On 3/15/2021 7:27 PM, Stephen Hemminger wrote:
Posix mutex are not by default safe for protecting for usage
from multiple processes. The flow ops mutex could be used by
both primary and secondary processes.
Bugzilla ID: 662
Signed-off-by: Stephen Hemminger <step...@networkplumber.org>
Fixes: 80d1a9aff7f6 ("ethdev: make flow API thread safe")
Cc: suanmi...@nvidia.com
---
lib/librte_ethdev/rte_ethdev.c | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/lib/librte_ethdev/rte_ethdev.c b/lib/librte_ethdev/rte_ethdev.c
index 6f514c388b4e..d1024df408a5 100644
--- a/lib/librte_ethdev/rte_ethdev.c
+++ b/lib/librte_ethdev/rte_ethdev.c
@@ -470,6 +470,7 @@ rte_eth_dev_allocate(const char *name)
{
uint16_t port_id;
struct rte_eth_dev *eth_dev = NULL;
+ pthread_mutexattr_t attr;
size_t name_len;
name_len = strnlen(name, RTE_ETH_NAME_MAX_LEN);
@@ -506,7 +507,10 @@ rte_eth_dev_allocate(const char *name)
strlcpy(eth_dev->data->name, name, sizeof(eth_dev->data->name));
eth_dev->data->port_id = port_id;
eth_dev->data->mtu = RTE_ETHER_MTU;
- pthread_mutex_init(ð_dev->data->flow_ops_mutex, NULL);
+
+ pthread_mutexattr_init(&attr);
+ pthread_mutexattr_setpshared(&attr, PTHREAD_PROCESS_SHARED);
+ pthread_mutex_init(ð_dev->data->flow_ops_mutex, &attr);
unlock:
rte_spinlock_unlock(ð_dev_shared_data->ownership_lock);
The problem looks legit, and there seems no obvious downside of the patch. Also
Huawei test result was good, hence:
Acked-by: Ferruh Yigit <ferruh.yi...@intel.com>
And the PMDs that set the 'RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE' flag already not
affected from this change at all, but how safe mutli process applications will
be for them is still not clear.