[PATCH 1/2] powerpc/pagetable: Add option to dump the linux pagetables

[PATCH 1/2] powerpc/pagetable: Add option to dump the linux pagetables

[Rashmica Gupta]

Useful to be able to dump the kernels page tables to check permissions
and memory types - derived from arm64's implementation.

Add a debugfs file to check the page tables. To use this the PPC_PTDUMP
config option must be selected.

Signed-off-by: Rashmica Gupta <rashmicy@gmail.com>
---
 arch/powerpc/Kconfig.debug             |  12 ++
 arch/powerpc/mm/Makefile               |   1 +
 arch/powerpc/mm/dump_linuxpagetables.c | 377 +++++++++++++++++++++++++++++++++
 3 files changed, 390 insertions(+)
 create mode 100644 arch/powerpc/mm/dump_linuxpagetables.c

diff --git a/arch/powerpc/Kconfig.debug b/arch/powerpc/Kconfig.debug
index 638f9ce740f5..26a60effea1a 100644
--- a/arch/powerpc/Kconfig.debug
+++ b/arch/powerpc/Kconfig.debug
@@ -344,4 +344,16 @@ config FAIL_IOMMU
 
 	  If you are unsure, say N.
 
+config PPC_PTDUMP
+        bool "Export kernel pagetable layout to userspace via debugfs"
+        depends on DEBUG_KERNEL
+        select DEBUG_FS
+        help
+	  This option exports the state of the kernel pagetables to a
+	  debugfs file. This is only useful for kernel developers who are
+	  working in architecture specific areas of the kernel - probably
+	  not a good idea to enable this feature in a production kernel.
+
+	  If you are unsure, say N.
+
 endmenu
diff --git a/arch/powerpc/mm/Makefile b/arch/powerpc/mm/Makefile
index adfee3f1aeb9..6935c6204fbc 100644
--- a/arch/powerpc/mm/Makefile
+++ b/arch/powerpc/mm/Makefile
@@ -41,3 +41,4 @@ obj-$(CONFIG_NOT_COHERENT_CACHE) += dma-noncoherent.o
 obj-$(CONFIG_HIGHMEM)		+= highmem.o
 obj-$(CONFIG_PPC_COPRO_BASE)	+= copro_fault.o
 obj-$(CONFIG_SPAPR_TCE_IOMMU)	+= mmu_context_iommu.o
+obj-$(CONFIG_PPC_PTDUMP)	+= dump_linuxpagetables.o
diff --git a/arch/powerpc/mm/dump_linuxpagetables.c b/arch/powerpc/mm/dump_linuxpagetables.c
new file mode 100644
index 000000000000..f97fbfdac4b9
--- /dev/null
+++ b/arch/powerpc/mm/dump_linuxpagetables.c
@@ -0,0 +1,377 @@
+/*
+ * Copyright 2016, Rashmica Gupta, IBM Corp.
+ *
+ * This traverses the kernel pagetables and dumps the
+ * information about the used sections of memory to
+ * /sys/kernel/debug/kernel_pagetables.
+ *
+ * Derived from the arm64 implementation:
+ * Copyright (c) 2014, The Linux Foundation, Laura Abbott.
+ * (C) Copyright 2008 Intel Corporation, Arjan van de Ven.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; version 2
+ * of the License.
+ */
+#include <linux/debugfs.h>
+#include <linux/fs.h>
+#include <linux/io.h>
+#include <linux/mm.h>
+#include <linux/sched.h>
+#include <linux/seq_file.h>
+#include <asm/fixmap.h>
+#include <asm/pgtable.h>
+#include <linux/const.h>
+#include <asm/page.h>
+#include <asm/pgalloc.h>
+
+struct addr_marker {
+	unsigned long start_address;
+	const char *name;
+};
+
+static struct addr_marker address_markers[] = {
+	{ VMALLOC_START,	"vmalloc() Area" },
+	{ VMALLOC_END,		"vmalloc() End" },
+	{ ISA_IO_BASE,		"isa I/O start" },
+	{ ISA_IO_END,		"isa I/O end" },
+	{ PHB_IO_BASE,		"phb I/O start" },
+	{ PHB_IO_END,		"phb I/O end" },
+	{ IOREMAP_BASE,		"I/O remap start" },
+	{ IOREMAP_END,		"I/O remap end" },
+	{ -1,			NULL },
+};
+
+/*
+ * To visualise what is happening,
+ *
+ *  - PTRS_PER_P** = how many entries there are in the corresponding P**
+ *  - P**_SHIFT = how many bits of the address we use to index into the
+ * corresponding P**
+ *  - P**_SIZE is how much memory we can access through the table - not the
+ * size of the table itself.
+ * P**={PGD, PUD, PMD, PTE}
+ *
+ *
+ * Each entry of the PGD points to a PUD. Each entry of a PUD points to a
+ * PMD. Each entry of a PMD points to a PTE. And every PTE entry points to
+ * a page.
+ *
+ * In the case where there are only 3 levels, the PUD is folded into the
+ * PGD: every PUD has only one entry which points to the PMD.
+ *
+ * The page dumper groups page table entries of the same type into a single
+ * description. It uses pg_state to track the range information while
+ * iterating over the PTE entries. When the continuity is broken it then
+ * dumps out a description of the range - ie PTEs that are virtually contiguous
+ * with the same PTE flags are chunked together. This is to make it clear how
+ * different areas of the kernel virtual memory are used.
+ *
+ */
+struct pg_state {
+	struct seq_file *seq;
+	const struct addr_marker *marker;
+	unsigned long start_address;
+	unsigned level;
+	u64 current_flags;
+};
+
+struct flag_info {
+	u64		mask;
+	u64		val;
+	const char	*set;
+	const char	*clear;
+};
+
+static const struct flag_info flag_array[] = {
+	{
+		.mask	= _PAGE_USER,
+		.val	= _PAGE_USER,
+		.set	= "user",
+		.clear	= "    ",
+	}, {
+		.mask	= _PAGE_RW,
+		.val	= _PAGE_RW,
+		.set	= "rw",
+		.clear	= "ro",
+	}, {
+		.mask	= _PAGE_EXEC,
+		.val	= _PAGE_EXEC,
+		.set	= " X ",
+		.clear	= "   ",
+	}, {
+		.mask	= _PAGE_PTE,
+		.val	= _PAGE_PTE,
+		.set	= "pte",
+		.clear	= "   ",
+	}, {
+		.mask	= _PAGE_PRESENT,
+		.val	= _PAGE_PRESENT,
+		.set	= "present",
+		.clear	= "       ",
+	}, {
+		.mask	= _PAGE_HASHPTE,
+		.val	= _PAGE_HASHPTE,
+		.set	= "hpte",
+		.clear	= "    ",
+	}, {
+		.mask	= _PAGE_GUARDED,
+		.val	= _PAGE_GUARDED,
+		.set	= "guarded",
+		.clear	= "       ",
+	}, {
+		.mask	= _PAGE_DIRTY,
+		.val	= _PAGE_DIRTY,
+		.set	= "dirty",
+		.clear	= "     ",
+	}, {
+		.mask	= _PAGE_ACCESSED,
+		.val	= _PAGE_ACCESSED,
+		.set	= "accessed",
+		.clear	= "        ",
+	}, {
+		.mask	= _PAGE_WRITETHRU,
+		.val	= _PAGE_WRITETHRU,
+		.set	= "write through",
+		.clear	= "             ",
+	}, {
+		.mask	= _PAGE_NO_CACHE,
+		.val	= _PAGE_NO_CACHE,
+		.set	= "no cache",
+		.clear	= "        ",
+	}, {
+		.mask	= _PAGE_BUSY,
+		.val	= _PAGE_BUSY,
+		.set	= "busy",
+	}, {
+		.mask	= _PAGE_COMBO,
+		.val	= _PAGE_COMBO,
+		.set	= "combo",
+	}, {
+		.mask	= _PAGE_F_GIX,
+		.val	= _PAGE_F_GIX,
+		.set	= "f_gix",
+	}, {
+		.mask	= _PAGE_F_SECOND,
+		.val	= _PAGE_F_SECOND,
+		.set	= "f_second",
+	}, {
+		.mask	= _PAGE_SPECIAL,
+		.val	= _PAGE_SPECIAL,
+		.set	= "special",
+	}, {
+		.mask	= _PAGE_4K_PFN,
+		.val	= _PAGE_4K_PFN,
+		.set	= "4K_pfn"
+	}
+};
+
+struct pgtable_level {
+	const struct flag_info *flag;
+	size_t num;
+	u64 mask;
+};
+
+static struct pgtable_level pg_level[] = {
+	{
+	}, { /* pgd */
+		.flag	= flag_array,
+		.num	= ARRAY_SIZE(flag_array),
+	}, { /* pud */
+		.flag	= flag_array,
+		.num	= ARRAY_SIZE(flag_array),
+	}, { /* pmd */
+		.flag	= flag_array,
+		.num	= ARRAY_SIZE(flag_array),
+	}, { /* pte */
+		.flag	= flag_array,
+		.num	= ARRAY_SIZE(flag_array),
+	},
+};
+
+static void dump_flag(struct pg_state *st, const struct flag_info *flag,
+			size_t num)
+{
+	unsigned i;
+
+	for (i = 0; i < num; i++, flag++) {
+		const char *s;
+
+		if ((st->current_flags & flag->mask) == flag->val)
+			s = flag->set;
+		else
+			s = flag->clear;
+
+		if (s)
+			seq_printf(st->seq, " %s", s);
+	}
+}
+
+static void dump_addr(struct pg_state *st, unsigned long addr)
+{
+	static const char units[] = "KMGTPE";
+	const char *unit = units;
+	unsigned long delta;
+
+	seq_printf(st->seq, "0x%016lx-0x%016lx   ", st->start_address, addr-1);
+	delta = (addr - st->start_address) >> 10;
+	/* Work out what appropriate unit to use */
+	while (!(delta & 1023) && unit[1]) {
+		delta >>= 10;
+		unit++;
+	}
+	seq_printf(st->seq, "%9lu%c", delta, *unit);
+
+}
+
+static void note_page(struct pg_state *st, unsigned long addr, unsigned level,
+				u64 val)
+{
+	u64 flag = val & pg_level[level].mask;
+
+	/* At first no level is set */
+	if (!st->level) {
+		st->level = level;
+		st->current_flags = flag;
+		st->start_address = addr;
+		seq_printf(st->seq, "---[ %s ]---\n", st->marker->name);
+	/*
+	 * Dump the section of virtual memory when:
+	 *   - the PTE flags from one entry to the next differs.
+	 *   - we change levels in the tree.
+	 *   - the address is in a different section of memory and is thus
+	 *   used for a different purpose, regardless of the flags.
+	 */
+	} else if (flag != st->current_flags || level != st->level ||
+		   addr >= st->marker[1].start_address) {
+
+		/* Check the PTE flags */
+		if (st->current_flags) {
+			dump_addr(st, addr);
+
+			/* Dump all the flags */
+			if (pg_level[st->level].flag)
+				dump_flag(st, pg_level[st->level].flag,
+					  pg_level[st->level].num);
+			seq_puts(st->seq, "\n");
+		}
+
+		/* Address indicates we have passed the end of the
+		 * current section of virtual memory
+		 */
+		while (addr >= st->marker[1].start_address) {
+			st->marker++;
+			seq_printf(st->seq, "---[ %s ]---\n", st->marker->name);
+		}
+		st->start_address = addr;
+		st->current_flags = flag;
+		st->level = level;
+	}
+}
+
+static void walk_pte(struct pg_state *st, pmd_t *pmd, unsigned long start)
+{
+	pte_t *pte = pte_offset_kernel(pmd, 0);
+	unsigned long addr;
+	unsigned i;
+
+	for (i = 0; i < PTRS_PER_PTE; i++, pte++) {
+		addr = start + i * PAGE_SIZE;
+		note_page(st, addr, 4, pte_val(*pte));
+	}
+}
+
+static void walk_pmd(struct pg_state *st, pud_t *pud, unsigned long start)
+{
+	pmd_t *pmd = pmd_offset(pud, 0);
+	unsigned long addr;
+	unsigned i;
+
+	for (i = 0; i < PTRS_PER_PMD; i++, pmd++) {
+		addr = start + i * PMD_SIZE;
+		if (!pmd_none(*pmd))
+			/* pmd exists */
+			walk_pte(st, pmd, addr);
+		else
+			note_page(st, addr, 3, pmd_val(*pmd));
+	}
+}
+
+static void walk_pud(struct pg_state *st, pgd_t *pgd, unsigned long start)
+{
+	pud_t *pud = pud_offset(pgd, 0);
+	unsigned long addr;
+	unsigned i;
+
+	for (i = 0; i < PTRS_PER_PUD; i++, pud++) {
+		addr = start + i * PUD_SIZE;
+		if (!pud_none(*pud))
+			/* pud exists */
+			walk_pmd(st, pud, addr);
+		else
+			note_page(st, addr, 2, pud_val(*pud));
+	}
+}
+
+static void walk_pgd(struct pg_state *st, unsigned long start)
+{
+	pgd_t *pgd = pgd_offset_k(0UL);
+	unsigned i;
+	unsigned long addr;
+
+	for (i = 0; i < PTRS_PER_PGD; i++, pgd++) {
+		addr = start + i * PGDIR_SIZE;
+		if (!pgd_none(*pgd))
+			/* pgd exists */
+			walk_pud(st, pgd, addr);
+		else
+			note_page(st, addr, 1, pgd_val(*pgd));
+	}
+}
+
+static int ptdump_show(struct seq_file *m, void *v)
+{
+	struct pg_state st = {
+		.seq = m,
+		.start_address = KERN_VIRT_START,
+		.marker = address_markers,
+	};
+	/* Traverse kernel page tables */
+	walk_pgd(&st, KERN_VIRT_START);
+	note_page(&st, 0, 0, 0);
+	return 0;
+}
+
+static int ptdump_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, ptdump_show, NULL);
+}
+
+static const struct file_operations ptdump_fops = {
+	.open		= ptdump_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+static void build_pgtable_complete_mask(void)
+{
+	unsigned i, j;
+
+	for (i = 0; i < ARRAY_SIZE(pg_level); i++)
+		if (pg_level[i].flag)
+			for (j = 0; j < pg_level[i].num; j++)
+				pg_level[i].mask |= pg_level[i].flag[j].mask;
+}
+
+static int ptdump_init(void)
+{
+	struct dentry *debugfs_file;
+
+	build_pgtable_complete_mask();
+	debugfs_file = debugfs_create_file("kernel_pagetables", 0400, NULL,
+			NULL, &ptdump_fops);
+	return debugfs_file ? 0 : -ENOMEM;
+}
+device_initcall(ptdump_init);
-- 
2.5.0

[PATCH 2/2] powerpc/pagetable: Add option to dump kernel hashpagetable

[Rashmica Gupta]

Useful to be able to dump the kernel hash page table to check
which pages are hashed along with their sizes and other details.

Add a debugfs file to check the page tables. To use this the PPC_PTDUMP
config option must be selected.

Signed-off-by: Rashmica Gupta <rashmicy@gmail.com>
---
 arch/powerpc/mm/Makefile             |   3 +-
 arch/powerpc/mm/dump_hashpagetable.c | 488 +++++++++++++++++++++++++++++++++++
 2 files changed, 490 insertions(+), 1 deletion(-)
 create mode 100644 arch/powerpc/mm/dump_hashpagetable.c

diff --git a/arch/powerpc/mm/Makefile b/arch/powerpc/mm/Makefile
index 6935c6204fbc..0bd48b345ff3 100644
--- a/arch/powerpc/mm/Makefile
+++ b/arch/powerpc/mm/Makefile
@@ -41,4 +41,5 @@ obj-$(CONFIG_NOT_COHERENT_CACHE) += dma-noncoherent.o
 obj-$(CONFIG_HIGHMEM)		+= highmem.o
 obj-$(CONFIG_PPC_COPRO_BASE)	+= copro_fault.o
 obj-$(CONFIG_SPAPR_TCE_IOMMU)	+= mmu_context_iommu.o
-obj-$(CONFIG_PPC_PTDUMP)	+= dump_linuxpagetables.o
+obj-$(CONFIG_PPC_PTDUMP)	+= dump_linuxpagetables.o \
+				   dump_hashpagetable.o
diff --git a/arch/powerpc/mm/dump_hashpagetable.c b/arch/powerpc/mm/dump_hashpagetable.c
new file mode 100644
index 000000000000..044fd13e4d07
--- /dev/null
+++ b/arch/powerpc/mm/dump_hashpagetable.c
@@ -0,0 +1,488 @@
+/*
+ * Copyright 2016, Rashmica Gupta, IBM Corp.
+ *
+ * This traverses the kernel virtual memory and dumps the pages that are in
+ * the hash pagetable, along with their flags to
+ * /sys/kernel/debug/kernel_hash_pagetable.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; version 2
+ * of the License.
+ */
+#include <linux/debugfs.h>
+#include <linux/fs.h>
+#include <linux/io.h>
+#include <linux/mm.h>
+#include <linux/sched.h>
+#include <linux/seq_file.h>
+#include <asm/fixmap.h>
+#include <asm/pgtable.h>
+#include <linux/const.h>
+#include <asm/page.h>
+#include <asm/pgalloc.h>
+#include <asm/plpar_wrappers.h>
+#include <linux/memblock.h>
+#include <asm/firmware.h>
+
+struct addr_marker {
+	unsigned long start_address;
+	const char *name;
+};
+
+static struct addr_marker address_markers[] = {
+	{ PAGE_OFFSET,		"Start of kernel VM"},
+	{ VMALLOC_START,	"vmalloc() Area" },
+	{ VMALLOC_END,		"vmalloc() End" },
+	{ ISA_IO_BASE,		"isa I/O start" },
+	{ ISA_IO_END,		"isa I/O end" },
+	{ PHB_IO_BASE,		"phb I/O start" },
+	{ PHB_IO_END,		"phb I/O end" },
+	{ IOREMAP_BASE,		"I/O remap start" },
+	{ IOREMAP_END,		"I/O remap end" },
+	{ VMEMMAP_BASE,		"vmemmap start" },
+	{ -1,			NULL },
+};
+
+struct pg_state {
+	struct seq_file *seq;
+	const struct addr_marker *marker;
+	unsigned long start_address;
+	unsigned level;
+	u64 current_flags;
+};
+
+struct flag_info {
+	u64		mask;
+	u64		val;
+	const char	*set;
+	const char	*clear;
+	bool		is_val;
+};
+
+static const struct flag_info v_flag_array[] = {
+	{
+		.mask   = SLB_VSID_B,
+		.val    = SLB_VSID_B_256M,
+		.set    = "ssize: 256M",
+		.clear  = "ssize: 1T  ",
+	}, {
+		.mask	= HPTE_V_SECONDARY,
+		.val	= HPTE_V_SECONDARY,
+		.set	= "secondary",
+		.clear	= "primary  ",
+	}, {
+		.mask	= HPTE_V_VALID,
+		.val	= HPTE_V_VALID,
+		.set	= "valid  ",
+		.clear	= "invalid",
+	}, {
+		.mask	= HPTE_V_BOLTED,
+		.val	= HPTE_V_BOLTED,
+		.set	= "bolted",
+		.clear	= "",
+	}
+};
+
+static const struct flag_info r_flag_array[] = {
+	{
+		.mask	= HPTE_R_PP0 | HPTE_R_PP,
+		.val	= HPTE_R_PP0 | HPTE_R_PP,
+		.set	= "prot",
+		.clear	= "",
+		.is_val = true,
+	}, {
+		.mask	= HPTE_R_KEY_HI | HPTE_R_KEY_LO,
+		.val	= HPTE_R_KEY_HI | HPTE_R_KEY_LO,
+		.set	= "key",
+		.clear	= "",
+		.is_val = true,
+	}, {
+		.mask	= HPTE_R_R,
+		.val	= HPTE_R_R,
+		.set	= "ref",
+		.clear	= "   ",
+	}, {
+		.mask	= HPTE_R_C,
+		.val	= HPTE_R_C,
+		.set	= "changed",
+		.clear	= "       ",
+	}, {
+		.mask	= HPTE_R_N,
+		.val	= HPTE_R_N,
+		.set	= "no execute",
+		.clear	= "",
+	}, {
+		.mask	= HPTE_R_WIMG,
+		.val	= HPTE_R_W,
+		.set	= "writethru",
+		.clear	= "",
+	}, {
+		.mask	= HPTE_R_WIMG,
+		.val	= HPTE_R_I,
+		.set	= "no cache",
+		.clear	= "",
+	}, {
+		.mask	= HPTE_R_WIMG,
+		.val	= HPTE_R_G,
+		.set	= "guarded",
+		.clear	= "",
+	}
+};
+
+static void dump_flag_info(struct pg_state *st, const struct flag_info
+		*flag, unsigned long pte, int num)
+{
+	unsigned i;
+
+	for (i = 0; i < num; i++, flag++) {
+		const char *s = NULL;
+
+		if (flag->is_val) {
+			seq_printf(st->seq, "  %s:%llx", flag->set, pte &
+					flag->val);
+		} else {
+			if ((pte & flag->mask) == flag->val)
+				s = flag->set;
+			else
+				s = flag->clear;
+			seq_printf(st->seq, "  %s", s);
+		}
+	}
+}
+
+static void dump_hpte_info(struct pg_state *st, unsigned long ea, unsigned long
+		v, unsigned long r, unsigned long rpn, int bps, int aps,
+		unsigned long lp)
+{
+	static const char units[] = "BKMGTPE";
+	const char *unit = units;
+
+	while (ea >= st->marker[1].start_address) {
+		st->marker++;
+		seq_printf(st->seq, "---[ %s ]---\n", st->marker->name);
+	}
+	seq_printf(st->seq, "0x%lx:\t", ea);
+	seq_printf(st->seq, "AVPN:%lx\t", HPTE_V_AVPN_VAL(v));
+	dump_flag_info(st, v_flag_array, v, ARRAY_SIZE(v_flag_array));
+	seq_printf(st->seq, "  rpn: %lx\t", rpn);
+	dump_flag_info(st, r_flag_array, r, ARRAY_SIZE(r_flag_array));
+
+	while (bps > 9 && unit[1]) {
+		bps -= 10;
+		unit++;
+	}
+	seq_printf(st->seq, "base_ps: %i%c\t", 1<<bps, *unit);
+	while (aps > 9 && unit[1]) {
+		aps -= 10;
+		unit++;
+	}
+	seq_printf(st->seq, "actual_ps: %i%c", 1<<aps, *unit);
+	if (lp != -1)
+		seq_printf(st->seq, "\tLP enc: %lx", lp);
+	seq_puts(st->seq, "\n");
+}
+
+static int native_find(unsigned long ea, int psize, bool primary, unsigned long
+		*v, unsigned long *r)
+{
+	struct hash_pte *hptep;
+	unsigned long hash, vsid, vpn, hpte_group, want_v, hpte_v;
+	int i, ssize = mmu_kernel_ssize;
+	unsigned long shift = mmu_psize_defs[psize].shift;
+
+	/* calculate hash */
+	vsid = get_kernel_vsid(ea, ssize);
+	vpn  = hpt_vpn(ea, vsid, ssize);
+	hash = hpt_hash(vpn, shift, ssize);
+	want_v = hpte_encode_avpn(vpn, psize, ssize);
+
+	/* to check in the secondary hash table, we invert the hash */
+	if (!primary)
+		hash = ~hash;
+	hpte_group = (hash & htab_hash_mask) * HPTES_PER_GROUP;
+	for (i = 0; i < HPTES_PER_GROUP; i++) {
+		hptep = htab_address + hpte_group;
+		hpte_v = be64_to_cpu(hptep->v);
+
+		if (HPTE_V_COMPARE(hpte_v, want_v) && (hpte_v & HPTE_V_VALID)) {
+			/* HPTE matches */
+			*v = be64_to_cpu(hptep->v);
+			*r = be64_to_cpu(hptep->r);
+			return 0;
+		}
+		++hpte_group;
+	}
+	return -1;
+}
+
+static int pseries_find(unsigned long ea, int psize, bool primary, unsigned
+		long *v, unsigned long *r)
+{
+	struct hash_pte ptes[4];
+	unsigned long vsid, vpn, hash, hpte_group, want_v;
+	int i, j, ssize = mmu_kernel_ssize;
+	long lpar_rc = 0;
+	unsigned long shift = mmu_psize_defs[psize].shift;
+
+	/* calculate hash */
+	vsid = get_kernel_vsid(ea, ssize);
+	vpn  = hpt_vpn(ea, vsid, ssize);
+	hash = hpt_hash(vpn, shift, ssize);
+	want_v = hpte_encode_avpn(vpn, psize, ssize);
+
+	/* to check in the secondary hash table, we invert the hash */
+	if (!primary)
+		hash = ~hash;
+	hpte_group = ((hash & htab_hash_mask) * HPTES_PER_GROUP) & ~0x7UL;
+	/* see if we can find an entry in the hpte with this hash */
+	for (i = 0; i < HPTES_PER_GROUP; i += 4, hpte_group += 4) {
+		lpar_rc = plpar_pte_read_4(0, hpte_group, (void *)ptes);
+
+		if (lpar_rc != H_SUCCESS)
+			continue;
+		for (j = 0; j < 4; j++) {
+			if (HPTE_V_COMPARE(ptes[j].v, want_v) &&
+					(ptes[j].v & HPTE_V_VALID)) {
+				/* HPTE matches */
+				*v = ptes[j].v;
+				*r = ptes[j].r;
+				return 0;
+			}
+		}
+	}
+	return -1;
+}
+
+static void decode_r(int bps, unsigned long r, unsigned long *rpn, int *aps,
+		unsigned long *lp_bits)
+{
+	struct mmu_psize_def entry;
+	unsigned long arpn, mask, lp;
+	int penc = -2, idx = 0, shift;
+
+	/*.
+	 * The LP field has 8 bits. Depending on the actual page size, some of
+	 * these bits are concatenated with the APRN to get the RPN. The rest
+	 * of the bits in the LP field is the LP value and is an encoding for
+	 * the base page size and the actual page size.
+	 *
+	 *  -	find the mmu entry for our base page size
+	 *  -	go through all page encodings and use the associated mask to
+	 *	find an encoding that matches our encoding in the LP field.
+	 */
+	arpn = (r & HPTE_R_RPN) >> HPTE_R_RPN_SHIFT;
+	lp = arpn & 0xff;
+
+	entry = mmu_psize_defs[bps];
+	while (idx < MMU_PAGE_COUNT) {
+		penc = entry.penc[idx];
+		if ((penc != -1) && (mmu_psize_defs[idx].shift)) {
+			shift = mmu_psize_defs[idx].shift -  HPTE_R_RPN_SHIFT;
+			mask = (0x1 << (shift)) - 1;
+			if ((lp & mask) == penc) {
+				*aps = mmu_psize_to_shift(idx);
+				*lp_bits = lp & mask;
+				*rpn = arpn >> shift;
+				return;
+			}
+		}
+		idx++;
+	}
+}
+
+static unsigned long hpte_find(struct pg_state *st, unsigned long ea, int psize)
+{
+	unsigned long slot;
+	unsigned long v  = 0, r = 0, rpn, lp_bits;
+	int base_psize = 0, actual_psize = 0;
+
+	if (ea <= PAGE_OFFSET)
+		return -1;
+
+	/* Look in primary table */
+	if (firmware_has_feature(FW_FEATURE_LPAR))
+		slot = pseries_find(ea, psize, true, &v, &r);
+	else
+		slot = native_find(ea, psize, true, &v, &r);
+
+	/* Look in secondary table */
+	if (slot == -1) {
+		if (firmware_has_feature(FW_FEATURE_LPAR))
+			slot = pseries_find(ea, psize, false, &v, &r);
+		else
+			slot = native_find(ea, psize, false, &v, &r);
+	}
+
+	/* No entry found */
+	if (slot == -1)
+		return -1;
+
+	/* We found an entry in the hash page table:
+	 *  - check that this has the same base page
+	 *  - find the actual page size
+	 *  - find the RPN
+	 */
+	base_psize = mmu_psize_to_shift(psize);
+
+	if ((v & HPTE_V_LARGE) == HPTE_V_LARGE) {
+		decode_r(psize, r, &rpn, &actual_psize, &lp_bits);
+	} else {
+		/* 4K actual page size */
+		actual_psize = 12;
+		rpn = (r & HPTE_R_RPN) >> HPTE_R_RPN_SHIFT;
+		/* In this case there are no LP bits */
+		lp_bits = -1;
+	}
+	/* We didn't find a matching encoding, so the PTE we found isn't for
+	 * this address.
+	 */
+	if (actual_psize == -1)
+		return -1;
+
+	dump_hpte_info(st, ea, v, r, rpn, base_psize, actual_psize, lp_bits);
+	return 0;
+}
+
+static void walk_pte(struct pg_state *st, pmd_t *pmd, unsigned long start)
+{
+	pte_t *pte = pte_offset_kernel(pmd, 0);
+	unsigned long addr, pteval, psize;
+	int i, status;
+
+	for (i = 0; i < PTRS_PER_PTE; i++, pte++) {
+		addr = start + i * PAGE_SIZE;
+		pteval = pte_val(*pte);
+
+		if (addr < VMALLOC_END)
+			psize = mmu_vmalloc_psize;
+		else
+			psize = mmu_io_psize;
+
+		/* check for secret 4K mappings */
+		if (((pteval & _PAGE_COMBO) == _PAGE_COMBO) ||
+			((pteval & _PAGE_4K_PFN) == _PAGE_4K_PFN))
+			psize = mmu_io_psize;
+
+		/* check for hashpte */
+		status = hpte_find(st, addr, psize);
+
+		if (((pteval & _PAGE_HASHPTE) != _PAGE_HASHPTE)
+				&& (status != -1)) {
+		/* found a hpte that is not in the linux page tables */
+			seq_printf(st->seq, "page probably bolted before linux"
+				" pagetables were set: addr:%lx, pteval:%lx\n",
+				addr, pteval);
+		}
+	}
+}
+
+static void walk_pmd(struct pg_state *st, pud_t *pud, unsigned long start)
+{
+	pmd_t *pmd = pmd_offset(pud, 0);
+	unsigned long addr;
+	unsigned i;
+
+	for (i = 0; i < PTRS_PER_PMD; i++, pmd++) {
+		addr = start + i * PMD_SIZE;
+		if (!pmd_none(*pmd))
+			/* pmd exists */
+			walk_pte(st, pmd, addr);
+	}
+}
+
+static void walk_pud(struct pg_state *st, pgd_t *pgd, unsigned long start)
+{
+	pud_t *pud = pud_offset(pgd, 0);
+	unsigned long addr;
+	unsigned i;
+
+	for (i = 0; i < PTRS_PER_PUD; i++, pud++) {
+		addr = start + i * PUD_SIZE;
+		if (!pud_none(*pud))
+			/* pud exists */
+			walk_pmd(st, pud, addr);
+	}
+}
+
+static void walk_linearmapping(struct pg_state *st)
+{
+	unsigned long addr;
+
+	/* Traverse the linear mapping section of virtual memory and dump pages
+	 * that are in the hash pagetable.
+	 */
+	for (addr = PAGE_OFFSET; addr < PAGE_OFFSET +
+			memblock_phys_mem_size(); addr += PAGE_SIZE)
+		hpte_find(st, addr, mmu_linear_psize);
+}
+
+static void walk_pagetables(struct pg_state *st)
+{
+	pgd_t *pgd = pgd_offset_k(0UL);
+	unsigned i;
+	unsigned long addr;
+
+	/* Traverse the linux pagetable structure and dump pages that are in
+	 * the hash pagetable.
+	 */
+	for (i = 0; i < PTRS_PER_PGD; i++, pgd++) {
+		addr = VMALLOC_START + i * PGDIR_SIZE;
+		if (!pgd_none(*pgd))
+			/* pgd exists */
+			walk_pud(st, pgd, addr);
+	}
+}
+
+static void walk_vmemmap(struct pg_state *st)
+{
+	struct vmemmap_backing *ptr = vmemmap_list;
+
+	/* Traverse the vmemmaped memory and dump pages that are in the hash
+	 * pagetable.
+	 */
+	while (ptr->list) {
+		hpte_find(st, ptr->virt_addr, mmu_vmemmap_psize);
+		ptr = ptr->list;
+	}
+	seq_puts(st->seq, "---[ vmemmap end ]---\n");
+}
+
+static int ptdump_show(struct seq_file *m, void *v)
+{
+	struct pg_state st = {
+		.seq = m,
+		.start_address = PAGE_OFFSET,
+		.marker = address_markers,
+	};
+	/* Traverse the 0xc, 0xd and 0xf areas of the kernel virtual memory and
+	 * dump pages that are in the hash pagetable.
+	 */
+	walk_linearmapping(&st);
+	walk_pagetables(&st);
+	walk_vmemmap(&st);
+	return 0;
+}
+
+static int ptdump_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, ptdump_show, NULL);
+}
+
+static const struct file_operations ptdump_fops = {
+	.open		= ptdump_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+
+static int ptdump_init(void)
+{
+	struct dentry *debugfs_file;
+
+	debugfs_file = debugfs_create_file("kernel_hash_pagetable", 0400,
+			NULL, NULL, &ptdump_fops);
+	return debugfs_file ? 0 : -ENOMEM;
+}
+device_initcall(ptdump_init);
-- 
2.5.0

Re: [PATCH 1/2] powerpc/pagetable: Add option to dump the linux pagetables

[kbuild test robot]

[-- Attachment #1: Type: text/plain, Size: 1371 bytes --]

Hi Rashmica,

[auto build test ERROR on powerpc/next]
[also build test ERROR on v4.5 next-20160322]
[if your patch is applied to the wrong git tree, please drop us a note to help improving the system]

url:    https://github.com/0day-ci/linux/commits/Rashmica-Gupta/powerpc-pagetable-Add-option-to-dump-the-linux-pagetables/20160322-060934
base:   https://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux.git next
config: powerpc-allyesconfig (attached as .config)
reproduce:
        wget https://git.kernel.org/cgit/linux/kernel/git/wfg/lkp-tests.git/plain/sbin/make.cross -O ~/bin/make.cross
        chmod +x ~/bin/make.cross
        # save the attached .config to linux build tree
        make.cross ARCH=powerpc 

All errors (new ones prefixed by >>):

>> arch/powerpc/mm/dump_linuxpagetables.c:148:11: error: '_PAGE_COMBO' undeclared here (not in a function)
      .mask = _PAGE_COMBO,
              ^

vim +/_PAGE_COMBO +148 arch/powerpc/mm/dump_linuxpagetables.c

   142			.clear	= "        ",
   143		}, {
   144			.mask	= _PAGE_BUSY,
   145			.val	= _PAGE_BUSY,
   146			.set	= "busy",
   147		}, {
 > 148			.mask	= _PAGE_COMBO,
   149			.val	= _PAGE_COMBO,
   150			.set	= "combo",
   151		}, {

---
0-DAY kernel test infrastructure                Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all                   Intel Corporation

[-- Attachment #2: .config.gz --]
[-- Type: application/octet-stream, Size: 47891 bytes --]

Re: [PATCH 2/2] powerpc/pagetable: Add option to dump kernel hashpagetable

[kbuild test robot]

[-- Attachment #1: Type: text/plain, Size: 1687 bytes --]

Hi Rashmica,

[auto build test ERROR on powerpc/next]
[also build test ERROR on v4.5 next-20160322]
[if your patch is applied to the wrong git tree, please drop us a note to help improving the system]

url:    https://github.com/0day-ci/linux/commits/Rashmica-Gupta/powerpc-pagetable-Add-option-to-dump-the-linux-pagetables/20160322-060934
base:   https://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux.git next
config: powerpc-allyesconfig (attached as .config)
reproduce:
        wget https://git.kernel.org/cgit/linux/kernel/git/wfg/lkp-tests.git/plain/sbin/make.cross -O ~/bin/make.cross
        chmod +x ~/bin/make.cross
        # save the attached .config to linux build tree
        make.cross ARCH=powerpc 

All errors (new ones prefixed by >>):

   arch/powerpc/mm/dump_hashpagetable.c: In function 'walk_pte':
>> arch/powerpc/mm/dump_hashpagetable.c:363:18: error: '_PAGE_COMBO' undeclared (first use in this function)
      if (((pteval & _PAGE_COMBO) == _PAGE_COMBO) ||
                     ^
   arch/powerpc/mm/dump_hashpagetable.c:363:18: note: each undeclared identifier is reported only once for each function it appears in

vim +/_PAGE_COMBO +363 arch/powerpc/mm/dump_hashpagetable.c

   357			if (addr < VMALLOC_END)
   358				psize = mmu_vmalloc_psize;
   359			else
   360				psize = mmu_io_psize;
   361	
   362			/* check for secret 4K mappings */
 > 363			if (((pteval & _PAGE_COMBO) == _PAGE_COMBO) ||
   364				((pteval & _PAGE_4K_PFN) == _PAGE_4K_PFN))
   365				psize = mmu_io_psize;
   366	

---
0-DAY kernel test infrastructure                Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all                   Intel Corporation

[-- Attachment #2: .config.gz --]
[-- Type: application/octet-stream, Size: 47891 bytes --]

[PATCH 2/2] powerpc/pagetable: Add option to dump kernel hashpagetable

[Rashmica Gupta]

Useful to be able to dump the kernel hash page table to check
which pages are hashed along with their sizes and other details.

Add a debugfs file to check the page tables. To use this the PPC_PTDUMP
config option must be selected.

Signed-off-by: Rashmica Gupta <rashmicy@gmail.com>
---
 arch/powerpc/mm/Makefile             |   3 +-
 arch/powerpc/mm/dump_hashpagetable.c | 501 +++++++++++++++++++++++++++++++++++
 2 files changed, 503 insertions(+), 1 deletion(-)
 create mode 100644 arch/powerpc/mm/dump_hashpagetable.c

diff --git a/arch/powerpc/mm/Makefile b/arch/powerpc/mm/Makefile
index 6935c6204fbc..0bd48b345ff3 100644
--- a/arch/powerpc/mm/Makefile
+++ b/arch/powerpc/mm/Makefile
@@ -41,4 +41,5 @@ obj-$(CONFIG_NOT_COHERENT_CACHE) += dma-noncoherent.o
 obj-$(CONFIG_HIGHMEM)		+= highmem.o
 obj-$(CONFIG_PPC_COPRO_BASE)	+= copro_fault.o
 obj-$(CONFIG_SPAPR_TCE_IOMMU)	+= mmu_context_iommu.o
-obj-$(CONFIG_PPC_PTDUMP)	+= dump_linuxpagetables.o
+obj-$(CONFIG_PPC_PTDUMP)	+= dump_linuxpagetables.o \
+				   dump_hashpagetable.o
diff --git a/arch/powerpc/mm/dump_hashpagetable.c b/arch/powerpc/mm/dump_hashpagetable.c
new file mode 100644
index 000000000000..fbd746ffe26a
--- /dev/null
+++ b/arch/powerpc/mm/dump_hashpagetable.c
@@ -0,0 +1,501 @@
+/*
+ * Copyright 2016, Rashmica Gupta, IBM Corp.
+ *
+ * This traverses the kernel virtual memory and dumps the pages that are in
+ * the hash pagetable, along with their flags to
+ * /sys/kernel/debug/kernel_hash_pagetable.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; version 2
+ * of the License.
+ */
+#include <linux/debugfs.h>
+#include <linux/fs.h>
+#include <linux/io.h>
+#include <linux/mm.h>
+#include <linux/sched.h>
+#include <linux/seq_file.h>
+#include <asm/fixmap.h>
+#include <asm/pgtable.h>
+#include <linux/const.h>
+#include <asm/page.h>
+#include <asm/pgalloc.h>
+#include <asm/plpar_wrappers.h>
+#include <linux/memblock.h>
+#include <asm/firmware.h>
+
+struct addr_marker {
+	unsigned long start_address;
+	const char *name;
+};
+
+static struct addr_marker address_markers[] = {
+	{ PAGE_OFFSET,		"Start of kernel VM"},
+	{ VMALLOC_START,	"vmalloc() Area" },
+	{ VMALLOC_END,		"vmalloc() End" },
+	{ ISA_IO_BASE,		"isa I/O start" },
+	{ ISA_IO_END,		"isa I/O end" },
+	{ PHB_IO_BASE,		"phb I/O start" },
+	{ PHB_IO_END,		"phb I/O end" },
+	{ IOREMAP_BASE,		"I/O remap start" },
+	{ IOREMAP_END,		"I/O remap end" },
+	{ VMEMMAP_BASE,		"vmemmap start" },
+	{ -1,			NULL },
+};
+
+struct pg_state {
+	struct seq_file *seq;
+	const struct addr_marker *marker;
+	unsigned long start_address;
+	unsigned level;
+	u64 current_flags;
+};
+
+struct flag_info {
+	u64		mask;
+	u64		val;
+	const char	*set;
+	const char	*clear;
+	bool		is_val;
+};
+
+static const struct flag_info v_flag_array[] = {
+	{
+		.mask   = SLB_VSID_B,
+		.val    = SLB_VSID_B_256M,
+		.set    = "ssize: 256M",
+		.clear  = "ssize: 1T  ",
+	}, {
+		.mask	= HPTE_V_SECONDARY,
+		.val	= HPTE_V_SECONDARY,
+		.set	= "secondary",
+		.clear	= "primary  ",
+	}, {
+		.mask	= HPTE_V_VALID,
+		.val	= HPTE_V_VALID,
+		.set	= "valid  ",
+		.clear	= "invalid",
+	}, {
+		.mask	= HPTE_V_BOLTED,
+		.val	= HPTE_V_BOLTED,
+		.set	= "bolted",
+		.clear	= "",
+	}
+};
+
+static const struct flag_info r_flag_array[] = {
+	{
+		.mask	= HPTE_R_PP0 | HPTE_R_PP,
+		.val	= PP_RWXX,
+		.set	= "prot:RW--",
+	}, {
+		.mask	= HPTE_R_PP0 | HPTE_R_PP,
+		.val	= PP_RWRX,
+		.set	= "prot:RWR-",
+	}, {
+		.mask	= HPTE_R_PP0 | HPTE_R_PP,
+		.val	= PP_RWRW,
+		.set	= "prot:RWRW",
+	}, {
+		.mask	= HPTE_R_PP0 | HPTE_R_PP,
+		.val	= PP_RXRX,
+		.set	= "prot:R-R-",
+	}, {
+		.mask	= HPTE_R_PP0 | HPTE_R_PP,
+		.val	= PP_RXXX,
+		.set	= "prot:R---",
+	}, {
+		.mask	= HPTE_R_KEY_HI | HPTE_R_KEY_LO,
+		.val	= HPTE_R_KEY_HI | HPTE_R_KEY_LO,
+		.set	= "key",
+		.clear	= "",
+		.is_val = true,
+	}, {
+		.mask	= HPTE_R_R,
+		.val	= HPTE_R_R,
+		.set	= "ref",
+		.clear	= "   ",
+	}, {
+		.mask	= HPTE_R_C,
+		.val	= HPTE_R_C,
+		.set	= "changed",
+		.clear	= "       ",
+	}, {
+		.mask	= HPTE_R_N,
+		.val	= HPTE_R_N,
+		.set	= "no execute",
+	}, {
+		.mask	= HPTE_R_WIMG,
+		.val	= HPTE_R_W,
+		.set	= "writethru",
+	}, {
+		.mask	= HPTE_R_WIMG,
+		.val	= HPTE_R_I,
+		.set	= "no cache",
+	}, {
+		.mask	= HPTE_R_WIMG,
+		.val	= HPTE_R_G,
+		.set	= "guarded",
+	}
+};
+
+static void dump_flag_info(struct pg_state *st, const struct flag_info
+		*flag, u64 pte, int num)
+{
+	unsigned i;
+
+	for (i = 0; i < num; i++, flag++) {
+		const char *s = NULL;
+
+		if (flag->is_val) {
+			seq_printf(st->seq, "  %s:%llx", flag->set, pte &
+					flag->val);
+		} else {
+			if ((pte & flag->mask) == flag->val)
+				s = flag->set;
+			else
+				s = flag->clear;
+			if (s)
+				seq_printf(st->seq, "  %s", s);
+		}
+	}
+}
+
+static void dump_hpte_info(struct pg_state *st, unsigned long ea, u64 v, u64 r,
+		unsigned long rpn, int bps, int aps, unsigned long lp)
+{
+	static const char units[] = "BKMGTPE";
+	const char *unit = units;
+
+	while (ea >= st->marker[1].start_address) {
+		st->marker++;
+		seq_printf(st->seq, "---[ %s ]---\n", st->marker->name);
+	}
+	seq_printf(st->seq, "0x%lx:\t", ea);
+	seq_printf(st->seq, "AVPN:%llx\t", HPTE_V_AVPN_VAL(v));
+	dump_flag_info(st, v_flag_array, v, ARRAY_SIZE(v_flag_array));
+	seq_printf(st->seq, "  rpn: %lx\t", rpn);
+	dump_flag_info(st, r_flag_array, r, ARRAY_SIZE(r_flag_array));
+
+	while (bps > 9 && unit[1]) {
+		bps -= 10;
+		unit++;
+	}
+	seq_printf(st->seq, "  base_ps: %i%c\t", 1<<bps, *unit);
+	unit = units;
+	while (aps > 9 && unit[1]) {
+		aps -= 10;
+		unit++;
+	}
+	seq_printf(st->seq, "actual_ps: %i%c", 1<<aps, *unit);
+	if (aps != 2)
+		seq_printf(st->seq, "\tLP enc: %lx", lp);
+	seq_puts(st->seq, "\n");
+}
+
+static int native_find(unsigned long ea, int psize, bool primary, u64 *v, u64
+		*r)
+{
+	struct hash_pte *hptep;
+	unsigned long hash, vsid, vpn, hpte_group, want_v, hpte_v;
+	int i, ssize = mmu_kernel_ssize;
+	unsigned long shift = mmu_psize_defs[psize].shift;
+
+	/* calculate hash */
+	vsid = get_kernel_vsid(ea, ssize);
+	vpn  = hpt_vpn(ea, vsid, ssize);
+	hash = hpt_hash(vpn, shift, ssize);
+	want_v = hpte_encode_avpn(vpn, psize, ssize);
+
+	/* to check in the secondary hash table, we invert the hash */
+	if (!primary)
+		hash = ~hash;
+	hpte_group = (hash & htab_hash_mask) * HPTES_PER_GROUP;
+	for (i = 0; i < HPTES_PER_GROUP; i++) {
+		hptep = htab_address + hpte_group;
+		hpte_v = be64_to_cpu(hptep->v);
+
+		if (HPTE_V_COMPARE(hpte_v, want_v) && (hpte_v & HPTE_V_VALID)) {
+			/* HPTE matches */
+			*v = be64_to_cpu(hptep->v);
+			*r = be64_to_cpu(hptep->r);
+			return 0;
+		}
+		++hpte_group;
+	}
+	return -1;
+}
+
+static int pseries_find(unsigned long ea, int psize, bool primary, u64 *v, u64
+		*r)
+{
+	struct hash_pte ptes[4];
+	unsigned long vsid, vpn, hash, hpte_group, want_v;
+	int i, j, ssize = mmu_kernel_ssize;
+	long lpar_rc = 0;
+	unsigned long shift = mmu_psize_defs[psize].shift;
+
+	/* calculate hash */
+	vsid = get_kernel_vsid(ea, ssize);
+	vpn  = hpt_vpn(ea, vsid, ssize);
+	hash = hpt_hash(vpn, shift, ssize);
+	want_v = hpte_encode_avpn(vpn, psize, ssize);
+
+	/* to check in the secondary hash table, we invert the hash */
+	if (!primary)
+		hash = ~hash;
+	hpte_group = ((hash & htab_hash_mask) * HPTES_PER_GROUP) & ~0x7UL;
+	/* see if we can find an entry in the hpte with this hash */
+	for (i = 0; i < HPTES_PER_GROUP; i += 4, hpte_group += 4) {
+		lpar_rc = plpar_pte_read_4(0, hpte_group, (void *)ptes);
+
+		if (lpar_rc != H_SUCCESS)
+			continue;
+		for (j = 0; j < 4; j++) {
+			if (HPTE_V_COMPARE(ptes[j].v, want_v) &&
+					(ptes[j].v & HPTE_V_VALID)) {
+				/* HPTE matches */
+				*v = ptes[j].v;
+				*r = ptes[j].r;
+				return 0;
+			}
+		}
+	}
+	return -1;
+}
+
+static void decode_r(int bps, unsigned long r, unsigned long *rpn, int *aps,
+		unsigned long *lp_bits)
+{
+	struct mmu_psize_def entry;
+	unsigned long arpn, mask, lp;
+	int penc = -2, idx = 0, shift;
+
+	/*.
+	 * The LP field has 8 bits. Depending on the actual page size, some of
+	 * these bits are concatenated with the APRN to get the RPN. The rest
+	 * of the bits in the LP field is the LP value and is an encoding for
+	 * the base page size and the actual page size.
+	 *
+	 *  -	find the mmu entry for our base page size
+	 *  -	go through all page encodings and use the associated mask to
+	 *	find an encoding that matches our encoding in the LP field.
+	 */
+	arpn = (r & HPTE_R_RPN) >> HPTE_R_RPN_SHIFT;
+	lp = arpn & 0xff;
+
+	entry = mmu_psize_defs[bps];
+	while (idx < MMU_PAGE_COUNT) {
+		penc = entry.penc[idx];
+		if ((penc != -1) && (mmu_psize_defs[idx].shift)) {
+			shift = mmu_psize_defs[idx].shift -  HPTE_R_RPN_SHIFT;
+			mask = (0x1 << (shift)) - 1;
+			if ((lp & mask) == penc) {
+				*aps = mmu_psize_to_shift(idx);
+				*lp_bits = lp & mask;
+				*rpn = arpn >> shift;
+				return;
+			}
+		}
+		idx++;
+	}
+}
+
+static unsigned long hpte_find(struct pg_state *st, unsigned long ea, int psize)
+{
+	unsigned long slot;
+	u64 v  = 0, r = 0;
+	unsigned long rpn, lp_bits;
+	int base_psize = 0, actual_psize = 0;
+
+	if (ea <= PAGE_OFFSET)
+		return -1;
+
+	/* Look in primary table */
+	if (firmware_has_feature(FW_FEATURE_LPAR))
+		slot = pseries_find(ea, psize, true, &v, &r);
+	else
+		slot = native_find(ea, psize, true, &v, &r);
+
+	/* Look in secondary table */
+	if (slot == -1) {
+		if (firmware_has_feature(FW_FEATURE_LPAR))
+			slot = pseries_find(ea, psize, false, &v, &r);
+		else
+			slot = native_find(ea, psize, false, &v, &r);
+	}
+
+	/* No entry found */
+	if (slot == -1)
+		return -1;
+
+	/* We found an entry in the hash page table:
+	 *  - check that this has the same base page
+	 *  - find the actual page size
+	 *  - find the RPN
+	 */
+	base_psize = mmu_psize_to_shift(psize);
+
+	if ((v & HPTE_V_LARGE) == HPTE_V_LARGE) {
+		decode_r(psize, r, &rpn, &actual_psize, &lp_bits);
+	} else {
+		/* 4K actual page size */
+		actual_psize = 12;
+		rpn = (r & HPTE_R_RPN) >> HPTE_R_RPN_SHIFT;
+		/* In this case there are no LP bits */
+		lp_bits = -1;
+	}
+	/* We didn't find a matching encoding, so the PTE we found isn't for
+	 * this address.
+	 */
+	if (actual_psize == -1)
+		return -1;
+
+	dump_hpte_info(st, ea, v, r, rpn, base_psize, actual_psize, lp_bits);
+	return 0;
+}
+
+static void walk_pte(struct pg_state *st, pmd_t *pmd, unsigned long start)
+{
+	pte_t *pte = pte_offset_kernel(pmd, 0);
+	unsigned long addr, pteval, psize;
+	int i, status;
+
+	for (i = 0; i < PTRS_PER_PTE; i++, pte++) {
+		addr = start + i * PAGE_SIZE;
+		pteval = pte_val(*pte);
+
+		if (addr < VMALLOC_END)
+			psize = mmu_vmalloc_psize;
+		else
+			psize = mmu_io_psize;
+#ifdef CONFIG_PPC_64K_PAGES
+		/* check for secret 4K mappings */
+		if (((pteval & _PAGE_COMBO) == _PAGE_COMBO) ||
+			((pteval & _PAGE_4K_PFN) == _PAGE_4K_PFN))
+			psize = mmu_io_psize;
+#endif
+		/* check for hashpte */
+		status = hpte_find(st, addr, psize);
+
+		if (((pteval & _PAGE_HASHPTE) != _PAGE_HASHPTE)
+				&& (status != -1)) {
+		/* found a hpte that is not in the linux page tables */
+			seq_printf(st->seq, "page probably bolted before linux"
+				" pagetables were set: addr:%lx, pteval:%lx\n",
+				addr, pteval);
+		}
+	}
+}
+
+static void walk_pmd(struct pg_state *st, pud_t *pud, unsigned long start)
+{
+	pmd_t *pmd = pmd_offset(pud, 0);
+	unsigned long addr;
+	unsigned i;
+
+	for (i = 0; i < PTRS_PER_PMD; i++, pmd++) {
+		addr = start + i * PMD_SIZE;
+		if (!pmd_none(*pmd))
+			/* pmd exists */
+			walk_pte(st, pmd, addr);
+	}
+}
+
+static void walk_pud(struct pg_state *st, pgd_t *pgd, unsigned long start)
+{
+	pud_t *pud = pud_offset(pgd, 0);
+	unsigned long addr;
+	unsigned i;
+
+	for (i = 0; i < PTRS_PER_PUD; i++, pud++) {
+		addr = start + i * PUD_SIZE;
+		if (!pud_none(*pud))
+			/* pud exists */
+			walk_pmd(st, pud, addr);
+	}
+}
+
+static void walk_linearmapping(struct pg_state *st)
+{
+	unsigned long addr;
+
+	/* Traverse the linear mapping section of virtual memory and dump pages
+	 * that are in the hash pagetable.
+	 */
+unsigned long psize = 1 << mmu_psize_defs[mmu_linear_psize].shift;
+	for (addr = PAGE_OFFSET; addr < PAGE_OFFSET +
+			memblock_phys_mem_size(); addr += psize)
+		hpte_find(st, addr, mmu_linear_psize);
+}
+
+static void walk_pagetables(struct pg_state *st)
+{
+	pgd_t *pgd = pgd_offset_k(0UL);
+	unsigned i;
+	unsigned long addr;
+
+	/* Traverse the linux pagetable structure and dump pages that are in
+	 * the hash pagetable.
+	 */
+	for (i = 0; i < PTRS_PER_PGD; i++, pgd++) {
+		addr = VMALLOC_START + i * PGDIR_SIZE;
+		if (!pgd_none(*pgd))
+			/* pgd exists */
+			walk_pud(st, pgd, addr);
+	}
+}
+
+static void walk_vmemmap(struct pg_state *st)
+{
+	struct vmemmap_backing *ptr = vmemmap_list;
+
+	/* Traverse the vmemmaped memory and dump pages that are in the hash
+	 * pagetable.
+	 */
+	while (ptr->list) {
+		hpte_find(st, ptr->virt_addr, mmu_vmemmap_psize);
+		ptr = ptr->list;
+	}
+	seq_puts(st->seq, "---[ vmemmap end ]---\n");
+}
+
+static int ptdump_show(struct seq_file *m, void *v)
+{
+	struct pg_state st = {
+		.seq = m,
+		.start_address = PAGE_OFFSET,
+		.marker = address_markers,
+	};
+	/* Traverse the 0xc, 0xd and 0xf areas of the kernel virtual memory and
+	 * dump pages that are in the hash pagetable.
+	 */
+	walk_linearmapping(&st);
+	walk_pagetables(&st);
+	walk_vmemmap(&st);
+	return 0;
+}
+
+static int ptdump_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, ptdump_show, NULL);
+}
+
+static const struct file_operations ptdump_fops = {
+	.open		= ptdump_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+
+static int ptdump_init(void)
+{
+	struct dentry *debugfs_file;
+
+	debugfs_file = debugfs_create_file("kernel_hash_pagetable", 0400,
+			NULL, NULL, &ptdump_fops);
+	return debugfs_file ? 0 : -ENOMEM;
+}
+device_initcall(ptdump_init);
-- 
2.5.0

Re: [PATCH 1/2] powerpc/pagetable: Add option to dump the linux pagetables

[Balbir Singh]

On 22/03/16 09:04, Rashmica Gupta wrote:
> Useful to be able to dump the kernels page tables to check permissions
> and memory types - derived from arm64's implementation.
>
> Add a debugfs file to check the page tables. To use this the PPC_PTDUMP
> config option must be selected.
>
> Signed-off-by: Rashmica Gupta <rashmicy@gmail.com>
> ---
>  arch/powerpc/Kconfig.debug             |  12 ++
>  arch/powerpc/mm/Makefile               |   1 +
>  arch/powerpc/mm/dump_linuxpagetables.c | 377 +++++++++++++++++++++++++++++++++
>  3 files changed, 390 insertions(+)
>  create mode 100644 arch/powerpc/mm/dump_linuxpagetables.c
>
> diff --git a/arch/powerpc/Kconfig.debug b/arch/powerpc/Kconfig.debug
> index 638f9ce740f5..26a60effea1a 100644
> --- a/arch/powerpc/Kconfig.debug
> +++ b/arch/powerpc/Kconfig.debug
> @@ -344,4 +344,16 @@ config FAIL_IOMMU
>  
>  	  If you are unsure, say N.
>  
> +config PPC_PTDUMP
> +        bool "Export kernel pagetable layout to userspace via debugfs"
> +        depends on DEBUG_KERNEL
> +        select DEBUG_FS
> +        help
> +	  This option exports the state of the kernel pagetables to a
> +	  debugfs file. This is only useful for kernel developers who are
> +	  working in architecture specific areas of the kernel - probably
> +	  not a good idea to enable this feature in a production kernel.
> +
> +	  If you are unsure, say N.
> +

Some minor comments below, but otherwise

Acked-by: Balbir Singh <bsingharora@gmail.com>

>  endmenu
> diff --git a/arch/powerpc/mm/Makefile b/arch/powerpc/mm/Makefile
> index adfee3f1aeb9..6935c6204fbc 100644
> --- a/arch/powerpc/mm/Makefile
> +++ b/arch/powerpc/mm/Makefile
> @@ -41,3 +41,4 @@ obj-$(CONFIG_NOT_COHERENT_CACHE) += dma-noncoherent.o
>  obj-$(CONFIG_HIGHMEM)		+= highmem.o
>  obj-$(CONFIG_PPC_COPRO_BASE)	+= copro_fault.o
>  obj-$(CONFIG_SPAPR_TCE_IOMMU)	+= mmu_context_iommu.o
> +obj-$(CONFIG_PPC_PTDUMP)	+= dump_linuxpagetables.o
> diff --git a/arch/powerpc/mm/dump_linuxpagetables.c b/arch/powerpc/mm/dump_linuxpagetables.c
> new file mode 100644
> index 000000000000..f97fbfdac4b9
> --- /dev/null
> +++ b/arch/powerpc/mm/dump_linuxpagetables.c
> @@ -0,0 +1,377 @@
> +/*
> + * Copyright 2016, Rashmica Gupta, IBM Corp.
> + *
> + * This traverses the kernel pagetables and dumps the
> + * information about the used sections of memory to
> + * /sys/kernel/debug/kernel_pagetables.
> + *
> + * Derived from the arm64 implementation:
> + * Copyright (c) 2014, The Linux Foundation, Laura Abbott.
> + * (C) Copyright 2008 Intel Corporation, Arjan van de Ven.
> + *
> + * This program is free software; you can redistribute it and/or
> + * modify it under the terms of the GNU General Public License
> + * as published by the Free Software Foundation; version 2
> + * of the License.
> + */
> +#include <linux/debugfs.h>
> +#include <linux/fs.h>
> +#include <linux/io.h>
> +#include <linux/mm.h>
> +#include <linux/sched.h>
> +#include <linux/seq_file.h>
> +#include <asm/fixmap.h>
> +#include <asm/pgtable.h>
> +#include <linux/const.h>
> +#include <asm/page.h>
> +#include <asm/pgalloc.h>
> +
> +struct addr_marker {
> +	unsigned long start_address;
> +	const char *name;
> +};
> +
> +static struct addr_marker address_markers[] = {
> +	{ VMALLOC_START,	"vmalloc() Area" },
> +	{ VMALLOC_END,		"vmalloc() End" },
> +	{ ISA_IO_BASE,		"isa I/O start" },
> +	{ ISA_IO_END,		"isa I/O end" },
> +	{ PHB_IO_BASE,		"phb I/O start" },
> +	{ PHB_IO_END,		"phb I/O end" },
> +	{ IOREMAP_BASE,		"I/O remap start" },
> +	{ IOREMAP_END,		"I/O remap end" },
> +	{ -1,			NULL },
> +};
> +
> +/*
> + * To visualise what is happening,
> + *
> + *  - PTRS_PER_P** = how many entries there are in the corresponding P**
> + *  - P**_SHIFT = how many bits of the address we use to index into the
> + * corresponding P**
> + *  - P**_SIZE is how much memory we can access through the table - not the
> + * size of the table itself.
> + * P**={PGD, PUD, PMD, PTE}
> + *
> + *
> + * Each entry of the PGD points to a PUD. Each entry of a PUD points to a
> + * PMD. Each entry of a PMD points to a PTE. And every PTE entry points to
> + * a page.
> + *
> + * In the case where there are only 3 levels, the PUD is folded into the
> + * PGD: every PUD has only one entry which points to the PMD.
> + *
> + * The page dumper groups page table entries of the same type into a single
> + * description. It uses pg_state to track the range information while
> + * iterating over the PTE entries. When the continuity is broken it then
> + * dumps out a description of the range - ie PTEs that are virtually contiguous
> + * with the same PTE flags are chunked together. This is to make it clear how
> + * different areas of the kernel virtual memory are used.
> + *
> + */
> +struct pg_state {
> +	struct seq_file *seq;
> +	const struct addr_marker *marker;
> +	unsigned long start_address;
> +	unsigned level;
> +	u64 current_flags;
> +};
> +
> +struct flag_info {
> +	u64		mask;
> +	u64		val;
> +	const char	*set;
> +	const char	*clear;
> +};
> +
> +static const struct flag_info flag_array[] = {
> +	{
> +		.mask	= _PAGE_USER,
> +		.val	= _PAGE_USER,
> +		.set	= "user",
> +		.clear	= "    ",
> +	}, {
> +		.mask	= _PAGE_RW,
> +		.val	= _PAGE_RW,
> +		.set	= "rw",
> +		.clear	= "ro",
> +	}, {
> +		.mask	= _PAGE_EXEC,
> +		.val	= _PAGE_EXEC,
> +		.set	= " X ",
Can we make this small "x" to sort of go in line with "rw"/"ro"
> +		.clear	= "   ",
> +	}, {
> +		.mask	= _PAGE_PTE,
> +		.val	= _PAGE_PTE,
> +		.set	= "pte",
> +		.clear	= "   ",
> +	}, {
> +		.mask	= _PAGE_PRESENT,
> +		.val	= _PAGE_PRESENT,
> +		.set	= "present",
> +		.clear	= "       ",
> +	}, {
> +		.mask	= _PAGE_HASHPTE,
> +		.val	= _PAGE_HASHPTE,
> +		.set	= "hpte",
> +		.clear	= "    ",
> +	}, {
> +		.mask	= _PAGE_GUARDED,
> +		.val	= _PAGE_GUARDED,
> +		.set	= "guarded",
> +		.clear	= "       ",
> +	}, {
> +		.mask	= _PAGE_DIRTY,
> +		.val	= _PAGE_DIRTY,
> +		.set	= "dirty",
> +		.clear	= "     ",
> +	}, {
> +		.mask	= _PAGE_ACCESSED,
> +		.val	= _PAGE_ACCESSED,
> +		.set	= "accessed",
> +		.clear	= "        ",
> +	}, {
> +		.mask	= _PAGE_WRITETHRU,
> +		.val	= _PAGE_WRITETHRU,
> +		.set	= "write through",
> +		.clear	= "             ",
> +	}, {
> +		.mask	= _PAGE_NO_CACHE,
> +		.val	= _PAGE_NO_CACHE,
> +		.set	= "no cache",
> +		.clear	= "        ",
> +	}, {
> +		.mask	= _PAGE_BUSY,
> +		.val	= _PAGE_BUSY,
> +		.set	= "busy",
> +	}, {
> +		.mask	= _PAGE_COMBO,
> +		.val	= _PAGE_COMBO,
> +		.set	= "combo",
> +	}, {
> +		.mask	= _PAGE_F_GIX,
> +		.val	= _PAGE_F_GIX,
> +		.set	= "f_gix",
Can we use "primary"?
> +	}, {
> +		.mask	= _PAGE_F_SECOND,
> +		.val	= _PAGE_F_SECOND,
> +		.set	= "f_second",
"secondary idx"
> +	}, {
> +		.mask	= _PAGE_SPECIAL,
> +		.val	= _PAGE_SPECIAL,
> +		.set	= "special",
> +	}, {
> +		.mask	= _PAGE_4K_PFN,
> +		.val	= _PAGE_4K_PFN,
> +		.set	= "4K_pfn"
> +	}
> +};
> +
> +struct pgtable_level {
> +	const struct flag_info *flag;
> +	size_t num;
> +	u64 mask;
> +};
> +
> +static struct pgtable_level pg_level[] = {
> +	{
> +	}, { /* pgd */
> +		.flag	= flag_array,
> +		.num	= ARRAY_SIZE(flag_array),
> +	}, { /* pud */
> +		.flag	= flag_array,
> +		.num	= ARRAY_SIZE(flag_array),
> +	}, { /* pmd */
> +		.flag	= flag_array,
> +		.num	= ARRAY_SIZE(flag_array),
> +	}, { /* pte */
> +		.flag	= flag_array,
> +		.num	= ARRAY_SIZE(flag_array),
> +	},
> +};
> +
> +static void dump_flag(struct pg_state *st, const struct flag_info *flag,
> +			size_t num)
> +{
> +	unsigned i;
> +
> +	for (i = 0; i < num; i++, flag++) {
> +		const char *s;
> +
> +		if ((st->current_flags & flag->mask) == flag->val)
> +			s = flag->set;
> +		else
> +			s = flag->clear;
> +
> +		if (s)
> +			seq_printf(st->seq, " %s", s);
> +	}
> +}
> +
> +static void dump_addr(struct pg_state *st, unsigned long addr)
> +{
> +	static const char units[] = "KMGTPE";
> +	const char *unit = units;
> +	unsigned long delta;
> +
> +	seq_printf(st->seq, "0x%016lx-0x%016lx   ", st->start_address, addr-1);
> +	delta = (addr - st->start_address) >> 10;
> +	/* Work out what appropriate unit to use */
> +	while (!(delta & 1023) && unit[1]) {

1023, I think ((1 << 10) - 1) is easier to understand but again you may find a better way of encapsulating this

> +		delta >>= 10;
> +		unit++;
> +	}
> +	seq_printf(st->seq, "%9lu%c", delta, *unit);
> +
> +}
> +
> +static void note_page(struct pg_state *st, unsigned long addr, unsigned level,
> +				u64 val)
> +{
> +	u64 flag = val & pg_level[level].mask;
> +
> +	/* At first no level is set */
> +	if (!st->level) {
> +		st->level = level;
> +		st->current_flags = flag;
> +		st->start_address = addr;
> +		seq_printf(st->seq, "---[ %s ]---\n", st->marker->name);
> +	/*
> +	 * Dump the section of virtual memory when:
> +	 *   - the PTE flags from one entry to the next differs.
> +	 *   - we change levels in the tree.
> +	 *   - the address is in a different section of memory and is thus
> +	 *   used for a different purpose, regardless of the flags.
> +	 */
> +	} else if (flag != st->current_flags || level != st->level ||
> +		   addr >= st->marker[1].start_address) {
> +
> +		/* Check the PTE flags */
> +		if (st->current_flags) {
> +			dump_addr(st, addr);
> +
> +			/* Dump all the flags */
> +			if (pg_level[st->level].flag)
> +				dump_flag(st, pg_level[st->level].flag,
> +					  pg_level[st->level].num);
> +			seq_puts(st->seq, "\n");
> +		}
> +
> +		/* Address indicates we have passed the end of the
> +		 * current section of virtual memory
> +		 */
> +		while (addr >= st->marker[1].start_address) {
> +			st->marker++;
> +			seq_printf(st->seq, "---[ %s ]---\n", st->marker->name);
> +		}
> +		st->start_address = addr;
> +		st->current_flags = flag;
> +		st->level = level;
> +	}
> +}
> +
> +static void walk_pte(struct pg_state *st, pmd_t *pmd, unsigned long start)
> +{
> +	pte_t *pte = pte_offset_kernel(pmd, 0);
> +	unsigned long addr;
> +	unsigned i;
> +
> +	for (i = 0; i < PTRS_PER_PTE; i++, pte++) {
> +		addr = start + i * PAGE_SIZE;
> +		note_page(st, addr, 4, pte_val(*pte));
> +	}
> +}
> +
> +static void walk_pmd(struct pg_state *st, pud_t *pud, unsigned long start)
> +{
> +	pmd_t *pmd = pmd_offset(pud, 0);
> +	unsigned long addr;
> +	unsigned i;
> +
> +	for (i = 0; i < PTRS_PER_PMD; i++, pmd++) {
> +		addr = start + i * PMD_SIZE;
> +		if (!pmd_none(*pmd))
> +			/* pmd exists */
> +			walk_pte(st, pmd, addr);
> +		else
> +			note_page(st, addr, 3, pmd_val(*pmd));
> +	}
> +}
> +
> +static void walk_pud(struct pg_state *st, pgd_t *pgd, unsigned long start)
> +{
> +	pud_t *pud = pud_offset(pgd, 0);
> +	unsigned long addr;
> +	unsigned i;
> +
> +	for (i = 0; i < PTRS_PER_PUD; i++, pud++) {
> +		addr = start + i * PUD_SIZE;
> +		if (!pud_none(*pud))
> +			/* pud exists */
> +			walk_pmd(st, pud, addr);
> +		else
> +			note_page(st, addr, 2, pud_val(*pud));
> +	}
> +}
> +
> +static void walk_pgd(struct pg_state *st, unsigned long start)
> +{
> +	pgd_t *pgd = pgd_offset_k(0UL);
> +	unsigned i;
> +	unsigned long addr;
> +
> +	for (i = 0; i < PTRS_PER_PGD; i++, pgd++) {
> +		addr = start + i * PGDIR_SIZE;
> +		if (!pgd_none(*pgd))
> +			/* pgd exists */
> +			walk_pud(st, pgd, addr);
> +		else
> +			note_page(st, addr, 1, pgd_val(*pgd));
> +	}
> +}
> +
> +static int ptdump_show(struct seq_file *m, void *v)
> +{
> +	struct pg_state st = {
> +		.seq = m,
> +		.start_address = KERN_VIRT_START,
> +		.marker = address_markers,
> +	};
> +	/* Traverse kernel page tables */
> +	walk_pgd(&st, KERN_VIRT_START);
> +	note_page(&st, 0, 0, 0);
> +	return 0;
> +}
> +
> +static int ptdump_open(struct inode *inode, struct file *file)
> +{
> +	return single_open(file, ptdump_show, NULL);
> +}
> +
> +static const struct file_operations ptdump_fops = {
> +	.open		= ptdump_open,
> +	.read		= seq_read,
> +	.llseek		= seq_lseek,
> +	.release	= single_release,
> +};
> +
> +static void build_pgtable_complete_mask(void)
> +{
> +	unsigned i, j;
> +
> +	for (i = 0; i < ARRAY_SIZE(pg_level); i++)
> +		if (pg_level[i].flag)
> +			for (j = 0; j < pg_level[i].num; j++)
> +				pg_level[i].mask |= pg_level[i].flag[j].mask;
> +}
> +
> +static int ptdump_init(void)
> +{
> +	struct dentry *debugfs_file;
> +
> +	build_pgtable_complete_mask();
> +	debugfs_file = debugfs_create_file("kernel_pagetables", 0400, NULL,
> +			NULL, &ptdump_fops);
> +	return debugfs_file ? 0 : -ENOMEM;
> +}
> +device_initcall(ptdump_init);

Re: [PATCH 2/2] powerpc/pagetable: Add option to dump kernel hashpagetable

[Balbir Singh]

On 22/03/16 09:04, Rashmica Gupta wrote:
> Useful to be able to dump the kernel hash page table to check
> which pages are hashed along with their sizes and other details.
>
> Add a debugfs file to check the page tables. To use this the PPC_PTDUMP
> config option must be selected.
>
> Signed-off-by: Rashmica Gupta <rashmicy@gmail.com>
> ---
>  arch/powerpc/mm/Makefile             |   3 +-
>  arch/powerpc/mm/dump_hashpagetable.c | 488 +++++++++++++++++++++++++++++++++++
>  2 files changed, 490 insertions(+), 1 deletion(-)
>  create mode 100644 arch/powerpc/mm/dump_hashpagetable.c
>
> diff --git a/arch/powerpc/mm/Makefile b/arch/powerpc/mm/Makefile
> index 6935c6204fbc..0bd48b345ff3 100644
> --- a/arch/powerpc/mm/Makefile
> +++ b/arch/powerpc/mm/Makefile
> @@ -41,4 +41,5 @@ obj-$(CONFIG_NOT_COHERENT_CACHE) += dma-noncoherent.o
>  obj-$(CONFIG_HIGHMEM)		+= highmem.o
>  obj-$(CONFIG_PPC_COPRO_BASE)	+= copro_fault.o
>  obj-$(CONFIG_SPAPR_TCE_IOMMU)	+= mmu_context_iommu.o
> -obj-$(CONFIG_PPC_PTDUMP)	+= dump_linuxpagetables.o
> +obj-$(CONFIG_PPC_PTDUMP)	+= dump_linuxpagetables.o \
> +				   dump_hashpagetable.o
> diff --git a/arch/powerpc/mm/dump_hashpagetable.c b/arch/powerpc/mm/dump_hashpagetable.c
> new file mode 100644
> index 000000000000..044fd13e4d07
> --- /dev/null
> +++ b/arch/powerpc/mm/dump_hashpagetable.c
> @@ -0,0 +1,488 @@
> +/*
> + * Copyright 2016, Rashmica Gupta, IBM Corp.
> + *
> + * This traverses the kernel virtual memory and dumps the pages that are in
> + * the hash pagetable, along with their flags to
> + * /sys/kernel/debug/kernel_hash_pagetable.
> + *
> + * This program is free software; you can redistribute it and/or
> + * modify it under the terms of the GNU General Public License
> + * as published by the Free Software Foundation; version 2
> + * of the License.
> + */
> +#include <linux/debugfs.h>
> +#include <linux/fs.h>
> +#include <linux/io.h>
> +#include <linux/mm.h>
> +#include <linux/sched.h>
> +#include <linux/seq_file.h>
> +#include <asm/fixmap.h>
> +#include <asm/pgtable.h>
> +#include <linux/const.h>
> +#include <asm/page.h>
> +#include <asm/pgalloc.h>
> +#include <asm/plpar_wrappers.h>
> +#include <linux/memblock.h>
> +#include <asm/firmware.h>
> +
> +struct addr_marker {
> +	unsigned long start_address;
> +	const char *name;
> +};
> +
> +static struct addr_marker address_markers[] = {
> +	{ PAGE_OFFSET,		"Start of kernel VM"},
> +	{ VMALLOC_START,	"vmalloc() Area" },
> +	{ VMALLOC_END,		"vmalloc() End" },
> +	{ ISA_IO_BASE,		"isa I/O start" },
> +	{ ISA_IO_END,		"isa I/O end" },
> +	{ PHB_IO_BASE,		"phb I/O start" },
> +	{ PHB_IO_END,		"phb I/O end" },
> +	{ IOREMAP_BASE,		"I/O remap start" },
> +	{ IOREMAP_END,		"I/O remap end" },
> +	{ VMEMMAP_BASE,		"vmemmap start" },
> +	{ -1,			NULL },
> +};
> +
> +struct pg_state {
> +	struct seq_file *seq;
> +	const struct addr_marker *marker;
> +	unsigned long start_address;
> +	unsigned level;
> +	u64 current_flags;
> +};
> +
> +struct flag_info {
> +	u64		mask;
> +	u64		val;
> +	const char	*set;
> +	const char	*clear;
> +	bool		is_val;
> +};
> +
> +static const struct flag_info v_flag_array[] = {
> +	{
> +		.mask   = SLB_VSID_B,
> +		.val    = SLB_VSID_B_256M,
> +		.set    = "ssize: 256M",
> +		.clear  = "ssize: 1T  ",
> +	}, {
> +		.mask	= HPTE_V_SECONDARY,
> +		.val	= HPTE_V_SECONDARY,
> +		.set	= "secondary",
> +		.clear	= "primary  ",
> +	}, {
> +		.mask	= HPTE_V_VALID,
> +		.val	= HPTE_V_VALID,
> +		.set	= "valid  ",
> +		.clear	= "invalid",
> +	}, {
> +		.mask	= HPTE_V_BOLTED,
> +		.val	= HPTE_V_BOLTED,
> +		.set	= "bolted",
> +		.clear	= "",
> +	}
> +};
> +
> +static const struct flag_info r_flag_array[] = {
> +	{
> +		.mask	= HPTE_R_PP0 | HPTE_R_PP,
> +		.val	= HPTE_R_PP0 | HPTE_R_PP,
> +		.set	= "prot",
> +		.clear	= "",
> +		.is_val = true,
> +	}, {
> +		.mask	= HPTE_R_KEY_HI | HPTE_R_KEY_LO,
> +		.val	= HPTE_R_KEY_HI | HPTE_R_KEY_LO,
> +		.set	= "key",
> +		.clear	= "",
> +		.is_val = true,
> +	}, {
> +		.mask	= HPTE_R_R,
> +		.val	= HPTE_R_R,
> +		.set	= "ref",
> +		.clear	= "   ",
> +	}, {
> +		.mask	= HPTE_R_C,
> +		.val	= HPTE_R_C,
> +		.set	= "changed",
> +		.clear	= "       ",
> +	}, {
> +		.mask	= HPTE_R_N,
> +		.val	= HPTE_R_N,
> +		.set	= "no execute",
> +		.clear	= "",
> +	}, {
> +		.mask	= HPTE_R_WIMG,
> +		.val	= HPTE_R_W,
> +		.set	= "writethru",
> +		.clear	= "",
> +	}, {
> +		.mask	= HPTE_R_WIMG,
> +		.val	= HPTE_R_I,
> +		.set	= "no cache",
> +		.clear	= "",
> +	}, {
> +		.mask	= HPTE_R_WIMG,
> +		.val	= HPTE_R_G,
> +		.set	= "guarded",
> +		.clear	= "",
> +	}
> +};
> +
> +static void dump_flag_info(struct pg_state *st, const struct flag_info
> +		*flag, unsigned long pte, int num)
> +{
> +	unsigned i;
> +
> +	for (i = 0; i < num; i++, flag++) {
> +		const char *s = NULL;
> +
> +		if (flag->is_val) {
> +			seq_printf(st->seq, "  %s:%llx", flag->set, pte &
> +					flag->val);
> +		} else {
> +			if ((pte & flag->mask) == flag->val)
> +				s = flag->set;
> +			else
> +				s = flag->clear;
> +			seq_printf(st->seq, "  %s", s);
> +		}
> +	}
> +}
> +
> +static void dump_hpte_info(struct pg_state *st, unsigned long ea, unsigned long
> +		v, unsigned long r, unsigned long rpn, int bps, int aps,
> +		unsigned long lp)
> +{
> +	static const char units[] = "BKMGTPE";
> +	const char *unit = units;
> +
> +	while (ea >= st->marker[1].start_address) {
> +		st->marker++;
> +		seq_printf(st->seq, "---[ %s ]---\n", st->marker->name);
> +	}
> +	seq_printf(st->seq, "0x%lx:\t", ea);
> +	seq_printf(st->seq, "AVPN:%lx\t", HPTE_V_AVPN_VAL(v));
> +	dump_flag_info(st, v_flag_array, v, ARRAY_SIZE(v_flag_array));
> +	seq_printf(st->seq, "  rpn: %lx\t", rpn);
> +	dump_flag_info(st, r_flag_array, r, ARRAY_SIZE(r_flag_array));
> +
> +	while (bps > 9 && unit[1]) {
> +		bps -= 10;
> +		unit++;
> +	}
> +	seq_printf(st->seq, "base_ps: %i%c\t", 1<<bps, *unit);
> +	while (aps > 9 && unit[1]) {
> +		aps -= 10;
> +		unit++;
> +	}
> +	seq_printf(st->seq, "actual_ps: %i%c", 1<<aps, *unit);
> +	if (lp != -1)
> +		seq_printf(st->seq, "\tLP enc: %lx", lp);
> +	seq_puts(st->seq, "\n");
> +}
> +
> +static int native_find(unsigned long ea, int psize, bool primary, unsigned long
> +		*v, unsigned long *r)
You probably need a #ifdef around this
> +{
> +	struct hash_pte *hptep;
> +	unsigned long hash, vsid, vpn, hpte_group, want_v, hpte_v;
> +	int i, ssize = mmu_kernel_ssize;
> +	unsigned long shift = mmu_psize_defs[psize].shift;
> +
> +	/* calculate hash */
> +	vsid = get_kernel_vsid(ea, ssize);
> +	vpn  = hpt_vpn(ea, vsid, ssize);
> +	hash = hpt_hash(vpn, shift, ssize);
> +	want_v = hpte_encode_avpn(vpn, psize, ssize);
> +
> +	/* to check in the secondary hash table, we invert the hash */
> +	if (!primary)
> +		hash = ~hash;
> +	hpte_group = (hash & htab_hash_mask) * HPTES_PER_GROUP;
> +	for (i = 0; i < HPTES_PER_GROUP; i++) {
> +		hptep = htab_address + hpte_group;
> +		hpte_v = be64_to_cpu(hptep->v);
> +
> +		if (HPTE_V_COMPARE(hpte_v, want_v) && (hpte_v & HPTE_V_VALID)) {
> +			/* HPTE matches */
> +			*v = be64_to_cpu(hptep->v);
> +			*r = be64_to_cpu(hptep->r);
> +			return 0;
> +		}
> +		++hpte_group;
> +	}
> +	return -1;
> +}
> +
> +static int pseries_find(unsigned long ea, int psize, bool primary, unsigned
> +		long *v, unsigned long *r)
> +{
Same here, should compile only if PLATFORM pseries is set, you need an #ifdef around this

> +	struct hash_pte ptes[4];
> +	unsigned long vsid, vpn, hash, hpte_group, want_v;
> +	int i, j, ssize = mmu_kernel_ssize;
> +	long lpar_rc = 0;
> +	unsigned long shift = mmu_psize_defs[psize].shift;
> +
> +	/* calculate hash */
> +	vsid = get_kernel_vsid(ea, ssize);
> +	vpn  = hpt_vpn(ea, vsid, ssize);
> +	hash = hpt_hash(vpn, shift, ssize);
> +	want_v = hpte_encode_avpn(vpn, psize, ssize);
> +
> +	/* to check in the secondary hash table, we invert the hash */
> +	if (!primary)
> +		hash = ~hash;
> +	hpte_group = ((hash & htab_hash_mask) * HPTES_PER_GROUP) & ~0x7UL;
> +	/* see if we can find an entry in the hpte with this hash */
> +	for (i = 0; i < HPTES_PER_GROUP; i += 4, hpte_group += 4) {
> +		lpar_rc = plpar_pte_read_4(0, hpte_group, (void *)ptes);
> +
> +		if (lpar_rc != H_SUCCESS)
> +			continue;
> +		for (j = 0; j < 4; j++) {
> +			if (HPTE_V_COMPARE(ptes[j].v, want_v) &&
> +					(ptes[j].v & HPTE_V_VALID)) {
> +				/* HPTE matches */
> +				*v = ptes[j].v;
> +				*r = ptes[j].r;
> +				return 0;
> +			}
> +		}
> +	}
> +	return -1;
> +}
> +
> +static void decode_r(int bps, unsigned long r, unsigned long *rpn, int *aps,
> +		unsigned long *lp_bits)
> +{
> +	struct mmu_psize_def entry;
> +	unsigned long arpn, mask, lp;
> +	int penc = -2, idx = 0, shift;
> +
> +	/*.
> +	 * The LP field has 8 bits. Depending on the actual page size, some of
> +	 * these bits are concatenated with the APRN to get the RPN. The rest
> +	 * of the bits in the LP field is the LP value and is an encoding for
> +	 * the base page size and the actual page size.
> +	 *
> +	 *  -	find the mmu entry for our base page size
> +	 *  -	go through all page encodings and use the associated mask to
> +	 *	find an encoding that matches our encoding in the LP field.
> +	 */
> +	arpn = (r & HPTE_R_RPN) >> HPTE_R_RPN_SHIFT;
> +	lp = arpn & 0xff;
> +
> +	entry = mmu_psize_defs[bps];
> +	while (idx < MMU_PAGE_COUNT) {
> +		penc = entry.penc[idx];
> +		if ((penc != -1) && (mmu_psize_defs[idx].shift)) {
> +			shift = mmu_psize_defs[idx].shift -  HPTE_R_RPN_SHIFT;
> +			mask = (0x1 << (shift)) - 1;
> +			if ((lp & mask) == penc) {
> +				*aps = mmu_psize_to_shift(idx);
> +				*lp_bits = lp & mask;
> +				*rpn = arpn >> shift;
> +				return;
> +			}
> +		}
> +		idx++;
> +	}
> +}
> +
> +static unsigned long hpte_find(struct pg_state *st, unsigned long ea, int psize)
> +{
> +	unsigned long slot;
> +	unsigned long v  = 0, r = 0, rpn, lp_bits;
> +	int base_psize = 0, actual_psize = 0;
> +
> +	if (ea <= PAGE_OFFSET)
> +		return -1;
> +
> +	/* Look in primary table */
> +	if (firmware_has_feature(FW_FEATURE_LPAR))
> +		slot = pseries_find(ea, psize, true, &v, &r);
> +	else
> +		slot = native_find(ea, psize, true, &v, &r);
> +
> +	/* Look in secondary table */
> +	if (slot == -1) {
> +		if (firmware_has_feature(FW_FEATURE_LPAR))
> +			slot = pseries_find(ea, psize, false, &v, &r);
> +		else
> +			slot = native_find(ea, psize, false, &v, &r);
> +	}
> +
> +	/* No entry found */
> +	if (slot == -1)
> +		return -1;
> +
> +	/* We found an entry in the hash page table:
> +	 *  - check that this has the same base page
> +	 *  - find the actual page size
> +	 *  - find the RPN
> +	 */
> +	base_psize = mmu_psize_to_shift(psize);
> +
> +	if ((v & HPTE_V_LARGE) == HPTE_V_LARGE) {
> +		decode_r(psize, r, &rpn, &actual_psize, &lp_bits);
> +	} else {
> +		/* 4K actual page size */
> +		actual_psize = 12;
> +		rpn = (r & HPTE_R_RPN) >> HPTE_R_RPN_SHIFT;
> +		/* In this case there are no LP bits */
> +		lp_bits = -1;
> +	}
> +	/* We didn't find a matching encoding, so the PTE we found isn't for
> +	 * this address.
> +	 */
> +	if (actual_psize == -1)
> +		return -1;
> +
> +	dump_hpte_info(st, ea, v, r, rpn, base_psize, actual_psize, lp_bits);
> +	return 0;
> +}
> +
> +static void walk_pte(struct pg_state *st, pmd_t *pmd, unsigned long start)
> +{
> +	pte_t *pte = pte_offset_kernel(pmd, 0);
> +	unsigned long addr, pteval, psize;
> +	int i, status;
> +
> +	for (i = 0; i < PTRS_PER_PTE; i++, pte++) {
> +		addr = start + i * PAGE_SIZE;
> +		pteval = pte_val(*pte);
> +
> +		if (addr < VMALLOC_END)
> +			psize = mmu_vmalloc_psize;
> +		else
> +			psize = mmu_io_psize;
> +
> +		/* check for secret 4K mappings */
> +		if (((pteval & _PAGE_COMBO) == _PAGE_COMBO) ||
> +			((pteval & _PAGE_4K_PFN) == _PAGE_4K_PFN))
> +			psize = mmu_io_psize;
> +
> +		/* check for hashpte */
> +		status = hpte_find(st, addr, psize);
> +
> +		if (((pteval & _PAGE_HASHPTE) != _PAGE_HASHPTE)
> +				&& (status != -1)) {
> +		/* found a hpte that is not in the linux page tables */
> +			seq_printf(st->seq, "page probably bolted before linux"
> +				" pagetables were set: addr:%lx, pteval:%lx\n",
> +				addr, pteval);
> +		}
> +	}
> +}
> +
> +static void walk_pmd(struct pg_state *st, pud_t *pud, unsigned long start)
> +{
> +	pmd_t *pmd = pmd_offset(pud, 0);
> +	unsigned long addr;
> +	unsigned i;
> +
> +	for (i = 0; i < PTRS_PER_PMD; i++, pmd++) {
> +		addr = start + i * PMD_SIZE;
> +		if (!pmd_none(*pmd))
> +			/* pmd exists */
> +			walk_pte(st, pmd, addr);
> +	}
> +}
> +
> +static void walk_pud(struct pg_state *st, pgd_t *pgd, unsigned long start)
> +{
> +	pud_t *pud = pud_offset(pgd, 0);
> +	unsigned long addr;
> +	unsigned i;
> +
> +	for (i = 0; i < PTRS_PER_PUD; i++, pud++) {
> +		addr = start + i * PUD_SIZE;
> +		if (!pud_none(*pud))
> +			/* pud exists */
> +			walk_pmd(st, pud, addr);
> +	}
> +}
> +
> +static void walk_linearmapping(struct pg_state *st)
> +{
> +	unsigned long addr;
> +
> +	/* Traverse the linear mapping section of virtual memory and dump pages
> +	 * that are in the hash pagetable.
> +	 */
> +	for (addr = PAGE_OFFSET; addr < PAGE_OFFSET +
> +			memblock_phys_mem_size(); addr += PAGE_SIZE)
> +		hpte_find(st, addr, mmu_linear_psize);
> +}
> +
> +static void walk_pagetables(struct pg_state *st)
> +{
> +	pgd_t *pgd = pgd_offset_k(0UL);
> +	unsigned i;
> +	unsigned long addr;
> +
> +	/* Traverse the linux pagetable structure and dump pages that are in
> +	 * the hash pagetable.
> +	 */
> +	for (i = 0; i < PTRS_PER_PGD; i++, pgd++) {
> +		addr = VMALLOC_START + i * PGDIR_SIZE;
> +		if (!pgd_none(*pgd))
> +			/* pgd exists */
> +			walk_pud(st, pgd, addr);
> +	}
> +}
> +
> +static void walk_vmemmap(struct pg_state *st)
> +{
> +	struct vmemmap_backing *ptr = vmemmap_list;
> +
> +	/* Traverse the vmemmaped memory and dump pages that are in the hash
> +	 * pagetable.
> +	 */
> +	while (ptr->list) {
> +		hpte_find(st, ptr->virt_addr, mmu_vmemmap_psize);
> +		ptr = ptr->list;
> +	}
> +	seq_puts(st->seq, "---[ vmemmap end ]---\n");
> +}
> +
> +static int ptdump_show(struct seq_file *m, void *v)
> +{
> +	struct pg_state st = {
> +		.seq = m,
> +		.start_address = PAGE_OFFSET,
> +		.marker = address_markers,
> +	};
> +	/* Traverse the 0xc, 0xd and 0xf areas of the kernel virtual memory and
> +	 * dump pages that are in the hash pagetable.
> +	 */
> +	walk_linearmapping(&st);
> +	walk_pagetables(&st);
> +	walk_vmemmap(&st);
> +	return 0;
> +}
> +
> +static int ptdump_open(struct inode *inode, struct file *file)
> +{
> +	return single_open(file, ptdump_show, NULL);
> +}
> +
> +static const struct file_operations ptdump_fops = {
> +	.open		= ptdump_open,
> +	.read		= seq_read,
> +	.llseek		= seq_lseek,
> +	.release	= single_release,
> +};
> +
> +
> +static int ptdump_init(void)
> +{
> +	struct dentry *debugfs_file;
> +
> +	debugfs_file = debugfs_create_file("kernel_hash_pagetable", 0400,
> +			NULL, NULL, &ptdump_fops);
> +	return debugfs_file ? 0 : -ENOMEM;
> +}
> +device_initcall(ptdump_init);

Otherwise (also check the _PAGE_COMBO build failure reports)


Acked-by: Balbir Singh <bsingharora@gmail.com>