GenWQE: Support blocking when DDCB queue is busy

When the GenWQE hardware queue was busy, the driver returned simply
-EBUSY. This caused polling by applications which increased the load
on the already busy system. This change implements the possiblity to
sleep on a waitqueue instead when the DDCB queue is busy. The
requestor is woken up when there is free space on the queue again.
The old way to get -EBUSY is still available if the device is openend
with O_NONBLOCKING. The default is now blocking behavior.
Signed-off-by: Frank Haverkamp <haver@linux.vnet.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drivers/misc/genwqe/card_base.h

@@ -201,7 +201,8 @@ static inline void genwqe_mapping_init(struct dma_mapping *m,
  * @ddcb_seq:          Sequence number of last DDCB
  * @ddcbs_in_flight:   Currently enqueued DDCBs
  * @ddcbs_completed:   Number of already completed DDCBs
- * @busy:              Number of -EBUSY returns
+ * @return_on_busy:    Number of -EBUSY returns on full queue
+ * @wait_on_busy:      Number of waits on full queue
  * @ddcb_daddr:        DMA address of first DDCB in the queue
  * @ddcb_vaddr:        Kernel virtual address of first DDCB in the queue
  * @ddcb_req:          Associated requests (one per DDCB)
@@ -218,7 +219,8 @@ struct ddcb_queue {
 	unsigned int ddcbs_in_flight;	/* number of ddcbs in processing */
 	unsigned int ddcbs_completed;
 	unsigned int ddcbs_max_in_flight;
-	unsigned int busy;		/* how many times -EBUSY? */
+	unsigned int return_on_busy;    /* how many times -EBUSY? */
+	unsigned int wait_on_busy;
 
 	dma_addr_t ddcb_daddr;		/* DMA address */
 	struct ddcb *ddcb_vaddr;	/* kernel virtual addr for DDCBs */
@@ -226,7 +228,7 @@ struct ddcb_queue {
 	wait_queue_head_t *ddcb_waitqs; /* waitqueue per ddcb */
 
 	spinlock_t ddcb_lock;		/* exclusive access to queue */
-	wait_queue_head_t ddcb_waitq;	/* wait for ddcb processing */
+	wait_queue_head_t busy_waitq;   /* wait for ddcb processing */
 
 	/* registers or the respective queue to be used */
 	u32 IO_QUEUE_CONFIG;
@@ -508,7 +510,7 @@ static inline bool dma_mapping_used(struct dma_mapping *m)
  * buildup and teardown.
  */
 int  __genwqe_execute_ddcb(struct genwqe_dev *cd,
-			   struct genwqe_ddcb_cmd *cmd);
+			   struct genwqe_ddcb_cmd *cmd, unsigned int f_flags);
 
 /**
  * __genwqe_execute_raw_ddcb() - Execute DDCB request without addr translation
@@ -520,9 +522,12 @@ int  __genwqe_execute_ddcb(struct genwqe_dev *cd,
  * modification.
  */
 int  __genwqe_execute_raw_ddcb(struct genwqe_dev *cd,
-			       struct genwqe_ddcb_cmd *cmd);
+			       struct genwqe_ddcb_cmd *cmd,
+			       unsigned int f_flags);
+int  __genwqe_enqueue_ddcb(struct genwqe_dev *cd,
+			   struct ddcb_requ *req,
+			   unsigned int f_flags);
 
-int  __genwqe_enqueue_ddcb(struct genwqe_dev *cd, struct ddcb_requ *req);
 int  __genwqe_wait_ddcb(struct genwqe_dev *cd, struct ddcb_requ *req);
 int  __genwqe_purge_ddcb(struct genwqe_dev *cd, struct ddcb_requ *req);
 

drivers/misc/genwqe/card_ddcb.c

@@ -448,8 +448,10 @@ static int genwqe_check_ddcb_queue(struct genwqe_dev *cd,
 		queue->ddcbs_completed++;
 		queue->ddcbs_in_flight--;
 
-		/* wake up process waiting for this DDCB */
+		/* wake up process waiting for this DDCB, and
+                   processes on the busy queue */
 		wake_up_interruptible(&queue->ddcb_waitqs[queue->ddcb_act]);
+		wake_up_interruptible(&queue->busy_waitq);
 
 pick_next_one:
 		queue->ddcb_act = (queue->ddcb_act + 1) % queue->ddcb_max;
@@ -747,12 +749,14 @@ int genwqe_init_debug_data(struct genwqe_dev *cd, struct genwqe_debug_data *d)
  * __genwqe_enqueue_ddcb() - Enqueue a DDCB
  * @cd:         pointer to genwqe device descriptor
  * @req:        pointer to DDCB execution request
+ * @f_flags:    file mode: blocking, non-blocking
  *
  * Return: 0 if enqueuing succeeded
  *         -EIO if card is unusable/PCIe problems
  *         -EBUSY if enqueuing failed
  */
-int __genwqe_enqueue_ddcb(struct genwqe_dev *cd, struct ddcb_requ *req)
+int __genwqe_enqueue_ddcb(struct genwqe_dev *cd, struct ddcb_requ *req,
+			  unsigned int f_flags)
 {
 	struct ddcb *pddcb;
 	unsigned long flags;
@@ -760,6 +764,7 @@ int __genwqe_enqueue_ddcb(struct genwqe_dev *cd, struct ddcb_requ *req)
 	struct pci_dev *pci_dev = cd->pci_dev;
 	u16 icrc;
 
+ retry:
 	if (cd->card_state != GENWQE_CARD_USED) {
 		printk_ratelimited(KERN_ERR
 			"%s %s: [%s] Card is unusable/PCIe problem Req#%d\n",
@@ -785,11 +790,26 @@ int __genwqe_enqueue_ddcb(struct genwqe_dev *cd, struct ddcb_requ *req)
 
 	pddcb = get_next_ddcb(cd, queue, &req->num);	/* get ptr and num */
 	if (pddcb == NULL) {
+		int rc;
+
 		spin_unlock_irqrestore(&queue->ddcb_lock, flags);
-		queue->busy++;
+
+		if (f_flags & O_NONBLOCK) {
+			queue->return_on_busy++;
 			return -EBUSY;
 		}
 
+		queue->wait_on_busy++;
+		rc = wait_event_interruptible(queue->busy_waitq,
+					      queue_free_ddcbs(queue) != 0);
+		dev_dbg(&pci_dev->dev, "[%s] waiting for free DDCB: rc=%d\n",
+			__func__, rc);
+		if (rc == -ERESTARTSYS)
+			return rc;  /* interrupted by a signal */
+
+		goto retry;
+	}
+
 	if (queue->ddcb_req[req->num] != NULL) {
 		spin_unlock_irqrestore(&queue->ddcb_lock, flags);
 
@@ -890,9 +910,11 @@ int __genwqe_enqueue_ddcb(struct genwqe_dev *cd, struct ddcb_requ *req)
  * __genwqe_execute_raw_ddcb() - Setup and execute DDCB
  * @cd:         pointer to genwqe device descriptor
  * @req:        user provided DDCB request
+ * @f_flags:    file mode: blocking, non-blocking
  */
 int __genwqe_execute_raw_ddcb(struct genwqe_dev *cd,
-			     struct genwqe_ddcb_cmd *cmd)
+			      struct genwqe_ddcb_cmd *cmd,
+			      unsigned int f_flags)
 {
 	int rc = 0;
 	struct pci_dev *pci_dev = cd->pci_dev;
@@ -908,7 +930,7 @@ int __genwqe_execute_raw_ddcb(struct genwqe_dev *cd,
 			__func__, cmd->asiv_length);
 		return -EINVAL;
 	}
-	rc = __genwqe_enqueue_ddcb(cd, req);
+	rc = __genwqe_enqueue_ddcb(cd, req, f_flags);
 	if (rc != 0)
 		return rc;
 
@@ -1014,7 +1036,8 @@ static int setup_ddcb_queue(struct genwqe_dev *cd, struct ddcb_queue *queue)
 	queue->ddcbs_in_flight = 0;  /* statistics */
 	queue->ddcbs_max_in_flight = 0;
 	queue->ddcbs_completed = 0;
-	queue->busy = 0;
+	queue->return_on_busy = 0;
+	queue->wait_on_busy = 0;
 
 	queue->ddcb_seq	  = 0x100; /* start sequence number */
 	queue->ddcb_max	  = genwqe_ddcb_max; /* module parameter */
@@ -1054,7 +1077,7 @@ static int setup_ddcb_queue(struct genwqe_dev *cd, struct ddcb_queue *queue)
 	queue->ddcb_next = 0;	/* queue is empty */
 
 	spin_lock_init(&queue->ddcb_lock);
-	init_waitqueue_head(&queue->ddcb_waitq);
+	init_waitqueue_head(&queue->busy_waitq);
 
 	val64 = ((u64)(queue->ddcb_max - 1) <<  8); /* lastptr */
 	__genwqe_writeq(cd, queue->IO_QUEUE_CONFIG,  0x07);  /* iCRC/vCRC */
@@ -1302,6 +1325,7 @@ static int queue_wake_up_all(struct genwqe_dev *cd)
 	for (i = 0; i < queue->ddcb_max; i++)
 		wake_up_interruptible(&queue->ddcb_waitqs[queue->ddcb_act]);
 
+	wake_up_interruptible(&queue->busy_waitq);
 	spin_unlock_irqrestore(&queue->ddcb_lock, flags);
 
 	return 0;

drivers/misc/genwqe/card_debugfs.c

@@ -244,14 +244,16 @@ static int genwqe_ddcb_info_show(struct seq_file *s, void *unused)
 		   "  ddcbs_in_flight:     %u\n"
 		   "  ddcbs_max_in_flight: %u\n"
 		   "  ddcbs_completed:     %u\n"
-		   "  busy:                %u\n"
+		   "  return_on_busy:      %u\n"
+		   "  wait_on_busy:        %u\n"
 		   "  irqs_processed:      %u\n",
 		   queue->ddcb_max, (long long)queue->ddcb_daddr,
 		   (long long)queue->ddcb_daddr +
 		   (queue->ddcb_max * DDCB_LENGTH),
 		   (long long)queue->ddcb_vaddr, queue->ddcbs_in_flight,
 		   queue->ddcbs_max_in_flight, queue->ddcbs_completed,
-		   queue->busy, cd->irqs_processed);
+		   queue->return_on_busy, queue->wait_on_busy,
+		   cd->irqs_processed);
 
 	/* Hardware State */
 	seq_printf(s, "  0x%08x 0x%016llx IO_QUEUE_CONFIG\n"

drivers/misc/genwqe/card_dev.c

@@ -516,6 +516,7 @@ static int do_flash_update(struct genwqe_file *cfile,
 	u32 crc;
 	u8 cmdopts;
 	struct genwqe_dev *cd = cfile->cd;
+	struct file *filp = cfile->filp;
 	struct pci_dev *pci_dev = cd->pci_dev;
 
 	if ((load->size & 0x3) != 0)
@@ -610,7 +611,7 @@ static int do_flash_update(struct genwqe_file *cfile,
 		/* For Genwqe5 we get back the calculated CRC */
 		*(u64 *)&req->asv[0] = 0ULL;			/* 0x80 */
 
-		rc = __genwqe_execute_raw_ddcb(cd, req);
+		rc = __genwqe_execute_raw_ddcb(cd, req, filp->f_flags);
 
 		load->retc = req->retc;
 		load->attn = req->attn;
@@ -650,6 +651,7 @@ static int do_flash_read(struct genwqe_file *cfile,
 	u8 *xbuf;
 	u8 cmdopts;
 	struct genwqe_dev *cd = cfile->cd;
+	struct file *filp = cfile->filp;
 	struct pci_dev *pci_dev = cd->pci_dev;
 	struct genwqe_ddcb_cmd *cmd;
 
@@ -727,7 +729,7 @@ static int do_flash_read(struct genwqe_file *cfile,
 		/* we only get back the calculated CRC */
 		*(u64 *)&cmd->asv[0] = 0ULL;	/* 0x80 */
 
-		rc = __genwqe_execute_raw_ddcb(cd, cmd);
+		rc = __genwqe_execute_raw_ddcb(cd, cmd, filp->f_flags);
 
 		load->retc = cmd->retc;
 		load->attn = cmd->attn;
@@ -988,13 +990,14 @@ static int genwqe_execute_ddcb(struct genwqe_file *cfile,
 {
 	int rc;
 	struct genwqe_dev *cd = cfile->cd;
+	struct file *filp = cfile->filp;
 	struct ddcb_requ *req = container_of(cmd, struct ddcb_requ, cmd);
 
 	rc = ddcb_cmd_fixups(cfile, req);
 	if (rc != 0)
 		return rc;
 
-	rc = __genwqe_execute_raw_ddcb(cd, cmd);
+	rc = __genwqe_execute_raw_ddcb(cd, cmd, filp->f_flags);
 	ddcb_cmd_cleanup(cfile, req);
 	return rc;
 }
@@ -1006,6 +1009,7 @@ static int do_execute_ddcb(struct genwqe_file *cfile,
 	struct genwqe_ddcb_cmd *cmd;
 	struct ddcb_requ *req;
 	struct genwqe_dev *cd = cfile->cd;
+	struct file *filp = cfile->filp;
 
 	cmd = ddcb_requ_alloc();
 	if (cmd == NULL)
@@ -1021,7 +1025,7 @@ static int do_execute_ddcb(struct genwqe_file *cfile,
 	if (!raw)
 		rc = genwqe_execute_ddcb(cfile, cmd);
 	else
-		rc = __genwqe_execute_raw_ddcb(cd, cmd);
+		rc = __genwqe_execute_raw_ddcb(cd, cmd, filp->f_flags);
 
 	/* Copy back only the modifed fields. Do not copy ASIV
 	   back since the copy got modified by the driver. */