Merge "mhi: core: improve bandwidth switch events processing"

This commit is contained in:
qctecmdr 2020-05-15 20:04:28 -07:00 committed by Gerrit - the friendly Code Review server
commit 3aa681b633

View File

@ -1440,29 +1440,12 @@ int mhi_process_bw_scale_ev_ring(struct mhi_controller *mhi_cntrl,
struct mhi_link_info link_info, *cur_info = &mhi_cntrl->mhi_link_info;
int result, ret = 0;
if (unlikely(MHI_EVENT_ACCESS_INVALID(mhi_cntrl->pm_state))) {
MHI_LOG("No EV access, PM_STATE:%s\n",
to_mhi_pm_state_str(mhi_cntrl->pm_state));
ret = -EIO;
goto exit_no_lock;
}
ret = __mhi_device_get_sync(mhi_cntrl);
if (ret)
goto exit_no_lock;
mutex_lock(&mhi_cntrl->pm_mutex);
spin_lock_bh(&mhi_event->lock);
dev_rp = mhi_to_virtual(ev_ring, er_ctxt->rp);
if (ev_ring->rp == dev_rp) {
spin_unlock_bh(&mhi_event->lock);
read_lock_bh(&mhi_cntrl->pm_lock);
mhi_cntrl->wake_put(mhi_cntrl, false);
read_unlock_bh(&mhi_cntrl->pm_lock);
MHI_VERB("no pending event found\n");
goto exit_bw_process;
goto exit_bw_scale_process;
}
/* if rp points to base, we need to wrap it around */
@ -1470,6 +1453,13 @@ int mhi_process_bw_scale_ev_ring(struct mhi_controller *mhi_cntrl,
dev_rp = ev_ring->base + ev_ring->len;
dev_rp--;
/* fast forward to currently processed element and recycle er */
ev_ring->rp = dev_rp;
ev_ring->wp = dev_rp - 1;
if (ev_ring->wp < ev_ring->base)
ev_ring->wp = ev_ring->base + ev_ring->len - ev_ring->el_size;
mhi_recycle_fwd_ev_ring_element(mhi_cntrl, ev_ring);
MHI_ASSERT(MHI_TRE_GET_EV_TYPE(dev_rp) != MHI_PKT_TYPE_BW_REQ_EVENT,
"!BW SCALE REQ event");
@ -1482,19 +1472,22 @@ int mhi_process_bw_scale_ev_ring(struct mhi_controller *mhi_cntrl,
link_info.target_link_speed,
link_info.target_link_width);
/* fast forward to currently processed element and recycle er */
ev_ring->rp = dev_rp;
ev_ring->wp = dev_rp - 1;
if (ev_ring->wp < ev_ring->base)
ev_ring->wp = ev_ring->base + ev_ring->len - ev_ring->el_size;
mhi_recycle_fwd_ev_ring_element(mhi_cntrl, ev_ring);
read_lock_bh(&mhi_cntrl->pm_lock);
if (likely(MHI_DB_ACCESS_VALID(mhi_cntrl)))
mhi_ring_er_db(mhi_event);
read_unlock_bh(&mhi_cntrl->pm_lock);
spin_unlock_bh(&mhi_event->lock);
atomic_inc(&mhi_cntrl->pending_pkts);
ret = mhi_device_get_sync(mhi_cntrl->mhi_dev,
MHI_VOTE_DEVICE | MHI_VOTE_BUS);
if (ret) {
atomic_dec(&mhi_cntrl->pending_pkts);
goto exit_bw_scale_process;
}
mutex_lock(&mhi_cntrl->pm_mutex);
ret = mhi_cntrl->bw_scale(mhi_cntrl, &link_info);
if (!ret)
*cur_info = link_info;
@ -1504,17 +1497,17 @@ int mhi_process_bw_scale_ev_ring(struct mhi_controller *mhi_cntrl,
read_lock_bh(&mhi_cntrl->pm_lock);
if (likely(MHI_DB_ACCESS_VALID(mhi_cntrl)))
mhi_cntrl->write_reg(mhi_cntrl, mhi_cntrl->bw_scale_db, 0,
MHI_BW_SCALE_RESULT(result,
link_info.sequence_num));
mhi_cntrl->wake_put(mhi_cntrl, false);
MHI_BW_SCALE_RESULT(result,
link_info.sequence_num));
read_unlock_bh(&mhi_cntrl->pm_lock);
exit_bw_process:
mhi_device_put(mhi_cntrl->mhi_dev, MHI_VOTE_DEVICE | MHI_VOTE_BUS);
atomic_dec(&mhi_cntrl->pending_pkts);
mutex_unlock(&mhi_cntrl->pm_mutex);
exit_no_lock:
MHI_VERB("exit er_index:%u\n", mhi_event->er_index);
exit_bw_scale_process:
MHI_VERB("exit er_index:%u ret:%d\n", mhi_event->er_index, ret);
return ret;
}