Skip to contentSkip to conversation

Support strict kernel memory permissions for security

PATCH Superseded Zong Li 2 participants 21 messages Mar 9, 2020 – Apr 7, 2020

Subsystems (2)

v3 of 5 Superseded Latest is v5 →

lore
Share & tools
lore

Zong Li

author
#0 lore raw

v3 cover Support strict kernel memory permissions for security

The main purpose of this patch series is changing the kernel mapping permission , make sure that code is not writeable, data is not executable, and read-only data is neither writable nor executable.

This patch series also supports the relevant implementations such as ARCH_HAS_SET_MEMORY, ARCH_HAS_SET_DIRECT_MAP, ARCH_SUPPORTS_DEBUG_PAGEALLOC and DEBUG_WX.

Changes in v3:
- Fix build error on nommu configuration. We already support nommu on RISC-V, so we should consider nommu case and test not only rv32/64, but also nommu.

Changes in v2:
- Use _data to specify the start of data section with write permission.
- Change ftrace patch text implementaion.
- Separate DEBUG_WX patch to another patchset.

Zong Li (9):
riscv: add ARCH_HAS_SET_MEMORY support
riscv: add ARCH_HAS_SET_DIRECT_MAP support
riscv: add ARCH_SUPPORTS_DEBUG_PAGEALLOC support
riscv: move exception table immediately after RO_DATA
riscv: add alignment for text, rodata and data sections
riscv: add STRICT_KERNEL_RWX support
riscv: add macro to get instruction length
riscv: introduce interfaces to patch kernel code
riscv: patch code by fixmap mapping

 arch/riscv/Kconfig                  |   6 +
 arch/riscv/include/asm/bug.h        |   8 ++
 arch/riscv/include/asm/fixmap.h     |   2 +
 arch/riscv/include/asm/patch.h      |  12 ++
 arch/riscv/include/asm/set_memory.h |  48 +++++++
 arch/riscv/kernel/Makefile          |   4 +-
 arch/riscv/kernel/ftrace.c          |  13 +-
 arch/riscv/kernel/patch.c           | 120 ++++++++++++++++++
 arch/riscv/kernel/traps.c           |   3 +-
 arch/riscv/kernel/vmlinux.lds.S     |  11 +-
 arch/riscv/mm/Makefile              |   2 +-
 arch/riscv/mm/init.c                |  44 +++++++
 arch/riscv/mm/pageattr.c            | 187 ++++++++++++++++++++++++++++
 13 files changed, 445 insertions(+), 15 deletions(-)
 create mode 100644 arch/riscv/include/asm/patch.h
 create mode 100644 arch/riscv/include/asm/set_memory.h
 create mode 100644 arch/riscv/kernel/patch.c
 create mode 100644 arch/riscv/mm/pageattr.c
Signature · 1 line
2.25.1
author
#1 lore raw

PATCH v3 1/9 riscv: add ARCH_HAS_SET_MEMORY support

Add set_memory_ro/rw/x/nx architecture hooks to change the page attribution.

Use own set_memory.h rather than generic set_memory.h (i.e. include/asm-generic/set_memory.h), because we want to add other function prototypes here.

---
 arch/riscv/Kconfig                  |   1 +
 arch/riscv/include/asm/set_memory.h |  24 +++++
 arch/riscv/mm/Makefile              |   2 +-
 arch/riscv/mm/pageattr.c            | 150 ++++++++++++++++++++++++++++
 4 files changed, 176 insertions(+), 1 deletion(-)
 create mode 100644 arch/riscv/include/asm/set_memory.h
 create mode 100644 arch/riscv/mm/pageattr.c
arch/riscv/Kconfig +1 -0
index af9b03609c5a..9ab592aa1b5f 100644
@@ -60,6 +60,7 @@ config RISCV
 	select HAVE_EBPF_JIT if 64BIT
 	select EDAC_SUPPORT
 	select ARCH_HAS_GIGANTIC_PAGE
+	select ARCH_HAS_SET_MEMORY
 	select ARCH_WANT_HUGE_PMD_SHARE if 64BIT
 	select SPARSEMEM_STATIC if 32BIT
 	select ARCH_WANT_DEFAULT_TOPDOWN_MMAP_LAYOUT if MMU
arch/riscv/include/asm/set_memory.h new file +24 -0
new file mode 100644
index 000000000000..79a810f0f38b
@@ -0,0 +1,24 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Copyright (C) 2019 SiFive
+ */
+
+#ifndef _ASM_RISCV_SET_MEMORY_H
+#define _ASM_RISCV_SET_MEMORY_H
+
+/*
+ * Functions to change memory attributes.
+ */
+#ifdef CONFIG_MMU
+int set_memory_ro(unsigned long addr, int numpages);
+int set_memory_rw(unsigned long addr, int numpages);
+int set_memory_x(unsigned long addr, int numpages);
+int set_memory_nx(unsigned long addr, int numpages);
+#else
+static inline int set_memory_ro(unsigned long addr, int numpages) { return 0; }
+static inline int set_memory_rw(unsigned long addr, int numpages) { return 0; }
+static inline int set_memory_x(unsigned long addr, int numpages) { return 0; }
+static inline int set_memory_nx(unsigned long addr, int numpages) { return 0; }
+#endif
+
+#endif /* _ASM_RISCV_SET_MEMORY_H */
arch/riscv/mm/Makefile +1 -1
index 814e16a8d68a..363ef01c30b1 100644
@@ -7,7 +7,7 @@ endif
 
 obj-y += init.o
 obj-y += extable.o
-obj-$(CONFIG_MMU) += fault.o
+obj-$(CONFIG_MMU) += fault.o pageattr.o
 obj-y += cacheflush.o
 obj-y += context.o
 
arch/riscv/mm/pageattr.c new file +150 -0

150 lines not on this page · raw message

Signature · 1 line
2.25.1
author
#2 lore raw

PATCH v3 2/9 riscv: add ARCH_HAS_SET_DIRECT_MAP support

In reply to Zong Li

Add set_direct_map_*() functions for setting the direct map alias for the page to its default permissions and to an invalid state that cannot be cached in a TLB. (See d253ca0c ("x86/mm/cpa: Add set_direct_map_*() functions")) Add a similar implementation for RISC-V.

---
 arch/riscv/Kconfig                  |  1 +
 arch/riscv/include/asm/set_memory.h |  3 +++
 arch/riscv/mm/pageattr.c            | 24 ++++++++++++++++++++++++
 3 files changed, 28 insertions(+)
arch/riscv/Kconfig +1 -0
index 9ab592aa1b5f..71fabb0ffdbe 100644
@@ -60,6 +60,7 @@ config RISCV
 	select HAVE_EBPF_JIT if 64BIT
 	select EDAC_SUPPORT
 	select ARCH_HAS_GIGANTIC_PAGE
+	select ARCH_HAS_SET_DIRECT_MAP
 	select ARCH_HAS_SET_MEMORY
 	select ARCH_WANT_HUGE_PMD_SHARE if 64BIT
 	select SPARSEMEM_STATIC if 32BIT
arch/riscv/include/asm/set_memory.h +3 -0
index 79a810f0f38b..620d81c372d9 100644
@@ -21,4 +21,7 @@ static inline int set_memory_x(unsigned long addr, int numpages) { return 0; }
 static inline int set_memory_nx(unsigned long addr, int numpages) { return 0; }
 #endif
 
+int set_direct_map_invalid_noflush(struct page *page);
+int set_direct_map_default_noflush(struct page *page);
+
 #endif /* _ASM_RISCV_SET_MEMORY_H */
arch/riscv/mm/pageattr.c +24 -0
index fcd59ef2835b..7be6cd67e2ef 100644
@@ -148,3 +148,27 @@ int set_memory_nx(unsigned long addr, int numpages)
 {
 	return __set_memory(addr, numpages, __pgprot(0), __pgprot(_PAGE_EXEC));
 }
+
+int set_direct_map_invalid_noflush(struct page *page)
+{
+	unsigned long start = (unsigned long)page_address(page);
+	unsigned long end = start + PAGE_SIZE;
+	struct pageattr_masks masks = {
+		.set_mask = __pgprot(0),
+		.clear_mask = __pgprot(_PAGE_PRESENT)
+	};
+
+	return walk_page_range(&init_mm, start, end, &pageattr_ops, &masks);
+}
+
+int set_direct_map_default_noflush(struct page *page)
+{
+	unsigned long start = (unsigned long)page_address(page);
+	unsigned long end = start + PAGE_SIZE;
+	struct pageattr_masks masks = {
+		.set_mask = PAGE_KERNEL,
+		.clear_mask = __pgprot(0)
+	};
+
+	return walk_page_range(&init_mm, start, end, &pageattr_ops, &masks);
+}
Signature · 1 line
2.25.1
author
#3 lore raw

PATCH v3 3/9 riscv: add ARCH_SUPPORTS_DEBUG_PAGEALLOC support

In reply to Zong Li

ARCH_SUPPORTS_DEBUG_PAGEALLOC provides a hook to map and unmap pages for debugging purposes. Implement the __kernel_map_pages functions to fill the poison pattern.

---
 arch/riscv/Kconfig       |  3 +++
 arch/riscv/mm/pageattr.c | 13 +++++++++++++
 2 files changed, 16 insertions(+)
arch/riscv/Kconfig +3 -0
index 71fabb0ffdbe..54437d7662a5 100644
@@ -133,6 +133,9 @@ config ARCH_SELECT_MEMORY_MODEL
 config ARCH_WANT_GENERAL_HUGETLB
 	def_bool y
 
+config ARCH_SUPPORTS_DEBUG_PAGEALLOC
+	def_bool y
+
 config SYS_SUPPORTS_HUGETLBFS
 	def_bool y
 
arch/riscv/mm/pageattr.c +13 -0
index 7be6cd67e2ef..728759eb530a 100644
@@ -172,3 +172,16 @@ int set_direct_map_default_noflush(struct page *page)
 
 	return walk_page_range(&init_mm, start, end, &pageattr_ops, &masks);
 }
+
+void __kernel_map_pages(struct page *page, int numpages, int enable)
+{
+	if (!debug_pagealloc_enabled())
+		return;
+
+	if (enable)
+		__set_memory((unsigned long)page_address(page), numpages,
+			     __pgprot(_PAGE_PRESENT), __pgprot(0));
+	else
+		__set_memory((unsigned long)page_address(page), numpages,
+			     __pgprot(0), __pgprot(_PAGE_PRESENT));
+}
Signature · 1 line
2.25.1
author
#4 lore raw

PATCH v3 4/9 riscv: move exception table immediately after RO_DATA

In reply to Zong Li

Move EXCEPTION_TABLE immediately after RO_DATA. Make it easy to set the attribution of the sections which should be read-only at a time. Add _data to specify the start of data section with write permission. This patch is prepared for STRICT_KERNEL_RWX support.

---
 arch/riscv/kernel/vmlinux.lds.S | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
arch/riscv/kernel/vmlinux.lds.S +4 -2
index 1e0193ded420..02e948b620af 100644
@@ -58,6 +58,10 @@ SECTIONS
 		*(.srodata*)
 	}
 
+	EXCEPTION_TABLE(0x10)
+
+	_data = .;
+
 	RW_DATA(L1_CACHE_BYTES, PAGE_SIZE, THREAD_SIZE)
 	.sdata : {
 		__global_pointer$ = . + 0x800;
@@ -69,8 +73,6 @@ SECTIONS
 
 	BSS_SECTION(PAGE_SIZE, PAGE_SIZE, 0)
 
-	EXCEPTION_TABLE(0x10)
-
 	.rel.dyn : {
 		*(.rel.dyn*)
 	}
Signature · 1 line
2.25.1
author
#5 lore raw

PATCH v3 5/9 riscv: add alignment for text, rodata and data sections

In reply to Zong Li

The kernel mapping will tried to optimize its mapping by using bigger size. In rv64, it tries to use PMD_SIZE, and tryies to use PGDIR_SIZE in rv32. To ensure that the start address of these sections could fit the mapping entry size, make them align to the biggest alignment.

Define a macro SECTION_ALIGN because the HPAGE_SIZE or PMD_SIZE, etc., are invisible in linker script.

This patch is prepared for STRICT_KERNEL_RWX support.

---
 arch/riscv/include/asm/set_memory.h | 13 +++++++++++++
 arch/riscv/kernel/vmlinux.lds.S     |  5 ++++-
 2 files changed, 17 insertions(+), 1 deletion(-)
arch/riscv/include/asm/set_memory.h +13 -0
index 620d81c372d9..4c5bae7ca01c 100644
@@ -6,6 +6,7 @@
 #ifndef _ASM_RISCV_SET_MEMORY_H
 #define _ASM_RISCV_SET_MEMORY_H
 
+#ifndef __ASSEMBLY__
 /*
  * Functions to change memory attributes.
  */
@@ -24,4 +25,16 @@ static inline int set_memory_nx(unsigned long addr, int numpages) { return 0; }
 int set_direct_map_invalid_noflush(struct page *page);
 int set_direct_map_default_noflush(struct page *page);
 
+#endif /* __ASSEMBLY__ */
+
+#ifdef CONFIG_ARCH_HAS_STRICT_KERNEL_RWX
+#ifdef CONFIG_64BIT
+#define SECTION_ALIGN (1 << 21)
+#else
+#define SECTION_ALIGN (1 << 22)
+#endif
+#else /* !CONFIG_ARCH_HAS_STRICT_KERNEL_RWX */
+#define SECTION_ALIGN L1_CACHE_BYTES
+#endif /* CONFIG_ARCH_HAS_STRICT_KERNEL_RWX */
+
 #endif /* _ASM_RISCV_SET_MEMORY_H */
arch/riscv/kernel/vmlinux.lds.S +4 -1
index 02e948b620af..ef87deea0350 100644
@@ -9,6 +9,7 @@
 #include <asm/page.h>
 #include <asm/cache.h>
 #include <asm/thread_info.h>
+#include <asm/set_memory.h>
 
 OUTPUT_ARCH(riscv)
 ENTRY(_start)
@@ -36,6 +37,7 @@ SECTIONS
 	PERCPU_SECTION(L1_CACHE_BYTES)
 	__init_end = .;
 
+	. = ALIGN(SECTION_ALIGN);
 	.text : {
 		_text = .;
 		_stext = .;
@@ -53,13 +55,14 @@ SECTIONS
 
 	/* Start of data section */
 	_sdata = .;
-	RO_DATA(L1_CACHE_BYTES)
+	RO_DATA(SECTION_ALIGN)
 	.srodata : {
 		*(.srodata*)
 	}
 
 	EXCEPTION_TABLE(0x10)
 
+	. = ALIGN(SECTION_ALIGN);
 	_data = .;
 
 	RW_DATA(L1_CACHE_BYTES, PAGE_SIZE, THREAD_SIZE)
Signature · 1 line
2.25.1
author
#6 lore raw

PATCH v3 6/9 riscv: add STRICT_KERNEL_RWX support

In reply to Zong Li

The commit contains that make text section as non-writable, rodata section as read-only, and data section as non-executable.

The init section should be changed to non-executable.

---
 arch/riscv/Kconfig                  |  1 +
 arch/riscv/include/asm/set_memory.h |  8 ++++++
 arch/riscv/mm/init.c                | 44 +++++++++++++++++++++++++++++
 3 files changed, 53 insertions(+)
arch/riscv/Kconfig +1 -0
index 54437d7662a5..5fbae6380b32 100644
@@ -62,6 +62,7 @@ config RISCV
 	select ARCH_HAS_GIGANTIC_PAGE
 	select ARCH_HAS_SET_DIRECT_MAP
 	select ARCH_HAS_SET_MEMORY
+	select ARCH_HAS_STRICT_KERNEL_RWX
 	select ARCH_WANT_HUGE_PMD_SHARE if 64BIT
 	select SPARSEMEM_STATIC if 32BIT
 	select ARCH_WANT_DEFAULT_TOPDOWN_MMAP_LAYOUT if MMU
arch/riscv/include/asm/set_memory.h +8 -0
index 4c5bae7ca01c..c38df4771c09 100644
@@ -22,6 +22,14 @@ static inline int set_memory_x(unsigned long addr, int numpages) { return 0; }
 static inline int set_memory_nx(unsigned long addr, int numpages) { return 0; }
 #endif
 
+#ifdef CONFIG_STRICT_KERNEL_RWX
+void set_kernel_text_ro(void);
+void set_kernel_text_rw(void);
+#else
+static inline void set_kernel_text_ro(void) { }
+static inline void set_kernel_text_rw(void) { }
+#endif
+
 int set_direct_map_invalid_noflush(struct page *page);
 int set_direct_map_default_noflush(struct page *page);
 
arch/riscv/mm/init.c +44 -0

62 lines not on this page · raw message

Signature · 1 line
2.25.1
author
#7 lore raw

PATCH v3 7/9 riscv: add macro to get instruction length

In reply to Zong Li

Extract the calculation of instruction length for common use.

---
 arch/riscv/include/asm/bug.h | 8 ++++++++
 arch/riscv/kernel/traps.c    | 3 ++-
 2 files changed, 10 insertions(+), 1 deletion(-)
arch/riscv/include/asm/bug.h +8 -0
index 75604fec1b1b..d6f1ec08d97b 100644
@@ -19,6 +19,14 @@
 #define __BUG_INSN_32	_UL(0x00100073) /* ebreak */
 #define __BUG_INSN_16	_UL(0x9002) /* c.ebreak */
 
+#define GET_INSN_LENGTH(insn)						\
+({									\
+	unsigned long __len;						\
+	__len = ((insn & __INSN_LENGTH_MASK) == __INSN_LENGTH_32) ?	\
+		4UL : 2UL;						\
+	__len;								\
+})
+
 typedef u32 bug_insn_t;
 
 #ifdef CONFIG_GENERIC_BUG_RELATIVE_POINTERS
arch/riscv/kernel/traps.c +2 -1
index ffb3d94bf0cc..a4d136355f78 100644
@@ -118,7 +118,8 @@ static inline unsigned long get_break_insn_length(unsigned long pc)
 
 	if (probe_kernel_address((bug_insn_t *)pc, insn))
 		return 0;
-	return (((insn & __INSN_LENGTH_MASK) == __INSN_LENGTH_32) ? 4UL : 2UL);
+
+	return GET_INSN_LENGTH(insn);
 }
 
 asmlinkage __visible void do_trap_break(struct pt_regs *regs)
Signature · 1 line
2.25.1
author
#8 lore raw

PATCH v3 8/9 riscv: introduce interfaces to patch kernel code

In reply to Zong Li

On strict kernel memory permission, we couldn't patch code without writable permission. Preserve two holes in fixmap area, so we can map the kernel code temporarily to fixmap area, then patch the instructions.

We need two pages here because we support the compressed instruction, so the instruction might be align to 2 bytes. When patching the 32-bit length instruction which is 2 bytes alignment, it will across two pages.

Introduce two interfaces to patch kernel code:
riscv_patch_text_nosync:
- patch code without synchronization, it's caller's responsibility to synchronize all CPUs if needed.
riscv_patch_text:
- patch code and always synchronize with stop_machine()

---
 arch/riscv/include/asm/fixmap.h |   2 +
 arch/riscv/include/asm/patch.h  |  12 ++++
 arch/riscv/kernel/Makefile      |   4 +-
 arch/riscv/kernel/patch.c       | 120 ++++++++++++++++++++++++++++++++
 4 files changed, 137 insertions(+), 1 deletion(-)
 create mode 100644 arch/riscv/include/asm/patch.h
 create mode 100644 arch/riscv/kernel/patch.c
arch/riscv/include/asm/fixmap.h +2 -0
index 42d2c42f3cc9..2368d49eb4ef 100644
@@ -27,6 +27,8 @@ enum fixed_addresses {
 	FIX_FDT = FIX_FDT_END + FIX_FDT_SIZE / PAGE_SIZE - 1,
 	FIX_PTE,
 	FIX_PMD,
+	FIX_TEXT_POKE1,
+	FIX_TEXT_POKE0,
 	FIX_EARLYCON_MEM_BASE,
 	__end_of_fixed_addresses
 };
arch/riscv/include/asm/patch.h new file +12 -0
new file mode 100644
index 000000000000..b5918a6e0615
@@ -0,0 +1,12 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Copyright (C) 2020 SiFive
+ */
+
+#ifndef _ASM_RISCV_PATCH_H
+#define _ASM_RISCV_PATCH_H
+
+int riscv_patch_text_nosync(void *addr, const void *insns, size_t len);
+int riscv_patch_text(void *addr, u32 insn);
+
+#endif /* _ASM_RISCV_PATCH_H */
arch/riscv/kernel/Makefile +3 -1
index f40205cb9a22..d189bd3d8501 100644
@@ -4,7 +4,8 @@
 #
 
 ifdef CONFIG_FTRACE
-CFLAGS_REMOVE_ftrace.o = -pg
+CFLAGS_REMOVE_ftrace.o	= -pg
+CFLAGS_REMOVE_patch.o	= -pg
 endif
 
 extra-y += head.o
@@ -26,6 +27,7 @@ obj-y	+= traps.o
 obj-y	+= riscv_ksyms.o
 obj-y	+= stacktrace.o
 obj-y	+= cacheinfo.o
+obj-y	+= patch.o
 obj-$(CONFIG_MMU) += vdso.o vdso/
 
 obj-$(CONFIG_RISCV_M_MODE)	+= clint.o
arch/riscv/kernel/patch.c new file +120 -0

120 lines not on this page · raw message

Signature · 1 line
2.25.1
maintainer
#9 lore raw

Re: [PATCH v3 8/9] riscv: introduce interfaces to patch kernel code

Hi,

Quoted 97 lines · On Tue, 10 Mar 2020 00:55:43 +0800 Zong Li <[email protected]> wrote:

On strict kernel memory permission, we couldn't patch code without writable permission. Preserve two holes in fixmap area, so we can map the kernel code temporarily to fixmap area, then patch the instructions.

We need two pages here because we support the compressed instruction, so the instruction might be align to 2 bytes. When patching the 32-bit length instruction which is 2 bytes alignment, it will across two pages.

Introduce two interfaces to patch kernel code:
riscv_patch_text_nosync:
- patch code without synchronization, it's caller's responsibility to synchronize all CPUs if needed.
riscv_patch_text:
- patch code and always synchronize with stop_machine()

---
 arch/riscv/include/asm/fixmap.h |   2 +
 arch/riscv/include/asm/patch.h  |  12 ++++
 arch/riscv/kernel/Makefile      |   4 +-
 arch/riscv/kernel/patch.c       | 120 ++++++++++++++++++++++++++++++++
 4 files changed, 137 insertions(+), 1 deletion(-)
 create mode 100644 arch/riscv/include/asm/patch.h
 create mode 100644 arch/riscv/kernel/patch.c
arch/riscv/include/asm/fixmap.h +2 -0
index 42d2c42f3cc9..2368d49eb4ef 100644
@@ -27,6 +27,8 @@ enum fixed_addresses {
 	FIX_FDT = FIX_FDT_END + FIX_FDT_SIZE / PAGE_SIZE - 1,
 	FIX_PTE,
 	FIX_PMD,
+	FIX_TEXT_POKE1,
+	FIX_TEXT_POKE0,
 	FIX_EARLYCON_MEM_BASE,
 	__end_of_fixed_addresses
 };
arch/riscv/include/asm/patch.h new file +12 -0
new file mode 100644
index 000000000000..b5918a6e0615
@@ -0,0 +1,12 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Copyright (C) 2020 SiFive
+ */
+
+#ifndef _ASM_RISCV_PATCH_H
+#define _ASM_RISCV_PATCH_H
+
+int riscv_patch_text_nosync(void *addr, const void *insns, size_t len);
+int riscv_patch_text(void *addr, u32 insn);
+
+#endif /* _ASM_RISCV_PATCH_H */
arch/riscv/kernel/Makefile +3 -1
index f40205cb9a22..d189bd3d8501 100644
@@ -4,7 +4,8 @@
 #
 
 ifdef CONFIG_FTRACE
-CFLAGS_REMOVE_ftrace.o = -pg
+CFLAGS_REMOVE_ftrace.o	= -pg
+CFLAGS_REMOVE_patch.o	= -pg
 endif
 
 extra-y += head.o
@@ -26,6 +27,7 @@ obj-y	+= traps.o
 obj-y	+= riscv_ksyms.o
 obj-y	+= stacktrace.o
 obj-y	+= cacheinfo.o
+obj-y	+= patch.o
 obj-$(CONFIG_MMU) += vdso.o vdso/
 
 obj-$(CONFIG_RISCV_M_MODE)	+= clint.o
arch/riscv/kernel/patch.c new file +13 -0
new file mode 100644
index 000000000000..8a4fc65ee022
@@ -0,0 +1,120 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (C) 2020 SiFive
+ */
+
+#include <linux/spinlock.h>
+#include <linux/mm.h>
+#include <linux/uaccess.h>
+#include <linux/stop_machine.h>
+#include <asm/kprobes.h>
+#include <asm/cacheflush.h>
+#include <asm/fixmap.h>
+

arch/riscv/kernel/patch.c line 23

Quoted 10 lines
arch/riscv/kernel/patch.c new file +10 -0
+struct riscv_insn_patch {
+	void *addr;
+	u32 insn;
+	atomic_t cpu_count;
+};
+
+#ifdef CONFIG_MMU
+static DEFINE_RAW_SPINLOCK(patch_lock);
+
+static void __kprobes *patch_map(void *addr, int fixmap)

Please use NOKPROBE_SYMBOL() instead of __kprobes. __kprobes is old style.

Quoted 13 lines
arch/riscv/kernel/patch.c new file +13 -0
+{
+	uintptr_t uintaddr = (uintptr_t) addr;
+	struct page *page;
+
+	if (core_kernel_text(uintaddr))
+		page = phys_to_page(__pa_symbol(addr));
+	else if (IS_ENABLED(CONFIG_STRICT_MODULE_RWX))
+		page = vmalloc_to_page(addr);
+	else
+		return addr;
+
+	BUG_ON(!page);
+

arch/riscv/kernel/patch.c line 46

Quoted 10 lines
arch/riscv/kernel/patch.c new file +10 -0
+	return (void *)set_fixmap_offset(fixmap, page_to_phys(page) +
+					 (uintaddr & ~PAGE_MASK));
+}
+
+static void __kprobes patch_unmap(int fixmap)
+{
+	clear_fixmap(fixmap);
+}
+
+static int __kprobes riscv_insn_write(void *addr, const void *insn, size_t len)

Why would you add "riscv_" prefix for those functions? It seems a bit odd.

arch/riscv/kernel/patch.c line 53

Quoted 7 lines
arch/riscv/kernel/patch.c new file +7 -0
+{
+	void *waddr = addr;
+	bool across_pages = (((uintptr_t) addr & ~PAGE_MASK) + len) > PAGE_SIZE;
+	unsigned long flags = 0;
+	int ret;
+
+	raw_spin_lock_irqsave(&patch_lock, flags);

This looks a bit odd since stop_machine() is protected by its own mutex, and also the irq is already disabled here.

Thank you,

Quoted 70 lines
arch/riscv/kernel/patch.c new file +67 -0
+
+	if (across_pages)
+		patch_map(addr + len, FIX_TEXT_POKE1);
+
+	waddr = patch_map(addr, FIX_TEXT_POKE0);
+
+	ret = probe_kernel_write(waddr, insn, len);
+
+	patch_unmap(FIX_TEXT_POKE0);
+
+	if (across_pages)
+		patch_unmap(FIX_TEXT_POKE1);
+
+	raw_spin_unlock_irqrestore(&patch_lock, flags);
+
+	return ret;
+}
+#else
+static int __kprobes riscv_insn_write(void *addr, const void *insn, size_t len)
+{
+	return probe_kernel_write(addr, insn, len);
+}
+#endif /* CONFIG_MMU */
+
+int __kprobes riscv_patch_text_nosync(void *addr, const void *insns, size_t len)
+{
+	u32 *tp = addr;
+	int ret;
+
+	ret = riscv_insn_write(tp, insns, len);
+
+	if (!ret)
+		flush_icache_range((uintptr_t) tp, (uintptr_t) tp + len);
+
+	return ret;
+}
+
+static int __kprobes riscv_patch_text_cb(void *data)
+{
+	struct riscv_insn_patch *patch = data;
+	int ret = 0;
+
+	if (atomic_inc_return(&patch->cpu_count) == 1) {
+		ret =
+		    riscv_patch_text_nosync(patch->addr, &patch->insn,
+					    GET_INSN_LENGTH(patch->insn));
+		atomic_inc(&patch->cpu_count);
+	} else {
+		while (atomic_read(&patch->cpu_count) <= num_online_cpus())
+			cpu_relax();
+		smp_mb();
+	}
+
+	return ret;
+}
+
+int __kprobes riscv_patch_text(void *addr, u32 insn)
+{
+	struct riscv_insn_patch patch = {
+		.addr = addr,
+		.insn = insn,
+		.cpu_count = ATOMIC_INIT(0),
+	};
+
+	return stop_machine_cpuslocked(riscv_patch_text_cb,
+				       &patch, cpu_online_mask);
+}
Signature · 1 line
2.25.1
Signature · 1 line
Masami Hiramatsu <[email protected]>
author
#10 lore raw

Re: [PATCH v3 8/9] riscv: introduce interfaces to patch kernel code

Quoted 133 lines · Masami Hiramatsu

Hi,

Quoted 97 lines · On Tue, 10 Mar 2020 00:55:43 +0800 Zong Li <[email protected]> wrote:

On strict kernel memory permission, we couldn't patch code without writable permission. Preserve two holes in fixmap area, so we can map the kernel code temporarily to fixmap area, then patch the instructions.

We need two pages here because we support the compressed instruction, so the instruction might be align to 2 bytes. When patching the 32-bit length instruction which is 2 bytes alignment, it will across two pages.

Introduce two interfaces to patch kernel code:
riscv_patch_text_nosync:
- patch code without synchronization, it's caller's responsibility to synchronize all CPUs if needed.
riscv_patch_text:
- patch code and always synchronize with stop_machine()

---
 arch/riscv/include/asm/fixmap.h |   2 +
 arch/riscv/include/asm/patch.h  |  12 ++++
 arch/riscv/kernel/Makefile      |   4 +-
 arch/riscv/kernel/patch.c       | 120 ++++++++++++++++++++++++++++++++
 4 files changed, 137 insertions(+), 1 deletion(-)
 create mode 100644 arch/riscv/include/asm/patch.h
 create mode 100644 arch/riscv/kernel/patch.c
arch/riscv/include/asm/fixmap.h +2 -0
index 42d2c42f3cc9..2368d49eb4ef 100644
@@ -27,6 +27,8 @@ enum fixed_addresses {
      FIX_FDT = FIX_FDT_END + FIX_FDT_SIZE / PAGE_SIZE - 1,
      FIX_PTE,
      FIX_PMD,
+     FIX_TEXT_POKE1,
+     FIX_TEXT_POKE0,
      FIX_EARLYCON_MEM_BASE,
      __end_of_fixed_addresses
 };
arch/riscv/include/asm/patch.h new file +12 -0
new file mode 100644
index 000000000000..b5918a6e0615
@@ -0,0 +1,12 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Copyright (C) 2020 SiFive
+ */
+
+#ifndef _ASM_RISCV_PATCH_H
+#define _ASM_RISCV_PATCH_H
+
+int riscv_patch_text_nosync(void *addr, const void *insns, size_t len);
+int riscv_patch_text(void *addr, u32 insn);
+
+#endif /* _ASM_RISCV_PATCH_H */
arch/riscv/kernel/Makefile +3 -1
index f40205cb9a22..d189bd3d8501 100644
@@ -4,7 +4,8 @@
 #
 
 ifdef CONFIG_FTRACE
-CFLAGS_REMOVE_ftrace.o = -pg
+CFLAGS_REMOVE_ftrace.o       = -pg
+CFLAGS_REMOVE_patch.o        = -pg
 endif
 
 extra-y += head.o
@@ -26,6 +27,7 @@ obj-y       += traps.o
 obj-y        += riscv_ksyms.o
 obj-y        += stacktrace.o
 obj-y        += cacheinfo.o
+obj-y        += patch.o
 obj-$(CONFIG_MMU) += vdso.o vdso/
 
 obj-$(CONFIG_RISCV_M_MODE)   += clint.o
arch/riscv/kernel/patch.c new file +13 -0
new file mode 100644
index 000000000000..8a4fc65ee022
@@ -0,0 +1,120 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (C) 2020 SiFive
+ */
+
+#include <linux/spinlock.h>
+#include <linux/mm.h>
+#include <linux/uaccess.h>
+#include <linux/stop_machine.h>
+#include <asm/kprobes.h>
+#include <asm/cacheflush.h>
+#include <asm/fixmap.h>
+

arch/riscv/kernel/patch.c line 23

Quoted 10 lines
arch/riscv/kernel/patch.c new file +10 -0
+struct riscv_insn_patch {
+     void *addr;
+     u32 insn;
+     atomic_t cpu_count;
+};
+
+#ifdef CONFIG_MMU
+static DEFINE_RAW_SPINLOCK(patch_lock);
+
+static void __kprobes *patch_map(void *addr, int fixmap)

Please use NOKPROBE_SYMBOL() instead of __kprobes. __kprobes is old style.

Quoted 17 lines
arch/riscv/kernel/patch.c new file +17 -0
+{
+     uintptr_t uintaddr = (uintptr_t) addr;
+     struct page *page;
+
+     if (core_kernel_text(uintaddr))
+             page = phys_to_page(__pa_symbol(addr));
+     else if (IS_ENABLED(CONFIG_STRICT_MODULE_RWX))
+             page = vmalloc_to_page(addr);
+     else
+             return addr;
+
+     BUG_ON(!page);
+
+     return (void *)set_fixmap_offset(fixmap, page_to_phys(page) +
+                                      (uintaddr & ~PAGE_MASK));
+}
+
Quoted 8 lines

arch/riscv/kernel/patch.c line 46

Quoted 6 lines
arch/riscv/kernel/patch.c new file +6 -0
+static void __kprobes patch_unmap(int fixmap)
+{
+     clear_fixmap(fixmap);
+}
+
+static int __kprobes riscv_insn_write(void *addr, const void *insn, size_t len)

Why would you add "riscv_" prefix for those functions? It seems a bit odd.

There is no particular reason, I just was used to adding a prefix for arch-related stuff. I have no preference here, it's OK to me to remove the prefix of these functions, do you think we need to remove them?

Quoted 11 lines

arch/riscv/kernel/patch.c line 53

Quoted 8 lines
arch/riscv/kernel/patch.c new file +7 -0
+{
+     void *waddr = addr;
+     bool across_pages = (((uintptr_t) addr & ~PAGE_MASK) + len) > PAGE_SIZE;
+     unsigned long flags = 0;
+     int ret;
+
+     raw_spin_lock_irqsave(&patch_lock, flags);

This looks a bit odd since stop_machine() is protected by its own mutex, and also the irq is already disabled here.

We need it because we don't always enter the riscv_patch_text_nosync() through stop_machine mechanism. If we call the riscv_patch_text_nosync() directly, we need a lock to protect the page.

Quoted 77 lines

Thank you,

Quoted 68 lines
arch/riscv/kernel/patch.c new file +67 -1
+
+     if (across_pages)
+             patch_map(addr + len, FIX_TEXT_POKE1);
+
+     waddr = patch_map(addr, FIX_TEXT_POKE0);
+
+     ret = probe_kernel_write(waddr, insn, len);
+
+     patch_unmap(FIX_TEXT_POKE0);
+
+     if (across_pages)
+             patch_unmap(FIX_TEXT_POKE1);
+
+     raw_spin_unlock_irqrestore(&patch_lock, flags);
+
+     return ret;
+}
+#else
+static int __kprobes riscv_insn_write(void *addr, const void *insn, size_t len)
+{
+     return probe_kernel_write(addr, insn, len);
+}
+#endif /* CONFIG_MMU */
+
+int __kprobes riscv_patch_text_nosync(void *addr, const void *insns, size_t len)
+{
+     u32 *tp = addr;
+     int ret;
+
+     ret = riscv_insn_write(tp, insns, len);
+
+     if (!ret)
+             flush_icache_range((uintptr_t) tp, (uintptr_t) tp + len);
+
+     return ret;
+}
+
+static int __kprobes riscv_patch_text_cb(void *data)
+{
+     struct riscv_insn_patch *patch = data;
+     int ret = 0;
+
+     if (atomic_inc_return(&patch->cpu_count) == 1) {
+             ret =
+                 riscv_patch_text_nosync(patch->addr, &patch->insn,
+                                         GET_INSN_LENGTH(patch->insn));
+             atomic_inc(&patch->cpu_count);
+     } else {
+             while (atomic_read(&patch->cpu_count) <= num_online_cpus())
+                     cpu_relax();
+             smp_mb();
+     }
+
+     return ret;
+}
+
+int __kprobes riscv_patch_text(void *addr, u32 insn)
+{
+     struct riscv_insn_patch patch = {
+             .addr = addr,
+             .insn = insn,
+             .cpu_count = ATOMIC_INIT(0),
+     };
+
+     return stop_machine_cpuslocked(riscv_patch_text_cb,
+                                    &patch, cpu_online_mask);
+}
--
Quoted 2 lines

2.25.1

--
Masami Hiramatsu <[email protected]>

maintainer
#11 lore raw

Re: [PATCH v3 8/9] riscv: introduce interfaces to patch kernel code

Hi,

Quoted 8 lines · On Wed, 1 Apr 2020 15:42:30 +0800 Zong Li <[email protected]> wrote:
Quoted 4 lines

Review

Quoted 2 lines
(quoted diff) +2 -0
+
+static int __kprobes riscv_insn_write(void *addr, const void *insn, size_t len)

Why would you add "riscv_" prefix for those functions? It seems a bit odd.

There is no particular reason, I just was used to adding a prefix for arch-related stuff. I have no preference here, it's OK to me to remove the prefix of these functions, do you think we need to remove them?

Yeah, it will be better, unless it can mixed up with arch-independent functions.

Quoted 11 lines
Quoted 10 lines

arch/riscv/kernel/patch.c line 53

Quoted 7 lines
arch/riscv/kernel/patch.c new file +7 -0
+{
+     void *waddr = addr;
+     bool across_pages = (((uintptr_t) addr & ~PAGE_MASK) + len) > PAGE_SIZE;
+     unsigned long flags = 0;
+     int ret;
+
+     raw_spin_lock_irqsave(&patch_lock, flags);

This looks a bit odd since stop_machine() is protected by its own mutex, and also the irq is already disabled here.

Quoted 4 lines

We need it because we don't always enter the riscv_patch_text_nosync() through stop_machine mechanism. If we call the riscv_patch_text_nosync() directly, we need a lock to protect the page.

Oh, OK, but it leads another question. Is that safe to patch the text without sync? Would you use it for UP system?
I think it is better to clarify "in what case user can call _nosync()" and add a comment on it.

Thank you,

Quoted 78 lines
Quoted 78 lines

Thank you,

Quoted 68 lines
arch/riscv/kernel/patch.c new file +67 -1
+
+     if (across_pages)
+             patch_map(addr + len, FIX_TEXT_POKE1);
+
+     waddr = patch_map(addr, FIX_TEXT_POKE0);
+
+     ret = probe_kernel_write(waddr, insn, len);
+
+     patch_unmap(FIX_TEXT_POKE0);
+
+     if (across_pages)
+             patch_unmap(FIX_TEXT_POKE1);
+
+     raw_spin_unlock_irqrestore(&patch_lock, flags);
+
+     return ret;
+}
+#else
+static int __kprobes riscv_insn_write(void *addr, const void *insn, size_t len)
+{
+     return probe_kernel_write(addr, insn, len);
+}
+#endif /* CONFIG_MMU */
+
+int __kprobes riscv_patch_text_nosync(void *addr, const void *insns, size_t len)
+{
+     u32 *tp = addr;
+     int ret;
+
+     ret = riscv_insn_write(tp, insns, len);
+
+     if (!ret)
+             flush_icache_range((uintptr_t) tp, (uintptr_t) tp + len);
+
+     return ret;
+}
+
+static int __kprobes riscv_patch_text_cb(void *data)
+{
+     struct riscv_insn_patch *patch = data;
+     int ret = 0;
+
+     if (atomic_inc_return(&patch->cpu_count) == 1) {
+             ret =
+                 riscv_patch_text_nosync(patch->addr, &patch->insn,
+                                         GET_INSN_LENGTH(patch->insn));
+             atomic_inc(&patch->cpu_count);
+     } else {
+             while (atomic_read(&patch->cpu_count) <= num_online_cpus())
+                     cpu_relax();
+             smp_mb();
+     }
+
+     return ret;
+}
+
+int __kprobes riscv_patch_text(void *addr, u32 insn)
+{
+     struct riscv_insn_patch patch = {
+             .addr = addr,
+             .insn = insn,
+             .cpu_count = ATOMIC_INIT(0),
+     };
+
+     return stop_machine_cpuslocked(riscv_patch_text_cb,
+                                    &patch, cpu_online_mask);
+}
--
Quoted 2 lines

2.25.1

--
Masami Hiramatsu <[email protected]>

Signature · 1 line
Masami Hiramatsu <[email protected]>
author
#12 lore raw

Re: [PATCH v3 8/9] riscv: introduce interfaces to patch kernel code

Quoted 11 lines · Masami Hiramatsu

Hi,

Quoted 5 lines · On Wed, 1 Apr 2020 15:42:30 +0800 Zong Li <[email protected]> wrote:
Quoted 4 lines

Review

Quoted 2 lines
(quoted diff) +2 -0
+
+static int __kprobes riscv_insn_write(void *addr, const void *insn, size_t len)

Why would you add "riscv_" prefix for those functions? It seems a bit odd.

Quoted 6 lines
Quoted 3 lines

There is no particular reason, I just was used to adding a prefix for arch-related stuff. I have no preference here, it's OK to me to remove the prefix of these functions, do you think we need to remove them?

Yeah, it will be better, unless it can mixed up with arch-independent functions.

OK. I'll remove it and use NOKPROBE_SYMBOL() instead of __kprobes annotation.

Quoted 17 lines
Quoted 12 lines
Quoted 11 lines

arch/riscv/kernel/patch.c line 53

Quoted 8 lines
arch/riscv/kernel/patch.c new file +7 -0
+{
+     void *waddr = addr;
+     bool across_pages = (((uintptr_t) addr & ~PAGE_MASK) + len) > PAGE_SIZE;
+     unsigned long flags = 0;
+     int ret;
+
+     raw_spin_lock_irqsave(&patch_lock, flags);

This looks a bit odd since stop_machine() is protected by its own mutex, and also the irq is already disabled here.

Quoted 4 lines

We need it because we don't always enter the riscv_patch_text_nosync() through stop_machine mechanism. If we call the riscv_patch_text_nosync() directly, we need a lock to protect the page.

Quoted 4 lines

Oh, OK, but it leads another question. Is that safe to patch the text without sync? Would you use it for UP system?
I think it is better to clarify "in what case user can call _nosync()" and add a comment on it.

The ftrace is one of the cases, as documentation of ftrace said, when dynamic ftrace is initialized, it calls kstop_machine to make the machine act like a uniprocessor so that it can freely modify code without worrying about other processors executing that same code. So the ftrace called the _nosync interface here directly.

Quoted 85 lines

Thank you,

Quoted 70 lines
Quoted 70 lines

Thank you,

Quoted 66 lines
arch/riscv/kernel/patch.c new file +66 -0
+
+     if (across_pages)
+             patch_map(addr + len, FIX_TEXT_POKE1);
+
+     waddr = patch_map(addr, FIX_TEXT_POKE0);
+
+     ret = probe_kernel_write(waddr, insn, len);
+
+     patch_unmap(FIX_TEXT_POKE0);
+
+     if (across_pages)
+             patch_unmap(FIX_TEXT_POKE1);
+
+     raw_spin_unlock_irqrestore(&patch_lock, flags);
+
+     return ret;
+}
+#else
+static int __kprobes riscv_insn_write(void *addr, const void *insn, size_t len)
+{
+     return probe_kernel_write(addr, insn, len);
+}
+#endif /* CONFIG_MMU */
+
+int __kprobes riscv_patch_text_nosync(void *addr, const void *insns, size_t len)
+{
+     u32 *tp = addr;
+     int ret;
+
+     ret = riscv_insn_write(tp, insns, len);
+
+     if (!ret)
+             flush_icache_range((uintptr_t) tp, (uintptr_t) tp + len);
+
+     return ret;
+}
+
+static int __kprobes riscv_patch_text_cb(void *data)
+{
+     struct riscv_insn_patch *patch = data;
+     int ret = 0;
+
+     if (atomic_inc_return(&patch->cpu_count) == 1) {
+             ret =
+                 riscv_patch_text_nosync(patch->addr, &patch->insn,
+                                         GET_INSN_LENGTH(patch->insn));
+             atomic_inc(&patch->cpu_count);
+     } else {
+             while (atomic_read(&patch->cpu_count) <= num_online_cpus())
+                     cpu_relax();
+             smp_mb();
+     }
+
+     return ret;
+}
+
+int __kprobes riscv_patch_text(void *addr, u32 insn)
+{
+     struct riscv_insn_patch patch = {
+             .addr = addr,
+             .insn = insn,
+             .cpu_count = ATOMIC_INIT(0),
+     };
+
+     return stop_machine_cpuslocked(riscv_patch_text_cb,
+                                    &patch, cpu_online_mask);
Quoted 8 lines
Quoted 8 lines
Quoted 4 lines

+}
--
2.25.1

--
Masami Hiramatsu <[email protected]>

--
Masami Hiramatsu <[email protected]>

maintainer
#13 lore raw

Re: [PATCH v3 8/9] riscv: introduce interfaces to patch kernel code

Hi Zong,

Quoted 21 lines · On Fri, 3 Apr 2020 17:04:51 +0800 Zong Li <[email protected]> wrote:
Quoted 16 lines
Quoted 11 lines
Quoted 10 lines

arch/riscv/kernel/patch.c line 53

Quoted 7 lines
arch/riscv/kernel/patch.c new file +7 -0
+{
+     void *waddr = addr;
+     bool across_pages = (((uintptr_t) addr & ~PAGE_MASK) + len) > PAGE_SIZE;
+     unsigned long flags = 0;
+     int ret;
+
+     raw_spin_lock_irqsave(&patch_lock, flags);

This looks a bit odd since stop_machine() is protected by its own mutex, and also the irq is already disabled here.

Quoted 4 lines

We need it because we don't always enter the riscv_patch_text_nosync() through stop_machine mechanism. If we call the riscv_patch_text_nosync() directly, we need a lock to protect the page.

Quoted 4 lines

Oh, OK, but it leads another question. Is that safe to patch the text without sync? Would you use it for UP system?
I think it is better to clarify "in what case user can call _nosync()" and add a comment on it.

Quoted 5 lines

The ftrace is one of the cases, as documentation of ftrace said, when dynamic ftrace is initialized, it calls kstop_machine to make the machine act like a uniprocessor so that it can freely modify code without worrying about other processors executing that same code. So the ftrace called the _nosync interface here directly.

Hmm, even though, since it already running under kstop_machine(), no other thread will run.
Could you consider to use text_mutex instead of that? The text_mutex is already widely used in x86 and kernel/kprobes.c etc.

(Hmm, it seems except for x86, alternative code don't care about
racing...)

Thank you,

Signature · 1 line
Masami Hiramatsu <[email protected]>
author
#14 lore raw

Re: [PATCH v3 8/9] riscv: introduce interfaces to patch kernel code

Quoted 33 lines · Masami Hiramatsu

Hi Zong,

Quoted 21 lines · On Fri, 3 Apr 2020 17:04:51 +0800 Zong Li <[email protected]> wrote:
Quoted 16 lines
Quoted 11 lines
Quoted 10 lines

arch/riscv/kernel/patch.c line 53

Quoted 7 lines
arch/riscv/kernel/patch.c new file +7 -0
+{
+     void *waddr = addr;
+     bool across_pages = (((uintptr_t) addr & ~PAGE_MASK) + len) > PAGE_SIZE;
+     unsigned long flags = 0;
+     int ret;
+
+     raw_spin_lock_irqsave(&patch_lock, flags);

This looks a bit odd since stop_machine() is protected by its own mutex, and also the irq is already disabled here.

Quoted 4 lines

We need it because we don't always enter the riscv_patch_text_nosync() through stop_machine mechanism. If we call the riscv_patch_text_nosync() directly, we need a lock to protect the page.

Quoted 4 lines

Oh, OK, but it leads another question. Is that safe to patch the text without sync? Would you use it for UP system?
I think it is better to clarify "in what case user can call _nosync()" and add a comment on it.

Quoted 5 lines

The ftrace is one of the cases, as documentation of ftrace said, when dynamic ftrace is initialized, it calls kstop_machine to make the machine act like a uniprocessor so that it can freely modify code without worrying about other processors executing that same code. So the ftrace called the _nosync interface here directly.

Quoted 8 lines

Hmm, even though, since it already running under kstop_machine(), no other thread will run.
Could you consider to use text_mutex instead of that? The text_mutex is already widely used in x86 and kernel/kprobes.c etc.

(Hmm, it seems except for x86, alternative code don't care about
racing...)

Yes, text_mutex seems to be great. I'll change to use text_mutex in the next version if it works fine after testing. Thanks.

Quoted 3 lines

Thank you,
--
Masami Hiramatsu <[email protected]>

author
#15 lore raw

Re: [PATCH v3 8/9] riscv: introduce interfaces to patch kernel code

Quoted 35 lines · Zong Li
Quoted 33 lines · Masami Hiramatsu

Hi Zong,

Quoted 26 lines · On Fri, 3 Apr 2020 17:04:51 +0800 Zong Li <[email protected]> wrote:
Quoted 16 lines
Quoted 11 lines
Quoted 10 lines

Review

Quoted 7 lines
(quoted diff) +7 -0
+{
+     void *waddr = addr;
+     bool across_pages = (((uintptr_t) addr & ~PAGE_MASK) + len) > PAGE_SIZE;
+     unsigned long flags = 0;
+     int ret;
+
+     raw_spin_lock_irqsave(&patch_lock, flags);

This looks a bit odd since stop_machine() is protected by its own mutex, and also the irq is already disabled here.

Quoted 4 lines

We need it because we don't always enter the riscv_patch_text_nosync() through stop_machine mechanism. If we call the riscv_patch_text_nosync() directly, we need a lock to protect the page.

Quoted 4 lines

Oh, OK, but it leads another question. Is that safe to patch the text without sync? Would you use it for UP system?
I think it is better to clarify "in what case user can call _nosync()" and add a comment on it.

The ftrace is one of the cases, as documentation of ftrace said, when dynamic ftrace is initialized, it calls kstop_machine to make the machine act like a uniprocessor so that it can freely modify code without worrying about other processors executing that same code. So the ftrace called the _nosync interface here directly.

Quoted 8 lines
Quoted 8 lines

Hmm, even though, since it already running under kstop_machine(), no other thread will run.
Could you consider to use text_mutex instead of that? The text_mutex is already widely used in x86 and kernel/kprobes.c etc.

(Hmm, it seems except for x86, alternative code don't care about
racing...)

The mutex_lock doesn't seem to work in ftrace context, I think it might be the reason why other architectures didn't use text_mutex in somewhere.

# echo function > current_tracer
[   28.198070] BUG: sleeping function called from invalid context at
kernel/locking/mutex.c:281
[   28.198663] in_atomic(): 1, irqs_disabled(): 1, non_block: 0, pid:
11, name: migration/0
[   28.199491] CPU: 0 PID: 11 Comm: migration/0 Not tainted
5.6.0-00012-gd6f56a7a4be2-dirty #10
[   28.200330] Call Trace:
[   28.200798] [<ffffffe00060319a>] walk_stackframe+0x0/0xcc
[   28.201395] [<ffffffe000603442>] show_stack+0x3c/0x46
[   28.200798] [<ffffffe00060319a>] walk_stackframe+0x0/0xcc
[   28.201395] [<ffffffe000603442>] show_stack+0x3c/0x46
[   28.201898] [<ffffffe000d498b0>] dump_stack+0x76/0x90
[   28.202329] [<ffffffe00062c3f0>] ___might_sleep+0x100/0x10e
[   28.202720] [<ffffffe00062c448>] __might_sleep+0x4a/0x78
[   28.203033] [<ffffffe000d61622>] mutex_lock+0x2c/0x54
[   28.203397] [<ffffffe00060393e>] patch_insn_write+0x32/0xd8
[   28.203780] [<ffffffe000603a94>] patch_text_nosync+0x10/0x32
[   28.204139] [<ffffffe0006051b0>] __ftrace_modify_call+0x5c/0x6c
[   28.204497] [<ffffffe0006052c6>] ftrace_update_ftrace_func+0x20/0x4a
[   28.204919] [<ffffffe000697742>] ftrace_modify_all_code+0xa0/0x148
[   28.205378] [<ffffffe0006977fc>] __ftrace_modify_code+0x12/0x1c
[   28.205793] [<ffffffe0006924b6>] multi_cpu_stop+0xa2/0x158
[   28.206147] [<ffffffe0006921b0>] cpu_stopper_thread+0xa4/0x13a
[   28.206510] [<ffffffe000629f38>] smpboot_thread_fn+0xf8/0x1da
[   28.206868] [<ffffffe000625f36>] kthread+0xfa/0x12a
[   28.207201] [<ffffffe0006017e2>] ret_from_exception+0x0/0xc
Quoted 7 lines

Yes, text_mutex seems to be great. I'll change to use text_mutex in the next version if it works fine after testing. Thanks.

Quoted 3 lines

Thank you,
--
Masami Hiramatsu <[email protected]>

maintainer
#16 lore raw

Re: [PATCH v3 8/9] riscv: introduce interfaces to patch kernel code

Quoted 41 lines · On Mon, 6 Apr 2020 18:36:42 +0800 Zong Li <[email protected]> wrote:
Quoted 40 lines · Zong Li
Quoted 38 lines · Masami Hiramatsu

Hi Zong,

Quoted 21 lines · On Fri, 3 Apr 2020 17:04:51 +0800 Zong Li <[email protected]> wrote:
Quoted 16 lines
Quoted 11 lines
Quoted 10 lines

Review

Quoted 7 lines
(quoted diff) +7 -0
+{
+     void *waddr = addr;
+     bool across_pages = (((uintptr_t) addr & ~PAGE_MASK) + len) > PAGE_SIZE;
+     unsigned long flags = 0;
+     int ret;
+
+     raw_spin_lock_irqsave(&patch_lock, flags);

This looks a bit odd since stop_machine() is protected by its own mutex, and also the irq is already disabled here.

Quoted 4 lines

We need it because we don't always enter the riscv_patch_text_nosync() through stop_machine mechanism. If we call the riscv_patch_text_nosync() directly, we need a lock to protect the page.

Quoted 4 lines

Oh, OK, but it leads another question. Is that safe to patch the text without sync? Would you use it for UP system?
I think it is better to clarify "in what case user can call _nosync()" and add a comment on it.

Quoted 5 lines

The ftrace is one of the cases, as documentation of ftrace said, when dynamic ftrace is initialized, it calls kstop_machine to make the machine act like a uniprocessor so that it can freely modify code without worrying about other processors executing that same code. So the ftrace called the _nosync interface here directly.

Hmm, even though, since it already running under kstop_machine(), no other thread will run.
Could you consider to use text_mutex instead of that? The text_mutex is already widely used in x86 and kernel/kprobes.c etc.

Quoted 7 lines
Quoted 3 lines
Quoted 3 lines

(Hmm, it seems except for x86, alternative code don't care about
racing...)

The mutex_lock doesn't seem to work in ftrace context, I think it might be the reason why other architectures didn't use text_mutex in somewhere.

Yes, you need to implement ftrace_arch_code_modify_prepare() and ftrace_arch_code_modify_post_process() in arch/riscv/kernel/ftrace.c. Please see arch/x86/kernel/ftrace.c.

Thank you,

Quoted 36 lines
# echo function > current_tracer
[   28.198070] BUG: sleeping function called from invalid context at
kernel/locking/mutex.c:281
[   28.198663] in_atomic(): 1, irqs_disabled(): 1, non_block: 0, pid:
11, name: migration/0
[   28.199491] CPU: 0 PID: 11 Comm: migration/0 Not tainted
5.6.0-00012-gd6f56a7a4be2-dirty #10
[   28.200330] Call Trace:
[   28.200798] [<ffffffe00060319a>] walk_stackframe+0x0/0xcc
[   28.201395] [<ffffffe000603442>] show_stack+0x3c/0x46
[   28.200798] [<ffffffe00060319a>] walk_stackframe+0x0/0xcc
[   28.201395] [<ffffffe000603442>] show_stack+0x3c/0x46
[   28.201898] [<ffffffe000d498b0>] dump_stack+0x76/0x90
[   28.202329] [<ffffffe00062c3f0>] ___might_sleep+0x100/0x10e
[   28.202720] [<ffffffe00062c448>] __might_sleep+0x4a/0x78
[   28.203033] [<ffffffe000d61622>] mutex_lock+0x2c/0x54
[   28.203397] [<ffffffe00060393e>] patch_insn_write+0x32/0xd8
[   28.203780] [<ffffffe000603a94>] patch_text_nosync+0x10/0x32
[   28.204139] [<ffffffe0006051b0>] __ftrace_modify_call+0x5c/0x6c
[   28.204497] [<ffffffe0006052c6>] ftrace_update_ftrace_func+0x20/0x4a
[   28.204919] [<ffffffe000697742>] ftrace_modify_all_code+0xa0/0x148
[   28.205378] [<ffffffe0006977fc>] __ftrace_modify_code+0x12/0x1c
[   28.205793] [<ffffffe0006924b6>] multi_cpu_stop+0xa2/0x158
[   28.206147] [<ffffffe0006921b0>] cpu_stopper_thread+0xa4/0x13a
[   28.206510] [<ffffffe000629f38>] smpboot_thread_fn+0xf8/0x1da
[   28.206868] [<ffffffe000625f36>] kthread+0xfa/0x12a
[   28.207201] [<ffffffe0006017e2>] ret_from_exception+0x0/0xc
Quoted 7 lines

Yes, text_mutex seems to be great. I'll change to use text_mutex in the next version if it works fine after testing. Thanks.

Quoted 3 lines

Thank you,
--
Masami Hiramatsu <[email protected]>

Signature · 1 line
Masami Hiramatsu <[email protected]>
author
#17 lore raw

Re: [PATCH v3 8/9] riscv: introduce interfaces to patch kernel code

Quoted 49 lines · Masami Hiramatsu
Quoted 45 lines · On Mon, 6 Apr 2020 18:36:42 +0800 Zong Li <[email protected]> wrote:
Quoted 43 lines · Zong Li
Quoted 41 lines · Masami Hiramatsu

Hi Zong,

Quoted 21 lines · On Fri, 3 Apr 2020 17:04:51 +0800 Zong Li <[email protected]> wrote:
Quoted 16 lines
Quoted 11 lines
Quoted 10 lines

Review

Quoted 7 lines
(quoted diff) +7 -0
+{
+     void *waddr = addr;
+     bool across_pages = (((uintptr_t) addr & ~PAGE_MASK) + len) > PAGE_SIZE;
+     unsigned long flags = 0;
+     int ret;
+
+     raw_spin_lock_irqsave(&patch_lock, flags);

This looks a bit odd since stop_machine() is protected by its own mutex, and also the irq is already disabled here.

Quoted 4 lines

We need it because we don't always enter the riscv_patch_text_nosync() through stop_machine mechanism. If we call the riscv_patch_text_nosync() directly, we need a lock to protect the page.

Quoted 4 lines

Oh, OK, but it leads another question. Is that safe to patch the text without sync? Would you use it for UP system?
I think it is better to clarify "in what case user can call _nosync()" and add a comment on it.

Quoted 5 lines

The ftrace is one of the cases, as documentation of ftrace said, when dynamic ftrace is initialized, it calls kstop_machine to make the machine act like a uniprocessor so that it can freely modify code without worrying about other processors executing that same code. So the ftrace called the _nosync interface here directly.

Hmm, even though, since it already running under kstop_machine(), no other thread will run.
Could you consider to use text_mutex instead of that? The text_mutex is already widely used in x86 and kernel/kprobes.c etc.

(Hmm, it seems except for x86, alternative code don't care about
racing...)

Quoted 8 lines
Quoted 3 lines

The mutex_lock doesn't seem to work in ftrace context, I think it might be the reason why other architectures didn't use text_mutex in somewhere.

Yes, you need to implement ftrace_arch_code_modify_prepare() and ftrace_arch_code_modify_post_process() in arch/riscv/kernel/ftrace.c. Please see arch/x86/kernel/ftrace.c.

Oh ok, I misunderstood it before, I just use text_mutex instead of patch_lock in patch.c. Thanks.

Quoted 42 lines

Thank you,

Quoted 20 lines
# echo function > current_tracer
[   28.198070] BUG: sleeping function called from invalid context at
kernel/locking/mutex.c:281
[   28.198663] in_atomic(): 1, irqs_disabled(): 1, non_block: 0, pid:
11, name: migration/0
[   28.199491] CPU: 0 PID: 11 Comm: migration/0 Not tainted
5.6.0-00012-gd6f56a7a4be2-dirty #10
[   28.200330] Call Trace:
[   28.200798] [<ffffffe00060319a>] walk_stackframe+0x0/0xcc
[   28.201395] [<ffffffe000603442>] show_stack+0x3c/0x46
[   28.200798] [<ffffffe00060319a>] walk_stackframe+0x0/0xcc
[   28.201395] [<ffffffe000603442>] show_stack+0x3c/0x46
[   28.201898] [<ffffffe000d498b0>] dump_stack+0x76/0x90
[   28.202329] [<ffffffe00062c3f0>] ___might_sleep+0x100/0x10e
[   28.202720] [<ffffffe00062c448>] __might_sleep+0x4a/0x78
[   28.203033] [<ffffffe000d61622>] mutex_lock+0x2c/0x54
[   28.203397] [<ffffffe00060393e>] patch_insn_write+0x32/0xd8
[   28.203780] [<ffffffe000603a94>] patch_text_nosync+0x10/0x32
[   28.204139] [<ffffffe0006051b0>] __ftrace_modify_call+0x5c/0x6c
Quoted 16 lines
[   28.204497] [<ffffffe0006052c6>] ftrace_update_ftrace_func+0x20/0x4a
[   28.204919] [<ffffffe000697742>] ftrace_modify_all_code+0xa0/0x148
[   28.205378] [<ffffffe0006977fc>] __ftrace_modify_code+0x12/0x1c
[   28.205793] [<ffffffe0006924b6>] multi_cpu_stop+0xa2/0x158
[   28.206147] [<ffffffe0006921b0>] cpu_stopper_thread+0xa4/0x13a
[   28.206510] [<ffffffe000629f38>] smpboot_thread_fn+0xf8/0x1da
[   28.206868] [<ffffffe000625f36>] kthread+0xfa/0x12a
[   28.207201] [<ffffffe0006017e2>] ret_from_exception+0x0/0xc
Quoted 7 lines

Yes, text_mutex seems to be great. I'll change to use text_mutex in the next version if it works fine after testing. Thanks.

Quoted 3 lines

Thank you,
--
Masami Hiramatsu <[email protected]>

--
Masami Hiramatsu <[email protected]>

author
#18 lore raw

PATCH v3 9/9 riscv: patch code by fixmap mapping

In reply to Zong Li

On strict kernel memory permission, the ftrace have to change the permission of text for dynamic patching the intructions. Use riscv_patch_text_nosync() to patch code instead of probe_kernel_write.

---
 arch/riscv/kernel/ftrace.c | 13 ++++---------
 1 file changed, 4 insertions(+), 9 deletions(-)
arch/riscv/kernel/ftrace.c +4 -9
index c40fdcdeb950..ce69b34ff55d 100644
@@ -8,6 +8,7 @@
 #include <linux/ftrace.h>
 #include <linux/uaccess.h>
 #include <asm/cacheflush.h>
+#include <asm/patch.h>
 
 #ifdef CONFIG_DYNAMIC_FTRACE
 static int ftrace_check_current_call(unsigned long hook_pos,
@@ -46,20 +47,14 @@ static int __ftrace_modify_call(unsigned long hook_pos, unsigned long target,
 {
 	unsigned int call[2];
 	unsigned int nops[2] = {NOP4, NOP4};
-	int ret = 0;
 
 	make_call(hook_pos, target, call);
 
-	/* replace the auipc-jalr pair at once */
-	ret = probe_kernel_write((void *)hook_pos, enable ? call : nops,
-				 MCOUNT_INSN_SIZE);
-	/* return must be -EPERM on write error */
-	if (ret)
+	/* Replace the auipc-jalr pair at once. Return -EPERM on write error. */
+	if (riscv_patch_text_nosync
+	    ((void *)hook_pos, enable ? call : nops, MCOUNT_INSN_SIZE))
 		return -EPERM;
 
-	smp_mb();
-	flush_icache_range((void *)hook_pos, (void *)hook_pos + MCOUNT_INSN_SIZE);
-
 	return 0;
 }
 
Signature · 1 line
2.25.1
maintainer
#19 lore raw
In reply to Zong Li

Hi Zong,

Quoted 7 lines · On Tue, 10 Mar 2020 00:55:35 +0800 Zong Li <[email protected]> wrote:

The main purpose of this patch series is changing the kernel mapping permission , make sure that code is not writeable, data is not executable, and read-only data is neither writable nor executable.

This patch series also supports the relevant implementations such as ARCH_HAS_SET_MEMORY, ARCH_HAS_SET_DIRECT_MAP, ARCH_SUPPORTS_DEBUG_PAGEALLOC and DEBUG_WX.

The order of the patches seems a bit strange. Since the first 7 patches makes kernel read-only, at that point ftrace is broken and it is fixed by the last 2 patches. That is not bisect-friendly. Can you move the last 2 patches to the top?

Thank you,

Quoted 44 lines

Changes in v3:
- Fix build error on nommu configuration. We already support nommu on RISC-V, so we should consider nommu case and test not only rv32/64, but also nommu.

Changes in v2:
- Use _data to specify the start of data section with write permission.
- Change ftrace patch text implementaion.
- Separate DEBUG_WX patch to another patchset.

Zong Li (9):
riscv: add ARCH_HAS_SET_MEMORY support
riscv: add ARCH_HAS_SET_DIRECT_MAP support
riscv: add ARCH_SUPPORTS_DEBUG_PAGEALLOC support
riscv: move exception table immediately after RO_DATA
riscv: add alignment for text, rodata and data sections
riscv: add STRICT_KERNEL_RWX support
riscv: add macro to get instruction length
riscv: introduce interfaces to patch kernel code
riscv: patch code by fixmap mapping

 arch/riscv/Kconfig                  |   6 +
 arch/riscv/include/asm/bug.h        |   8 ++
 arch/riscv/include/asm/fixmap.h     |   2 +
 arch/riscv/include/asm/patch.h      |  12 ++
 arch/riscv/include/asm/set_memory.h |  48 +++++++
 arch/riscv/kernel/Makefile          |   4 +-
 arch/riscv/kernel/ftrace.c          |  13 +-
 arch/riscv/kernel/patch.c           | 120 ++++++++++++++++++
 arch/riscv/kernel/traps.c           |   3 +-
 arch/riscv/kernel/vmlinux.lds.S     |  11 +-
 arch/riscv/mm/Makefile              |   2 +-
 arch/riscv/mm/init.c                |  44 +++++++
 arch/riscv/mm/pageattr.c            | 187 ++++++++++++++++++++++++++++
 13 files changed, 445 insertions(+), 15 deletions(-)
 create mode 100644 arch/riscv/include/asm/patch.h
 create mode 100644 arch/riscv/include/asm/set_memory.h
 create mode 100644 arch/riscv/kernel/patch.c
 create mode 100644 arch/riscv/mm/pageattr.c
Signature · 1 line
2.25.1
Signature · 1 line
Masami Hiramatsu <[email protected]>
author
#20 lore raw
Quoted 10 lines · Masami Hiramatsu

Hi Zong,

Quoted 4 lines · On Tue, 10 Mar 2020 00:55:35 +0800 Zong Li <[email protected]> wrote:

The main purpose of this patch series is changing the kernel mapping permission , make sure that code is not writeable, data is not executable, and read-only data is neither writable nor executable.

Quoted 9 lines
Quoted 3 lines

This patch series also supports the relevant implementations such as ARCH_HAS_SET_MEMORY, ARCH_HAS_SET_DIRECT_MAP, ARCH_SUPPORTS_DEBUG_PAGEALLOC and DEBUG_WX.

The order of the patches seems a bit strange. Since the first 7 patches makes kernel read-only, at that point ftrace is broken and it is fixed by the last 2 patches. That is not bisect-friendly. Can you move the last 2 patches to the top?

OK. Let me rearrange the order. Thanks.

Quoted 50 lines

Thank you,

Quoted 41 lines

Changes in v3:
- Fix build error on nommu configuration. We already support nommu on RISC-V, so we should consider nommu case and test not only rv32/64, but also nommu.

Changes in v2:
- Use _data to specify the start of data section with write permission.
- Change ftrace patch text implementaion.
- Separate DEBUG_WX patch to another patchset.

Zong Li (9):
riscv: add ARCH_HAS_SET_MEMORY support
riscv: add ARCH_HAS_SET_DIRECT_MAP support
riscv: add ARCH_SUPPORTS_DEBUG_PAGEALLOC support
riscv: move exception table immediately after RO_DATA
riscv: add alignment for text, rodata and data sections
riscv: add STRICT_KERNEL_RWX support
riscv: add macro to get instruction length
riscv: introduce interfaces to patch kernel code
riscv: patch code by fixmap mapping

 arch/riscv/Kconfig                  |   6 +
 arch/riscv/include/asm/bug.h        |   8 ++
 arch/riscv/include/asm/fixmap.h     |   2 +
 arch/riscv/include/asm/patch.h      |  12 ++
 arch/riscv/include/asm/set_memory.h |  48 +++++++
 arch/riscv/kernel/Makefile          |   4 +-
 arch/riscv/kernel/ftrace.c          |  13 +-
 arch/riscv/kernel/patch.c           | 120 ++++++++++++++++++
 arch/riscv/kernel/traps.c           |   3 +-
 arch/riscv/kernel/vmlinux.lds.S     |  11 +-
 arch/riscv/mm/Makefile              |   2 +-
 arch/riscv/mm/init.c                |  44 +++++++
 arch/riscv/mm/pageattr.c            | 187 ++++++++++++++++++++++++++++
 13 files changed, 445 insertions(+), 15 deletions(-)
 create mode 100644 arch/riscv/include/asm/patch.h
 create mode 100644 arch/riscv/include/asm/set_memory.h
 create mode 100644 arch/riscv/kernel/patch.c
 create mode 100644 arch/riscv/mm/pageattr.c
Quoted 3 lines

--
2.25.1

--
Masami Hiramatsu <[email protected]>