ipath_mad.c

来自「linux 内核源代码」· C语言 代码 · 共 1,571 行 · 第 1/3 页

/*
 * Copyright (c) 2006, 2007 QLogic Corporation. All rights reserved.
 * Copyright (c) 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 <rdma/ib_smi.h>

#include "ipath_kernel.h"
#include "ipath_verbs.h"
#include "ipath_common.h"

#define IB_SMP_UNSUP_VERSION	__constant_htons(0x0004)
#define IB_SMP_UNSUP_METHOD	__constant_htons(0x0008)
#define IB_SMP_UNSUP_METH_ATTR	__constant_htons(0x000C)
#define IB_SMP_INVALID_FIELD	__constant_htons(0x001C)

static int reply(struct ib_smp *smp)
{
	/*
	 * The verbs framework will handle the directed/LID route
	 * packet changes.
	 */
	smp->method = IB_MGMT_METHOD_GET_RESP;
	if (smp->mgmt_class == IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE)
		smp->status |= IB_SMP_DIRECTION;
	return IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_REPLY;
}

static int recv_subn_get_nodedescription(struct ib_smp *smp,
					 struct ib_device *ibdev)
{
	if (smp->attr_mod)
		smp->status |= IB_SMP_INVALID_FIELD;

	strncpy(smp->data, ibdev->node_desc, sizeof(smp->data));

	return reply(smp);
}

struct nodeinfo {
	u8 base_version;
	u8 class_version;
	u8 node_type;
	u8 num_ports;
	__be64 sys_guid;
	__be64 node_guid;
	__be64 port_guid;
	__be16 partition_cap;
	__be16 device_id;
	__be32 revision;
	u8 local_port_num;
	u8 vendor_id[3];
} __attribute__ ((packed));

static int recv_subn_get_nodeinfo(struct ib_smp *smp,
				  struct ib_device *ibdev, u8 port)
{
	struct nodeinfo *nip = (struct nodeinfo *)&smp->data;
	struct ipath_devdata *dd = to_idev(ibdev)->dd;
	u32 vendor, majrev, minrev;

	/* GUID 0 is illegal */
	if (smp->attr_mod || (dd->ipath_guid == 0))
		smp->status |= IB_SMP_INVALID_FIELD;

	nip->base_version = 1;
	nip->class_version = 1;
	nip->node_type = 1;	/* channel adapter */
	/*
	 * XXX The num_ports value will need a layer function to get
	 * the value if we ever have more than one IB port on a chip.
	 * We will also need to get the GUID for the port.
	 */
	nip->num_ports = ibdev->phys_port_cnt;
	/* This is already in network order */
	nip->sys_guid = to_idev(ibdev)->sys_image_guid;
	nip->node_guid = dd->ipath_guid;
	nip->port_guid = dd->ipath_guid;
	nip->partition_cap = cpu_to_be16(ipath_get_npkeys(dd));
	nip->device_id = cpu_to_be16(dd->ipath_deviceid);
	majrev = dd->ipath_majrev;
	minrev = dd->ipath_minrev;
	nip->revision = cpu_to_be32((majrev << 16) | minrev);
	nip->local_port_num = port;
	vendor = dd->ipath_vendorid;
	nip->vendor_id[0] = 0;
	nip->vendor_id[1] = vendor >> 8;
	nip->vendor_id[2] = vendor;

	return reply(smp);
}

static int recv_subn_get_guidinfo(struct ib_smp *smp,
				  struct ib_device *ibdev)
{
	u32 startgx = 8 * be32_to_cpu(smp->attr_mod);
	__be64 *p = (__be64 *) smp->data;

	/* 32 blocks of 8 64-bit GUIDs per block */

	memset(smp->data, 0, sizeof(smp->data));

	/*
	 * We only support one GUID for now.  If this changes, the
	 * portinfo.guid_cap field needs to be updated too.
	 */
	if (startgx == 0) {
		__be64 g = to_idev(ibdev)->dd->ipath_guid;
		if (g == 0)
			/* GUID 0 is illegal */
			smp->status |= IB_SMP_INVALID_FIELD;
		else
			/* The first is a copy of the read-only HW GUID. */
			*p = g;
	} else
		smp->status |= IB_SMP_INVALID_FIELD;

	return reply(smp);
}


static int get_overrunthreshold(struct ipath_devdata *dd)
{
	return (dd->ipath_ibcctrl >>
		INFINIPATH_IBCC_OVERRUNTHRESHOLD_SHIFT) &
		INFINIPATH_IBCC_OVERRUNTHRESHOLD_MASK;
}

/**
 * set_overrunthreshold - set the overrun threshold
 * @dd: the infinipath device
 * @n: the new threshold
 *
 * Note that this will only take effect when the link state changes.
 */
static int set_overrunthreshold(struct ipath_devdata *dd, unsigned n)
{
	unsigned v;

	v = (dd->ipath_ibcctrl >> INFINIPATH_IBCC_OVERRUNTHRESHOLD_SHIFT) &
		INFINIPATH_IBCC_OVERRUNTHRESHOLD_MASK;
	if (v != n) {
		dd->ipath_ibcctrl &=
			~(INFINIPATH_IBCC_OVERRUNTHRESHOLD_MASK <<
			  INFINIPATH_IBCC_OVERRUNTHRESHOLD_SHIFT);
		dd->ipath_ibcctrl |=
			(u64) n << INFINIPATH_IBCC_OVERRUNTHRESHOLD_SHIFT;
		ipath_write_kreg(dd, dd->ipath_kregs->kr_ibcctrl,
				 dd->ipath_ibcctrl);
	}
	return 0;
}

static int get_phyerrthreshold(struct ipath_devdata *dd)
{
	return (dd->ipath_ibcctrl >>
		INFINIPATH_IBCC_PHYERRTHRESHOLD_SHIFT) &
		INFINIPATH_IBCC_PHYERRTHRESHOLD_MASK;
}

/**
 * set_phyerrthreshold - set the physical error threshold
 * @dd: the infinipath device
 * @n: the new threshold
 *
 * Note that this will only take effect when the link state changes.
 */
static int set_phyerrthreshold(struct ipath_devdata *dd, unsigned n)
{
	unsigned v;

	v = (dd->ipath_ibcctrl >> INFINIPATH_IBCC_PHYERRTHRESHOLD_SHIFT) &
		INFINIPATH_IBCC_PHYERRTHRESHOLD_MASK;
	if (v != n) {
		dd->ipath_ibcctrl &=
			~(INFINIPATH_IBCC_PHYERRTHRESHOLD_MASK <<
			  INFINIPATH_IBCC_PHYERRTHRESHOLD_SHIFT);
		dd->ipath_ibcctrl |=
			(u64) n << INFINIPATH_IBCC_PHYERRTHRESHOLD_SHIFT;
		ipath_write_kreg(dd, dd->ipath_kregs->kr_ibcctrl,
				 dd->ipath_ibcctrl);
	}
	return 0;
}

/**
 * get_linkdowndefaultstate - get the default linkdown state
 * @dd: the infinipath device
 *
 * Returns zero if the default is POLL, 1 if the default is SLEEP.
 */
static int get_linkdowndefaultstate(struct ipath_devdata *dd)
{
	return !!(dd->ipath_ibcctrl & INFINIPATH_IBCC_LINKDOWNDEFAULTSTATE);
}

static int recv_subn_get_portinfo(struct ib_smp *smp,
				  struct ib_device *ibdev, u8 port)
{
	struct ipath_ibdev *dev;
	struct ib_port_info *pip = (struct ib_port_info *)smp->data;
	u16 lid;
	u8 ibcstat;
	u8 mtu;
	int ret;

	if (be32_to_cpu(smp->attr_mod) > ibdev->phys_port_cnt) {
		smp->status |= IB_SMP_INVALID_FIELD;
		ret = reply(smp);
		goto bail;
	}

	dev = to_idev(ibdev);

	/* Clear all fields.  Only set the non-zero fields. */
	memset(smp->data, 0, sizeof(smp->data));

	/* Only return the mkey if the protection field allows it. */
	if (smp->method == IB_MGMT_METHOD_SET || dev->mkey == smp->mkey ||
	    dev->mkeyprot == 0)
		pip->mkey = dev->mkey;
	pip->gid_prefix = dev->gid_prefix;
	lid = dev->dd->ipath_lid;
	pip->lid = lid ? cpu_to_be16(lid) : IB_LID_PERMISSIVE;
	pip->sm_lid = cpu_to_be16(dev->sm_lid);
	pip->cap_mask = cpu_to_be32(dev->port_cap_flags);
	/* pip->diag_code; */
	pip->mkey_lease_period = cpu_to_be16(dev->mkey_lease_period);
	pip->local_port_num = port;
	pip->link_width_enabled = dev->link_width_enabled;
	pip->link_width_supported = 3;	/* 1x or 4x */
	pip->link_width_active = 2;	/* 4x */
	pip->linkspeed_portstate = 0x10;	/* 2.5Gbps */
	ibcstat = dev->dd->ipath_lastibcstat;
	pip->linkspeed_portstate |= ((ibcstat >> 4) & 0x3) + 1;
	pip->portphysstate_linkdown =
		(ipath_cvt_physportstate[ibcstat & 0xf] << 4) |
		(get_linkdowndefaultstate(dev->dd) ? 1 : 2);
	pip->mkeyprot_resv_lmc = (dev->mkeyprot << 6) | dev->dd->ipath_lmc;
	pip->linkspeedactive_enabled = 0x11;	/* 2.5Gbps, 2.5Gbps */
	switch (dev->dd->ipath_ibmtu) {
	case 4096:
		mtu = IB_MTU_4096;
		break;
	case 2048:
		mtu = IB_MTU_2048;
		break;
	case 1024:
		mtu = IB_MTU_1024;
		break;
	case 512:
		mtu = IB_MTU_512;
		break;
	case 256:
		mtu = IB_MTU_256;
		break;
	default:		/* oops, something is wrong */
		mtu = IB_MTU_2048;
		break;
	}
	pip->neighbormtu_mastersmsl = (mtu << 4) | dev->sm_sl;
	pip->vlcap_inittype = 0x10;	/* VLCap = VL0, InitType = 0 */
	pip->vl_high_limit = dev->vl_high_limit;
	/* pip->vl_arb_high_cap; // only one VL */
	/* pip->vl_arb_low_cap; // only one VL */
	/* InitTypeReply = 0 */
	/*
	 * Note: the chips support a maximum MTU of 4096, but the driver
	 * hasn't implemented this feature yet, so set the maximum value
	 * to 2048.
	 */
	pip->inittypereply_mtucap = IB_MTU_2048;
	// HCAs ignore VLStallCount and HOQLife
	/* pip->vlstallcnt_hoqlife; */
	pip->operationalvl_pei_peo_fpi_fpo = 0x10;	/* OVLs = 1 */
	pip->mkey_violations = cpu_to_be16(dev->mkey_violations);
	/* P_KeyViolations are counted by hardware. */
	pip->pkey_violations =
		cpu_to_be16((ipath_get_cr_errpkey(dev->dd) -
			     dev->z_pkey_violations) & 0xFFFF);
	pip->qkey_violations = cpu_to_be16(dev->qkey_violations);
	/* Only the hardware GUID is supported for now */
	pip->guid_cap = 1;
	pip->clientrereg_resv_subnetto = dev->subnet_timeout;
	/* 32.768 usec. response time (guessing) */
	pip->resv_resptimevalue = 3;
	pip->localphyerrors_overrunerrors =
		(get_phyerrthreshold(dev->dd) << 4) |
		get_overrunthreshold(dev->dd);
	/* pip->max_credit_hint; */
	/* pip->link_roundtrip_latency[3]; */

	ret = reply(smp);

bail:
	return ret;
}

/**
 * get_pkeys - return the PKEY table for port 0
 * @dd: the infinipath device
 * @pkeys: the pkey table is placed here
 */
static int get_pkeys(struct ipath_devdata *dd, u16 * pkeys)
{
	struct ipath_portdata *pd = dd->ipath_pd[0];

	memcpy(pkeys, pd->port_pkeys, sizeof(pd->port_pkeys));

	return 0;
}

static int recv_subn_get_pkeytable(struct ib_smp *smp,
				   struct ib_device *ibdev)
{
	u32 startpx = 32 * (be32_to_cpu(smp->attr_mod) & 0xffff);
	u16 *p = (u16 *) smp->data;
	__be16 *q = (__be16 *) smp->data;

	/* 64 blocks of 32 16-bit P_Key entries */

	memset(smp->data, 0, sizeof(smp->data));
	if (startpx == 0) {
		struct ipath_ibdev *dev = to_idev(ibdev);
		unsigned i, n = ipath_get_npkeys(dev->dd);

		get_pkeys(dev->dd, p);

		for (i = 0; i < n; i++)
			q[i] = cpu_to_be16(p[i]);
	} else
		smp->status |= IB_SMP_INVALID_FIELD;

	return reply(smp);
}

static int recv_subn_set_guidinfo(struct ib_smp *smp,
				  struct ib_device *ibdev)
{
	/* The only GUID we support is the first read-only entry. */
	return recv_subn_get_guidinfo(smp, ibdev);
}

/**
 * set_linkdowndefaultstate - set the default linkdown state
 * @dd: the infinipath device
 * @sleep: the new state
 *
 * Note that this will only take effect when the link state changes.
 */
static int set_linkdowndefaultstate(struct ipath_devdata *dd, int sleep)
{
	if (sleep)
		dd->ipath_ibcctrl |= INFINIPATH_IBCC_LINKDOWNDEFAULTSTATE;
	else
		dd->ipath_ibcctrl &= ~INFINIPATH_IBCC_LINKDOWNDEFAULTSTATE;
	ipath_write_kreg(dd, dd->ipath_kregs->kr_ibcctrl,
			 dd->ipath_ibcctrl);
	return 0;
}

/**
 * recv_subn_set_portinfo - set port information
 * @smp: the incoming SM packet
 * @ibdev: the infiniband device
 * @port: the port on the device
 *
 * Set Portinfo (see ch. 14.2.5.6).
 */
static int recv_subn_set_portinfo(struct ib_smp *smp,
				  struct ib_device *ibdev, u8 port)
{
	struct ib_port_info *pip = (struct ib_port_info *)smp->data;
	struct ib_event event;
	struct ipath_ibdev *dev;
	struct ipath_devdata *dd;
	char clientrereg = 0;
	u16 lid, smlid;
	u8 lwe;
	u8 lse;
	u8 state;
	u16 lstate;
	u32 mtu;
	int ret, ore;

	if (be32_to_cpu(smp->attr_mod) > ibdev->phys_port_cnt)
		goto err;

	dev = to_idev(ibdev);
	dd = dev->dd;
	event.device = ibdev;
	event.element.port_num = port;

	dev->mkey = pip->mkey;
	dev->gid_prefix = pip->gid_prefix;
	dev->mkey_lease_period = be16_to_cpu(pip->mkey_lease_period);

	lid = be16_to_cpu(pip->lid);
	if (dd->ipath_lid != lid ||
	    dd->ipath_lmc != (pip->mkeyprot_resv_lmc & 7)) {
		/* Must be a valid unicast LID address. */
		if (lid == 0 || lid >= IPATH_MULTICAST_LID_BASE)
			goto err;
		ipath_set_lid(dd, lid, pip->mkeyprot_resv_lmc & 7);
		event.event = IB_EVENT_LID_CHANGE;
		ib_dispatch_event(&event);
	}

	smlid = be16_to_cpu(pip->sm_lid);
	if (smlid != dev->sm_lid) {
		/* Must be a valid unicast LID address. */
		if (smlid == 0 || smlid >= IPATH_MULTICAST_LID_BASE)
			goto err;
		dev->sm_lid = smlid;
		event.event = IB_EVENT_SM_CHANGE;
		ib_dispatch_event(&event);
	}

	/* Only 4x supported but allow 1x or 4x to be set (see 14.2.6.6). */
	lwe = pip->link_width_enabled;
	if ((lwe >= 4 && lwe <= 8) || (lwe >= 0xC && lwe <= 0xFE))
		goto err;
	if (lwe == 0xFF)
		dev->link_width_enabled = 3;	/* 1x or 4x */
	else if (lwe)
		dev->link_width_enabled = lwe;

	/* Only 2.5 Gbs supported. */
	lse = pip->linkspeedactive_enabled & 0xF;
	if (lse >= 2 && lse <= 0xE)
		goto err;

	/* Set link down default state. */
	switch (pip->portphysstate_linkdown & 0xF) {
	case 0: /* NOP */
		break;
	case 1: /* SLEEP */
		if (set_linkdowndefaultstate(dd, 1))
			goto err;
		break;
	case 2: /* POLL */
		if (set_linkdowndefaultstate(dd, 0))
			goto err;
		break;
	default:
		goto err;
	}

	dev->mkeyprot = pip->mkeyprot_resv_lmc >> 6;
	dev->vl_high_limit = pip->vl_high_limit;

	switch ((pip->neighbormtu_mastersmsl >> 4) & 0xF) {
	case IB_MTU_256:
		mtu = 256;
		break;
	case IB_MTU_512:
		mtu = 512;
		break;
	case IB_MTU_1024:
		mtu = 1024;
		break;
	case IB_MTU_2048:
		mtu = 2048;
		break;
	case IB_MTU_4096:
		mtu = 4096;
		break;
	default:
		/* XXX We have already partially updated our state! */
		goto err;
	}
	ipath_set_mtu(dd, mtu);

	dev->sm_sl = pip->neighbormtu_mastersmsl & 0xF;

	/* We only support VL0 */
	if (((pip->operationalvl_pei_peo_fpi_fpo >> 4) & 0xF) > 1)
		goto err;

	if (pip->mkey_violations == 0)
		dev->mkey_violations = 0;

	/*
	 * Hardware counter can't be reset so snapshot and subtract
	 * later.
	 */
	if (pip->pkey_violations == 0)
		dev->z_pkey_violations = ipath_get_cr_errpkey(dd);

	if (pip->qkey_violations == 0)
		dev->qkey_violations = 0;

	ore = pip->localphyerrors_overrunerrors;
	if (set_phyerrthreshold(dd, (ore >> 4) & 0xF))
		goto err;