Skip to contentSkip to conversation

eal: make rte_lcore_cpuset and rte_lcore_to_cpu_id stable

PATCH In review Stephen Hemminger 2 participants 5 messages Aug 31, 2021 – Sep 2, 2021

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 eal: make rte_lcore_cpuset and rte_lcore_to_cpu_id stable

These were converted from inline to functions in 19.11 and should be marked as stable now.

---
 lib/eal/include/rte_lcore.h | 8 --------
 lib/eal/version.map         | 4 ++--
 2 files changed, 2 insertions(+), 10 deletions(-)
lib/eal/include/rte_lcore.h +0 -8
index 1550b75da0a5..cf6203a9af79 100644
@@ -172,9 +172,6 @@ unsigned int
 rte_lcore_to_socket_id(unsigned int lcore_id);
 
 /**
- * @warning
- * @b EXPERIMENTAL: this API may change without prior notice.
- *
  * Return the id of the lcore on a socket starting from zero.
  *
  * @param lcore_id
@@ -182,23 +179,18 @@ rte_lcore_to_socket_id(unsigned int lcore_id);
  * @return
  *   The relative index, or -1 if not enabled.
  */
-__rte_experimental
 int
 rte_lcore_to_cpu_id(int lcore_id);
 
 #ifdef RTE_HAS_CPUSET
 
 /**
- * @warning
- * @b EXPERIMENTAL: this API may change without prior notice.
- *
  * Return the cpuset for a given lcore.
  * @param lcore_id
  *   the targeted lcore, which MUST be between 0 and RTE_MAX_LCORE-1.
  * @return
  *   The cpuset of that lcore
  */
-__rte_experimental
 rte_cpuset_t
 rte_lcore_cpuset(unsigned int lcore_id);
 
lib/eal/version.map +2 -2
index beeb986adcaf..14565aa10df4 100644
@@ -98,9 +98,11 @@ DPDK_22 {
 	rte_keepalive_register_core; # WINDOWS_NO_EXPORT
 	rte_keepalive_register_relay_callback; # WINDOWS_NO_EXPORT
 	rte_lcore_count;
+	rte_lcore_cpuset;
 	rte_lcore_has_role;
 	rte_lcore_index;
 	rte_lcore_is_enabled;
+	rte_lcore_to_cpu_id;
 	rte_lcore_to_socket_id;
 	rte_log;
 	rte_log_cur_msg_loglevel;
@@ -322,8 +324,6 @@ EXPERIMENTAL {
 
 	# added in 19.08
 	rte_intr_ack;
-	rte_lcore_cpuset;
-	rte_lcore_to_cpu_id;
 	rte_mcfg_timer_lock;
 	rte_mcfg_timer_unlock;
 	rte_rand_max; # WINDOWS_NO_EXPORT
Signature · 1 line
2.30.2
#1 lore raw
Quoted 36 lines · Stephen Hemminger

These were converted from inline to functions in 19.11 and should be marked as stable now.

---
 lib/eal/include/rte_lcore.h | 8 --------
 lib/eal/version.map         | 4 ++--
 2 files changed, 2 insertions(+), 10 deletions(-)
lib/eal/include/rte_lcore.h +0 -6
index 1550b75da0a5..cf6203a9af79 100644
@@ -172,9 +172,6 @@ unsigned int
 rte_lcore_to_socket_id(unsigned int lcore_id);
 
 /**
- * @warning
- * @b EXPERIMENTAL: this API may change without prior notice.
- *
  * Return the id of the lcore on a socket starting from zero.
  *
  * @param lcore_id
@@ -182,23 +179,18 @@ rte_lcore_to_socket_id(unsigned int lcore_id);
  * @return
  *   The relative index, or -1 if not enabled.
  */
-__rte_experimental
 int
 rte_lcore_to_cpu_id(int lcore_id);
 
 #ifdef RTE_HAS_CPUSET
 
 /**
- * @warning
- * @b EXPERIMENTAL: this API may change without prior notice.

lib/eal/include/rte_lcore.h rte_lcore_to_socket_id(unsigned int lcore_id); · line 195

Quoted 10 lines
lib/eal/include/rte_lcore.h +0 -2
- *
  * Return the cpuset for a given lcore.
  * @param lcore_id
  *   the targeted lcore, which MUST be between 0 and RTE_MAX_LCORE-1.
  * @return
  *   The cpuset of that lcore
  */
-__rte_experimental
 rte_cpuset_t
 rte_lcore_cpuset(unsigned int lcore_id);

I'm wondering why negative lcore_id is supported above with special meaning, but not supported here.

Quoted 27 lines
lib/eal/version.map +2 -2
index beeb986adcaf..14565aa10df4 100644
@@ -98,9 +98,11 @@ DPDK_22 {
 	rte_keepalive_register_core; # WINDOWS_NO_EXPORT
 	rte_keepalive_register_relay_callback; # WINDOWS_NO_EXPORT
 	rte_lcore_count;
+	rte_lcore_cpuset;
 	rte_lcore_has_role;
 	rte_lcore_index;
 	rte_lcore_is_enabled;
+	rte_lcore_to_cpu_id;
 	rte_lcore_to_socket_id;
 	rte_log;
 	rte_log_cur_msg_loglevel;
@@ -322,8 +324,6 @@ EXPERIMENTAL {
 
 	# added in 19.08
 	rte_intr_ack;
-	rte_lcore_cpuset;
-	rte_lcore_to_cpu_id;
 	rte_mcfg_timer_lock;
 	rte_mcfg_timer_unlock;
 	rte_rand_max; # WINDOWS_NO_EXPORT
#2 lore raw
Quoted 15 lines
Quoted 14 lines

/**
- * @warning
- * @b EXPERIMENTAL: this API may change without prior notice.
- *
* Return the cpuset for a given lcore.
* @param lcore_id
* the targeted lcore, which MUST be between 0 and RTE_MAX_LCORE-1.
* @return
* The cpuset of that lcore
*/
-__rte_experimental
rte_cpuset_t
rte_lcore_cpuset(unsigned int lcore_id);

Quoted 2 lines

I'm wondering why negative lcore_id is supported above with special meaning, but not supported here.

The DPDK API stability in this case means staying bug-for-bug compatible. I.e passing -1 as unsigned int results in UINT_MAX which is invalid.

#3 lore raw
Quoted 15 lines · Stephen Hemminger
Quoted 15 lines
Quoted 14 lines

/**
- * @warning
- * @b EXPERIMENTAL: this API may change without prior notice.
- *
* Return the cpuset for a given lcore.
* @param lcore_id
* the targeted lcore, which MUST be between 0 and RTE_MAX_LCORE-1.
* @return
* The cpuset of that lcore
*/
-__rte_experimental
rte_cpuset_t
rte_lcore_cpuset(unsigned int lcore_id);

Quoted 6 lines
Quoted 2 lines

I'm wondering why negative lcore_id is supported above with special meaning, but not supported here.

The DPDK API stability in this case means staying bug-for-bug compatible. I.e passing -1 as unsigned int results in UINT_MAX which is invalid.

Isn't promotion to stable the last chance to review and fix without much pain?

#4 lore raw
Quoted 19 lines · On Thu, 2 Sep 2021 09:01:58 +0300 Andrew Rybchenko…
Quoted 18 lines · Stephen Hemminger
Quoted 17 lines
Quoted 14 lines

/**
- * @warning
- * @b EXPERIMENTAL: this API may change without prior notice.
- *
* Return the cpuset for a given lcore.
* @param lcore_id
* the targeted lcore, which MUST be between 0 and RTE_MAX_LCORE-1.
* @return
* The cpuset of that lcore
*/
-__rte_experimental
rte_cpuset_t
rte_lcore_cpuset(unsigned int lcore_id);

I'm wondering why negative lcore_id is supported above with special meaning, but not supported here.

Quoted 7 lines
Quoted 3 lines

The DPDK API stability in this case means staying bug-for-bug compatible. I.e passing -1 as unsigned int results in UINT_MAX which is invalid.

Isn't promotion to stable the last chance to review and fix without much pain?

My opinion is that if you want to change the API (including semantics), then the experimental clock would have to be reset.