Merge branch 'hfi1' into k.o/for-next

Signed-off-by: NDoug Ledford <dledford@redhat.com>

drivers/infiniband/hw/hfi1/chip.c

@@ -9956,7 +9956,7 @@ int hfi1_get_ib_cfg(struct hfi1_pportdata *ppd, int which)
 		val = ppd->phy_error_threshold;
 		break;
 	case HFI1_IB_CFG_LINKDEFAULT: /* IB link default (sleep/poll) */
-		val = dd->link_default;
+		val = HLS_DEFAULT;
 		break;
 
 	case HFI1_IB_CFG_HRTBT: /* Heartbeat off/enable/auto */
@@ -10159,6 +10159,10 @@ static const char * const state_complete_reasons[] = {
 	[0x33] =
 	  "Link partner completed the VerifyCap state, but the passing lanes do not meet the local link width policy",
 	[0x34] = tx_out_of_policy,
+	[0x35] = "Negotiated link width is mutually exclusive",
+	[0x36] =
+	  "Timed out before receiving verifycap frames in VerifyCap.Exchange",
+	[0x37] = "Unable to resolve secure data exchange",
 };
 
 static const char *state_complete_reason_code_string(struct hfi1_pportdata *ppd,
@@ -10547,7 +10551,7 @@ int set_link_state(struct hfi1_pportdata *ppd, u32 state)
 
 	orig_new_state = state;
 	if (state == HLS_DN_DOWNDEF)
-		state = dd->link_default;
+		state = HLS_DEFAULT;
 
 	/* interpret poll -> poll as a link bounce */
 	poll_bounce = ppd->host_link_state == HLS_DN_POLL &&
@@ -12925,7 +12929,7 @@ static void clean_up_interrupts(struct hfi1_devdata *dd)
 			if (!me->arg) /* => no irq, no affinity */
 				continue;
 			hfi1_put_irq_affinity(dd, me);
-			free_irq(me->irq, me->arg);
+			pci_free_irq(dd->pcidev, i, me->arg);
 		}
 
 		/* clean structures */
@@ -12935,7 +12939,7 @@ static void clean_up_interrupts(struct hfi1_devdata *dd)
 	} else {
 		/* INTx */
 		if (dd->requested_intx_irq) {
-			free_irq(dd->pcidev->irq, dd);
+			pci_free_irq(dd->pcidev, 0, dd);
 			dd->requested_intx_irq = 0;
 		}
 		disable_intx(dd->pcidev);
@@ -12994,10 +12998,8 @@ static int request_intx_irq(struct hfi1_devdata *dd)
 {
 	int ret;
 
-	snprintf(dd->intx_name, sizeof(dd->intx_name), DRIVER_NAME "_%d",
-		 dd->unit);
-	ret = request_irq(dd->pcidev->irq, general_interrupt,
-			  IRQF_SHARED, dd->intx_name, dd);
+	ret = pci_request_irq(dd->pcidev, 0, general_interrupt, NULL, dd,
+			      DRIVER_NAME "_%d", dd->unit);
 	if (ret)
 		dd_dev_err(dd, "unable to request INTx interrupt, err %d\n",
 			   ret);
@@ -13040,13 +13042,14 @@ static int request_msix_irqs(struct hfi1_devdata *dd)
 		int idx;
 		struct hfi1_ctxtdata *rcd = NULL;
 		struct sdma_engine *sde = NULL;
+		char name[MAX_NAME_SIZE];
 
-		/* obtain the arguments to request_irq */
+		/* obtain the arguments to pci_request_irq */
 		if (first_general <= i && i < last_general) {
 			idx = i - first_general;
 			handler = general_interrupt;
 			arg = dd;
-			snprintf(me->name, sizeof(me->name),
+			snprintf(name, sizeof(name),
 				 DRIVER_NAME "_%d", dd->unit);
 			err_info = "general";
 			me->type = IRQ_GENERAL;
@@ -13055,14 +13058,14 @@ static int request_msix_irqs(struct hfi1_devdata *dd)
 			sde = &dd->per_sdma[idx];
 			handler = sdma_interrupt;
 			arg = sde;
-			snprintf(me->name, sizeof(me->name),
+			snprintf(name, sizeof(name),
 				 DRIVER_NAME "_%d sdma%d", dd->unit, idx);
 			err_info = "sdma";
 			remap_sdma_interrupts(dd, idx, i);
 			me->type = IRQ_SDMA;
 		} else if (first_rx <= i && i < last_rx) {
 			idx = i - first_rx;
-			rcd = hfi1_rcd_get_by_index(dd, idx);
+			rcd = hfi1_rcd_get_by_index_safe(dd, idx);
 			if (rcd) {
 				/*
 				 * Set the interrupt register and mask for this
@@ -13074,7 +13077,7 @@ static int request_msix_irqs(struct hfi1_devdata *dd)
 				handler = receive_context_interrupt;
 				thread = receive_context_thread;
 				arg = rcd;
-				snprintf(me->name, sizeof(me->name),
+				snprintf(name, sizeof(name),
 					 DRIVER_NAME "_%d kctxt%d",
 					 dd->unit, idx);
 				err_info = "receive context";
@@ -13095,18 +13098,10 @@ static int request_msix_irqs(struct hfi1_devdata *dd)
 		if (!arg)
 			continue;
 		/* make sure the name is terminated */
-		me->name[sizeof(me->name) - 1] = 0;
+		name[sizeof(name) - 1] = 0;
 		me->irq = pci_irq_vector(dd->pcidev, i);
-		/*
-		 * On err return me->irq.  Don't need to clear this
-		 * because 'arg' has not been set, and cleanup will
-		 * do the right thing.
-		 */
-		if (me->irq < 0)
-			return me->irq;
-
-		ret = request_threaded_irq(me->irq, handler, thread, 0,
-					   me->name, arg);
+		ret = pci_request_irq(dd->pcidev, i, handler, thread, arg,
+				      name);
 		if (ret) {
 			dd_dev_err(dd,
 				   "unable to allocate %s interrupt, irq %d, index %d, err %d\n",
@@ -13114,7 +13109,7 @@ static int request_msix_irqs(struct hfi1_devdata *dd)
 			return ret;
 		}
 		/*
-		 * assign arg after request_irq call, so it will be
+		 * assign arg after pci_request_irq call, so it will be
 		 * cleaned up
 		 */
 		me->arg = arg;
@@ -13132,7 +13127,7 @@ void hfi1_vnic_synchronize_irq(struct hfi1_devdata *dd)
 	int i;
 
 	if (!dd->num_msix_entries) {
-		synchronize_irq(dd->pcidev->irq);
+		synchronize_irq(pci_irq_vector(dd->pcidev, 0));
 		return;
 	}
 
@@ -13153,7 +13148,7 @@ void hfi1_reset_vnic_msix_info(struct hfi1_ctxtdata *rcd)
 		return;
 
 	hfi1_put_irq_affinity(dd, me);
-	free_irq(me->irq, me->arg);
+	pci_free_irq(dd->pcidev, rcd->msix_intr, me->arg);
 
 	me->arg = NULL;
 }
@@ -13176,28 +13171,21 @@ void hfi1_set_vnic_msix_info(struct hfi1_ctxtdata *rcd)
 	rcd->ireg = (IS_RCVAVAIL_START + idx) / 64;
 	rcd->imask = ((u64)1) <<
 		  ((IS_RCVAVAIL_START + idx) % 64);
-
-	snprintf(me->name, sizeof(me->name),
-		 DRIVER_NAME "_%d kctxt%d", dd->unit, idx);
-	me->name[sizeof(me->name) - 1] = 0;
 	me->type = IRQ_RCVCTXT;
 	me->irq = pci_irq_vector(dd->pcidev, rcd->msix_intr);
-	if (me->irq < 0) {
-		dd_dev_err(dd, "vnic irq vector request (idx %d) fail %d\n",
-			   idx, me->irq);
-		return;
-	}
 	remap_intr(dd, IS_RCVAVAIL_START + idx, rcd->msix_intr);
 
-	ret = request_threaded_irq(me->irq, receive_context_interrupt,
-				   receive_context_thread, 0, me->name, arg);
+	ret = pci_request_irq(dd->pcidev, rcd->msix_intr,
+			      receive_context_interrupt,
+			      receive_context_thread, arg,
+			      DRIVER_NAME "_%d kctxt%d", dd->unit, idx);
 	if (ret) {
 		dd_dev_err(dd, "vnic irq request (irq %d, idx %d) fail %d\n",
 			   me->irq, idx, ret);
 		return;
 	}
 	/*
-	 * assign arg after request_irq call, so it will be
+	 * assign arg after pci_request_irq call, so it will be
 	 * cleaned up
 	 */
 	me->arg = arg;
@@ -13206,7 +13194,7 @@ void hfi1_set_vnic_msix_info(struct hfi1_ctxtdata *rcd)
 	if (ret) {
 		dd_dev_err(dd,
 			   "unable to pin IRQ %d\n", ret);
-		free_irq(me->irq, me->arg);
+		pci_free_irq(dd->pcidev, rcd->msix_intr, me->arg);
 	}
 }
 
@@ -14907,8 +14895,6 @@ struct hfi1_devdata *hfi1_init_dd(struct pci_dev *pdev,
 		init_vl_arb_caches(ppd);
 	}
 
-	dd->link_default = HLS_DN_POLL;
-
 	/*
 	 * Do remaining PCIe setup and save PCIe values in dd.
 	 * Any error printing is already done by the init code.

drivers/infiniband/hw/hfi1/debugfs.c

@@ -243,7 +243,7 @@ static int _ctx_stats_seq_show(struct seq_file *s, void *v)
 	spos = v;
 	i = *spos;
 
-	rcd = hfi1_rcd_get_by_index(dd, i);
+	rcd = hfi1_rcd_get_by_index_safe(dd, i);
 	if (!rcd)
 		return SEQ_SKIP;
 
@@ -402,7 +402,7 @@ static int _rcds_seq_show(struct seq_file *s, void *v)
 	loff_t *spos = v;
 	loff_t i = *spos;
 
-	rcd = hfi1_rcd_get_by_index(dd, i);
+	rcd = hfi1_rcd_get_by_index_safe(dd, i);
 	if (rcd)
 		seqfile_dump_rcd(s, rcd);
 	hfi1_rcd_put(rcd);

drivers/infiniband/hw/hfi1/driver.c

@@ -866,7 +866,7 @@ static inline void set_nodma_rtail(struct hfi1_devdata *dd, u16 ctxt)
 	 * interrupt handler for all statically allocated kernel contexts.
 	 */
 	if (ctxt >= dd->first_dyn_alloc_ctxt) {
-		rcd = hfi1_rcd_get_by_index(dd, ctxt);
+		rcd = hfi1_rcd_get_by_index_safe(dd, ctxt);
 		if (rcd) {
 			rcd->do_interrupt =
 				&handle_receive_interrupt_nodma_rtail;
@@ -895,7 +895,7 @@ static inline void set_dma_rtail(struct hfi1_devdata *dd, u16 ctxt)
 	 * interrupt handler for all statically allocated kernel contexts.
 	 */
 	if (ctxt >= dd->first_dyn_alloc_ctxt) {
-		rcd = hfi1_rcd_get_by_index(dd, ctxt);
+		rcd = hfi1_rcd_get_by_index_safe(dd, ctxt);
 		if (rcd) {
 			rcd->do_interrupt =
 				&handle_receive_interrupt_dma_rtail;

drivers/infiniband/hw/hfi1/file_ops.c

@@ -78,16 +78,20 @@ static unsigned int hfi1_poll(struct file *fp, struct poll_table_struct *pt);
 static int hfi1_file_mmap(struct file *fp, struct vm_area_struct *vma);
 
 static u64 kvirt_to_phys(void *addr);
-static int assign_ctxt(struct hfi1_filedata *fd, struct hfi1_user_info *uinfo);
+static int assign_ctxt(struct hfi1_filedata *fd, unsigned long arg, u32 len);
 static void init_subctxts(struct hfi1_ctxtdata *uctxt,
 			  const struct hfi1_user_info *uinfo);
 static int init_user_ctxt(struct hfi1_filedata *fd,
 			  struct hfi1_ctxtdata *uctxt);
 static void user_init(struct hfi1_ctxtdata *uctxt);
-static int get_ctxt_info(struct hfi1_filedata *fd, void __user *ubase,
-			 __u32 len);
-static int get_base_info(struct hfi1_filedata *fd, void __user *ubase,
-			 __u32 len);
+static int get_ctxt_info(struct hfi1_filedata *fd, unsigned long arg, u32 len);
+static int get_base_info(struct hfi1_filedata *fd, unsigned long arg, u32 len);
+static int user_exp_rcv_setup(struct hfi1_filedata *fd, unsigned long arg,
+			      u32 len);
+static int user_exp_rcv_clear(struct hfi1_filedata *fd, unsigned long arg,
+			      u32 len);
+static int user_exp_rcv_invalid(struct hfi1_filedata *fd, unsigned long arg,
+				u32 len);
 static int setup_base_ctxt(struct hfi1_filedata *fd,
 			   struct hfi1_ctxtdata *uctxt);
 static int setup_subctxt(struct hfi1_ctxtdata *uctxt);
@@ -101,10 +105,11 @@ static void deallocate_ctxt(struct hfi1_ctxtdata *uctxt);
 static unsigned int poll_urgent(struct file *fp, struct poll_table_struct *pt);
 static unsigned int poll_next(struct file *fp, struct poll_table_struct *pt);
 static int user_event_ack(struct hfi1_ctxtdata *uctxt, u16 subctxt,
-			  unsigned long events);
-static int set_ctxt_pkey(struct hfi1_ctxtdata *uctxt, u16 subctxt, u16 pkey);
+			  unsigned long arg);
+static int set_ctxt_pkey(struct hfi1_ctxtdata *uctxt, unsigned long arg);
+static int ctxt_reset(struct hfi1_ctxtdata *uctxt);
 static int manage_rcvq(struct hfi1_ctxtdata *uctxt, u16 subctxt,
-		       int start_stop);
+		       unsigned long arg);
 static int vma_fault(struct vm_fault *vmf);
 static long hfi1_file_ioctl(struct file *fp, unsigned int cmd,
 			    unsigned long arg);
@@ -221,13 +226,8 @@ static long hfi1_file_ioctl(struct file *fp, unsigned int cmd,
 {
 	struct hfi1_filedata *fd = fp->private_data;
 	struct hfi1_ctxtdata *uctxt = fd->uctxt;
-	struct hfi1_user_info uinfo;
-	struct hfi1_tid_info tinfo;
 	int ret = 0;
-	unsigned long addr;
 	int uval = 0;
-	unsigned long ul_uval = 0;
-	u16 uval16 = 0;
 
 	hfi1_cdbg(IOCTL, "IOCTL recv: 0x%x", cmd);
 	if (cmd != HFI1_IOCTL_ASSIGN_CTXT &&
@@ -237,171 +237,55 @@ static long hfi1_file_ioctl(struct file *fp, unsigned int cmd,
 
 	switch (cmd) {
 	case HFI1_IOCTL_ASSIGN_CTXT:
-		if (uctxt)
-			return -EINVAL;
-
-		if (copy_from_user(&uinfo,
-				   (struct hfi1_user_info __user *)arg,
-				   sizeof(uinfo)))
-			return -EFAULT;
-
-		ret = assign_ctxt(fd, &uinfo);
+		ret = assign_ctxt(fd, arg, _IOC_SIZE(cmd));
 		break;
+
 	case HFI1_IOCTL_CTXT_INFO:
-		ret = get_ctxt_info(fd, (void __user *)(unsigned long)arg,
-				    sizeof(struct hfi1_ctxt_info));
+		ret = get_ctxt_info(fd, arg, _IOC_SIZE(cmd));
 		break;
+
 	case HFI1_IOCTL_USER_INFO:
-		ret = get_base_info(fd, (void __user *)(unsigned long)arg,
-				    sizeof(struct hfi1_base_info));
+		ret = get_base_info(fd, arg, _IOC_SIZE(cmd));
 		break;
+
 	case HFI1_IOCTL_CREDIT_UPD:
 		if (uctxt)
 			sc_return_credits(uctxt->sc);
 		break;
 
 	case HFI1_IOCTL_TID_UPDATE:
-		if (copy_from_user(&tinfo,
-				   (struct hfi11_tid_info __user *)arg,
-				   sizeof(tinfo)))
-			return -EFAULT;
-
-		ret = hfi1_user_exp_rcv_setup(fd, &tinfo);
-		if (!ret) {
-			/*
-			 * Copy the number of tidlist entries we used
-			 * and the length of the buffer we registered.
-			 */
-			addr = arg + offsetof(struct hfi1_tid_info, tidcnt);
-			if (copy_to_user((void __user *)addr, &tinfo.tidcnt,
-					 sizeof(tinfo.tidcnt)))
-				return -EFAULT;
-
-			addr = arg + offsetof(struct hfi1_tid_info, length);
-			if (copy_to_user((void __user *)addr, &tinfo.length,
-					 sizeof(tinfo.length)))
-				ret = -EFAULT;
-		}
+		ret = user_exp_rcv_setup(fd, arg, _IOC_SIZE(cmd));
 		break;
 
 	case HFI1_IOCTL_TID_FREE:
-		if (copy_from_user(&tinfo,
-				   (struct hfi11_tid_info __user *)arg,
-				   sizeof(tinfo)))
-			return -EFAULT;
-
-		ret = hfi1_user_exp_rcv_clear(fd, &tinfo);
-		if (ret)
-			break;
-		addr = arg + offsetof(struct hfi1_tid_info, tidcnt);
-		if (copy_to_user((void __user *)addr, &tinfo.tidcnt,
-				 sizeof(tinfo.tidcnt)))
-			ret = -EFAULT;
+		ret = user_exp_rcv_clear(fd, arg, _IOC_SIZE(cmd));
 		break;
 
 	case HFI1_IOCTL_TID_INVAL_READ:
-		if (copy_from_user(&tinfo,
-				   (struct hfi11_tid_info __user *)arg,
-				   sizeof(tinfo)))
-			return -EFAULT;
-
-		ret = hfi1_user_exp_rcv_invalid(fd, &tinfo);
-		if (ret)
-			break;
-		addr = arg + offsetof(struct hfi1_tid_info, tidcnt);
-		if (copy_to_user((void __user *)addr, &tinfo.tidcnt,
-				 sizeof(tinfo.tidcnt)))
-			ret = -EFAULT;
+		ret = user_exp_rcv_invalid(fd, arg, _IOC_SIZE(cmd));
 		break;
 
 	case HFI1_IOCTL_RECV_CTRL:
-		ret = get_user(uval, (int __user *)arg);
-		if (ret != 0)
-			return -EFAULT;
-		ret = manage_rcvq(uctxt, fd->subctxt, uval);
+		ret = manage_rcvq(uctxt, fd->subctxt, arg);
 		break;
 
 	case HFI1_IOCTL_POLL_TYPE:
-		ret = get_user(uval, (int __user *)arg);
-		if (ret != 0)
+		if (get_user(uval, (int __user *)arg))
 			return -EFAULT;
 		uctxt->poll_type = (typeof(uctxt->poll_type))uval;
 		break;
 
 	case HFI1_IOCTL_ACK_EVENT:
-		ret = get_user(ul_uval, (unsigned long __user *)arg);
-		if (ret != 0)
-			return -EFAULT;
-		ret = user_event_ack(uctxt, fd->subctxt, ul_uval);
+		ret = user_event_ack(uctxt, fd->subctxt, arg);
 		break;
 
 	case HFI1_IOCTL_SET_PKEY:
-		ret = get_user(uval16, (u16 __user *)arg);
-		if (ret != 0)
-			return -EFAULT;
-		if (HFI1_CAP_IS_USET(PKEY_CHECK))
-			ret = set_ctxt_pkey(uctxt, fd->subctxt, uval16);
-		else
-			return -EPERM;
+		ret = set_ctxt_pkey(uctxt, arg);
 		break;
 
-	case HFI1_IOCTL_CTXT_RESET: {
-		struct send_context *sc;
-		struct hfi1_devdata *dd;
-
-		if (!uctxt || !uctxt->dd || !uctxt->sc)
-			return -EINVAL;
-
-		/*
-		 * There is no protection here. User level has to
-		 * guarantee that no one will be writing to the send
-		 * context while it is being re-initialized.
-		 * If user level breaks that guarantee, it will break
-		 * it's own context and no one else's.
-		 */
-		dd = uctxt->dd;
-		sc = uctxt->sc;
-		/*
-		 * Wait until the interrupt handler has marked the
-		 * context as halted or frozen. Report error if we time
-		 * out.
-		 */
-		wait_event_interruptible_timeout(
-			sc->halt_wait, (sc->flags & SCF_HALTED),
-			msecs_to_jiffies(SEND_CTXT_HALT_TIMEOUT));
-		if (!(sc->flags & SCF_HALTED))
-			return -ENOLCK;
-
-		/*
-		 * If the send context was halted due to a Freeze,
-		 * wait until the device has been "unfrozen" before
-		 * resetting the context.
-		 */
-		if (sc->flags & SCF_FROZEN) {
-			wait_event_interruptible_timeout(
-				dd->event_queue,
-				!(ACCESS_ONCE(dd->flags) & HFI1_FROZEN),
-				msecs_to_jiffies(SEND_CTXT_HALT_TIMEOUT));
-			if (dd->flags & HFI1_FROZEN)
-				return -ENOLCK;
-
-			if (dd->flags & HFI1_FORCED_FREEZE)
-				/*
-				 * Don't allow context reset if we are into
-				 * forced freeze
-				 */
-				return -ENODEV;
-
-			sc_disable(sc);
-			ret = sc_enable(sc);
-			hfi1_rcvctrl(dd, HFI1_RCVCTRL_CTXT_ENB, uctxt);
-		} else {
-			ret = sc_restart(sc);
-		}
-		if (!ret)
-			sc_return_credits(sc);
+	case HFI1_IOCTL_CTXT_RESET:
+		ret = ctxt_reset(uctxt);
 		break;
-	}
 
 	case HFI1_IOCTL_GET_VERS:
 		uval = HFI1_USER_SWVERSION;
@@ -595,9 +479,8 @@ static int hfi1_file_mmap(struct file *fp, struct vm_area_struct *vma)
 		 * Use the page where this context's flags are. User level
 		 * knows where it's own bitmap is within the page.
 		 */
-		memaddr = (unsigned long)(dd->events +
-				  ((uctxt->ctxt - dd->first_dyn_alloc_ctxt) *
-				   HFI1_MAX_SHARED_CTXTS)) & PAGE_MASK;
+		memaddr = (unsigned long)
+			(dd->events + uctxt_offset(uctxt)) & PAGE_MASK;
 		memlen = PAGE_SIZE;
 		/*
 		 * v3.7 removes VM_RESERVED but the effect is kept by
@@ -779,8 +662,7 @@ static int hfi1_file_close(struct inode *inode, struct file *fp)
 	 * Clear any left over, unhandled events so the next process that
 	 * gets this context doesn't get confused.
 	 */
-	ev = dd->events + ((uctxt->ctxt - dd->first_dyn_alloc_ctxt) *
-			   HFI1_MAX_SHARED_CTXTS) + fdata->subctxt;
+	ev = dd->events + uctxt_offset(uctxt) + fdata->subctxt;
 	*ev = 0;
 
 	spin_lock_irqsave(&dd->uctxt_lock, flags);
@@ -891,20 +773,30 @@ static int complete_subctxt(struct hfi1_filedata *fd)
 	return ret;
 }
 
-static int assign_ctxt(struct hfi1_filedata *fd, struct hfi1_user_info *uinfo)
+static int assign_ctxt(struct hfi1_filedata *fd, unsigned long arg, u32 len)
 {
 	int ret;
 	unsigned int swmajor, swminor;
 	struct hfi1_ctxtdata *uctxt = NULL;
+	struct hfi1_user_info uinfo;
+
+	if (fd->uctxt)
+		return -EINVAL;
 
-	swmajor = uinfo->userversion >> 16;
+	if (sizeof(uinfo) != len)
+		return -EINVAL;
+
+	if (copy_from_user(&uinfo, (void __user *)arg, sizeof(uinfo)))
+		return -EFAULT;
+
+	swmajor = uinfo.userversion >> 16;
 	if (swmajor != HFI1_USER_SWMAJOR)
 		return -ENODEV;
 
-	if (uinfo->subctxt_cnt > HFI1_MAX_SHARED_CTXTS)
+	if (uinfo.subctxt_cnt > HFI1_MAX_SHARED_CTXTS)
 		return -EINVAL;
 
-	swminor = uinfo->userversion & 0xffff;
+	swminor = uinfo.userversion & 0xffff;
 
 	/*
 	 * Acquire the mutex to protect against multiple creations of what
@@ -915,14 +807,14 @@ static int assign_ctxt(struct hfi1_filedata *fd, struct hfi1_user_info *uinfo)
 	 * Get a sub context if available  (fd->uctxt will be set).
 	 * ret < 0 error, 0 no context, 1 sub-context found
 	 */
-	ret = find_sub_ctxt(fd, uinfo);
+	ret = find_sub_ctxt(fd, &uinfo);
 
 	/*
 	 * Allocate a base context if context sharing is not required or a
 	 * sub context wasn't found.
 	 */
 	if (!ret)
-		ret = allocate_ctxt(fd, fd->dd, uinfo, &uctxt);
+		ret = allocate_ctxt(fd, fd->dd, &uinfo, &uctxt);
 
 	mutex_unlock(&hfi1_mutex);
 
@@ -1237,12 +1129,13 @@ static void user_init(struct hfi1_ctxtdata *uctxt)
 	hfi1_rcvctrl(uctxt->dd, rcvctrl_ops, uctxt);
 }
 
-static int get_ctxt_info(struct hfi1_filedata *fd, void __user *ubase,
-			 __u32 len)
+static int get_ctxt_info(struct hfi1_filedata *fd, unsigned long arg, u32 len)
 {
 	struct hfi1_ctxt_info cinfo;
 	struct hfi1_ctxtdata *uctxt = fd->uctxt;
-	int ret = 0;
+
+	if (sizeof(cinfo) != len)
+		return -EINVAL;
 
 	memset(&cinfo, 0, sizeof(cinfo));
 	cinfo.runtime_flags = (((uctxt->flags >> HFI1_CAP_MISC_SHIFT) &
@@ -1272,10 +1165,10 @@ static int get_ctxt_info(struct hfi1_filedata *fd, void __user *ubase,
 	cinfo.rcvegr_size = uctxt->egrbufs.rcvtid_size;
 
 	trace_hfi1_ctxt_info(uctxt->dd, uctxt->ctxt, fd->subctxt, cinfo);
-	if (copy_to_user(ubase, &cinfo, sizeof(cinfo)))
-		ret = -EFAULT;
+	if (copy_to_user((void __user *)arg, &cinfo, len))
+		return -EFAULT;
 
-	return ret;
+	return 0;
 }
 
 static int init_user_ctxt(struct hfi1_filedata *fd,
@@ -1341,18 +1234,18 @@ static int setup_base_ctxt(struct hfi1_filedata *fd,
 	return ret;
 }
 
-static int get_base_info(struct hfi1_filedata *fd, void __user *ubase,
-			 __u32 len)
+static int get_base_info(struct hfi1_filedata *fd, unsigned long arg, u32 len)
 {
 	struct hfi1_base_info binfo;
 	struct hfi1_ctxtdata *uctxt = fd->uctxt;
 	struct hfi1_devdata *dd = uctxt->dd;
-	ssize_t sz;
 	unsigned offset;
-	int ret = 0;
 
 	trace_hfi1_uctxtdata(uctxt->dd, uctxt, fd->subctxt);
 
+	if (sizeof(binfo) != len)
+		return -EINVAL;
+
 	memset(&binfo, 0, sizeof(binfo));
 	binfo.hw_version = dd->revision;
 	binfo.sw_version = HFI1_KERN_SWVERSION;
@@ -1389,8 +1282,7 @@ static int get_base_info(struct hfi1_filedata *fd, void __user *ubase,
 	 */
 	binfo.user_regbase = HFI1_MMAP_TOKEN(UREGS, uctxt->ctxt,
 					     fd->subctxt, 0);
-	offset = offset_in_page((((uctxt->ctxt - dd->first_dyn_alloc_ctxt) *
-		    HFI1_MAX_SHARED_CTXTS) + fd->subctxt) *
+	offset = offset_in_page((uctxt_offset(uctxt) + fd->subctxt) *
 				sizeof(*dd->events));
 	binfo.events_bufbase = HFI1_MMAP_TOKEN(EVENTS, uctxt->ctxt,
 					       fd->subctxt,
@@ -1412,9 +1304,123 @@ static int get_base_info(struct hfi1_filedata *fd, void __user *ubase,
 							  uctxt->ctxt,
 							  fd->subctxt, 0);
 	}
-	sz = (len < sizeof(binfo)) ? len : sizeof(binfo);
-	if (copy_to_user(ubase, &binfo, sz))
+
+	if (copy_to_user((void __user *)arg, &binfo, len))
+		return -EFAULT;
+
+	return 0;
+}
+
+/**
+ * user_exp_rcv_setup - Set up the given tid rcv list
+ * @fd: file data of the current driver instance
+ * @arg: ioctl argumnent for user space information
+ * @len: length of data structure associated with ioctl command
+ *
+ * Wrapper to validate ioctl information before doing _rcv_setup.
+ *
+ */
+static int user_exp_rcv_setup(struct hfi1_filedata *fd, unsigned long arg,
+			      u32 len)
+{
+	int ret;
+	unsigned long addr;
+	struct hfi1_tid_info tinfo;
+
+	if (sizeof(tinfo) != len)
+		return -EINVAL;
+
+	if (copy_from_user(&tinfo, (void __user *)arg, (sizeof(tinfo))))
+		return -EFAULT;
+
+	ret = hfi1_user_exp_rcv_setup(fd, &tinfo);
+	if (!ret) {
+		/*
+		 * Copy the number of tidlist entries we used
+		 * and the length of the buffer we registered.
+		 */
+		addr = arg + offsetof(struct hfi1_tid_info, tidcnt);
+		if (copy_to_user((void __user *)addr, &tinfo.tidcnt,
+				 sizeof(tinfo.tidcnt)))
+			return -EFAULT;
+
+		addr = arg + offsetof(struct hfi1_tid_info, length);
+		if (copy_to_user((void __user *)addr, &tinfo.length,
+				 sizeof(tinfo.length)))
+			ret = -EFAULT;
+	}
+
+	return ret;
+}
+
+/**
+ * user_exp_rcv_clear - Clear the given tid rcv list
+ * @fd: file data of the current driver instance
+ * @arg: ioctl argumnent for user space information
+ * @len: length of data structure associated with ioctl command
+ *
+ * The hfi1_user_exp_rcv_clear() can be called from the error path.  Because
+ * of this, we need to use this wrapper to copy the user space information
+ * before doing the clear.
+ */
+static int user_exp_rcv_clear(struct hfi1_filedata *fd, unsigned long arg,
+			      u32 len)
+{
+	int ret;
+	unsigned long addr;
+	struct hfi1_tid_info tinfo;
+
+	if (sizeof(tinfo) != len)
+		return -EINVAL;
+
+	if (copy_from_user(&tinfo, (void __user *)arg, (sizeof(tinfo))))
+		return -EFAULT;
+
+	ret = hfi1_user_exp_rcv_clear(fd, &tinfo);
+	if (!ret) {
+		addr = arg + offsetof(struct hfi1_tid_info, tidcnt);
+		if (copy_to_user((void __user *)addr, &tinfo.tidcnt,
+				 sizeof(tinfo.tidcnt)))
+			return -EFAULT;
+	}
+
+	return ret;
+}
+
+/**
+ * user_exp_rcv_invalid - Invalidate the given tid rcv list
+ * @fd: file data of the current driver instance
+ * @arg: ioctl argumnent for user space information
+ * @len: length of data structure associated with ioctl command
+ *
+ * Wrapper to validate ioctl information before doing _rcv_invalid.
+ *
+ */
+static int user_exp_rcv_invalid(struct hfi1_filedata *fd, unsigned long arg,
+				u32 len)
+{
+	int ret;
+	unsigned long addr;
+	struct hfi1_tid_info tinfo;
+
+	if (sizeof(tinfo) != len)
+		return -EINVAL;
+
+	if (!fd->invalid_tids)
+		return -EINVAL;
+
+	if (copy_from_user(&tinfo, (void __user *)arg, (sizeof(tinfo))))
+		return -EFAULT;
+
+	ret = hfi1_user_exp_rcv_invalid(fd, &tinfo);
+	if (ret)
+		return ret;
+
+	addr = arg + offsetof(struct hfi1_tid_info, tidcnt);
+	if (copy_to_user((void __user *)addr, &tinfo.tidcnt,
+			 sizeof(tinfo.tidcnt)))
 		ret = -EFAULT;
+
 	return ret;
 }
 
@@ -1482,14 +1488,13 @@ int hfi1_set_uevent_bits(struct hfi1_pportdata *ppd, const int evtbit)
 	     ctxt++) {
 		uctxt = hfi1_rcd_get_by_index(dd, ctxt);
 		if (uctxt) {
-			unsigned long *evs = dd->events +
-				(uctxt->ctxt - dd->first_dyn_alloc_ctxt) *
-				HFI1_MAX_SHARED_CTXTS;
+			unsigned long *evs;
 			int i;
 			/*
 			 * subctxt_cnt is 0 if not shared, so do base
 			 * separately, first, then remaining subctxt, if any
 			 */
+			evs = dd->events + uctxt_offset(uctxt);
 			set_bit(evtbit, evs);
 			for (i = 1; i < uctxt->subctxt_cnt; i++)
 				set_bit(evtbit, evs + i);
@@ -1511,13 +1516,18 @@ int hfi1_set_uevent_bits(struct hfi1_pportdata *ppd, const int evtbit)
  * re-init the software copy of the head register
  */
 static int manage_rcvq(struct hfi1_ctxtdata *uctxt, u16 subctxt,
-		       int start_stop)
+		       unsigned long arg)
 {
 	struct hfi1_devdata *dd = uctxt->dd;
 	unsigned int rcvctrl_op;
+	int start_stop;
 
 	if (subctxt)
-		goto bail;
+		return 0;
+
+	if (get_user(start_stop, (int __user *)arg))
+		return -EFAULT;
+
 	/* atomically clear receive enable ctxt. */
 	if (start_stop) {
 		/*
@@ -1536,7 +1546,7 @@ static int manage_rcvq(struct hfi1_ctxtdata *uctxt, u16 subctxt,
 	}
 	hfi1_rcvctrl(dd, rcvctrl_op, uctxt);
 	/* always; new head should be equal to new tail; see above */
-bail:
+
 	return 0;
 }
 
@@ -1546,17 +1556,20 @@ static int manage_rcvq(struct hfi1_ctxtdata *uctxt, u16 subctxt,
  * set, if desired, and checks again in future.
  */
 static int user_event_ack(struct hfi1_ctxtdata *uctxt, u16 subctxt,
-			  unsigned long events)
+			  unsigned long arg)
 {
 	int i;
 	struct hfi1_devdata *dd = uctxt->dd;
 	unsigned long *evs;
+	unsigned long events;
 
 	if (!dd->events)
 		return 0;
 
-	evs = dd->events + ((uctxt->ctxt - dd->first_dyn_alloc_ctxt) *
-			    HFI1_MAX_SHARED_CTXTS) + subctxt;
+	if (get_user(events, (unsigned long __user *)arg))
+		return -EFAULT;
+
+	evs = dd->events + uctxt_offset(uctxt) + subctxt;
 
 	for (i = 0; i <= _HFI1_MAX_EVENT_BIT; i++) {
 		if (!test_bit(i, &events))
@@ -1566,26 +1579,89 @@ static int user_event_ack(struct hfi1_ctxtdata *uctxt, u16 subctxt,
 	return 0;
 }
 
-static int set_ctxt_pkey(struct hfi1_ctxtdata *uctxt, u16 subctxt, u16 pkey)
+static int set_ctxt_pkey(struct hfi1_ctxtdata *uctxt, unsigned long arg)
 {
-	int ret = -ENOENT, i, intable = 0;
+	int i;
 	struct hfi1_pportdata *ppd = uctxt->ppd;
 	struct hfi1_devdata *dd = uctxt->dd;
+	u16 pkey;
 
-	if (pkey == LIM_MGMT_P_KEY || pkey == FULL_MGMT_P_KEY) {
-		ret = -EINVAL;
-		goto done;
-	}
+	if (!HFI1_CAP_IS_USET(PKEY_CHECK))
+		return -EPERM;
+
+	if (get_user(pkey, (u16 __user *)arg))
+		return -EFAULT;
+
+	if (pkey == LIM_MGMT_P_KEY || pkey == FULL_MGMT_P_KEY)
+		return -EINVAL;
 
 	for (i = 0; i < ARRAY_SIZE(ppd->pkeys); i++)
-		if (pkey == ppd->pkeys[i]) {
-			intable = 1;
-			break;
+		if (pkey == ppd->pkeys[i])
+			return hfi1_set_ctxt_pkey(dd, uctxt, pkey);
+
+	return -ENOENT;
+}
+
+/**
+ * ctxt_reset - Reset the user context
+ * @uctxt: valid user context
+ */
+static int ctxt_reset(struct hfi1_ctxtdata *uctxt)
+{
+	struct send_context *sc;
+	struct hfi1_devdata *dd;
+	int ret = 0;
+
+	if (!uctxt || !uctxt->dd || !uctxt->sc)
+		return -EINVAL;
+
+	/*
+	 * There is no protection here. User level has to guarantee that
+	 * no one will be writing to the send context while it is being
+	 * re-initialized.  If user level breaks that guarantee, it will
+	 * break it's own context and no one else's.
+	 */
+	dd = uctxt->dd;
+	sc = uctxt->sc;
+
+	/*
+	 * Wait until the interrupt handler has marked the context as
+	 * halted or frozen. Report error if we time out.
+	 */
+	wait_event_interruptible_timeout(
+		sc->halt_wait, (sc->flags & SCF_HALTED),
+		msecs_to_jiffies(SEND_CTXT_HALT_TIMEOUT));
+	if (!(sc->flags & SCF_HALTED))
+		return -ENOLCK;
+
+	/*
+	 * If the send context was halted due to a Freeze, wait until the
+	 * device has been "unfrozen" before resetting the context.
+	 */
+	if (sc->flags & SCF_FROZEN) {
+		wait_event_interruptible_timeout(
+			dd->event_queue,
+			!(READ_ONCE(dd->flags) & HFI1_FROZEN),
+			msecs_to_jiffies(SEND_CTXT_HALT_TIMEOUT));
+		if (dd->flags & HFI1_FROZEN)
+			return -ENOLCK;
+
+		if (dd->flags & HFI1_FORCED_FREEZE)
+			/*
+			 * Don't allow context reset if we are into
+			 * forced freeze
+			 */
+			return -ENODEV;
+
+		sc_disable(sc);
+		ret = sc_enable(sc);
+		hfi1_rcvctrl(dd, HFI1_RCVCTRL_CTXT_ENB, uctxt);
+	} else {
+		ret = sc_restart(sc);
 	}
+	if (!ret)
+		sc_return_credits(sc);
 
-	if (intable)
-		ret = hfi1_set_ctxt_pkey(dd, uctxt, pkey);
-done:
 	return ret;
 }
 

drivers/infiniband/hw/hfi1/hfi.h

@@ -535,6 +535,8 @@ struct rvt_sge_state;
 #define HLS_UP (HLS_UP_INIT | HLS_UP_ARMED | HLS_UP_ACTIVE)
 #define HLS_DOWN ~(HLS_UP)
 
+#define HLS_DEFAULT HLS_DN_POLL
+
 /* use this MTU size if none other is given */
 #define HFI1_DEFAULT_ACTIVE_MTU 10240
 /* use this MTU size as the default maximum */
@@ -616,7 +618,6 @@ struct hfi1_msix_entry {
 	enum irq_type type;
 	int irq;
 	void *arg;
-	char name[MAX_NAME_SIZE];
 	cpumask_t mask;
 	struct irq_affinity_notify notify;
 };
@@ -1109,8 +1110,7 @@ struct hfi1_devdata {
 	u16 rcvegrbufsize_shift;
 	/* both sides of the PCIe link are gen3 capable */
 	u8 link_gen3_capable;
-	/* default link down value (poll/sleep) */
-	u8 link_default;
+	u8 dc_shutdown;
 	/* localbus width (1, 2,4,8,16,32) from config space  */
 	u32 lbus_width;
 	/* localbus speed in MHz */
@@ -1183,7 +1183,6 @@ struct hfi1_devdata {
 
 	/* INTx information */
 	u32 requested_intx_irq;		/* did we request one? */
-	char intx_name[MAX_NAME_SIZE];	/* INTx name */
 
 	/* general interrupt: mask of handled interrupts */
 	u64 gi_mask[CCE_NUM_INT_CSRS];
@@ -1295,7 +1294,6 @@ struct hfi1_devdata {
 	u8 oui1;
 	u8 oui2;
 	u8 oui3;
-	u8 dc_shutdown;
 
 	/* Timer and counter used to detect RcvBufOvflCnt changes */
 	struct timer_list rcverr_timer;
@@ -1373,8 +1371,12 @@ struct hfi1_filedata {
 extern struct list_head hfi1_dev_list;
 extern spinlock_t hfi1_devs_lock;
 struct hfi1_devdata *hfi1_lookup(int unit);
-extern u32 hfi1_cpulist_count;
-extern unsigned long *hfi1_cpulist;
+
+static inline unsigned long uctxt_offset(struct hfi1_ctxtdata *uctxt)
+{
+	return (uctxt->ctxt - uctxt->dd->first_dyn_alloc_ctxt) *
+		HFI1_MAX_SHARED_CTXTS;
+}
 
 int hfi1_init(struct hfi1_devdata *dd, int reinit);
 int hfi1_count_active_units(void);
@@ -1396,6 +1398,8 @@ void hfi1_init_pportdata(struct pci_dev *pdev, struct hfi1_pportdata *ppd,
 void hfi1_free_ctxtdata(struct hfi1_devdata *dd, struct hfi1_ctxtdata *rcd);
 int hfi1_rcd_put(struct hfi1_ctxtdata *rcd);
 void hfi1_rcd_get(struct hfi1_ctxtdata *rcd);
+struct hfi1_ctxtdata *hfi1_rcd_get_by_index_safe(struct hfi1_devdata *dd,
+						 u16 ctxt);
 struct hfi1_ctxtdata *hfi1_rcd_get_by_index(struct hfi1_devdata *dd, u16 ctxt);
 int handle_receive_interrupt(struct hfi1_ctxtdata *rcd, int thread);
 int handle_receive_interrupt_nodma_rtail(struct hfi1_ctxtdata *rcd, int thread);

drivers/infiniband/hw/hfi1/init.c

@@ -123,8 +123,6 @@ MODULE_PARM_DESC(user_credit_return_threshold, "Credit return threshold for user
 static inline u64 encode_rcv_header_entry_size(u16 size);
 
 static struct idr hfi1_unit_table;
-u32 hfi1_cpulist_count;
-unsigned long *hfi1_cpulist;
 
 static int hfi1_create_kctxt(struct hfi1_devdata *dd,
 			     struct hfi1_pportdata *ppd)
@@ -285,6 +283,27 @@ static int allocate_rcd_index(struct hfi1_devdata *dd,
 	return 0;
 }
 
+/**
+ * hfi1_rcd_get_by_index_safe - validate the ctxt index before accessing the
+ * array
+ * @dd: pointer to a valid devdata structure
+ * @ctxt: the index of an possilbe rcd
+ *
+ * This is a wrapper for hfi1_rcd_get_by_index() to validate that the given
+ * ctxt index is valid.
+ *
+ * The caller is responsible for making the _put().
+ *
+ */
+struct hfi1_ctxtdata *hfi1_rcd_get_by_index_safe(struct hfi1_devdata *dd,
+						 u16 ctxt)
+{
+	if (ctxt < dd->num_rcv_contexts)
+		return hfi1_rcd_get_by_index(dd, ctxt);
+
+	return NULL;
+}
+
 /**
  * hfi1_rcd_get_by_index
  * @dd: pointer to a valid devdata structure
@@ -1272,39 +1291,21 @@ struct hfi1_devdata *hfi1_alloc_devdata(struct pci_dev *pdev, size_t extra)
 	dd->int_counter = alloc_percpu(u64);
 	if (!dd->int_counter) {
 		ret = -ENOMEM;
-		hfi1_early_err(&pdev->dev,
-			       "Could not allocate per-cpu int_counter\n");
 		goto bail;
 	}
 
 	dd->rcv_limit = alloc_percpu(u64);
 	if (!dd->rcv_limit) {
 		ret = -ENOMEM;
-		hfi1_early_err(&pdev->dev,
-			       "Could not allocate per-cpu rcv_limit\n");
 		goto bail;
 	}
 
 	dd->send_schedule = alloc_percpu(u64);
 	if (!dd->send_schedule) {
 		ret = -ENOMEM;
-		hfi1_early_err(&pdev->dev,
-			       "Could not allocate per-cpu int_counter\n");
 		goto bail;
 	}
 
-	if (!hfi1_cpulist_count) {
-		u32 count = num_online_cpus();
-
-		hfi1_cpulist = kcalloc(BITS_TO_LONGS(count), sizeof(long),
-				       GFP_KERNEL);
-		if (hfi1_cpulist)
-			hfi1_cpulist_count = count;
-		else
-			hfi1_early_err(
-			&pdev->dev,
-			"Could not alloc cpulist info, cpu affinity might be wrong\n");
-	}
 	kobject_init(&dd->kobj, &hfi1_devdata_type);
 	return dd;
 
@@ -1477,8 +1478,6 @@ static void __exit hfi1_mod_cleanup(void)
 	node_affinity_destroy();
 	hfi1_wss_exit();
 	hfi1_dbg_exit();
-	hfi1_cpulist_count = 0;
-	kfree(hfi1_cpulist);
 
 	idr_destroy(&hfi1_unit_table);
 	dispose_firmware();	/* asymmetric with obtain_firmware() */

drivers/infiniband/hw/hfi1/sdma.c

@@ -1392,6 +1392,13 @@ int sdma_init(struct hfi1_devdata *dd, u8 port)
 		return ret;
 
 	idle_cnt = ns_to_cclock(dd, idle_cnt);
+	if (idle_cnt)
+		dd->default_desc1 =
+			SDMA_DESC1_HEAD_TO_HOST_FLAG;
+	else
+		dd->default_desc1 =
+			SDMA_DESC1_INT_REQ_FLAG;
+
 	if (!sdma_desct_intr)
 		sdma_desct_intr = SDMA_DESC_INTR;
 
@@ -1436,13 +1443,6 @@ int sdma_init(struct hfi1_devdata *dd, u8 port)
 		sde->tail_csr =
 			get_kctxt_csr_addr(dd, this_idx, SD(TAIL));
 
-		if (idle_cnt)
-			dd->default_desc1 =
-				SDMA_DESC1_HEAD_TO_HOST_FLAG;
-		else
-			dd->default_desc1 =
-				SDMA_DESC1_INT_REQ_FLAG;
-
 		tasklet_init(&sde->sdma_hw_clean_up_task, sdma_hw_clean_up_task,
 			     (unsigned long)sde);
 

drivers/infiniband/hw/hfi1/trace.c

 /*
- * Copyright(c) 2015, 2016 Intel Corporation.
+ * Copyright(c) 2015 - 2017 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
@@ -91,12 +91,17 @@ u8 hfi1_trace_opa_hdr_len(struct hfi1_opa_header *opa_hdr)
 		return __get_16b_hdr_len(&opa_hdr->opah);
 }
 
-const char *hfi1_trace_get_packet_str(struct hfi1_packet *packet)
+const char *hfi1_trace_get_packet_l4_str(u8 l4)
 {
-	if (packet->etype != RHF_RCV_TYPE_BYPASS)
-		return "IB";
+	if (l4)
+		return "16B";
+	else
+		return "9B";
+}
 
-	switch (hfi1_16B_get_l2(packet->hdr)) {
+const char *hfi1_trace_get_packet_l2_str(u8 l2)
+{
+	switch (l2) {
 	case 0:
 		return "0";
 	case 1:
@@ -109,14 +114,6 @@ const char *hfi1_trace_get_packet_str(struct hfi1_packet *packet)
 	return "";
 }
 
-const char *hfi1_trace_get_packet_type_str(u8 l4)
-{
-	if (l4)
-		return "16B";
-	else
-		return "9B";
-}
-
 #define IMM_PRN  "imm:%d"
 #define RETH_PRN "reth vaddr:0x%.16llx rkey:0x%.8x dlen:0x%.8x"
 #define AETH_PRN "aeth syn:0x%.2x %s msn:0x%.8x"

drivers/infiniband/hw/hfi1/trace.h

@@ -44,6 +44,16 @@
  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  *
  */
+
+#define packettype_name(etype) { RHF_RCV_TYPE_##etype, #etype }
+#define show_packettype(etype)                  \
+__print_symbolic(etype,                         \
+	packettype_name(EXPECTED),              \
+	packettype_name(EAGER),                 \
+	packettype_name(IB),                    \
+	packettype_name(ERROR),                 \
+	packettype_name(BYPASS))
+
 #include "trace_dbg.h"
 #include "trace_misc.h"
 #include "trace_ctxts.h"

drivers/infiniband/hw/hfi1/trace_ibhdrs.h

 /*
- * Copyright(c) 2015, 2016 Intel Corporation.
+ * Copyright(c) 2015 - 2017 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
@@ -99,8 +99,7 @@ u8 ibhdr_exhdr_len(struct ib_header *hdr);
 const char *parse_everbs_hdrs(struct trace_seq *p, u8 opcode, void *ehdrs);
 u8 hfi1_trace_opa_hdr_len(struct hfi1_opa_header *opah);
 u8 hfi1_trace_packet_hdr_len(struct hfi1_packet *packet);
-const char *hfi1_trace_get_packet_type_str(u8 l4);
-const char *hfi1_trace_get_packet_str(struct hfi1_packet *packet);
+const char *hfi1_trace_get_packet_l4_str(u8 l4);
 void hfi1_trace_parse_9b_bth(struct ib_other_headers *ohdr,
 			     u8 *ack, u8 *becn, u8 *fecn, u8 *mig,
 			     u8 *se, u8 *pad, u8 *opcode, u8 *tver,
@@ -129,6 +128,8 @@ const char *hfi1_trace_fmt_bth(struct trace_seq *p, bool bypass,
 			       u8 se, u8 pad, u8 opcode, const char *opname,
 			       u8 tver, u16 pkey, u32 psn, u32 qpn);
 
+const char *hfi1_trace_get_packet_l2_str(u8 l2);
+
 #define __parse_ib_ehdrs(op, ehdrs) parse_everbs_hdrs(p, op, ehdrs)
 
 #define lrh_name(lrh) { HFI1_##lrh, #lrh }
@@ -136,8 +137,6 @@ const char *hfi1_trace_fmt_bth(struct trace_seq *p, bool bypass,
 __print_symbolic(lrh,                    \
 	lrh_name(LRH_BTH),               \
 	lrh_name(LRH_GRH))
-#define PKT_ENTRY(pkt)	__string(ptype,  hfi1_trace_get_packet_str(packet))
-#define PKT_ASSIGN(pkt) __assign_str(ptype, hfi1_trace_get_packet_str(packet))
 
 DECLARE_EVENT_CLASS(hfi1_input_ibhdr_template,
 		    TP_PROTO(struct hfi1_devdata *dd,
@@ -146,12 +145,12 @@ DECLARE_EVENT_CLASS(hfi1_input_ibhdr_template,
 		    TP_ARGS(dd, packet, sc5),
 		    TP_STRUCT__entry(
 			DD_DEV_ENTRY(dd)
-			PKT_ENTRY(packet)
-			__field(bool, bypass)
+			__field(u8, etype)
 			__field(u8, ack)
 			__field(u8, age)
 			__field(u8, becn)
 			__field(u8, fecn)
+			__field(u8, l2)
 			__field(u8, l4)
 			__field(u8, lnh)
 			__field(u8, lver)
@@ -176,10 +175,10 @@ DECLARE_EVENT_CLASS(hfi1_input_ibhdr_template,
 			),
 		    TP_fast_assign(
 			DD_DEV_ASSIGN(dd);
-			PKT_ASSIGN(packet);
 
-			if (packet->etype == RHF_RCV_TYPE_BYPASS) {
-				__entry->bypass = true;
+			__entry->etype = packet->etype;
+			__entry->l2 = hfi1_16B_get_l2(packet->hdr);
+			if (__entry->etype == RHF_RCV_TYPE_BYPASS) {
 				hfi1_trace_parse_16b_hdr(packet->hdr,
 							 &__entry->age,
 							 &__entry->becn,
@@ -203,7 +202,6 @@ DECLARE_EVENT_CLASS(hfi1_input_ibhdr_template,
 							   &__entry->psn,
 							   &__entry->qpn);
 			} else {
-				__entry->bypass = false;
 				hfi1_trace_parse_9b_hdr(packet->hdr, sc5,
 							&__entry->lnh,
 							&__entry->lver,
@@ -233,9 +231,13 @@ DECLARE_EVENT_CLASS(hfi1_input_ibhdr_template,
 			 ),
 		    TP_printk("[%s] (%s) %s %s hlen:%d %s",
 			      __get_str(dev),
-			      __get_str(ptype),
+			      __entry->etype != RHF_RCV_TYPE_BYPASS ?
+					show_packettype(__entry->etype) :
+					hfi1_trace_get_packet_l2_str(
+						__entry->l2),
 			      hfi1_trace_fmt_lrh(p,
-						 __entry->bypass,
+						 __entry->etype ==
+							RHF_RCV_TYPE_BYPASS,
 						 __entry->age,
 						 __entry->becn,
 						 __entry->fecn,
@@ -252,7 +254,8 @@ DECLARE_EVENT_CLASS(hfi1_input_ibhdr_template,
 						 __entry->dlid,
 						 __entry->slid),
 			      hfi1_trace_fmt_bth(p,
-						 __entry->bypass,
+						 __entry->etype ==
+							RHF_RCV_TYPE_BYPASS,
 						 __entry->ack,
 						 __entry->becn,
 						 __entry->fecn,
@@ -284,7 +287,7 @@ DECLARE_EVENT_CLASS(hfi1_output_ibhdr_template,
 		    TP_ARGS(dd, opah, sc5),
 		    TP_STRUCT__entry(
 			DD_DEV_ENTRY(dd)
-			__field(bool, bypass)
+			__field(u8, hdr_type)
 			__field(u8, ack)
 			__field(u8, age)
 			__field(u8, becn)
@@ -316,8 +319,8 @@ DECLARE_EVENT_CLASS(hfi1_output_ibhdr_template,
 
 			DD_DEV_ASSIGN(dd);
 
-			if (opah->hdr_type)  {
-				__entry->bypass = true;
+			__entry->hdr_type = opah->hdr_type;
+			if (__entry->hdr_type)  {
 				hfi1_trace_parse_16b_hdr(&opah->opah,
 							 &__entry->age,
 							 &__entry->becn,
@@ -345,7 +348,6 @@ DECLARE_EVENT_CLASS(hfi1_output_ibhdr_template,
 							 &__entry->psn,
 							 &__entry->qpn);
 			} else {
-				__entry->bypass = false;
 				hfi1_trace_parse_9b_hdr(&opah->ibh, sc5,
 							&__entry->lnh,
 							&__entry->lver,
@@ -378,9 +380,9 @@ DECLARE_EVENT_CLASS(hfi1_output_ibhdr_template,
 		    ),
 		    TP_printk("[%s] (%s) %s %s hlen:%d %s",
 			      __get_str(dev),
-			      hfi1_trace_get_packet_type_str(__entry->l4),
+			      hfi1_trace_get_packet_l4_str(__entry->l4),
 			      hfi1_trace_fmt_lrh(p,
-						 __entry->bypass,
+						 !!__entry->hdr_type,
 						 __entry->age,
 						 __entry->becn,
 						 __entry->fecn,
@@ -397,7 +399,7 @@ DECLARE_EVENT_CLASS(hfi1_output_ibhdr_template,
 						 __entry->dlid,
 						 __entry->slid),
 			      hfi1_trace_fmt_bth(p,
-						 __entry->bypass,
+						 !!__entry->hdr_type,
 						 __entry->ack,
 						 __entry->becn,
 						 __entry->fecn,

drivers/infiniband/hw/hfi1/trace_rx.h

 /*
- * Copyright(c) 2015, 2016 Intel Corporation.
+ * Copyright(c) 2015 - 2017 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
@@ -62,15 +62,6 @@ __print_symbolic(type,                       \
 #undef TRACE_SYSTEM
 #define TRACE_SYSTEM hfi1_rx
 
-#define packettype_name(etype) { RHF_RCV_TYPE_##etype, #etype }
-#define show_packettype(etype)                  \
-__print_symbolic(etype,                         \
-	packettype_name(EXPECTED),              \
-	packettype_name(EAGER),                 \
-	packettype_name(IB),                    \
-	packettype_name(ERROR),                 \
-	packettype_name(BYPASS))
-
 TRACE_EVENT(hfi1_rcvhdr,
 	    TP_PROTO(struct hfi1_devdata *dd,
 		     u32 ctxt,

drivers/infiniband/hw/hfi1/user_exp_rcv.c

@@ -542,14 +542,10 @@ int hfi1_user_exp_rcv_invalid(struct hfi1_filedata *fd,
 {
 	struct hfi1_ctxtdata *uctxt = fd->uctxt;
 	unsigned long *ev = uctxt->dd->events +
-		(((uctxt->ctxt - uctxt->dd->first_dyn_alloc_ctxt) *
-		  HFI1_MAX_SHARED_CTXTS) + fd->subctxt);
+		(uctxt_offset(uctxt) + fd->subctxt);
 	u32 *array;
 	int ret = 0;
 
-	if (!fd->invalid_tids)
-		return -EINVAL;
-
 	/*
 	 * copy_to_user() can sleep, which will leave the invalid_lock
 	 * locked and cause the MMU notifier to be blocked on the lock
@@ -942,8 +938,7 @@ static int tid_rb_invalidate(void *arg, struct mmu_rb_node *mnode)
 			 * process in question.
 			 */
 			ev = uctxt->dd->events +
-			  (((uctxt->ctxt - uctxt->dd->first_dyn_alloc_ctxt) *
-			    HFI1_MAX_SHARED_CTXTS) + fdata->subctxt);
+				(uctxt_offset(uctxt) + fdata->subctxt);
 			set_bit(_HFI1_EVENT_TID_MMU_NOTIFY_BIT, ev);
 		}
 		fdata->invalid_tid_idx++;

drivers/infiniband/hw/hfi1/user_sdma.c

@@ -956,10 +956,8 @@ static int pin_sdma_pages(struct user_sdma_request *req,
 	struct hfi1_user_sdma_pkt_q *pq = req->pq;
 
 	pages = kcalloc(npages, sizeof(*pages), GFP_KERNEL);
-	if (!pages) {
-		SDMA_DBG(req, "Failed page array alloc");
+	if (!pages)
 		return -ENOMEM;
-	}
 	memcpy(pages, node->pages, node->npages * sizeof(*pages));
 
 	npages -= node->npages;
@@ -1254,20 +1252,25 @@ static int set_txreq_header_ahg(struct user_sdma_request *req,
 				struct user_sdma_txreq *tx, u32 datalen)
 {
 	u32 ahg[AHG_KDETH_ARRAY_SIZE];
-	int diff = 0;
+	int idx = 0;
 	u8 omfactor; /* KDETH.OM */
 	struct hfi1_user_sdma_pkt_q *pq = req->pq;
 	struct hfi1_pkt_header *hdr = &req->hdr;
 	u16 pbclen = le16_to_cpu(hdr->pbc[0]);
 	u32 val32, tidval = 0, lrhlen = get_lrh_len(*hdr, pad_len(datalen));
+	size_t array_size = ARRAY_SIZE(ahg);
 
 	if (PBC2LRH(pbclen) != lrhlen) {
 		/* PBC.PbcLengthDWs */
-		AHG_HEADER_SET(ahg, diff, 0, 0, 12,
-			       cpu_to_le16(LRH2PBC(lrhlen)));
+		idx = ahg_header_set(ahg, idx, array_size, 0, 0, 12,
+				     (__force u16)cpu_to_le16(LRH2PBC(lrhlen)));
+		if (idx < 0)
+			return idx;
 		/* LRH.PktLen (we need the full 16 bits due to byte swap) */
-		AHG_HEADER_SET(ahg, diff, 3, 0, 16,
-			       cpu_to_be16(lrhlen >> 2));
+		idx = ahg_header_set(ahg, idx, array_size, 3, 0, 16,
+				     (__force u16)cpu_to_be16(lrhlen >> 2));
+		if (idx < 0)
+			return idx;
 	}
 
 	/*
@@ -1278,12 +1281,23 @@ static int set_txreq_header_ahg(struct user_sdma_request *req,
 		(HFI1_CAP_IS_KSET(EXTENDED_PSN) ? 0x7fffffff : 0xffffff);
 	if (unlikely(tx->flags & TXREQ_FLAGS_REQ_ACK))
 		val32 |= 1UL << 31;
-	AHG_HEADER_SET(ahg, diff, 6, 0, 16, cpu_to_be16(val32 >> 16));
-	AHG_HEADER_SET(ahg, diff, 6, 16, 16, cpu_to_be16(val32 & 0xffff));
+	idx = ahg_header_set(ahg, idx, array_size, 6, 0, 16,
+			     (__force u16)cpu_to_be16(val32 >> 16));
+	if (idx < 0)
+		return idx;
+	idx = ahg_header_set(ahg, idx, array_size, 6, 16, 16,
+			     (__force u16)cpu_to_be16(val32 & 0xffff));
+	if (idx < 0)
+		return idx;
 	/* KDETH.Offset */
-	AHG_HEADER_SET(ahg, diff, 15, 0, 16,
-		       cpu_to_le16(req->koffset & 0xffff));
-	AHG_HEADER_SET(ahg, diff, 15, 16, 16, cpu_to_le16(req->koffset >> 16));
+	idx = ahg_header_set(ahg, idx, array_size, 15, 0, 16,
+			     (__force u16)cpu_to_le16(req->koffset & 0xffff));
+	if (idx < 0)
+		return idx;
+	idx = ahg_header_set(ahg, idx, array_size, 15, 16, 16,
+			     (__force u16)cpu_to_le16(req->koffset >> 16));
+	if (idx < 0)
+		return idx;
 	if (req_opcode(req->info.ctrl) == EXPECTED) {
 		__le16 val;
 
@@ -1310,10 +1324,13 @@ static int set_txreq_header_ahg(struct user_sdma_request *req,
 				 KDETH_OM_MAX_SIZE) ? KDETH_OM_LARGE_SHIFT :
 				 KDETH_OM_SMALL_SHIFT;
 		/* KDETH.OM and KDETH.OFFSET (TID) */
-		AHG_HEADER_SET(ahg, diff, 7, 0, 16,
+		idx = ahg_header_set(
+				ahg, idx, array_size, 7, 0, 16,
 				((!!(omfactor - KDETH_OM_SMALL_SHIFT)) << 15 |
 				((req->tidoffset >> omfactor)
 				& 0x7fff)));
+		if (idx < 0)
+			return idx;
 		/* KDETH.TIDCtrl, KDETH.TID, KDETH.Intr, KDETH.SH */
 		val = cpu_to_le16(((EXP_TID_GET(tidval, CTRL) & 0x3) << 10) |
 				   (EXP_TID_GET(tidval, IDX) & 0x3ff));
@@ -1330,21 +1347,22 @@ static int set_txreq_header_ahg(struct user_sdma_request *req,
 					     AHG_KDETH_INTR_SHIFT));
 		}
 
-		AHG_HEADER_SET(ahg, diff, 7, 16, 14, val);
+		idx = ahg_header_set(ahg, idx, array_size,
+				     7, 16, 14, (__force u16)val);
+		if (idx < 0)
+			return idx;
 	}
-	if (diff < 0)
-		return diff;
 
 	trace_hfi1_sdma_user_header_ahg(pq->dd, pq->ctxt, pq->subctxt,
 					req->info.comp_idx, req->sde->this_idx,
-					req->ahg_idx, ahg, diff, tidval);
+					req->ahg_idx, ahg, idx, tidval);
 	sdma_txinit_ahg(&tx->txreq,
 			SDMA_TXREQ_F_USE_AHG,
-			datalen, req->ahg_idx, diff,
+			datalen, req->ahg_idx, idx,
 			ahg, sizeof(req->hdr),
 			user_sdma_txreq_cb);
 
-	return diff;
+	return idx;
 }
 
 /*

drivers/infiniband/hw/hfi1/user_sdma.h

@@ -80,15 +80,26 @@
 #define PBC2LRH(x) ((((x) & 0xfff) << 2) - 4)
 #define LRH2PBC(x) ((((x) >> 2) + 1) & 0xfff)
 
-#define AHG_HEADER_SET(arr, idx, dw, bit, width, value)			\
-	do {								\
-		if ((idx) < ARRAY_SIZE((arr)))				\
-			(arr)[(idx++)] = sdma_build_ahg_descriptor(	\
-				(__force u16)(value), (dw), (bit),	\
-							(width));	\
-		else							\
-			return -ERANGE;					\
-	} while (0)
+/**
+ * Build an SDMA AHG header update descriptor and save it to an array.
+ * @arr        - Array to save the descriptor to.
+ * @idx        - Index of the array at which the descriptor will be saved.
+ * @array_size - Size of the array arr.
+ * @dw         - Update index into the header in DWs.
+ * @bit        - Start bit.
+ * @width      - Field width.
+ * @value      - 16 bits of immediate data to write into the field.
+ * Returns -ERANGE if idx is invalid. If successful, returns the next index
+ * (idx + 1) of the array to be used for the next descriptor.
+ */
+static inline int ahg_header_set(u32 *arr, int idx, size_t array_size,
+				 u8 dw, u8 bit, u8 width, u16 value)
+{
+	if ((size_t)idx >= array_size)
+		return -ERANGE;
+	arr[idx++] = sdma_build_ahg_descriptor(value, dw, bit, width);
+	return idx;
+}
 
 /* Tx request flag bits */
 #define TXREQ_FLAGS_REQ_ACK   BIT(0)      /* Set the ACK bit in the header */

drivers/infiniband/hw/qib/qib.h

@@ -443,14 +443,12 @@ struct qib_irq_notify;
 #endif
 
 struct qib_msix_entry {
-	int irq;
 	void *arg;
 #ifdef CONFIG_INFINIBAND_QIB_DCA
 	int dca;
 	int rcv;
 	struct qib_irq_notify *notifier;
 #endif
-	char name[MAX_NAME_SIZE];
 	cpumask_var_t mask;
 };
 
@@ -1434,10 +1432,8 @@ int qib_pcie_ddinit(struct qib_devdata *, struct pci_dev *,
 		    const struct pci_device_id *);
 void qib_pcie_ddcleanup(struct qib_devdata *);
 int qib_pcie_params(struct qib_devdata *dd, u32 minw, u32 *nent);
-int qib_reinit_intr(struct qib_devdata *);
-void qib_enable_intx(struct qib_devdata *dd);
-void qib_nomsi(struct qib_devdata *);
-void qib_nomsix(struct qib_devdata *);
+void qib_free_irq(struct qib_devdata *dd);
+int qib_reinit_intr(struct qib_devdata *dd);
 void qib_pcie_getcmd(struct qib_devdata *, u16 *, u8 *, u8 *);
 void qib_pcie_reenable(struct qib_devdata *, u16, u8, u8);
 /* interrupts for device */

drivers/infiniband/hw/qib/qib_7220.h

@@ -67,7 +67,6 @@ struct qib_chip_specific {
 	u32 lastbuf_for_pio;
 	u32 updthresh; /* current AvailUpdThld */
 	u32 updthresh_dflt; /* default AvailUpdThld */
-	int irq;
 	u8 presets_needed;
 	u8 relock_timer_active;
 	char emsgbuf[128];

drivers/infiniband/hw/qib/qib_iba6120.c

@@ -245,7 +245,6 @@ struct qib_chip_specific {
 	u64 iblnkerrsnap;
 	u64 ibcctrl; /* shadow for kr_ibcctrl */
 	u32 lastlinkrecov; /* link recovery issue */
-	int irq;
 	u32 cntrnamelen;
 	u32 portcntrnamelen;
 	u32 ncntrs;
@@ -1485,15 +1484,6 @@ static void qib_6120_setup_setextled(struct qib_pportdata *ppd, u32 on)
 	spin_unlock_irqrestore(&dd->cspec->gpio_lock, flags);
 }
 
-static void qib_6120_free_irq(struct qib_devdata *dd)
-{
-	if (dd->cspec->irq) {
-		free_irq(dd->cspec->irq, dd);
-		dd->cspec->irq = 0;
-	}
-	qib_nomsi(dd);
-}
-
 /**
  * qib_6120_setup_cleanup - clean up any per-chip chip-specific stuff
  * @dd: the qlogic_ib device
@@ -1502,7 +1492,7 @@ static void qib_6120_free_irq(struct qib_devdata *dd)
 */
 static void qib_6120_setup_cleanup(struct qib_devdata *dd)
 {
-	qib_6120_free_irq(dd);
+	qib_free_irq(dd);
 	kfree(dd->cspec->cntrs);
 	kfree(dd->cspec->portcntrs);
 	if (dd->cspec->dummy_hdrq) {
@@ -1706,6 +1696,8 @@ static irqreturn_t qib_6120intr(int irq, void *data)
  */
 static void qib_setup_6120_interrupt(struct qib_devdata *dd)
 {
+	int ret;
+
 	/*
 	 * If the chip supports added error indication via GPIO pins,
 	 * enable interrupts on those bits so the interrupt routine
@@ -1719,19 +1711,12 @@ static void qib_setup_6120_interrupt(struct qib_devdata *dd)
 		qib_write_kreg(dd, kr_gpio_mask, dd->cspec->gpio_mask);
 	}
 
-	if (!dd->cspec->irq)
-		qib_dev_err(dd,
-			"irq is 0, BIOS error?  Interrupts won't work\n");
-	else {
-		int ret;
-
-		ret = request_irq(dd->cspec->irq, qib_6120intr, 0,
-				  QIB_DRV_NAME, dd);
+	ret = pci_request_irq(dd->pcidev, 0, qib_6120intr, NULL, dd,
+			      QIB_DRV_NAME);
 	if (ret)
 		qib_dev_err(dd,
 			    "Couldn't setup interrupt (irq=%d): %d\n",
-				dd->cspec->irq, ret);
-	}
+			    pci_irq_vector(dd->pcidev, 0), ret);
 }
 
 /**
@@ -3490,7 +3475,7 @@ struct qib_devdata *qib_init_iba6120_funcs(struct pci_dev *pdev,
 	dd->f_bringup_serdes    = qib_6120_bringup_serdes;
 	dd->f_cleanup           = qib_6120_setup_cleanup;
 	dd->f_clear_tids        = qib_6120_clear_tids;
-	dd->f_free_irq          = qib_6120_free_irq;
+	dd->f_free_irq          = qib_free_irq;
 	dd->f_get_base_info     = qib_6120_get_base_info;
 	dd->f_get_msgheader     = qib_6120_get_msgheader;
 	dd->f_getsendbuf        = qib_6120_getsendbuf;
@@ -3559,8 +3544,6 @@ struct qib_devdata *qib_init_iba6120_funcs(struct pci_dev *pdev,
 	if (qib_pcie_params(dd, 8, NULL))
 		qib_dev_err(dd,
 			"Failed to setup PCIe or interrupts; continuing anyway\n");
-	dd->cspec->irq = pdev->irq; /* save IRQ */
-
 	/* clear diagctrl register, in case diags were running and crashed */
 	qib_write_kreg(dd, kr_hwdiagctrl, 0);
 

drivers/infiniband/hw/qib/qib_iba7220.c

@@ -1780,15 +1780,6 @@ static void qib_setup_7220_setextled(struct qib_pportdata *ppd, u32 on)
 		qib_write_kreg(dd, kr_rcvpktledcnt, ledblink);
 }
 
-static void qib_7220_free_irq(struct qib_devdata *dd)
-{
-	if (dd->cspec->irq) {
-		free_irq(dd->cspec->irq, dd);
-		dd->cspec->irq = 0;
-	}
-	qib_nomsi(dd);
-}
-
 /*
  * qib_setup_7220_cleanup - clean up any per-chip chip-specific stuff
  * @dd: the qlogic_ib device
@@ -1798,7 +1789,7 @@ static void qib_7220_free_irq(struct qib_devdata *dd)
  */
 static void qib_setup_7220_cleanup(struct qib_devdata *dd)
 {
-	qib_7220_free_irq(dd);
+	qib_free_irq(dd);
 	kfree(dd->cspec->cntrs);
 	kfree(dd->cspec->portcntrs);
 }
@@ -2026,20 +2017,14 @@ static irqreturn_t qib_7220intr(int irq, void *data)
  */
 static void qib_setup_7220_interrupt(struct qib_devdata *dd)
 {
-	if (!dd->cspec->irq)
-		qib_dev_err(dd,
-			"irq is 0, BIOS error?  Interrupts won't work\n");
-	else {
-		int ret = request_irq(dd->cspec->irq, qib_7220intr,
-			dd->msi_lo ? 0 : IRQF_SHARED,
-			QIB_DRV_NAME, dd);
+	int ret;
 
+	ret = pci_request_irq(dd->pcidev, 0, qib_7220intr, NULL, dd,
+			      QIB_DRV_NAME);
 	if (ret)
-			qib_dev_err(dd,
-				"Couldn't setup %s interrupt (irq=%d): %d\n",
-				dd->msi_lo ?  "MSI" : "INTx",
-				dd->cspec->irq, ret);
-	}
+		qib_dev_err(dd, "Couldn't setup %s interrupt (irq=%d): %d\n",
+			    dd->pcidev->msi_enabled ?  "MSI" : "INTx",
+			    pci_irq_vector(dd->pcidev, 0), ret);
 }
 
 /**
@@ -3303,15 +3288,11 @@ static int qib_7220_intr_fallback(struct qib_devdata *dd)
 
 	qib_devinfo(dd->pcidev,
 		    "MSI interrupt not detected, trying INTx interrupts\n");
-	qib_7220_free_irq(dd);
-	qib_enable_intx(dd);
-	/*
-	 * Some newer kernels require free_irq before disable_msi,
-	 * and irq can be changed during disable and INTx enable
-	 * and we need to therefore use the pcidev->irq value,
-	 * not our saved MSI value.
-	 */
-	dd->cspec->irq = dd->pcidev->irq;
+
+	qib_free_irq(dd);
+	dd->msi_lo = 0;
+	if (pci_alloc_irq_vectors(dd->pcidev, 1, 1, PCI_IRQ_LEGACY) < 0)
+		qib_dev_err(dd, "Failed to enable INTx\n");
 	qib_setup_7220_interrupt(dd);
 	return 1;
 }
@@ -4535,7 +4516,7 @@ struct qib_devdata *qib_init_iba7220_funcs(struct pci_dev *pdev,
 	dd->f_bringup_serdes    = qib_7220_bringup_serdes;
 	dd->f_cleanup           = qib_setup_7220_cleanup;
 	dd->f_clear_tids        = qib_7220_clear_tids;
-	dd->f_free_irq          = qib_7220_free_irq;
+	dd->f_free_irq          = qib_free_irq;
 	dd->f_get_base_info     = qib_7220_get_base_info;
 	dd->f_get_msgheader     = qib_7220_get_msgheader;
 	dd->f_getsendbuf        = qib_7220_getsendbuf;
@@ -4618,9 +4599,6 @@ struct qib_devdata *qib_init_iba7220_funcs(struct pci_dev *pdev,
 		qib_dev_err(dd,
 			"Failed to setup PCIe or interrupts; continuing anyway\n");
 
-	/* save IRQ for possible later use */
-	dd->cspec->irq = pdev->irq;
-
 	if (qib_read_kreg64(dd, kr_hwerrstatus) &
 	    QLOGIC_IB_HWE_SERDESPLLFAILED)
 		qib_write_kreg(dd, kr_hwerrclear,

drivers/infiniband/hw/qib/qib_iba7322.c

@@ -553,7 +553,6 @@ struct qib_chip_specific {
 	u32 updthresh; /* current AvailUpdThld */
 	u32 updthresh_dflt; /* default AvailUpdThld */
 	u32 r1;
-	int irq;
 	u32 num_msix_entries;
 	u32 sdmabufcnt;
 	u32 lastbuf_for_pio;
@@ -756,10 +755,8 @@ static void check_7322_rxe_status(struct qib_pportdata *);
 static u32 __iomem *qib_7322_getsendbuf(struct qib_pportdata *, u64, u32 *);
 #ifdef CONFIG_INFINIBAND_QIB_DCA
 static void qib_setup_dca(struct qib_devdata *dd);
-static void setup_dca_notifier(struct qib_devdata *dd,
-			       struct qib_msix_entry *m);
-static void reset_dca_notifier(struct qib_devdata *dd,
-			       struct qib_msix_entry *m);
+static void setup_dca_notifier(struct qib_devdata *dd, int msixnum);
+static void reset_dca_notifier(struct qib_devdata *dd, int msixnum);
 #endif
 
 /**
@@ -2778,7 +2775,7 @@ static void qib_setup_dca(struct qib_devdata *dd)
 		qib_write_kreg(dd, KREG_IDX(DCACtrlB) + i,
 			       cspec->dca_rcvhdr_ctrl[i]);
 	for (i = 0; i < cspec->num_msix_entries; i++)
-		setup_dca_notifier(dd, &cspec->msix_entries[i]);
+		setup_dca_notifier(dd, i);
 }
 
 static void qib_irq_notifier_notify(struct irq_affinity_notify *notify,
@@ -2820,49 +2817,41 @@ static void qib_irq_notifier_release(struct kref *ref)
 }
 #endif
 
-/*
- * Disable MSIx interrupt if enabled, call generic MSIx code
- * to cleanup, and clear pending MSIx interrupts.
- * Used for fallback to INTx, after reset, and when MSIx setup fails.
- */
-static void qib_7322_nomsix(struct qib_devdata *dd)
+static void qib_7322_free_irq(struct qib_devdata *dd)
 {
 	u64 intgranted;
-	int n;
+	int i;
 
 	dd->cspec->main_int_mask = ~0ULL;
-	n = dd->cspec->num_msix_entries;
-	if (n) {
-		int i;
 
-		dd->cspec->num_msix_entries = 0;
-		for (i = 0; i < n; i++) {
+	for (i = 0; i < dd->cspec->num_msix_entries; i++) {
+		/* only free IRQs that were allocated */
+		if (dd->cspec->msix_entries[i].arg) {
 #ifdef CONFIG_INFINIBAND_QIB_DCA
-			reset_dca_notifier(dd, &dd->cspec->msix_entries[i]);
+			reset_dca_notifier(dd, i);
 #endif
-			irq_set_affinity_hint(
-				dd->cspec->msix_entries[i].irq, NULL);
+			irq_set_affinity_hint(pci_irq_vector(dd->pcidev, i),
+					      NULL);
 			free_cpumask_var(dd->cspec->msix_entries[i].mask);
-			free_irq(dd->cspec->msix_entries[i].irq,
+			pci_free_irq(dd->pcidev, i,
 				     dd->cspec->msix_entries[i].arg);
 		}
-		qib_nomsix(dd);
 	}
+
+	/* If num_msix_entries was 0, disable the INTx IRQ */
+	if (!dd->cspec->num_msix_entries)
+		pci_free_irq(dd->pcidev, 0, dd);
+	else
+		dd->cspec->num_msix_entries = 0;
+
+	pci_free_irq_vectors(dd->pcidev);
+
 	/* make sure no MSIx interrupts are left pending */
 	intgranted = qib_read_kreg64(dd, kr_intgranted);
 	if (intgranted)
 		qib_write_kreg(dd, kr_intgranted, intgranted);
 }
 
-static void qib_7322_free_irq(struct qib_devdata *dd)
-{
-	if (dd->cspec->irq) {
-		free_irq(dd->cspec->irq, dd);
-		dd->cspec->irq = 0;
-	}
-	qib_7322_nomsix(dd);
-}
-
 static void qib_setup_7322_cleanup(struct qib_devdata *dd)
 {
 	int i;
@@ -3329,22 +3318,20 @@ static irqreturn_t sdma_cleanup_intr(int irq, void *data)
 
 #ifdef CONFIG_INFINIBAND_QIB_DCA
 
-static void reset_dca_notifier(struct qib_devdata *dd, struct qib_msix_entry *m)
+static void reset_dca_notifier(struct qib_devdata *dd, int msixnum)
 {
-	if (!m->dca)
+	if (!dd->cspec->msix_entries[msixnum].dca)
 		return;
-	qib_devinfo(dd->pcidev,
-		"Disabling notifier on HCA %d irq %d\n",
-		dd->unit,
-		m->irq);
-	irq_set_affinity_notifier(
-		m->irq,
-		NULL);
-	m->notifier = NULL;
+
+	qib_devinfo(dd->pcidev, "Disabling notifier on HCA %d irq %d\n",
+		    dd->unit, pci_irq_vector(dd->pcidev, msixnum));
+	irq_set_affinity_notifier(pci_irq_vector(dd->pcidev, msixnum), NULL);
+	dd->cspec->msix_entries[msixnum].notifier = NULL;
 }
 
-static void setup_dca_notifier(struct qib_devdata *dd, struct qib_msix_entry *m)
+static void setup_dca_notifier(struct qib_devdata *dd, int msixnum)
 {
+	struct qib_msix_entry *m = &dd->cspec->msix_entries[msixnum];
 	struct qib_irq_notify *n;
 
 	if (!m->dca)
@@ -3354,7 +3341,7 @@ static void setup_dca_notifier(struct qib_devdata *dd, struct qib_msix_entry *m)
 		int ret;
 
 		m->notifier = n;
-		n->notify.irq = m->irq;
+		n->notify.irq = pci_irq_vector(dd->pcidev, msixnum);
 		n->notify.notify = qib_irq_notifier_notify;
 		n->notify.release = qib_irq_notifier_release;
 		n->arg = m->arg;
@@ -3415,22 +3402,17 @@ static void qib_setup_7322_interrupt(struct qib_devdata *dd, int clearpend)
 	if (!dd->cspec->num_msix_entries) {
 		/* Try to get INTx interrupt */
 try_intx:
-		if (!dd->pcidev->irq) {
-			qib_dev_err(dd,
-				"irq is 0, BIOS error?  Interrupts won't work\n");
-			goto bail;
-		}
-		ret = request_irq(dd->pcidev->irq, qib_7322intr,
-				  IRQF_SHARED, QIB_DRV_NAME, dd);
+		ret = pci_request_irq(dd->pcidev, 0, qib_7322intr, NULL, dd,
+				      QIB_DRV_NAME);
 		if (ret) {
-			qib_dev_err(dd,
+			qib_dev_err(
+				dd,
 				"Couldn't setup INTx interrupt (irq=%d): %d\n",
-				dd->pcidev->irq, ret);
-			goto bail;
+				pci_irq_vector(dd->pcidev, 0), ret);
+			return;
 		}
-		dd->cspec->irq = dd->pcidev->irq;
 		dd->cspec->main_int_mask = ~0ULL;
-		goto bail;
+		return;
 	}
 
 	/* Try to get MSIx interrupts */
@@ -3458,10 +3440,6 @@ static void qib_setup_7322_interrupt(struct qib_devdata *dd, int clearpend)
 #ifdef CONFIG_INFINIBAND_QIB_DCA
 		int dca = 0;
 #endif
-
-		dd->cspec->msix_entries[msixnum].
-			name[sizeof(dd->cspec->msix_entries[msixnum].name) - 1]
-			= '\0';
 		if (i < ARRAY_SIZE(irq_table)) {
 			if (irq_table[i].port) {
 				/* skip if for a non-configured port */
@@ -3475,10 +3453,9 @@ static void qib_setup_7322_interrupt(struct qib_devdata *dd, int clearpend)
 #endif
 			lsb = irq_table[i].lsb;
 			handler = irq_table[i].handler;
-			snprintf(dd->cspec->msix_entries[msixnum].name,
-				sizeof(dd->cspec->msix_entries[msixnum].name)
-				 - 1,
-				QIB_DRV_NAME "%d%s", dd->unit,
+			ret = pci_request_irq(dd->pcidev, msixnum, handler,
+					      NULL, arg, QIB_DRV_NAME "%d%s",
+					      dd->unit,
 					      irq_table[i].name);
 		} else {
 			unsigned ctxt;
@@ -3495,26 +3472,12 @@ static void qib_setup_7322_interrupt(struct qib_devdata *dd, int clearpend)
 #endif
 			lsb = QIB_I_RCVAVAIL_LSB + ctxt;
 			handler = qib_7322pintr;
-			snprintf(dd->cspec->msix_entries[msixnum].name,
-				sizeof(dd->cspec->msix_entries[msixnum].name)
-				 - 1,
-				QIB_DRV_NAME "%d (kctx)", dd->unit);
+			ret = pci_request_irq(dd->pcidev, msixnum, handler,
+					      NULL, arg,
+					      QIB_DRV_NAME "%d (kctx)",
+					      dd->unit);
 		}
 
-		dd->cspec->msix_entries[msixnum].irq = pci_irq_vector(
-			dd->pcidev, msixnum);
-		if (dd->cspec->msix_entries[msixnum].irq < 0) {
-			qib_dev_err(dd,
-				    "Couldn't get MSIx irq (vec=%d): %d\n",
-				    msixnum,
-				    dd->cspec->msix_entries[msixnum].irq);
-			qib_7322_nomsix(dd);
-			goto try_intx;
-		}
-		ret = request_irq(dd->cspec->msix_entries[msixnum].irq,
-				  handler, 0,
-				  dd->cspec->msix_entries[msixnum].name,
-				  arg);
 		if (ret) {
 			/*
 			 * Shouldn't happen since the enable said we could
@@ -3523,9 +3486,11 @@ static void qib_setup_7322_interrupt(struct qib_devdata *dd, int clearpend)
 			qib_dev_err(dd,
 				    "Couldn't setup MSIx interrupt (vec=%d, irq=%d): %d\n",
 				    msixnum,
-				dd->cspec->msix_entries[msixnum].irq,
+				    pci_irq_vector(dd->pcidev, msixnum),
 				    ret);
-			qib_7322_nomsix(dd);
+			qib_7322_free_irq(dd);
+			pci_alloc_irq_vectors(dd->pcidev, 1, 1,
+					      PCI_IRQ_LEGACY);
 			goto try_intx;
 		}
 		dd->cspec->msix_entries[msixnum].arg = arg;
@@ -3559,7 +3524,7 @@ static void qib_setup_7322_interrupt(struct qib_devdata *dd, int clearpend)
 					dd->cspec->msix_entries[msixnum].mask);
 			}
 			irq_set_affinity_hint(
-				dd->cspec->msix_entries[msixnum].irq,
+				pci_irq_vector(dd->pcidev, msixnum),
 				dd->cspec->msix_entries[msixnum].mask);
 		}
 		msixnum++;
@@ -3570,7 +3535,6 @@ static void qib_setup_7322_interrupt(struct qib_devdata *dd, int clearpend)
 	dd->cspec->main_int_mask = mask;
 	tasklet_init(&dd->error_tasklet, qib_error_tasklet,
 		(unsigned long)dd);
-bail:;
 }
 
 /**
@@ -3674,8 +3638,9 @@ static int qib_do_7322_reset(struct qib_devdata *dd)
 	/* no interrupts till re-initted */
 	qib_7322_set_intr_state(dd, 0);
 
+	qib_7322_free_irq(dd);
+
 	if (msix_entries) {
-		qib_7322_nomsix(dd);
 		/* can be up to 512 bytes, too big for stack */
 		msix_vecsave = kmalloc(2 * dd->cspec->num_msix_entries *
 			sizeof(u64), GFP_KERNEL);
@@ -3765,11 +3730,11 @@ static int qib_do_7322_reset(struct qib_devdata *dd)
 		write_7322_init_portregs(&dd->pport[i]);
 	write_7322_initregs(dd);
 
-	if (qib_pcie_params(dd, dd->lbus_width,
-			    &dd->cspec->num_msix_entries))
+	if (qib_pcie_params(dd, dd->lbus_width, &msix_entries))
 		qib_dev_err(dd,
 			"Reset failed to setup PCIe or interrupts; continuing anyway\n");
 
+	dd->cspec->num_msix_entries = msix_entries;
 	qib_setup_7322_interrupt(dd, 1);
 
 	for (i = 0; i < dd->num_pports; ++i) {
@@ -5197,8 +5162,9 @@ static int qib_7322_intr_fallback(struct qib_devdata *dd)
 
 	qib_devinfo(dd->pcidev,
 		"MSIx interrupt not detected, trying INTx interrupts\n");
-	qib_7322_nomsix(dd);
-	qib_enable_intx(dd);
+	qib_7322_free_irq(dd);
+	if (pci_alloc_irq_vectors(dd->pcidev, 1, 1, PCI_IRQ_LEGACY) < 0)
+		qib_dev_err(dd, "Failed to enable INTx\n");
 	qib_setup_7322_interrupt(dd, 0);
 	return 1;
 }

drivers/infiniband/hw/qib/qib_pcie.c

@@ -193,7 +193,7 @@ void qib_pcie_ddcleanup(struct qib_devdata *dd)
  * chip reset (the kernel PCI infrastructure doesn't yet handle that
  * correctly.
  */
-static void qib_msi_setup(struct qib_devdata *dd, int pos)
+static void qib_cache_msi_info(struct qib_devdata *dd, int pos)
 {
 	struct pci_dev *pdev = dd->pcidev;
 	u16 control;
@@ -208,64 +208,39 @@ static void qib_msi_setup(struct qib_devdata *dd, int pos)
 			     &dd->msi_data);
 }
 
-static int qib_allocate_irqs(struct qib_devdata *dd, u32 maxvec)
-{
-	unsigned int flags = PCI_IRQ_LEGACY;
-
-	/* Check our capabilities */
-	if (dd->pcidev->msix_cap) {
-		flags |= PCI_IRQ_MSIX;
-	} else {
-		if (dd->pcidev->msi_cap) {
-			flags |= PCI_IRQ_MSI;
-			/* Get msi_lo and msi_hi */
-			qib_msi_setup(dd, dd->pcidev->msi_cap);
-		}
-	}
-
-	if (!(flags & (PCI_IRQ_MSIX | PCI_IRQ_MSI)))
-		qib_dev_err(dd, "No PCI MSI or MSIx capability!\n");
-
-	return pci_alloc_irq_vectors(dd->pcidev, 1, maxvec, flags);
-}
-
 int qib_pcie_params(struct qib_devdata *dd, u32 minw, u32 *nent)
 {
 	u16 linkstat, speed;
 	int nvec;
 	int maxvec;
-	int ret = 0;
+	unsigned int flags = PCI_IRQ_MSIX | PCI_IRQ_MSI;
 
 	if (!pci_is_pcie(dd->pcidev)) {
 		qib_dev_err(dd, "Can't find PCI Express capability!\n");
 		/* set up something... */
 		dd->lbus_width = 1;
 		dd->lbus_speed = 2500; /* Gen1, 2.5GHz */
-		ret = -1;
+		nvec = -1;
 		goto bail;
 	}
 
+	if (dd->flags & QIB_HAS_INTX)
+		flags |= PCI_IRQ_LEGACY;
 	maxvec = (nent && *nent) ? *nent : 1;
-	nvec = qib_allocate_irqs(dd, maxvec);
-	if (nvec < 0) {
-		ret = nvec;
+	nvec = pci_alloc_irq_vectors(dd->pcidev, 1, maxvec, flags);
+	if (nvec < 0)
 		goto bail;
-	}
 
 	/*
-	 * If nent exists, make sure to record how many vectors were allocated
+	 * If nent exists, make sure to record how many vectors were allocated.
+	 * If msix_enabled is false, return 0 so the fallback code works
+	 * correctly.
 	 */
-	if (nent) {
-		*nent = nvec;
+	if (nent)
+		*nent = !dd->pcidev->msix_enabled ? 0 : nvec;
 
-		/*
-		 * If we requested (nent) MSIX, but msix_enabled is not set,
-		 * pci_alloc_irq_vectors() enabled INTx.
-		 */
-		if (!dd->pcidev->msix_enabled)
-			qib_dev_err(dd,
-				    "no msix vectors allocated, using INTx\n");
-	}
+	if (dd->pcidev->msi_enabled)
+		qib_cache_msi_info(dd, dd->pcidev->msi_cap);
 
 	pcie_capability_read_word(dd->pcidev, PCI_EXP_LNKSTA, &linkstat);
 	/*
@@ -306,7 +281,21 @@ int qib_pcie_params(struct qib_devdata *dd, u32 minw, u32 *nent)
 	/* fill in string, even on errors */
 	snprintf(dd->lbus_info, sizeof(dd->lbus_info),
 		 "PCIe,%uMHz,x%u\n", dd->lbus_speed, dd->lbus_width);
-	return ret;
+	return nvec < 0 ? nvec : 0;
+}
+
+/**
+ * qib_free_irq - Cleanup INTx and MSI interrupts
+ * @dd: valid pointer to qib dev data
+ *
+ * Since cleanup for INTx and MSI interrupts is trivial, have a common
+ * routine.
+ *
+ */
+void qib_free_irq(struct qib_devdata *dd)
+{
+	pci_free_irq(dd->pcidev, 0, dd);
+	pci_free_irq_vectors(dd->pcidev);
 }
 
 /*
@@ -351,10 +340,10 @@ int qib_reinit_intr(struct qib_devdata *dd)
 			      dd->msi_data);
 	ret = 1;
 bail:
-	if (!ret && (dd->flags & QIB_HAS_INTX)) {
-		qib_enable_intx(dd);
+	qib_free_irq(dd);
+
+	if (!ret && (dd->flags & QIB_HAS_INTX))
 		ret = 1;
-	}
 
 	/* and now set the pci master bit again */
 	pci_set_master(dd->pcidev);
@@ -362,56 +351,6 @@ int qib_reinit_intr(struct qib_devdata *dd)
 	return ret;
 }
 
-/*
- * Disable msi interrupt if enabled, and clear msi_lo.
- * This is used primarily for the fallback to INTx, but
- * is also used in reinit after reset, and during cleanup.
- */
-void qib_nomsi(struct qib_devdata *dd)
-{
-	dd->msi_lo = 0;
-	pci_free_irq_vectors(dd->pcidev);
-}
-
-/*
- * Same as qib_nosmi, but for MSIx.
- */
-void qib_nomsix(struct qib_devdata *dd)
-{
-	pci_free_irq_vectors(dd->pcidev);
-}
-
-/*
- * Similar to pci_intx(pdev, 1), except that we make sure
- * msi(x) is off.
- */
-void qib_enable_intx(struct qib_devdata *dd)
-{
-	u16 cw, new;
-	int pos;
-	struct pci_dev *pdev = dd->pcidev;
-
-	if (pci_alloc_irq_vectors(pdev, 1, 1, PCI_IRQ_LEGACY) < 0)
-		qib_dev_err(dd,	"Failed to enable INTx\n");
-
-	pos = pdev->msi_cap;
-	if (pos) {
-		/* then turn off MSI */
-		pci_read_config_word(pdev, pos + PCI_MSI_FLAGS, &cw);
-		new = cw & ~PCI_MSI_FLAGS_ENABLE;
-		if (new != cw)
-			pci_write_config_word(pdev, pos + PCI_MSI_FLAGS, new);
-	}
-	pos = pdev->msix_cap;
-	if (pos) {
-		/* then turn off MSIx */
-		pci_read_config_word(pdev, pos + PCI_MSIX_FLAGS, &cw);
-		new = cw & ~PCI_MSIX_FLAGS_ENABLE;
-		if (new != cw)
-			pci_write_config_word(pdev, pos + PCI_MSIX_FLAGS, new);
-	}
-}
-
 /*
  * These two routines are helper routines for the device reset code
  * to move all the pcie code out of the chip-specific driver code.