Re: [PATCH v5 2/9] riscv: introduce interfaces to patch kernel code
From: Palmer Dabbelt
Date: Fri Apr 17 2020 - 20:23:13 EST
On Wed, 08 Apr 2020 00:56:57 PDT (-0700), zong.li@xxxxxxxxxx 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()
Signed-off-by: Zong Li <zong.li@xxxxxxxxxx>
Suggested-by: Masami Hiramatsu <mhiramat@xxxxxxxxxx>
---
arch/riscv/include/asm/fixmap.h | 2 +
arch/riscv/include/asm/patch.h | 12 +++
arch/riscv/kernel/Makefile | 4 +-
arch/riscv/kernel/patch.c | 128 ++++++++++++++++++++++++++++++++
4 files changed, 145 insertions(+), 1 deletion(-)
create mode 100644 arch/riscv/include/asm/patch.h
create mode 100644 arch/riscv/kernel/patch.c
diff --git a/arch/riscv/include/asm/fixmap.h b/arch/riscv/include/asm/fixmap.h
index 42d2c42f3cc9..2368d49eb4ef 100644
--- a/arch/riscv/include/asm/fixmap.h
+++ b/arch/riscv/include/asm/fixmap.h
@@ -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
};
diff --git a/arch/riscv/include/asm/patch.h b/arch/riscv/include/asm/patch.h
new file mode 100644
index 000000000000..9a7d7346001e
--- /dev/null
+++ b/arch/riscv/include/asm/patch.h
@@ -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 patch_text_nosync(void *addr, const void *insns, size_t len);
+int patch_text(void *addr, u32 insn);
+
+#endif /* _ASM_RISCV_PATCH_H */
diff --git a/arch/riscv/kernel/Makefile b/arch/riscv/kernel/Makefile
index f40205cb9a22..d189bd3d8501 100644
--- a/arch/riscv/kernel/Makefile
+++ b/arch/riscv/kernel/Makefile
@@ -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
diff --git a/arch/riscv/kernel/patch.c b/arch/riscv/kernel/patch.c
new file mode 100644
index 000000000000..f16466123947
--- /dev/null
+++ b/arch/riscv/kernel/patch.c
@@ -0,0 +1,128 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (C) 2020 SiFive
+ */
+
+#include <linux/mm.h>
+#include <linux/memory.h>
+#include <linux/uaccess.h>
+#include <linux/spinlock.h>
+#include <linux/stop_machine.h>
+#include <asm/fixmap.h>
+#include <asm/kprobes.h>
+#include <asm/cacheflush.h>
+
+struct patch_insn_patch {
+ void *addr;
+ u32 insn;
+ atomic_t cpu_count;
+};
+
+#ifdef CONFIG_MMU
+static void *patch_map(void *addr, int fixmap)
+{
+ 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));
+}
+NOKPROBE_SYMBOL(patch_map);
+
+static void patch_unmap(int fixmap)
+{
+ clear_fixmap(fixmap);
+}
+NOKPROBE_SYMBOL(patch_unmap);
+
+static int patch_insn_write(void *addr, const void *insn, size_t len)
+{
+ void *waddr = addr;
+ bool across_pages = (((uintptr_t) addr & ~PAGE_MASK) + len) > PAGE_SIZE;
+ int ret;
+
+ /*
+ * Before reaching here, it was expected to lock the text_mutex
+ * already, so we don't need to give another lock here and could
+ * ensure that it was safe between each cores.
+ */
+ lockdep_assert_held(&text_mutex);
+
+ 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);
+
+ return ret;
+}
+NOKPROBE_SYMBOL(patch_insn_write);
+#else
+static int patch_insn_write(void *addr, const void *insn, size_t len)
+{
+ return probe_kernel_write(addr, insn, len);
+}
+NOKPROBE_SYMBOL(patch_insn_write);
+#endif /* CONFIG_MMU */
+
+int patch_text_nosync(void *addr, const void *insns, size_t len)
+{
+ u32 *tp = addr;
+ int ret;
+
+ ret = patch_insn_write(tp, insns, len);
+
+ if (!ret)
+ flush_icache_range((uintptr_t) tp, (uintptr_t) tp + len);
+
+ return ret;
+}
+NOKPROBE_SYMBOL(patch_text_nosync);
+
+static int patch_text_cb(void *data)
+{
+ struct patch_insn_patch *patch = data;
+ int ret = 0;
+
+ if (atomic_inc_return(&patch->cpu_count) == 1) {
+ ret =
+ 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;
+}
+NOKPROBE_SYMBOL(patch_text_cb);
+
+int patch_text(void *addr, u32 insn)
+{
+ struct patch_insn_patch patch = {
+ .addr = addr,
+ .insn = insn,
+ .cpu_count = ATOMIC_INIT(0),
+ };
+
+ return stop_machine_cpuslocked(patch_text_cb,
+ &patch, cpu_online_mask);
+}
+NOKPROBE_SYMBOL(patch_text);
Reviewed-by: Palmer Dabbelt <palmerdabbelt@xxxxxxxxxx>