[PATCH v10 03/14] unwind_user: Add compat mode frame pointer support

Steven Rostedt posted 14 patches 4 months ago
There is a newer version of this series
[PATCH v10 03/14] unwind_user: Add compat mode frame pointer support
Posted by Steven Rostedt 4 months ago
From: Josh Poimboeuf <jpoimboe@kernel.org>

Add optional support for user space compat mode frame pointer unwinding.
If supported, the arch needs to enable CONFIG_HAVE_UNWIND_USER_COMPAT_FP
and define ARCH_INIT_USER_COMPAT_FP_FRAME.

Signed-off-by: Josh Poimboeuf <jpoimboe@kernel.org>
Co-developed-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
---
Changes since v9: https://lore.kernel.org/linux-trace-kernel/20250513223551.459986355@goodmis.org/

As asm-generic headers are not included when an architecture defines the
header, having more than one #ifndef and setting variables does not work
with those checks in the asm-generic header and the architecture header
does not define all the values.

- Move #indef arch_unwind_user_state to linux/user_unwind_types.h

- Move the following to linux/unwind_user.h:
   #ifndef ARCH_INIT_USER_COMPAT_FP_FRAME
   #ifndef arch_unwind_user_init
   #ifndef arch_unwind_user_next

- Changed UNWIND_GET_USER_LONG() to use "unsigned long" instead of u64 as
  this can be called on 32 bit architectures and just because
  "compat_state()" returns false doesn't mean that the value is 64 bit.

 arch/Kconfig                            |  4 +++
 include/asm-generic/Kbuild              |  1 +
 include/asm-generic/unwind_user_types.h |  5 ++++
 include/linux/unwind_user.h             | 13 +++++++++
 include/linux/unwind_user_types.h       |  7 +++++
 kernel/unwind/user.c                    | 36 ++++++++++++++++++++++---
 6 files changed, 62 insertions(+), 4 deletions(-)
 create mode 100644 include/asm-generic/unwind_user_types.h

diff --git a/arch/Kconfig b/arch/Kconfig
index 8e3fd723bd74..2c41d3072910 100644
--- a/arch/Kconfig
+++ b/arch/Kconfig
@@ -442,6 +442,10 @@ config HAVE_UNWIND_USER_FP
 	bool
 	select UNWIND_USER
 
+config HAVE_UNWIND_USER_COMPAT_FP
+	bool
+	depends on HAVE_UNWIND_USER_FP
+
 config HAVE_PERF_REGS
 	bool
 	help
diff --git a/include/asm-generic/Kbuild b/include/asm-generic/Kbuild
index 295c94a3ccc1..b797a2434396 100644
--- a/include/asm-generic/Kbuild
+++ b/include/asm-generic/Kbuild
@@ -60,6 +60,7 @@ mandatory-y += topology.h
 mandatory-y += trace_clock.h
 mandatory-y += uaccess.h
 mandatory-y += unwind_user.h
+mandatory-y += unwind_user_types.h
 mandatory-y += vermagic.h
 mandatory-y += vga.h
 mandatory-y += video.h
diff --git a/include/asm-generic/unwind_user_types.h b/include/asm-generic/unwind_user_types.h
new file mode 100644
index 000000000000..f568b82e52cd
--- /dev/null
+++ b/include/asm-generic/unwind_user_types.h
@@ -0,0 +1,5 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _ASM_GENERIC_UNWIND_USER_TYPES_H
+#define _ASM_GENERIC_UNWIND_USER_TYPES_H
+
+#endif /* _ASM_GENERIC_UNWIND_USER_TYPES_H */
diff --git a/include/linux/unwind_user.h b/include/linux/unwind_user.h
index a405111c41b0..c70da8f7e54c 100644
--- a/include/linux/unwind_user.h
+++ b/include/linux/unwind_user.h
@@ -9,6 +9,19 @@
  #define ARCH_INIT_USER_FP_FRAME
 #endif
 
+#ifndef ARCH_INIT_USER_COMPAT_FP_FRAME
+ #define ARCH_INIT_USER_COMPAT_FP_FRAME
+ #define in_compat_mode(regs) false
+#endif
+
+#ifndef arch_unwind_user_init
+static inline void arch_unwind_user_init(struct unwind_user_state *state, struct pt_regs *reg) {}
+#endif
+
+#ifndef arch_unwind_user_next
+static inline void arch_unwind_user_next(struct unwind_user_state *state) {}
+#endif
+
 int unwind_user_start(struct unwind_user_state *state);
 int unwind_user_next(struct unwind_user_state *state);
 
diff --git a/include/linux/unwind_user_types.h b/include/linux/unwind_user_types.h
index 65bd070eb6b0..0b6563951ca4 100644
--- a/include/linux/unwind_user_types.h
+++ b/include/linux/unwind_user_types.h
@@ -3,10 +3,16 @@
 #define _LINUX_UNWIND_USER_TYPES_H
 
 #include <linux/types.h>
+#include <asm/unwind_user_types.h>
+
+#ifndef arch_unwind_user_state
+struct arch_unwind_user_state {};
+#endif
 
 enum unwind_user_type {
 	UNWIND_USER_TYPE_NONE,
 	UNWIND_USER_TYPE_FP,
+	UNWIND_USER_TYPE_COMPAT_FP,
 };
 
 struct unwind_stacktrace {
@@ -25,6 +31,7 @@ struct unwind_user_state {
 	unsigned long ip;
 	unsigned long sp;
 	unsigned long fp;
+	struct arch_unwind_user_state arch;
 	enum unwind_user_type type;
 	bool done;
 };
diff --git a/kernel/unwind/user.c b/kernel/unwind/user.c
index 4fc550356b33..29e1f497a26e 100644
--- a/kernel/unwind/user.c
+++ b/kernel/unwind/user.c
@@ -12,12 +12,32 @@ static struct unwind_user_frame fp_frame = {
 	ARCH_INIT_USER_FP_FRAME
 };
 
+static struct unwind_user_frame compat_fp_frame = {
+	ARCH_INIT_USER_COMPAT_FP_FRAME
+};
+
 static inline bool fp_state(struct unwind_user_state *state)
 {
 	return IS_ENABLED(CONFIG_HAVE_UNWIND_USER_FP) &&
 	       state->type == UNWIND_USER_TYPE_FP;
 }
 
+static inline bool compat_state(struct unwind_user_state *state)
+{
+	return IS_ENABLED(CONFIG_HAVE_UNWIND_USER_COMPAT_FP) &&
+	       state->type == UNWIND_USER_TYPE_COMPAT_FP;
+}
+
+#define UNWIND_GET_USER_LONG(to, from, state)				\
+({									\
+	int __ret;							\
+	if (compat_state(state))					\
+		__ret = get_user(to, (u32 __user *)(from));		\
+	else								\
+		__ret = get_user(to, (unsigned long __user *)(from));	\
+	__ret;								\
+})
+
 int unwind_user_next(struct unwind_user_state *state)
 {
 	struct unwind_user_frame *frame;
@@ -26,7 +46,9 @@ int unwind_user_next(struct unwind_user_state *state)
 	if (state->done)
 		return -EINVAL;
 
-	if (fp_state(state))
+	if (compat_state(state))
+		frame = &compat_fp_frame;
+	else if (fp_state(state))
 		frame = &fp_frame;
 	else
 		goto the_end;
@@ -37,10 +59,10 @@ int unwind_user_next(struct unwind_user_state *state)
 	if (cfa <= state->sp)
 		goto the_end;
 
-	if (get_user(ra, (unsigned long *)(cfa + frame->ra_off)))
+	if (UNWIND_GET_USER_LONG(ra, cfa + frame->ra_off, state))
 		goto the_end;
 
-	if (frame->fp_off && get_user(fp, (unsigned long __user *)(cfa + frame->fp_off)))
+	if (frame->fp_off && UNWIND_GET_USER_LONG(fp, cfa + frame->fp_off, state))
 		goto the_end;
 
 	state->ip = ra;
@@ -48,6 +70,8 @@ int unwind_user_next(struct unwind_user_state *state)
 	if (frame->fp_off)
 		state->fp = fp;
 
+	arch_unwind_user_next(state);
+
 	return 0;
 
 the_end:
@@ -66,7 +90,9 @@ int unwind_user_start(struct unwind_user_state *state)
 		return -EINVAL;
 	}
 
-	if (IS_ENABLED(CONFIG_HAVE_UNWIND_USER_FP))
+	if (IS_ENABLED(CONFIG_HAVE_UNWIND_USER_COMPAT_FP) && in_compat_mode(regs))
+		state->type = UNWIND_USER_TYPE_COMPAT_FP;
+	else if (IS_ENABLED(CONFIG_HAVE_UNWIND_USER_FP))
 		state->type = UNWIND_USER_TYPE_FP;
 	else
 		state->type = UNWIND_USER_TYPE_NONE;
@@ -75,6 +101,8 @@ int unwind_user_start(struct unwind_user_state *state)
 	state->sp = user_stack_pointer(regs);
 	state->fp = frame_pointer(regs);
 
+	arch_unwind_user_init(state, regs);
+
 	return 0;
 }
 
-- 
2.47.2
Re: [PATCH v10 03/14] unwind_user: Add compat mode frame pointer support
Posted by Peter Zijlstra 3 months, 3 weeks ago
On Tue, Jun 10, 2025 at 08:54:24PM -0400, Steven Rostedt wrote:

> +#ifndef arch_unwind_user_init
> +static inline void arch_unwind_user_init(struct unwind_user_state *state, struct pt_regs *reg) {}
> +#endif
> +
> +#ifndef arch_unwind_user_next
> +static inline void arch_unwind_user_next(struct unwind_user_state *state) {}
> +#endif

The purpose of these arch hooks is so far mysterious. No comments, no
changelog, no nothing.
Re: [PATCH v10 03/14] unwind_user: Add compat mode frame pointer support
Posted by Steven Rostedt 3 months, 3 weeks ago
On Wed, 18 Jun 2025 15:47:58 +0200
Peter Zijlstra <peterz@infradead.org> wrote:

> On Tue, Jun 10, 2025 at 08:54:24PM -0400, Steven Rostedt wrote:
> 
> > +#ifndef arch_unwind_user_init
> > +static inline void arch_unwind_user_init(struct unwind_user_state *state, struct pt_regs *reg) {}
> > +#endif
> > +
> > +#ifndef arch_unwind_user_next
> > +static inline void arch_unwind_user_next(struct unwind_user_state *state) {}
> > +#endif  
> 
> The purpose of these arch hooks is so far mysterious. No comments, no
> changelog, no nothing.

I'll add comments.

It's used later in the x86 compat code to allow the architecture to do any
special initialization or to handling moving to the next frame.

From patch 14:

+#define in_compat_mode(regs) !user_64bit_mode(regs)
+
+static inline void arch_unwind_user_init(struct unwind_user_state *state,
+					 struct pt_regs *regs)
+{
+	unsigned long cs_base, ss_base;
+
+	if (state->type != UNWIND_USER_TYPE_COMPAT_FP)
+		return;
+
+	scoped_guard(irqsave) {
+		cs_base = segment_base_address(regs->cs);
+		ss_base = segment_base_address(regs->ss);
+	}
+
+	state->arch.cs_base = cs_base;
+	state->arch.ss_base = ss_base;
+
+	state->ip += cs_base;
+	state->sp += ss_base;
+	state->fp += ss_base;
+}
+#define arch_unwind_user_init arch_unwind_user_init
+
+static inline void arch_unwind_user_next(struct unwind_user_state *state)
+{
+	if (state->type != UNWIND_USER_TYPE_COMPAT_FP)
+		return;
+
+	state->ip += state->arch.cs_base;
+	state->fp += state->arch.ss_base;
+}
+#define arch_unwind_user_next arch_unwind_user_next

-- Steve
Re: [PATCH v10 03/14] unwind_user: Add compat mode frame pointer support
Posted by Peter Zijlstra 3 months, 3 weeks ago
On Wed, Jun 18, 2025 at 11:18:40AM -0400, Steven Rostedt wrote:
> On Wed, 18 Jun 2025 15:47:58 +0200
> Peter Zijlstra <peterz@infradead.org> wrote:
> 
> > On Tue, Jun 10, 2025 at 08:54:24PM -0400, Steven Rostedt wrote:
> > 
> > > +#ifndef arch_unwind_user_init
> > > +static inline void arch_unwind_user_init(struct unwind_user_state *state, struct pt_regs *reg) {}
> > > +#endif
> > > +
> > > +#ifndef arch_unwind_user_next
> > > +static inline void arch_unwind_user_next(struct unwind_user_state *state) {}
> > > +#endif  
> > 
> > The purpose of these arch hooks is so far mysterious. No comments, no
> > changelog, no nothing.
> 
> I'll add comments.

How about you introduce the hooks when they're actually needed instead?
Re: [PATCH v10 03/14] unwind_user: Add compat mode frame pointer support
Posted by Steven Rostedt 3 months, 3 weeks ago
On Thu, 19 Jun 2025 09:51:03 +0200
Peter Zijlstra <peterz@infradead.org> wrote:

> > > 
> > > The purpose of these arch hooks is so far mysterious. No comments, no
> > > changelog, no nothing.  
> > 
> > I'll add comments.  
> 
> How about you introduce the hooks when they're actually needed instead?

OK. Then I'll move the hooks to the later patch.

-- Steve
Re: [PATCH v10 03/14] unwind_user: Add compat mode frame pointer support
Posted by Peter Zijlstra 3 months, 3 weeks ago
On Tue, Jun 10, 2025 at 08:54:24PM -0400, Steven Rostedt wrote:
> diff --git a/kernel/unwind/user.c b/kernel/unwind/user.c
> index 4fc550356b33..29e1f497a26e 100644
> --- a/kernel/unwind/user.c
> +++ b/kernel/unwind/user.c
> @@ -12,12 +12,32 @@ static struct unwind_user_frame fp_frame = {
>  	ARCH_INIT_USER_FP_FRAME
>  };
>  
> +static struct unwind_user_frame compat_fp_frame = {
> +	ARCH_INIT_USER_COMPAT_FP_FRAME
> +};
> +
>  static inline bool fp_state(struct unwind_user_state *state)
>  {
>  	return IS_ENABLED(CONFIG_HAVE_UNWIND_USER_FP) &&
>  	       state->type == UNWIND_USER_TYPE_FP;
>  }
>  
> +static inline bool compat_state(struct unwind_user_state *state)

Consistency would mandate this thing be called: compat_fp_state().

> +{
> +	return IS_ENABLED(CONFIG_HAVE_UNWIND_USER_COMPAT_FP) &&
> +	       state->type == UNWIND_USER_TYPE_COMPAT_FP;
> +}
> +
> +#define UNWIND_GET_USER_LONG(to, from, state)				\

Do we have to shout this?

> +({									\
> +	int __ret;							\
> +	if (compat_state(state))					\
> +		__ret = get_user(to, (u32 __user *)(from));		\
> +	else								\
> +		__ret = get_user(to, (unsigned long __user *)(from));	\
> +	__ret;								\
> +})
> +
>  int unwind_user_next(struct unwind_user_state *state)
>  {
>  	struct unwind_user_frame *frame;
Re: [PATCH v10 03/14] unwind_user: Add compat mode frame pointer support
Posted by Steven Rostedt 3 months, 3 weeks ago
On Wed, 18 Jun 2025 15:46:41 +0200
Peter Zijlstra <peterz@infradead.org> wrote:

> On Tue, Jun 10, 2025 at 08:54:24PM -0400, Steven Rostedt wrote:
> > diff --git a/kernel/unwind/user.c b/kernel/unwind/user.c
> > index 4fc550356b33..29e1f497a26e 100644
> > --- a/kernel/unwind/user.c
> > +++ b/kernel/unwind/user.c
> > @@ -12,12 +12,32 @@ static struct unwind_user_frame fp_frame = {
> >  	ARCH_INIT_USER_FP_FRAME
> >  };
> >  
> > +static struct unwind_user_frame compat_fp_frame = {
> > +	ARCH_INIT_USER_COMPAT_FP_FRAME
> > +};
> > +
> >  static inline bool fp_state(struct unwind_user_state *state)
> >  {
> >  	return IS_ENABLED(CONFIG_HAVE_UNWIND_USER_FP) &&
> >  	       state->type == UNWIND_USER_TYPE_FP;
> >  }
> >  
> > +static inline bool compat_state(struct unwind_user_state *state)  
> 
> Consistency would mandate this thing be called: compat_fp_state().

Sure. Will update it.

> 
> > +{
> > +	return IS_ENABLED(CONFIG_HAVE_UNWIND_USER_COMPAT_FP) &&
> > +	       state->type == UNWIND_USER_TYPE_COMPAT_FP;
> > +}
> > +
> > +#define UNWIND_GET_USER_LONG(to, from, state)				\  
> 
> Do we have to shout this?

Don't we usually shout macros?

-- Steve

> 
> > +({									\
> > +	int __ret;							\
> > +	if (compat_state(state))					\
> > +		__ret = get_user(to, (u32 __user *)(from));		\
> > +	else								\
> > +		__ret = get_user(to, (unsigned long __user *)(from));	\
> > +	__ret;								\
> > +})
> > +
> >  int unwind_user_next(struct unwind_user_state *state)
> >  {
> >  	struct unwind_user_frame *frame;
Re: [PATCH v10 03/14] unwind_user: Add compat mode frame pointer support
Posted by Linus Torvalds 3 months, 3 weeks ago
On Wed, 18 Jun 2025 at 08:10, Steven Rostedt <rostedt@goodmis.org> wrote:
>
> Don't we usually shout macros?

Not really. It's more about "we shout about macros when they _behave_
like macros".

So typically we do all upper-case for things that are constants
(whether enums or macros, actually) or when they don't act like normal
functions.

But when it's not particularly important that it's a macro, and it
could have been a function (but maybe it was just simpler to use a
macro for whatever reason), we typically don't use all upper-case.

In this case, I have to agree with PeterZ that this just looks odd:

-       if (get_user(ra, (unsigned long *)(cfa + frame->ra_off)))
+       if (UNWIND_GET_USER_LONG(ra, cfa + frame->ra_off, state))

why is UNWIND_GET_USER_LONG() so loud when it just replaces "get_user()"?

Note that "get_user()" itself is a macro, and is lower-case, even
though you couldn't actually do it as a function (because it changes
its first argument in place).

             Linus
Re: [PATCH v10 03/14] unwind_user: Add compat mode frame pointer support
Posted by Steven Rostedt 3 months, 3 weeks ago
On Wed, 18 Jun 2025 10:52:22 -0700
Linus Torvalds <torvalds@linux-foundation.org> wrote:

> In this case, I have to agree with PeterZ that this just looks odd:
> 
> -       if (get_user(ra, (unsigned long *)(cfa + frame->ra_off)))
> +       if (UNWIND_GET_USER_LONG(ra, cfa + frame->ra_off, state))
> 
> why is UNWIND_GET_USER_LONG() so loud when it just replaces "get_user()"?
> 
> Note that "get_user()" itself is a macro, and is lower-case, even
> though you couldn't actually do it as a function (because it changes
> its first argument in place).

OK, I'll make it lower case.

Thanks!

-- Steve