IB/qib: Optimize pio ack buffer allocation

This patch optimizes pio buffer allocation in the kernel.

For qib, kernel pio buffers are used for sending acks.  The code to
allocate the buffer would always start at 0 until it found a buffer.

This means that an average of 64 comparisions were done on each
allocate, since the busy bit won't be cleared until the bits are
refreshed when buffers are exhausted.

This patch adds two new fields in the devdata struct, last_pio and
min_kernel_pio.  last_pio is the last buffer that was allocated.
min_kernel_pio is the lowest potential available buffer.

min_kernel_pio is modifed as contexts are allocated and deallocted.
Reviewed-by: NRamkrishna Vepa <ramkrishna.vepa@intel.com>
Signed-off-by: NMike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

drivers/infiniband/hw/qib/qib.h

@@ -873,7 +873,14 @@ struct qib_devdata {
 	 * pio_writing.
 	 */
 	spinlock_t pioavail_lock;
-
+	/*
+	 * index of last buffer to optimize search for next
+	 */
+	u32 last_pio;
+	/*
+	 * min kernel pio buffer to optimize search
+	 */
+	u32 min_kernel_pio;
 	/*
 	 * Shadow copies of registers; size indicates read access size.
 	 * Most of them are readonly, but some are write-only register,

drivers/infiniband/hw/qib/qib_iba6120.c

@@ -3132,6 +3132,7 @@ static void get_6120_chip_params(struct qib_devdata *dd)
 	val = qib_read_kreg64(dd, kr_sendpiobufcnt);
 	dd->piobcnt2k = val & ~0U;
 	dd->piobcnt4k = val >> 32;
+	dd->last_pio = dd->piobcnt4k + dd->piobcnt2k - 1;
 	/* these may be adjusted in init_chip_wc_pat() */
 	dd->pio2kbase = (u32 __iomem *)
 		(((char __iomem *)dd->kregbase) + dd->pio2k_bufbase);

drivers/infiniband/hw/qib/qib_iba7220.c

@@ -4157,6 +4157,7 @@ static int qib_init_7220_variables(struct qib_devdata *dd)
 		dd->cspec->sdmabufcnt;
 	dd->lastctxt_piobuf = dd->cspec->lastbuf_for_pio - sbufs;
 	dd->cspec->lastbuf_for_pio--; /* range is <= , not < */
+	dd->last_pio = dd->cspec->lastbuf_for_pio;
 	dd->pbufsctxt = dd->lastctxt_piobuf /
 		(dd->cfgctxts - dd->first_user_ctxt);
 

drivers/infiniband/hw/qib/qib_iba7322.c

@@ -6379,6 +6379,7 @@ static int qib_init_7322_variables(struct qib_devdata *dd)
 		dd->cspec->sdmabufcnt;
 	dd->lastctxt_piobuf = dd->cspec->lastbuf_for_pio - sbufs;
 	dd->cspec->lastbuf_for_pio--; /* range is <= , not < */
+	dd->last_pio = dd->cspec->lastbuf_for_pio;
 	dd->pbufsctxt = (dd->cfgctxts > dd->first_user_ctxt) ?
 		dd->lastctxt_piobuf / (dd->cfgctxts - dd->first_user_ctxt) : 0;
 

drivers/infiniband/hw/qib/qib_tx.c

@@ -295,6 +295,7 @@ u32 __iomem *qib_getsendbuf_range(struct qib_devdata *dd, u32 *pbufnum,
 
 	nbufs = last - first + 1; /* number in range to check */
 	if (dd->upd_pio_shadow) {
+update_shadow:
 		/*
 		 * Minor optimization.  If we had no buffers on last call,
 		 * start out by doing the update; continue and do scan even
@@ -304,37 +305,39 @@ u32 __iomem *qib_getsendbuf_range(struct qib_devdata *dd, u32 *pbufnum,
 		updated++;
 	}
 	i = first;
-rescan:
 	/*
 	 * While test_and_set_bit() is atomic, we do that and then the
 	 * change_bit(), and the pair is not.  See if this is the cause
 	 * of the remaining armlaunch errors.
 	 */
 	spin_lock_irqsave(&dd->pioavail_lock, flags);
+	if (dd->last_pio >= first && dd->last_pio <= last)
+		i = dd->last_pio + 1;
+	if (!first)
+		/* adjust to min possible  */
+		nbufs = last - dd->min_kernel_pio + 1;
 	for (j = 0; j < nbufs; j++, i++) {
 		if (i > last)
-			i = first;
+			i = !first ? dd->min_kernel_pio : first;
 		if (__test_and_set_bit((2 * i) + 1, shadow))
 			continue;
 		/* flip generation bit */
 		__change_bit(2 * i, shadow);
 		/* remember that the buffer can be written to now */
 		__set_bit(i, dd->pio_writing);
+		if (!first && first != last) /* first == last on VL15, avoid */
+			dd->last_pio = i;
 		break;
 	}
 	spin_unlock_irqrestore(&dd->pioavail_lock, flags);
 
 	if (j == nbufs) {
-		if (!updated) {
+		if (!updated)
 			/*
 			 * First time through; shadow exhausted, but may be
 			 * buffers available, try an update and then rescan.
 			 */
-			update_send_bufs(dd);
-			updated++;
-			i = first;
-			goto rescan;
-		}
+			goto update_shadow;
 		no_send_bufs(dd);
 		buf = NULL;
 	} else {
@@ -422,14 +425,20 @@ void qib_chg_pioavailkernel(struct qib_devdata *dd, unsigned start,
 				__clear_bit(QLOGIC_IB_SENDPIOAVAIL_CHECK_SHIFT
 					    + start, dd->pioavailshadow);
 			__set_bit(start, dd->pioavailkernel);
+			if ((start >> 1) < dd->min_kernel_pio)
+				dd->min_kernel_pio = start >> 1;
 		} else {
 			__set_bit(start + QLOGIC_IB_SENDPIOAVAIL_BUSY_SHIFT,
 				  dd->pioavailshadow);
 			__clear_bit(start, dd->pioavailkernel);
+			if ((start >> 1) > dd->min_kernel_pio)
+				dd->min_kernel_pio = start >> 1;
 		}
 		start += 2;
 	}
 
+	if (dd->min_kernel_pio > 0 && dd->last_pio < dd->min_kernel_pio - 1)
+		dd->last_pio = dd->min_kernel_pio - 1;
 	spin_unlock_irqrestore(&dd->pioavail_lock, flags);
 
 	dd->f_txchk_change(dd, ostart, len, avail, rcd);