crypto: caam - use NAPI instead of tasklet

This patch updates the current tasklet implementation to NAPI so as
the system is more balanced in the terms that the packet submission
and the packet forwarding after being processed can be done at
the same priority

Signed-off-by: Naveen Burmi <naveenburmi@freescale.com>

rebased and tuned NAPI_WEIGHT.

Signed-off-by: Kim Phillips <kim.phillips@freescale.com>
(cherry picked from commit c74e14d7ff270f8d85c7988e9286f64b721f34ee)
Change-Id: I3a31db49a1a6060b3ad5cd0fc4ee4044858438bc
Reviewed-on: http://git.am.freescale.net:8181/520
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
Tested-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>

Signed-off-by: Nitesh Lal <NiteshNarayanLal@freescale.com>
Change-Id: I685d687d89a53387287912cd2273f8c1d6a6e4e4
Reviewed-on: http://git.am.freescale.net:8181/9753
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Vakul Garg <vakul@freescale.com>
Reviewed-by: Jose Rivera <German.Rivera@freescale.com>

3 files changed, 46 insertions, 11 deletions

diff --git a/drivers/crypto/caam/Kconfig b/drivers/crypto/caam/Kconfig
index 01a4bd3..97927a5 100644
--- a/drivers/crypto/caam/Kconfig
+++ b/drivers/crypto/caam/Kconfig
@@ -75,7 +75,7 @@ config CRYPTO_DEV_FSL_CAAM_INTC_TIME_THLD
 
 config CRYPTO_DEV_FSL_CAAM_CRYPTO_API
 	tristate "Register algorithm implementations with the Crypto API"
-	depends on CRYPTO_DEV_FSL_CAAM && CRYPTO_DEV_FSL_CAAM_JR
+	depends on CRYPTO_DEV_FSL_CAAM && CRYPTO_DEV_FSL_CAAM_JR && NET
 	default y
 	select CRYPTO_ALGAPI
 	select CRYPTO_AUTHENC
diff --git a/drivers/crypto/caam/intern.h b/drivers/crypto/caam/intern.h
index 6d85fcc..4ca6722 100644
--- a/drivers/crypto/caam/intern.h
+++ b/drivers/crypto/caam/intern.h
@@ -23,6 +23,8 @@
 #define JOBR_INTC_COUNT_THLD 0
 #endif
 
+#define CAAM_NAPI_WEIGHT	63
+
 /*
  * Storage for tracking each in-process entry moving across a ring
  * Each entry on an output ring needs one of these
@@ -41,7 +43,8 @@ struct caam_drv_private_jr {
 	struct device		*dev;
 	int ridx;
 	struct caam_job_ring __iomem *rregs;	/* JobR's register space */
-	struct tasklet_struct irqtask;
+	struct napi_struct __percpu *irqtask;
+	struct net_device __percpu *net_dev;
 	int irq;			/* One per queue */
 
 	/* Number of scatterlist crypt transforms active on the JobR */
diff --git a/drivers/crypto/caam/jr.c b/drivers/crypto/caam/jr.c
index 1d80bd3..0875f85 100644
--- a/drivers/crypto/caam/jr.c
+++ b/drivers/crypto/caam/jr.c
@@ -69,11 +69,17 @@ int caam_jr_shutdown(struct device *dev)
 {
 	struct caam_drv_private_jr *jrp = dev_get_drvdata(dev);
 	dma_addr_t inpbusaddr, outbusaddr;
-	int ret;
+	int i, ret;
 
 	ret = caam_reset_hw_jr(dev);
 
-	tasklet_kill(&jrp->irqtask);
+	for_each_possible_cpu(i) {
+		napi_disable(per_cpu_ptr(jrp->irqtask, i));
+		netif_napi_del(per_cpu_ptr(jrp->irqtask, i));
+	}
+
+	free_percpu(jrp->irqtask);
+	free_percpu(jrp->net_dev);
 
 	/* Release interrupt */
 	free_irq(jrp->irq, dev);
@@ -153,23 +159,24 @@ static irqreturn_t caam_jr_interrupt(int irq, void *st_dev)
 	wr_reg32(&jrp->rregs->jrintstatus, irqstate);
 
 	preempt_disable();
-	tasklet_schedule(&jrp->irqtask);
+	napi_schedule(per_cpu_ptr(jrp->irqtask, smp_processor_id()));
 	preempt_enable();
 
 	return IRQ_HANDLED;
 }
 
 /* Deferred service handler, run as interrupt-fired tasklet */
-static void caam_jr_dequeue(unsigned long devarg)
+static int caam_jr_dequeue(struct napi_struct *napi, int budget)
 {
 	int hw_idx, sw_idx, i, head, tail;
-	struct device *dev = (struct device *)devarg;
+	struct device *dev = &napi->dev->dev;
 	struct caam_drv_private_jr *jrp = dev_get_drvdata(dev);
 	void (*usercall)(struct device *dev, u32 *desc, u32 status, void *arg);
 	u32 *userdesc, userstatus;
 	void *userarg;
+	int cleaned = 0;
 
-	while (rd_reg32(&jrp->rregs->outring_used)) {
+	while (rd_reg32(&jrp->rregs->outring_used) && cleaned < budget) {
 
 		head = ACCESS_ONCE(jrp->head);
 
@@ -229,10 +236,15 @@ static void caam_jr_dequeue(unsigned long devarg)
 
 		/* Finally, execute user's callback */
 		usercall(dev, userdesc, userstatus, userarg);
+		cleaned++;
+	}
+	if (cleaned < budget) {
+		napi_complete(per_cpu_ptr(jrp->irqtask, smp_processor_id()));
+		/* reenable / unmask IRQs */
+		clrbits32(&jrp->rregs->rconfig_lo, JRCFG_IMSK);
 	}
 
-	/* reenable / unmask IRQs */
-	clrbits32(&jrp->rregs->rconfig_lo, JRCFG_IMSK);
+	return cleaned;
 }
 
 /**
@@ -378,8 +390,28 @@ static int caam_jr_init(struct device *dev)
 	int i, error;
 
 	jrp = dev_get_drvdata(dev);
+	/* Connect job ring interrupt handler. */
+	jrp->irqtask = alloc_percpu(struct napi_struct);
+	if (!jrp->irqtask) {
+		dev_err(dev, "can't allocate percpu irqtask memory\n");
+		return -ENOMEM;
+	}
+
+	jrp->net_dev = alloc_percpu(struct net_device);
+	if (!jrp->net_dev) {
+		dev_err(dev, "can't allocate percpu net_dev memory\n");
+		free_percpu(jrp->irqtask);
+		return -ENOMEM;
+	}
 
-	tasklet_init(&jrp->irqtask, caam_jr_dequeue, (unsigned long)dev);
+	for_each_possible_cpu(i) {
+		(per_cpu_ptr(jrp->net_dev, i))->dev = *dev;
+		INIT_LIST_HEAD(&per_cpu_ptr(jrp->net_dev, i)->napi_list);
+		netif_napi_add(per_cpu_ptr(jrp->net_dev, i),
+			       per_cpu_ptr(jrp->irqtask, i),
+			       caam_jr_dequeue, CAAM_NAPI_WEIGHT);
+		napi_enable(per_cpu_ptr(jrp->irqtask, i));
+	}
 
 	/* Connect job ring interrupt handler. */
 	error = request_irq(jrp->irq, caam_jr_interrupt, IRQF_SHARED,