On Fri, Mar 17, 2023 at 01:19:45PM -0700, Tyler Retzlaff wrote: > Replace the use of rte_atomic.h types and functions, instead use GCC > supplied C++11 memory model builtins. > > Signed-off-by: Tyler Retzlaff <roret...@linux.microsoft.com> > --- > drivers/net/ice/ice_dcf.c | 1 - > drivers/net/ice/ice_dcf_ethdev.c | 1 - > drivers/net/ice/ice_ethdev.c | 10 ++++++---- > 3 files changed, 6 insertions(+), 6 deletions(-) > > diff --git a/drivers/net/ice/ice_dcf.c b/drivers/net/ice/ice_dcf.c > index 1c3d22a..80d2cbd 100644 > --- a/drivers/net/ice/ice_dcf.c > +++ b/drivers/net/ice/ice_dcf.c > @@ -14,7 +14,6 @@ > #include <rte_common.h> > > #include <rte_pci.h> > -#include <rte_atomic.h> > #include <rte_eal.h> > #include <rte_ether.h> > #include <ethdev_driver.h> > diff --git a/drivers/net/ice/ice_dcf_ethdev.c > b/drivers/net/ice/ice_dcf_ethdev.c > index dcbf2af..13ff245 100644 > --- a/drivers/net/ice/ice_dcf_ethdev.c > +++ b/drivers/net/ice/ice_dcf_ethdev.c > @@ -11,7 +11,6 @@ > #include <rte_interrupts.h> > #include <rte_debug.h> > #include <rte_pci.h> > -#include <rte_atomic.h> > #include <rte_eal.h> > #include <rte_ether.h> > #include <ethdev_pci.h> > diff --git a/drivers/net/ice/ice_ethdev.c b/drivers/net/ice/ice_ethdev.c > index 9a88cf9..bdf4569 100644 > --- a/drivers/net/ice/ice_ethdev.c > +++ b/drivers/net/ice/ice_ethdev.c > @@ -3927,8 +3927,9 @@ static int ice_init_rss(struct ice_pf *pf) > struct rte_eth_link *dst = link; > struct rte_eth_link *src = &dev->data->dev_link; > > - if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst, > - *(uint64_t *)src) == 0) > + if (!__atomic_compare_exchange_n((uint64_t *)dst, > + (uint64_t *)dst, *(uint64_t *)src, 0, > + __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST)) > return -1; > > return 0; > @@ -3941,8 +3942,9 @@ static int ice_init_rss(struct ice_pf *pf) > struct rte_eth_link *dst = &dev->data->dev_link; > struct rte_eth_link *src = link; > > - if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst, > - *(uint64_t *)src) == 0) > + if (!__atomic_compare_exchange_n((uint64_t *)dst, > + (uint64_t *)dst, *(uint64_t *)src, 0, > + __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST)) > return -1; >
*(uint64_t *)dst for the second parameter look like a bug to me, a non-atomic load will be generated. probably this code should be corrected by performing __atomic_load_n(dst, ...) to a stack variable and then performing the cmpset/compare_exchange.
