IB/hfi1: Rework the IRQ API to be more flexible

The current IRQ API is an all or nothing interface.  This has two
problems:

  1. All IRQs are enabled regardless of use
  2. Moving from general interrupt to MSIx handling is difficult

Introduce a new API to enable/disable specific IRQs or a range of IRQs.

Do not enable and disable all IRQs in one step.

Rework various modules to enable/disable IRQs when needed.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Reviewed-by: Sadanand Warrier <sadanand.warrier@intel.com>
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

drivers/infiniband/hw/hfi1/chip.c

@@ -1098,9 +1098,9 @@ struct err_reg_info {
 	const char *desc;
 };
 
-#define NUM_MISC_ERRS (IS_GENERAL_ERR_END - IS_GENERAL_ERR_START)
-#define NUM_DC_ERRS (IS_DC_END - IS_DC_START)
-#define NUM_VARIOUS (IS_VARIOUS_END - IS_VARIOUS_START)
+#define NUM_MISC_ERRS (IS_GENERAL_ERR_END + 1 - IS_GENERAL_ERR_START)
+#define NUM_DC_ERRS (IS_DC_END + 1 - IS_DC_START)
+#define NUM_VARIOUS (IS_VARIOUS_END + 1 - IS_VARIOUS_START)
 
 /*
  * Helpers for building HFI and DC error interrupt table entries.  Different
@@ -8176,7 +8176,7 @@ static void is_rcv_avail_int(struct hfi1_devdata *dd, unsigned int source)
 /**
  * is_rcv_urgent_int() - User receive context urgent IRQ handler
  * @dd: valid dd
- * @source: logical IRQ source (ofse from IS_RCVURGENT_START)
+ * @source: logical IRQ source (offset from IS_RCVURGENT_START)
  *
  * RX block receive urgent interrupt.  Source is < 160.
  *
@@ -8226,7 +8226,7 @@ static const struct is_table is_table[] = {
 				is_sdma_eng_err_name,	is_sdma_eng_err_int },
 { IS_SENDCTXT_ERR_START, IS_SENDCTXT_ERR_END,
 				is_sendctxt_err_name,	is_sendctxt_err_int },
-{ IS_SDMA_START,	     IS_SDMA_END,
+{ IS_SDMA_START,	     IS_SDMA_IDLE_END,
 				is_sdma_eng_name,	is_sdma_eng_int },
 { IS_VARIOUS_START,	     IS_VARIOUS_END,
 				is_various_name,	is_various_int },
@@ -8252,7 +8252,7 @@ static void is_interrupt(struct hfi1_devdata *dd, unsigned int source)
 
 	/* avoids a double compare by walking the table in-order */
 	for (entry = &is_table[0]; entry->is_name; entry++) {
-		if (source < entry->end) {
+		if (source <= entry->end) {
 			trace_hfi1_interrupt(dd, entry, source);
 			entry->is_int(dd, source - entry->start);
 			return;
@@ -9646,30 +9646,10 @@ void qsfp_event(struct work_struct *work)
 	}
 }
 
-static void init_qsfp_int(struct hfi1_devdata *dd)
+void init_qsfp_int(struct hfi1_devdata *dd)
 {
 	struct hfi1_pportdata *ppd = dd->pport;
-	u64 qsfp_mask, cce_int_mask;
-	const int qsfp1_int_smask = QSFP1_INT % 64;
-	const int qsfp2_int_smask = QSFP2_INT % 64;
-
-	/*
-	 * disable QSFP1 interrupts for HFI1, QSFP2 interrupts for HFI0
-	 * Qsfp1Int and Qsfp2Int are adjacent bits in the same CSR,
-	 * therefore just one of QSFP1_INT/QSFP2_INT can be used to find
-	 * the index of the appropriate CSR in the CCEIntMask CSR array
-	 */
-	cce_int_mask = read_csr(dd, CCE_INT_MASK +
-				(8 * (QSFP1_INT / 64)));
-	if (dd->hfi1_id) {
-		cce_int_mask &= ~((u64)1 << qsfp1_int_smask);
-		write_csr(dd, CCE_INT_MASK + (8 * (QSFP1_INT / 64)),
-			  cce_int_mask);
-	} else {
-		cce_int_mask &= ~((u64)1 << qsfp2_int_smask);
-		write_csr(dd, CCE_INT_MASK + (8 * (QSFP2_INT / 64)),
-			  cce_int_mask);
-	}
+	u64 qsfp_mask;
 
 	qsfp_mask = (u64)(QSFP_HFI0_INT_N | QSFP_HFI0_MODPRST_N);
 	/* Clear current status to avoid spurious interrupts */
@@ -9686,6 +9666,12 @@ static void init_qsfp_int(struct hfi1_devdata *dd)
 	write_csr(dd,
 		  dd->hfi1_id ? ASIC_QSFP2_INVERT : ASIC_QSFP1_INVERT,
 		  qsfp_mask);
+
+	/* Enable the appropriate QSFP IRQ source */
+	if (!dd->hfi1_id)
+		set_intr_bits(dd, QSFP1_INT, QSFP1_INT, true);
+	else
+		set_intr_bits(dd, QSFP2_INT, QSFP2_INT, true);
 }
 
 /*
@@ -11926,10 +11912,16 @@ void hfi1_rcvctrl(struct hfi1_devdata *dd, unsigned int op,
 
 		rcvctrl &= ~RCV_CTXT_CTRL_ENABLE_SMASK;
 	}
-	if (op & HFI1_RCVCTRL_INTRAVAIL_ENB)
+	if (op & HFI1_RCVCTRL_INTRAVAIL_ENB) {
+		set_intr_bits(dd, IS_RCVAVAIL_START + rcd->ctxt,
+			      IS_RCVAVAIL_START + rcd->ctxt, true);
 		rcvctrl |= RCV_CTXT_CTRL_INTR_AVAIL_SMASK;
-	if (op & HFI1_RCVCTRL_INTRAVAIL_DIS)
+	}
+	if (op & HFI1_RCVCTRL_INTRAVAIL_DIS) {
+		set_intr_bits(dd, IS_RCVAVAIL_START + rcd->ctxt,
+			      IS_RCVAVAIL_START + rcd->ctxt, false);
 		rcvctrl &= ~RCV_CTXT_CTRL_INTR_AVAIL_SMASK;
+	}
 	if ((op & HFI1_RCVCTRL_TAILUPD_ENB) && rcd->rcvhdrtail_kvaddr)
 		rcvctrl |= RCV_CTXT_CTRL_TAIL_UPD_SMASK;
 	if (op & HFI1_RCVCTRL_TAILUPD_DIS) {
@@ -12957,57 +12949,65 @@ int hfi1_tempsense_rd(struct hfi1_devdata *dd, struct hfi1_temp *temp)
 	return ret;
 }
 
+/* ========================================================================= */
+
 /**
- * get_int_mask - get 64 bit int mask
- * @dd - the devdata
- * @i - the csr (relative to CCE_INT_MASK)
+ * read_mod_write() - Calculate the IRQ register index and set/clear the bits
+ * @dd: valid devdata
+ * @src: IRQ source to determine register index from
+ * @bits: the bits to set or clear
+ * @set: true == set the bits, false == clear the bits
  *
- * Returns the mask with the urgent interrupt mask
- * bit clear for kernel receive contexts.
  */
-static u64 get_int_mask(struct hfi1_devdata *dd, u32 i)
+static void read_mod_write(struct hfi1_devdata *dd, u16 src, u64 bits,
+			   bool set)
 {
-	u64 mask = U64_MAX; /* default to no change */
-
-	if (i >= (IS_RCVURGENT_START / 64) && i < (IS_RCVURGENT_END / 64)) {
-		int j = (i - (IS_RCVURGENT_START / 64)) * 64;
-		int k = !j ? IS_RCVURGENT_START % 64 : 0;
+	u64 reg;
+	u16 idx = src / BITS_PER_REGISTER;
 
-		if (j)
-			j -= IS_RCVURGENT_START % 64;
-		/* j = 0..dd->first_dyn_alloc_ctxt - 1,k = 0..63 */
-		for (; j < dd->first_dyn_alloc_ctxt && k < 64; j++, k++)
-			/* convert to bit in mask and clear */
-			mask &= ~BIT_ULL(k);
-	}
-	return mask;
+	spin_lock(&dd->irq_src_lock);
+	reg = read_csr(dd, CCE_INT_MASK + (8 * idx));
+	if (set)
+		reg |= bits;
+	else
+		reg &= ~bits;
+	write_csr(dd, CCE_INT_MASK + (8 * idx), reg);
+	spin_unlock(&dd->irq_src_lock);
 }
 
-/* ========================================================================= */
-
-/*
- * Enable/disable chip from delivering interrupts.
+/**
+ * set_intr_bits() - Enable/disable a range (one or more) IRQ sources
+ * @dd: valid devdata
+ * @first: first IRQ source to set/clear
+ * @last: last IRQ source (inclusive) to set/clear
+ * @set: true == set the bits, false == clear the bits
+ *
+ * If first == last, set the exact source.
  */
-void set_intr_state(struct hfi1_devdata *dd, u32 enable)
+int set_intr_bits(struct hfi1_devdata *dd, u16 first, u16 last, bool set)
 {
-	int i;
+	u64 bits = 0;
+	u64 bit;
+	u16 src;
 
-	/*
-	 * In HFI, the mask needs to be 1 to allow interrupts.
-	 */
-	if (enable) {
-		/* enable all interrupts but urgent on kernel contexts */
-		for (i = 0; i < CCE_NUM_INT_CSRS; i++) {
-			u64 mask = get_int_mask(dd, i);
+	if (first > NUM_INTERRUPT_SOURCES || last > NUM_INTERRUPT_SOURCES)
+		return -EINVAL;
 
-			write_csr(dd, CCE_INT_MASK + (8 * i), mask);
-		}
+	if (last < first)
+		return -ERANGE;
 
-		init_qsfp_int(dd);
-	} else {
-		for (i = 0; i < CCE_NUM_INT_CSRS; i++)
-			write_csr(dd, CCE_INT_MASK + (8 * i), 0ull);
+	for (src = first; src <= last; src++) {
+		bit = src % BITS_PER_REGISTER;
+		/* wrapped to next register? */
+		if (!bit && bits) {
+			read_mod_write(dd, src - 1, bits, set);
+			bits = 0;
 		}
+		bits |= BIT_ULL(bit);
+	}
+	read_mod_write(dd, src, bits, set);
+
+	return 0;
 }
 
 /*
@@ -13074,12 +13074,9 @@ void remap_sdma_interrupts(struct hfi1_devdata *dd, int engine, int msix_intr)
 	 *	SDMAProgress
 	 *	SDMAIdle
 	 */
-	remap_intr(dd, IS_SDMA_START + 0 * TXE_NUM_SDMA_ENGINES + engine,
-		   msix_intr);
-	remap_intr(dd, IS_SDMA_START + 1 * TXE_NUM_SDMA_ENGINES + engine,
-		   msix_intr);
-	remap_intr(dd, IS_SDMA_START + 2 * TXE_NUM_SDMA_ENGINES + engine,
-		   msix_intr);
+	remap_intr(dd, IS_SDMA_START + engine, msix_intr);
+	remap_intr(dd, IS_SDMA_PROGRESS_START + engine, msix_intr);
+	remap_intr(dd, IS_SDMA_IDLE_START + engine, msix_intr);
 }
 
 /*
@@ -13109,7 +13106,8 @@ static int set_up_interrupts(struct hfi1_devdata *dd)
 	int ret;
 
 	/* mask all interrupts */
-	set_intr_state(dd, 0);
+	set_intr_bits(dd, IS_FIRST_SOURCE, IS_LAST_SOURCE, false);
+
 	/* clear all pending interrupts */
 	clear_all_interrupts(dd);
 

drivers/infiniband/hw/hfi1/chip.h

@@ -52,9 +52,7 @@
  */
 
 /* sizes */
-#define CCE_NUM_MSIX_VECTORS 256
-#define CCE_NUM_INT_CSRS 12
-#define CCE_NUM_INT_MAP_CSRS 96
+#define BITS_PER_REGISTER (BITS_PER_BYTE * sizeof(u64))
 #define NUM_INTERRUPT_SOURCES 768
 #define RXE_NUM_CONTEXTS 160
 #define RXE_PER_CONTEXT_SIZE 0x1000	/* 4k */
@@ -161,34 +159,49 @@
 	(CR_CREDIT_RETURN_DUE_TO_FORCE_MASK << \
 	CR_CREDIT_RETURN_DUE_TO_FORCE_SHIFT)
 
-/* interrupt source numbers */
+/* Specific IRQ sources */
+#define CCE_ERR_INT		  0
+#define RXE_ERR_INT		  1
+#define MISC_ERR_INT		  2
+#define PIO_ERR_INT		  4
+#define SDMA_ERR_INT		  5
+#define EGRESS_ERR_INT		  6
+#define TXE_ERR_INT		  7
+#define PBC_INT			240
+#define GPIO_ASSERT_INT		241
+#define QSFP1_INT		242
+#define QSFP2_INT		243
+#define TCRIT_INT		244
+
+/* interrupt source ranges */
+#define IS_FIRST_SOURCE		CCE_ERR_INT
 #define IS_GENERAL_ERR_START		  0
 #define IS_SDMAENG_ERR_START		 16
 #define IS_SENDCTXT_ERR_START		 32
-#define IS_SDMA_START		192 /* includes SDmaProgress,SDmaIdle */
+#define IS_SDMA_START			192
+#define IS_SDMA_PROGRESS_START		208
+#define IS_SDMA_IDLE_START		224
 #define IS_VARIOUS_START		240
 #define IS_DC_START			248
 #define IS_RCVAVAIL_START		256
 #define IS_RCVURGENT_START		416
 #define IS_SENDCREDIT_START		576
 #define IS_RESERVED_START		736
-#define IS_MAX_SOURCES		768
+#define IS_LAST_SOURCE			767
 
 /* derived interrupt source values */
-#define IS_GENERAL_ERR_END		IS_SDMAENG_ERR_START
-#define IS_SDMAENG_ERR_END		IS_SENDCTXT_ERR_START
-#define IS_SENDCTXT_ERR_END		IS_SDMA_START
-#define IS_SDMA_END			IS_VARIOUS_START
-#define IS_VARIOUS_END		IS_DC_START
-#define IS_DC_END			IS_RCVAVAIL_START
-#define IS_RCVAVAIL_END		IS_RCVURGENT_START
-#define IS_RCVURGENT_END		IS_SENDCREDIT_START
-#define IS_SENDCREDIT_END		IS_RESERVED_START
-#define IS_RESERVED_END		IS_MAX_SOURCES
-
-/* absolute interrupt numbers for QSFP1Int and QSFP2Int */
-#define QSFP1_INT		242
-#define QSFP2_INT		243
+#define IS_GENERAL_ERR_END		7
+#define IS_SDMAENG_ERR_END		31
+#define IS_SENDCTXT_ERR_END		191
+#define IS_SDMA_END                     207
+#define IS_SDMA_PROGRESS_END            223
+#define IS_SDMA_IDLE_END		239
+#define IS_VARIOUS_END			244
+#define IS_DC_END			255
+#define IS_RCVAVAIL_END			415
+#define IS_RCVURGENT_END		575
+#define IS_SENDCREDIT_END		735
+#define IS_RESERVED_END			IS_LAST_SOURCE
 
 /* DCC_CFG_PORT_CONFIG logical link states */
 #define LSTATE_DOWN    0x1
@@ -1421,6 +1434,8 @@ irqreturn_t sdma_interrupt(int irq, void *data);
 irqreturn_t receive_context_interrupt(int irq, void *data);
 irqreturn_t receive_context_thread(int irq, void *data);
 
+int set_intr_bits(struct hfi1_devdata *dd, u16 first, u16 last, bool set);
+void init_qsfp_int(struct hfi1_devdata *dd);
 void clear_all_interrupts(struct hfi1_devdata *dd);
 void remap_intr(struct hfi1_devdata *dd, int isrc, int msix_intr);
 void remap_sdma_interrupts(struct hfi1_devdata *dd, int engine, int msix_intr);

drivers/infiniband/hw/hfi1/file_ops.c

@@ -639,6 +639,9 @@ static int hfi1_file_close(struct inode *inode, struct file *fp)
 
 	hfi1_cdbg(PROC, "closing ctxt %u:%u", uctxt->ctxt, fdata->subctxt);
 
+	set_intr_bits(dd, IS_RCVURGENT_START + uctxt->ctxt,
+		      IS_RCVURGENT_START + uctxt->ctxt, false);
+
 	flush_wc();
 	/* drain user sdma queue */
 	hfi1_user_sdma_free_queues(fdata, uctxt);
@@ -1217,6 +1220,10 @@ static int setup_base_ctxt(struct hfi1_filedata *fd,
 	fd->uctxt = uctxt;
 	hfi1_rcd_get(uctxt);
 
+	/* Enable the Urgent IRQ for this user context */
+	set_intr_bits(dd, IS_RCVURGENT_START + uctxt->ctxt,
+		      IS_RCVURGENT_START + uctxt->ctxt, true);
+
 done:
 	if (uctxt->subctxt_cnt) {
 		/*

drivers/infiniband/hw/hfi1/hfi.h

@@ -1213,9 +1213,6 @@ struct hfi1_devdata {
 
 	struct diag_client *diag_client;
 
-	/* MSI-X information */
-	struct hfi1_msix_info msix_info;
-
 	/* general interrupt: mask of handled interrupts */
 	u64 gi_mask[CCE_NUM_INT_CSRS];
 
@@ -1229,6 +1226,9 @@ struct hfi1_devdata {
 	 */
 	struct timer_list synth_stats_timer;
 
+	/* MSI-X information */
+	struct hfi1_msix_info msix_info;
+
 	/*
 	 * device counters
 	 */
@@ -1355,6 +1355,8 @@ struct hfi1_devdata {
 
 	/* vnic data */
 	struct hfi1_vnic_data vnic;
+	/* Lock to protect IRQ SRC register access */
+	spinlock_t irq_src_lock;
 };
 
 static inline bool hfi1_vnic_is_rsm_full(struct hfi1_devdata *dd, int spare)

drivers/infiniband/hw/hfi1/init.c

@@ -831,6 +831,23 @@ static int create_workqueues(struct hfi1_devdata *dd)
 	return -ENOMEM;
 }
 
+/**
+ * enable_general_intr() - Enable the IRQs that will be handled by the
+ * general interrupt handler.
+ * @dd: valid devdata
+ *
+ */
+static void enable_general_intr(struct hfi1_devdata *dd)
+{
+	set_intr_bits(dd, CCE_ERR_INT, MISC_ERR_INT, true);
+	set_intr_bits(dd, PIO_ERR_INT, TXE_ERR_INT, true);
+	set_intr_bits(dd, IS_SENDCTXT_ERR_START, IS_SENDCTXT_ERR_END, true);
+	set_intr_bits(dd, PBC_INT, GPIO_ASSERT_INT, true);
+	set_intr_bits(dd, TCRIT_INT, TCRIT_INT, true);
+	set_intr_bits(dd, IS_DC_START, IS_DC_END, true);
+	set_intr_bits(dd, IS_SENDCREDIT_START, IS_SENDCREDIT_END, true);
+}
+
 /**
  * hfi1_init - do the actual initialization sequence on the chip
  * @dd: the hfi1_ib device
@@ -915,6 +932,7 @@ int hfi1_init(struct hfi1_devdata *dd, int reinit)
 				   "failed to allocate kernel ctxt's rcvhdrq and/or egr bufs\n");
 			ret = lastfail;
 		}
+		/* enable IRQ */
 		hfi1_rcd_put(rcd);
 	}
 
@@ -953,7 +971,8 @@ int hfi1_init(struct hfi1_devdata *dd, int reinit)
 			HFI1_STATUS_INITTED;
 	if (!ret) {
 		/* enable all interrupts from the chip */
-		set_intr_state(dd, 1);
+		enable_general_intr(dd);
+		init_qsfp_int(dd);
 
 		/* chip is OK for user apps; mark it as initialized */
 		for (pidx = 0; pidx < dd->num_pports; ++pidx) {
@@ -1050,8 +1069,8 @@ static void shutdown_device(struct hfi1_devdata *dd)
 	}
 	dd->flags &= ~HFI1_INITTED;
 
-	/* mask and clean up interrupts, but not errors */
-	set_intr_state(dd, 0);
+	/* mask and clean up interrupts */
+	set_intr_bits(dd, IS_FIRST_SOURCE, IS_LAST_SOURCE, false);
 	msix_clean_up_interrupts(dd);
 
 	for (pidx = 0; pidx < dd->num_pports; ++pidx) {
@@ -1312,6 +1331,7 @@ static struct hfi1_devdata *hfi1_alloc_devdata(struct pci_dev *pdev,
 	spin_lock_init(&dd->pio_map_lock);
 	mutex_init(&dd->dc8051_lock);
 	init_waitqueue_head(&dd->event_queue);
+	spin_lock_init(&dd->irq_src_lock);
 
 	dd->int_counter = alloc_percpu(u64);
 	if (!dd->int_counter) {

drivers/infiniband/hw/hfi1/msix.c

@@ -240,6 +240,21 @@ int msix_request_sdma_irq(struct sdma_engine *sde)
 	return 0;
 }
 
+/**
+ * enable_sdma_src() - Helper to enable SDMA IRQ srcs
+ * @dd: valid devdata structure
+ * @i: index of SDMA engine
+ */
+static void enable_sdma_srcs(struct hfi1_devdata *dd, int i)
+{
+	set_intr_bits(dd, IS_SDMA_START + i, IS_SDMA_START + i, true);
+	set_intr_bits(dd, IS_SDMA_PROGRESS_START + i,
+		      IS_SDMA_PROGRESS_START + i, true);
+	set_intr_bits(dd, IS_SDMA_IDLE_START + i, IS_SDMA_IDLE_START + i, true);
+	set_intr_bits(dd, IS_SDMAENG_ERR_START + i, IS_SDMAENG_ERR_START + i,
+		      true);
+}
+
 /**
  * msix_request_irqs() - Allocate all MSIx IRQs
  * @dd: valid devdata structure
@@ -262,6 +277,7 @@ int msix_request_irqs(struct hfi1_devdata *dd)
 		ret = msix_request_sdma_irq(sde);
 		if (ret)
 			return ret;
+		enable_sdma_srcs(sde->dd, i);
 	}
 
 	for (i = 0; i < dd->n_krcv_queues; i++) {