xref: /illumos-gate/usr/src/lib/fm/topo/modules/i86pc/chip/chip_amd.c (revision 20c794b3)

/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License (the "License").
 * You may not use this file except in compliance with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */

/*
 * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
 * Use is subject to license terms.
 */

#pragma ident	"%Z%%M%	%I%	%E% SMI"

/*
 * AMD memory enumeration
 */

#include <sys/types.h>
#include <unistd.h>
#include <stropts.h>
#include <sys/fm/protocol.h>
#include <sys/mc.h>
#include <sys/mc_amd.h>
#include <fm/topo_mod.h>
#include <strings.h>
#include <sys/stat.h>
#include <fcntl.h>

#include "chip.h"

#define	MAX_CHANNUM	1
#define	MAX_DIMMNUM	7
#define	MAX_CSNUM	7

static const topo_pgroup_info_t cs_pgroup =
	{ PGNAME(CS), TOPO_STABILITY_PRIVATE, TOPO_STABILITY_PRIVATE, 1 };
static const topo_pgroup_info_t dimm_pgroup =
	{ PGNAME(DIMM), TOPO_STABILITY_PRIVATE, TOPO_STABILITY_PRIVATE, 1 };
static const topo_pgroup_info_t mc_pgroup =
	{ PGNAME(MCT), TOPO_STABILITY_PRIVATE, TOPO_STABILITY_PRIVATE, 1 };
static const topo_pgroup_info_t rank_pgroup =
	{ PGNAME(RANK), TOPO_STABILITY_PRIVATE, TOPO_STABILITY_PRIVATE, 1 };
static const topo_pgroup_info_t chan_pgroup =
	{ PGNAME(CHAN), TOPO_STABILITY_PRIVATE, TOPO_STABILITY_PRIVATE, 1 };

static const topo_method_t dimm_methods[] = {
	{ SIMPLE_DIMM_LBL, "Property method", 0,
	    TOPO_STABILITY_INTERNAL, simple_dimm_label},
	{ SIMPLE_DIMM_LBL_MP, "Property method", 0,
	    TOPO_STABILITY_INTERNAL, simple_dimm_label_mp},
	{ SEQ_DIMM_LBL, "Property method", 0,
	    TOPO_STABILITY_INTERNAL, seq_dimm_label},
	{ NULL }
};

static const topo_method_t rank_methods[] = {
	{ TOPO_METH_ASRU_COMPUTE, TOPO_METH_ASRU_COMPUTE_DESC,
	    TOPO_METH_ASRU_COMPUTE_VERSION, TOPO_STABILITY_INTERNAL,
	    mem_asru_compute },
	{ NULL }
};

static const topo_method_t gen_cs_methods[] = {
	{ TOPO_METH_ASRU_COMPUTE, TOPO_METH_ASRU_COMPUTE_DESC,
	    TOPO_METH_ASRU_COMPUTE_VERSION, TOPO_STABILITY_INTERNAL,
	    mem_asru_compute },
	{ NULL }
};

static nvlist_t *cs_fmri[MC_CHIP_NCS];

/*
 * Called when there is no memory-controller driver to provide topology
 * information.  Generate a maximal memory topology that is appropriate
 * for the chip revision.  The memory-controller node has already been
 * bound as mcnode, and the parent of that is cnode.
 *
 * We create a tree of dram-channel and chip-select nodes below the
 * memory-controller node.  There will be two dram channels and 8 chip-selects
 * below each, regardless of actual socket type, processor revision and so on.
 * This is adequate for generic diagnosis up to family 0x10 revision C.
 * When support for revision D is implemented (or maybe C) we should take
 * the opportunity to rework the topology tree completely (socket change will
 * mean there can be no diagnosis history tied to the topology).
 */
/*ARGSUSED*/
static int
amd_generic_mc_create(topo_mod_t *mod, tnode_t *cnode, tnode_t *mcnode,
    int family, int model, int stepping, nvlist_t *auth)
{
	int chan, cs;

	/*
	 * Elsewhere we have already returned for families less than 0xf.
	 * This "generic" topology is adequate for all of family 0xf and
	 * for revisions A, B and C of family 0x10 (A = model 0, B = model 1,
	 * we'll guess C = model 3 at this point).
	 */
	if (family > 0x10 || (family == 0x10 && model > 3))
		return (1);

	if (topo_node_range_create(mod, mcnode, CHAN_NODE_NAME, 0,
	    MAX_CHANNUM) < 0) {
		whinge(mod, NULL, "amd_generic_mc_create: range create for "
		    "channels failed\n");
		return (-1);
	}

	for (chan = 0; chan <= MAX_CHANNUM; chan++) {
		tnode_t *chnode;
		nvlist_t *fmri;
		int err;

		if (mkrsrc(mod, mcnode, CHAN_NODE_NAME, chan, auth,
		    &fmri) != 0) {
			whinge(mod, NULL, "amd_generic_mc_create: mkrsrc "
			    "failed\n");
			return (-1);
		}

		if ((chnode = topo_node_bind(mod, mcnode, CHAN_NODE_NAME,
		    chan, fmri)) == NULL) {
			nvlist_free(fmri);
			whinge(mod, NULL, "amd_generic_mc_create: node "
			    "bind failed\n");
			return (-1);
		}

		nvlist_free(fmri);

		(void) topo_pgroup_create(chnode, &chan_pgroup, &err);

		(void) topo_prop_set_string(chnode, PGNAME(CHAN), "channel",
		    TOPO_PROP_IMMUTABLE, chan == 0 ? "A" : "B", &err);

		if (topo_node_range_create(mod, chnode, CS_NODE_NAME,
		    0, MAX_CSNUM) < 0) {
			whinge(mod, NULL, "amd_generic_mc_create: "
			    "range create for cs failed\n");
			return (-1);
		}

		for (cs = 0; cs <= MAX_CSNUM; cs++) {
			tnode_t *csnode;

			if (mkrsrc(mod, chnode, CS_NODE_NAME, cs, auth,
			    &fmri) != 0) {
				whinge(mod, NULL, "amd_generic_mc_create: "
				    "mkrsrc for cs failed\n");
				return (-1);
			}

			if ((csnode = topo_node_bind(mod, chnode, CS_NODE_NAME,
			    cs, fmri)) == NULL) {
				nvlist_free(fmri);
				whinge(mod, NULL, "amd_generic_mc_create: "
				    "bind for cs failed\n");
				return (-1);
			}

			/*
			 * Dynamic ASRU for page faults within a chip-select.
			 * The topology does not represent pages (there are
			 * too many) so when a page is faulted we generate
			 * an ASRU to represent the individual page.
			 */
			if (topo_method_register(mod, csnode,
			    gen_cs_methods) < 0)
				whinge(mod, NULL, "amd_generic_mc_create: "
				    "method registration failed\n");

			(void) topo_node_asru_set(csnode, fmri,
			    TOPO_ASRU_COMPUTE, &err);

			nvlist_free(fmri);
		}
	}

	return (0);
}

static nvlist_t *
amd_lookup_by_mcid(topo_mod_t *mod, topo_instance_t id)
{
	mc_snapshot_info_t mcs;
	void *buf = NULL;
	uint8_t ver;

	nvlist_t *nvl = NULL;
	char path[64];
	int fd, err;

	(void) snprintf(path, sizeof (path), "/dev/mc/mc%d", id);
	fd = open(path, O_RDONLY);

	if (fd == -1) {
		/*
		 * Some v20z and v40z systems may have had the 3rd-party
		 * NWSnps packagae installed which installs a /dev/mc
		 * link.  So try again via /devices.
		 */
		(void) snprintf(path, sizeof (path),
		    "/devices/pci@0,0/pci1022,1102@%x,2:mc-amd",
		    MC_AMD_DEV_OFFSET + id);
		fd = open(path, O_RDONLY);
	}

	if (fd == -1)
		return (NULL);	/* do not whinge */

	if (ioctl(fd, MC_IOC_SNAPSHOT_INFO, &mcs) == -1 ||
	    (buf = topo_mod_alloc(mod, mcs.mcs_size)) == NULL ||
	    ioctl(fd, MC_IOC_SNAPSHOT, buf) == -1) {

		whinge(mod, NULL, "mc failed to snapshot %s: %s\n",
		    path, strerror(errno));

		free(buf);
		(void) close(fd);
		return (NULL);
	}

	(void) close(fd);
	err = nvlist_unpack(buf, mcs.mcs_size, &nvl, 0);
	topo_mod_free(mod, buf, mcs.mcs_size);


	if (nvlist_lookup_uint8(nvl, MC_NVLIST_VERSTR, &ver) != 0) {
		whinge(mod, NULL, "mc nvlist is not versioned\n");
		nvlist_free(nvl);
		return (NULL);
	} else if (ver != MC_NVLIST_VERS1) {
		whinge(mod, NULL, "mc nvlist version mismatch\n");
		nvlist_free(nvl);
		return (NULL);
	}

	return (err ? NULL : nvl);
}

int
amd_rank_create(topo_mod_t *mod, tnode_t *pnode, nvlist_t *dimmnvl,
    nvlist_t *auth)
{
	uint64_t *csnumarr;
	char **csnamearr;
	uint_t ncs, ncsname;
	tnode_t *ranknode;
	nvlist_t *fmri, *pfmri = NULL;
	uint64_t dsz, rsz;
	int nerr = 0;
	int err;
	int i;

	if (nvlist_lookup_uint64_array(dimmnvl, "csnums", &csnumarr,
	    &ncs) != 0 || nvlist_lookup_string_array(dimmnvl, "csnames",
	    &csnamearr, &ncsname) != 0 || ncs != ncsname) {
		whinge(mod, &nerr, "amd_rank_create: "
		    "csnums/csnames extraction failed\n");
		return (nerr);
	}

	if (topo_node_resource(pnode, &pfmri, &err) < 0) {
		whinge(mod, &nerr, "amd_rank_create: parent fmri lookup "
		    "failed\n");
		return (nerr);
	}

	if (topo_node_range_create(mod, pnode, RANK_NODE_NAME, 0, ncs) < 0) {
		whinge(mod, &nerr, "amd_rank_create: range create failed\n");
		nvlist_free(pfmri);
		return (nerr);
	}

	if (topo_prop_get_uint64(pnode, PGNAME(DIMM), "size", &dsz,
	    &err) == 0) {
		rsz = dsz / ncs;
	} else {
		whinge(mod, &nerr, "amd_rank_create: parent dimm has no "
		    "size\n");
		return (nerr);
	}

	for (i = 0; i < ncs; i++) {
		if (mkrsrc(mod, pnode, RANK_NODE_NAME, i, auth, &fmri) < 0) {
			whinge(mod, &nerr, "amd_rank_create: mkrsrc failed\n");
			continue;
		}

		if ((ranknode = topo_node_bind(mod, pnode, RANK_NODE_NAME, i,
		    fmri)) == NULL) {
			nvlist_free(fmri);
			whinge(mod, &nerr, "amd_rank_create: node bind "
			    "failed\n");
			continue;
		}

		nvlist_free(fmri);

		(void) topo_node_fru_set(ranknode, pfmri, 0, &err);

		/*
		 * If a rank is faulted the asru is the associated
		 * chip-select, but if a page within a rank is faulted
		 * the asru is just that page.  Hence the dual preconstructed
		 * and computed ASRU.
		 */
		if (topo_method_register(mod, ranknode, rank_methods) < 0)
			whinge(mod, &nerr, "amd_rank_create: "
			    "topo_method_register failed");

		(void) topo_node_asru_set(ranknode, cs_fmri[csnumarr[i]],
		    TOPO_ASRU_COMPUTE, &err);

		(void) topo_pgroup_create(ranknode, &rank_pgroup, &err);

		(void) topo_prop_set_uint64(ranknode, PGNAME(RANK), "size",
		    TOPO_PROP_IMMUTABLE, rsz, &err);

		(void) topo_prop_set_string(ranknode, PGNAME(RANK), "csname",
		    TOPO_PROP_IMMUTABLE, csnamearr[i], &err);

		(void) topo_prop_set_uint64(ranknode, PGNAME(RANK), "csnum",
		    TOPO_PROP_IMMUTABLE, csnumarr[i], &err);
	}

	nvlist_free(pfmri);

	return (nerr);
}

static int
amd_dimm_create(topo_mod_t *mod, tnode_t *pnode, const char *name,
    nvlist_t *mc, nvlist_t *auth)
{
	int i, err, nerr = 0;
	nvpair_t *nvp;
	tnode_t *dimmnode;
	nvlist_t *fmri, *asru, **dimmarr = NULL;
	uint64_t num;
	uint_t ndimm;

	if (nvlist_lookup_nvlist_array(mc, "dimmlist", &dimmarr, &ndimm) != 0) {
		whinge(mod, NULL, "amd_dimm_create: dimmlist lookup failed\n");
		return (-1);
	}

	if (ndimm == 0)
		return (0);	/* no dimms present on this node */

	if (topo_node_range_create(mod, pnode, name, 0, MAX_DIMMNUM) < 0) {
		whinge(mod, NULL, "amd_dimm_create: range create failed\n");
		return (-1);
	}

	for (i = 0; i < ndimm; i++) {
		if (nvlist_lookup_uint64(dimmarr[i], "num", &num) != 0) {
			whinge(mod, &nerr, "amd_dimm_create: dimm num property "
			    "missing\n");
			continue;
		}

		if (mkrsrc(mod, pnode, name, num, auth, &fmri) < 0) {
			whinge(mod, &nerr, "amd_dimm_create: mkrsrc failed\n");
			continue;
		}

		if ((dimmnode = topo_node_bind(mod, pnode, name, num, fmri))
		    == NULL) {
			nvlist_free(fmri);
			whinge(mod, &nerr, "amd_dimm_create: node bind "
			    "failed\n");
			continue;
		}

		if (topo_method_register(mod, dimmnode, dimm_methods) < 0)
			whinge(mod, &nerr, "dimm_create: "
			    "topo_method_register failed");

		/*
		 * Use the mem computation method directly to publish the asru
		 * in the "mem" scheme.
		 */
		if (mem_asru_create(mod, fmri, &asru) == 0) {
			(void) topo_node_asru_set(dimmnode, asru, 0, &err);
			nvlist_free(asru);
		} else {

			nvlist_free(fmri);
			whinge(mod, &nerr, "amd_dimm_create: "
			    "mem_asru_create failed\n");
			continue;
		}

		(void) topo_node_fru_set(dimmnode, fmri, 0, &err);

		nvlist_free(fmri);

		(void) topo_pgroup_create(dimmnode, &dimm_pgroup, &err);

		for (nvp = nvlist_next_nvpair(dimmarr[i], NULL); nvp != NULL;
		    nvp = nvlist_next_nvpair(dimmarr[i], nvp)) {
			if (nvpair_type(nvp) == DATA_TYPE_UINT64_ARRAY &&
			    strcmp(nvpair_name(nvp), "csnums") == 0 ||
			    nvpair_type(nvp) == DATA_TYPE_STRING_ARRAY &&
			    strcmp(nvpair_name(nvp), "csnames") == 0)
				continue;	/* used in amd_rank_create() */

			nerr += nvprop_add(mod, nvp, PGNAME(DIMM), dimmnode);
		}

		nerr += amd_rank_create(mod, dimmnode, dimmarr[i], auth);
	}

	return (nerr == 0 ? 0 : -1);
}

static int
amd_cs_create(topo_mod_t *mod, tnode_t *pnode, const char *name, nvlist_t *mc,
    nvlist_t *auth)
{
	int i, err, nerr = 0;
	nvpair_t *nvp;
	tnode_t *csnode;
	nvlist_t *fmri, **csarr = NULL;
	uint64_t csnum;
	uint_t ncs;

	if (nvlist_lookup_nvlist_array(mc, "cslist", &csarr, &ncs) != 0)
		return (-1);

	if (ncs == 0)
		return (0);	/* no chip-selects configured on this node */

	if (topo_node_range_create(mod, pnode, name, 0, MAX_CSNUM) < 0)
		return (-1);

	for (i = 0; i < ncs; i++) {
		if (nvlist_lookup_uint64(csarr[i], "num", &csnum) != 0) {
			whinge(mod, &nerr, "amd_cs_create: cs num property "
			    "missing\n");
			continue;
		}

		if (mkrsrc(mod, pnode, name, csnum, auth, &fmri) != 0) {
			whinge(mod, &nerr, "amd_cs_create: mkrsrc failed\n");
			continue;
		}

		if ((csnode = topo_node_bind(mod, pnode, name, csnum, fmri))
		    == NULL) {
			nvlist_free(fmri);
			whinge(mod, &nerr, "amd_cs_create: node bind failed\n");
			continue;
		}

		cs_fmri[csnum] = fmri;	/* nvlist will be freed in mc_create */

		(void) topo_node_asru_set(csnode, fmri, 0, &err);

		(void) topo_pgroup_create(csnode, &cs_pgroup, &err);

		for (nvp = nvlist_next_nvpair(csarr[i], NULL); nvp != NULL;
		    nvp = nvlist_next_nvpair(csarr[i], nvp)) {
			nerr += nvprop_add(mod, nvp, PGNAME(CS), csnode);
		}
	}

	return (nerr == 0 ? 0 : -1);
}

static int
amd_dramchan_create(topo_mod_t *mod, tnode_t *pnode, const char *name,
    nvlist_t *auth)
{
	tnode_t *chnode;
	nvlist_t *fmri;
	char *socket;
	int i, nchan;
	int err, nerr = 0;

	/*
	 * We will enumerate the number of channels present even if only
	 * channel A is in use (i.e., running in 64-bit mode).  Only
	 * the socket 754 package has a single channel.
	 */
	if (topo_prop_get_string(pnode, PGNAME(MCT), "socket",
	    &socket, &err) == 0 && strcmp(socket, "Socket 754") == 0)
		nchan = 1;
	else
		nchan = 2;

	topo_mod_strfree(mod, socket);

	if (topo_node_range_create(mod, pnode, name, 0, nchan - 1) < 0)
		return (-1);

	for (i = 0; i < nchan; i++) {
		if (mkrsrc(mod, pnode, name, i, auth, &fmri) != 0) {
			whinge(mod, &nerr, "amd_dramchan_create: mkrsrc "
			    "failed\n");
			continue;
		}

		if ((chnode = topo_node_bind(mod, pnode, name, i, fmri))
		    == NULL) {
			nvlist_free(fmri);
			whinge(mod, &nerr, "amd_dramchan_create: node bind "
			    "failed\n");
			continue;
		}

		nvlist_free(fmri);

		(void) topo_pgroup_create(chnode, &chan_pgroup, &err);

		(void) topo_prop_set_string(chnode, PGNAME(CHAN), "channel",
		    TOPO_PROP_IMMUTABLE, i == 0 ? "A" : "B", &err);
	}

	return (nerr == 0 ? 0 : -1);
}

static int
amd_htconfig(topo_mod_t *mod, tnode_t *cnode, nvlist_t *htnvl)
{
	nvpair_t *nvp;
	int nerr = 0;

	if (strcmp(topo_node_name(cnode), CHIP_NODE_NAME) != 0) {
		whinge(mod, &nerr, "amd_htconfig: must pass a chip node!");
		return (-1);
	}

	for (nvp = nvlist_next_nvpair(htnvl, NULL); nvp != NULL;
	    nvp = nvlist_next_nvpair(htnvl, nvp)) {
		if (nvprop_add(mod, nvp, PGNAME(CHIP), cnode) != 0)
			nerr++;
	}

	return (nerr == 0 ? 0 : -1);
}

void
amd_mc_create(topo_mod_t *mod, tnode_t *pnode, const char *name, nvlist_t *auth,
    int family, int model, int stepping, int *nerrp)
{
	tnode_t *mcnode;
	nvlist_t *fmri;
	nvpair_t *nvp;
	nvlist_t *mc = NULL;
	int i;

	/*
	 * Return with no error for anything before AMD family 0xf - we
	 * won't generate even a generic memory topolofy for earlier
	 * families.
	 */
	if (family < 0xf)
		return;

	if (mkrsrc(mod, pnode, name, 0, auth, &fmri) != 0) {
		whinge(mod, nerrp, "mc_create: mkrsrc failed\n");
		return;
	}

	if (topo_node_range_create(mod, pnode, name, 0, 0) < 0) {
		nvlist_free(fmri);
		whinge(mod, nerrp, "mc_create: node range create failed\n");
		return;
	}

	if ((mcnode = topo_node_bind(mod, pnode, name, 0,
	    fmri)) == NULL) {
		nvlist_free(mc);
		topo_node_range_destroy(pnode, name);
		nvlist_free(fmri);
		whinge(mod, nerrp, "mc_create: mc bind failed\n");
		return;
	}
	(void) topo_node_fru_set(mcnode, NULL, 0, nerrp);
	nvlist_free(fmri);

	if ((mc = amd_lookup_by_mcid(mod, topo_node_instance(pnode))) == NULL) {
		/*
		 * If a memory-controller driver exists for this chip model
		 * it has not attached or has otherwise malfunctioned;
		 * alternatively no memory-controller driver exists for this
		 * (presumably newly-released) cpu model.  We fallback to
		 * creating a generic maximal topology.
		 */
		if (amd_generic_mc_create(mod, pnode, mcnode,
		    family, model, stepping, auth) != 0)
			++*nerrp;
		return;
	}

	/*
	 * Add memory controller properties
	 */
	(void) topo_pgroup_create(mcnode, &mc_pgroup, nerrp);

	for (nvp = nvlist_next_nvpair(mc, NULL); nvp != NULL;
	    nvp = nvlist_next_nvpair(mc, nvp)) {
		char *name = nvpair_name(nvp);
		data_type_t type = nvpair_type(nvp);

		if (type == DATA_TYPE_NVLIST_ARRAY &&
		    (strcmp(name, "cslist") == 0 ||
		    strcmp(name, "dimmlist") == 0)) {
			continue;
		} else if (type == DATA_TYPE_UINT8 &&
		    strcmp(name, MC_NVLIST_VERSTR) == 0) {
			continue;
		} else if (type == DATA_TYPE_NVLIST &&
		    strcmp(name, "htconfig") == 0) {
			nvlist_t *htnvl;

			(void) nvpair_value_nvlist(nvp, &htnvl);
			if (amd_htconfig(mod, pnode, htnvl) != 0)
				++*nerrp;
		} else {
			if (nvprop_add(mod, nvp, PGNAME(MCT), mcnode) != 0)
				++*nerrp;
		}
	}

	if (amd_dramchan_create(mod, mcnode, CHAN_NODE_NAME, auth) != 0 ||
	    amd_cs_create(mod, mcnode, CS_NODE_NAME, mc, auth) != 0 ||
	    amd_dimm_create(mod, mcnode, DIMM_NODE_NAME, mc, auth) != 0)
		++*nerrp;

	/*
	 * Free the fmris for the chip-selects allocated in amd_cs_create
	 */
	for (i = 0; i < MC_CHIP_NCS; i++) {
		if (cs_fmri[i] != NULL) {
			nvlist_free(cs_fmri[i]);
			cs_fmri[i] = NULL;
		}
	}

	nvlist_free(mc);
}