From: Masami Hiramatsu (Google) <mhiramat@kernel.org>
This introduces ftrace_get_symaddr() which tries to convert fentry_ip
passed by ftrace or fgraph callback to symaddr without calling
kallsyms API. It returns the symbol address or 0 if it fails to
convert it.
Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202412061423.K79V55Hd-lkp@intel.com/
Closes: https://lore.kernel.org/oe-kbuild-all/202412061804.5VRzF14E-lkp@intel.com/
---
Changes in v21:
- On arm64, fix the macro name to ftrace_get_symaddr() correctly.
- Define ftrace_get_symaddr() outside of CONFIG_DYNAMIC_FTRACE.
Changes in v19:
- Newly added.
---
arch/arm64/include/asm/ftrace.h | 2 +
arch/arm64/kernel/ftrace.c | 63 +++++++++++++++++++++++++++++++++++++++
arch/x86/include/asm/ftrace.h | 21 +++++++++++++
include/linux/ftrace.h | 13 ++++++++
4 files changed, 99 insertions(+)
diff --git a/arch/arm64/include/asm/ftrace.h b/arch/arm64/include/asm/ftrace.h
index 876e88ad4119..bfe3ce9df197 100644
--- a/arch/arm64/include/asm/ftrace.h
+++ b/arch/arm64/include/asm/ftrace.h
@@ -52,6 +52,8 @@ extern unsigned long ftrace_graph_call;
extern void return_to_handler(void);
unsigned long ftrace_call_adjust(unsigned long addr);
+unsigned long arch_ftrace_get_symaddr(unsigned long fentry_ip);
+#define ftrace_get_symaddr(fentry_ip) arch_ftrace_get_symaddr(fentry_ip)
#ifdef CONFIG_DYNAMIC_FTRACE_WITH_ARGS
#define HAVE_ARCH_FTRACE_REGS
diff --git a/arch/arm64/kernel/ftrace.c b/arch/arm64/kernel/ftrace.c
index 570c38be833c..d7c0d023dfe5 100644
--- a/arch/arm64/kernel/ftrace.c
+++ b/arch/arm64/kernel/ftrace.c
@@ -143,6 +143,69 @@ unsigned long ftrace_call_adjust(unsigned long addr)
return addr;
}
+/* Convert fentry_ip to the symbol address without kallsyms */
+unsigned long arch_ftrace_get_symaddr(unsigned long fentry_ip)
+{
+ u32 insn;
+
+ /*
+ * When using patchable-function-entry without pre-function NOPS, ftrace
+ * entry is the address of the first NOP after the function entry point.
+ *
+ * The compiler has either generated:
+ *
+ * func+00: func: NOP // To be patched to MOV X9, LR
+ * func+04: NOP // To be patched to BL <caller>
+ *
+ * Or:
+ *
+ * func-04: BTI C
+ * func+00: func: NOP // To be patched to MOV X9, LR
+ * func+04: NOP // To be patched to BL <caller>
+ *
+ * The fentry_ip is the address of `BL <caller>` which is at `func + 4`
+ * bytes in either case.
+ */
+ if (!IS_ENABLED(CONFIG_DYNAMIC_FTRACE_WITH_CALL_OPS))
+ return fentry_ip - AARCH64_INSN_SIZE;
+
+ /*
+ * When using patchable-function-entry with pre-function NOPs, BTI is
+ * a bit different.
+ *
+ * func+00: func: NOP // To be patched to MOV X9, LR
+ * func+04: NOP // To be patched to BL <caller>
+ *
+ * Or:
+ *
+ * func+00: func: BTI C
+ * func+04: NOP // To be patched to MOV X9, LR
+ * func+08: NOP // To be patched to BL <caller>
+ *
+ * The fentry_ip is the address of `BL <caller>` which is at either
+ * `func + 4` or `func + 8` depends on whether there is a BTI.
+ */
+
+ /* If there is no BTI, the func address should be one instruction before. */
+ if (!IS_ENABLED(CONFIG_ARM64_BTI_KERNEL))
+ return fentry_ip - AARCH64_INSN_SIZE;
+
+ /* We want to be extra safe in case entry ip is on the page edge,
+ * but otherwise we need to avoid get_kernel_nofault()'s overhead.
+ */
+ if ((fentry_ip & ~PAGE_MASK) < AARCH64_INSN_SIZE * 2) {
+ if (get_kernel_nofault(insn, (u32 *)(fentry_ip - AARCH64_INSN_SIZE * 2)))
+ return 0;
+ } else {
+ insn = *(u32 *)(fentry_ip - AARCH64_INSN_SIZE * 2);
+ }
+
+ if (aarch64_insn_is_bti(le32_to_cpu((__le32)insn)))
+ return fentry_ip - AARCH64_INSN_SIZE * 2;
+
+ return fentry_ip - AARCH64_INSN_SIZE;
+}
+
/*
* Replace a single instruction, which may be a branch or NOP.
* If @validate == true, a replaced instruction is checked against 'old'.
diff --git a/arch/x86/include/asm/ftrace.h b/arch/x86/include/asm/ftrace.h
index cc92c99ef276..f9cb4d07df58 100644
--- a/arch/x86/include/asm/ftrace.h
+++ b/arch/x86/include/asm/ftrace.h
@@ -34,6 +34,27 @@ static inline unsigned long ftrace_call_adjust(unsigned long addr)
return addr;
}
+static inline unsigned long arch_ftrace_get_symaddr(unsigned long fentry_ip)
+{
+#ifdef CONFIG_X86_KERNEL_IBT
+ u32 instr;
+
+ /* We want to be extra safe in case entry ip is on the page edge,
+ * but otherwise we need to avoid get_kernel_nofault()'s overhead.
+ */
+ if ((fentry_ip & ~PAGE_MASK) < ENDBR_INSN_SIZE) {
+ if (get_kernel_nofault(instr, (u32 *)(fentry_ip - ENDBR_INSN_SIZE)))
+ return fentry_ip;
+ } else {
+ instr = *(u32 *)(fentry_ip - ENDBR_INSN_SIZE);
+ }
+ if (is_endbr(instr))
+ fentry_ip -= ENDBR_INSN_SIZE;
+#endif
+ return fentry_ip;
+}
+#define ftrace_get_symaddr(fentry_ip) arch_ftrace_get_symaddr(fentry_ip)
+
#ifdef CONFIG_HAVE_DYNAMIC_FTRACE_WITH_ARGS
#include <linux/ftrace_regs.h>
diff --git a/include/linux/ftrace.h b/include/linux/ftrace.h
index 4c553fe9c026..07092dfb21a4 100644
--- a/include/linux/ftrace.h
+++ b/include/linux/ftrace.h
@@ -622,6 +622,19 @@ enum {
FTRACE_MAY_SLEEP = (1 << 5),
};
+/* Arches can override ftrace_get_symaddr() to convert fentry_ip to symaddr. */
+#ifndef ftrace_get_symaddr
+/**
+ * ftrace_get_symaddr - return the symbol address from fentry_ip
+ * @fentry_ip: the address of ftrace location
+ *
+ * Get the symbol address from @fentry_ip (fast path). If there is no fast
+ * search path, this returns 0.
+ * User may need to use kallsyms API to find the symbol address.
+ */
+#define ftrace_get_symaddr(fentry_ip) (0)
+#endif
+
#ifdef CONFIG_DYNAMIC_FTRACE
void ftrace_arch_code_modify_prepare(void);
Hello,
I got errors building Linux 6.14-rc1 that were solved by reverting this
patch and the one after (19/20 and 20/20).
Errors look like:
In file included from ./arch/x86/include/asm/asm-prototypes.h:2,
from <stdin>:3:
./arch/x86/include/asm/ftrace.h: In function 'arch_ftrace_get_symaddr':
./arch/x86/include/asm/ftrace.h:46:21: error: implicit declaration of
function 'get_kernel_nofault' [-Wimplicit-function-declaration]
46 | if (get_kernel_nofault(instr, (u32 *)(fentry_ip
- ENDBR_INSN_SIZE)))
| ^~~~~~~~~~~~~~~~~~
Will send .config on request if needed.
Le 26/12/2024 à 06:15, Masami Hiramatsu (Google) a écrit :
> From: Masami Hiramatsu (Google) <mhiramat@kernel.org>
>
> This introduces ftrace_get_symaddr() which tries to convert fentry_ip
> passed by ftrace or fgraph callback to symaddr without calling
> kallsyms API. It returns the symbol address or 0 if it fails to
> convert it.
>
> Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
> Reported-by: kernel test robot <lkp@intel.com>
> Closes: https://lore.kernel.org/oe-kbuild-all/202412061423.K79V55Hd-lkp@intel.com/
> Closes: https://lore.kernel.org/oe-kbuild-all/202412061804.5VRzF14E-lkp@intel.com/
> ---
> Changes in v21:
> - On arm64, fix the macro name to ftrace_get_symaddr() correctly.
> - Define ftrace_get_symaddr() outside of CONFIG_DYNAMIC_FTRACE.
> Changes in v19:
> - Newly added.
> ---
> arch/arm64/include/asm/ftrace.h | 2 +
> arch/arm64/kernel/ftrace.c | 63 +++++++++++++++++++++++++++++++++++++++
> arch/x86/include/asm/ftrace.h | 21 +++++++++++++
> include/linux/ftrace.h | 13 ++++++++
> 4 files changed, 99 insertions(+)
>
> diff --git a/arch/arm64/include/asm/ftrace.h b/arch/arm64/include/asm/ftrace.h
> index 876e88ad4119..bfe3ce9df197 100644
> --- a/arch/arm64/include/asm/ftrace.h
> +++ b/arch/arm64/include/asm/ftrace.h
> @@ -52,6 +52,8 @@ extern unsigned long ftrace_graph_call;
> extern void return_to_handler(void);
>
> unsigned long ftrace_call_adjust(unsigned long addr);
> +unsigned long arch_ftrace_get_symaddr(unsigned long fentry_ip);
> +#define ftrace_get_symaddr(fentry_ip) arch_ftrace_get_symaddr(fentry_ip)
>
> #ifdef CONFIG_DYNAMIC_FTRACE_WITH_ARGS
> #define HAVE_ARCH_FTRACE_REGS
> diff --git a/arch/arm64/kernel/ftrace.c b/arch/arm64/kernel/ftrace.c
> index 570c38be833c..d7c0d023dfe5 100644
> --- a/arch/arm64/kernel/ftrace.c
> +++ b/arch/arm64/kernel/ftrace.c
> @@ -143,6 +143,69 @@ unsigned long ftrace_call_adjust(unsigned long addr)
> return addr;
> }
>
> +/* Convert fentry_ip to the symbol address without kallsyms */
> +unsigned long arch_ftrace_get_symaddr(unsigned long fentry_ip)
> +{
> + u32 insn;
> +
> + /*
> + * When using patchable-function-entry without pre-function NOPS, ftrace
> + * entry is the address of the first NOP after the function entry point.
> + *
> + * The compiler has either generated:
> + *
> + * func+00: func: NOP // To be patched to MOV X9, LR
> + * func+04: NOP // To be patched to BL <caller>
> + *
> + * Or:
> + *
> + * func-04: BTI C
> + * func+00: func: NOP // To be patched to MOV X9, LR
> + * func+04: NOP // To be patched to BL <caller>
> + *
> + * The fentry_ip is the address of `BL <caller>` which is at `func + 4`
> + * bytes in either case.
> + */
> + if (!IS_ENABLED(CONFIG_DYNAMIC_FTRACE_WITH_CALL_OPS))
> + return fentry_ip - AARCH64_INSN_SIZE;
> +
> + /*
> + * When using patchable-function-entry with pre-function NOPs, BTI is
> + * a bit different.
> + *
> + * func+00: func: NOP // To be patched to MOV X9, LR
> + * func+04: NOP // To be patched to BL <caller>
> + *
> + * Or:
> + *
> + * func+00: func: BTI C
> + * func+04: NOP // To be patched to MOV X9, LR
> + * func+08: NOP // To be patched to BL <caller>
> + *
> + * The fentry_ip is the address of `BL <caller>` which is at either
> + * `func + 4` or `func + 8` depends on whether there is a BTI.
> + */
> +
> + /* If there is no BTI, the func address should be one instruction before. */
> + if (!IS_ENABLED(CONFIG_ARM64_BTI_KERNEL))
> + return fentry_ip - AARCH64_INSN_SIZE;
> +
> + /* We want to be extra safe in case entry ip is on the page edge,
> + * but otherwise we need to avoid get_kernel_nofault()'s overhead.
> + */
> + if ((fentry_ip & ~PAGE_MASK) < AARCH64_INSN_SIZE * 2) {
> + if (get_kernel_nofault(insn, (u32 *)(fentry_ip - AARCH64_INSN_SIZE * 2)))
> + return 0;
> + } else {
> + insn = *(u32 *)(fentry_ip - AARCH64_INSN_SIZE * 2);
> + }
> +
> + if (aarch64_insn_is_bti(le32_to_cpu((__le32)insn)))
> + return fentry_ip - AARCH64_INSN_SIZE * 2;
> +
> + return fentry_ip - AARCH64_INSN_SIZE;
> +}
> +
> /*
> * Replace a single instruction, which may be a branch or NOP.
> * If @validate == true, a replaced instruction is checked against 'old'.
> diff --git a/arch/x86/include/asm/ftrace.h b/arch/x86/include/asm/ftrace.h
> index cc92c99ef276..f9cb4d07df58 100644
> --- a/arch/x86/include/asm/ftrace.h
> +++ b/arch/x86/include/asm/ftrace.h
> @@ -34,6 +34,27 @@ static inline unsigned long ftrace_call_adjust(unsigned long addr)
> return addr;
> }
>
> +static inline unsigned long arch_ftrace_get_symaddr(unsigned long fentry_ip)
> +{
> +#ifdef CONFIG_X86_KERNEL_IBT
> + u32 instr;
> +
> + /* We want to be extra safe in case entry ip is on the page edge,
> + * but otherwise we need to avoid get_kernel_nofault()'s overhead.
> + */
> + if ((fentry_ip & ~PAGE_MASK) < ENDBR_INSN_SIZE) {
> + if (get_kernel_nofault(instr, (u32 *)(fentry_ip - ENDBR_INSN_SIZE)))
> + return fentry_ip;
> + } else {
> + instr = *(u32 *)(fentry_ip - ENDBR_INSN_SIZE);
> + }
> + if (is_endbr(instr))
> + fentry_ip -= ENDBR_INSN_SIZE;
> +#endif
> + return fentry_ip;
> +}
> +#define ftrace_get_symaddr(fentry_ip) arch_ftrace_get_symaddr(fentry_ip)
> +
> #ifdef CONFIG_HAVE_DYNAMIC_FTRACE_WITH_ARGS
>
> #include <linux/ftrace_regs.h>
> diff --git a/include/linux/ftrace.h b/include/linux/ftrace.h
> index 4c553fe9c026..07092dfb21a4 100644
> --- a/include/linux/ftrace.h
> +++ b/include/linux/ftrace.h
> @@ -622,6 +622,19 @@ enum {
> FTRACE_MAY_SLEEP = (1 << 5),
> };
>
> +/* Arches can override ftrace_get_symaddr() to convert fentry_ip to symaddr. */
> +#ifndef ftrace_get_symaddr
> +/**
> + * ftrace_get_symaddr - return the symbol address from fentry_ip
> + * @fentry_ip: the address of ftrace location
> + *
> + * Get the symbol address from @fentry_ip (fast path). If there is no fast
> + * search path, this returns 0.
> + * User may need to use kallsyms API to find the symbol address.
> + */
> +#define ftrace_get_symaddr(fentry_ip) (0)
> +#endif
> +
> #ifdef CONFIG_DYNAMIC_FTRACE
>
> void ftrace_arch_code_modify_prepare(void);
>
>
On Mon, 3 Feb 2025 22:33:48 +0100
Gabriel de Perthuis <g2p.code@gmail.com> wrote:
> Hello,
>
> I got errors building Linux 6.14-rc1 that were solved by reverting this
> patch and the one after (19/20 and 20/20).
>
> Errors look like:
>
> In file included from ./arch/x86/include/asm/asm-prototypes.h:2,
> from <stdin>:3:
> ./arch/x86/include/asm/ftrace.h: In function 'arch_ftrace_get_symaddr':
> ./arch/x86/include/asm/ftrace.h:46:21: error: implicit declaration of
> function 'get_kernel_nofault' [-Wimplicit-function-declaration]
> 46 | if (get_kernel_nofault(instr, (u32 *)(fentry_ip
> - ENDBR_INSN_SIZE)))
> | ^~~~~~~~~~~~~~~~~~
>
> Will send .config on request if needed.
Thanks for the report!
-------<arch/x86/include/asm/asm-prototypes.h>
/* SPDX-License-Identifier: GPL-2.0 */
#include <asm/ftrace.h>
#include <linux/uaccess.h>
-----
Ah, that's why... get_kernel_nofault() is defined in linux/uaccess.h.
I also found that is_endbr() is in asm/ibt.h.
Can you try this?
Thank you,
diff --git a/arch/x86/include/asm/ftrace.h b/arch/x86/include/asm/ftrace.h
index f9cb4d07df58..d24d7c71253f 100644
--- a/arch/x86/include/asm/ftrace.h
+++ b/arch/x86/include/asm/ftrace.h
@@ -16,24 +16,9 @@
# include <asm/ibt.h>
/* Add offset for endbr64 if IBT enabled */
# define FTRACE_MCOUNT_MAX_OFFSET ENDBR_INSN_SIZE
-#endif
-
-#ifdef CONFIG_DYNAMIC_FTRACE
-#define ARCH_SUPPORTS_FTRACE_OPS 1
-#endif
-
-#ifndef __ASSEMBLY__
-extern void __fentry__(void);
-
-static inline unsigned long ftrace_call_adjust(unsigned long addr)
-{
- /*
- * addr is the address of the mcount call instruction.
- * recordmcount does the necessary offset calculation.
- */
- return addr;
-}
+#include <linux/uaccess.h>
+/* This only supports fentry based ftrace. */
static inline unsigned long arch_ftrace_get_symaddr(unsigned long fentry_ip)
{
#ifdef CONFIG_X86_KERNEL_IBT
@@ -55,6 +40,24 @@ static inline unsigned long arch_ftrace_get_symaddr(unsigned long fentry_ip)
}
#define ftrace_get_symaddr(fentry_ip) arch_ftrace_get_symaddr(fentry_ip)
+#endif
+
+#ifdef CONFIG_DYNAMIC_FTRACE
+#define ARCH_SUPPORTS_FTRACE_OPS 1
+#endif
+
+#ifndef __ASSEMBLY__
+extern void __fentry__(void);
+
+static inline unsigned long ftrace_call_adjust(unsigned long addr)
+{
+ /*
+ * addr is the address of the mcount call instruction.
+ * recordmcount does the necessary offset calculation.
+ */
+ return addr;
+}
+
#ifdef CONFIG_HAVE_DYNAMIC_FTRACE_WITH_ARGS
#include <linux/ftrace_regs.h>
--
Masami Hiramatsu (Google) <mhiramat@kernel.org>
© 2016 - 2025 Red Hat, Inc.