>
> It is tiny probable to collide with running timer, and we can probe colliding,
> then the overhead of unlock/lock is removed.
>
> Any comment is appreciated.
>
> Thanks
>
> Signed-off-by: Hillf Danton <[email protected]>
> ---
> drivers/scsi/libfc/fc_fcp.c | 12 ++++++------
> 1 files changed, 6 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/scsi/libfc/fc_fcp.c b/drivers/scsi/libfc/fc_fcp.c
> index 9cd2149..725f308 100644
> --- a/drivers/scsi/libfc/fc_fcp.c
> +++ b/drivers/scsi/libfc/fc_fcp.c
> @@ -1272,11 +1272,11 @@ static int fc_lun_reset(struct fc_lport
> *lport, struct fc_fcp_pkt *fsp,
>
> spin_lock_bh(&fsp->scsi_pkt_lock);
> fsp->state |= FC_SRB_COMPL;
> - spin_unlock_bh(&fsp->scsi_pkt_lock);
> -
> - del_timer_sync(&fsp->timer);
> -
> - spin_lock_bh(&fsp->scsi_pkt_lock);
> + if (try_to_del_timer_sync(&fsp->timer) < 0) {
> + spin_unlock_bh(&fsp->scsi_pkt_lock);
> + del_timer_sync(&fsp->timer);
> + spin_lock_bh(&fsp->scsi_pkt_lock);
> + }
I am not sure if we will get any benefit of this. This fsp is for
lun reset, i.e., not the i/o from queuecommand, it seems to me we would
always end up do the unlock/del_timer_sync/lock, which is even more
overhead.
yi
> if (fsp->seq_ptr) {
> lport->tt.exch_done(fsp->seq_ptr);
> fsp->seq_ptr = NULL;
> @@ -1890,7 +1890,7 @@ static void fc_io_compl(struct fc_fcp_pkt *fsp)
> fc_fcp_ddp_done(fsp);
>
> fsp->state |= FC_SRB_COMPL;
> - if (!(fsp->state & FC_SRB_FCP_PROCESSING_TMO)) {
> + if (try_to_del_timer_sync(&fsp->timer) < 0) {
> spin_unlock_bh(&fsp->scsi_pkt_lock);
> del_timer_sync(&fsp->timer);
> spin_lock_bh(&fsp->scsi_pkt_lock);
> _______________________________________________
> devel mailing list
> [email protected]
> https://lists.open-fcoe.org/mailman/listinfo/devel
_______________________________________________
devel mailing list
[email protected]
https://lists.open-fcoe.org/mailman/listinfo/devel
