) PROJECT_LOGO = $(SRCDIR)/doc/images/ODP-Logo-HQ.svg INPUT = $(SRCDIR)/doc/application-api-guide \ $(SRCDIR)/doc/platform-api-guide \ + include/odp/api \ $(SRCDIR)/include/odp/api \ $(SRCDIR)/platform/$(WITH_PLATFORM)/doc \ - $(SRCDIR)/platform/$(WITH_PLATFORM)/include/odp/api \ - $(SRCDIR)/platform/$(WITH_PLATFORM)/arch/$(WITH_ARCH) + $(SRCDIR)/platform/$(WITH_PLATFORM)/include/odp/api EXAMPLE_PATH = $(SRCDIR)/example $(SRCDIR)/platform $(SRCDIR) commit f46ecd688b8c0de5f273ad92478f473eea167112 Author: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov(a)linaro.org> Date: Mon Oct 22 17:00:29 2018 +0300 configure.ac: set DX environment WITH_PLATFORM from configure.ac Rather than manually modifying ax_prog_doxygen.m4 set variable from configure.ac. Signed-off-by: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov(a)linaro.org> Reviewed-by: Bill Fischofer <bill.fischofer(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/configure.ac b/configure.ac index 3b71ef510..68da20df5 100644 --- a/configure.ac +++ b/configure.ac @@ -283,6 +283,8 @@ AM_CONDITIONAL([ARCH_IS_X86_64], [test "x${ARCH_ABI}" = "xx86_64-linux"]) DX_HTML_FEATURE(ON) DX_PDF_FEATURE(OFF) DX_PS_FEATURE(OFF) +DX_ENV_APPEND(WITH_PLATFORM, $with_platform) + DX_INIT_DOXYGEN($PACKAGE_NAME, ${srcdir}/doc/application-api-guide/Doxyfile, ${builddir}/doc/application-api-guide/output, ----------------------------------------------------------------------- Summary of changes: configure.ac | 2 + doc/application-api-guide/Doxyfile | 1 + doc/platform-api-guide/Doxyfile | 4 +- m4/ax_prog_doxygen.m4 | 177 ++++++++++++++++--------------------- 4 files changed, 81 insertions(+), 103 deletions(-) hooks/post-receive --

7 years, 3 months

[COMMIT] branch master updated. v1.19.0.2_tigermoth-208-g0e53dabac

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 0e53dabacd38643eb17834b15b5ae8e459700a4f (commit) via bf5c002a2d9549fa0120d03588684096145f9fcc (commit) via 3f3997ff3af3f85c4ce150f41c9d95b7fd9e9f57 (commit) via 8f9f3784b900297f555f3a17d5ce47584846c77c (commit) via c5ee737dca5375e2f9fab38ffc768223b2d9eb64 (commit) via 1aca253a08ee916a66486bb6eca05cfc4567e38f (commit) via 4443213d11e3d9feeecb6d2f5b3bad80d8d73c4b (commit) via b6f8955f3f70a5e09bdd86c408373c0b8caeb3eb (commit) via 3cae562cf7e11f87d00c19f96b7b29f98a8dddc3 (commit) via 8c72eb46fb1ee47340ffc7af322215fb78894b37 (commit) from 044cae351a1e60e4381cdbae37e2a759772c2258 (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 0e53dabacd38643eb17834b15b5ae8e459700a4f Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Mon Oct 15 15:30:54 2018 +0300 travis: run all test with gcc first Change build matrix format from N x 2 to 2 x N. Run first all tests with GCC, then with clang. Signed-off-by: Petri Savolainen <petri.savolainen(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 e07942872..4e29102fa 100644 --- a/.travis.yml +++ b/.travis.yml @@ -36,6 +36,10 @@ cache: - netmap - $HOME/doxygen-install +compiler: + - gcc + - clang + env: global: # @@ -74,10 +78,6 @@ matrix: - compiler: gcc env: BUILD_ONLY=1 ARCH="i386" -compiler: - - gcc - - clang - install: - if [ ${NETMAP} -eq 1 ] ; then echo "Installing NETMAP"; commit bf5c002a2d9549fa0120d03588684096145f9fcc Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Tue Oct 16 15:57:01 2018 +0300 travis: explicit netmap test cases Build, install and test with netmap only on couple of test cases. This saves travis test time and validates that ODP passes tests also without netmap (on x86). Signed-off-by: Petri Savolainen <petri.savolainen(a)linaro.org> Reviewed-by: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/.travis.yml b/.travis.yml index 18bf47902..e07942872 100644 --- a/.travis.yml +++ b/.travis.yml @@ -45,9 +45,11 @@ env: - CODECOV_TOKEN=a733c34c-5f5c-4ff1-af4b-e9f5edb1ab5e - UBUNTU_VERS="16.04" - BUILD_ONLY=0 + - NETMAP=0 matrix: - CONF="" - CONF="--disable-abi-compat" + - NETMAP=1 CONF="" - BUILD_ONLY=1 ARCH="arm64" - BUILD_ONLY=1 ARCH="armhf" - BUILD_ONLY=1 ARCH="powerpc" @@ -58,7 +60,7 @@ env: - BUILD_ONLY=1 ARCH="i386" CONF="--disable-abi-compat" - CONF="--enable-deprecated" - CONF="--enable-dpdk-zero-copy --disable-static-applications" - - CONF="--disable-static-applications" + - NETMAP=1 CONF="--disable-static-applications" - CONF="--disable-host-optimization" - CONF="--disable-host-optimization --disable-abi-compat" - BUILD_ONLY=1 ARCH="x86_64" CONF="--enable-pcapng-support" @@ -77,13 +79,13 @@ compiler: - clang install: - - sudo apt-get install linux-headers-`uname -r` - - if [ -z "${ARCH}" -o "${ARCH}" == "x86_64" ] ; then - echo "compilling netmap"; + - if [ ${NETMAP} -eq 1 ] ; then + echo "Installing NETMAP"; + sudo apt-get install linux-headers-`uname -r` ; CDIR=`pwd` ; git -c advice.detachedHead=false clone -q --depth=1 --single-branch --branch=v11.2 https://github.com/luigirizzo/netmap.git; pushd netmap/LINUX; - ./configure; + ./configure --drivers= ; make -j $(nproc); popd; sudo insmod ./netmap/LINUX/netmap.ko; commit 3f3997ff3af3f85c4ce150f41c9d95b7fd9e9f57 Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Tue Oct 16 15:28:56 2018 +0300 travis: start coverage test early Coverage test takes a long time. Start it early, so that other jobs run in parallel with it. Signed-off-by: Petri Savolainen <petri.savolainen(a)linaro.org> Reviewed-by: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/.travis.yml b/.travis.yml index ffafe650a..18bf47902 100644 --- a/.travis.yml +++ b/.travis.yml @@ -106,6 +106,16 @@ script: fi jobs: include: + - stage: test + env: TEST=coverage + compiler: gcc + script: + - if [ -z "${DOCKER_NAMESPACE}" ] ; then export DOCKER_NAMESPACE="opendataplane"; fi + - docker run --privileged -i -t + -v `pwd`:/odp --shm-size 8g + -e CODECOV_TOKEN="${CODECOV_TOKEN}" + -e CC="${CC}" + ${DOCKER_NAMESPACE}/travis-odp-lng-ubuntu_16.04 /odp/scripts/ci/coverage.sh - stage: test env: TEST=scheduler_sp compiler: gcc @@ -142,17 +152,6 @@ jobs: -e ODP_CONFIG_FILE=/odp/platform/linux-generic/test/process-mode.conf -e ODPH_PROC_MODE=1 ${DOCKER_NAMESPACE}/travis-odp-lng-ubuntu_16.04 /odp/scripts/ci/check.sh - - stage: test - env: TEST=coverage - compiler: gcc - script: - - if [ -z "${DOCKER_NAMESPACE}" ] ; then export DOCKER_NAMESPACE="opendataplane"; fi - - docker run --privileged -i -t - -v `pwd`:/odp --shm-size 8g - -e CODECOV_TOKEN="${CODECOV_TOKEN}" - -e CC="${CC}" - -e CONF="${CONF}" - ${DOCKER_NAMESPACE}/travis-odp-lng-ubuntu_16.04 /odp/scripts/ci/coverage.sh - stage: test env: TEST=distcheck compiler: gcc commit 8f9f3784b900297f555f3a17d5ce47584846c77c Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Tue Oct 16 12:48:30 2018 +0300 travis: exclude duplicate tests arm64/i386 with GCC build tests are excluded from the matrix as those are executed in the first phase. Signed-off-by: Petri Savolainen <petri.savolainen(a)linaro.org> Reviewed-by: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/.travis.yml b/.travis.yml index ea26b0533..ffafe650a 100644 --- a/.travis.yml +++ b/.travis.yml @@ -65,6 +65,13 @@ env: - CONF="--without-openssl" - CONF="" UBUNTU_VERS="18.04" +matrix: + exclude: + - compiler: gcc + env: BUILD_ONLY=1 ARCH="arm64" + - compiler: gcc + env: BUILD_ONLY=1 ARCH="i386" + compiler: - gcc - clang commit c5ee737dca5375e2f9fab38ffc768223b2d9eb64 Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Tue Oct 16 10:24:53 2018 +0300 travis: change pcapng test to build only Pcapng does not have specific validation tests, so running 'make check' on it is not important. Change it to a build only test to save travis execution time. Signed-off-by: Petri Savolainen <petri.savolainen(a)linaro.org> Reviewed-by: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/.travis.yml b/.travis.yml index 8c369c535..ea26b0533 100644 --- a/.travis.yml +++ b/.travis.yml @@ -61,7 +61,7 @@ env: - CONF="--disable-static-applications" - CONF="--disable-host-optimization" - CONF="--disable-host-optimization --disable-abi-compat" - - CONF="--enable-pcapng-support" + - BUILD_ONLY=1 ARCH="x86_64" CONF="--enable-pcapng-support" - CONF="--without-openssl" - CONF="" UBUNTU_VERS="18.04" @@ -85,7 +85,7 @@ install: script: - if [ -z "${DOCKER_NAMESPACE}" ] ; then export DOCKER_NAMESPACE="opendataplane"; fi - if [ ${BUILD_ONLY} -eq 1 ] ; then - docker run -i -t -v `pwd`:/odp + docker run -i -t -v `pwd`:/odp --shm-size 8g -e CC="${CC}" -e CONF="${CONF}" ${DOCKER_NAMESPACE}/travis-odp-lng-ubuntu_${UBUNTU_VERS} /odp/scripts/ci/build_${ARCH}.sh ; commit 1aca253a08ee916a66486bb6eca05cfc4567e38f Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Tue Oct 16 10:16:16 2018 +0300 travis: explicit build only tests Use new BUILD_ONLY variable to select test cases that are built only, and not checked with 'make check'. This enables to mark some x86_64 test cases to "build only". Signed-off-by: Petri Savolainen <petri.savolainen(a)linaro.org> Reviewed-by: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/.travis.yml b/.travis.yml index 7b912af21..8c369c535 100644 --- a/.travis.yml +++ b/.travis.yml @@ -44,17 +44,18 @@ env: # you need generated new one at https://codecov.io specific for your repo. - CODECOV_TOKEN=a733c34c-5f5c-4ff1-af4b-e9f5edb1ab5e - UBUNTU_VERS="16.04" + - BUILD_ONLY=0 matrix: - CONF="" - CONF="--disable-abi-compat" - - CROSS_ARCH="arm64" - - CROSS_ARCH="armhf" - - CROSS_ARCH="powerpc" - - CROSS_ARCH="i386" - - CROSS_ARCH="arm64" CONF="--disable-abi-compat" - - CROSS_ARCH="armhf" CONF="--disable-abi-compat" - - CROSS_ARCH="powerpc" CONF="--disable-abi-compat" - - CROSS_ARCH="i386" CONF="--disable-abi-compat" + - BUILD_ONLY=1 ARCH="arm64" + - BUILD_ONLY=1 ARCH="armhf" + - BUILD_ONLY=1 ARCH="powerpc" + - BUILD_ONLY=1 ARCH="i386" + - BUILD_ONLY=1 ARCH="arm64" CONF="--disable-abi-compat" + - BUILD_ONLY=1 ARCH="armhf" CONF="--disable-abi-compat" + - BUILD_ONLY=1 ARCH="powerpc" CONF="--disable-abi-compat" + - BUILD_ONLY=1 ARCH="i386" CONF="--disable-abi-compat" - CONF="--enable-deprecated" - CONF="--enable-dpdk-zero-copy --disable-static-applications" - CONF="--disable-static-applications" @@ -70,7 +71,7 @@ compiler: install: - sudo apt-get install linux-headers-`uname -r` - - if [ -z "${CROSS_ARCH}" ] ; then + - if [ -z "${ARCH}" -o "${ARCH}" == "x86_64" ] ; then echo "compilling netmap"; CDIR=`pwd` ; git -c advice.detachedHead=false clone -q --depth=1 --single-branch --branch=v11.2 https://github.com/luigirizzo/netmap.git; @@ -83,11 +84,11 @@ install: fi script: - if [ -z "${DOCKER_NAMESPACE}" ] ; then export DOCKER_NAMESPACE="opendataplane"; fi - - if [ -n "${CROSS_ARCH}" ] ; then + - if [ ${BUILD_ONLY} -eq 1 ] ; then docker run -i -t -v `pwd`:/odp -e CC="${CC}" -e CONF="${CONF}" - ${DOCKER_NAMESPACE}/travis-odp-lng-ubuntu_${UBUNTU_VERS} /odp/scripts/ci/build_${CROSS_ARCH}.sh ; + ${DOCKER_NAMESPACE}/travis-odp-lng-ubuntu_${UBUNTU_VERS} /odp/scripts/ci/build_${ARCH}.sh ; else echo "Running test" ; docker run --privileged -i -t commit 4443213d11e3d9feeecb6d2f5b3bad80d8d73c4b Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Tue Oct 16 08:55:38 2018 +0300 travis: clean process mode test script Removed extra environment variable. Use default configure. Explicitly use ubuntu 16.04 as other tests under jobs. Signed-off-by: Petri Savolainen <petri.savolainen(a)linaro.org> Reviewed-by: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/.travis.yml b/.travis.yml index d39cf93dd..7b912af21 100644 --- a/.travis.yml +++ b/.travis.yml @@ -129,12 +129,11 @@ jobs: - if [ -z "${DOCKER_NAMESPACE}" ] ; then export DOCKER_NAMESPACE="opendataplane"; fi - docker run --privileged -i -t -v `pwd`:/odp --shm-size 8g - -e CODECOV_TOKEN="${CODECOV_TOKEN}" -e CC="${CC}" - -e CONF="${CONF}" + -e CONF="" -e ODP_CONFIG_FILE=/odp/platform/linux-generic/test/process-mode.conf -e ODPH_PROC_MODE=1 - ${DOCKER_NAMESPACE}/travis-odp-lng-ubuntu_${UBUNTU_VERS} /odp/scripts/ci/check.sh ; + ${DOCKER_NAMESPACE}/travis-odp-lng-ubuntu_16.04 /odp/scripts/ci/check.sh - stage: test env: TEST=coverage compiler: gcc commit b6f8955f3f70a5e09bdd86c408373c0b8caeb3eb Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Tue Oct 16 08:47:50 2018 +0300 travis: split distcheck test Distcheck test takes over 20 minutes. Split it into two test cases, so that those can run in parallel. Signed-off-by: Petri Savolainen <petri.savolainen(a)linaro.org> Reviewed-by: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/.travis.yml b/.travis.yml index f6e0e49ff..d39cf93dd 100644 --- a/.travis.yml +++ b/.travis.yml @@ -154,7 +154,17 @@ jobs: - docker run --privileged -i -t -v `pwd`:/odp --shm-size 8g -e CC="${CC}" - -e CONF="${CONF}" + -e CONF="--enable-user-guides" + ${DOCKER_NAMESPACE}/travis-odp-lng-ubuntu_16.04 /odp/scripts/ci/distcheck.sh + - stage: test + env: TEST=distcheck_nonabi + compiler: gcc + script: + - if [ -z "${DOCKER_NAMESPACE}" ] ; then export DOCKER_NAMESPACE="opendataplane"; fi + - docker run --privileged -i -t + -v `pwd`:/odp --shm-size 8g + -e CC="${CC}" + -e CONF="--enable-user-guides --disable-abi-compat" ${DOCKER_NAMESPACE}/travis-odp-lng-ubuntu_16.04 /odp/scripts/ci/distcheck.sh - stage: "build only" env: TEST=doxygen diff --git a/scripts/ci/distcheck.sh b/scripts/ci/distcheck.sh index c78349e2d..9d45536f4 100755 --- a/scripts/ci/distcheck.sh +++ b/scripts/ci/distcheck.sh @@ -7,15 +7,13 @@ fi cd "$(dirname "$0")"/../.. ./bootstrap -./configure \ - --enable-user-guides +./configure ${CONF} # Ignore possible failures there because these tests depends on measurements # and systems might differ in performance. export CI="true" -make distcheck - -make clean +# Additional configure flags for distcheck +export DISTCHECK_CONFIGURE_FLAGS="${CONF}" -make distcheck DISTCHECK__CONFIGURE_FLAGS=--disable-abi-compat +make distcheck commit 3cae562cf7e11f87d00c19f96b7b29f98a8dddc3 Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Mon Oct 15 17:12:10 2018 +0300 travis: add test cases for optional schedulers Added separate tests for schedulers, so that those can be executed in parallel. Schedulers are tested with default compiler flags. Previously, optional schedulers were tested only without compiler optimizations (-O0) in coverage test. Signed-off-by: Petri Savolainen <petri.savolainen(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 16e3a3103..f6e0e49ff 100644 --- a/.travis.yml +++ b/.travis.yml @@ -98,6 +98,28 @@ script: fi jobs: include: + - stage: test + env: TEST=scheduler_sp + compiler: gcc + script: + - if [ -z "${DOCKER_NAMESPACE}" ] ; then export DOCKER_NAMESPACE="opendataplane"; fi + - docker run --privileged -i -t + -v `pwd`:/odp --shm-size 8g + -e CC="${CC}" + -e CONF="" + -e ODP_SCHEDULER=sp + ${DOCKER_NAMESPACE}/travis-odp-lng-ubuntu_16.04 /odp/scripts/ci/check.sh + - stage: test + env: TEST=scheduler_scalable + compiler: gcc + script: + - if [ -z "${DOCKER_NAMESPACE}" ] ; then export DOCKER_NAMESPACE="opendataplane"; fi + - docker run --privileged -i -t + -v `pwd`:/odp --shm-size 8g + -e CC="${CC}" + -e CONF="" + -e ODP_SCHEDULER=scalable + ${DOCKER_NAMESPACE}/travis-odp-lng-ubuntu_16.04 /odp/scripts/ci/check.sh - stage: test env: TEST=process_mode install: commit 8c72eb46fb1ee47340ffc7af322215fb78894b37 Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Mon Oct 15 14:42:55 2018 +0300 travis: build ODP on multiple threads Build ODP on multiple threads with 'make -j $(nproc)'. Signed-off-by: Petri Savolainen <petri.savolainen(a)linaro.org> Reviewed-by: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/scripts/ci/build.sh b/scripts/ci/build.sh index 995c797b2..21922a493 100755 --- a/scripts/ci/build.sh +++ b/scripts/ci/build.sh @@ -9,7 +9,7 @@ cd "$(dirname "$0")"/../.. --prefix=/opt/odp \ ${CONF} -make -j 8 +make -j $(nproc) make install ----------------------------------------------------------------------- Summary of changes: .travis.yml | 98 ++++++++++++++++++++++++++++++++++--------------- scripts/ci/build.sh | 2 +- scripts/ci/distcheck.sh | 10 ++--- 3 files changed, 74 insertions(+), 36 deletions(-) hooks/post-receive --

7 years, 3 months

[COMMIT] branch master updated. v1.19.0.2_tigermoth-198-g044cae351

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 044cae351a1e60e4381cdbae37e2a759772c2258 (commit) via 95cbee5ac9cdcb696d48e6f820593c813f535f8a (commit) via 0c8a2c0027e35184cbc6eeadc3a8aa8bd2e8605b (commit) via 2c41f8f678bbb96a63dedb4e4d948c5dd779f65d (commit) via 7e188187a77fa9bab56f1407eb660c324736cd34 (commit) via 031802ec0c5e3e44d70aee7b19e0cd1055c9fe0b (commit) from 690bacc61c55e4f8f3195faaac068da83790ebf4 (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 044cae351a1e60e4381cdbae37e2a759772c2258 Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Wed Oct 17 16:51:49 2018 +0300 linux-gen: config: maximum pool size 1M Change maximum pool size back to 1M. Maximum packet pool capability is defined in the config file. This is the upper limit for that config. Also this is max capability of buffer and tmo pools. Signed-off-by: Petri Savolainen <petri.savolainen(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/platform/linux-generic/include/odp_config_internal.h b/platform/linux-generic/include/odp_config_internal.h index 7a573bd76..65f751972 100644 --- a/platform/linux-generic/include/odp_config_internal.h +++ b/platform/linux-generic/include/odp_config_internal.h @@ -150,7 +150,7 @@ extern "C" { * Maximum number of events in a pool. Power of two minus one results optimal * memory usage for the ring. */ -#define CONFIG_POOL_MAX_NUM ((256 * 1024) - 1) +#define CONFIG_POOL_MAX_NUM ((1024 * 1024) - 1) /* * Maximum number of events in a thread local pool cache commit 95cbee5ac9cdcb696d48e6f820593c813f535f8a Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Wed Oct 17 16:41:52 2018 +0300 linux-gen: pool: add max num packets in config file This config is used to for maximum capability. The default capability needs to be modest so that system memory limit is not exceeded. User may increase maximum number of packets when system memory size allows (and SHM single VA is not used). Signed-off-by: Petri Savolainen <petri.savolainen(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/config/odp-linux-generic.conf b/config/odp-linux-generic.conf index 4db9ed489..af651d7f6 100644 --- a/config/odp-linux-generic.conf +++ b/config/odp-linux-generic.conf @@ -16,7 +16,7 @@ # Mandatory fields odp_implementation = "linux-generic" -config_file_version = "0.1.0" +config_file_version = "0.1.1" # Shared memory options shm: { @@ -39,6 +39,16 @@ shm: { single_va = 0 } +# Pool options +pool: { + # Packet pool options + pkt: { + # Maximum number of packets per pool. Power of two minus one + # results optimal memory usage (e.g. (256 * 1024) - 1). + max_num = 262143 + } +} + # DPDK pktio options pktio_dpdk: { # Default options diff --git a/platform/linux-generic/include/odp_pool_internal.h b/platform/linux-generic/include/odp_pool_internal.h index 2696e8eec..94f859de0 100644 --- a/platform/linux-generic/include/odp_pool_internal.h +++ b/platform/linux-generic/include/odp_pool_internal.h @@ -89,6 +89,11 @@ typedef struct pool_t { typedef struct pool_table_t { pool_t pool[ODP_CONFIG_POOLS]; odp_shm_t shm; + + struct { + uint32_t pkt_max_num; + } config; + } pool_table_t; extern pool_table_t *pool_tbl; diff --git a/platform/linux-generic/odp_pool.c b/platform/linux-generic/odp_pool.c index db7a8da3f..d08be4378 100644 --- a/platform/linux-generic/odp_pool.c +++ b/platform/linux-generic/odp_pool.c @@ -22,6 +22,7 @@ #include <odp_ring_internal.h> #include <odp_shm_internal.h> #include <odp_global_data.h> +#include <odp_libconfig_internal.h> #include <string.h> #include <stdio.h> @@ -33,8 +34,9 @@ #define UNLOCK(a) odp_ticketlock_unlock(a) #define LOCK_INIT(a) odp_ticketlock_init(a) -#define CACHE_BURST 32 -#define RING_SIZE_MIN (2 * CACHE_BURST) +#define CACHE_BURST 32 +#define RING_SIZE_MIN (2 * CACHE_BURST) +#define POOL_MAX_NUM_MIN RING_SIZE_MIN /* Make sure packet buffers don't cross huge page boundaries starting from this * page size. 2MB is typically the smallest used huge page size. */ @@ -83,6 +85,32 @@ static inline pool_t *pool_from_buf(odp_buffer_t buf) return buf_hdr->pool_ptr; } +static int read_config_file(pool_table_t *pool_tbl) +{ + const char *str; + int val = 0; + + ODP_PRINT("Pool config:\n"); + + str = "pool.pkt.max_num"; + if (!_odp_libconfig_lookup_int(str, &val)) { + ODP_ERR("Config option '%s' not found.\n", str); + return -1; + } + + if (val > CONFIG_POOL_MAX_NUM || val < POOL_MAX_NUM_MIN) { + ODP_ERR("Bad value %s = %u\n", str, val); + return -1; + } + + pool_tbl->config.pkt_max_num = val; + ODP_PRINT(" %s: %i\n", str, val); + + ODP_PRINT("\n"); + + return 0; +} + int odp_pool_init_global(void) { uint32_t i; @@ -101,6 +129,11 @@ int odp_pool_init_global(void) memset(pool_tbl, 0, sizeof(pool_table_t)); pool_tbl->shm = shm; + if (read_config_file(pool_tbl)) { + odp_shm_free(shm); + return -1; + } + for (i = 0; i < ODP_CONFIG_POOLS; i++) { pool_t *pool = pool_entry(i); @@ -950,7 +983,7 @@ int odp_pool_capability(odp_pool_capability_t *capa) /* Packet pools */ capa->pkt.max_pools = ODP_CONFIG_POOLS; capa->pkt.max_len = CONFIG_PACKET_MAX_LEN; - capa->pkt.max_num = CONFIG_POOL_MAX_NUM; + capa->pkt.max_num = pool_tbl->config.pkt_max_num; capa->pkt.min_headroom = CONFIG_PACKET_HEADROOM; capa->pkt.max_headroom = CONFIG_PACKET_HEADROOM; capa->pkt.min_tailroom = CONFIG_PACKET_TAILROOM; diff --git a/platform/linux-generic/test/process-mode.conf b/platform/linux-generic/test/process-mode.conf index 8618f2fdb..d80df25c4 100644 --- a/platform/linux-generic/test/process-mode.conf +++ b/platform/linux-generic/test/process-mode.conf @@ -1,6 +1,6 @@ # Mandatory fields odp_implementation = "linux-generic" -config_file_version = "0.1.0" +config_file_version = "0.1.1" # Shared memory options shm: { commit 0c8a2c0027e35184cbc6eeadc3a8aa8bd2e8605b Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Wed Oct 17 14:43:43 2018 +0300 linux-gen: pool: add packet param checks Do also checks against maximum number of packets and max packet headroom size. Signed-off-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_pool.c b/platform/linux-generic/odp_pool.c index 544348189..db7a8da3f 100644 --- a/platform/linux-generic/odp_pool.c +++ b/platform/linux-generic/odp_pool.c @@ -572,6 +572,17 @@ static int check_params(odp_pool_param_t *params) break; case ODP_POOL_PACKET: + if (params->pkt.num > capa.pkt.max_num) { + ODP_ERR("pkt.num too large %u\n", params->pkt.num); + return -1; + } + + if (params->pkt.max_num > capa.pkt.max_num) { + ODP_ERR("pkt.max_num too large %u\n", + params->pkt.max_num); + return -1; + } + if (params->pkt.len > capa.pkt.max_len) { ODP_ERR("pkt.len too large %u\n", params->pkt.len); return -1; @@ -595,6 +606,12 @@ static int check_params(odp_pool_param_t *params) return -1; } + if (params->pkt.headroom > capa.pkt.max_headroom) { + ODP_ERR("pkt.headroom too large %u\n", + params->pkt.headroom); + return -1; + } + break; case ODP_POOL_TIMEOUT: commit 2c41f8f678bbb96a63dedb4e4d948c5dd779f65d Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Wed Oct 17 14:38:16 2018 +0300 linux-gen: pool: output error on pool create Change from debug to error message on pool create checks. These are slow path errors and it's helpful to see those also when not debugging. Signed-off-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_pool.c b/platform/linux-generic/odp_pool.c index 7a4a9eb99..544348189 100644 --- a/platform/linux-generic/odp_pool.c +++ b/platform/linux-generic/odp_pool.c @@ -555,17 +555,17 @@ static int check_params(odp_pool_param_t *params) switch (params->type) { case ODP_POOL_BUFFER: if (params->buf.num > capa.buf.max_num) { - ODP_DBG("buf.num too large %u\n", params->buf.num); + ODP_ERR("buf.num too large %u\n", params->buf.num); return -1; } if (params->buf.size > capa.buf.max_size) { - ODP_DBG("buf.size too large %u\n", params->buf.size); + ODP_ERR("buf.size too large %u\n", params->buf.size); return -1; } if (params->buf.align > capa.buf.max_align) { - ODP_DBG("buf.align too large %u\n", params->buf.align); + ODP_ERR("buf.align too large %u\n", params->buf.align); return -1; } @@ -573,24 +573,24 @@ static int check_params(odp_pool_param_t *params) case ODP_POOL_PACKET: if (params->pkt.len > capa.pkt.max_len) { - ODP_DBG("pkt.len too large %u\n", params->pkt.len); + ODP_ERR("pkt.len too large %u\n", params->pkt.len); return -1; } if (params->pkt.max_len > capa.pkt.max_len) { - ODP_DBG("pkt.max_len too large %u\n", + ODP_ERR("pkt.max_len too large %u\n", params->pkt.max_len); return -1; } if (params->pkt.seg_len > capa.pkt.max_seg_len) { - ODP_DBG("pkt.seg_len too large %u\n", + ODP_ERR("pkt.seg_len too large %u\n", params->pkt.seg_len); return -1; } if (params->pkt.uarea_size > capa.pkt.max_uarea_size) { - ODP_DBG("pkt.uarea_size too large %u\n", + ODP_ERR("pkt.uarea_size too large %u\n", params->pkt.uarea_size); return -1; } @@ -599,13 +599,13 @@ static int check_params(odp_pool_param_t *params) case ODP_POOL_TIMEOUT: if (params->tmo.num > capa.tmo.max_num) { - ODP_DBG("tmo.num too large %u\n", params->tmo.num); + ODP_ERR("tmo.num too large %u\n", params->tmo.num); return -1; } break; default: - ODP_DBG("bad pool type %i\n", params->type); + ODP_ERR("bad pool type %i\n", params->type); return -1; } commit 7e188187a77fa9bab56f1407eb660c324736cd34 Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Wed Oct 17 13:13:20 2018 +0300 linux-gen: config: move queue size config to scalable Only scalable queues use build time the queue size config. Move it to scalable config file. Signed-off-by: Petri Savolainen <petri.savolainen(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/platform/linux-generic/include/odp_config_internal.h b/platform/linux-generic/include/odp_config_internal.h index 8ceb325b2..7a573bd76 100644 --- a/platform/linux-generic/include/odp_config_internal.h +++ b/platform/linux-generic/include/odp_config_internal.h @@ -26,13 +26,6 @@ extern "C" { */ #define ODP_CONFIG_QUEUES 1024 -/* - * Maximum queue depth. Maximum number of elements that can be stored in a - * queue. This value is used only when the size is not explicitly provided - * during queue creation. - */ -#define CONFIG_QUEUE_SIZE 4096 - /* * Maximum number of ordered locks per queue */ diff --git a/platform/linux-generic/include/odp_schedule_scalable_config.h b/platform/linux-generic/include/odp_schedule_scalable_config.h index a84dc0724..3462d047b 100644 --- a/platform/linux-generic/include/odp_schedule_scalable_config.h +++ b/platform/linux-generic/include/odp_schedule_scalable_config.h @@ -9,6 +9,9 @@ #ifndef ODP_SCHEDULE_SCALABLE_CONFIG_H_ #define ODP_SCHEDULE_SCALABLE_CONFIG_H_ +/* Maximum number of events that can be stored in a queue */ +#define CONFIG_SCAL_QUEUE_SIZE 4096 + /* * Default scaling factor for the scheduler group * diff --git a/platform/linux-generic/odp_queue_scalable.c b/platform/linux-generic/odp_queue_scalable.c index bbc57e44d..b7ff2195f 100644 --- a/platform/linux-generic/odp_queue_scalable.c +++ b/platform/linux-generic/odp_queue_scalable.c @@ -110,7 +110,7 @@ static int queue_init(queue_entry_t *queue, const char *name, sched_elem = &queue->s.sched_elem; ring_size = param->size > 0 ? - ROUNDUP_POWER2_U32(param->size) : CONFIG_QUEUE_SIZE; + ROUNDUP_POWER2_U32(param->size) : CONFIG_SCAL_QUEUE_SIZE; strncpy(queue->s.name, 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)); @@ -212,15 +212,16 @@ static int queue_init_global(void) /* Add size of the array holding the queues */ pool_size = sizeof(queue_table_t); /* Add storage required for queues */ - pool_size += (CONFIG_QUEUE_SIZE * sizeof(odp_buffer_hdr_t *)) * - ODP_CONFIG_QUEUES; + pool_size += (CONFIG_SCAL_QUEUE_SIZE * + sizeof(odp_buffer_hdr_t *)) * ODP_CONFIG_QUEUES; + /* Add the reorder window size */ pool_size += sizeof(reorder_window_t) * ODP_CONFIG_QUEUES; /* Choose min_alloc and max_alloc such that buddy allocator is * is selected. */ min_alloc = 0; - max_alloc = CONFIG_QUEUE_SIZE * sizeof(odp_buffer_hdr_t *); + max_alloc = CONFIG_SCAL_QUEUE_SIZE * sizeof(odp_buffer_hdr_t *); queue_shm_pool = _odp_ishm_pool_create("queue_shm_pool", pool_size, min_alloc, max_alloc, commit 031802ec0c5e3e44d70aee7b19e0cd1055c9fe0b Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Thu Oct 18 09:37:47 2018 +0300 linux-gen: config: improve config file check error output It was hard to notice from log that config file version has a mismatch. This is a common error when config file version has updated, but a user has not updated his own config file. Improve error check output, so that failure reason is better highlighted. Signed-off-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_libconfig.c b/platform/linux-generic/odp_libconfig.c index baf825bbe..014409e2b 100644 --- a/platform/linux-generic/odp_libconfig.c +++ b/platform/linux-generic/odp_libconfig.c @@ -21,10 +21,12 @@ int _odp_libconfig_init_global(void) const char *filename; const char *vers; const char *vers_rt; - const char *ipml; - const char *ipml_rt; + const char *impl; + const char *impl_rt; config_t *config = &odp_global_ro.libconfig_default; config_t *config_rt = &odp_global_ro.libconfig_runtime; + const char *impl_field = "odp_implementation"; + const char *vers_field = "config_file_version"; config_init(config); config_init(config_rt); @@ -40,34 +42,45 @@ int _odp_libconfig_init_global(void) if (filename == NULL) return 0; - ODP_PRINT("CONFIG FILE: %s\n", filename); + ODP_PRINT("ODP CONFIG FILE: %s\n", filename); if (!config_read_file(config_rt, filename)) { - ODP_ERR("Failed to read config file: %s(%d): %s\n", - config_error_file(config_rt), - config_error_line(config_rt), - config_error_text(config_rt)); + ODP_PRINT(" ERROR: failed to read config file: %s(%d): %s\n\n", + config_error_file(config_rt), + config_error_line(config_rt), + config_error_text(config_rt)); goto fail; } /* Check runtime configuration's implementation name and version */ - if (!config_lookup_string(config, "odp_implementation", &ipml) || - !config_lookup_string(config_rt, "odp_implementation", &ipml_rt)) { - ODP_ERR("Configuration missing 'odp_implementation' field\n"); + if (!config_lookup_string(config, impl_field, &impl) || + !config_lookup_string(config_rt, impl_field, &impl_rt)) { + ODP_PRINT(" ERROR: missing mandatory field: %s\n\n", + impl_field); goto fail; } - if (!config_lookup_string(config, "config_file_version", &vers) || - !config_lookup_string(config_rt, "config_file_version", &vers_rt)) { - ODP_ERR("Configuration missing 'config_file_version' field\n"); + if (!config_lookup_string(config, vers_field, &vers) || + !config_lookup_string(config_rt, vers_field, &vers_rt)) { + ODP_PRINT(" ERROR: missing mandatory field: %s\n\n", + vers_field); goto fail; } - if (strcmp(vers, vers_rt) || strcmp(ipml, ipml_rt)) { - ODP_ERR("Runtime configuration mismatch\n"); + if (strcmp(impl, impl_rt)) { + ODP_PRINT(" ERROR: ODP implementation name mismatch:\n" + " Expected: \"%s\"\n" + " Found: \"%s\"\n\n", impl, impl_rt); + goto fail; + } + if (strcmp(vers, vers_rt)) { + ODP_PRINT(" ERROR: config file version number mismatch:\n" + " Expected: \"%s\"\n" + " Found: \"%s\"\n\n", vers, vers_rt); goto fail; } return 0; fail: + ODP_ERR("Config file failure\n"); config_destroy(config); config_destroy(config_rt); return -1; ----------------------------------------------------------------------- Summary of changes: config/odp-linux-generic.conf | 12 +++- .../linux-generic/include/odp_config_internal.h | 9 +-- platform/linux-generic/include/odp_pool_internal.h | 5 ++ .../include/odp_schedule_scalable_config.h | 3 + platform/linux-generic/odp_libconfig.c | 43 ++++++++----- platform/linux-generic/odp_pool.c | 74 ++++++++++++++++++---- platform/linux-generic/odp_queue_scalable.c | 9 +-- platform/linux-generic/test/process-mode.conf | 2 +- 8 files changed, 116 insertions(+), 41 deletions(-) hooks/post-receive --

7 years, 3 months

[COMMIT] branch master updated. v1.19.0.2_tigermoth-192-g690bacc61

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 690bacc61c55e4f8f3195faaac068da83790ebf4 (commit) from 0287f1432eda1cf3e7fb8323e0fda838cf744816 (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 690bacc61c55e4f8f3195faaac068da83790ebf4 Author: Matias Elo <matias.elo(a)nokia.com> Date: Thu Oct 18 09:49:01 2018 +0300 linux-gen: pool: increase minimum packet segment length Some DPDK NICs need at least 2176 byte buffers (2048B + headroom) to not segment standard ethernet frames. Increase minimum segment length to avoid this and add matching check to zero-copy dpdk pktio pool create. Reported-by: P. Gyanesh Kumar Patra <pgyanesh.patra(a)gmail.com> Signed-off-by: Matias Elo <matias.elo(a)nokia.com> Tested-by: P. Gyanesh Kumar Patra <pgyanesh.patra(a)gmail.com> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/platform/linux-generic/include/odp_config_internal.h b/platform/linux-generic/include/odp_config_internal.h index a94012acc..8ceb325b2 100644 --- a/platform/linux-generic/include/odp_config_internal.h +++ b/platform/linux-generic/include/odp_config_internal.h @@ -125,8 +125,8 @@ extern "C" { * defined segment length (seg_len in odp_pool_param_t) will be rounded up into * this value. */ -#define CONFIG_PACKET_SEG_LEN_MIN ((2 * 1024) - \ - CONFIG_PACKET_HEADROOM - \ +#define CONFIG_PACKET_SEG_LEN_MIN ((2 * 1024) + \ + CONFIG_PACKET_HEADROOM + \ CONFIG_PACKET_TAILROOM) /* Maximum number of shared memory blocks. diff --git a/platform/linux-generic/pktio/dpdk.c b/platform/linux-generic/pktio/dpdk.c index 23f39445e..4253451f5 100644 --- a/platform/linux-generic/pktio/dpdk.c +++ b/platform/linux-generic/pktio/dpdk.c @@ -327,6 +327,13 @@ static struct rte_mempool *mbuf_pool_create(const char *name, goto fail; } + if (pool_entry->seg_len < RTE_MBUF_DEFAULT_BUF_SIZE) { + ODP_ERR("Some NICs need at least %dB buffers to not segment " + "standard ethernet frames. Increase pool seg_len.\n", + RTE_MBUF_DEFAULT_BUF_SIZE); + goto fail; + } + total_size = rte_mempool_calc_obj_size(elt_size, MEMPOOL_F_NO_SPREAD, &sz); if (total_size != pool_entry->block_size) { ----------------------------------------------------------------------- Summary of changes: platform/linux-generic/include/odp_config_internal.h | 4 ++-- platform/linux-generic/pktio/dpdk.c | 7 +++++++ 2 files changed, 9 insertions(+), 2 deletions(-) hooks/post-receive --

7 years, 3 months

[COMMIT] branch master updated. v1.19.0.2_tigermoth-191-g0287f1432

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 0287f1432eda1cf3e7fb8323e0fda838cf744816 (commit) via f8a9b873ffd6b1a512eb87ebc3ab183bc71157b0 (commit) via df04f4ccb940fea8cf3d7b5d41e31a381bb0ec08 (commit) from 570758a22fd0d6e2b2a73eb8ed0a8360a5b0ef32 (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 0287f1432eda1cf3e7fb8323e0fda838cf744816 Author: Matias Elo <matias.elo(a)nokia.com> Date: Thu Oct 18 13:23:56 2018 +0300 example: generator: remove print from packet tx loop The printf() would flood output when using small tx interval. Increase global stats print interval instead. Signed-off-by: Matias Elo <matias.elo(a)nokia.com> Reviewed-by: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/example/generator/odp_generator.c b/example/generator/odp_generator.c index 8544da554..54194c664 100644 --- a/example/generator/odp_generator.c +++ b/example/generator/odp_generator.c @@ -31,6 +31,7 @@ #define MAX_UDP_TX_BURST 512 #define DEFAULT_RX_BURST 32 #define MAX_RX_BURST 512 +#define STATS_INTERVAL 10 /* Interval between stats prints (sec) */ #define APPL_MODE_UDP 0 /**< UDP mode */ #define APPL_MODE_PING 1 /**< ping mode */ @@ -775,14 +776,9 @@ static int gen_send_thread(void *arg) counters->ctr_pkt_snd += pkt_array_size - burst_size; - if (args->appl.interval != 0) { - printf(" [%02i] send pkt no:%ju seq %ju\n", - thr, - counters->ctr_seq, - counters->ctr_seq % 0xffff); + if (args->appl.interval != 0) odp_time_wait_ns((uint64_t)args->appl.interval * ODP_TIME_MSEC_IN_NS); - } counters->ctr_seq += seq_step; } @@ -1014,7 +1010,7 @@ static void print_global_stats(int num_workers) uint64_t pkts_rcv = 0, pkts_rcv_prev = 0; uint64_t pps_rcv = 0, maximum_pps_rcv = 0; uint64_t stall, pkts_snd_drop; - int verbose_interval = 20, i; + int verbose_interval = STATS_INTERVAL, i; odp_thrmask_t thrd_mask; odp_barrier_wait(&args->barrier); commit f8a9b873ffd6b1a512eb87ebc3ab183bc71157b0 Author: Matias Elo <matias.elo(a)nokia.com> Date: Mon Oct 15 12:49:47 2018 +0300 example: generator: use odp_wait_time_ns() instead of timers Simplify code by using odp_wait_time_ns() instead of timers. Signed-off-by: Matias Elo <matias.elo(a)nokia.com> Reviewed-by: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/example/generator/odp_generator.c b/example/generator/odp_generator.c index 4e06115c2..8544da554 100644 --- a/example/generator/odp_generator.c +++ b/example/generator/odp_generator.c @@ -132,10 +132,6 @@ typedef struct { } rx; }; odp_pool_t pool; /**< Pool for packet IO */ - odp_timer_pool_t tp; /**< Timer pool handle */ - odp_queue_t tq; /**< Queue for timeouts */ - odp_timer_t tim; /**< Timer handle */ - odp_timeout_t tmo_ev; /**< Timeout event */ int mode; /**< Thread mode */ } thread_args_t; @@ -171,27 +167,6 @@ static void usage(char *progname); static int scan_ip(char *buf, unsigned int *paddr); static void print_global_stats(int num_workers); -/** - * Sleep for the specified amount of milliseconds - * Use ODP timer, busy wait until timer expired and timeout event received - */ -static void millisleep(uint32_t ms, - odp_timer_pool_t tp, - odp_timer_t tim, - odp_queue_t q, - odp_timeout_t tmo) -{ - uint64_t ticks = odp_timer_ns_to_tick(tp, 1000000ULL * ms); - odp_event_t ev = odp_timeout_to_event(tmo); - int rc = odp_timer_set_rel(tim, ticks, &ev); - - if (rc != ODP_TIMER_SUCCESS) - EXAMPLE_ABORT("odp_timer_set_rel() failed\n"); - /* Spin waiting for timeout event */ - while ((ev = odp_queue_deq(q)) == ODP_EVENT_INVALID) - (void)0; -} - /** * Scan ip * Parse ip address. @@ -805,11 +780,8 @@ static int gen_send_thread(void *arg) thr, counters->ctr_seq, counters->ctr_seq % 0xffff); - millisleep(args->appl.interval, - thr_args->tp, - thr_args->tim, - thr_args->tq, - thr_args->tmo_ev); + odp_time_wait_ns((uint64_t)args->appl.interval * + ODP_TIME_MSEC_IN_NS); } counters->ctr_seq += seq_step; } @@ -1131,15 +1103,9 @@ int main(int argc, char *argv[]) odp_cpumask_t cpumask; char cpumaskstr[ODP_CPUMASK_STR_SIZE]; odp_pool_param_t params; - odp_timer_pool_param_t tparams; - odp_timer_pool_t tp; - odp_pool_t tmop; - odp_queue_t tq; - odp_event_t ev; interface_t *ifs; odp_instance_t instance; odph_odpthread_params_t thr_params; - odp_timer_capability_t timer_capa; /* Init ODP before calling anything else */ if (odp_init_global(&instance, NULL, NULL)) { @@ -1227,36 +1193,6 @@ int main(int argc, char *argv[]) } odp_pool_print(pool); - /* Create timer pool */ - if (odp_timer_capability(ODP_CLOCK_CPU, &timer_capa)) { - EXAMPLE_ERR("Error: get timer capacity failed.\n"); - exit(EXIT_FAILURE); - } - tparams.res_ns = MAX(1 * ODP_TIME_MSEC_IN_NS, - timer_capa.highest_res_ns); - tparams.min_tmo = 0; - tparams.max_tmo = 10000 * ODP_TIME_SEC_IN_NS; - tparams.num_timers = num_workers; /* One timer per worker */ - tparams.priv = 0; /* Shared */ - tparams.clk_src = ODP_CLOCK_CPU; - tp = odp_timer_pool_create("timer_pool", &tparams); - if (tp == ODP_TIMER_POOL_INVALID) { - EXAMPLE_ERR("Timer pool create failed.\n"); - exit(EXIT_FAILURE); - } - odp_timer_pool_start(); - - /* Create timeout pool */ - odp_pool_param_init(&params); - params.tmo.num = tparams.num_timers; /* One timeout per timer */ - params.type = ODP_POOL_TIMEOUT; - - tmop = odp_pool_create("timeout_pool", &params); - if (tmop == ODP_POOL_INVALID) { - EXAMPLE_ERR("Error: timeout pool create failed.\n"); - exit(EXIT_FAILURE); - } - ifs = malloc(sizeof(interface_t) * args->appl.if_count); for (i = 0; i < args->appl.if_count; ++i) { @@ -1303,27 +1239,10 @@ int main(int argc, char *argv[]) cpu_first = odp_cpumask_first(&cpumask); odp_cpumask_set(&cpu_mask, cpu_first); - tq = odp_queue_create("", NULL); - if (tq == ODP_QUEUE_INVALID) { - EXAMPLE_ERR("queue_create failed\n"); - abort(); - } thr_args = &args->thread[PING_THR_RX]; if (!args->appl.sched) thr_args->rx.pktin = ifs[0].pktin[0]; thr_args->pool = pool; - thr_args->tp = tp; - thr_args->tq = tq; - thr_args->tim = odp_timer_alloc(tp, tq, NULL); - if (thr_args->tim == ODP_TIMER_INVALID) { - EXAMPLE_ERR("timer_alloc failed\n"); - abort(); - } - thr_args->tmo_ev = odp_timeout_alloc(tmop); - if (thr_args->tmo_ev == ODP_TIMEOUT_INVALID) { - EXAMPLE_ERR("timeout_alloc failed\n"); - abort(); - } thr_args->mode = args->appl.mode; memset(&thr_params, 0, sizeof(thr_params)); @@ -1338,27 +1257,10 @@ int main(int argc, char *argv[]) odph_odpthreads_create(&thread_tbl[PING_THR_RX], &cpu_mask, &thr_params); - tq = odp_queue_create("", NULL); - if (tq == ODP_QUEUE_INVALID) { - EXAMPLE_ERR("queue_create failed\n"); - abort(); - } thr_args = &args->thread[PING_THR_TX]; thr_args->tx.pktout = ifs[0].pktout[0]; thr_args->tx.pktout_cfg = &ifs[0].config.pktout; thr_args->pool = pool; - thr_args->tp = tp; - thr_args->tq = tq; - thr_args->tim = odp_timer_alloc(tp, tq, NULL); - if (thr_args->tim == ODP_TIMER_INVALID) { - EXAMPLE_ERR("timer_alloc failed\n"); - abort(); - } - thr_args->tmo_ev = odp_timeout_alloc(tmop); - if (thr_args->tmo_ev == ODP_TIMEOUT_INVALID) { - EXAMPLE_ERR("timeout_alloc failed\n"); - abort(); - } thr_args->mode = args->appl.mode; cpu_next = odp_cpumask_next(&cpumask, cpu_first); odp_cpumask_zero(&cpu_mask); @@ -1427,24 +1329,7 @@ int main(int argc, char *argv[]) args->thread[i].counters.ctr_seq = start_seq; } - tq = odp_queue_create("", NULL); - if (tq == ODP_QUEUE_INVALID) { - EXAMPLE_ERR("queue_create failed\n"); - abort(); - } args->thread[i].pool = pool; - args->thread[i].tp = tp; - args->thread[i].tq = tq; - args->thread[i].tim = odp_timer_alloc(tp, tq, NULL); - if (args->thread[i].tim == ODP_TIMER_INVALID) { - EXAMPLE_ERR("timer_alloc failed\n"); - abort(); - } - args->thread[i].tmo_ev = odp_timeout_alloc(tmop); - if (args->thread[i].tmo_ev == ODP_TIMEOUT_INVALID) { - EXAMPLE_ERR("timeout_alloc failed\n"); - abort(); - } args->thread[i].mode = args->appl.mode; if (args->appl.mode == APPL_MODE_UDP) { @@ -1484,22 +1369,6 @@ int main(int argc, char *argv[]) for (i = 0; i < args->appl.if_count; ++i) odp_pktio_stop(ifs[i].pktio); - for (i = 0; i < num_workers; ++i) { - odp_timer_cancel(args->thread[i].tim, &ev); - odp_timer_free(args->thread[i].tim); - odp_timeout_free(args->thread[i].tmo_ev); - } - - for (i = 0; i < num_workers; ++i) { - while (1) { - ev = odp_queue_deq(args->thread[i].tq); - if (ev == ODP_EVENT_INVALID) - break; - odp_event_free(ev); - } - odp_queue_destroy(args->thread[i].tq); - } - for (i = 0; i < args->appl.if_count; ++i) odp_pktio_close(ifs[i].pktio); free(ifs); @@ -1507,9 +1376,6 @@ int main(int argc, char *argv[]) free(args->appl.if_str); if (0 != odp_pool_destroy(pool)) fprintf(stderr, "unable to destroy pool \"pool\"\n"); - odp_timer_pool_destroy(tp); - if (0 != odp_pool_destroy(tmop)) - fprintf(stderr, "unable to destroy pool \"tmop\"\n"); if (0 != odp_shm_free(shm)) fprintf(stderr, "unable to free \"shm\"\n"); odp_term_local(); commit df04f4ccb940fea8cf3d7b5d41e31a381bb0ec08 Author: Matias Elo <matias.elo(a)nokia.com> Date: Wed Oct 17 10:40:53 2018 +0300 example: stop and close pktio devices on exit Stop and close used pktio devices on exit to free used resources. Signed-off-by: Matias Elo <matias.elo(a)nokia.com> Reviewed-by: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/example/ipsec/odp_ipsec.c b/example/ipsec/odp_ipsec.c index ee63451c2..b45a82d74 100644 --- a/example/ipsec/odp_ipsec.c +++ b/example/ipsec/odp_ipsec.c @@ -1325,6 +1325,20 @@ main(int argc, char *argv[]) odph_odpthreads_join(thread_tbl); } + /* Stop and close used pktio devices */ + for (i = 0; i < global->appl.if_count; i++) { + odp_pktio_t pktio = odp_pktio_lookup(global->appl.if_names[i]); + + if (pktio == ODP_PKTIO_INVALID) + continue; + + if (odp_pktio_stop(pktio) || odp_pktio_close(pktio)) { + EXAMPLE_ERR("Error: failed to close pktio %s\n", + global->appl.if_names[i]); + exit(EXIT_FAILURE); + } + } + free(global->appl.if_names); free(global->appl.if_str); diff --git a/example/ipsec_api/odp_ipsec.c b/example/ipsec_api/odp_ipsec.c index b402002c7..d72109cf5 100644 --- a/example/ipsec_api/odp_ipsec.c +++ b/example/ipsec_api/odp_ipsec.c @@ -1025,6 +1025,20 @@ main(int argc, char *argv[]) odph_odpthreads_join(thread_tbl); } + /* Stop and close used pktio devices */ + for (i = 0; i < global->appl.if_count; i++) { + odp_pktio_t pktio = odp_pktio_lookup(global->appl.if_names[i]); + + if (pktio == ODP_PKTIO_INVALID) + continue; + + if (odp_pktio_stop(pktio) || odp_pktio_close(pktio)) { + EXAMPLE_ERR("Error: failed to close pktio %s\n", + global->appl.if_names[i]); + exit(EXIT_FAILURE); + } + } + free(global->appl.if_names); free(global->appl.if_str); diff --git a/example/ipsec_offload/odp_ipsec_offload.c b/example/ipsec_offload/odp_ipsec_offload.c index 89b9dddf1..90b3f6409 100644 --- a/example/ipsec_offload/odp_ipsec_offload.c +++ b/example/ipsec_offload/odp_ipsec_offload.c @@ -627,6 +627,20 @@ main(int argc, char *argv[]) &thr_params); odph_odpthreads_join(thread_tbl); + /* Stop and close used pktio devices */ + for (i = 0; i < global->appl.if_count; i++) { + odp_pktio_t pktio = odp_pktio_lookup(global->appl.if_names[i]); + + if (pktio == ODP_PKTIO_INVALID) + continue; + + if (odp_pktio_stop(pktio) || odp_pktio_close(pktio)) { + EXAMPLE_ERR("Error: failed to close pktio %s\n", + global->appl.if_names[i]); + exit(EXIT_FAILURE); + } + } + free(global->appl.if_names); free(global->appl.if_str); diff --git a/example/l2fwd_simple/odp_l2fwd_simple.c b/example/l2fwd_simple/odp_l2fwd_simple.c index 6835e5bb1..8daeec4c0 100644 --- a/example/l2fwd_simple/odp_l2fwd_simple.c +++ b/example/l2fwd_simple/odp_l2fwd_simple.c @@ -240,6 +240,15 @@ int main(int argc, char **argv) odph_odpthreads_create(thd, &cpumask, &thr_params); odph_odpthreads_join(thd); + if (odp_pktio_stop(global->if0) || odp_pktio_close(global->if0)) { + printf("Error: failed to close interface %s\n", argv[1]); + exit(EXIT_FAILURE); + } + if (odp_pktio_stop(global->if1) || odp_pktio_close(global->if1)) { + printf("Error: failed to close interface %s\n", argv[2]); + exit(EXIT_FAILURE); + } + if (odp_pool_destroy(pool)) { printf("Error: pool destroy\n"); exit(EXIT_FAILURE); diff --git a/example/l3fwd/odp_l3fwd.c b/example/l3fwd/odp_l3fwd.c index f72cf3734..708c4df90 100644 --- a/example/l3fwd/odp_l3fwd.c +++ b/example/l3fwd/odp_l3fwd.c @@ -1126,6 +1126,16 @@ int main(int argc, char **argv) for (i = 0; i < nb_worker; i++) odph_odpthreads_join(&thread_tbl[i]); + /* Stop and close used pktio devices */ + for (i = 0; i < args->if_count; i++) { + odp_pktio_t pktio = global->l3fwd_pktios[i].pktio; + + if (odp_pktio_stop(pktio) || odp_pktio_close(pktio)) { + printf("Error: failed to close pktio\n"); + exit(EXIT_FAILURE); + } + } + /* if_names share a single buffer, so only one free */ free(args->if_names[0]); diff --git a/example/switch/odp_switch.c b/example/switch/odp_switch.c index dab60e833..a67fa180f 100644 --- a/example/switch/odp_switch.c +++ b/example/switch/odp_switch.c @@ -1019,6 +1019,16 @@ int main(int argc, char **argv) for (i = 0; i < num_workers; ++i) odph_odpthreads_join(&thread_tbl[i]); + /* Stop and close used pktio devices */ + for (i = 0; i < if_count; i++) { + odp_pktio_t pktio = gbl_args->pktios[i].pktio; + + if (odp_pktio_stop(pktio) || odp_pktio_close(pktio)) { + printf("Error: failed to close pktio\n"); + exit(EXIT_FAILURE); + } + } + free(gbl_args->appl.if_names); free(gbl_args->appl.if_str); diff --git a/platform/linux-generic/test/mmap_vlan_ins/mmap_vlan_ins.c b/platform/linux-generic/test/mmap_vlan_ins/mmap_vlan_ins.c index 5066eede7..faa51dad3 100644 --- a/platform/linux-generic/test/mmap_vlan_ins/mmap_vlan_ins.c +++ b/platform/linux-generic/test/mmap_vlan_ins/mmap_vlan_ins.c @@ -203,6 +203,15 @@ int main(int argc, char **argv) ret = global->g_ret; + if (odp_pktio_stop(global->if0) || odp_pktio_close(global->if0)) { + printf("Error: failed to close interface %s\n", argv[1]); + exit(EXIT_FAILURE); + } + if (odp_pktio_stop(global->if1) || odp_pktio_close(global->if1)) { + printf("Error: failed to close interface %s\n", argv[2]); + exit(EXIT_FAILURE); + } + if (odp_pool_destroy(pool)) { printf("Error: pool destroy\n"); exit(EXIT_FAILURE); ----------------------------------------------------------------------- Summary of changes: example/generator/odp_generator.c | 148 +-------------------- example/ipsec/odp_ipsec.c | 14 ++ example/ipsec_api/odp_ipsec.c | 14 ++ example/ipsec_offload/odp_ipsec_offload.c | 14 ++ example/l2fwd_simple/odp_l2fwd_simple.c | 9 ++ example/l3fwd/odp_l3fwd.c | 10 ++ example/switch/odp_switch.c | 10 ++ .../test/mmap_vlan_ins/mmap_vlan_ins.c | 9 ++ 8 files changed, 85 insertions(+), 143 deletions(-) hooks/post-receive --

7 years, 3 months

[COMMIT] branch master updated. v1.19.0.2_tigermoth-188-g570758a22

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 570758a22fd0d6e2b2a73eb8ed0a8360a5b0ef32 (commit) via 4a877ef354b51189acc631c9b0ef669cf16045cc (commit) from dc6c26b9ef60ac0f10dca34b9f84ad440cf2c39e (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 570758a22fd0d6e2b2a73eb8ed0a8360a5b0ef32 Author: Matias Elo <matias.elo(a)nokia.com> Date: Tue Oct 2 14:13:35 2018 +0300 linux-gen: ring: allocate global data from shm Enables using ipc in process mode. Signed-off-by: Matias Elo <matias.elo(a)nokia.com> Reviewed-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_packet_io_ring_internal.h b/platform/linux-generic/include/odp_packet_io_ring_internal.h index 889a65597..6b4e06a42 100644 --- a/platform/linux-generic/include/odp_packet_io_ring_internal.h +++ b/platform/linux-generic/include/odp_packet_io_ring_internal.h @@ -398,9 +398,14 @@ int _ring_mc_dequeue_burst(_ring_t *r, void **obj_table, unsigned n); void _ring_list_dump(void); /** - * initialise ring tailq + * Initialize ring tailq */ -void _ring_tailq_init(void); +int _ring_tailq_init(void); + +/** + * Terminate ring tailq + */ +int _ring_tailq_term(void); #ifdef __cplusplus } diff --git a/platform/linux-generic/pktio/ipc.c b/platform/linux-generic/pktio/ipc.c index 74af0b975..fc5654432 100644 --- a/platform/linux-generic/pktio/ipc.c +++ b/platform/linux-generic/pktio/ipc.c @@ -826,9 +826,13 @@ static int ipc_close(pktio_entry_t *pktio_entry) static int ipc_pktio_init_global(void) { - _ring_tailq_init(); - ODP_PRINT("PKTIO: initialized ipc interface.\n"); - return 0; + ODP_DBG("PKTIO: initializing ipc interface.\n"); + return _ring_tailq_init(); +} + +static int ipc_pktio_term_global(void) +{ + return _ring_tailq_term(); } const pktio_if_ops_t ipc_pktio_ops = { @@ -836,7 +840,7 @@ const pktio_if_ops_t ipc_pktio_ops = { .print = NULL, .init_global = ipc_pktio_init_global, .init_local = NULL, - .term = NULL, + .term = ipc_pktio_term_global, .open = ipc_pktio_open, .close = ipc_close, .recv = ipc_pktio_recv, diff --git a/platform/linux-generic/pktio/ring.c b/platform/linux-generic/pktio/ring.c index 488dc1a1f..6a354cb1e 100644 --- a/platform/linux-generic/pktio/ring.c +++ b/platform/linux-generic/pktio/ring.c @@ -80,13 +80,21 @@ #include <odp_packet_io_ring_internal.h> #include <odp_errno_define.h> #include <odp_global_data.h> +#include <odp_shm_internal.h> #include <odp/api/plat/cpu_inlines.h> -static TAILQ_HEAD(, _ring) odp_ring_list; - #define RING_VAL_IS_POWER_2(x) ((((x) - 1) & (x)) == 0) +typedef struct { + TAILQ_HEAD(, _ring) ring_list; + /* Rings tailq lock */ + odp_rwlock_t qlock; + odp_shm_t shm; +} global_data_t; + +static global_data_t *global; + /* * the enqueue of pointers on the ring. */ @@ -149,13 +157,37 @@ static TAILQ_HEAD(, _ring) odp_ring_list; } \ } while (0) -static odp_rwlock_t qlock; /* rings tailq lock */ -/* init tailq_ring */ -void _ring_tailq_init(void) +/* Initialize tailq_ring */ +int _ring_tailq_init(void) +{ odp_shm_t shm; + + /* Allocate globally shared memory */ + shm = odp_shm_reserve("_odp_ring_global", sizeof(global_data_t), + ODP_CACHE_LINE_SIZE, _ODP_SHM_NO_HP); + if (ODP_SHM_INVALID == shm) { + ODP_ERR("Shm reserve failed for pktio ring\n"); + return -1; + } + + global = odp_shm_addr(shm); + memset(global, 0, sizeof(global_data_t)); + global->shm = shm; + + TAILQ_INIT(&global->ring_list); + odp_rwlock_init(&global->qlock); + + return 0; +} + +/* Terminate tailq_ring */ +int _ring_tailq_term(void) { - TAILQ_INIT(&odp_ring_list); - odp_rwlock_init(&qlock); + if (odp_shm_free(global->shm)) { + ODP_ERR("Shm free failed for pktio ring\n"); + return -1; + } + return 0; } /* create the ring */ @@ -187,7 +219,7 @@ _ring_create(const char *name, unsigned count, unsigned flags) snprintf(ring_name, sizeof(ring_name), "%s", name); ring_size = count * sizeof(void *) + sizeof(_ring_t); - odp_rwlock_write_lock(&qlock); + odp_rwlock_write_lock(&global->qlock); /* reserve a memory zone for this ring.*/ shm = odp_shm_reserve(ring_name, ring_size, ODP_CACHE_LINE_SIZE, shm_flag); @@ -208,13 +240,13 @@ _ring_create(const char *name, unsigned count, unsigned flags) r->cons.tail = 0; if (!(flags & _RING_NO_LIST)) - TAILQ_INSERT_TAIL(&odp_ring_list, r, next); + TAILQ_INSERT_TAIL(&global->ring_list, r, next); } else { __odp_errno = ENOMEM; ODP_ERR("Cannot reserve memory\n"); } - odp_rwlock_write_unlock(&qlock); + odp_rwlock_write_unlock(&global->qlock); return r; } @@ -225,10 +257,10 @@ int _ring_destroy(const char *name) if (shm != ODP_SHM_INVALID) { _ring_t *r = odp_shm_addr(shm); - odp_rwlock_write_lock(&qlock); + odp_rwlock_write_lock(&global->qlock); if (!(r->flags & _RING_NO_LIST)) - TAILQ_REMOVE(&odp_ring_list, r, next); - odp_rwlock_write_unlock(&qlock); + TAILQ_REMOVE(&global->ring_list, r, next); + odp_rwlock_write_unlock(&global->qlock); return odp_shm_free(shm); } @@ -442,13 +474,13 @@ void _ring_list_dump(void) { const _ring_t *mp = NULL; - odp_rwlock_read_lock(&qlock); + odp_rwlock_read_lock(&global->qlock); - TAILQ_FOREACH(mp, &odp_ring_list, next) { + TAILQ_FOREACH(mp, &global->ring_list, next) { _ring_dump(mp); } - odp_rwlock_read_unlock(&qlock); + odp_rwlock_read_unlock(&global->qlock); } /* search a ring from its name */ @@ -456,12 +488,12 @@ _ring_t *_ring_lookup(const char *name) { _ring_t *r; - odp_rwlock_read_lock(&qlock); - TAILQ_FOREACH(r, &odp_ring_list, next) { + odp_rwlock_read_lock(&global->qlock); + TAILQ_FOREACH(r, &global->ring_list, next) { if (strncmp(name, r->name, _RING_NAMESIZE) == 0) break; } - odp_rwlock_read_unlock(&qlock); + odp_rwlock_read_unlock(&global->qlock); return r; } commit 4a877ef354b51189acc631c9b0ef669cf16045cc Author: Matias Elo <matias.elo(a)nokia.com> Date: Tue Oct 2 11:20:23 2018 +0300 linux-gen: sched scalable: allocate global data from shm Enables using scalable scheduler in process mode. Signed-off-by: Matias Elo <matias.elo(a)nokia.com> Reviewed-by: Bill Fischofer <bill.fischofer(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/platform/linux-generic/odp_schedule_scalable.c b/platform/linux-generic/odp_schedule_scalable.c index 9acb997a1..a21728d8b 100644 --- a/platform/linux-generic/odp_schedule_scalable.c +++ b/platform/linux-generic/odp_schedule_scalable.c @@ -32,6 +32,7 @@ #include <odp_schedule_if.h> #include <odp_bitset.h> #include <odp_packet_io_internal.h> +#include <odp_shm_internal.h> #include <odp_timer_internal.h> #include <limits.h> @@ -46,8 +47,6 @@ #define FLAG_PKTIN 0x80 -static _odp_ishm_pool_t *sched_shm_pool; - ODP_STATIC_ASSERT(ODP_SCHED_PRIO_LOWEST == (ODP_SCHED_PRIO_NUM - 2), "lowest_prio_does_not_match_with_num_prios"); @@ -58,22 +57,24 @@ ODP_STATIC_ASSERT((ODP_SCHED_PRIO_NORMAL > 0) && ODP_STATIC_ASSERT(CHECK_IS_POWER2(ODP_CONFIG_QUEUES), "Number_of_queues_is_not_power_of_two"); -/* - * Scheduler group related variables. - */ -/* Currently used scheduler groups */ -static sched_group_mask_t sg_free; -static sched_group_t *sg_vec[MAX_SCHED_GROUP]; -/* Group lock for MT-safe APIs */ -static odp_spinlock_t sched_grp_lock; - #define SCHED_GROUP_JOIN 0 #define SCHED_GROUP_LEAVE 1 -/* - * Per thread state - */ -static sched_scalable_thread_state_t thread_state[MAXTHREADS]; +typedef struct { + odp_shm_t shm; + _odp_ishm_pool_t *sched_shm_pool; + /** Currently used scheduler groups */ + sched_group_mask_t sg_free; + sched_group_t *sg_vec[MAX_SCHED_GROUP]; + /** Group lock for MT-safe APIs */ + odp_spinlock_t sched_grp_lock; + /** Per thread state */ + sched_scalable_thread_state_t thread_state[MAXTHREADS]; + uint16_t poll_count[ODP_CONFIG_PKTIO_ENTRIES]; +} sched_global_t; + +static sched_global_t *global; + __thread sched_scalable_thread_state_t *sched_ts; static int thread_state_init(int tidx) @@ -82,7 +83,7 @@ static int thread_state_init(int tidx) uint32_t i; ODP_ASSERT(tidx < MAXTHREADS); - ts = &thread_state[tidx]; + ts = &global->thread_state[tidx]; ts->atomq = NULL; ts->src_schedq = NULL; ts->rctx = NULL; @@ -523,9 +524,9 @@ static void signal_threads_add(sched_group_t *sg, uint32_t sgi, uint32_t prio) /* Notify the thread about membership in this * group/priority. */ - atom_bitset_set(&thread_state[thr].sg_wanted[prio], + atom_bitset_set(&global->thread_state[thr].sg_wanted[prio], sgi, __ATOMIC_RELEASE); - __atomic_store_n(&thread_state[thr].sg_sem, 1, + __atomic_store_n(&global->thread_state[thr].sg_sem, 1, __ATOMIC_RELEASE); } } @@ -537,11 +538,11 @@ sched_queue_t *sched_queue_add(odp_schedule_group_t grp, uint32_t prio) uint32_t x; ODP_ASSERT(grp >= 0 && grp < (odp_schedule_group_t)MAX_SCHED_GROUP); - ODP_ASSERT((sg_free & (1ULL << grp)) == 0); + ODP_ASSERT((global->sg_free & (1ULL << grp)) == 0); ODP_ASSERT(prio < ODP_SCHED_PRIO_NUM); sgi = grp; - sg = sg_vec[sgi]; + sg = global->sg_vec[sgi]; /* Use xcount to spread queues over the xfactor schedq's * per priority. @@ -563,11 +564,11 @@ static uint32_t sched_pktin_add(odp_schedule_group_t grp, uint32_t prio) sched_group_t *sg; ODP_ASSERT(grp >= 0 && grp < (odp_schedule_group_t)MAX_SCHED_GROUP); - ODP_ASSERT((sg_free & (1ULL << grp)) == 0); + ODP_ASSERT((global->sg_free & (1ULL << grp)) == 0); ODP_ASSERT(prio < ODP_SCHED_PRIO_NUM); sgi = grp; - sg = sg_vec[sgi]; + sg = global->sg_vec[sgi]; (void)sched_queue_add(grp, ODP_SCHED_PRIO_PKTIN); return (ODP_SCHED_PRIO_PKTIN - prio) * sg->xfactor; @@ -584,9 +585,9 @@ static void signal_threads_rem(sched_group_t *sg, uint32_t sgi, uint32_t prio) /* Notify the thread about membership in this * group/priority. */ - atom_bitset_clr(&thread_state[thr].sg_wanted[prio], + atom_bitset_clr(&global->thread_state[thr].sg_wanted[prio], sgi, __ATOMIC_RELEASE); - __atomic_store_n(&thread_state[thr].sg_sem, 1, + __atomic_store_n(&global->thread_state[thr].sg_sem, 1, __ATOMIC_RELEASE); } } @@ -598,11 +599,11 @@ void sched_queue_rem(odp_schedule_group_t grp, uint32_t prio) uint32_t x; ODP_ASSERT(grp >= 0 && grp < (odp_schedule_group_t)MAX_SCHED_GROUP); - ODP_ASSERT((sg_free & (1ULL << grp)) == 0); + ODP_ASSERT((global->sg_free & (1ULL << grp)) == 0); ODP_ASSERT(prio < ODP_SCHED_PRIO_NUM); sgi = grp; - sg = sg_vec[sgi]; + sg = global->sg_vec[sgi]; x = __atomic_sub_fetch(&sg->xcount[prio], 1, __ATOMIC_RELAXED); if (x == 0) { @@ -631,7 +632,7 @@ static void update_sg_add(sched_scalable_thread_state_t *ts, added = bitset_andn(sg_wanted, ts->sg_actual[p]); while (!bitset_is_null(added)) { sgi = bitset_ffs(added) - 1; - sg = sg_vec[sgi]; + sg = global->sg_vec[sgi]; for (x = 0; x < sg->xfactor; x++) { /* Include our thread index to shift * (rotate) the order of schedq's @@ -657,7 +658,7 @@ static void update_sg_rem(sched_scalable_thread_state_t *ts, removed = bitset_andn(ts->sg_actual[p], sg_wanted); while (!bitset_is_null(removed)) { sgi = bitset_ffs(removed) - 1; - sg = sg_vec[sgi]; + sg = global->sg_vec[sgi]; for (x = 0; x < sg->xfactor; x++) { remove_schedq_from_list(ts, &sg->schedq[p * @@ -710,8 +711,6 @@ static inline void _schedule_release_ordered(sched_scalable_thread_state_t *ts) ts->rctx = NULL; } -static uint16_t poll_count[ODP_CONFIG_PKTIO_ENTRIES]; - static void pktio_start(int pktio_idx, int num_in_queue, int in_queue_idx[], @@ -725,7 +724,8 @@ static void pktio_start(int pktio_idx, for (i = 0; i < num_in_queue; i++) { rxq = in_queue_idx[i]; ODP_ASSERT(rxq < PKTIO_MAX_QUEUES); - __atomic_fetch_add(&poll_count[pktio_idx], 1, __ATOMIC_RELAXED); + __atomic_fetch_add(&global->poll_count[pktio_idx], 1, + __ATOMIC_RELAXED); qentry = qentry_from_ext(odpq[i]); elem = &qentry->s.sched_elem; elem->cons_type |= FLAG_PKTIN; /* Set pktin queue flag */ @@ -742,7 +742,7 @@ static void pktio_stop(sched_elem_t *elem) { elem->cons_type &= ~FLAG_PKTIN; /* Clear pktin queue flag */ sched_pktin_rem(elem->sched_grp); - if (__atomic_sub_fetch(&poll_count[elem->pktio_idx], + if (__atomic_sub_fetch(&global->poll_count[elem->pktio_idx], 1, __ATOMIC_RELAXED) == 0) { /* Call stop_finalize when all queues * of the pktio have been removed */ @@ -1389,22 +1389,21 @@ static int schedule_group_update(sched_group_t *sg, atom_bitset_clr(&sg->thr_wanted, thr, __ATOMIC_RELAXED); for (p = 0; p < ODP_SCHED_PRIO_NUM; p++) { if (sg->xcount[p] != 0) { + sched_scalable_thread_state_t *state; + + state = &global->thread_state[thr]; + /* This priority level has ODP queues * Notify the thread about membership in * this group/priority */ if (join_leave == SCHED_GROUP_JOIN) - atom_bitset_set( - &thread_state[thr].sg_wanted[p], - sgi, - __ATOMIC_RELEASE); + atom_bitset_set(&state->sg_wanted[p], + sgi, __ATOMIC_RELEASE); else - atom_bitset_clr( - &thread_state[thr].sg_wanted[p], - sgi, - __ATOMIC_RELEASE); - __atomic_store_n(&thread_state[thr].sg_sem, - 1, + atom_bitset_clr(&state->sg_wanted[p], + sgi, __ATOMIC_RELEASE); + __atomic_store_n(&state->sg_sem, 1, __ATOMIC_RELEASE); } } @@ -1447,10 +1446,10 @@ static odp_schedule_group_t schedule_group_create(const char *name, if (mask == NULL) ODP_ABORT("mask is NULL\n"); - odp_spinlock_lock(&sched_grp_lock); + odp_spinlock_lock(&global->sched_grp_lock); /* Allocate a scheduler group */ - free = atom_bitset_load(&sg_free, __ATOMIC_RELAXED); + free = atom_bitset_load(&global->sg_free, __ATOMIC_RELAXED); do { /* All sched_groups in use */ if (bitset_is_null(free)) @@ -1460,7 +1459,7 @@ static odp_schedule_group_t schedule_group_create(const char *name, /* All sched_groups in use */ if (sgi >= MAX_SCHED_GROUP) goto no_free_sched_group; - } while (!atom_bitset_cmpxchg(&sg_free, + } while (!atom_bitset_cmpxchg(&global->sg_free, &free, bitset_clr(free, sgi), true, @@ -1477,12 +1476,13 @@ static odp_schedule_group_t schedule_group_create(const char *name, size = sizeof(sched_group_t) + (ODP_SCHED_PRIO_NUM * xfactor - 1) * sizeof(sched_queue_t); - sg = (sched_group_t *)shm_pool_alloc_align(sched_shm_pool, size); + sg = (sched_group_t *)shm_pool_alloc_align(global->sched_shm_pool, + size); if (sg == NULL) goto shm_pool_alloc_failed; strncpy(sg->name, name ? name : "", ODP_SCHED_GROUP_NAME_LEN - 1); - sg_vec[sgi] = sg; + global->sg_vec[sgi] = sg; memset(sg->thr_actual, 0, sizeof(sg->thr_actual)); sg->thr_wanted = bitset_null(); sg->xfactor = xfactor; @@ -1494,16 +1494,16 @@ static odp_schedule_group_t schedule_group_create(const char *name, if (odp_thrmask_count(mask) != 0) schedule_group_update(sg, sgi, mask, SCHED_GROUP_JOIN); - odp_spinlock_unlock(&sched_grp_lock); + odp_spinlock_unlock(&global->sched_grp_lock); return (odp_schedule_group_t)(sgi); shm_pool_alloc_failed: /* Free the allocated group index */ - atom_bitset_set(&sg_free, sgi, __ATOMIC_RELAXED); + atom_bitset_set(&global->sg_free, sgi, __ATOMIC_RELAXED); no_free_sched_group: - odp_spinlock_unlock(&sched_grp_lock); + odp_spinlock_unlock(&global->sched_grp_lock); return ODP_SCHED_GROUP_INVALID; } @@ -1529,15 +1529,15 @@ static int schedule_group_destroy(odp_schedule_group_t group) sched_ts->sg_sem = 0; update_sg_membership(sched_ts); } - odp_spinlock_lock(&sched_grp_lock); + odp_spinlock_lock(&global->sched_grp_lock); sgi = (uint32_t)group; - if (bitset_is_set(sg_free, sgi)) { + if (bitset_is_set(global->sg_free, sgi)) { ret = -1; goto group_not_found; } - sg = sg_vec[sgi]; + sg = global->sg_vec[sgi]; /* First ensure all threads have processed group_join/group_leave * requests. */ @@ -1570,18 +1570,18 @@ static int schedule_group_destroy(odp_schedule_group_t group) } } - _odp_ishm_pool_free(sched_shm_pool, sg); - sg_vec[sgi] = NULL; - atom_bitset_set(&sg_free, sgi, __ATOMIC_RELEASE); + _odp_ishm_pool_free(global->sched_shm_pool, sg); + global->sg_vec[sgi] = NULL; + atom_bitset_set(&global->sg_free, sgi, __ATOMIC_RELEASE); - odp_spinlock_unlock(&sched_grp_lock); + odp_spinlock_unlock(&global->sched_grp_lock); return ret; thrd_q_present_in_group: group_not_found: - odp_spinlock_unlock(&sched_grp_lock); + odp_spinlock_unlock(&global->sched_grp_lock); invalid_group: @@ -1599,19 +1599,19 @@ static odp_schedule_group_t schedule_group_lookup(const char *name) group = ODP_SCHED_GROUP_INVALID; - odp_spinlock_lock(&sched_grp_lock); + odp_spinlock_lock(&global->sched_grp_lock); /* Scan through the schedule group array */ for (sgi = 0; sgi < MAX_SCHED_GROUP; sgi++) { - if ((sg_vec[sgi] != NULL) && - (strncmp(name, sg_vec[sgi]->name, + if ((global->sg_vec[sgi] != NULL) && + (strncmp(name, global->sg_vec[sgi]->name, ODP_SCHED_GROUP_NAME_LEN) == 0)) { group = (odp_schedule_group_t)sgi; break; } } - odp_spinlock_unlock(&sched_grp_lock); + odp_spinlock_unlock(&global->sched_grp_lock); return group; } @@ -1630,18 +1630,18 @@ static int schedule_group_join(odp_schedule_group_t group, if (mask == NULL) ODP_ABORT("name or mask is NULL\n"); - odp_spinlock_lock(&sched_grp_lock); + odp_spinlock_lock(&global->sched_grp_lock); sgi = (uint32_t)group; - if (bitset_is_set(sg_free, sgi)) { - odp_spinlock_unlock(&sched_grp_lock); + if (bitset_is_set(global->sg_free, sgi)) { + odp_spinlock_unlock(&global->sched_grp_lock); return -1; } - sg = sg_vec[sgi]; + sg = global->sg_vec[sgi]; ret = schedule_group_update(sg, sgi, mask, SCHED_GROUP_JOIN); - odp_spinlock_unlock(&sched_grp_lock); + odp_spinlock_unlock(&global->sched_grp_lock); return ret; } @@ -1662,24 +1662,24 @@ static int schedule_group_leave(odp_schedule_group_t group, if (mask == NULL) ODP_ABORT("name or mask is NULL\n"); - odp_spinlock_lock(&sched_grp_lock); + odp_spinlock_lock(&global->sched_grp_lock); sgi = (uint32_t)group; - if (bitset_is_set(sg_free, sgi)) { + if (bitset_is_set(global->sg_free, sgi)) { ret = -1; goto group_not_found; } - sg = sg_vec[sgi]; + sg = global->sg_vec[sgi]; ret = schedule_group_update(sg, sgi, mask, SCHED_GROUP_LEAVE); - odp_spinlock_unlock(&sched_grp_lock); + odp_spinlock_unlock(&global->sched_grp_lock); return ret; group_not_found: - odp_spinlock_unlock(&sched_grp_lock); + odp_spinlock_unlock(&global->sched_grp_lock); invalid_group: return ret; @@ -1701,23 +1701,23 @@ static int schedule_group_thrmask(odp_schedule_group_t group, if (mask == NULL) ODP_ABORT("name or mask is NULL\n"); - odp_spinlock_lock(&sched_grp_lock); + odp_spinlock_lock(&global->sched_grp_lock); sgi = (uint32_t)group; - if (bitset_is_set(sg_free, sgi)) { + if (bitset_is_set(global->sg_free, sgi)) { ret = -1; goto group_not_found; } - sg = sg_vec[sgi]; + sg = global->sg_vec[sgi]; ret = _schedule_group_thrmask(sg, mask); - odp_spinlock_unlock(&sched_grp_lock); + odp_spinlock_unlock(&global->sched_grp_lock); return ret; group_not_found: - odp_spinlock_unlock(&sched_grp_lock); + odp_spinlock_unlock(&global->sched_grp_lock); invalid_group: return ret; @@ -1739,26 +1739,26 @@ static int schedule_group_info(odp_schedule_group_t group, if (info == NULL) ODP_ABORT("name or mask is NULL\n"); - odp_spinlock_lock(&sched_grp_lock); + odp_spinlock_lock(&global->sched_grp_lock); sgi = (uint32_t)group; - if (bitset_is_set(sg_free, sgi)) { + if (bitset_is_set(global->sg_free, sgi)) { ret = -1; goto group_not_found; } - sg = sg_vec[sgi]; + sg = global->sg_vec[sgi]; ret = _schedule_group_thrmask(sg, &info->thrmask); info->name = sg->name; - odp_spinlock_unlock(&sched_grp_lock); + odp_spinlock_unlock(&global->sched_grp_lock); return ret; group_not_found: - odp_spinlock_unlock(&sched_grp_lock); + odp_spinlock_unlock(&global->sched_grp_lock); invalid_group: return ret; @@ -1770,14 +1770,30 @@ static int schedule_init_global(void) odp_schedule_group_t tmp_all; odp_schedule_group_t tmp_wrkr; odp_schedule_group_t tmp_ctrl; + odp_shm_t shm; uint32_t bits; uint32_t pool_size; uint64_t min_alloc; uint64_t max_alloc; + shm = odp_shm_reserve("_odp_sched_scalable", + sizeof(sched_global_t), + ODP_CACHE_LINE_SIZE, _ODP_SHM_NO_HP); + + global = odp_shm_addr(shm); + if (global == NULL) { + ODP_ERR("Schedule init: Shm reserve failed.\n"); + return -1; + } + + memset(global, 0, sizeof(sched_global_t)); + global->shm = shm; + /* Attach to the pool if it exists */ - sched_shm_pool = _odp_ishm_pool_lookup("sched_shm_pool"); - if (sched_shm_pool == NULL) { + global->sched_shm_pool = _odp_ishm_pool_lookup("sched_shm_pool"); + if (global->sched_shm_pool == NULL) { + _odp_ishm_pool_t *pool; + /* Add storage required for sched groups. Assume worst case * xfactor of MAXTHREADS. */ @@ -1791,31 +1807,31 @@ static int schedule_init_global(void) (ODP_SCHED_PRIO_NUM * MAXTHREADS - 1) * sizeof(sched_queue_t); max_alloc = min_alloc; - sched_shm_pool = _odp_ishm_pool_create("sched_shm_pool", - pool_size, - min_alloc, max_alloc, - _ODP_ISHM_SINGLE_VA); - if (sched_shm_pool == NULL) { + pool = _odp_ishm_pool_create("sched_shm_pool", pool_size, + min_alloc, max_alloc, + _ODP_ISHM_SINGLE_VA); + if (pool == NULL) { ODP_ERR("Failed to allocate shared memory pool " "for sched\n"); goto failed_sched_shm_pool_create; } + global->sched_shm_pool = pool; } - odp_spinlock_init(&sched_grp_lock); + odp_spinlock_init(&global->sched_grp_lock); bits = MAX_SCHED_GROUP; - if (MAX_SCHED_GROUP == sizeof(sg_free) * CHAR_BIT) - sg_free = ~0; + if (MAX_SCHED_GROUP == sizeof(global->sg_free) * CHAR_BIT) + global->sg_free = ~0; else - sg_free = (1 << bits) - 1; + global->sg_free = (1 << bits) - 1; for (uint32_t i = 0; i < MAX_SCHED_GROUP; i++) - sg_vec[i] = NULL; + global->sg_vec[i] = NULL; for (uint32_t i = 0; i < MAXTHREADS; i++) { - thread_state[i].sg_sem = 0; + global->thread_state[i].sg_sem = 0; for (uint32_t j = 0; j < ODP_SCHED_PRIO_NUM; j++) - thread_state[i].sg_wanted[j] = bitset_null(); + global->thread_state[i].sg_wanted[j] = bitset_null(); } /* Create sched groups for default GROUP_ALL, GROUP_WORKER and @@ -1871,7 +1887,12 @@ static int schedule_term_global(void) if (odp_schedule_group_destroy(ODP_SCHED_GROUP_CONTROL) != 0) ODP_ERR("Failed to destroy ODP_SCHED_GROUP_CONTROL\n"); - _odp_ishm_pool_destroy(sched_shm_pool); + _odp_ishm_pool_destroy(global->sched_shm_pool); + + if (odp_shm_free(global->shm)) { + ODP_ERR("Shm free failed for scalable scheduler"); + return -1; + } return 0; } ----------------------------------------------------------------------- Summary of changes: .../include/odp_packet_io_ring_internal.h | 9 +- platform/linux-generic/odp_schedule_scalable.c | 215 +++++++++++---------- platform/linux-generic/pktio/ipc.c | 12 +- platform/linux-generic/pktio/ring.c | 70 +++++-- 4 files changed, 184 insertions(+), 122 deletions(-) hooks/post-receive --

7 years, 3 months

[COMMIT] branch api-next updated. v1.19.0.2_tigermoth-192-g79f23e85b

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 79f23e85b2c7e8f977efc2d10bb06639f08edbaf (commit) via 602e207726055100b068eff1442c28fba95b8fac (commit) via fad375346ab5cab78b44e7de8417527ca77e7ca6 (commit) via 59debc98d1d5ced95da8c95cdbc714184128f1e6 (commit) from e4f37812327c8bef175ea8425e37bf634dab0528 (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 79f23e85b2c7e8f977efc2d10bb06639f08edbaf Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Mon Oct 15 10:00:06 2018 +0300 linux-gen: sched: remove unused schedule interface functions These functions are not needed after iquery scheduler was removed. Signed-off-by: Petri Savolainen <petri.savolainen(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/platform/linux-generic/include/odp_schedule_if.h b/platform/linux-generic/include/odp_schedule_if.h index 30481773d..13e209f2c 100644 --- a/platform/linux-generic/include/odp_schedule_if.h +++ b/platform/linux-generic/include/odp_schedule_if.h @@ -47,7 +47,6 @@ typedef uint32_t (*schedule_max_ordered_locks_fn_t)(void); typedef void (*schedule_save_context_fn_t)(uint32_t queue_index); typedef struct schedule_fn_t { - int status_sync; schedule_pktio_start_fn_t pktio_start; schedule_thr_add_fn_t thr_add; schedule_thr_rem_fn_t thr_rem; @@ -67,10 +66,6 @@ typedef struct schedule_fn_t { schedule_order_unlock_lock_fn_t order_unlock_lock; schedule_max_ordered_locks_fn_t max_ordered_locks; - /* Called only when status_sync is set */ - schedule_unsched_queue_fn_t unsched_queue; - schedule_save_context_fn_t save_context; - } schedule_fn_t; /* Interface towards the scheduler */ diff --git a/platform/linux-generic/odp_queue_basic.c b/platform/linux-generic/odp_queue_basic.c index 96409b753..ccd97da1e 100644 --- a/platform/linux-generic/odp_queue_basic.c +++ b/platform/linux-generic/odp_queue_basic.c @@ -698,7 +698,6 @@ int sched_queue_deq(uint32_t queue_index, odp_event_t ev[], int max_num, int num_deq, status; ring_st_t *ring_st; queue_entry_t *queue = qentry_from_index(queue_index); - int status_sync = sched_fn->status_sync; uint32_t buf_idx[max_num]; ring_st = &queue->s.ring_st; @@ -724,21 +723,14 @@ int sched_queue_deq(uint32_t queue_index, odp_event_t ev[], int max_num, if (num_deq == 0) { /* Already empty queue */ - if (update_status && status == QUEUE_STATUS_SCHED) { + if (update_status && status == QUEUE_STATUS_SCHED) queue->s.status = QUEUE_STATUS_NOTSCHED; - if (odp_unlikely(status_sync)) - sched_fn->unsched_queue(queue->s.index); - } - UNLOCK(queue); return 0; } - if (odp_unlikely(status_sync)) - sched_fn->save_context(queue->s.index); - UNLOCK(queue); buffer_index_to_buf((odp_buffer_hdr_t **)ev, buf_idx, num_deq); diff --git a/platform/linux-generic/odp_schedule_basic.c b/platform/linux-generic/odp_schedule_basic.c index f9f456707..a2eea6b1a 100644 --- a/platform/linux-generic/odp_schedule_basic.c +++ b/platform/linux-generic/odp_schedule_basic.c @@ -1469,7 +1469,6 @@ static int schedule_num_grps(void) /* Fill in scheduler interface */ const schedule_fn_t schedule_basic_fn = { - .status_sync = 0, .pktio_start = schedule_pktio_start, .thr_add = schedule_thr_add, .thr_rem = schedule_thr_rem, @@ -1484,9 +1483,7 @@ const schedule_fn_t schedule_basic_fn = { .term_local = schedule_term_local, .order_lock = order_lock, .order_unlock = order_unlock, - .max_ordered_locks = schedule_max_ordered_locks, - .unsched_queue = NULL, - .save_context = NULL + .max_ordered_locks = schedule_max_ordered_locks }; /* Fill in scheduler API calls */ diff --git a/platform/linux-generic/odp_schedule_sp.c b/platform/linux-generic/odp_schedule_sp.c index 6b9431b69..e63bd0618 100644 --- a/platform/linux-generic/odp_schedule_sp.c +++ b/platform/linux-generic/odp_schedule_sp.c @@ -853,7 +853,6 @@ static void order_unlock(void) /* Fill in scheduler interface */ const schedule_fn_t schedule_sp_fn = { - .status_sync = 0, .pktio_start = pktio_start, .thr_add = thr_add, .thr_rem = thr_rem, @@ -868,9 +867,7 @@ const schedule_fn_t schedule_sp_fn = { .term_local = term_local, .order_lock = order_lock, .order_unlock = order_unlock, - .max_ordered_locks = max_ordered_locks, - .unsched_queue = NULL, - .save_context = NULL + .max_ordered_locks = max_ordered_locks }; /* Fill in scheduler API calls */ commit 602e207726055100b068eff1442c28fba95b8fac Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Mon Oct 15 09:49:27 2018 +0300 linux-gen: remove iquery scheduler implementation Iquery scheduler has no advantage compared to basic and scalable scheduler. It has not been developed for a long time. Remove it to reduce maintenance overhead. Signed-off-by: Petri Savolainen <petri.savolainen(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/platform/linux-generic/Makefile.am b/platform/linux-generic/Makefile.am index ab0b755d5..b973bc301 100644 --- a/platform/linux-generic/Makefile.am +++ b/platform/linux-generic/Makefile.am @@ -189,7 +189,6 @@ __LIB__libodp_linux_la_SOURCES = \ odp_rwlock_recursive.c \ odp_schedule_basic.c \ odp_schedule_if.c \ - odp_schedule_iquery.c \ odp_schedule_scalable.c \ odp_schedule_scalable_ordered.c \ odp_schedule_sp.c \ diff --git a/platform/linux-generic/odp_schedule_iquery.c b/platform/linux-generic/odp_schedule_iquery.c deleted file mode 100644 index e1ef10c46..000000000 --- a/platform/linux-generic/odp_schedule_iquery.c +++ /dev/null @@ -1,1587 +0,0 @@ -/* Copyright (c) 2016-2018, Linaro Limited - * All rights reserved. - * - * SPDX-License-Identifier: BSD-3-Clause - */ - -#include "config.h" - -#include <odp/api/schedule.h> -#include <odp_schedule_if.h> -#include <odp/api/align.h> -#include <odp/api/queue.h> -#include <odp/api/shared_memory.h> -#include <odp_debug_internal.h> -#include <odp_ring_internal.h> -#include <odp_buffer_internal.h> -#include <odp_bitmap_internal.h> -#include <odp/api/thread.h> -#include <odp/api/plat/thread_inlines.h> -#include <odp/api/time.h> -#include <odp/api/rwlock.h> -#include <odp/api/hints.h> -#include <odp/api/cpu.h> -#include <odp/api/thrmask.h> -#include <odp/api/packet_io.h> -#include <odp_config_internal.h> -#include <odp_timer_internal.h> -#include <odp_queue_basic_internal.h> -#include <odp/api/plat/queue_inlines.h> - -/* Number of priority levels */ -#define NUM_SCHED_PRIO 8 - -ODP_STATIC_ASSERT(ODP_SCHED_PRIO_LOWEST == (NUM_SCHED_PRIO - 1), - "lowest_prio_does_not_match_with_num_prios"); - -ODP_STATIC_ASSERT((ODP_SCHED_PRIO_NORMAL > 0) && - (ODP_SCHED_PRIO_NORMAL < (NUM_SCHED_PRIO - 1)), - "normal_prio_is_not_between_highest_and_lowest"); - -/* Number of scheduling groups */ -#define NUM_SCHED_GRPS 256 - -/* Start of named groups in group mask arrays */ -#define SCHED_GROUP_NAMED (ODP_SCHED_GROUP_CONTROL + 1) - -/* Instantiate a WAPL bitmap to be used as queue index bitmap */ -typedef WAPL_BITMAP(ODP_CONFIG_QUEUES) queue_index_bitmap_t; - -typedef struct { - odp_rwlock_t lock; - queue_index_bitmap_t queues; /* queues in this priority level */ -} sched_prio_t; - -typedef struct { - odp_rwlock_t lock; - bool allocated; - odp_thrmask_t threads; /* threads subscribe to this group */ - queue_index_bitmap_t queues; /* queues in this group */ - char name[ODP_SCHED_GROUP_NAME_LEN]; -} sched_group_t; - -/* Packet input poll command queues */ -#define PKTIO_CMD_QUEUES 4 - -/* Maximum number of packet input queues per command */ -#define MAX_PKTIN 16 - -/* Maximum number of packet IO interfaces */ -#define NUM_PKTIO ODP_CONFIG_PKTIO_ENTRIES - -/* Maximum number of pktio poll commands */ -#define NUM_PKTIO_CMD (MAX_PKTIN * NUM_PKTIO) - -/* Not a valid index */ -#define NULL_INDEX ((uint32_t)-1) -/* Pktio command is free */ -#define PKTIO_CMD_FREE ((uint32_t)-1) - -/* Packet IO poll queue ring size. In worst case, all pktios - * have all pktins enabled and one poll command is created per - * pktin queue. The ring size must be larger than or equal to - * NUM_PKTIO_CMD / PKTIO_CMD_QUEUES, so that it can hold all - * poll commands in the worst case. - */ -#define PKTIO_RING_SIZE (NUM_PKTIO_CMD / PKTIO_CMD_QUEUES) - -/* Mask for wrapping around pktio poll command index */ -#define PKTIO_RING_MASK (PKTIO_RING_SIZE - 1) - -/* Maximum number of dequeues */ -#define MAX_DEQ CONFIG_BURST_SIZE - -/* Instantiate a RING data structure as pktio command queue */ -typedef struct ODP_ALIGNED_CACHE { - /* Ring header */ - ring_t ring; - - /* Ring data: pktio poll command indexes */ - uint32_t cmd_index[PKTIO_RING_SIZE]; -} pktio_cmd_queue_t; - -/* Packet IO poll command */ -typedef struct { - int pktio; - int count; - int pktin[MAX_PKTIN]; - uint32_t index; -} pktio_cmd_t; - -/* Collect the pktio poll resources */ -typedef struct { - odp_rwlock_t lock; - /* count active commands per pktio interface */ - int actives[NUM_PKTIO]; - pktio_cmd_t commands[NUM_PKTIO_CMD]; - pktio_cmd_queue_t queues[PKTIO_CMD_QUEUES]; -} pktio_poll_t; - -/* Forward declaration */ -typedef struct sched_thread_local sched_thread_local_t; - -/* Order context of a queue */ -typedef struct ODP_ALIGNED_CACHE { - /* Current ordered context id */ - odp_atomic_u64_t ODP_ALIGNED_CACHE ctx; - - /* Next unallocated context id */ - odp_atomic_u64_t next_ctx; - - /* Array of ordered locks */ - odp_atomic_u64_t lock[CONFIG_QUEUE_MAX_ORD_LOCKS]; - -} order_context_t; - -typedef struct { - odp_shm_t selfie; - - /* Schedule priorities */ - sched_prio_t prios[NUM_SCHED_PRIO]; - - /* Schedule groups */ - sched_group_t groups[NUM_SCHED_GRPS]; - - /* Cache queue parameters for easy reference */ - odp_schedule_param_t queues[ODP_CONFIG_QUEUES]; - - /* Poll pktio inputs in spare time */ - pktio_poll_t pktio_poll; - - /* Queues send or unwind their availability indications - * for scheduling, the bool value also serves as a focal - * point for atomic competition. */ - bool availables[ODP_CONFIG_QUEUES]; - - /* Quick reference to per thread context */ - sched_thread_local_t *threads[ODP_THREAD_COUNT_MAX]; - - order_context_t order[ODP_CONFIG_QUEUES]; -} sched_global_t; - -/* Per thread events cache */ -typedef struct { - int count; - odp_queue_t queue; - odp_event_t stash[MAX_DEQ], *top; -} event_cache_t; - -/* Ordered stash size */ -#define MAX_ORDERED_STASH 512 - -/* Storage for stashed enqueue operation arguments */ -typedef struct { - odp_buffer_hdr_t *buf_hdr[QUEUE_MULTI_MAX]; - odp_queue_t queue; - int num; -} ordered_stash_t; - -/* Ordered lock states */ -typedef union { - uint8_t u8[CONFIG_QUEUE_MAX_ORD_LOCKS]; - uint32_t all; -} lock_called_t; - -ODP_STATIC_ASSERT(sizeof(lock_called_t) == sizeof(uint32_t), - "Lock_called_values_do_not_fit_in_uint32"); - -/* Instantiate a sparse bitmap to store thread's interested - * queue indexes per priority. - */ -typedef SPARSE_BITMAP(ODP_CONFIG_QUEUES) queue_index_sparse_t; - -struct sched_thread_local { - int thread; - bool pause; - - /* Cache events only for atomic queue */ - event_cache_t cache; - - /* Saved atomic context */ - bool *atomic; - - /* Record the pktio polls have done */ - uint16_t pktin_polls; - - /* Interested queue indexes to be checked by thread - * at each priority level for scheduling, and a round - * robin iterator to improve fairness between queues - * in the same priority level. - */ - odp_rwlock_t lock; - int r_locked; - queue_index_sparse_t indexes[NUM_SCHED_PRIO]; - sparse_bitmap_iterator_t iterators[NUM_SCHED_PRIO]; - - struct { - /* Source queue index */ - uint32_t src_queue; - uint64_t ctx; /**< Ordered context id */ - int stash_num; /**< Number of stashed enqueue operations */ - uint8_t in_order; /**< Order status */ - lock_called_t lock_called; /**< States of ordered locks */ - /** Storage for stashed enqueue operations */ - ordered_stash_t stash[MAX_ORDERED_STASH]; - } ordered; -}; - -/* Global scheduler context */ -static sched_global_t *sched; - -/* Thread local scheduler context */ -static __thread sched_thread_local_t thread_local; - -static int schedule_init_global(void) -{ - odp_shm_t shm; - int i, k, prio, group; - - ODP_DBG("Schedule[iquery] init ... "); - - shm = odp_shm_reserve("odp_scheduler_iquery", - sizeof(sched_global_t), - ODP_CACHE_LINE_SIZE, 0); - - sched = odp_shm_addr(shm); - - if (sched == NULL) { - ODP_ERR("Schedule[iquery] " - "init: shm reserve.\n"); - return -1; - } - - memset(sched, 0, sizeof(sched_global_t)); - - sched->selfie = shm; - - for (prio = 0; prio < NUM_SCHED_PRIO; prio++) - odp_rwlock_init(&sched->prios[prio].lock); - - for (group = 0; group < NUM_SCHED_GRPS; group++) { - sched->groups[group].allocated = false; - odp_rwlock_init(&sched->groups[group].lock); - } - - odp_rwlock_init(&sched->pktio_poll.lock); - - for (i = 0; i < PKTIO_CMD_QUEUES; i++) { - pktio_cmd_queue_t *queue = - &sched->pktio_poll.queues[i]; - - ring_init(&queue->ring); - - for (k = 0; k < PKTIO_RING_SIZE; k++) - queue->cmd_index[k] = -1; - } - - for (i = 0; i < NUM_PKTIO_CMD; i++) - sched->pktio_poll.commands[i].index = PKTIO_CMD_FREE; - - ODP_DBG("done\n"); - return 0; -} - -static int schedule_term_global(void) -{ - uint32_t i; - odp_shm_t shm = sched->selfie; - - for (i = 0; i < ODP_CONFIG_QUEUES; i++) { - int count = 0; - odp_event_t events[1]; - - if (sched->availables[i]) - count = sched_queue_deq(i, events, 1, 1); - - if (count > 0) - ODP_ERR("Queue (%d) not empty\n", i); - } - - memset(sched, 0, sizeof(sched_global_t)); - - if (odp_shm_free(shm) < 0) { - ODP_ERR("Schedule[iquery] " - "term: shm release.\n"); - return -1; - } - return 0; -} - -/* - * These APIs are used to manipulate thread's interests. - */ -static void thread_set_interest(sched_thread_local_t *thread, - unsigned int queue_index, int prio); - -static void thread_clear_interest(sched_thread_local_t *thread, - unsigned int queue_index, int prio); - -static void thread_set_interests(sched_thread_local_t *thread, - queue_index_bitmap_t *set); - -static void thread_clear_interests(sched_thread_local_t *thread, - queue_index_bitmap_t *clear); - -static void sched_thread_local_reset(void) -{ - int prio; - queue_index_sparse_t *index; - sparse_bitmap_iterator_t *iterator; - - memset(&thread_local, 0, sizeof(sched_thread_local_t)); - - thread_local.thread = odp_thread_id(); - thread_local.cache.queue = ODP_QUEUE_INVALID; - thread_local.ordered.src_queue = NULL_INDEX; - - odp_rwlock_init(&thread_local.lock); - - for (prio = 0; prio < NUM_SCHED_PRIO; prio++) { - index = &thread_local.indexes[prio]; - iterator = &thread_local.iterators[prio]; - - sparse_bitmap_zero(index); - sparse_bitmap_iterator(iterator, index); - } -} - -static int schedule_init_local(void) -{ - int group; - sched_group_t *G; - queue_index_bitmap_t collect; - - wapl_bitmap_zero(&collect); - sched_thread_local_reset(); - - /* Collect all queue indexes of the schedule groups - * which this thread has subscribed - */ - for (group = 0; group < NUM_SCHED_GRPS; group++) { - G = &sched->groups[group]; - odp_rwlock_read_lock(&G->lock); - - if ((group < SCHED_GROUP_NAMED || G->allocated) && - odp_thrmask_isset(&G->threads, thread_local.thread)) - wapl_bitmap_or(&collect, &collect, &G->queues); - - odp_rwlock_read_unlock(&G->lock); - } - - /* Distribute the above collected queue indexes into - * thread local interests per priority level. - */ - thread_set_interests(&thread_local, &collect); - - /* "Night gathers, and now my watch begins..." */ - sched->threads[thread_local.thread] = &thread_local; - return 0; -} - -static inline void schedule_release_context(void); - -static int schedule_term_local(void) -{ - int group; - sched_group_t *G; - - if (thread_local.cache.count) { - ODP_ERR("Locally pre-scheduled events exist.\n"); - return -1; - } - - schedule_release_context(); - - /* Unsubscribe all named schedule groups */ - for (group = SCHED_GROUP_NAMED; - group < NUM_SCHED_GRPS; group++) { - G = &sched->groups[group]; - odp_rwlock_write_lock(&G->lock); - - if (G->allocated && odp_thrmask_isset( - &G->threads, thread_local.thread)) - odp_thrmask_clr(&G->threads, thread_local.thread); - - odp_rwlock_write_unlock(&G->lock); - } - - /* "...for this night and all the nights to come." */ - sched->threads[thread_local.thread] = NULL; - sched_thread_local_reset(); - return 0; -} - -static int init_sched_queue(uint32_t queue_index, - const odp_schedule_param_t *sched_param) -{ - int prio, group, thread, i; - sched_prio_t *P; - sched_group_t *G; - sched_thread_local_t *local; - - prio = sched_param->prio; - group = sched_param->group; - - G = &sched->groups[group]; - odp_rwlock_write_lock(&G->lock); - - /* Named schedule group must be created prior - * to queue creation to this group. - */ - if (group >= SCHED_GROUP_NAMED && !G->allocated) { - odp_rwlock_write_unlock(&G->lock); - return -1; - } - - /* Record the queue in its priority level globally */ - P = &sched->prios[prio]; - - odp_rwlock_write_lock(&P->lock); - wapl_bitmap_set(&P->queues, queue_index); - odp_rwlock_write_unlock(&P->lock); - - /* Record the queue in its schedule group */ - wapl_bitmap_set(&G->queues, queue_index); - - /* Cache queue parameters for easy reference */ - memcpy(&sched->queues[queue_index], - sched_param, sizeof(odp_schedule_param_t)); - - odp_atomic_init_u64(&sched->order[queue_index].ctx, 0); - odp_atomic_init_u64(&sched->order[queue_index].next_ctx, 0); - - for (i = 0; i < CONFIG_QUEUE_MAX_ORD_LOCKS; i++) - odp_atomic_init_u64(&sched->order[queue_index].lock[i], 0); - - /* Update all threads in this schedule group to - * start check this queue index upon scheduling. - */ - thread = odp_thrmask_first(&G->threads); - while (thread >= 0) { - local = sched->threads[thread]; - thread_set_interest(local, queue_index, prio); - thread = odp_thrmask_next(&G->threads, thread); - } - - odp_rwlock_write_unlock(&G->lock); - return 0; -} - -/* - * Must be called with schedule group's rwlock held. - * This is also being used in destroy_schedule_group() - * to destroy all orphan queues while destroying a whole - * schedule group. - */ -static void __destroy_sched_queue( - sched_group_t *G, uint32_t queue_index) -{ - int prio, thread; - sched_prio_t *P; - sched_thread_local_t *local; - - prio = sched->queues[queue_index].prio; - - /* Forget the queue in its schedule group */ - wapl_bitmap_clear(&G->queues, queue_index); - - /* Forget queue schedule parameters */ - memset(&sched->queues[queue_index], - 0, sizeof(odp_schedule_param_t)); - - /* Update all threads in this schedule group to - * stop check this queue index upon scheduling. - */ - thread = odp_thrmask_first(&G->threads); - while (thread >= 0) { - local = sched->threads[thread]; - thread_clear_interest(local, queue_index, prio); - thread = odp_thrmask_next(&G->threads, thread); - } - - /* Forget the queue in its priority level globally */ - P = &sched->prios[prio]; - - odp_rwlock_write_lock(&P->lock); - wapl_bitmap_clear(&P->queues, queue_index); - odp_rwlock_write_unlock(&P->lock); -} - -static void destroy_sched_queue(uint32_t queue_index) -{ - int group; - sched_group_t *G; - - group = sched->queues[queue_index].group; - - G = &sched->groups[group]; - odp_rwlock_write_lock(&G->lock); - - /* Named schedule group could have been destroyed - * earlier and left these orphan queues. - */ - if (group >= SCHED_GROUP_NAMED && !G->allocated) { - odp_rwlock_write_unlock(&G->lock); - return; - } - - if (thread_local.r_locked) - odp_rwlock_read_unlock(&thread_local.lock); - - __destroy_sched_queue(G, queue_index); - - if (thread_local.r_locked) - odp_rwlock_read_lock(&thread_local.lock); - - odp_rwlock_write_unlock(&G->lock); - - if (sched->queues[queue_index].sync == ODP_SCHED_SYNC_ORDERED && - odp_atomic_load_u64(&sched->order[queue_index].ctx) != - odp_atomic_load_u64(&sched->order[queue_index].next_ctx)) - ODP_ERR("queue reorder incomplete\n"); -} - -static int pktio_cmd_queue_hash(int pktio, int pktin) -{ - return (pktio ^ pktin) % PKTIO_CMD_QUEUES; -} - -static inline pktio_cmd_t *alloc_pktio_cmd(void) -{ - int i; - pktio_cmd_t *cmd = NULL; - - odp_rwlock_write_lock(&sched->pktio_poll.lock); - - /* Find next free command */ - for (i = 0; i < NUM_PKTIO_CMD; i++) { - if (sched->pktio_poll.commands[i].index - == PKTIO_CMD_FREE) { - cmd = &sched->pktio_poll.commands[i]; - cmd->index = i; - break; - } - } - - odp_rwlock_write_unlock(&sched->pktio_poll.lock); - return cmd; -} - -static inline void free_pktio_cmd(pktio_cmd_t *cmd) -{ - odp_rwlock_write_lock(&sched->pktio_poll.lock); - - cmd->index = PKTIO_CMD_FREE; - - odp_rwlock_write_unlock(&sched->pktio_poll.lock); -} - -static void schedule_pktio_start(int pktio, - int count, - int pktin[], - odp_queue_t odpq[] ODP_UNUSED) -{ - int i, index; - pktio_cmd_t *cmd; - - if (count > MAX_PKTIN) - ODP_ABORT("Too many input queues for scheduler\n"); - - /* Record the active commands count per pktio interface */ - sched->pktio_poll.actives[pktio] = count; - - /* Create a pktio poll command per pktin */ - for (i = 0; i < count; i++) { - cmd = alloc_pktio_cmd(); - - if (cmd == NULL) - ODP_ABORT("Scheduler out of pktio commands\n"); - - index = pktio_cmd_queue_hash(pktio, pktin[i]); - - cmd->pktio = pktio; - cmd->count = 1; - cmd->pktin[0] = pktin[i]; - ring_enq(&sched->pktio_poll.queues[index].ring, - PKTIO_RING_MASK, cmd->index); - } -} - -static int schedule_pktio_stop(int pktio, int pktin ODP_UNUSED) -{ - int remains; - - odp_rwlock_write_lock(&sched->pktio_poll.lock); - - sched->pktio_poll.actives[pktio]--; - remains = sched->pktio_poll.actives[pktio]; - - odp_rwlock_write_unlock(&sched->pktio_poll.lock); - return remains; -} - -static inline bool do_schedule_prio(int prio); - -static inline int pop_cache_events(odp_event_t ev[], unsigned int max) -{ - int k = 0; - event_cache_t *cache; - - cache = &thread_local.cache; - while (cache->count && max) { - ev[k] = *cache->top++; - k++; - max--; - cache->count--; - } - - return k; -} - -static inline void assign_queue_handle(odp_queue_t *handle) -{ - if (handle) - *handle = thread_local.cache.queue; -} - -static inline void pktio_poll_input(void) -{ - int i, hash; - uint32_t index; - - ring_t *ring; - pktio_cmd_t *cmd; - - /* - * Each thread starts the search for a poll command - * from the hash(threadID) queue to mitigate contentions. - * If the queue is empty, it moves to other queues. - * - * Most of the times, the search stops on the first - * command found to optimize multi-threaded performance. - * A small portion of polls have to do full iteration to - * avoid packet input starvation when there are less - * threads than command queues. - */ - hash = thread_local.thread % PKTIO_CMD_QUEUES; - - for (i = 0; i < PKTIO_CMD_QUEUES; i++, - hash = (hash + 1) % PKTIO_CMD_QUEUES) { - ring = &sched->pktio_poll.queues[hash].ring; - - if (odp_unlikely(ring_deq(ring, PKTIO_RING_MASK, &index) == 0)) - continue; - - cmd = &sched->pktio_poll.commands[index]; - - /* Poll packet input */ - if (odp_unlikely(sched_cb_pktin_poll_old(cmd->pktio, - cmd->count, - cmd->pktin))) { - /* Pktio stopped or closed. Remove poll - * command and call stop_finalize when all - * commands of the pktio has been removed. - */ - if (schedule_pktio_stop(cmd->pktio, - cmd->pktin[0]) == 0) - sched_cb_pktio_stop_finalize(cmd->pktio); - - free_pktio_cmd(cmd); - } else { - /* Continue scheduling the pktio */ - ring_enq(ring, PKTIO_RING_MASK, index); - - /* Do not iterate through all pktin poll - * command queues every time. - */ - if (odp_likely(thread_local.pktin_polls & 0xF)) - break; - } - } - - thread_local.pktin_polls++; -} - -/* - * Schedule queues - */ -static int do_schedule(odp_queue_t *out_queue, - odp_event_t out_ev[], unsigned int max_num) -{ - int prio, count; - - /* Consume locally cached events */ - count = pop_cache_events(out_ev, max_num); - if (count > 0) { - assign_queue_handle(out_queue); - return count; - } - - schedule_release_context(); - - if (odp_unlikely(thread_local.pause)) - return count; - - odp_rwlock_read_lock(&thread_local.lock); - thread_local.r_locked = 1; - - /* Schedule events */ - for (prio = 0; prio < NUM_SCHED_PRIO; prio++) { - /* Round robin iterate the interested queue - * indexes in this priority level to compete - * and consume available queues - */ - if (!do_schedule_prio(prio)) - continue; - - count = pop_cache_events(out_ev, max_num); - assign_queue_handle(out_queue); - - odp_rwlock_read_unlock(&thread_local.lock); - thread_local.r_locked = 0; - return count; - } - - odp_rwlock_read_unlock(&thread_local.lock); - thread_local.r_locked = 0; - - /* Poll packet input when there are no events */ - pktio_poll_input(); - return 0; -} - -static int schedule_loop(odp_queue_t *out_queue, uint64_t wait, - odp_event_t out_ev[], unsigned int max_num) -{ - int count, first = 1; - odp_time_t next, wtime; - - while (1) { - timer_run(); - - count = do_schedule(out_queue, out_ev, max_num); - - if (count) - break; - - if (wait == ODP_SCHED_WAIT) - continue; - - if (wait == ODP_SCHED_NO_WAIT) - break; - - if (first) { - wtime = odp_time_local_from_ns(wait); - next = odp_time_sum(odp_time_local(), wtime); - first = 0; - continue; - } - - if (odp_time_cmp(next, odp_time_local()) < 0) - break; - } - - return count; -} - -static odp_event_t schedule(odp_queue_t *out_queue, uint64_t wait) -{ - odp_event_t ev; - - ev = ODP_EVENT_INVALID; - - schedule_loop(out_queue, wait, &ev, 1); - - return ev; -} - -static int schedule_multi(odp_queue_t *out_queue, uint64_t wait, - odp_event_t events[], int num) -{ - return schedule_loop(out_queue, wait, events, num); -} - -static void schedule_pause(void) -{ - thread_local.pause = 1; -} - -static void schedule_resume(void) -{ - thread_local.pause = 0; -} - -static uint64_t schedule_wait_time(uint64_t ns) -{ - return ns; -} - -static int number_of_priorites(void) -{ - return NUM_SCHED_PRIO; -} - -/* - * Create a named schedule group with pre-defined - * set of subscription threads. - * - * Sched queues belonging to this group must be - * created after the group creation. Upon creation - * the group holds 0 sched queues. - */ -static odp_schedule_group_t schedule_group_create( - const char *name, const odp_thrmask_t *mask) -{ - int group; - sched_group_t *G; - - for (group = SCHED_GROUP_NAMED; - group < NUM_SCHED_GRPS; group++) { - G = &sched->groups[group]; - - odp_rwlock_write_lock(&G->lock); - if (!G->allocated) { - strncpy(G->name, name ? name : "", - ODP_SCHED_GROUP_NAME_LEN - 1); - odp_thrmask_copy(&G->threads, mask); - wapl_bitmap_zero(&G->queues); - - G->allocated = true; - odp_rwlock_write_unlock(&G->lock); - return (odp_schedule_group_t)group; - } - odp_rwlock_write_unlock(&G->lock); - } - - return ODP_SCHED_GROUP_INVALID; -} - -static inline void __destroy_group_queues(sched_group_t *group) -{ - unsigned int index; - queue_index_bitmap_t queues; - wapl_bitmap_iterator_t it; - - /* Constructor */ - wapl_bitmap_zero(&queues); - wapl_bitmap_copy(&queues, &group->queues); - wapl_bitmap_iterator(&it, &queues); - - /* Walk through the queue index bitmap */ - for (it.start(&it); it.has_next(&it);) { - index = it.next(&it); - __destroy_sched_queue(group, index); - } -} - -/* - * Destroy a named schedule group. - */ -static int schedule_group_destroy(odp_schedule_group_t group) -{ - int done = -1; - sched_group_t *G; - - if (group < SCHED_GROUP_NAMED || - group >= NUM_SCHED_GRPS) - return -1; - - G = &sched->groups[group]; - odp_rwlock_write_lock(&G->lock); - - if (G->allocated) { - /* Destroy all queues in this schedule group - * and leave no orphan queues. - */ - __destroy_group_queues(G); - - done = 0; - G->allocated = false; - wapl_bitmap_zero(&G->queues); - odp_thrmask_zero(&G->threads); - memset(G->name, 0, ODP_SCHED_GROUP_NAME_LEN); - } - - odp_rwlock_write_unlock(&G->lock); - return done; -} - -static odp_schedule_group_t schedule_group_lookup(const char *name) -{ - int group; - sched_group_t *G; - - for (group = SCHED_GROUP_NAMED; - group < NUM_SCHED_GRPS; group++) { - G = &sched->groups[group]; - - odp_rwlock_read_lock(&G->lock); - if (strcmp(name, G->name) == 0) { - odp_rwlock_read_unlock(&G->lock); - return (odp_schedule_group_t)group; - } - odp_rwlock_read_unlock(&G->lock); - } - - return ODP_SCHED_GROUP_INVALID; -} - -static int schedule_group_join(odp_schedule_group_t group, - const odp_thrmask_t *mask) -{ - int done = -1, thread; - sched_group_t *G; - sched_thread_local_t *local; - - /* Named schedule group only */ - if (group < SCHED_GROUP_NAMED || - group >= NUM_SCHED_GRPS) - return done; - - G = &sched->groups[group]; - odp_rwlock_write_lock(&G->lock); - - if (G->allocated) { - /* Make new joined threads to start check - * queue indexes in this schedule group - */ - thread = odp_thrmask_first(mask); - while (thread >= 0) { - local = sched->threads[thread]; - thread_set_interests(local, &G->queues); - - odp_thrmask_set(&G->threads, thread); - thread = odp_thrmask_next(mask, thread); - } - done = 0; - } - - odp_rwlock_write_unlock(&G->lock); - return done; -} - -static int schedule_group_leave(odp_schedule_group_t group, - const odp_thrmask_t *mask) -{ - int done = -1, thread; - sched_group_t *G; - sched_thread_local_t *local; - - /* Named schedule group only */ - if (group < SCHED_GROUP_NAMED || - group >= NUM_SCHED_GRPS) - return done; - - G = &sched->groups[group]; - odp_rwlock_write_lock(&G->lock); - - if (G->allocated) { - /* Make leaving threads to stop check - * queue indexes in this schedule group - */ - thread = odp_thrmask_first(mask); - while (thread >= 0) { - local = sched->threads[thread]; - thread_clear_interests(local, &G->queues); - - odp_thrmask_clr(&G->threads, thread); - thread = odp_thrmask_next(mask, thread); - } - done = 0; - } - - odp_rwlock_write_unlock(&G->lock); - return done; -} - -static int schedule_group_thrmask(odp_schedule_group_t group, - odp_thrmask_t *thrmask) -{ - int done = -1; - sched_group_t *G; - - /* Named schedule group only */ - if (group < SCHED_GROUP_NAMED || - group >= NUM_SCHED_GRPS) - return done; - - G = &sched->groups[group]; - odp_rwlock_read_lock(&G->lock); - - if (G->allocated && thrmask != NULL) { - done = 0; - odp_thrmask_copy(thrmask, &G->threads); - } - - odp_rwlock_read_unlock(&G->lock); - return done; -} - -static int schedule_group_info(odp_schedule_group_t group, - odp_schedule_group_info_t *info) -{ - int done = -1; - sched_group_t *G; - - /* Named schedule group only */ - if (group < SCHED_GROUP_NAMED || - group >= NUM_SCHED_GRPS) - return done; - - G = &sched->groups[group]; - odp_rwlock_read_lock(&G->lock); - - if (G->allocated && info != NULL) { - done = 0; - info->name = G->name; - odp_thrmask_copy(&info->thrmask, &G->threads); - } - - odp_rwlock_read_unlock(&G->lock); - return done; -} - -/* This function is a no-op */ -static void schedule_prefetch(int num ODP_UNUSED) -{ -} - -/* - * Limited to join and leave pre-defined schedule groups - * before and after thread local initialization or termination. - */ -static int group_add_thread(odp_schedule_group_t group, int thread) -{ - sched_group_t *G; - - if (group < 0 || group >= SCHED_GROUP_NAMED) - return -1; - - G = &sched->groups[group]; - - odp_rwlock_write_lock(&G->lock); - odp_thrmask_set(&G->threads, thread); - odp_rwlock_write_unlock(&G->lock); - return 0; -} - -static int group_remove_thread(odp_schedule_group_t group, int thread) -{ - sched_group_t *G; - - if (group < 0 || group >= SCHED_GROUP_NAMED) - return -1; - - G = &sched->groups[group]; - - odp_rwlock_write_lock(&G->lock); - odp_thrmask_clr(&G->threads, thread); - odp_rwlock_write_unlock(&G->lock); - return 0; -} - -static int number_of_groups(void) -{ - return NUM_SCHED_GRPS; -} - -static int schedule_sched_queue(uint32_t queue_index) -{ - /* Set available indications globally */ - sched->availables[queue_index] = true; - return 0; -} - -static int schedule_unsched_queue(uint32_t queue_index) -{ - /* Clear available indications globally */ - sched->availables[queue_index] = false; - return 0; -} - -static void schedule_release_atomic(void) -{ - unsigned int queue_index; - - if ((thread_local.atomic != NULL) && - (thread_local.cache.count == 0)) { - queue_index = thread_local.atomic - sched->availables; - thread_local.atomic = NULL; - sched->availables[queue_index] = true; - } -} - -static inline int ordered_own_turn(uint32_t queue_index) -{ - uint64_t ctx; - - ctx = odp_atomic_load_acq_u64(&sched->order[queue_index].ctx); - - return ctx == thread_local.ordered.ctx; -} - -static inline void wait_for_order(uint32_t queue_index) -{ - /* Busy loop to synchronize ordered processing */ - while (1) { - if (ordered_own_turn(queue_index)) - break; - odp_cpu_pause(); - } -} - -/** - * Perform stashed enqueue operations - * - * Should be called only when already in order. - */ -static inline void ordered_stash_release(void) -{ - int i; - - for (i = 0; i < thread_local.ordered.stash_num; i++) { - odp_queue_t queue; - odp_buffer_hdr_t **buf_hdr; - int num, num_enq; - - queue = thread_local.ordered.stash[i].queue; - buf_hdr = thread_local.ordered.stash[i].buf_hdr; - num = thread_local.ordered.stash[i].num; - - num_enq = odp_queue_enq_multi(queue, - (odp_event_t *)buf_hdr, num); - - if (odp_unlikely(num_enq < num)) { - if (odp_unlikely(num_enq < 0)) - num_enq = 0; - - ODP_DBG("Dropped %i packets\n", num - num_enq); - buffer_free_multi(&buf_hdr[num_enq], num - num_enq); - } - } - thread_local.ordered.stash_num = 0; -} - -static inline void release_ordered(void) -{ - uint32_t qi; - uint32_t i; - - qi = thread_local.ordered.src_queue; - - wait_for_order(qi); - - /* Release all ordered locks */ - for (i = 0; i < sched->queues[qi].lock_count; i++) { - if (!thread_local.ordered.lock_called.u8[i]) - odp_atomic_store_rel_u64(&sched->order[qi].lock[i], - thread_local.ordered.ctx + 1); - } - - thread_local.ordered.lock_called.all = 0; - thread_local.ordered.src_queue = NULL_INDEX; - thread_local.ordered.in_order = 0; - - ordered_stash_release(); - - /* Next thread can continue processing */ - odp_atomic_add_rel_u64(&sched->order[qi].ctx, 1); -} - -static void schedule_release_ordered(void) -{ - uint32_t queue_index; - - queue_index = thread_local.ordered.src_queue; - - if (odp_unlikely((queue_index == NULL_INDEX) || - thread_local.cache.count)) - return; - - release_ordered(); -} - -static inline void schedule_release_context(void) -{ - if (thread_local.ordered.src_queue != NULL_INDEX) - release_ordered(); - else - schedule_release_atomic(); -} - -static int schedule_ord_enq_multi(odp_queue_t dst_queue, void *buf_hdr[], - int num, int *ret) -{ - int i; - uint32_t stash_num = thread_local.ordered.stash_num; - queue_entry_t *dst_qentry = qentry_from_handle(dst_queue); - uint32_t src_queue = thread_local.ordered.src_queue; - - if ((src_queue == NULL_INDEX) || thread_local.ordered.in_order) - return 0; - - if (ordered_own_turn(src_queue)) { - /* Own turn, so can do enqueue directly. */ - thread_local.ordered.in_order = 1; - ordered_stash_release(); - return 0; - } - - /* Pktout may drop packets, so the operation cannot be stashed. */ - if (dst_qentry->s.pktout.pktio != ODP_PKTIO_INVALID || - odp_unlikely(stash_num >= MAX_ORDERED_STASH)) { - /* If the local stash is full, wait until it is our turn and - * then release the stash and do enqueue directly. */ - wait_for_order(src_queue); - - thread_local.ordered.in_order = 1; - - ordered_stash_release(); - return 0; - } - - thread_local.ordered.stash[stash_num].queue = dst_queue; - thread_local.ordered.stash[stash_num].num = num; - for (i = 0; i < num; i++) - thread_local.ordered.stash[stash_num].buf_hdr[i] = buf_hdr[i]; - - thread_local.ordered.stash_num++; - - *ret = num; - return 1; -} - -static void order_lock(void) -{ - uint32_t queue_index; - - queue_index = thread_local.ordered.src_queue; - - if (queue_index == NULL_INDEX) - return; - - wait_for_order(queue_index); -} - -static void order_unlock(void) -{ -} - -static void schedule_order_lock(uint32_t lock_index) -{ - odp_atomic_u64_t *ord_lock; - uint32_t queue_index; - - queue_index = thread_local.ordered.src_queue; - - ODP_ASSERT(queue_index != NULL_INDEX && - lock_index <= sched->queues[queue_index].lock_count && - !thread_local.ordered.lock_called.u8[lock_index]); - - ord_lock = &sched->order[queue_index].lock[lock_index]; - - /* Busy loop to synchronize ordered processing */ - while (1) { - uint64_t lock_seq; - - lock_seq = odp_atomic_load_acq_u64(ord_lock); - - if (lock_seq == thread_local.ordered.ctx) { - thread_local.ordered.lock_called.u8[lock_index] = 1; - return; - } - odp_cpu_pause(); - } -} - -static void schedule_order_unlock(uint32_t lock_index) -{ - odp_atomic_u64_t *ord_lock; - uint32_t queue_index; - - queue_index = thread_local.ordered.src_queue; - - ODP_ASSERT(queue_index != NULL_INDEX && - lock_index <= sched->queues[queue_index].lock_count); - - ord_lock = &sched->order[queue_index].lock[lock_index]; - - ODP_ASSERT(thread_local.ordered.ctx == odp_atomic_load_u64(ord_lock)); - - odp_atomic_store_rel_u64(ord_lock, thread_local.ordered.ctx + 1); -} - -static void schedule_order_unlock_lock(uint32_t unlock_index, - uint32_t lock_index) -{ - schedule_order_unlock(unlock_index); - schedule_order_lock(lock_index); -} - -static uint32_t schedule_max_ordered_locks(void) -{ - return CONFIG_QUEUE_MAX_ORD_LOCKS; -} - -static void schedule_order_lock_start(uint32_t lock_index) -{ - (void)lock_index; -} - -static void schedule_order_lock_wait(uint32_t lock_index) -{ - schedule_order_lock(lock_index); -} - -static inline bool is_atomic_queue(unsigned int queue_index) -{ - return (sched->queues[queue_index].sync == ODP_SCHED_SYNC_ATOMIC); -} - -static inline bool is_ordered_queue(unsigned int queue_index) -{ - return (sched->queues[queue_index].sync == ODP_SCHED_SYNC_ORDERED); -} - -static void schedule_save_context(uint32_t queue_index) -{ - if (is_atomic_queue(queue_index)) { - thread_local.atomic = &sched->availables[queue_index]; - } else if (is_ordered_queue(queue_index)) { - uint64_t ctx; - odp_atomic_u64_t *next_ctx; - - next_ctx = &sched->order[queue_index].next_ctx; - ctx = odp_atomic_fetch_inc_u64(next_ctx); - - thread_local.ordered.ctx = ctx; - thread_local.ordered.src_queue = queue_index; - } -} - -/* Fill in scheduler interface */ -const schedule_fn_t schedule_iquery_fn = { - .status_sync = 1, - .pktio_start = schedule_pktio_start, - .thr_add = group_add_thread, - .thr_rem = group_remove_thread, - .num_grps = number_of_groups, - .init_queue = init_sched_queue, - .destroy_queue = destroy_sched_queue, - .sched_queue = schedule_sched_queue, - .ord_enq_multi = schedule_ord_enq_multi, - .init_global = schedule_init_global, - .term_global = schedule_term_global, - .init_local = schedule_init_local, - .term_local = schedule_term_local, - .order_lock = order_lock, - .order_unlock = order_unlock, - .max_ordered_locks = schedule_max_ordered_locks, - .unsched_queue = schedule_unsched_queue, - .save_context = schedule_save_context -}; - -/* Fill in scheduler API calls */ -const schedule_api_t schedule_iquery_api = { - .schedule_wait_time = schedule_wait_time, - .schedule = schedule, - .schedule_multi = schedule_multi, - .schedule_pause = schedule_pause, - .schedule_resume = schedule_resume, - .schedule_release_atomic = schedule_release_atomic, - .schedule_release_ordered = schedule_release_ordered, - .schedule_prefetch = schedule_prefetch, - .schedule_num_prio = number_of_priorites, - .schedule_group_create = schedule_group_create, - .schedule_group_destroy = schedule_group_destroy, - .schedule_group_lookup = schedule_group_lookup, - .schedule_group_join = schedule_group_join, - .schedule_group_leave = schedule_group_leave, - .schedule_group_thrmask = schedule_group_thrmask, - .schedule_group_info = schedule_group_info, - .schedule_order_lock = schedule_order_lock, - .schedule_order_unlock = schedule_order_unlock, - .schedule_order_unlock_lock = schedule_order_unlock_lock, - .schedule_order_lock_start = schedule_order_lock_start, - .schedule_order_lock_wait = schedule_order_lock_wait -}; - -static void thread_set_interest(sched_thread_local_t *thread, - unsigned int queue_index, int prio) -{ - queue_index_sparse_t *index; - - if (thread == NULL) - return; - - if (prio >= NUM_SCHED_PRIO) - return; - - index = &thread->indexes[prio]; - - odp_rwlock_write_lock(&thread->lock); - sparse_bitmap_set(index, queue_index); - odp_rwlock_write_unlock(&thread->lock); -} - -static void thread_clear_interest(sched_thread_local_t *thread, - unsigned int queue_index, int prio) -{ - queue_index_sparse_t *index; - - if (thread == NULL) - return; - - if (prio >= NUM_SCHED_PRIO) - return; - - index = &thread->indexes[prio]; - - odp_rwlock_write_lock(&thread->lock); - sparse_bitmap_clear(index, queue_index); - odp_rwlock_write_unlock(&thread->lock); -} - -static void thread_set_interests(sched_thread_local_t *thread, - queue_index_bitmap_t *set) -{ - int prio; - sched_prio_t *P; - unsigned int queue_index; - queue_index_bitmap_t subset; - wapl_bitmap_iterator_t it; - - if (thread == NULL || set == NULL) - return; - - for (prio = 0; prio < NUM_SCHED_PRIO; prio++) { - P = &sched->prios[prio]; - odp_rwlock_read_lock(&P->lock); - - /* The collection of queue indexes in 'set' - * may belong to several priority levels. - */ - wapl_bitmap_zero(&subset); - wapl_bitmap_and(&subset, &P->queues, set); - - odp_rwlock_read_unlock(&P->lock); - - /* Add the subset to local indexes */ - wapl_bitmap_iterator(&it, &subset); - for (it.start(&it); it.has_next(&it);) { - queue_index = it.next(&it); - thread_set_interest(thread, queue_index, prio); - } - } -} - -static void thread_clear_interests(sched_thread_local_t *thread, - queue_index_bitmap_t *clear) -{ - int prio; - sched_prio_t *P; - unsigned int queue_index; - queue_index_bitmap_t subset; - wapl_bitmap_iterator_t it; - - if (thread == NULL || clear == NULL) - return; - - for (prio = 0; prio < NUM_SCHED_PRIO; prio++) { - P = &sched->prios[prio]; - odp_rwlock_read_lock(&P->lock); - - /* The collection of queue indexes in 'clear' - * may belong to several priority levels. - */ - wapl_bitmap_zero(&subset); - wapl_bitmap_and(&subset, &P->queues, clear); - - odp_rwlock_read_unlock(&P->lock); - - /* Remove the subset from local indexes */ - wapl_bitmap_iterator(&it, &subset); - for (it.start(&it); it.has_next(&it);) { - queue_index = it.next(&it); - thread_clear_interest(thread, queue_index, prio); - } - } -} - -static inline bool compete_atomic_queue(unsigned int queue_index) -{ - bool expected = sched->availables[queue_index]; - - if (expected && is_atomic_queue(queue_index)) { - expected = __atomic_compare_exchange_n( - &sched->availables[queue_index], - &expected, false, 0, - __ATOMIC_RELEASE, __ATOMIC_RELAXED); - } - - return expected; -} - -static inline int consume_queue(int prio, unsigned int queue_index) -{ - int count; - unsigned int max = MAX_DEQ; - event_cache_t *cache = &thread_local.cache; - - /* Low priorities have smaller batch size to limit - * head of line blocking latency. - */ - if (odp_unlikely(MAX_DEQ > 1 && prio > ODP_SCHED_PRIO_DEFAULT)) - max = MAX_DEQ / 2; - - /* For ordered queues we want consecutive events to - * be dispatched to separate threads, so do not cache - * them locally. - */ - if (is_ordered_queue(queue_index)) - max = 1; - - count = sched_queue_deq(queue_index, cache->stash, max, 1); - - if (count <= 0) - return 0; - - cache->top = &cache->stash[0]; - cache->count = count; - cache->queue = queue_from_index(queue_index); - return count; -} - -static inline bool do_schedule_prio(int prio) -{ - int nbits, next, end; - unsigned int queue_index; - sparse_bitmap_iterator_t *it; - - it = &thread_local.iterators[prio]; - nbits = (int)*it->_base.last; - - /* No interests at all! */ - if (nbits <= 0) - return false; - - /* In critical path, cannot afford iterator calls, - * do it manually with internal knowledge - */ - it->_start = (it->_start + 1) % nbits; - end = it->_start + nbits; - - for (next = it->_start; next < end; next++) { - queue_index = it->_base.il[next % nbits]; - - if (!compete_atomic_queue(queue_index)) - continue; - - if (!consume_queue(prio, queue_index)) - continue; - - return true; - } - - return false; -} commit fad375346ab5cab78b44e7de8417527ca77e7ca6 Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Mon Oct 15 09:44:10 2018 +0300 linux-gen: sched: remove iquery from interface Remove iquery from scheduler and queue interface. Signed-off-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_queue_if.c b/platform/linux-generic/odp_queue_if.c index 65f65006e..5ff8bbac5 100644 --- a/platform/linux-generic/odp_queue_if.c +++ b/platform/linux-generic/odp_queue_if.c @@ -103,9 +103,7 @@ int _odp_queue_init_global(void) if (sched == NULL || !strcmp(sched, "default")) sched = ODP_SCHEDULE_DEFAULT; - if (!strcmp(sched, "basic") || - !strcmp(sched, "sp") || - !strcmp(sched, "iquery")) { + if (!strcmp(sched, "basic") || !strcmp(sched, "sp")) { queue_fn = &queue_basic_fn; _odp_queue_api = &queue_basic_api; } else if (!strcmp(sched, "scalable")) { diff --git a/platform/linux-generic/odp_schedule_if.c b/platform/linux-generic/odp_schedule_if.c index df1ee2c26..c88b35fd5 100644 --- a/platform/linux-generic/odp_schedule_if.c +++ b/platform/linux-generic/odp_schedule_if.c @@ -19,9 +19,6 @@ extern const schedule_api_t schedule_sp_api; extern const schedule_fn_t schedule_basic_fn; extern const schedule_api_t schedule_basic_api; -extern const schedule_fn_t schedule_iquery_fn; -extern const schedule_api_t schedule_iquery_api; - extern const schedule_fn_t schedule_scalable_fn; extern const schedule_api_t schedule_scalable_api; @@ -154,9 +151,6 @@ int _odp_schedule_init_global(void) } else if (!strcmp(sched, "sp")) { sched_fn = &schedule_sp_fn; sched_api = &schedule_sp_api; - } else if (!strcmp(sched, "iquery")) { - sched_fn = &schedule_iquery_fn; - sched_api = &schedule_iquery_api; } else if (!strcmp(sched, "scalable")) { sched_fn = &schedule_scalable_fn; sched_api = &schedule_scalable_api; commit 59debc98d1d5ced95da8c95cdbc714184128f1e6 Author: Petri Savolainen <petri.savolainen(a)linaro.org> Date: Mon Oct 15 09:39:03 2018 +0300 ci: remove iquery tests Remove iquery scheduler from CI test scripts. Signed-off-by: Petri Savolainen <petri.savolainen(a)linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov(a)linaro.org> diff --git a/.shippable.yml b/.shippable.yml index 53e094bcb..5696be1e7 100644 --- a/.shippable.yml +++ b/.shippable.yml @@ -7,8 +7,6 @@ compiler: env: - CONF="--disable-test-perf --disable-test-perf-proc" - CONF="--disable-abi-compat --disable-test-perf --disable-test-perf-proc" - # - CONF="--enable-schedule-sp" - # - CONF="--enable-schedule-iquery" # - CONF="--enable-dpdk-zero-copy" # - CROSS_ARCH="arm64" # - CROSS_ARCH="armhf" CFLAGS="-march=armv7-a" @@ -39,8 +37,6 @@ build: - ./scripts/shippable-post.sh basic - ODP_SCHEDULER=sp make check - ./scripts/shippable-post.sh sp - - ODP_SCHEDULER=iquery make check - - ./scripts/shippable-post.sh iquery - ODP_SCHEDULER=scalable make check - ./scripts/shippable-post.sh scalable - rm -rf $HOME/odp-shmdir diff --git a/scripts/ci/coverage.sh b/scripts/ci/coverage.sh index 18e2e9806..9640b7114 100755 --- a/scripts/ci/coverage.sh +++ b/scripts/ci/coverage.sh @@ -23,7 +23,6 @@ export CI="true" ODP_SCHEDULER=basic make check ODP_SCHEDULER=sp make check -ODP_SCHEDULER=iquery make check ODP_SCHEDULER=scalable make check bash <(curl -s https://codecov.io/bash) -X coveragepy ----------------------------------------------------------------------- Summary of changes: .shippable.yml | 4 - platform/linux-generic/Makefile.am | 1 - platform/linux-generic/include/odp_schedule_if.h | 5 - platform/linux-generic/odp_queue_basic.c | 10 +- platform/linux-generic/odp_queue_if.c | 4 +- platform/linux-generic/odp_schedule_basic.c | 5 +- platform/linux-generic/odp_schedule_if.c | 6 - platform/linux-generic/odp_schedule_iquery.c | 1587 ---------------------- platform/linux-generic/odp_schedule_sp.c | 5 +- scripts/ci/coverage.sh | 1 - 10 files changed, 4 insertions(+), 1624 deletions(-) delete mode 100644 platform/linux-generic/odp_schedule_iquery.c hooks/post-receive --

7 years, 3 months

Jump to page:

2026

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

git-lng-odp