[PATCH net-next] net/smc: Optimize the timing of unlocking in smc_listen_work

[PATCH net-next] net/smc: Optimize the timing of unlocking in smc_listen_work

[liqiang]

The optimized code is equivalent to the original process, and it releases the
lock early.

Signed-off-by: liqiang <liqiang64@huawei.com>
---
 net/smc/af_smc.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c
index 9d76e902fd77..7fa80be1ea93 100644
--- a/net/smc/af_smc.c
+++ b/net/smc/af_smc.c
@@ -2526,9 +2526,9 @@ static void smc_listen_work(struct work_struct *work)
 	if (!ini->is_smcd) {
 		rc = smc_listen_rdma_finish(new_smc, cclc,
 					    ini->first_contact_local, ini);
-		if (rc)
-			goto out_unlock;
 		mutex_unlock(&smc_server_lgr_pending);
+		if (rc)
+			goto out_decl;
 	}
 	smc_conn_save_peer_info(new_smc, cclc);
 
-- 
2.43.0

Re: [PATCH net-next] net/smc: Optimize the timing of unlocking in smc_listen_work

[Jakub Kicinski]

On Sat, 30 Nov 2024 16:26:30 +0800 liqiang wrote:
> The optimized code is equivalent to the original process, and it releases the
> lock early.

By a single clock cycle? You need to provide much more detailed
justification, otherwise this looks like churn for no real gain.
-- 
pw-bot: cr

Re: [PATCH net-next] net/smc: Optimize the timing of unlocking in smc_listen_work

[Li Qiang]

在 2024/12/1 3:04, Jakub Kicinski 写道:
> On Sat, 30 Nov 2024 16:26:30 +0800 liqiang wrote:
>> The optimized code is equivalent to the original process, and it releases the
>> lock early.
> 
> By a single clock cycle? You need to provide much more detailed
> justification, otherwise this looks like churn for no real gain.

I don't have any more information about this patch, I saw that
the lock will be released regardless of branch true or false,
so brought it forward.

Thanks for your reply, Jakub!
:)

-- 
Cheers,
Li Qiang