alarm related patches
cover alarm related patches
This is the set of things which stood out as needing fixing when looking at the rte_eal_alarm tests.
Stephen Hemminger (5):
eal: add missing parameter check to rte_eal_alarm_set on Windows Revert "test/alarm: disable bad time cases on Windows"
test: support alarm test on FreeBSD
test/alarm: rewrite the alarm test
eal: simplify eal alarm cancel by using LIST_FOREACH_SAFE
app/test/test_alarm.c | 84 +++++++++++++++++++++---------------- lib/eal/linux/eal_alarm.c | 51 +++++++++------------- lib/eal/windows/eal_alarm.c | 7 ++++ 3 files changed, 75 insertions(+), 67 deletions(-)
Signature · 1 line
2.43.0
PATCH 1/5 eal: add missing parameter check to rte_eal_alarm_set on Windows
Both Linux and FreeBSD, check parameters to rte_eal_alarm_set, but Windows missed this. And the test was "fixed" instead.
--- lib/eal/windows/eal_alarm.c | 7 +++++++ 1 file changed, 7 insertions(+)
lib/eal/windows/eal_alarm.c +7 -0
@@ -92,6 +92,13 @@ rte_eal_alarm_set(uint64_t us, rte_eal_alarm_callback cb_fn, void *cb_arg)LARGE_INTEGER deadline;int ret; +/* Check parameters, including that us won't cause a uint64_t overflow */ +if (us <1 || us > (UINT64_MAX -US_PER_S )) { +EAL_LOG (ERR ,"Invalid alarm interval" ); + ret = -EINVAL ; + goto exit; + } +if (cb_fn ==NULL ) {EAL_LOG (ERR ,"NULL callback" ); ret = -EINVAL ;
Signature · 1 line
2.43.0
PATCH 2/5 Revert "test/alarm: disable bad time cases on Windows"
In reply to Stephen HemmingerThis reverts commit a089d320338d708f5b7126dab5fd6861c82e6347.
Windows EAL should have been fixed rather than papering over the bug.
--- app/test/test_alarm.c | 4 ---- 1 file changed, 4 deletions(-)
app/test/test_alarm.c +0 -4
@@ -10,7 +10,6 @@#include "test.h" -#ifndef RTE_EXEC_ENV_WINDOWS static volatile int flag; static void
@@ -19,7 +18,6 @@ test_alarm_callback(void *cb_arg) flag =1 ;printf ("Callback setting flag - OK. [cb_arg = %p]\n" , cb_arg); } -#endif static int test_alarm (void )
@@ -29,7 +27,6 @@ test_alarm(void)return 0 ;#endif -#ifndef RTE_EXEC_ENV_WINDOWS /* check if it will fail to set alarm with wrong us value */ printf ("check if it will fail to set alarm with wrong ms values\n" ); if (rte_eal_alarm_set(0, test_alarm_callback,
@@ -42,7 +39,6 @@ test_alarm(void)printf ("should not be successful with (UINT64_MAX-1) us value\n" );return -1 ; } -#endif /* check if it will fail to set alarm with null callback parameter */ printf ("check if it will fail to set alarm with null callback parameter\n" );
Signature · 1 line
2.43.0
Re: [PATCH 2/5] Revert "test/alarm: disable bad time cases on Windows"
2024-08-08 12:46 (UTC-0700), Stephen Hemminger:
Quoted 4 lines
This reverts commit
a089d320338d708f5b7126dab5fd6861c82e6347.Windows EAL should have been fixed rather than papering over the bug.
Linux and FreeBSD alarm implementations use the same approach that limits possible timeout range in API.
Test cases in question check that these values are rejected. Windows EAL can accept any values.
I think the proper fix would be documenting Unix limitations at API level (it never worked the other way, so no real breakage), then adding the same checks to Windows EAL only for consistency.
Re: [PATCH 2/5] Revert "test/alarm: disable bad time cases on Windows"
2024-08-09 10:23 (UTC+0300), Dmitry Kozlyuk:
Quoted 6 lines
2024-08-08 12:46 (UTC-0700), Stephen Hemminger:
Quoted 4 lines
This reverts commit
a089d320338d708f5b7126dab5fd6861c82e6347.Windows EAL should have been fixed rather than papering over the bug.
Quoted 7 lines
Linux and FreeBSD alarm implementations use the same approach that limits possible timeout range in API.
Test cases in question check that these values are rejected. Windows EAL can accept any values.
I think the proper fix would be documenting Unix limitations at API level (it never worked the other way, so no real breakage), then adding the same checks to Windows EAL only for consistency.
Oops, I got patch 1/5 in email later somehow. It is doing just that. Sorry for the noise.
Re: [PATCH 2/5] Revert "test/alarm: disable bad time cases on Windows"
In reply to Dmitry KozlyukQuoted 10 lines · On Fri, 9 Aug 2024 10:23:24 +0300 Dmitry Kozlyuk <[email protected]>…
2024-08-08 12:46 (UTC-0700), Stephen Hemminger:
Quoted 4 lines
This reverts commit
a089d320338d708f5b7126dab5fd6861c82e6347.Windows EAL should have been fixed rather than papering over the bug.
Linux and FreeBSD alarm implementations use the same approach that limits possible timeout range in API.
Test cases in question check that these values are rejected. Windows EAL can accept any values.
That was a bug. There should not be different behavior on different OS.
Quoted 3 lines
I think the proper fix would be documenting Unix limitations at API level (it never worked the other way, so no real breakage), then adding the same checks to Windows EAL only for consistency.
No. All EAL should behave the same.
PATCH 3/5 test: support alarm test on FreeBSD
In reply to Stephen HemmingerYes the alarm API is supported on FreeBSD.
--- app/test/test_alarm.c | 5 ----- 1 file changed, 5 deletions(-)
app/test/test_alarm.c +0 -5
@@ -22,11 +22,6 @@ test_alarm_callback(void *cb_arg)static int test_alarm (void ) { -#ifdef RTE_EXEC_ENV_FREEBSD -printf ("The alarm API is not supported on FreeBSD\n" ); -return 0 ; -#endif -/* check if it will fail to set alarm with wrong us value */ printf ("check if it will fail to set alarm with wrong ms values\n" ); if (rte_eal_alarm_set(0, test_alarm_callback,
Signature · 1 line
2.43.0
PATCH 4/5 test/alarm: rewrite the alarm test
In reply to Stephen HemmingerThe alarm test had several issues:
- only negative tests were done, no positive test to make sure API worked
- the set/cancel test had too short an interval and the alarm might expire
- the test should have used the standard TEST macros.
--- app/test/test_alarm.c | 77 ++++++++++++++++++++++++++++--------------- 1 file changed, 50 insertions(+), 27 deletions(-)
app/test/test_alarm.c +50 -27
@@ -3,53 +3,76 @@*/ #include <stdio.h> +#include <stdbool.h> #include <stdint.h> -#include <rte_common.h> #include <rte_alarm.h> +#include <rte_atomic.h> +#include <rte_common.h> +#include <rte_cycles.h> +#include <rte_errno.h> #include "test.h" -static volatile int flag; +#ifndef US_PER_SEC +#define US_PER_SEC 1000000u +#endif static void test_alarm_callback (void *cb_arg) { - flag =1 ; -printf ("Callback setting flag - OK. [cb_arg = %p]\n" , cb_arg); +bool *triggered = cb_arg; + +rte_atomic_store_explicit (triggered,1 , rte_memory_order_release); } +static int test_alarm (void ) { -/* check if it will fail to set alarm with wrong us value */ -printf ("check if it will fail to set alarm with wrong ms values\n" ); -if (rte_eal_alarm_set (0 , test_alarm_callback, -NULL ) >=0 ) { -printf ("should not be successful with 0 us value\n" ); -return -1 ; - } -if (rte_eal_alarm_set (UINT64_MAX -1 , test_alarm_callback, -NULL ) >=0 ) { -printf ("should not be successful with (UINT64_MAX-1) us value\n" ); -return -1 ; - } +bool triggered = false; +bool later = false; +int ret; + +/* check if it will fail to set alarm with wrong us values */ +TEST_ASSERT_FAIL (rte_eal_alarm_set (0 , test_alarm_callback,NULL ), +"Expected rte_eal_alarm_set to fail with 0 us value" ); + +/* check it if will fail with a very large timeout value */ +TEST_ASSERT_FAIL (rte_eal_alarm_set (UINT64_MAX -1 , test_alarm_callback,NULL ), +"Expected rte_eal_alarm_set to fail with (UINT64_MAX-1) us value" );/* check if it will fail to set alarm with null callback parameter */ -printf ("check if it will fail to set alarm with null callback parameter\n" ); -if (rte_eal_alarm_set (10 /* ms */ ,NULL ,NULL ) >=0 ) { -printf ("should not be successful to set alarm with null callback parameter\n" ); -return -1 ; - } +TEST_ASSERT_FAIL (rte_eal_alarm_set (US_PER_SEC ,NULL ,NULL ), +"Expected rte_eal_alarm_set to fail with null callback parameter" );/* check if it will fail to remove alarm with null callback parameter */ -printf ("check if it will fail to remove alarm with null callback parameter\n" ); -if (rte_eal_alarm_cancel (NULL ,NULL ) ==0 ) { -printf ("should not be successful to remove alarm with null callback parameter" ); -return -1 ; - } +TEST_ASSERT_FAIL (rte_eal_alarm_cancel (NULL ,NULL ), +"Expected rte_eal_alarm_cancel to fail with null callback parameter" ); + +/* check if can set a alarm for one second */ +TEST_ASSERT_SUCCESS (rte_eal_alarm_set (US_PER_SEC , test_alarm_callback, &triggered), +"Setting one second alarm failed" ); + +/* set a longer alarm that will be canceled. */ +TEST_ASSERT_SUCCESS (rte_eal_alarm_set (10 *US_PER_SEC , test_alarm_callback, &later), +"Setting ten second alarm failed" ); + +/* wait for alarm to happen */ +while (rte_atomic_load_explicit (&triggered, rte_memory_order_acquire) == false) +rte_delay_us_sleep (US_PER_SEC /10 ); + +TEST_ASSERT (!rte_atomic_load_explicit (&later, rte_memory_order_acquire), +"Only one alarm should have fired." ); + + ret =rte_eal_alarm_cancel (test_alarm_callback, &triggered); +TEST_ASSERT (ret ==0 && rte_errno ==ENOENT , +"Canceling alarm after run ret %d: %s" , ret,rte_strerror (rte_errno)); + + ret =rte_eal_alarm_cancel (test_alarm_callback, &later); +TEST_ASSERT (ret ==1 ,"Canceling ten second alarm failed %d: %s" , ret,rte_strerror (rte_errno));return 0 ; } -REGISTER_TEST_COMMAND (alarm_autotest, test_alarm); +REGISTER_FAST_TEST (alarm_autotest, true, true, test_alarm);
Signature · 1 line
2.43.0
PATCH 5/5 eal: simplify eal alarm cancel by using LIST_FOREACH_SAFE
In reply to Stephen HemmingerThe code for rte_eal_alarm_cancel was using two loops and duplicate code to handle cancel at the start of the list. Introduce the LIST_FOREACH_SAFE() macro from FreeBsd which makes writing the loop cleaner and simpler.
--- lib/eal/linux/eal_alarm.c | 51 +++++++++++++++------------------------ 1 file changed, 19 insertions(+), 32 deletions(-)
lib/eal/linux/eal_alarm.c +19 -32
@@ -53,6 +53,13 @@ static struct rte_intr_handle *intr_handle;static int handler_registered =0 ;static void eal_alarm_callback (void *arg); +#ifndef LIST_FOREACH_SAFE +#define LIST_FOREACH_SAFE (var, head, field, tvar) \ + for ((var) = LIST_FIRST((head)); \ + (var) && ((tvar) = LIST_NEXT((var), field), 1); \ + (var) = (tvar)) +#endif +void rte_eal_alarm_cleanup (void ) {
@@ -194,7 +201,7 @@ rte_eal_alarm_set(uint64_t us, rte_eal_alarm_callback cb_fn, void *cb_arg)int rte_eal_alarm_cancel (rte_eal_alarm_callback cb_fn,void *cb_arg) { -struct alarm_entry *ap, *ap_prev; +struct alarm_entry *ap, *tmp;int count =0 ;int err =0 ;int executing;
@@ -207,46 +214,26 @@ rte_eal_alarm_cancel(rte_eal_alarm_callback cb_fn, void *cb_arg)do { executing =0 ;rte_spinlock_lock (&alarm_list_lk); -/* remove any matches at the start of the list */ -while ((ap =LIST_FIRST (&alarm_list)) !=NULL && - cb_fn == ap->cb_fn && - (cb_arg == (void *)-1 || cb_arg == ap->cb_arg)) { - -if (ap->executing ==0 ) { -LIST_REMOVE (ap, next); -free (ap); - count++; - }else { -/* If calling from other context, mark that alarm is executing -* so loop can spin till it finish. Otherwise we are trying to -* cancel our self - mark it by EINPROGRESS */ -if (pthread_equal (ap->executing_id,pthread_self ()) ==0 ) - executing++; -else - err =EINPROGRESS ; - -break ; - } - } - ap_prev = ap; -/* now go through list, removing entries not at start */ - LIST_FOREACH(ap, &alarm_list, next) { + LIST_FOREACH_SAFE(ap, &alarm_list, next, tmp) {/* this won't be true first time through */ if (cb_fn == ap->cb_fn && (cb_arg == (void *)-1 || cb_arg == ap->cb_arg)) { -if (ap->executing ==0 ) {LIST_REMOVE (ap, next);free (ap); count++; - ap = ap_prev; - }else if (pthread_equal (ap->executing_id,pthread_self ()) ==0 ) - executing++; -else - err =EINPROGRESS ; + }else { +/* If calling from other context, mark that alarm is executing +* so loop can spin till it finish. Otherwise we are trying to +* cancel our self - mark it by EINPROGRESS +*/ +if (pthread_equal (ap->executing_id,pthread_self ()) ==0 ) + executing++; +else + err =EINPROGRESS ; + } } - ap_prev = ap; }rte_spinlock_unlock (&alarm_list_lk); }while (executing !=0 );
Signature · 1 line
2.43.0
RE: [PATCH 5/5] eal: simplify eal alarm cancel by using LIST_FOREACH_SAFE
lib/eal/linux/eal_alarm.c static struct rte_intr_handle *intr_handle; · line 61
Quoted 11 lines
+++
b/lib/eal/linux/eal_alarm.clib/eal/linux/eal_alarm.c +6
-0@@ -53,6 +53,13 @@ static struct rte_intr_handle *intr_handle;static int handler_registered =0 ;static void eal_alarm_callback (void *arg); +#ifndef LIST_FOREACH_SAFE +#define LIST_FOREACH_SAFE (var, head, field, tvar) \ + for ((var) = LIST_FIRST((head)); \ + (var) && ((tvar) = LIST_NEXT((var), field), 1); \ + (var) = (tvar)) +#endif
This macro is already defined for Windows [1]; isn't it also defined in some Linux/BSD standard header file already?
[1]: https://elixir.bootlin.com/dpdk/v24.07/source/lib/eal/windows/include/sys/queue.h#L515
Re: [PATCH 5/5] eal: simplify eal alarm cancel by using LIST_FOREACH_SAFE
Quoted 14 lines · On Fri, 9 Aug 2024 10:33:49 +0200 Morten Brørup <[email protected]>…
lib/eal/linux/eal_alarm.c static struct rte_intr_handle *intr_handle; · line 61
Quoted 11 lines
+++
b/lib/eal/linux/eal_alarm.clib/eal/linux/eal_alarm.c +6
-0@@ -53,6 +53,13 @@ static struct rte_intr_handle *intr_handle;static int handler_registered =0 ;static void eal_alarm_callback (void *arg); +#ifndef LIST_FOREACH_SAFE +#define LIST_FOREACH_SAFE (var, head, field, tvar) \ + for ((var) = LIST_FIRST((head)); \ + (var) && ((tvar) = LIST_NEXT((var), field), 1); \ + (var) = (tvar)) +#endif This macro is already defined for Windows [1]; isn't it also defined in some Linux/BSD standard header file already?
Quoted 2 lines
[1]: https://elixir.bootlin.com/dpdk/v24.07/source/lib/eal/windows/include/sys/queue.h#L515
Unfortunately, sys/queue.h on Linux is stuck on some old version. There maybe more complete version on some libbsd headers but don't want to get into having that dependency.
v2 cover alarm test fixes
In reply to Stephen HemmingerThis is the set of things which stood out as needing fixing when looking at the rte_eal_alarm tests.
v2 - drop cleanup of eal alarm code for later
- add RTE_ATOMIC() to fix clang build
Stephen Hemminger (3):
eal: add missing parameter check to rte_eal_alarm_set on Windows
test: support alarm test on FreeBSD
test/alarm: rewrite the alarm test
app/test/test_alarm.c | 85 ++++++++++++++++++++++--------------- lib/eal/windows/eal_alarm.c | 7 +++ 2 files changed, 57 insertions(+), 35 deletions(-)
Signature · 1 line
2.43.0
PATCH v2 1/3 eal: add missing parameter check to rte_eal_alarm_set on Windows
Both Linux and FreeBSD, check parameters to rte_eal_alarm_set, but Windows missed this. And the test was "fixed" instead.
This reverts commit a089d320338d708f5b7126dab5fd6861c82e6347.
Windows EAL should have been fixed rather than papering over the bug.
--- app/test/test_alarm.c | 4 ---- lib/eal/windows/eal_alarm.c | 7 +++++++ 2 files changed, 7 insertions(+), 4 deletions(-)
app/test/test_alarm.c +0 -4
@@ -10,7 +10,6 @@#include "test.h" -#ifndef RTE_EXEC_ENV_WINDOWS static volatile int flag; static void
@@ -19,7 +18,6 @@ test_alarm_callback(void *cb_arg) flag =1 ;printf ("Callback setting flag - OK. [cb_arg = %p]\n" , cb_arg); } -#endif static int test_alarm (void )
@@ -29,7 +27,6 @@ test_alarm(void)return 0 ;#endif -#ifndef RTE_EXEC_ENV_WINDOWS /* check if it will fail to set alarm with wrong us value */ printf ("check if it will fail to set alarm with wrong ms values\n" ); if (rte_eal_alarm_set(0, test_alarm_callback,
@@ -42,7 +39,6 @@ test_alarm(void)printf ("should not be successful with (UINT64_MAX-1) us value\n" );return -1 ; } -#endif /* check if it will fail to set alarm with null callback parameter */ printf ("check if it will fail to set alarm with null callback parameter\n" );
lib/eal/windows/eal_alarm.c +7 -0
@@ -92,6 +92,13 @@ rte_eal_alarm_set(uint64_t us, rte_eal_alarm_callback cb_fn, void *cb_arg)LARGE_INTEGER deadline;int ret; +/* Check parameters, including that us won't cause a uint64_t overflow */ +if (us <1 || us > (UINT64_MAX -US_PER_S )) { +EAL_LOG (ERR ,"Invalid alarm interval" ); + ret = -EINVAL ; + goto exit; + } +if (cb_fn ==NULL ) {EAL_LOG (ERR ,"NULL callback" ); ret = -EINVAL ;
Signature · 1 line
2.43.0
PATCH v2 2/3 test: support alarm test on FreeBSD
In reply to Stephen HemmingerYes the alarm API is supported on FreeBSD.
--- app/test/test_alarm.c | 5 ----- 1 file changed, 5 deletions(-)
app/test/test_alarm.c +0 -5
@@ -22,11 +22,6 @@ test_alarm_callback(void *cb_arg)static int test_alarm (void ) { -#ifdef RTE_EXEC_ENV_FREEBSD -printf ("The alarm API is not supported on FreeBSD\n" ); -return 0 ; -#endif -/* check if it will fail to set alarm with wrong us value */ printf ("check if it will fail to set alarm with wrong ms values\n" ); if (rte_eal_alarm_set(0, test_alarm_callback,
Signature · 1 line
2.43.0
PATCH v2 3/3 test/alarm: rewrite the alarm test
In reply to Stephen HemmingerThe alarm test had several issues:
- only negative tests were done, no positive test to make sure API worked
- the set/cancel test had too short an interval and the alarm might expire
- the test should have used the standard TEST macros.
--- app/test/test_alarm.c | 78 ++++++++++++++++++++++++++++--------------- 1 file changed, 51 insertions(+), 27 deletions(-)
app/test/test_alarm.c +51 -27
@@ -3,53 +3,77 @@*/ #include <stdio.h> +#include <stdbool.h> #include <stdint.h> -#include <rte_common.h> #include <rte_alarm.h> +#include <rte_atomic.h> +#include <rte_common.h> +#include <rte_cycles.h> +#include <rte_errno.h> #include "test.h" -static volatile int flag; +#ifndef US_PER_SEC +#define US_PER_SEC 1000000u +#endif static void test_alarm_callback (void *cb_arg) { - flag =1 ; -printf ("Callback setting flag - OK. [cb_arg = %p]\n" , cb_arg); +RTE_ATOMIC (bool ) *triggered = cb_arg; + +rte_atomic_store_explicit (triggered,1 , rte_memory_order_release); } +static int test_alarm (void ) { -/* check if it will fail to set alarm with wrong us value */ -printf ("check if it will fail to set alarm with wrong ms values\n" ); -if (rte_eal_alarm_set (0 , test_alarm_callback, -NULL ) >=0 ) { -printf ("should not be successful with 0 us value\n" ); -return -1 ; - } -if (rte_eal_alarm_set (UINT64_MAX -1 , test_alarm_callback, -NULL ) >=0 ) { -printf ("should not be successful with (UINT64_MAX-1) us value\n" ); -return -1 ; - } +RTE_ATOMIC (bool ) triggered = false; +RTE_ATOMIC (bool ) later = false; +int ret; + +/* check if it will fail to set alarm with wrong us values */ +TEST_ASSERT_FAIL (rte_eal_alarm_set (0 , test_alarm_callback,NULL ), +"Expected rte_eal_alarm_set to fail with 0 us value" ); + +/* check it if will fail with a very large timeout value */ +TEST_ASSERT_FAIL (rte_eal_alarm_set (UINT64_MAX -1 , test_alarm_callback,NULL ), +"Expected rte_eal_alarm_set to fail with (UINT64_MAX-1) us value" );/* check if it will fail to set alarm with null callback parameter */ -printf ("check if it will fail to set alarm with null callback parameter\n" ); -if (rte_eal_alarm_set (10 /* ms */ ,NULL ,NULL ) >=0 ) { -printf ("should not be successful to set alarm with null callback parameter\n" ); -return -1 ; - } +TEST_ASSERT_FAIL (rte_eal_alarm_set (US_PER_SEC ,NULL ,NULL ), +"Expected rte_eal_alarm_set to fail with null callback parameter" );/* check if it will fail to remove alarm with null callback parameter */ -printf ("check if it will fail to remove alarm with null callback parameter\n" ); -if (rte_eal_alarm_cancel (NULL ,NULL ) ==0 ) { -printf ("should not be successful to remove alarm with null callback parameter" ); -return -1 ; - } +TEST_ASSERT_FAIL (rte_eal_alarm_cancel (NULL ,NULL ), +"Expected rte_eal_alarm_cancel to fail with null callback parameter" ); + +/* check if can set a alarm for one second */ +TEST_ASSERT_SUCCESS (rte_eal_alarm_set (US_PER_SEC , test_alarm_callback, &triggered), +"Setting one second alarm failed" ); + +/* set a longer alarm that will be canceled. */ +TEST_ASSERT_SUCCESS (rte_eal_alarm_set (10 *US_PER_SEC , test_alarm_callback, &later), +"Setting ten second alarm failed" ); + +/* wait for alarm to happen */ +while (rte_atomic_load_explicit (&triggered, rte_memory_order_acquire) == false) +rte_delay_us_sleep (US_PER_SEC /10 ); + +TEST_ASSERT (!rte_atomic_load_explicit (&later, rte_memory_order_acquire), +"Only one alarm should have fired." ); + + ret =rte_eal_alarm_cancel (test_alarm_callback, &triggered); +TEST_ASSERT (ret ==0 && rte_errno ==ENOENT , +"Canceling alarm after run ret %d: %s" , ret,rte_strerror (rte_errno)); + + ret =rte_eal_alarm_cancel (test_alarm_callback, &later); +TEST_ASSERT (ret ==1 ,"Canceling ten second alarm failed %d: %s" , + ret,rte_strerror (rte_errno));return 0 ; } -REGISTER_TEST_COMMAND (alarm_autotest, test_alarm); +REGISTER_FAST_TEST (alarm_autotest, true, true, test_alarm);
Signature · 1 line
2.43.0
Re: [PATCH v2 0/3] alarm test fixes
In reply to Stephen HemmingerLGTM
Quoted 15 lines · Stephen Hemminger
This is the set of things which stood out as needing fixing when looking at the
rte_eal_alarmtests.v2 - drop cleanup of eal alarm code for later
- addRTE_ATOMIC()to fix clang buildStephen Hemminger (3):
eal: add missing parameter check torte_eal_alarm_seton Windows
test: support alarm test on FreeBSD
test/alarm: rewrite the alarm testapp/test/test_alarm.c | 85 ++++++++++++++++++++++--------------- lib/eal/windows/eal_alarm.c | 7 +++ 2 files changed, 57 insertions(+), 35 deletions(-)
Re: [PATCH v2 0/3] alarm test fixes
In reply to Stephen HemmingerQuoted 7 lines · Stephen Hemminger
This is the set of things which stood out as needing fixing when looking at the
rte_eal_alarmtests.v2 - drop cleanup of eal alarm code for later
- addRTE_ATOMIC()to fix clang build
Quoted 8 lines
Stephen Hemminger (3):
eal: add missing parameter check torte_eal_alarm_seton Windows
test: support alarm test on FreeBSD
test/alarm: rewrite the alarm testapp/test/test_alarm.c | 85 ++++++++++++++++++++++--------------- lib/eal/windows/eal_alarm.c | 7 +++ 2 files changed, 57 insertions(+), 35 deletions(-)
Thanks for the cleanup, and reviews.
Series applied.
Signature · 1 line
David Marchand