ipath_driver.c

来自「linux 内核源代码」· C语言 代码 · 共 2,286 行 · 第 1/5 页

			ipath_dbg("No 64bit DMA mask, used 32 bit mask\n");
			ret = pci_set_consistent_dma_mask(pdev, DMA_32BIT_MASK);
			if (ret)
				dev_info(&pdev->dev,
					"Unable to set DMA consistent mask "
					"for unit %u: %d\n",
					dd->ipath_unit, ret);

		}
	}
	else {
		ret = pci_set_consistent_dma_mask(pdev, DMA_64BIT_MASK);
		if (ret)
			dev_info(&pdev->dev,
				"Unable to set DMA consistent mask "
				"for unit %u: %d\n",
				dd->ipath_unit, ret);
	}

	pci_set_master(pdev);

	/*
	 * Save BARs to rewrite after device reset.  Save all 64 bits of
	 * BAR, just in case.
	 */
	dd->ipath_pcibar0 = addr;
	dd->ipath_pcibar1 = addr >> 32;
	dd->ipath_deviceid = ent->device;	/* save for later use */
	dd->ipath_vendorid = ent->vendor;

	/* setup the chip-specific functions, as early as possible. */
	switch (ent->device) {
	case PCI_DEVICE_ID_INFINIPATH_HT:
#ifdef CONFIG_HT_IRQ
		ipath_init_iba6110_funcs(dd);
		break;
#else
		ipath_dev_err(dd, "QLogic HT device 0x%x cannot work if "
			      "CONFIG_HT_IRQ is not enabled\n", ent->device);
		return -ENODEV;
#endif
	case PCI_DEVICE_ID_INFINIPATH_PE800:
#ifdef CONFIG_PCI_MSI
		ipath_init_iba6120_funcs(dd);
		break;
#else
		ipath_dev_err(dd, "QLogic PCIE device 0x%x cannot work if "
			      "CONFIG_PCI_MSI is not enabled\n", ent->device);
		return -ENODEV;
#endif
	default:
		ipath_dev_err(dd, "Found unknown QLogic deviceid 0x%x, "
			      "failing\n", ent->device);
		return -ENODEV;
	}

	for (j = 0; j < 6; j++) {
		if (!pdev->resource[j].start)
			continue;
		ipath_cdbg(VERBOSE, "BAR %d start %llx, end %llx, len %llx\n",
			   j, (unsigned long long)pdev->resource[j].start,
			   (unsigned long long)pdev->resource[j].end,
			   (unsigned long long)pci_resource_len(pdev, j));
	}

	if (!addr) {
		ipath_dev_err(dd, "No valid address in BAR 0!\n");
		ret = -ENODEV;
		goto bail_regions;
	}

	dd->ipath_pcirev = pdev->revision;

#if defined(__powerpc__)
	/* There isn't a generic way to specify writethrough mappings */
	dd->ipath_kregbase = __ioremap(addr, len,
		(_PAGE_NO_CACHE|_PAGE_WRITETHRU));
#else
	dd->ipath_kregbase = ioremap_nocache(addr, len);
#endif

	if (!dd->ipath_kregbase) {
		ipath_dbg("Unable to map io addr %llx to kvirt, failing\n",
			  addr);
		ret = -ENOMEM;
		goto bail_iounmap;
	}
	dd->ipath_kregend = (u64 __iomem *)
		((void __iomem *)dd->ipath_kregbase + len);
	dd->ipath_physaddr = addr;	/* used for io_remap, etc. */
	/* for user mmap */
	ipath_cdbg(VERBOSE, "mapped io addr %llx to kregbase %p\n",
		   addr, dd->ipath_kregbase);

	/*
	 * clear ipath_flags here instead of in ipath_init_chip as it is set
	 * by ipath_setup_htconfig.
	 */
	dd->ipath_flags = 0;
	dd->ipath_lli_counter = 0;
	dd->ipath_lli_errors = 0;

	if (dd->ipath_f_bus(dd, pdev))
		ipath_dev_err(dd, "Failed to setup config space; "
			      "continuing anyway\n");

	/*
	 * set up our interrupt handler; IRQF_SHARED probably not needed,
	 * since MSI interrupts shouldn't be shared but won't  hurt for now.
	 * check 0 irq after we return from chip-specific bus setup, since
	 * that can affect this due to setup
	 */
	if (!dd->ipath_irq)
		ipath_dev_err(dd, "irq is 0, BIOS error?  Interrupts won't "
			      "work\n");
	else {
		ret = request_irq(dd->ipath_irq, ipath_intr, IRQF_SHARED,
				  IPATH_DRV_NAME, dd);
		if (ret) {
			ipath_dev_err(dd, "Couldn't setup irq handler, "
				      "irq=%d: %d\n", dd->ipath_irq, ret);
			goto bail_iounmap;
		}
	}

	ret = ipath_init_chip(dd, 0);	/* do the chip-specific init */
	if (ret)
		goto bail_irqsetup;

	ret = ipath_enable_wc(dd);

	if (ret) {
		ipath_dev_err(dd, "Write combining not enabled "
			      "(err %d): performance may be poor\n",
			      -ret);
		ret = 0;
	}

	ipath_verify_pioperf(dd);

	ipath_device_create_group(&pdev->dev, dd);
	ipathfs_add_device(dd);
	ipath_user_add(dd);
	ipath_diag_add(dd);
	ipath_register_ib_device(dd);

	/* Check that card status in STATUS_TIMEOUT seconds. */
	schedule_delayed_work(&dd->status_work, HZ * STATUS_TIMEOUT);

	goto bail;

bail_irqsetup:
	if (pdev->irq) free_irq(pdev->irq, dd);

bail_iounmap:
	iounmap((volatile void __iomem *) dd->ipath_kregbase);

bail_regions:
	pci_release_regions(pdev);

bail_disable:
	pci_disable_device(pdev);

bail_devdata:
	ipath_free_devdata(pdev, dd);

bail:
	return ret;
}

static void __devexit cleanup_device(struct ipath_devdata *dd)
{
	int port;

	if (*dd->ipath_statusp & IPATH_STATUS_CHIP_PRESENT) {
		/* can't do anything more with chip; needs re-init */
		*dd->ipath_statusp &= ~IPATH_STATUS_CHIP_PRESENT;
		if (dd->ipath_kregbase) {
			/*
			 * if we haven't already cleaned up before these are
			 * to ensure any register reads/writes "fail" until
			 * re-init
			 */
			dd->ipath_kregbase = NULL;
			dd->ipath_uregbase = 0;
			dd->ipath_sregbase = 0;
			dd->ipath_cregbase = 0;
			dd->ipath_kregsize = 0;
		}
		ipath_disable_wc(dd);
	}

	if (dd->ipath_pioavailregs_dma) {
		dma_free_coherent(&dd->pcidev->dev, PAGE_SIZE,
				  (void *) dd->ipath_pioavailregs_dma,
				  dd->ipath_pioavailregs_phys);
		dd->ipath_pioavailregs_dma = NULL;
	}
	if (dd->ipath_dummy_hdrq) {
		dma_free_coherent(&dd->pcidev->dev,
			dd->ipath_pd[0]->port_rcvhdrq_size,
			dd->ipath_dummy_hdrq, dd->ipath_dummy_hdrq_phys);
		dd->ipath_dummy_hdrq = NULL;
	}

	if (dd->ipath_pageshadow) {
		struct page **tmpp = dd->ipath_pageshadow;
		dma_addr_t *tmpd = dd->ipath_physshadow;
		int i, cnt = 0;

		ipath_cdbg(VERBOSE, "Unlocking any expTID pages still "
			   "locked\n");
		for (port = 0; port < dd->ipath_cfgports; port++) {
			int port_tidbase = port * dd->ipath_rcvtidcnt;
			int maxtid = port_tidbase + dd->ipath_rcvtidcnt;
			for (i = port_tidbase; i < maxtid; i++) {
				if (!tmpp[i])
					continue;
				pci_unmap_page(dd->pcidev, tmpd[i],
					PAGE_SIZE, PCI_DMA_FROMDEVICE);
				ipath_release_user_pages(&tmpp[i], 1);
				tmpp[i] = NULL;
				cnt++;
			}
		}
		if (cnt) {
			ipath_stats.sps_pageunlocks += cnt;
			ipath_cdbg(VERBOSE, "There were still %u expTID "
				   "entries locked\n", cnt);
		}
		if (ipath_stats.sps_pagelocks ||
		    ipath_stats.sps_pageunlocks)
			ipath_cdbg(VERBOSE, "%llu pages locked, %llu "
				   "unlocked via ipath_m{un}lock\n",
				   (unsigned long long)
				   ipath_stats.sps_pagelocks,
				   (unsigned long long)
				   ipath_stats.sps_pageunlocks);

		ipath_cdbg(VERBOSE, "Free shadow page tid array at %p\n",
			   dd->ipath_pageshadow);
		tmpp = dd->ipath_pageshadow;
		dd->ipath_pageshadow = NULL;
		vfree(tmpp);
	}

	/*
	 * free any resources still in use (usually just kernel ports)
	 * at unload; we do for portcnt, not cfgports, because cfgports
	 * could have changed while we were loaded.
	 */
	for (port = 0; port < dd->ipath_portcnt; port++) {
		struct ipath_portdata *pd = dd->ipath_pd[port];
		dd->ipath_pd[port] = NULL;
		ipath_free_pddata(dd, pd);
	}
	kfree(dd->ipath_pd);
	/*
	 * debuggability, in case some cleanup path tries to use it
	 * after this
	 */
	dd->ipath_pd = NULL;
}

static void __devexit ipath_remove_one(struct pci_dev *pdev)
{
	struct ipath_devdata *dd = pci_get_drvdata(pdev);

	ipath_cdbg(VERBOSE, "removing, pdev=%p, dd=%p\n", pdev, dd);

	/*
	 * disable the IB link early, to be sure no new packets arrive, which
	 * complicates the shutdown process
	 */
	ipath_shutdown_device(dd);

	cancel_delayed_work(&dd->status_work);
	flush_scheduled_work();

	if (dd->verbs_dev)
		ipath_unregister_ib_device(dd->verbs_dev);

	ipath_diag_remove(dd);
	ipath_user_remove(dd);
	ipathfs_remove_device(dd);
	ipath_device_remove_group(&pdev->dev, dd);

	ipath_cdbg(VERBOSE, "Releasing pci memory regions, dd %p, "
		   "unit %u\n", dd, (u32) dd->ipath_unit);

	cleanup_device(dd);

	/*
	 * turn off rcv, send, and interrupts for all ports, all drivers
	 * should also hard reset the chip here?
	 * free up port 0 (kernel) rcvhdr, egr bufs, and eventually tid bufs
	 * for all versions of the driver, if they were allocated
	 */
	if (dd->ipath_irq) {
		ipath_cdbg(VERBOSE, "unit %u free irq %d\n",
			   dd->ipath_unit, dd->ipath_irq);
		dd->ipath_f_free_irq(dd);
	} else
		ipath_dbg("irq is 0, not doing free_irq "
			  "for unit %u\n", dd->ipath_unit);
	/*
	 * we check for NULL here, because it's outside
	 * the kregbase check, and we need to call it
	 * after the free_irq.	Thus it's possible that
	 * the function pointers were never initialized.
	 */
	if (dd->ipath_f_cleanup)
		/* clean up chip-specific stuff */
		dd->ipath_f_cleanup(dd);

	ipath_cdbg(VERBOSE, "Unmapping kregbase %p\n", dd->ipath_kregbase);
	iounmap((volatile void __iomem *) dd->ipath_kregbase);
	pci_release_regions(pdev);
	ipath_cdbg(VERBOSE, "calling pci_disable_device\n");
	pci_disable_device(pdev);

	ipath_free_devdata(pdev, dd);
}

/* general driver use */
DEFINE_MUTEX(ipath_mutex);

static DEFINE_SPINLOCK(ipath_pioavail_lock);

/**
 * ipath_disarm_piobufs - cancel a range of PIO buffers
 * @dd: the infinipath device
 * @first: the first PIO buffer to cancel
 * @cnt: the number of PIO buffers to cancel
 *
 * cancel a range of PIO buffers, used when they might be armed, but
 * not triggered.  Used at init to ensure buffer state, and also user
 * process close, in case it died while writing to a PIO buffer
 * Also after errors.
 */
void ipath_disarm_piobufs(struct ipath_devdata *dd, unsigned first,
			  unsigned cnt)
{
	unsigned i, last = first + cnt;
	u64 sendctrl, sendorig;

	ipath_cdbg(PKT, "disarm %u PIObufs first=%u\n", cnt, first);
	sendorig = dd->ipath_sendctrl;
	for (i = first; i < last; i++) {
		sendctrl = sendorig  | INFINIPATH_S_DISARM |
			(i << INFINIPATH_S_DISARMPIOBUF_SHIFT);
		ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl,
				 sendctrl);
	}

	/*
	 * Write it again with current value, in case ipath_sendctrl changed
	 * while we were looping; no critical bits that would require
	 * locking.
	 *
	 * disable PIOAVAILUPD, then re-enable, reading scratch in
	 * between.  This seems to avoid a chip timing race that causes
	 * pioavail updates to memory to stop.
	 */
	ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl,
			 sendorig & ~INFINIPATH_S_PIOBUFAVAILUPD);
	sendorig = ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
	ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl,
			 dd->ipath_sendctrl);
}

/**
 * ipath_wait_linkstate - wait for an IB link state change to occur
 * @dd: the infinipath 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
 * ipath_set_linkstate.  Returns 0 if state reached, otherwise
 * -ETIMEDOUT state can have multiple states set, for any of several
 * transitions.
 */
static int ipath_wait_linkstate(struct ipath_devdata *dd, u32 state,
				int msecs)
{
	dd->ipath_state_wanted = state;
	wait_event_interruptible_timeout(ipath_state_wait,
					 (dd->ipath_flags & state),
					 msecs_to_jiffies(msecs));
	dd->ipath_state_wanted = 0;

	if (!(dd->ipath_flags & state)) {
		u64 val;
		ipath_cdbg(VERBOSE, "Didn't reach linkstate %s within %u"
			   " ms\n",
			   /* test INIT ahead of DOWN, both can be set */
			   (state & IPATH_LINKINIT) ? "INIT" :
			   ((state & IPATH_LINKDOWN) ? "DOWN" :
			    ((state & IPATH_LINKARMED) ? "ARM" : "ACTIVE")),
			   msecs);
		val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_ibcstatus);
		ipath_cdbg(VERBOSE, "ibcc=%llx ibcstatus=%llx (%s)\n",
			   (unsigned long long) ipath_read_kreg64(
				   dd, dd->ipath_kregs->kr_ibcctrl),
			   (unsigned long long) val,
			   ipath_ibcstatus_str[val & 0xf]);
	}
	return (dd->ipath_flags & state) ? 0 : -ETIMEDOUT;
}

/*
 * Decode the error status into strings, deciding whether to always
 * print * it or not depending on "normal packet errors" vs everything
 * else.   Return 1 if "real" errors, otherwise 0 if only packet
 * errors, so caller can decide what to print with the string.
 */
int ipath_decode_err(char *buf, size_t blen, ipath_err_t err)
{
	int iserr = 1;
	*buf = '\0';
	if (err & INFINIPATH_E_PKTERRS) {
		if (!(err & ~INFINIPATH_E_PKTERRS))
			iserr = 0; // if only packet errors.
		if (ipath_debug & __IPATH_ERRPKTDBG) {
			if (err & INFINIPATH_E_REBP)
				strlcat(buf, "EBP ", blen);
			if (err & INFINIPATH_E_RVCRC)
				strlcat(buf, "VCRC ", blen);
			if (err & INFINIPATH_E_RICRC) {
				strlcat(buf, "CRC ", blen);
				// clear for check below, so only once
				err &= INFINIPATH_E_RICRC;
			}
			if (err & INFINIPATH_E_RSHORTPKTLEN)
				strlcat(buf, "rshortpktlen ", blen);
			if (err & INFINIPATH_E_SDROPPEDDATAPKT)
				strlcat(buf, "sdroppeddatapkt ", blen);
			if (err & INFINIPATH_E_SPKTLEN)
				strlcat(buf, "spktlen ", blen);
		}
		if ((err & INFINIPATH_E_RICRC) &&
			!(err&(INFINIPATH_E_RVCRC|INFINIPATH_E_REBP)))
			strlcat(buf, "CRC ", blen);
		if (!iserr)
			goto done;
	}
	if (err & INFINIPATH_E_RHDRLEN)
		strlcat(buf, "rhdrlen ", blen);
	if (err & INFINIPATH_E_RBADTID)
		strlcat(buf, "rbadtid ", blen);
	if (err & INFINIPATH_E_RBADVERSION)
		strlcat(buf, "rbadversion ", blen);
	if (err & INFINIPATH_E_RHDR)
		strlcat(buf, "rhdr ", blen);
	if (err & INFINIPATH_E_RLONGPKTLEN)
		strlcat(buf, "rlongpktlen ", blen);
	if (err & INFINIPATH_E_RMAXPKTLEN)