[v3 PATCH 1/2] powerpc/mm: Dump linux pagetables

[v3 PATCH 1/2] powerpc/mm: Dump 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>
---
v2 -> v3: Changed to be compatible with P9. 

 arch/powerpc/Kconfig.debug             |  12 +
 arch/powerpc/mm/Makefile               |   1 +
 arch/powerpc/mm/dump_linuxpagetables.c | 440 +++++++++++++++++++++++++++++++++
 3 files changed, 453 insertions(+)
 create mode 100644 arch/powerpc/mm/dump_linuxpagetables.c

diff --git a/arch/powerpc/Kconfig.debug b/arch/powerpc/Kconfig.debug
index 2be221874de8..d19f496a7e12 100644
--- a/arch/powerpc/Kconfig.debug
+++ b/arch/powerpc/Kconfig.debug
@@ -350,4 +350,16 @@ 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 f2cea6d5e764..d3dfa02082b1 100644
--- a/arch/powerpc/mm/Makefile
+++ b/arch/powerpc/mm/Makefile
@@ -43,3 +43,4 @@ 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..d242bc79ae64
--- /dev/null
+++ b/arch/powerpc/mm/dump_linuxpagetables.c
@@ -0,0 +1,440 @@
+/*
+ * 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>
+
+/*
+ * 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 int level;
+	u64 current_flags;
+};
+
+struct addr_marker {
+	unsigned long start_address;
+	const char *name;
+};
+
+static struct addr_marker address_markers[] = {
+	{ 0,	"Start of kernel VM" },
+	{ 0,	"vmalloc() Area" },
+	{ 0,	"vmalloc() End" },
+	{ 0,	"isa I/O start" },
+	{ 0,	"isa I/O end" },
+	{ 0,	"phb I/O start" },
+	{ 0,	"phb I/O end" },
+	{ 0,	"I/O remap start" },
+	{ 0,	"I/O remap end" },
+	{ 0,	"vmemmap start" },
+	{ -1,	NULL },
+};
+
+struct flag_info {
+	u64		mask;
+	u64		val;
+	const char	*set;
+	const char	*clear;
+	bool		is_val;
+	int		shift;
+};
+
+static const struct flag_info flag_array[] = {
+	{
+#ifdef CONFIG_PPC_STD_MMU_64
+		.mask	= _PAGE_PRIVILEGED,
+		.val	= 0,
+#else
+		.mask	= _PAGE_USER,
+		.val	= _PAGE_USER,
+#endif
+		.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	= "       ",
+	}, {
+#ifdef CONFIG_PPC_STD_MMU_64
+		.mask	= H_PAGE_HASHPTE,
+		.val	= H_PAGE_HASHPTE,
+#else
+		.mask	= _PAGE_HASHPTE,
+		.val	= _PAGE_HASHPTE,
+#endif
+		.set	= "hpte",
+		.clear	= "    ",
+	}, {
+#ifndef CONFIG_PPC_STD_MMU_64
+		.mask	= _PAGE_GUARDED,
+		.val	= _PAGE_GUARDED,
+		.set	= "guarded",
+		.clear	= "       ",
+	}, {
+#endif
+		.mask	= _PAGE_DIRTY,
+		.val	= _PAGE_DIRTY,
+		.set	= "dirty",
+		.clear	= "     ",
+	}, {
+		.mask	= _PAGE_ACCESSED,
+		.val	= _PAGE_ACCESSED,
+		.set	= "accessed",
+		.clear	= "        ",
+	}, {
+#ifndef CONFIG_PPC_STD_MMU_64
+		.mask	= _PAGE_WRITETHRU,
+		.val	= _PAGE_WRITETHRU,
+		.set	= "write through",
+		.clear	= "             ",
+	}, {
+#endif
+		.mask	= _PAGE_NO_CACHE,
+		.val	= _PAGE_NO_CACHE,
+		.set	= "no cache",
+		.clear	= "        ",
+	}, {
+		.mask	= H_PAGE_BUSY,
+		.val	= H_PAGE_BUSY,
+		.set	= "busy",
+	}, {
+#ifdef CONFIG_PPC_64K_PAGES
+		.mask	= H_PAGE_COMBO,
+		.val	= H_PAGE_COMBO,
+		.set	= "combo",
+	}, {
+		.mask	= H_PAGE_4K_PFN,
+		.val	= H_PAGE_4K_PFN,
+		.set	= "4K_pfn",
+	}, {
+#endif
+		.mask	= H_PAGE_F_GIX,
+		.val	= H_PAGE_F_GIX,
+		.set	= "f_gix",
+		.is_val	= true,
+		.shift	= H_PAGE_F_GIX_SHIFT,
+	}, {
+		.mask	= H_PAGE_F_SECOND,
+		.val	= H_PAGE_F_SECOND,
+		.set	= "f_second",
+	}, {
+		.mask	= _PAGE_SPECIAL,
+		.val	= _PAGE_SPECIAL,
+		.set	= "special",
+	}
+};
+
+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_info(struct pg_state *st, const struct flag_info
+		*flag, u64 pte, int num)
+{
+	unsigned int i;
+
+	for (i = 0; i < num; i++, flag++) {
+		const char *s = NULL;
+		u64 val;
+
+		/* flag not defined so don't check it */
+		if (flag->mask == 0)
+			continue;
+		/* Some 'flags' are actually values */
+		if (flag->is_val) {
+			val = pte & flag->val;
+			if (flag->shift)
+				val = val >> flag->shift;
+			seq_printf(st->seq, "  %s:%llx", flag->set, val);
+		} else {
+			if ((pte & flag->mask) == flag->val)
+				s = flag->set;
+			else
+				s = flag->clear;
+			if (s)
+				seq_printf(st->seq, "  %s", s);
+		}
+		st->current_flags &= ~flag->mask;
+	}
+	if (st->current_flags != 0)
+		seq_printf(st->seq, "  unknown flags:%llx", st->current_flags);
+}
+
+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 int 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_info(st, pg_level[st->level].flag,
+					  st->current_flags,
+					  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 int 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 int 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 int 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_pagetables(struct pg_state *st)
+{
+	pgd_t *pgd = pgd_offset_k(0UL);
+	unsigned int 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 = KERN_VIRT_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 void populate_markers(void)
+{
+	address_markers[0].start_address = PAGE_OFFSET;
+	address_markers[1].start_address = VMALLOC_START;
+	address_markers[2].start_address = VMALLOC_END;
+	address_markers[3].start_address = ISA_IO_BASE;
+	address_markers[4].start_address = ISA_IO_END;
+	address_markers[5].start_address = PHB_IO_BASE;
+	address_markers[6].start_address = PHB_IO_END;
+	address_markers[7].start_address = IOREMAP_BASE;
+	address_markers[8].start_address = IOREMAP_END;
+#ifdef CONFIG_PPC_STD_MMU_64
+	address_markers[9].start_address =  H_VMEMMAP_BASE;
+#else
+	address_markers[9].start_address =  VMEMMAP_BASE;
+#endif
+}
+
+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_pagetables(&st);
+	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 int 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;
+
+	populate_markers();
+	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

[v3 PATCH 2/2] powerpc/mm: Dump hash table

[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 hash page table. If radix is enabled
(and so there is no hash page table) then this file doesn't exist. To
use this the PPC_PTDUMP config option must be selected.

Signed-off-by: Rashmica Gupta <rashmicy@gmail.com>
---
v2 -> v3: Changed to be compatible with P9.

 arch/powerpc/mm/Makefile             |   3 +-
 arch/powerpc/mm/dump_hashpagetable.c | 543 +++++++++++++++++++++++++++++++++++
 2 files changed, 545 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 d3dfa02082b1..df4fbf56aa7b 100644
--- a/arch/powerpc/mm/Makefile
+++ b/arch/powerpc/mm/Makefile
@@ -43,4 +43,5 @@ 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..f8ccc224eb2c
--- /dev/null
+++ b/arch/powerpc/mm/dump_hashpagetable.c
@@ -0,0 +1,543 @@
+/*
+ * 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.
+ *
+ * If radix is enabled then there is no hash page table and so no debugfs file
+ * is generated.
+ *
+ * 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 pg_state {
+	struct seq_file *seq;
+	const struct addr_marker *marker;
+	unsigned long start_address;
+	unsigned int level;
+	u64 current_flags;
+};
+
+struct addr_marker {
+	unsigned long start_address;
+	const char *name;
+};
+
+static struct addr_marker address_markers[] = {
+	{ 0,	"Start of kernel VM" },
+	{ 0,	"vmalloc() Area" },
+	{ 0,	"vmalloc() End" },
+	{ 0,	"isa I/O start" },
+	{ 0,	"isa I/O end" },
+	{ 0,	"phb I/O start" },
+	{ 0,	"phb I/O end" },
+	{ 0,	"I/O remap start" },
+	{ 0,	"I/O remap end" },
+	{ 0,	"vmemmap start" },
+	{ -1,	NULL },
+};
+
+struct flag_info {
+	u64		mask;
+	u64		val;
+	const char	*set;
+	const char	*clear;
+	bool		is_val;
+	int		shift;
+};
+
+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 int calculate_pagesize(struct pg_state *st, int ps, char s[])
+{
+	static const char units[] = "BKMGTPE";
+	const char *unit = units;
+
+	while (ps > 9 && unit[1]) {
+		ps -= 10;
+		unit++;
+	}
+	seq_printf(st->seq, "  %s_ps: %i%c\t", s, 1<<ps, *unit);
+	return ps;
+}
+
+static void dump_flag_info(struct pg_state *st, const struct flag_info
+		*flag, u64 pte, int num)
+{
+	unsigned int i;
+
+	for (i = 0; i < num; i++, flag++) {
+		const char *s = NULL;
+		u64 val;
+
+		/* flag not defined so don't check it */
+		if (flag->mask == 0)
+			continue;
+		/* Some 'flags' are actually values */
+		if (flag->is_val) {
+			val = pte & flag->val;
+			if (flag->shift)
+				val = val >> flag->shift;
+			seq_printf(st->seq, "  %s:%llx", flag->set, 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)
+{
+	int aps_index;
+
+	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));
+
+	calculate_pagesize(st, bps, "base");
+	aps_index = calculate_pagesize(st, aps, "actual");
+	if (aps_index != 2)
+		seq_printf(st->seq, "LP 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 & H_PAGE_COMBO) == H_PAGE_COMBO) ||
+			((pteval & H_PAGE_4K_PFN) == H_PAGE_4K_PFN))
+			psize = mmu_io_psize;
+#endif
+		/* check for hashpte */
+		status = hpte_find(st, addr, psize);
+
+		if (((pteval & H_PAGE_HASHPTE) != H_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 int 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 int 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_pagetables(struct pg_state *st)
+{
+	pgd_t *pgd = pgd_offset_k(0UL);
+	unsigned int 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 = KERN_VIRT_START + i * PGDIR_SIZE;
+		if (!pgd_none(*pgd))
+			/* pgd exists */
+			walk_pud(st, pgd, 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_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 void populate_markers(void)
+{
+	address_markers[0].start_address = PAGE_OFFSET;
+	address_markers[1].start_address = VMALLOC_START;
+	address_markers[2].start_address = VMALLOC_END;
+	address_markers[3].start_address = ISA_IO_BASE;
+	address_markers[4].start_address = ISA_IO_END;
+	address_markers[5].start_address = PHB_IO_BASE;
+	address_markers[6].start_address = PHB_IO_END;
+	address_markers[7].start_address = IOREMAP_BASE;
+	address_markers[8].start_address = IOREMAP_END;
+#ifdef CONFIG_PPC_STD_MMU_64
+	address_markers[9].start_address =  H_VMEMMAP_BASE;
+#else
+	address_markers[9].start_address =  VMEMMAP_BASE;
+#endif
+}
+
+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;
+
+	if (!radix_enabled()) {
+		populate_markers();
+		debugfs_file = debugfs_create_file("kernel_hash_pagetable",
+				0400, NULL, NULL, &ptdump_fops);
+		return debugfs_file ? 0 : -ENOMEM;
+	}
+	return 0;
+}
+device_initcall(ptdump_init);
-- 
2.5.0

Re: [v3,2/2] powerpc/mm: Dump hash table

[Michael Ellerman]

On Fri, 2016-27-05 at 05:49:00 UTC, 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 hash page table. If radix is enabled
> (and so there is no hash page table) then this file doesn't exist. To
> use this the PPC_PTDUMP config option must be selected.
> 
> Signed-off-by: Rashmica Gupta <rashmicy@gmail.com>

Applied to powerpc next, thanks.

https://git.kernel.org/powerpc/c/1515ab932156257afd8a5864506dab

cheers

Re: [v3 PATCH 2/2] powerpc/mm: Dump hash table

[Aneesh Kumar K.V]

Rashmica Gupta <rashmicy@gmail.com> writes:

> 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 hash page table. If radix is enabled
> (and so there is no hash page table) then this file doesn't exist. To
> use this the PPC_PTDUMP config option must be selected.
>
> Signed-off-by: Rashmica Gupta <rashmicy@gmail.com>

I completely missed this when it was sent to the list.

> ---
> v2 -> v3: Changed to be compatible with P9.
>
>  arch/powerpc/mm/Makefile             |   3 +-
>  arch/powerpc/mm/dump_hashpagetable.c | 543 +++++++++++++++++++++++++++++++++++
>  2 files changed, 545 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 d3dfa02082b1..df4fbf56aa7b 100644
> --- a/arch/powerpc/mm/Makefile
> +++ b/arch/powerpc/mm/Makefile
> @@ -43,4 +43,5 @@ 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..f8ccc224eb2c
> --- /dev/null
> +++ b/arch/powerpc/mm/dump_hashpagetable.c
> @@ -0,0 +1,543 @@
> +/*
> + * 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.
> + *
> + * If radix is enabled then there is no hash page table and so no debugfs file
> + * is generated.
> + *
> + * 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 pg_state {
> +	struct seq_file *seq;
> +	const struct addr_marker *marker;
> +	unsigned long start_address;
> +	unsigned int level;
> +	u64 current_flags;
> +};
> +
> +struct addr_marker {
> +	unsigned long start_address;
> +	const char *name;
> +};
> +
> +static struct addr_marker address_markers[] = {
> +	{ 0,	"Start of kernel VM" },
> +	{ 0,	"vmalloc() Area" },
> +	{ 0,	"vmalloc() End" },
> +	{ 0,	"isa I/O start" },
> +	{ 0,	"isa I/O end" },
> +	{ 0,	"phb I/O start" },
> +	{ 0,	"phb I/O end" },
> +	{ 0,	"I/O remap start" },
> +	{ 0,	"I/O remap end" },
> +	{ 0,	"vmemmap start" },
> +	{ -1,	NULL },
> +};
> +
> +struct flag_info {
> +	u64		mask;
> +	u64		val;
> +	const char	*set;
> +	const char	*clear;
> +	bool		is_val;
> +	int		shift;
> +};
> +
> +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 int calculate_pagesize(struct pg_state *st, int ps, char s[])
> +{
> +	static const char units[] = "BKMGTPE";
> +	const char *unit = units;
> +
> +	while (ps > 9 && unit[1]) {
> +		ps -= 10;
> +		unit++;
> +	}
> +	seq_printf(st->seq, "  %s_ps: %i%c\t", s, 1<<ps, *unit);
> +	return ps;
> +}
> +
> +static void dump_flag_info(struct pg_state *st, const struct flag_info
> +		*flag, u64 pte, int num)
> +{
> +	unsigned int i;
> +
> +	for (i = 0; i < num; i++, flag++) {
> +		const char *s = NULL;
> +		u64 val;
> +
> +		/* flag not defined so don't check it */
> +		if (flag->mask == 0)
> +			continue;
> +		/* Some 'flags' are actually values */
> +		if (flag->is_val) {
> +			val = pte & flag->val;
> +			if (flag->shift)
> +				val = val >> flag->shift;
> +			seq_printf(st->seq, "  %s:%llx", flag->set, 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)
> +{
> +	int aps_index;
> +
> +	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));
> +
> +	calculate_pagesize(st, bps, "base");
> +	aps_index = calculate_pagesize(st, aps, "actual");
> +	if (aps_index != 2)
> +		seq_printf(st->seq, "LP 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 & H_PAGE_COMBO) == H_PAGE_COMBO) ||
> +			((pteval & H_PAGE_4K_PFN) == H_PAGE_4K_PFN))
> +			psize = mmu_io_psize;
> +#endif
> +		/* check for hashpte */
> +		status = hpte_find(st, addr, psize);
> +
> +		if (((pteval & H_PAGE_HASHPTE) != H_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 int 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);

Is that correct ? we need to handle bad_pmd, hugetlb entries, THP
entries etc right ? Also while walking we also need to make sure it is
not getting split. 



> +	}
> +}
> +
> +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 int 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_pagetables(struct pg_state *st)
> +{
> +	pgd_t *pgd = pgd_offset_k(0UL);
> +	unsigned int 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 = KERN_VIRT_START + i * PGDIR_SIZE;
> +		if (!pgd_none(*pgd))
> +			/* pgd exists */
> +			walk_pud(st, pgd, 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_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 void populate_markers(void)
> +{
> +	address_markers[0].start_address = PAGE_OFFSET;
> +	address_markers[1].start_address = VMALLOC_START;
> +	address_markers[2].start_address = VMALLOC_END;
> +	address_markers[3].start_address = ISA_IO_BASE;
> +	address_markers[4].start_address = ISA_IO_END;
> +	address_markers[5].start_address = PHB_IO_BASE;
> +	address_markers[6].start_address = PHB_IO_END;
> +	address_markers[7].start_address = IOREMAP_BASE;
> +	address_markers[8].start_address = IOREMAP_END;
> +#ifdef CONFIG_PPC_STD_MMU_64
> +	address_markers[9].start_address =  H_VMEMMAP_BASE;
> +#else
> +	address_markers[9].start_address =  VMEMMAP_BASE;
> +#endif
> +}
> +
> +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;
> +
> +	if (!radix_enabled()) {
> +		populate_markers();
> +		debugfs_file = debugfs_create_file("kernel_hash_pagetable",
> +				0400, NULL, NULL, &ptdump_fops);
> +		return debugfs_file ? 0 : -ENOMEM;
> +	}
> +	return 0;
> +}
> +device_initcall(ptdump_init);
> -- 
> 2.5.0
>
> _______________________________________________
> Linuxppc-dev mailing list
> Linuxppc-dev@lists.ozlabs.org
> https://lists.ozlabs.org/listinfo/linuxppc-dev

Re: [v3,1/2] powerpc/mm: Dump linux pagetables

[Michael Ellerman]

On Fri, 2016-27-05 at 05:48:59 UTC, 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>

Applied to powerpc next, thanks.

https://git.kernel.org/powerpc/c/8eb07b187000d48152c4ef97f075bd

cheers

Re: [v3,1/2] powerpc/mm: Dump linux pagetables

[Christophe LEROY]

Hi Rashmica,

Le 17/11/2016 à 13:03, Michael Ellerman a écrit :
> On Fri, 2016-27-05 at 05:48:59 UTC, 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>
>
> Applied to powerpc next, thanks.
>
> https://git.kernel.org/powerpc/c/8eb07b187000d48152c4ef97f075bd

For your information, compilation fails on mpc885_ads_defconfig when we 
activate PPC_PTDUMP:

   CC      arch/powerpc/mm/dump_linuxpagetables.o
arch/powerpc/mm/dump_linuxpagetables.c: In function 'walk_pagetables':
arch/powerpc/mm/dump_linuxpagetables.c:369:10: error: 'KERN_VIRT_START' 
undeclared (first use in this function)
arch/powerpc/mm/dump_linuxpagetables.c:369:10: note: each undeclared 
identifier is reported only once for each function it appears in
arch/powerpc/mm/dump_linuxpagetables.c: In function 'populate_markers':
arch/powerpc/mm/dump_linuxpagetables.c:383:37: error: 'ISA_IO_BASE' 
undeclared (first use in this function)
arch/powerpc/mm/dump_linuxpagetables.c:384:37: error: 'ISA_IO_END' 
undeclared (first use in this function)
arch/powerpc/mm/dump_linuxpagetables.c:385:37: error: 'PHB_IO_BASE' 
undeclared (first use in this function)
arch/powerpc/mm/dump_linuxpagetables.c:386:37: error: 'PHB_IO_END' 
undeclared (first use in this function)
arch/powerpc/mm/dump_linuxpagetables.c:387:37: error: 'IOREMAP_BASE' 
undeclared (first use in this function)
arch/powerpc/mm/dump_linuxpagetables.c:388:37: error: 'IOREMAP_END' 
undeclared (first use in this function)
arch/powerpc/mm/dump_linuxpagetables.c:392:38: error: 'VMEMMAP_BASE' 
undeclared (first use in this function)
arch/powerpc/mm/dump_linuxpagetables.c: In function 'ptdump_show':
arch/powerpc/mm/dump_linuxpagetables.c:400:20: error: 'KERN_VIRT_START' 
undeclared (first use in this function)
make[1]: *** [arch/powerpc/mm/dump_linuxpagetables.o] Error 1
make: *** [arch/powerpc/mm] Error 2



Christophe

Re: [v3,1/2] powerpc/mm: Dump linux pagetables

[Michael Ellerman]

Christophe LEROY <christophe.leroy@c-s.fr> writes:

> Hi Rashmica,
>
> Le 17/11/2016 =C3=A0 13:03, Michael Ellerman a =C3=A9crit :
>> On Fri, 2016-27-05 at 05:48:59 UTC, 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>
>>
>> Applied to powerpc next, thanks.
>>
>> https://git.kernel.org/powerpc/c/8eb07b187000d48152c4ef97f075bd
>
> For your information, compilation fails on mpc885_ads_defconfig when we=20
> activate PPC_PTDUMP:
>
>    CC      arch/powerpc/mm/dump_linuxpagetables.o
> arch/powerpc/mm/dump_linuxpagetables.c: In function 'walk_pagetables':
> arch/powerpc/mm/dump_linuxpagetables.c:369:10: error: 'KERN_VIRT_START'=20
> undeclared (first use in this function)

Yeah it's basically 64-bit only at the moment.

I have a TODO for it: https://github.com/linuxppc/linux/issues/88

But I haven't got around to doing it, maybe someone who has more
interest in 32-bit will have time before me ;)

cheers

Re: [v3,1/2] powerpc/mm: Dump linux pagetables

[Christophe LEROY]

Le 13/04/2017 à 13:31, Michael Ellerman a écrit :
> Christophe LEROY <christophe.leroy@c-s.fr> writes:
>
>> Hi Rashmica,
>>
>> Le 17/11/2016 à 13:03, Michael Ellerman a écrit :
>>> On Fri, 2016-27-05 at 05:48:59 UTC, 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>
>>>
>>> Applied to powerpc next, thanks.
>>>
>>> https://git.kernel.org/powerpc/c/8eb07b187000d48152c4ef97f075bd
>>
>> For your information, compilation fails on mpc885_ads_defconfig when we
>> activate PPC_PTDUMP:
>>
>>    CC      arch/powerpc/mm/dump_linuxpagetables.o
>> arch/powerpc/mm/dump_linuxpagetables.c: In function 'walk_pagetables':
>> arch/powerpc/mm/dump_linuxpagetables.c:369:10: error: 'KERN_VIRT_START'
>> undeclared (first use in this function)
>
> Yeah it's basically 64-bit only at the moment.
>
> I have a TODO for it: https://github.com/linuxppc/linux/issues/88
>
> But I haven't got around to doing it, maybe someone who has more
> interest in 32-bit will have time before me ;)

Well, I just did it a few minutes ago.

Christophe

Re: [v3,1/2] powerpc/mm: Dump linux pagetables

[Michael Ellerman]

Christophe LEROY <christophe.leroy@c-s.fr> writes:
> Le 13/04/2017 =C3=A0 13:31, Michael Ellerman a =C3=A9crit :
>> Christophe LEROY <christophe.leroy@c-s.fr> writes:
>>> Le 17/11/2016 =C3=A0 13:03, Michael Ellerman a =C3=A9crit :
>>>> On Fri, 2016-27-05 at 05:48:59 UTC, 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_PTDU=
MP
>>>>> config option must be selected.
>>>>>
>>>>> Signed-off-by: Rashmica Gupta <rashmicy@gmail.com>
>>>>
>>>> Applied to powerpc next, thanks.
>>>>
>>>> https://git.kernel.org/powerpc/c/8eb07b187000d48152c4ef97f075bd
>>>
>>> For your information, compilation fails on mpc885_ads_defconfig when we
>>> activate PPC_PTDUMP:
>>>
>>>    CC      arch/powerpc/mm/dump_linuxpagetables.o
>>> arch/powerpc/mm/dump_linuxpagetables.c: In function 'walk_pagetables':
>>> arch/powerpc/mm/dump_linuxpagetables.c:369:10: error: 'KERN_VIRT_START'
>>> undeclared (first use in this function)
>>
>> Yeah it's basically 64-bit only at the moment.
>>
>> I have a TODO for it: https://github.com/linuxppc/linux/issues/88
>>
>> But I haven't got around to doing it, maybe someone who has more
>> interest in 32-bit will have time before me ;)
>
> Well, I just did it a few minutes ago.

It's like magic! :)

cheers