x86/kprobes: Fix crash when probing CS CALL instructionsgrafted

When using eBPF to probe CS CALL instructions within a function,
a crash can be triggered.

The eBPF tool probes offset 257 of the __hrtimer_run_queues()
function:

<__hrtimer_run_queues+249>:  nopl   0x0(%rax,%rax,1)
<__hrtimer_run_queues+254>:  mov    %r14,%rdi
<__hrtimer_run_queues+257>:  cs call <__x86_indirect_thunk_r12>
<__hrtimer_run_queues+263>:  mov    %eax,%r12d
<__hrtimer_run_queues+266>:  xchg   %ax,%ax
<__hrtimer_run_queues+268>:  mov    %r13,%rdi

Which triggers this crash:

  BUG: unable to handle page fault for address: 00000000000f41c9
  #PF: supervisor write access in kernel mode
  #PF: error_code(0x0002) - not-present page
  PGD 0 P4D 0
  Oops: 0002 [#1] SMP NOPTI
  CPU: 1 PID: 0 Comm: swapper/1 Kdump: loaded Tainted: P
  RIP: 0010:__hrtimer_run_queues+0x106/0x230

Note that __hrtimer_run_queues+0x106 is __hrtimer_run_queues+262, which is
at the 6th byte of the above CS CALL instruction. Since the CS CALL
instruction occupies 6 bytes, the exception occurred in the middle of that
call instruction.

The root cause is that when using eBPF tools to probe in the middle of a
function, a kprobe with INT3 is used as the underlying implementation.

During single-step emulation of the original CALL instruction,
int3_emulate_call() assumes that the probed CALL instruction is 5 bytes
long. However, the actual CS-prefixed CALL instruction occupies 6 bytes,
so it constructs an incorrect exception return address. When the CPU
returns from the kprobe handler, the next instruction to be executed is at
the address of the last byte of that CS CALL instruction. Coincidentally,
starting from that address, the CPU fetches and decodes a completely
different instruction, which ultimately triggers a kernel crash.

Fix the issue by using the actual instruction length obtained from
the instruction decoder when constructing the exception return
address, rather than relying on the hardcoded CALL_INSN_SIZE macro.

[ mingo: Refined the changelog ]

Fixes: 6256e668b7af ("x86/kprobes: Use int3 instead of debug trap for single-step")
Suggested-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Signed-off-by: Jinke Han <jinkehan@didiglobal.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Acked-by: Yafang Shao <laoar.shao@gmail.com>
Acked-by: Borislav Petkov <bp@alien8.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://patch.msgid.link/20260908073742.GA10517@didi-ThinkCentre-M920t-N000

6 files changed, 916 insertions, 0 deletions

diff --git a/arch/microblaze/mm/Makefile b/arch/microblaze/mm/Makefile
new file mode 100644
index 000000000..75edfc110
--- /dev/null
+++ b/arch/microblaze/mm/Makefile
@@ -0,0 +1,6 @@
+# SPDX-License-Identifier: GPL-2.0-only
+#
+# Makefile
+#
+
+obj-y := consistent.o init.o pgtable.o mmu_context.o fault.o
diff --git a/arch/microblaze/mm/consistent.c b/arch/microblaze/mm/consistent.c
new file mode 100644
index 000000000..b7ad4a986
--- /dev/null
+++ b/arch/microblaze/mm/consistent.c
@@ -0,0 +1,23 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Microblaze support for cache consistent memory.
+ * Copyright (C) 2010 Michal Simek <monstr@monstr.eu>
+ * Copyright (C) 2010 PetaLogix
+ * Copyright (C) 2005 John Williams <jwilliams@itee.uq.edu.au>
+ */
+
+#include <linux/kernel.h>
+#include <linux/string.h>
+#include <linux/types.h>
+#include <linux/mm.h>
+#include <linux/init.h>
+#include <linux/dma-map-ops.h>
+#include <asm/cpuinfo.h>
+#include <asm/cacheflush.h>
+
+void arch_dma_prep_coherent(struct page *page, size_t size)
+{
+	phys_addr_t paddr = page_to_phys(page);
+
+	flush_dcache_range(paddr, paddr + size);
+}
diff --git a/arch/microblaze/mm/fault.c b/arch/microblaze/mm/fault.c
new file mode 100644
index 000000000..d3c3c33b7
--- /dev/null
+++ b/arch/microblaze/mm/fault.c
@@ -0,0 +1,299 @@
+/*
+ *  arch/microblaze/mm/fault.c
+ *
+ *    Copyright (C) 2007 Xilinx, Inc.  All rights reserved.
+ *
+ *  Derived from "arch/ppc/mm/fault.c"
+ *    Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
+ *
+ *  Derived from "arch/i386/mm/fault.c"
+ *    Copyright (C) 1991, 1992, 1993, 1994  Linus Torvalds
+ *
+ *  Modified by Cort Dougan and Paul Mackerras.
+ *
+ * This file is subject to the terms and conditions of the GNU General
+ * Public License.  See the file COPYING in the main directory of this
+ * archive for more details.
+ *
+ */
+
+#include <linux/extable.h>
+#include <linux/signal.h>
+#include <linux/sched.h>
+#include <linux/kernel.h>
+#include <linux/errno.h>
+#include <linux/string.h>
+#include <linux/types.h>
+#include <linux/ptrace.h>
+#include <linux/mman.h>
+#include <linux/mm.h>
+#include <linux/interrupt.h>
+#include <linux/perf_event.h>
+
+#include <asm/page.h>
+#include <asm/mmu.h>
+#include <linux/mmu_context.h>
+#include <linux/uaccess.h>
+#include <asm/exceptions.h>
+
+static unsigned long pte_misses;	/* updated by do_page_fault() */
+static unsigned long pte_errors;	/* updated by do_page_fault() */
+
+/*
+ * Check whether the instruction at regs->pc is a store using
+ * an update addressing form which will update r1.
+ */
+static int store_updates_sp(struct pt_regs *regs)
+{
+	unsigned int inst;
+
+	if (get_user(inst, (unsigned int __user *)regs->pc))
+		return 0;
+	/* check for 1 in the rD field */
+	if (((inst >> 21) & 0x1f) != 1)
+		return 0;
+	/* check for store opcodes */
+	if ((inst & 0xd0000000) == 0xd0000000)
+		return 1;
+	return 0;
+}
+
+
+/*
+ * bad_page_fault is called when we have a bad access from the kernel.
+ * It is called from do_page_fault above and from some of the procedures
+ * in traps.c.
+ */
+void bad_page_fault(struct pt_regs *regs, unsigned long address, int sig)
+{
+	const struct exception_table_entry *fixup;
+/* MS: no context */
+	/* Are we prepared to handle this fault?  */
+	fixup = search_exception_tables(regs->pc);
+	if (fixup) {
+		regs->pc = fixup->fixup;
+		return;
+	}
+
+	/* kernel has accessed a bad area */
+	die("kernel access of bad area", regs, sig);
+}
+
+/*
+ * The error_code parameter is ESR for a data fault,
+ * 0 for an instruction fault.
+ */
+void do_page_fault(struct pt_regs *regs, unsigned long address,
+		   unsigned long error_code)
+{
+	struct vm_area_struct *vma;
+	struct mm_struct *mm = current->mm;
+	int code = SEGV_MAPERR;
+	int is_write = error_code & ESR_S;
+	vm_fault_t fault;
+	unsigned int flags = FAULT_FLAG_DEFAULT;
+
+	regs->ear = address;
+	regs->esr = error_code;
+
+	/* On a kernel SLB miss we can only check for a valid exception entry */
+	if (unlikely(kernel_mode(regs) && (address >= TASK_SIZE))) {
+		pr_warn("kernel task_size exceed");
+		_exception(SIGSEGV, regs, code, address);
+	}
+
+	/* for instr TLB miss and instr storage exception ESR_S is undefined */
+	if ((error_code & 0x13) == 0x13 || (error_code & 0x11) == 0x11)
+		is_write = 0;
+
+	if (unlikely(faulthandler_disabled() || !mm)) {
+		if (kernel_mode(regs))
+			goto bad_area_nosemaphore;
+
+		/* faulthandler_disabled() in user mode is really bad,
+		   as is current->mm == NULL. */
+		pr_emerg("Page fault in user mode with faulthandler_disabled(), mm = %p\n",
+			 mm);
+		pr_emerg("r15 = %lx  MSR = %lx\n",
+		       regs->r15, regs->msr);
+		die("Weird page fault", regs, SIGSEGV);
+	}
+
+	if (user_mode(regs))
+		flags |= FAULT_FLAG_USER;
+
+	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address);
+
+	/* When running in the kernel we expect faults to occur only to
+	 * addresses in user space.  All other faults represent errors in the
+	 * kernel and should generate an OOPS.  Unfortunately, in the case of an
+	 * erroneous fault occurring in a code path which already holds mmap_lock
+	 * we will deadlock attempting to validate the fault against the
+	 * address space.  Luckily the kernel only validly references user
+	 * space from well defined areas of code, which are listed in the
+	 * exceptions table.
+	 *
+	 * As the vast majority of faults will be valid we will only perform
+	 * the source reference check when there is a possibility of a deadlock.
+	 * Attempt to lock the address space, if we cannot we then validate the
+	 * source.  If this is invalid we can skip the address space check,
+	 * thus avoiding the deadlock.
+	 */
+	if (unlikely(!mmap_read_trylock(mm))) {
+		if (kernel_mode(regs) && !search_exception_tables(regs->pc))
+			goto bad_area_nosemaphore;
+
+retry:
+		mmap_read_lock(mm);
+	}
+
+	vma = find_vma(mm, address);
+	if (unlikely(!vma))
+		goto bad_area;
+
+	if (vma->vm_start <= address)
+		goto good_area;
+
+	if (unlikely(!(vma->vm_flags & VM_GROWSDOWN)))
+		goto bad_area;
+
+	if (unlikely(!is_write))
+		goto bad_area;
+
+	/*
+	 * N.B. The ABI allows programs to access up to
+	 * a few hundred bytes below the stack pointer (TBD).
+	 * The kernel signal delivery code writes up to about 1.5kB
+	 * below the stack pointer (r1) before decrementing it.
+	 * The exec code can write slightly over 640kB to the stack
+	 * before setting the user r1.  Thus we allow the stack to
+	 * expand to 1MB without further checks.
+	 */
+	if (unlikely(address + 0x100000 < vma->vm_end)) {
+
+		/* get user regs even if this fault is in kernel mode */
+		struct pt_regs *uregs = current->thread.regs;
+		if (uregs == NULL)
+			goto bad_area;
+
+		/*
+		 * A user-mode access to an address a long way below
+		 * the stack pointer is only valid if the instruction
+		 * is one which would update the stack pointer to the
+		 * address accessed if the instruction completed,
+		 * i.e. either stwu rs,n(r1) or stwux rs,r1,rb
+		 * (or the byte, halfword, float or double forms).
+		 *
+		 * If we don't check this then any write to the area
+		 * between the last mapped region and the stack will
+		 * expand the stack rather than segfaulting.
+		 */
+		if (address + 2048 < uregs->r1
+			&& (kernel_mode(regs) || !store_updates_sp(regs)))
+				goto bad_area;
+	}
+	vma = expand_stack(mm, address);
+	if (!vma)
+		goto bad_area_nosemaphore;
+
+good_area:
+	code = SEGV_ACCERR;
+
+	/* a write */
+	if (unlikely(is_write)) {
+		if (unlikely(!(vma->vm_flags & VM_WRITE)))
+			goto bad_area;
+		flags |= FAULT_FLAG_WRITE;
+	/* a read */
+	} else {
+		/* protection fault */
+		if (unlikely(error_code & 0x08000000))
+			goto bad_area;
+		if (unlikely(!(vma->vm_flags & (VM_READ | VM_EXEC))))
+			goto bad_area;
+	}
+
+	/*
+	 * If for any reason at all we couldn't handle the fault,
+	 * make sure we exit gracefully rather than endlessly redo
+	 * the fault.
+	 */
+	fault = handle_mm_fault(vma, address, flags, regs);
+
+	if (fault_signal_pending(fault, regs)) {
+		if (!user_mode(regs))
+			bad_page_fault(regs, address, SIGBUS);
+		return;
+	}
+
+	/* The fault is fully completed (including releasing mmap lock) */
+	if (fault & VM_FAULT_COMPLETED)
+		return;
+
+	if (unlikely(fault & VM_FAULT_ERROR)) {
+		if (fault & VM_FAULT_OOM)
+			goto out_of_memory;
+		else if (fault & VM_FAULT_SIGSEGV)
+			goto bad_area;
+		else if (fault & VM_FAULT_SIGBUS)
+			goto do_sigbus;
+		BUG();
+	}
+
+	if (fault & VM_FAULT_RETRY) {
+		flags |= FAULT_FLAG_TRIED;
+
+		/*
+		 * No need to mmap_read_unlock(mm) as we would
+		 * have already released it in __lock_page_or_retry
+		 * in mm/filemap.c.
+		 */
+
+		goto retry;
+	}
+
+	mmap_read_unlock(mm);
+
+	/*
+	 * keep track of tlb+htab misses that are good addrs but
+	 * just need pte's created via handle_mm_fault()
+	 * -- Cort
+	 */
+	pte_misses++;
+	return;
+
+bad_area:
+	mmap_read_unlock(mm);
+
+bad_area_nosemaphore:
+	pte_errors++;
+
+	/* User mode accesses cause a SIGSEGV */
+	if (user_mode(regs)) {
+		_exception(SIGSEGV, regs, code, address);
+		return;
+	}
+
+	bad_page_fault(regs, address, SIGSEGV);
+	return;
+
+/*
+ * We ran out of memory, or some other thing happened to us that made
+ * us unable to handle the page fault gracefully.
+ */
+out_of_memory:
+	mmap_read_unlock(mm);
+	if (!user_mode(regs))
+		bad_page_fault(regs, address, SIGKILL);
+	else
+		pagefault_out_of_memory();
+	return;
+
+do_sigbus:
+	mmap_read_unlock(mm);
+	if (user_mode(regs)) {
+		force_sig_fault(SIGBUS, BUS_ADRERR, (void __user *)address);
+		return;
+	}
+	bad_page_fault(regs, address, SIGBUS);
+}
diff --git a/arch/microblaze/mm/init.c b/arch/microblaze/mm/init.c
new file mode 100644
index 000000000..848cdee13
--- /dev/null
+++ b/arch/microblaze/mm/init.c
@@ -0,0 +1,259 @@
+/*
+ * Copyright (C) 2007-2008 Michal Simek <monstr@monstr.eu>
+ * Copyright (C) 2006 Atmark Techno, Inc.
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License. See the file "COPYING" in the main directory of this archive
+ * for more details.
+ */
+
+#include <linux/dma-map-ops.h>
+#include <linux/memblock.h>
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/mm.h> /* mem_init */
+#include <linux/initrd.h>
+#include <linux/of_fdt.h>
+#include <linux/pagemap.h>
+#include <linux/pfn.h>
+#include <linux/slab.h>
+#include <linux/swap.h>
+#include <linux/export.h>
+
+#include <asm/page.h>
+#include <asm/mmu_context.h>
+#include <asm/pgalloc.h>
+#include <asm/sections.h>
+#include <asm/tlb.h>
+#include <asm/fixmap.h>
+
+/* Use for MMU and noMMU because of PCI generic code */
+int mem_init_done;
+
+char *klimit = _end;
+
+/*
+ * Initialize the bootmem system and give it all the memory we
+ * have available.
+ */
+unsigned long memory_start;
+EXPORT_SYMBOL(memory_start);
+unsigned long memory_size;
+EXPORT_SYMBOL(memory_size);
+unsigned long lowmem_size;
+
+EXPORT_SYMBOL(min_low_pfn);
+EXPORT_SYMBOL(max_low_pfn);
+
+#ifdef CONFIG_HIGHMEM
+static void __init highmem_init(void)
+{
+	pr_debug("%x\n", (u32)PKMAP_BASE);
+	map_page(PKMAP_BASE, 0, 0);	/* XXX gross */
+	pkmap_page_table = virt_to_kpte(PKMAP_BASE);
+}
+#endif /* CONFIG_HIGHMEM */
+
+void __init arch_zone_limits_init(unsigned long *max_zone_pfns)
+{
+#ifdef CONFIG_HIGHMEM
+	max_zone_pfns[ZONE_DMA] = max_low_pfn;
+	max_zone_pfns[ZONE_HIGHMEM] = max_pfn;
+#else
+	max_zone_pfns[ZONE_DMA] = max_pfn;
+#endif
+}
+
+/*
+ * paging_init() sets up the page tables - in fact we've already done this.
+ */
+static void __init paging_init(void)
+{
+	int idx;
+
+	/* Setup fixmaps */
+	for (idx = 0; idx < __end_of_fixed_addresses; idx++)
+		clear_fixmap(idx);
+
+#ifdef CONFIG_HIGHMEM
+	highmem_init();
+#endif
+}
+
+void __init setup_memory(void)
+{
+	/*
+	 * Kernel:
+	 * start: base phys address of kernel - page align
+	 * end: base phys address of kernel - page align
+	 *
+	 * min_low_pfn - the first page (mm/bootmem.c - node_boot_start)
+	 * max_low_pfn
+	 */
+
+	/* memory start is from the kernel end (aligned) to higher addr */
+	min_low_pfn = memory_start >> PAGE_SHIFT; /* minimum for allocation */
+	max_low_pfn = ((u64)memory_start + (u64)lowmem_size) >> PAGE_SHIFT;
+	max_pfn = ((u64)memory_start + (u64)memory_size) >> PAGE_SHIFT;
+
+	pr_info("%s: min_low_pfn: %#lx\n", __func__, min_low_pfn);
+	pr_info("%s: max_low_pfn: %#lx\n", __func__, max_low_pfn);
+	pr_info("%s: max_pfn: %#lx\n", __func__, max_pfn);
+
+	paging_init();
+}
+
+void __init mem_init(void)
+{
+	mem_init_done = 1;
+}
+
+int page_is_ram(unsigned long pfn)
+{
+	return pfn < max_low_pfn;
+}
+
+/*
+ * Check for command-line options that affect what MMU_init will do.
+ */
+static void __init mm_cmdline_setup(void)
+{
+	unsigned long maxmem = 0;
+	char *p = cmd_line;
+
+	/* Look for mem= option on command line */
+	p = strstr(cmd_line, "mem=");
+	if (p) {
+		p += 4;
+		maxmem = memparse(p, &p);
+		if (maxmem && memory_size > maxmem) {
+			memory_size = maxmem;
+			memblock.memory.regions[0].size = memory_size;
+		}
+	}
+}
+
+/*
+ * MMU_init_hw does the chip-specific initialization of the MMU hardware.
+ */
+static void __init mmu_init_hw(void)
+{
+	/*
+	 * The Zone Protection Register (ZPR) defines how protection will
+	 * be applied to every page which is a member of a given zone. At
+	 * present, we utilize only two of the zones.
+	 * The zone index bits (of ZSEL) in the PTE are used for software
+	 * indicators, except the LSB.  For user access, zone 1 is used,
+	 * for kernel access, zone 0 is used.  We set all but zone 1
+	 * to zero, allowing only kernel access as indicated in the PTE.
+	 * For zone 1, we set a 01 binary (a value of 10 will not work)
+	 * to allow user access as indicated in the PTE.  This also allows
+	 * kernel access as indicated in the PTE.
+	 */
+	__asm__ __volatile__ ("ori r11, r0, 0x10000000;" \
+			"mts rzpr, r11;"
+			: : : "r11");
+}
+
+/*
+ * MMU_init sets up the basic memory mappings for the kernel,
+ * including both RAM and possibly some I/O regions,
+ * and sets up the page tables and the MMU hardware ready to go.
+ */
+
+/* called from head.S */
+asmlinkage void __init mmu_init(void)
+{
+	unsigned int kstart, ksize;
+
+	if ((u32) memblock.memory.regions[0].size < 0x400000) {
+		pr_emerg("Memory must be greater than 4MB\n");
+		machine_restart(NULL);
+	}
+
+	if ((u32) memblock.memory.regions[0].size < kernel_tlb) {
+		pr_emerg("Kernel size is greater than memory node\n");
+		machine_restart(NULL);
+	}
+
+	/* Find main memory where the kernel is */
+	memory_start = (u32) memblock.memory.regions[0].base;
+	lowmem_size = memory_size = (u32) memblock.memory.regions[0].size;
+
+	if (lowmem_size > CONFIG_LOWMEM_SIZE) {
+		lowmem_size = CONFIG_LOWMEM_SIZE;
+#ifndef CONFIG_HIGHMEM
+		memory_size = lowmem_size;
+#endif
+	}
+
+	mm_cmdline_setup(); /* FIXME parse args from command line - not used */
+
+	/*
+	 * Map out the kernel text/data/bss from the available physical
+	 * memory.
+	 */
+	kstart = __pa(CONFIG_KERNEL_START); /* kernel start */
+	/* kernel size */
+	ksize = PAGE_ALIGN(((u32)_end - (u32)CONFIG_KERNEL_START));
+	memblock_reserve(kstart, ksize);
+
+#if defined(CONFIG_BLK_DEV_INITRD)
+	/* Remove the init RAM disk from the available memory. */
+	if (initrd_start) {
+		unsigned long size;
+		size = initrd_end - initrd_start;
+		memblock_reserve(__virt_to_phys(initrd_start), size);
+	}
+#endif /* CONFIG_BLK_DEV_INITRD */
+
+	/* Initialize the MMU hardware */
+	mmu_init_hw();
+
+	/* Map in all of RAM starting at CONFIG_KERNEL_START */
+	mapin_ram();
+
+	/* Extend vmalloc and ioremap area as big as possible */
+#ifdef CONFIG_HIGHMEM
+	ioremap_base = ioremap_bot = PKMAP_BASE;
+#else
+	ioremap_base = ioremap_bot = FIXADDR_START;
+#endif
+
+	/* Initialize the context management stuff */
+	mmu_context_init();
+
+	/* Shortly after that, the entire linear mapping will be available */
+	/* This will also cause that unflatten device tree will be allocated
+	 * inside 768MB limit */
+	memblock_set_current_limit(memory_start + lowmem_size - 1);
+
+	parse_early_param();
+
+	early_init_fdt_scan_reserved_mem();
+
+	/* CMA initialization */
+	dma_contiguous_reserve(memory_start + lowmem_size - 1);
+
+	memblock_dump_all();
+}
+
+static const pgprot_t protection_map[16] = {
+	[VM_NONE]					= PAGE_NONE,
+	[VM_READ]					= PAGE_READONLY_X,
+	[VM_WRITE]					= PAGE_COPY,
+	[VM_WRITE | VM_READ]				= PAGE_COPY_X,
+	[VM_EXEC]					= PAGE_READONLY,
+	[VM_EXEC | VM_READ]				= PAGE_READONLY_X,
+	[VM_EXEC | VM_WRITE]				= PAGE_COPY,
+	[VM_EXEC | VM_WRITE | VM_READ]			= PAGE_COPY_X,
+	[VM_SHARED]					= PAGE_NONE,
+	[VM_SHARED | VM_READ]				= PAGE_READONLY_X,
+	[VM_SHARED | VM_WRITE]				= PAGE_SHARED,
+	[VM_SHARED | VM_WRITE | VM_READ]		= PAGE_SHARED_X,
+	[VM_SHARED | VM_EXEC]				= PAGE_READONLY,
+	[VM_SHARED | VM_EXEC | VM_READ]			= PAGE_READONLY_X,
+	[VM_SHARED | VM_EXEC | VM_WRITE]		= PAGE_SHARED,
+	[VM_SHARED | VM_EXEC | VM_WRITE | VM_READ]	= PAGE_SHARED_X
+};
+DECLARE_VM_GET_PAGE_PROT
diff --git a/arch/microblaze/mm/mmu_context.c b/arch/microblaze/mm/mmu_context.c
new file mode 100644
index 000000000..cbc234816
--- /dev/null
+++ b/arch/microblaze/mm/mmu_context.c
@@ -0,0 +1,65 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * This file contains the routines for handling the MMU.
+ *
+ *    Copyright (C) 2007 Xilinx, Inc.  All rights reserved.
+ *
+ *  Derived from arch/ppc/mm/4xx_mmu.c:
+ *  -- paulus
+ *
+ *  Derived from arch/ppc/mm/init.c:
+ *    Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
+ *
+ *  Modifications by Paul Mackerras (PowerMac) (paulus@cs.anu.edu.au)
+ *  and Cort Dougan (PReP) (cort@cs.nmt.edu)
+ *    Copyright (C) 1996 Paul Mackerras
+ *  Amiga/APUS changes by Jesper Skov (jskov@cygnus.co.uk).
+ *
+ *  Derived from "arch/i386/mm/init.c"
+ *    Copyright (C) 1991, 1992, 1993, 1994  Linus Torvalds
+ */
+
+#include <linux/mm.h>
+#include <linux/init.h>
+
+#include <asm/tlbflush.h>
+#include <asm/mmu_context.h>
+
+mm_context_t next_mmu_context;
+unsigned long context_map[LAST_CONTEXT / BITS_PER_LONG + 1];
+atomic_t nr_free_contexts;
+struct mm_struct *context_mm[LAST_CONTEXT+1];
+
+/*
+ * Initialize the context management stuff.
+ */
+void __init mmu_context_init(void)
+{
+	/*
+	 * The use of context zero is reserved for the kernel.
+	 * This code assumes FIRST_CONTEXT < 32.
+	 */
+	context_map[0] = (1 << FIRST_CONTEXT) - 1;
+	next_mmu_context = FIRST_CONTEXT;
+	atomic_set(&nr_free_contexts, LAST_CONTEXT - FIRST_CONTEXT + 1);
+}
+
+/*
+ * Steal a context from a task that has one at the moment.
+ *
+ * This isn't an LRU system, it just frees up each context in
+ * turn (sort-of pseudo-random replacement :).  This would be the
+ * place to implement an LRU scheme if anyone were motivated to do it.
+ */
+void steal_context(void)
+{
+	struct mm_struct *mm;
+
+	/* free up context `next_mmu_context' */
+	/* if we shouldn't free context 0, don't... */
+	if (next_mmu_context < FIRST_CONTEXT)
+		next_mmu_context = FIRST_CONTEXT;
+	mm = context_mm[next_mmu_context];
+	flush_tlb_mm(mm);
+	destroy_context(mm);
+}
diff --git a/arch/microblaze/mm/pgtable.c b/arch/microblaze/mm/pgtable.c
new file mode 100644
index 000000000..e96dd1b7a
--- /dev/null
+++ b/arch/microblaze/mm/pgtable.c
@@ -0,0 +1,264 @@
+/*
+ *  This file contains the routines setting up the linux page tables.
+ *
+ * Copyright (C) 2008 Michal Simek
+ * Copyright (C) 2008 PetaLogix
+ *
+ *    Copyright (C) 2007 Xilinx, Inc.  All rights reserved.
+ *
+ *  Derived from arch/ppc/mm/pgtable.c:
+ *    -- paulus
+ *
+ *  Derived from arch/ppc/mm/init.c:
+ *    Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
+ *
+ *  Modifications by Paul Mackerras (PowerMac) (paulus@cs.anu.edu.au)
+ *  and Cort Dougan (PReP) (cort@cs.nmt.edu)
+ *    Copyright (C) 1996 Paul Mackerras
+ *  Amiga/APUS changes by Jesper Skov (jskov@cygnus.co.uk).
+ *
+ *  Derived from "arch/i386/mm/init.c"
+ *    Copyright (C) 1991, 1992, 1993, 1994  Linus Torvalds
+ *
+ *  This file is subject to the terms and conditions of the GNU General
+ *  Public License.  See the file COPYING in the main directory of this
+ *  archive for more details.
+ *
+ */
+
+#include <linux/export.h>
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/vmalloc.h>
+#include <linux/init.h>
+#include <linux/mm_types.h>
+#include <linux/pgtable.h>
+#include <linux/memblock.h>
+#include <linux/kallsyms.h>
+
+#include <asm/pgalloc.h>
+#include <linux/io.h>
+#include <asm/mmu.h>
+#include <asm/sections.h>
+#include <asm/fixmap.h>
+
+unsigned long ioremap_base;
+unsigned long ioremap_bot;
+EXPORT_SYMBOL(ioremap_bot);
+
+static void __iomem *__ioremap(phys_addr_t addr, unsigned long size,
+		unsigned long flags)
+{
+	unsigned long v, i;
+	phys_addr_t p;
+	int err;
+
+	/*
+	 * Choose an address to map it to.
+	 * Once the vmalloc system is running, we use it.
+	 * Before then, we use space going down from ioremap_base
+	 * (ioremap_bot records where we're up to).
+	 */
+	p = addr & PAGE_MASK;
+	size = PAGE_ALIGN(addr + size) - p;
+
+	/*
+	 * Don't allow anybody to remap normal RAM that we're using.
+	 * mem_init() sets high_memory so only do the check after that.
+	 *
+	 * However, allow remap of rootfs: TBD
+	 */
+
+	if (mem_init_done &&
+		p >= memory_start && p < virt_to_phys(high_memory) &&
+		!(p >= __virt_to_phys((phys_addr_t)__bss_stop) &&
+		p < __virt_to_phys((phys_addr_t)__bss_stop))) {
+		pr_warn("__ioremap(): phys addr "PTE_FMT" is RAM lr %ps\n",
+			(unsigned long)p, __builtin_return_address(0));
+		return NULL;
+	}
+
+	if (size == 0)
+		return NULL;
+
+	/*
+	 * Is it already mapped? If the whole area is mapped then we're
+	 * done, otherwise remap it since we want to keep the virt addrs for
+	 * each request contiguous.
+	 *
+	 * We make the assumption here that if the bottom and top
+	 * of the range we want are mapped then it's mapped to the
+	 * same virt address (and this is contiguous).
+	 *  -- Cort
+	 */
+
+	if (mem_init_done) {
+		struct vm_struct *area;
+		area = get_vm_area(size, VM_IOREMAP);
+		if (area == NULL)
+			return NULL;
+		v = (unsigned long) area->addr;
+	} else {
+		v = (ioremap_bot -= size);
+	}
+
+	if ((flags & _PAGE_PRESENT) == 0)
+		flags |= _PAGE_KERNEL;
+	if (flags & _PAGE_NO_CACHE)
+		flags |= _PAGE_GUARDED;
+
+	err = 0;
+	for (i = 0; i < size && err == 0; i += PAGE_SIZE)
+		err = map_page(v + i, p + i, flags);
+	if (err) {
+		if (mem_init_done)
+			vfree((void *)v);
+		return NULL;
+	}
+
+	return (void __iomem *) (v + ((unsigned long)addr & ~PAGE_MASK));
+}
+
+void __iomem *ioremap(phys_addr_t addr, unsigned long size)
+{
+	return __ioremap(addr, size, _PAGE_NO_CACHE);
+}
+EXPORT_SYMBOL(ioremap);
+
+void iounmap(volatile void __iomem *addr)
+{
+	if ((__force void *)addr > high_memory &&
+					(unsigned long) addr < ioremap_bot)
+		vfree((void *) (PAGE_MASK & (unsigned long) addr));
+}
+EXPORT_SYMBOL(iounmap);
+
+
+int map_page(unsigned long va, phys_addr_t pa, int flags)
+{
+	p4d_t *p4d;
+	pud_t *pud;
+	pmd_t *pd;
+	pte_t *pg;
+	int err = -ENOMEM;
+
+	/* Use upper 10 bits of VA to index the first level map */
+	p4d = p4d_offset(pgd_offset_k(va), va);
+	pud = pud_offset(p4d, va);
+	pd = pmd_offset(pud, va);
+	/* Use middle 10 bits of VA to index the second-level map */
+	pg = pte_alloc_kernel(pd, va); /* from powerpc - pgtable.c */
+	/* pg = pte_alloc_kernel(&init_mm, pd, va); */
+
+	if (pg != NULL) {
+		err = 0;
+		set_pte_at(&init_mm, va, pg, pfn_pte(pa >> PAGE_SHIFT,
+				__pgprot(flags)));
+		if (unlikely(mem_init_done))
+			_tlbie(va);
+	}
+	return err;
+}
+
+/*
+ * Map in all of physical memory starting at CONFIG_KERNEL_START.
+ */
+void __init mapin_ram(void)
+{
+	unsigned long v, p, s, f;
+
+	v = CONFIG_KERNEL_START;
+	p = memory_start;
+	for (s = 0; s < lowmem_size; s += PAGE_SIZE) {
+		f = _PAGE_PRESENT | _PAGE_ACCESSED |
+				_PAGE_SHARED | _PAGE_HWEXEC;
+		if (!is_kernel_text(v))
+			f |= _PAGE_WRENABLE;
+		else
+			/* On the MicroBlaze, no user access
+			   forces R/W kernel access */
+			f |= _PAGE_USER;
+		map_page(v, p, f);
+		v += PAGE_SIZE;
+		p += PAGE_SIZE;
+	}
+}
+
+/* is x a power of 2? */
+#define is_power_of_2(x)	((x) != 0 && (((x) & ((x) - 1)) == 0))
+
+/* Scan the real Linux page tables and return a PTE pointer for
+ * a virtual address in a context.
+ * Returns true (1) if PTE was found, zero otherwise.  The pointer to
+ * the PTE pointer is unmodified if PTE is not found.
+ */
+static int get_pteptr(struct mm_struct *mm, unsigned long addr, pte_t **ptep)
+{
+	pgd_t	*pgd;
+	p4d_t	*p4d;
+	pud_t	*pud;
+	pmd_t	*pmd;
+	pte_t	*pte;
+	int     retval = 0;
+
+	pgd = pgd_offset(mm, addr & PAGE_MASK);
+	if (pgd) {
+		p4d = p4d_offset(pgd, addr & PAGE_MASK);
+		pud = pud_offset(p4d, addr & PAGE_MASK);
+		pmd = pmd_offset(pud, addr & PAGE_MASK);
+		if (pmd_present(*pmd)) {
+			pte = pte_offset_kernel(pmd, addr & PAGE_MASK);
+			if (pte) {
+				retval = 1;
+				*ptep = pte;
+			}
+		}
+	}
+	return retval;
+}
+
+/* Find physical address for this virtual address.  Normally used by
+ * I/O functions, but anyone can call it.
+ */
+unsigned long iopa(unsigned long addr)
+{
+	unsigned long pa;
+
+	pte_t *pte;
+	struct mm_struct *mm;
+
+	/* Allow mapping of user addresses (within the thread)
+	 * for DMA if necessary.
+	 */
+	if (addr < TASK_SIZE)
+		mm = current->mm;
+	else
+		mm = &init_mm;
+
+	pa = 0;
+	if (get_pteptr(mm, addr, &pte))
+		pa = (pte_val(*pte) & PAGE_MASK) | (addr & ~PAGE_MASK);
+
+	return pa;
+}
+
+__ref pte_t *pte_alloc_one_kernel(struct mm_struct *mm)
+{
+	if (mem_init_done)
+		return __pte_alloc_one_kernel(mm);
+	else
+		return memblock_alloc_try_nid(PAGE_SIZE, PAGE_SIZE,
+					      MEMBLOCK_LOW_LIMIT,
+					      memory_start + kernel_tlb,
+					      NUMA_NO_NODE);
+}
+
+void __set_fixmap(enum fixed_addresses idx, phys_addr_t phys, pgprot_t flags)
+{
+	unsigned long address = __fix_to_virt(idx);
+
+	if (idx >= __end_of_fixed_addresses)
+		BUG();
+
+	map_page(address, phys, pgprot_val(flags));
+}