[PATCH v3 13/17] dmaengine: bcm-sba-raid: Re-factor sba_process_deferred_requests()

From: Anup Patel
Date: Tue Aug 22 2017 - 05:59:00 EST


Currently, sba_process_deferred_requests() handles both pending
and completed sba_request which is unnecessary overhead for
sba_issue_pending() because completed sba_request handling is
not required in sba_issue_pending().

This patch breaks sba_process_deferred_requests() into two parts
sba_process_received_request() and _sba_process_pending_requests().

The sba_issue_pending() will only process pending sba_request
by calling _sba_process_pending_requests(). This will improve
sba_issue_pending().

The sba_receive_message() will only process received sba_request
by calling sba_process_received_request() for each received
sba_request. The sba_process_received_request() will also call
_sba_process_pending_requests() after handling received sba_request
because we might have pending sba_request not submitted by previous
call to sba_issue_pending().

Signed-off-by: Anup Patel <anup.patel@xxxxxxxxxxxx>
Reviewed-by: Scott Branden <scott.branden@xxxxxxxxxxxx>
---
drivers/dma/bcm-sba-raid.c | 109 +++++++++++++++++++--------------------------
1 file changed, 47 insertions(+), 62 deletions(-)

diff --git a/drivers/dma/bcm-sba-raid.c b/drivers/dma/bcm-sba-raid.c
index 2f444cc..cb6b2e2 100644
--- a/drivers/dma/bcm-sba-raid.c
+++ b/drivers/dma/bcm-sba-raid.c
@@ -419,22 +419,20 @@ static int sba_send_mbox_request(struct sba_device *sba,
return 0;
}

-static void sba_process_deferred_requests(struct sba_device *sba)
+/* Note: Must be called with sba->reqs_lock held */
+static void _sba_process_pending_requests(struct sba_device *sba)
{
int ret;
u32 count;
- unsigned long flags;
struct sba_request *req;
- struct dma_async_tx_descriptor *tx;
-
- spin_lock_irqsave(&sba->reqs_lock, flags);
-
- /* Count pending requests */
- count = 0;
- list_for_each_entry(req, &sba->reqs_pending_list, node)
- count++;

- /* Process pending requests */
+ /*
+ * Process few pending requests
+ *
+ * For now, we process (<number_of_mailbox_channels> * 8)
+ * number of requests at a time.
+ */
+ count = sba->mchans_count * 8;
while (!list_empty(&sba->reqs_pending_list) && count) {
/* Get the first pending request */
req = list_first_entry(&sba->reqs_pending_list,
@@ -445,11 +443,7 @@ static void sba_process_deferred_requests(struct sba_device *sba)
break;

/* Send request to mailbox channel */
- spin_unlock_irqrestore(&sba->reqs_lock, flags);
ret = sba_send_mbox_request(sba, req);
- spin_lock_irqsave(&sba->reqs_lock, flags);
-
- /* If something went wrong then keep request pending */
if (ret < 0) {
_sba_pending_request(sba, req);
break;
@@ -457,20 +451,18 @@ static void sba_process_deferred_requests(struct sba_device *sba)

count--;
}
+}

- /* Count completed requests */
- count = 0;
- list_for_each_entry(req, &sba->reqs_completed_list, node)
- count++;
-
- /* Process completed requests */
- while (!list_empty(&sba->reqs_completed_list) && count) {
- req = list_first_entry(&sba->reqs_completed_list,
- struct sba_request, node);
- list_del_init(&req->node);
- tx = &req->tx;
+static void sba_process_received_request(struct sba_device *sba,
+ struct sba_request *req)
+{
+ unsigned long flags;
+ struct dma_async_tx_descriptor *tx;
+ struct sba_request *nreq, *first = req->first;

- spin_unlock_irqrestore(&sba->reqs_lock, flags);
+ /* Process only after all chained requests are received */
+ if (!atomic_dec_return(&first->next_pending_count)) {
+ tx = &first->tx;

WARN_ON(tx->cookie < 0);
if (tx->cookie > 0) {
@@ -485,41 +477,34 @@ static void sba_process_deferred_requests(struct sba_device *sba)

spin_lock_irqsave(&sba->reqs_lock, flags);

- /* If waiting for 'ack' then move to completed list */
- if (!async_tx_test_ack(&req->tx))
- _sba_complete_request(sba, req);
- else
- _sba_free_request(sba, req);
+ /* Free all requests chained to first request */
+ list_for_each_entry(nreq, &first->next, next)
+ _sba_free_request(sba, nreq);
+ INIT_LIST_HEAD(&first->next);

- count--;
- }
+ /* Mark request as received */
+ _sba_received_request(sba, first);

- /* Re-check pending and completed work */
- count = 0;
- if (!list_empty(&sba->reqs_pending_list) ||
- !list_empty(&sba->reqs_completed_list))
- count = 1;
-
- spin_unlock_irqrestore(&sba->reqs_lock, flags);
-}
-
-static void sba_process_received_request(struct sba_device *sba,
- struct sba_request *req)
-{
- unsigned long flags;
-
- spin_lock_irqsave(&sba->reqs_lock, flags);
+ /* The client is allowed to attach dependent operations
+ * until 'ack' is set
+ */
+ if (!async_tx_test_ack(tx))
+ _sba_complete_request(sba, first);
+ else
+ _sba_free_request(sba, first);

- /* Mark request as received */
- _sba_received_request(sba, req);
+ /* Cleanup completed requests */
+ list_for_each_entry_safe(req, nreq,
+ &sba->reqs_completed_list, node) {
+ if (async_tx_test_ack(&req->tx))
+ _sba_free_request(sba, req);
+ }

- /* Update request */
- if (!atomic_dec_return(&req->first->next_pending_count))
- _sba_complete_request(sba, req->first);
- if (req->first != req)
- _sba_free_request(sba, req);
+ /* Process pending requests */
+ _sba_process_pending_requests(sba);

- spin_unlock_irqrestore(&sba->reqs_lock, flags);
+ spin_unlock_irqrestore(&sba->reqs_lock, flags);
+ }
}

/* ====== DMAENGINE callbacks ===== */
@@ -544,10 +529,13 @@ static int sba_device_terminate_all(struct dma_chan *dchan)

static void sba_issue_pending(struct dma_chan *dchan)
{
+ unsigned long flags;
struct sba_device *sba = to_sba_device(dchan);

- /* Process deferred requests */
- sba_process_deferred_requests(sba);
+ /* Process pending requests */
+ spin_lock_irqsave(&sba->reqs_lock, flags);
+ _sba_process_pending_requests(sba);
+ spin_unlock_irqrestore(&sba->reqs_lock, flags);
}

static dma_cookie_t sba_tx_submit(struct dma_async_tx_descriptor *tx)
@@ -1482,9 +1470,6 @@ static void sba_receive_message(struct mbox_client *cl, void *msg)

/* Process received request */
sba_process_received_request(sba, req);
-
- /* Process deferred requests */
- sba_process_deferred_requests(sba);
}

/* ====== Platform driver routines ===== */
--
2.7.4