modules/genunix/memory.c

/*
 * 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 2008 Sun Microsystems, Inc.  All rights reserved.
 * Use is subject to license terms.
 */

#include <mdb/mdb_modapi.h>
#include <sys/types.h>
#include <vm/page.h>
#include <sys/thread.h>
#include <sys/swap.h>
#include <sys/memlist.h>
#if defined(__i386) || defined(__amd64)
#include <sys/balloon_impl.h>
#endif

/*
 * Page walker.
 * By default, this will walk all pages in the system.  If given an
 * address, it will walk all pages belonging to the vnode at that
 * address.
 */

/*
 * page_walk_data
 *
 * pw_hashleft is set to -1 when walking a vnode's pages, and holds the
 * number of hash locations remaining in the page hash table when
 * walking all pages.
 *
 * The astute reader will notice that pw_hashloc is only used when
 * reading all pages (to hold a pointer to our location in the page
 * hash table), and that pw_first is only used when reading the pages
 * belonging to a particular vnode (to hold a pointer to the first
 * page).  While these could be combined to be a single pointer, they
 * are left separate for clarity.
 */
typedef struct page_walk_data {
	long		pw_hashleft;
	void		**pw_hashloc;
	uintptr_t	pw_first;
} page_walk_data_t;

int
page_walk_init(mdb_walk_state_t *wsp)
{
	page_walk_data_t	*pwd;
	void	**ptr;
	size_t	hashsz;
	vnode_t	vn;

	if (wsp->walk_addr == NULL) {

		/*
		 * Walk all pages
		 */

		if ((mdb_readvar(&ptr, "page_hash") == -1) ||
		    (mdb_readvar(&hashsz, "page_hashsz") == -1) ||
		    (ptr == NULL) || (hashsz == 0)) {
			mdb_warn("page_hash, page_hashsz not found or invalid");
			return (WALK_ERR);
		}

		/*
		 * Since we are walking all pages, initialize hashleft
		 * to be the remaining number of entries in the page
		 * hash.  hashloc is set the start of the page hash
		 * table.  Setting the walk address to 0 indicates that
		 * we aren't currently following a hash chain, and that
		 * we need to scan the page hash table for a page.
		 */
		pwd = mdb_alloc(sizeof (page_walk_data_t), UM_SLEEP);
		pwd->pw_hashleft = hashsz;
		pwd->pw_hashloc = ptr;
		wsp->walk_addr = 0;
	} else {

		/*
		 * Walk just this vnode
		 */

		if (mdb_vread(&vn, sizeof (vnode_t), wsp->walk_addr) == -1) {
			mdb_warn("unable to read vnode_t at %#lx",
			    wsp->walk_addr);
			return (WALK_ERR);
		}

		/*
		 * We set hashleft to -1 to indicate that we are
		 * walking a vnode, and initialize first to 0 (it is
		 * used to terminate the walk, so it must not be set
		 * until after we have walked the first page).  The
		 * walk address is set to the first page.
		 */
		pwd = mdb_alloc(sizeof (page_walk_data_t), UM_SLEEP);
		pwd->pw_hashleft = -1;
		pwd->pw_first = 0;

		wsp->walk_addr = (uintptr_t)vn.v_pages;
	}

	wsp->walk_data = pwd;

	return (WALK_NEXT);
}

int
page_walk_step(mdb_walk_state_t *wsp)
{
	page_walk_data_t	*pwd = wsp->walk_data;
	page_t		page;
	uintptr_t	pp;

	pp = wsp->walk_addr;

	if (pwd->pw_hashleft < 0) {

		/* We're walking a vnode's pages */

		/*
		 * If we don't have any pages to walk, we have come
		 * back around to the first one (we finished), or we
		 * can't read the page we're looking at, we are done.
		 */
		if (pp == NULL || pp == pwd->pw_first)
			return (WALK_DONE);
		if (mdb_vread(&page, sizeof (page_t), pp) == -1) {
			mdb_warn("unable to read page_t at %#lx", pp);
			return (WALK_ERR);
		}

		/*
		 * Set the walk address to the next page, and if the
		 * first page hasn't been set yet (i.e. we are on the
		 * first page), set it.
		 */
		wsp->walk_addr = (uintptr_t)page.p_vpnext;
		if (pwd->pw_first == NULL)
			pwd->pw_first = pp;

	} else if (pwd->pw_hashleft > 0) {

		/* We're walking all pages */

		/*
		 * If pp (the walk address) is NULL, we scan through
		 * the page hash table until we find a page.
		 */
		if (pp == NULL) {

			/*
			 * Iterate through the page hash table until we
			 * find a page or reach the end.
			 */
			do {
				if (mdb_vread(&pp, sizeof (uintptr_t),
				    (uintptr_t)pwd->pw_hashloc) == -1) {
					mdb_warn("unable to read from %#p",
					    pwd->pw_hashloc);
					return (WALK_ERR);
				}
				pwd->pw_hashleft--;
				pwd->pw_hashloc++;
			} while (pwd->pw_hashleft && (pp == NULL));

			/*
			 * We've reached the end; exit.
			 */
			if (pp == NULL)
				return (WALK_DONE);
		}

		if (mdb_vread(&page, sizeof (page_t), pp) == -1) {
			mdb_warn("unable to read page_t at %#lx", pp);
			return (WALK_ERR);
		}

		/*
		 * Set the walk address to the next page.
		 */
		wsp->walk_addr = (uintptr_t)page.p_hash;

	} else {
		/* We've finished walking all pages. */
		return (WALK_DONE);
	}

	return (wsp->walk_callback(pp, &page, wsp->walk_cbdata));
}

void
page_walk_fini(mdb_walk_state_t *wsp)
{
	mdb_free(wsp->walk_data, sizeof (page_walk_data_t));
}

/* Summary statistics of pages */
typedef struct memstat {
	struct vnode    *ms_kvp;	/* Cached address of kernel vnode */
	struct vnode    *ms_zvp;	/* Cached address of zio vnode    */
	uint64_t	ms_kmem;	/* Pages of kernel memory	  */
	uint64_t	ms_zfs_data;	/* Pages of zfs data		  */
	uint64_t	ms_anon;	/* Pages of anonymous memory	  */
	uint64_t	ms_vnode;	/* Pages of named (vnode) memory  */
	uint64_t	ms_exec;	/* Pages of exec/library memory	  */
	uint64_t	ms_cachelist;	/* Pages on the cachelist (free)  */
	uint64_t	ms_total;	/* Pages on page hash		  */
} memstat_t;

#define	MS_PP_ISKAS(pp, stats)				\
	((pp)->p_vnode == (stats)->ms_kvp)

#define	MS_PP_ISZFS_DATA(pp, stats)			\
	(((stats)->ms_zvp != NULL) && ((pp)->p_vnode == (stats)->ms_zvp))

/*
 * Summarize pages by type; called from page walker.
 */

/* ARGSUSED */
static int
memstat_callback(page_t *page, page_t *pp, memstat_t *stats)
{
	struct vnode vn, *vp;
	uintptr_t ptr;

	/* read page's vnode pointer */
	if ((ptr = (uintptr_t)(pp->p_vnode)) != NULL) {
		if (mdb_vread(&vn, sizeof (vnode_t), ptr) == -1) {
			mdb_warn("unable to read vnode_t at %#lx",
			    ptr);
			return (WALK_ERR);
		}
		vp = &vn;
	} else
		vp = NULL;

	if (PP_ISFREE(pp))
		stats->ms_cachelist++;
	else if (vp && IS_SWAPFSVP(vp))
		stats->ms_anon++;
	else if (MS_PP_ISZFS_DATA(pp, stats))
		stats->ms_zfs_data++;
	else if (MS_PP_ISKAS(pp, stats))
		stats->ms_kmem++;
	else if (vp && (((vp)->v_flag & VVMEXEC)) != 0)
		stats->ms_exec++;
	else
		stats->ms_vnode++;

	stats->ms_total++;

	return (WALK_NEXT);
}

/* ARGSUSED */
int
memstat(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
{
	ulong_t pagesize;
	pgcnt_t total_pages, physmem;
	ulong_t freemem;
	memstat_t stats;
	memstat_t unused_stats;
	GElf_Sym sym;
#if defined(__i386) || defined(__amd64)
	bln_stats_t bln_stats;
	ssize_t bln_size;
#endif

	bzero(&stats, sizeof (memstat_t));
	bzero(&unused_stats, sizeof (memstat_t));

	if (argc != 0 || (flags & DCMD_ADDRSPEC))
		return (DCMD_USAGE);

	/* Grab base page size */
	if (mdb_readvar(&pagesize, "_pagesize") == -1) {
		mdb_warn("unable to read _pagesize");
		return (DCMD_ERR);
	}

	/* Total physical memory */
	if (mdb_readvar(&total_pages, "total_pages") == -1) {
		mdb_warn("unable to read total_pages");
		return (DCMD_ERR);
	}

	/* Artificially limited memory */
	if (mdb_readvar(&physmem, "physmem") == -1) {
		mdb_warn("unable to read physmem");
		return (DCMD_ERR);
	}

	/* read kernel vnode pointer */
	if (mdb_lookup_by_obj(MDB_OBJ_EXEC, "kvp",
	    (GElf_Sym *)&sym) == -1) {
		mdb_warn("unable to read kvp");
		return (DCMD_ERR);
	}

	stats.ms_kvp = (struct vnode *)(uintptr_t)sym.st_value;

	/*
	 * Read the zio vnode pointer.  It may not exist on all kernels, so it
	 * it isn't found, it's not a fatal error.
	 */
	if (mdb_lookup_by_obj(MDB_OBJ_EXEC, "zvp",
	    (GElf_Sym *)&sym) == -1) {
		stats.ms_zvp = NULL;
	} else {
		stats.ms_zvp = (struct vnode *)(uintptr_t)sym.st_value;
	}

	/* Walk page structures, summarizing usage */
	if (mdb_walk("page", (mdb_walk_cb_t)memstat_callback,
	    &stats) == -1) {
		mdb_warn("can't walk pages");
		return (DCMD_ERR);
	}

	/* read unused pages vnode */
	if (mdb_lookup_by_obj(MDB_OBJ_EXEC, "unused_pages_vp",
	    (GElf_Sym *)&sym) == -1) {
		mdb_warn("unable to read unused_pages_vp");
		return (DCMD_ERR);
	}

	unused_stats.ms_kvp = (struct vnode *)(uintptr_t)sym.st_value;

	/* Find unused pages */
	if (mdb_walk("page", (mdb_walk_cb_t)memstat_callback,
	    &unused_stats) == -1) {
		mdb_warn("can't walk pages");
		return (DCMD_ERR);
	}

	/*
	 * If physmem != total_pages, then the administrator has limited the
	 * number of pages available in the system.  In order to account for
	 * this, we reduce the amount normally attributed to the page cache.
	 */
	stats.ms_vnode -= unused_stats.ms_kmem;
	stats.ms_total -= unused_stats.ms_kmem;

#define	MS_PCT_TOTAL(x)	((ulong_t)((((5 * total_pages) + ((x) * 1000ull))) / \
		((physmem) * 10)))

	mdb_printf("Page Summary                Pages                MB"
	    "  %%Tot\n");
	mdb_printf("------------     ----------------  ----------------"
	    "  ----\n");
	mdb_printf("Kernel           %16llu  %16llu  %3lu%%\n",
	    stats.ms_kmem,
	    (uint64_t)stats.ms_kmem * pagesize / (1024 * 1024),
	    MS_PCT_TOTAL(stats.ms_kmem));

	if (stats.ms_zfs_data != 0)
		mdb_printf("ZFS File Data    %16llu  %16llu  %3lu%%\n",
		    stats.ms_zfs_data,
		    (uint64_t)stats.ms_zfs_data * pagesize / (1024 * 1024),
		    MS_PCT_TOTAL(stats.ms_zfs_data));

	mdb_printf("Anon             %16llu  %16llu  %3lu%%\n",
	    stats.ms_anon,
	    (uint64_t)stats.ms_anon * pagesize / (1024 * 1024),
	    MS_PCT_TOTAL(stats.ms_anon));
	mdb_printf("Exec and libs    %16llu  %16llu  %3lu%%\n",
	    stats.ms_exec,
	    (uint64_t)stats.ms_exec * pagesize / (1024 * 1024),
	    MS_PCT_TOTAL(stats.ms_exec));
	mdb_printf("Page cache       %16llu  %16llu  %3lu%%\n",
	    stats.ms_vnode,
	    (uint64_t)stats.ms_vnode * pagesize / (1024 * 1024),
	    MS_PCT_TOTAL(stats.ms_vnode));
	mdb_printf("Free (cachelist) %16llu  %16llu  %3lu%%\n",
	    stats.ms_cachelist,
	    (uint64_t)stats.ms_cachelist * pagesize / (1024 * 1024),
	    MS_PCT_TOTAL(stats.ms_cachelist));

	/*
	 * occasionally, we double count pages above.  To avoid printing
	 * absurdly large values for freemem, we clamp it at zero.
	 */
	if (physmem > stats.ms_total)
		freemem = physmem - stats.ms_total;
	else
		freemem = 0;

#if defined(__i386) || defined(__amd64)
	/* Are we running under Xen?  If so, get balloon memory usage. */
	if ((bln_size = mdb_readvar(&bln_stats, "bln_stats")) != -1) {
		if (freemem > bln_stats.bln_hv_pages)
			freemem -= bln_stats.bln_hv_pages;
		else
			freemem = 0;
	}
#endif

	mdb_printf("Free (freelist)  %16lu  %16llu  %3lu%%\n", freemem,
	    (uint64_t)freemem * pagesize / (1024 * 1024),
	    MS_PCT_TOTAL(freemem));

#if defined(__i386) || defined(__amd64)
	if (bln_size != -1) {
		mdb_printf("Balloon          %16lu  %16llu  %3lu%%\n",
		    bln_stats.bln_hv_pages,
		    (uint64_t)bln_stats.bln_hv_pages * pagesize / (1024 * 1024),
		    MS_PCT_TOTAL(bln_stats.bln_hv_pages));
	}
#endif

	mdb_printf("\nTotal            %16lu  %16lu\n",
	    physmem,
	    (uint64_t)physmem * pagesize / (1024 * 1024));

	if (physmem != total_pages) {
		mdb_printf("Physical         %16lu  %16lu\n",
		    total_pages,
		    (uint64_t)total_pages * pagesize / (1024 * 1024));
	}

#undef MS_PCT_TOTAL

	return (DCMD_OK);
}

int
page(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
{
	page_t	p;

	if (!(flags & DCMD_ADDRSPEC)) {
		if (mdb_walk_dcmd("page", "page", argc, argv) == -1) {
			mdb_warn("can't walk pages");
			return (DCMD_ERR);
		}
		return (DCMD_OK);
	}

	if (DCMD_HDRSPEC(flags)) {
		mdb_printf("%<u>%?s %?s %16s %8s %3s %3s %2s %2s %2s%</u>\n",
		    "PAGE", "VNODE", "OFFSET", "SELOCK",
		    "LCT", "COW", "IO", "FS", "ST");
	}

	if (mdb_vread(&p, sizeof (page_t), addr) == -1) {
		mdb_warn("can't read page_t at %#lx", addr);
		return (DCMD_ERR);
	}

	mdb_printf("%0?lx %?p %16llx %8x %3d %3d %2x %2x %2x\n",
	    addr, p.p_vnode, p.p_offset, p.p_selock, p.p_lckcnt, p.p_cowcnt,
	    p.p_iolock_state, p.p_fsdata, p.p_state);

	return (DCMD_OK);
}

int
swap_walk_init(mdb_walk_state_t *wsp)
{
	void	*ptr;

	if ((mdb_readvar(&ptr, "swapinfo") == -1) || ptr == NULL) {
		mdb_warn("swapinfo not found or invalid");
		return (WALK_ERR);
	}

	wsp->walk_addr = (uintptr_t)ptr;

	return (WALK_NEXT);
}

int
swap_walk_step(mdb_walk_state_t *wsp)
{
	uintptr_t	sip;
	struct swapinfo	si;

	sip = wsp->walk_addr;

	if (sip == NULL)
		return (WALK_DONE);

	if (mdb_vread(&si, sizeof (struct swapinfo), sip) == -1) {
		mdb_warn("unable to read swapinfo at %#lx", sip);
		return (WALK_ERR);
	}

	wsp->walk_addr = (uintptr_t)si.si_next;

	return (wsp->walk_callback(sip, &si, wsp->walk_cbdata));
}

int
swapinfof(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
{
	struct swapinfo	si;
	char		*name;

	if (!(flags & DCMD_ADDRSPEC)) {
		if (mdb_walk_dcmd("swapinfo", "swapinfo", argc, argv) == -1) {
			mdb_warn("can't walk swapinfo");
			return (DCMD_ERR);
		}
		return (DCMD_OK);
	}

	if (DCMD_HDRSPEC(flags)) {
		mdb_printf("%<u>%?s %?s %9s %9s %s%</u>\n",
		    "ADDR", "VNODE", "PAGES", "FREE", "NAME");
	}

	if (mdb_vread(&si, sizeof (struct swapinfo), addr) == -1) {
		mdb_warn("can't read swapinfo at %#lx", addr);
		return (DCMD_ERR);
	}

	name = mdb_alloc(si.si_pnamelen, UM_SLEEP | UM_GC);
	if (mdb_vread(name, si.si_pnamelen, (uintptr_t)si.si_pname) == -1)
		name = "*error*";

	mdb_printf("%0?lx %?p %9d %9d %s\n",
	    addr, si.si_vp, si.si_npgs, si.si_nfpgs, name);

	return (DCMD_OK);
}

int
memlist_walk_step(mdb_walk_state_t *wsp)
{
	uintptr_t	mlp;
	struct memlist	ml;

	mlp = wsp->walk_addr;

	if (mlp == NULL)
		return (WALK_DONE);

	if (mdb_vread(&ml, sizeof (struct memlist), mlp) == -1) {
		mdb_warn("unable to read memlist at %#lx", mlp);
		return (WALK_ERR);
	}

	wsp->walk_addr = (uintptr_t)ml.next;

	return (wsp->walk_callback(mlp, &ml, wsp->walk_cbdata));
}

int
memlist(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
{
	struct memlist	ml;

	if (!(flags & DCMD_ADDRSPEC)) {
		uintptr_t ptr;
		uint_t list = 0;
		int i;
		static const char *lists[] = {
			"phys_install",
			"phys_avail",
			"virt_avail"
		};

		if (mdb_getopts(argc, argv,
		    'i', MDB_OPT_SETBITS, (1 << 0), &list,
		    'a', MDB_OPT_SETBITS, (1 << 1), &list,
		    'v', MDB_OPT_SETBITS, (1 << 2), &list, NULL) != argc)
			return (DCMD_USAGE);

		if (!list)
			list = 1;

		for (i = 0; list; i++, list >>= 1) {
			if (!(list & 1))
				continue;
			if ((mdb_readvar(&ptr, lists[i]) == -1) ||
			    (ptr == NULL)) {
				mdb_warn("%s not found or invalid", lists[i]);
				return (DCMD_ERR);
			}

			mdb_printf("%s:\n", lists[i]);
			if (mdb_pwalk_dcmd("memlist", "memlist", 0, NULL,
			    ptr) == -1) {
				mdb_warn("can't walk memlist");
				return (DCMD_ERR);
			}
		}
		return (DCMD_OK);
	}

	if (DCMD_HDRSPEC(flags))
		mdb_printf("%<u>%?s %16s %16s%</u>\n", "ADDR", "BASE", "SIZE");

	if (mdb_vread(&ml, sizeof (struct memlist), addr) == -1) {
		mdb_warn("can't read memlist at %#lx", addr);
		return (DCMD_ERR);
	}

	mdb_printf("%0?lx %16llx %16llx\n", addr, ml.address, ml.size);

	return (DCMD_OK);
}