can: rx-offload: Add support for HW fifo based irq offloading
authorDavid Jander <david@protonic.nl>
Fri, 10 Oct 2014 15:30:10 +0000 (17:30 +0200)
committerMarc Kleine-Budde <mkl@pengutronix.de>
Mon, 6 Feb 2017 14:13:23 +0000 (15:13 +0100)
Some CAN controllers have a usable FIFO already but can still benefit
from off-loading the CAN controller FIFO. The CAN frames of the FIFO are
read and put into a skb queue during interrupt and then transmitted in a
NAPI context.

Signed-off-by: David Jander <david@protonic.nl>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
drivers/net/can/Makefile
drivers/net/can/rx-offload.c [new file with mode: 0644]
include/linux/can/rx-offload.h [new file with mode: 0644]

index 7a85495dbb0c436f1f8ba30285c2f304c08a21ac..0da4f2f5c7e3eaef08f9a3ee3e5a0ab8081a80a9 100644 (file)
@@ -6,7 +6,8 @@ obj-$(CONFIG_CAN_VCAN)          += vcan.o
 obj-$(CONFIG_CAN_SLCAN)                += slcan.o
 
 obj-$(CONFIG_CAN_DEV)          += can-dev.o
-can-dev-y                      := dev.o
+can-dev-y                      += dev.o
+can-dev-y                      += rx-offload.o
 
 can-dev-$(CONFIG_CAN_LEDS)     += led.o
 
diff --git a/drivers/net/can/rx-offload.c b/drivers/net/can/rx-offload.c
new file mode 100644 (file)
index 0000000..7267c61
--- /dev/null
@@ -0,0 +1,156 @@
+/*
+ * Copyright (c) 2014 David Jander, Protonic Holland
+ * Copyright (C) 2014-2017 Pengutronix, Marc Kleine-Budde <kernel@pengutronix.de>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the version 2 of the GNU General Public License
+ * as published by the Free Software Foundation
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/can/dev.h>
+#include <linux/can/rx-offload.h>
+
+static int can_rx_offload_napi_poll(struct napi_struct *napi, int quota)
+{
+       struct can_rx_offload *offload = container_of(napi, struct can_rx_offload, napi);
+       struct net_device *dev = offload->dev;
+       struct net_device_stats *stats = &dev->stats;
+       struct sk_buff *skb;
+       int work_done = 0;
+
+       while ((work_done < quota) &&
+              (skb = skb_dequeue(&offload->skb_queue))) {
+               struct can_frame *cf = (struct can_frame *)skb->data;
+
+               work_done++;
+               stats->rx_packets++;
+               stats->rx_bytes += cf->can_dlc;
+               netif_receive_skb(skb);
+       }
+
+       if (work_done < quota) {
+               napi_complete_done(napi, work_done);
+
+               /* Check if there was another interrupt */
+               if (!skb_queue_empty(&offload->skb_queue))
+                       napi_reschedule(&offload->napi);
+       }
+
+       can_led_event(offload->dev, CAN_LED_EVENT_RX);
+
+       return work_done;
+}
+
+static struct sk_buff *can_rx_offload_offload_one(struct can_rx_offload *offload, unsigned int n)
+{
+       struct sk_buff *skb = NULL;
+       struct can_frame *cf;
+       int ret;
+
+       /* If queue is full or skb not available, read to discard mailbox */
+       if (likely(skb_queue_len(&offload->skb_queue) <=
+                  offload->skb_queue_len_max))
+               skb = alloc_can_skb(offload->dev, &cf);
+
+       if (!skb) {
+               struct can_frame cf_overflow;
+
+               ret = offload->mailbox_read(offload, &cf_overflow, n);
+               if (ret)
+                       offload->dev->stats.rx_dropped++;
+
+               return NULL;
+       }
+
+       ret = offload->mailbox_read(offload, cf, n);
+       if (!ret) {
+               kfree_skb(skb);
+               return NULL;
+       }
+
+       return skb;
+}
+
+int can_rx_offload_irq_offload_fifo(struct can_rx_offload *offload)
+{
+       struct sk_buff *skb;
+       int received = 0;
+
+       while ((skb = can_rx_offload_offload_one(offload, 0))) {
+               skb_queue_tail(&offload->skb_queue, skb);
+               received++;
+       }
+
+       if (received)
+               can_rx_offload_schedule(offload);
+
+       return received;
+}
+EXPORT_SYMBOL_GPL(can_rx_offload_irq_offload_fifo);
+
+int can_rx_offload_irq_queue_err_skb(struct can_rx_offload *offload, struct sk_buff *skb)
+{
+       if (skb_queue_len(&offload->skb_queue) >
+           offload->skb_queue_len_max)
+               return -ENOMEM;
+
+       skb_queue_tail(&offload->skb_queue, skb);
+       can_rx_offload_schedule(offload);
+
+       return 0;
+}
+EXPORT_SYMBOL_GPL(can_rx_offload_irq_queue_err_skb);
+
+static int can_rx_offload_init_queue(struct net_device *dev, struct can_rx_offload *offload, unsigned int weight)
+{
+       offload->dev = dev;
+
+       /* Limit queue len to 4x the weight (rounted to next power of two) */
+       offload->skb_queue_len_max = 2 << fls(weight);
+       offload->skb_queue_len_max *= 4;
+       skb_queue_head_init(&offload->skb_queue);
+
+       can_rx_offload_reset(offload);
+       netif_napi_add(dev, &offload->napi, can_rx_offload_napi_poll, weight);
+
+       dev_dbg(dev->dev.parent, "%s: skb_queue_len_max=%d\n",
+               __func__, offload->skb_queue_len_max);
+
+       return 0;
+}
+
+int can_rx_offload_add_fifo(struct net_device *dev, struct can_rx_offload *offload, unsigned int weight)
+{
+       if (!offload->mailbox_read)
+               return -EINVAL;
+
+       return can_rx_offload_init_queue(dev, offload, weight);
+}
+EXPORT_SYMBOL_GPL(can_rx_offload_add_fifo);
+
+void can_rx_offload_enable(struct can_rx_offload *offload)
+{
+       can_rx_offload_reset(offload);
+       napi_enable(&offload->napi);
+}
+EXPORT_SYMBOL_GPL(can_rx_offload_enable);
+
+void can_rx_offload_del(struct can_rx_offload *offload)
+{
+       netif_napi_del(&offload->napi);
+       skb_queue_purge(&offload->skb_queue);
+}
+EXPORT_SYMBOL_GPL(can_rx_offload_del);
+
+void can_rx_offload_reset(struct can_rx_offload *offload)
+{
+}
+EXPORT_SYMBOL_GPL(can_rx_offload_reset);
diff --git a/include/linux/can/rx-offload.h b/include/linux/can/rx-offload.h
new file mode 100644 (file)
index 0000000..cadfe98
--- /dev/null
@@ -0,0 +1,51 @@
+/*
+ * linux/can/rx-offload.h
+ *
+ * Copyright (c) 2014 David Jander, Protonic Holland
+ * Copyright (c) 2014-2017 Pengutronix, Marc Kleine-Budde <kernel@pengutronix.de>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the version 2 of the GNU General Public License
+ * as published by the Free Software Foundation
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ */
+
+#ifndef _CAN_RX_OFFLOAD_H
+#define _CAN_RX_OFFLOAD_H
+
+#include <linux/netdevice.h>
+#include <linux/can.h>
+
+struct can_rx_offload {
+       struct net_device *dev;
+
+       unsigned int (*mailbox_read)(struct can_rx_offload *offload, struct can_frame *cf, unsigned int mb);
+
+       struct sk_buff_head skb_queue;
+       u32 skb_queue_len_max;
+
+       struct napi_struct napi;
+};
+
+int can_rx_offload_add_fifo(struct net_device *dev, struct can_rx_offload *offload, unsigned int weight);
+int can_rx_offload_irq_offload_fifo(struct can_rx_offload *offload);
+int can_rx_offload_irq_queue_err_skb(struct can_rx_offload *offload, struct sk_buff *skb);
+void can_rx_offload_reset(struct can_rx_offload *offload);
+void can_rx_offload_del(struct can_rx_offload *offload);
+void can_rx_offload_enable(struct can_rx_offload *offload);
+
+static inline void can_rx_offload_schedule(struct can_rx_offload *offload)
+{
+       napi_schedule(&offload->napi);
+}
+
+static inline void can_rx_offload_disable(struct can_rx_offload *offload)
+{
+       napi_disable(&offload->napi);
+}
+
+#endif /* !_CAN_RX_OFFLOAD_H */