Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband:
  RDMA/nes: Fix incorrect unlock in nes_process_mac_intr()
  RDMA/nes: Async event for closed QP causes crash
  RDMA/nes: Have ethtool read hardware registers for rx/tx stats
  RDMA/cxgb4: Only insert sq qid in lookup table
  RDMA/cxgb4: Support IB_WR_READ_WITH_INV opcode
  RDMA/cxgb4: Set fence flag for inv-local-stag work requests
  RDMA/cxgb4: Update some HW limits
  RDMA/cxgb4: Don't limit fastreg page list depth
  RDMA/cxgb4: Return proper errors in fastreg mr/pbl allocation
  RDMA/cxgb4: Fix overflow bug in CQ arm
  RDMA/cxgb4: Optimize CQ overflow detection
  RDMA/cxgb4: CQ size must be IQ size - 2
  RDMA/cxgb4: Register RDMA provider based on LLD state_change events
  RDMA/cxgb4: Detach from the LLD after unregistering RDMA device
  IB/ipath: Remove support for QLogic PCIe QLE devices
  IB/qib: Add new qib driver for QLogic PCIe InfiniBand adapters
  IB/mad: Make needlessly global mad_sendq_size/mad_recvq_size static
  IB/core: Allow device-specific per-port sysfs files
  mlx4_core: Clean up mlx4_alloc_icm() a bit
  mlx4_core: Fix possible chunk sg list overflow in mlx4_alloc_icm()

1 files changed, 665 insertions, 0 deletions

diff --git a/drivers/infiniband/hw/qib/qib_driver.c b/drivers/infiniband/hw/qib/qib_driver.c
new file mode 100644
index 000000000000..f15ce076ac49
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_driver.c
@@ -0,0 +1,665 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
+ * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/spinlock.h>
+#include <linux/pci.h>
+#include <linux/io.h>
+#include <linux/delay.h>
+#include <linux/netdevice.h>
+#include <linux/vmalloc.h>
+
+#include "qib.h"
+
+/*
+ * The size has to be longer than this string, so we can append
+ * board/chip information to it in the init code.
+ */
+const char ib_qib_version[] = QIB_IDSTR "\n";
+
+DEFINE_SPINLOCK(qib_devs_lock);
+LIST_HEAD(qib_dev_list);
+DEFINE_MUTEX(qib_mutex);	/* general driver use */
+
+unsigned qib_ibmtu;
+module_param_named(ibmtu, qib_ibmtu, uint, S_IRUGO);
+MODULE_PARM_DESC(ibmtu, "Set max IB MTU (0=2KB, 1=256, 2=512, ... 5=4096");
+
+unsigned qib_compat_ddr_negotiate = 1;
+module_param_named(compat_ddr_negotiate, qib_compat_ddr_negotiate, uint,
+		   S_IWUSR | S_IRUGO);
+MODULE_PARM_DESC(compat_ddr_negotiate,
+		 "Attempt pre-IBTA 1.2 DDR speed negotiation");
+
+MODULE_LICENSE("Dual BSD/GPL");
+MODULE_AUTHOR("QLogic <support@qlogic.com>");
+MODULE_DESCRIPTION("QLogic IB driver");
+
+/*
+ * QIB_PIO_MAXIBHDR is the max IB header size allowed for in our
+ * PIO send buffers.  This is well beyond anything currently
+ * defined in the InfiniBand spec.
+ */
+#define QIB_PIO_MAXIBHDR 128
+
+struct qlogic_ib_stats qib_stats;
+
+const char *qib_get_unit_name(int unit)
+{
+	static char iname[16];
+
+	snprintf(iname, sizeof iname, "infinipath%u", unit);
+	return iname;
+}
+
+/*
+ * Return count of units with at least one port ACTIVE.
+ */
+int qib_count_active_units(void)
+{
+	struct qib_devdata *dd;
+	struct qib_pportdata *ppd;
+	unsigned long flags;
+	int pidx, nunits_active = 0;
+
+	spin_lock_irqsave(&qib_devs_lock, flags);
+	list_for_each_entry(dd, &qib_dev_list, list) {
+		if (!(dd->flags & QIB_PRESENT) || !dd->kregbase)
+			continue;
+		for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+			ppd = dd->pport + pidx;
+			if (ppd->lid && (ppd->lflags & (QIBL_LINKINIT |
+					 QIBL_LINKARMED | QIBL_LINKACTIVE))) {
+				nunits_active++;
+				break;
+			}
+		}
+	}
+	spin_unlock_irqrestore(&qib_devs_lock, flags);
+	return nunits_active;
+}
+
+/*
+ * Return count of all units, optionally return in arguments
+ * the number of usable (present) units, and the number of
+ * ports that are up.
+ */
+int qib_count_units(int *npresentp, int *nupp)
+{
+	int nunits = 0, npresent = 0, nup = 0;
+	struct qib_devdata *dd;
+	unsigned long flags;
+	int pidx;
+	struct qib_pportdata *ppd;
+
+	spin_lock_irqsave(&qib_devs_lock, flags);
+
+	list_for_each_entry(dd, &qib_dev_list, list) {
+		nunits++;
+		if ((dd->flags & QIB_PRESENT) && dd->kregbase)
+			npresent++;
+		for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+			ppd = dd->pport + pidx;
+			if (ppd->lid && (ppd->lflags & (QIBL_LINKINIT |
+					 QIBL_LINKARMED | QIBL_LINKACTIVE)))
+				nup++;
+		}
+	}
+
+	spin_unlock_irqrestore(&qib_devs_lock, flags);
+
+	if (npresentp)
+		*npresentp = npresent;
+	if (nupp)
+		*nupp = nup;
+
+	return nunits;
+}
+
+/**
+ * qib_wait_linkstate - wait for an IB link state change to occur
+ * @dd: the qlogic_ib device
+ * @state: the state to wait for
+ * @msecs: the number of milliseconds to wait
+ *
+ * wait up to msecs milliseconds for IB link state change to occur for
+ * now, take the easy polling route.  Currently used only by
+ * qib_set_linkstate.  Returns 0 if state reached, otherwise
+ * -ETIMEDOUT state can have multiple states set, for any of several
+ * transitions.
+ */
+int qib_wait_linkstate(struct qib_pportdata *ppd, u32 state, int msecs)
+{
+	int ret;
+	unsigned long flags;
+
+	spin_lock_irqsave(&ppd->lflags_lock, flags);
+	if (ppd->state_wanted) {
+		spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+		ret = -EBUSY;
+		goto bail;
+	}
+	ppd->state_wanted = state;
+	spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+	wait_event_interruptible_timeout(ppd->state_wait,
+					 (ppd->lflags & state),
+					 msecs_to_jiffies(msecs));
+	spin_lock_irqsave(&ppd->lflags_lock, flags);
+	ppd->state_wanted = 0;
+	spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+
+	if (!(ppd->lflags & state))
+		ret = -ETIMEDOUT;
+	else
+		ret = 0;
+bail:
+	return ret;
+}
+
+int qib_set_linkstate(struct qib_pportdata *ppd, u8 newstate)
+{
+	u32 lstate;
+	int ret;
+	struct qib_devdata *dd = ppd->dd;
+	unsigned long flags;
+
+	switch (newstate) {
+	case QIB_IB_LINKDOWN_ONLY:
+		dd->f_set_ib_cfg(ppd, QIB_IB_CFG_LSTATE,
+				 IB_LINKCMD_DOWN | IB_LINKINITCMD_NOP);
+		/* don't wait */
+		ret = 0;
+		goto bail;
+
+	case QIB_IB_LINKDOWN:
+		dd->f_set_ib_cfg(ppd, QIB_IB_CFG_LSTATE,
+				 IB_LINKCMD_DOWN | IB_LINKINITCMD_POLL);
+		/* don't wait */
+		ret = 0;
+		goto bail;
+
+	case QIB_IB_LINKDOWN_SLEEP:
+		dd->f_set_ib_cfg(ppd, QIB_IB_CFG_LSTATE,
+				 IB_LINKCMD_DOWN | IB_LINKINITCMD_SLEEP);
+		/* don't wait */
+		ret = 0;
+		goto bail;
+
+	case QIB_IB_LINKDOWN_DISABLE:
+		dd->f_set_ib_cfg(ppd, QIB_IB_CFG_LSTATE,
+				 IB_LINKCMD_DOWN | IB_LINKINITCMD_DISABLE);
+		/* don't wait */
+		ret = 0;
+		goto bail;
+
+	case QIB_IB_LINKARM:
+		if (ppd->lflags & QIBL_LINKARMED) {
+			ret = 0;
+			goto bail;
+		}
+		if (!(ppd->lflags & (QIBL_LINKINIT | QIBL_LINKACTIVE))) {
+			ret = -EINVAL;
+			goto bail;
+		}
+		/*
+		 * Since the port can be ACTIVE when we ask for ARMED,
+		 * clear QIBL_LINKV so we can wait for a transition.
+		 * If the link isn't ARMED, then something else happened
+		 * and there is no point waiting for ARMED.
+		 */
+		spin_lock_irqsave(&ppd->lflags_lock, flags);
+		ppd->lflags &= ~QIBL_LINKV;
+		spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+		dd->f_set_ib_cfg(ppd, QIB_IB_CFG_LSTATE,
+				 IB_LINKCMD_ARMED | IB_LINKINITCMD_NOP);
+		lstate = QIBL_LINKV;
+		break;
+
+	case QIB_IB_LINKACTIVE:
+		if (ppd->lflags & QIBL_LINKACTIVE) {
+			ret = 0;
+			goto bail;
+		}
+		if (!(ppd->lflags & QIBL_LINKARMED)) {
+			ret = -EINVAL;
+			goto bail;
+		}
+		dd->f_set_ib_cfg(ppd, QIB_IB_CFG_LSTATE,
+				 IB_LINKCMD_ACTIVE | IB_LINKINITCMD_NOP);
+		lstate = QIBL_LINKACTIVE;
+		break;
+
+	default:
+		ret = -EINVAL;
+		goto bail;
+	}
+	ret = qib_wait_linkstate(ppd, lstate, 10);
+
+bail:
+	return ret;
+}
+
+/*
+ * Get address of eager buffer from it's index (allocated in chunks, not
+ * contiguous).
+ */
+static inline void *qib_get_egrbuf(const struct qib_ctxtdata *rcd, u32 etail)
+{
+	const u32 chunk = etail / rcd->rcvegrbufs_perchunk;
+	const u32 idx =  etail % rcd->rcvegrbufs_perchunk;
+
+	return rcd->rcvegrbuf[chunk] + idx * rcd->dd->rcvegrbufsize;
+}
+
+/*
+ * Returns 1 if error was a CRC, else 0.
+ * Needed for some chip's synthesized error counters.
+ */
+static u32 qib_rcv_hdrerr(struct qib_pportdata *ppd, u32 ctxt,
+			  u32 eflags, u32 l, u32 etail, __le32 *rhf_addr,
+			  struct qib_message_header *hdr)
+{
+	u32 ret = 0;
+
+	if (eflags & (QLOGIC_IB_RHF_H_ICRCERR | QLOGIC_IB_RHF_H_VCRCERR))
+		ret = 1;
+	return ret;
+}
+
+/*
+ * qib_kreceive - receive a packet
+ * @rcd: the qlogic_ib context
+ * @llic: gets count of good packets needed to clear lli,
+ *          (used with chips that need need to track crcs for lli)
+ *
+ * called from interrupt handler for errors or receive interrupt
+ * Returns number of CRC error packets, needed by some chips for
+ * local link integrity tracking.   crcs are adjusted down by following
+ * good packets, if any, and count of good packets is also tracked.
+ */
+u32 qib_kreceive(struct qib_ctxtdata *rcd, u32 *llic, u32 *npkts)
+{
+	struct qib_devdata *dd = rcd->dd;
+	struct qib_pportdata *ppd = rcd->ppd;
+	__le32 *rhf_addr;
+	void *ebuf;
+	const u32 rsize = dd->rcvhdrentsize;        /* words */
+	const u32 maxcnt = dd->rcvhdrcnt * rsize;   /* words */
+	u32 etail = -1, l, hdrqtail;
+	struct qib_message_header *hdr;
+	u32 eflags, etype, tlen, i = 0, updegr = 0, crcs = 0;
+	int last;
+	u64 lval;
+	struct qib_qp *qp, *nqp;
+
+	l = rcd->head;
+	rhf_addr = (__le32 *) rcd->rcvhdrq + l + dd->rhf_offset;
+	if (dd->flags & QIB_NODMA_RTAIL) {
+		u32 seq = qib_hdrget_seq(rhf_addr);
+		if (seq != rcd->seq_cnt)
+			goto bail;
+		hdrqtail = 0;
+	} else {
+		hdrqtail = qib_get_rcvhdrtail(rcd);
+		if (l == hdrqtail)
+			goto bail;
+		smp_rmb();  /* prevent speculative reads of dma'ed hdrq */
+	}
+
+	for (last = 0, i = 1; !last; i += !last) {
+		hdr = dd->f_get_msgheader(dd, rhf_addr);
+		eflags = qib_hdrget_err_flags(rhf_addr);
+		etype = qib_hdrget_rcv_type(rhf_addr);
+		/* total length */
+		tlen = qib_hdrget_length_in_bytes(rhf_addr);
+		ebuf = NULL;
+		if ((dd->flags & QIB_NODMA_RTAIL) ?
+		    qib_hdrget_use_egr_buf(rhf_addr) :
+		    (etype != RCVHQ_RCV_TYPE_EXPECTED)) {
+			etail = qib_hdrget_index(rhf_addr);
+			updegr = 1;
+			if (tlen > sizeof(*hdr) ||
+			    etype >= RCVHQ_RCV_TYPE_NON_KD)
+				ebuf = qib_get_egrbuf(rcd, etail);
+		}
+		if (!eflags) {
+			u16 lrh_len = be16_to_cpu(hdr->lrh[2]) << 2;
+
+			if (lrh_len != tlen) {
+				qib_stats.sps_lenerrs++;
+				goto move_along;
+			}
+		}
+		if (etype == RCVHQ_RCV_TYPE_NON_KD && !eflags &&
+		    ebuf == NULL &&
+		    tlen > (dd->rcvhdrentsize - 2 + 1 -
+				qib_hdrget_offset(rhf_addr)) << 2) {
+			goto move_along;
+		}
+
+		/*
+		 * Both tiderr and qibhdrerr are set for all plain IB
+		 * packets; only qibhdrerr should be set.
+		 */
+		if (unlikely(eflags))
+			crcs += qib_rcv_hdrerr(ppd, rcd->ctxt, eflags, l,
+					       etail, rhf_addr, hdr);
+		else if (etype == RCVHQ_RCV_TYPE_NON_KD) {
+			qib_ib_rcv(rcd, hdr, ebuf, tlen);
+			if (crcs)
+				crcs--;
+			else if (llic && *llic)
+				--*llic;
+		}
+move_along:
+		l += rsize;
+		if (l >= maxcnt)
+			l = 0;
+		rhf_addr = (__le32 *) rcd->rcvhdrq + l + dd->rhf_offset;
+		if (dd->flags & QIB_NODMA_RTAIL) {
+			u32 seq = qib_hdrget_seq(rhf_addr);
+
+			if (++rcd->seq_cnt > 13)
+				rcd->seq_cnt = 1;
+			if (seq != rcd->seq_cnt)
+				last = 1;
+		} else if (l == hdrqtail)
+			last = 1;
+		/*
+		 * Update head regs etc., every 16 packets, if not last pkt,
+		 * to help prevent rcvhdrq overflows, when many packets
+		 * are processed and queue is nearly full.
+		 * Don't request an interrupt for intermediate updates.
+		 */
+		lval = l;
+		if (!last && !(i & 0xf)) {
+			dd->f_update_usrhead(rcd, lval, updegr, etail);
+			updegr = 0;
+		}
+	}
+
+	rcd->head = l;
+	rcd->pkt_count += i;
+
+	/*
+	 * Iterate over all QPs waiting to respond.
+	 * The list won't change since the IRQ is only run on one CPU.
+	 */
+	list_for_each_entry_safe(qp, nqp, &rcd->qp_wait_list, rspwait) {
+		list_del_init(&qp->rspwait);
+		if (qp->r_flags & QIB_R_RSP_NAK) {
+			qp->r_flags &= ~QIB_R_RSP_NAK;
+			qib_send_rc_ack(qp);
+		}
+		if (qp->r_flags & QIB_R_RSP_SEND) {
+			unsigned long flags;
+
+			qp->r_flags &= ~QIB_R_RSP_SEND;
+			spin_lock_irqsave(&qp->s_lock, flags);
+			if (ib_qib_state_ops[qp->state] &
+					QIB_PROCESS_OR_FLUSH_SEND)
+				qib_schedule_send(qp);
+			spin_unlock_irqrestore(&qp->s_lock, flags);
+		}
+		if (atomic_dec_and_test(&qp->refcount))
+			wake_up(&qp->wait);
+	}
+
+bail:
+	/* Report number of packets consumed */
+	if (npkts)
+		*npkts = i;
+
+	/*
+	 * Always write head at end, and setup rcv interrupt, even
+	 * if no packets were processed.
+	 */
+	lval = (u64)rcd->head | dd->rhdrhead_intr_off;
+	dd->f_update_usrhead(rcd, lval, updegr, etail);
+	return crcs;
+}
+
+/**
+ * qib_set_mtu - set the MTU
+ * @ppd: the perport data
+ * @arg: the new MTU
+ *
+ * We can handle "any" incoming size, the issue here is whether we
+ * need to restrict our outgoing size.   For now, we don't do any
+ * sanity checking on this, and we don't deal with what happens to
+ * programs that are already running when the size changes.
+ * NOTE: changing the MTU will usually cause the IBC to go back to
+ * link INIT state...
+ */
+int qib_set_mtu(struct qib_pportdata *ppd, u16 arg)
+{
+	u32 piosize;
+	int ret, chk;
+
+	if (arg != 256 && arg != 512 && arg != 1024 && arg != 2048 &&
+	    arg != 4096) {
+		ret = -EINVAL;
+		goto bail;
+	}
+	chk = ib_mtu_enum_to_int(qib_ibmtu);
+	if (chk > 0 && arg > chk) {
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	piosize = ppd->ibmaxlen;
+	ppd->ibmtu = arg;
+
+	if (arg >= (piosize - QIB_PIO_MAXIBHDR)) {
+		/* Only if it's not the initial value (or reset to it) */
+		if (piosize != ppd->init_ibmaxlen) {
+			if (arg > piosize && arg <= ppd->init_ibmaxlen)
+				piosize = ppd->init_ibmaxlen - 2 * sizeof(u32);
+			ppd->ibmaxlen = piosize;
+		}
+	} else if ((arg + QIB_PIO_MAXIBHDR) != ppd->ibmaxlen) {
+		piosize = arg + QIB_PIO_MAXIBHDR - 2 * sizeof(u32);
+		ppd->ibmaxlen = piosize;
+	}
+
+	ppd->dd->f_set_ib_cfg(ppd, QIB_IB_CFG_MTU, 0);
+
+	ret = 0;
+
+bail:
+	return ret;
+}
+
+int qib_set_lid(struct qib_pportdata *ppd, u32 lid, u8 lmc)
+{
+	struct qib_devdata *dd = ppd->dd;
+	ppd->lid = lid;
+	ppd->lmc = lmc;
+
+	dd->f_set_ib_cfg(ppd, QIB_IB_CFG_LIDLMC,
+			 lid | (~((1U << lmc) - 1)) << 16);
+
+	qib_devinfo(dd->pcidev, "IB%u:%u got a lid: 0x%x\n",
+		    dd->unit, ppd->port, lid);
+
+	return 0;
+}
+
+/*
+ * Following deal with the "obviously simple" task of overriding the state
+ * of the LEDS, which normally indicate link physical and logical status.
+ * The complications arise in dealing with different hardware mappings
+ * and the board-dependent routine being called from interrupts.
+ * and then there's the requirement to _flash_ them.
+ */
+#define LED_OVER_FREQ_SHIFT 8
+#define LED_OVER_FREQ_MASK (0xFF<<LED_OVER_FREQ_SHIFT)
+/* Below is "non-zero" to force override, but both actual LEDs are off */
+#define LED_OVER_BOTH_OFF (8)
+
+static void qib_run_led_override(unsigned long opaque)
+{
+	struct qib_pportdata *ppd = (struct qib_pportdata *)opaque;
+	struct qib_devdata *dd = ppd->dd;
+	int timeoff;
+	int ph_idx;
+
+	if (!(dd->flags & QIB_INITTED))
+		return;
+
+	ph_idx = ppd->led_override_phase++ & 1;
+	ppd->led_override = ppd->led_override_vals[ph_idx];
+	timeoff = ppd->led_override_timeoff;
+
+	dd->f_setextled(ppd, 1);
+	/*
+	 * don't re-fire the timer if user asked for it to be off; we let
+	 * it fire one more time after they turn it off to simplify
+	 */
+	if (ppd->led_override_vals[0] || ppd->led_override_vals[1])
+		mod_timer(&ppd->led_override_timer, jiffies + timeoff);
+}
+
+void qib_set_led_override(struct qib_pportdata *ppd, unsigned int val)
+{
+	struct qib_devdata *dd = ppd->dd;
+	int timeoff, freq;
+
+	if (!(dd->flags & QIB_INITTED))
+		return;
+
+	/* First check if we are blinking. If not, use 1HZ polling */
+	timeoff = HZ;
+	freq = (val & LED_OVER_FREQ_MASK) >> LED_OVER_FREQ_SHIFT;
+
+	if (freq) {
+		/* For blink, set each phase from one nybble of val */
+		ppd->led_override_vals[0] = val & 0xF;
+		ppd->led_override_vals[1] = (val >> 4) & 0xF;
+		timeoff = (HZ << 4)/freq;
+	} else {
+		/* Non-blink set both phases the same. */
+		ppd->led_override_vals[0] = val & 0xF;
+		ppd->led_override_vals[1] = val & 0xF;
+	}
+	ppd->led_override_timeoff = timeoff;
+
+	/*
+	 * If the timer has not already been started, do so. Use a "quick"
+	 * timeout so the function will be called soon, to look at our request.
+	 */
+	if (atomic_inc_return(&ppd->led_override_timer_active) == 1) {
+		/* Need to start timer */
+		init_timer(&ppd->led_override_timer);
+		ppd->led_override_timer.function = qib_run_led_override;
+		ppd->led_override_timer.data = (unsigned long) ppd;
+		ppd->led_override_timer.expires = jiffies + 1;
+		add_timer(&ppd->led_override_timer);
+	} else {
+		if (ppd->led_override_vals[0] || ppd->led_override_vals[1])
+			mod_timer(&ppd->led_override_timer, jiffies + 1);
+		atomic_dec(&ppd->led_override_timer_active);
+	}
+}
+
+/**
+ * qib_reset_device - reset the chip if possible
+ * @unit: the device to reset
+ *
+ * Whether or not reset is successful, we attempt to re-initialize the chip
+ * (that is, much like a driver unload/reload).  We clear the INITTED flag
+ * so that the various entry points will fail until we reinitialize.  For
+ * now, we only allow this if no user contexts are open that use chip resources
+ */
+int qib_reset_device(int unit)
+{
+	int ret, i;
+	struct qib_devdata *dd = qib_lookup(unit);
+	struct qib_pportdata *ppd;
+	unsigned long flags;
+	int pidx;
+
+	if (!dd) {
+		ret = -ENODEV;
+		goto bail;
+	}
+
+	qib_devinfo(dd->pcidev, "Reset on unit %u requested\n", unit);
+
+	if (!dd->kregbase || !(dd->flags & QIB_PRESENT)) {
+		qib_devinfo(dd->pcidev, "Invalid unit number %u or "
+			    "not initialized or not present\n", unit);
+		ret = -ENXIO;
+		goto bail;
+	}
+
+	spin_lock_irqsave(&dd->uctxt_lock, flags);
+	if (dd->rcd)
+		for (i = dd->first_user_ctxt; i < dd->cfgctxts; i++) {
+			if (!dd->rcd[i] || !dd->rcd[i]->cnt)
+				continue;
+			spin_unlock_irqrestore(&dd->uctxt_lock, flags);
+			ret = -EBUSY;
+			goto bail;
+		}
+	spin_unlock_irqrestore(&dd->uctxt_lock, flags);
+
+	for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+		ppd = dd->pport + pidx;
+		if (atomic_read(&ppd->led_override_timer_active)) {
+			/* Need to stop LED timer, _then_ shut off LEDs */
+			del_timer_sync(&ppd->led_override_timer);
+			atomic_set(&ppd->led_override_timer_active, 0);
+		}
+
+		/* Shut off LEDs after we are sure timer is not running */
+		ppd->led_override = LED_OVER_BOTH_OFF;
+		dd->f_setextled(ppd, 0);
+		if (dd->flags & QIB_HAS_SEND_DMA)
+			qib_teardown_sdma(ppd);
+	}
+
+	ret = dd->f_reset(dd);
+	if (ret == 1)
+		ret = qib_init(dd, 1);
+	else
+		ret = -EAGAIN;
+	if (ret)
+		qib_dev_err(dd, "Reinitialize unit %u after "
+			    "reset failed with %d\n", unit, ret);
+	else
+		qib_devinfo(dd->pcidev, "Reinitialized unit %u after "
+			    "resetting\n", unit);
+
+bail:
+	return ret;
+}