Skip to contentSkip to conversation

unwind_user: x86: Deferred unwinding infrastructure

PATCH In review Steven Rostedt 7 participants 53 messages Jul 1, 2025 – Jul 7, 2025

Subsystems (10)
x86 BPF MM core Tracing +6 more subsystems
lore
Share & tools
lore

Steven Rostedt

author maintainer
#0 lore raw

v12 cover unwind_user: x86: Deferred unwinding infrastructure

[
UPDATE: Florian Weimer is looking to having Fedora built with SFrames
so that once this becomes available in the kernel, it will also be usable in Fedora.
]

This is the first patch series of a set that will make it possible to be able to use SFrames[1] in the Linux kernel. A quick recap of the motivation for doing this.

Currently the only way to get a user space stack trace from a stack walk (and not just copying large amount of user stack into the kernel ring buffer) is to use frame pointers. This has a few issues. The biggest one is that compiling frame pointers into every application and library has been shown to cause performance overhead.

Another issue is that the format of the frames may not always be consistent between different compilers and some architectures (s390) has no defined format to do a reliable stack walk. The only way to perform user space profiling on these architectures is to copy the user stack into the kernel buffer.

SFrames is now supported in gcc binutils and soon will also be supported by LLVM. SFrames acts more like ORC, and lives in the ELF executable file as its own section. Like ORC it has two tables where the first table is sorted by instruction pointers (IP) and using the current IP and finding it's entry in the first table, it will take you to the second table which will tell you where the return address of the current function is located and then you can use that address to look it up in the first table to find the return address of that function, and so on. This performs a user space stack walk.

Now because the SFrame section lives in the ELF file it needs to be faulted into memory when it is used. This means that walking the user space stack requires being in a faultable context. As profilers like perf request a stack trace in interrupt or NMI context, it cannot do the walking when it is requested. Instead it must be deferred until it is safe to fault in user space. One place this is known to be safe is when the task is about to return back to user space.

Josh originally wrote the PoC of this code and his last version he posted was back in January:

https://lore.kernel.org/all/[email protected]/

That series contained everything from adding a new faultable user space stack walking code, deferring the stack walk, implementing sframes, fixing up x86 (VDSO), and even added both the kernel and user space side of perf to make it work. But Josh also ran out of time to work on it and I picked it up. As there's several parts to this series, I also broke it out. Especially since there's parts of his series that do not depend on each other.

This series contains only the core infrastructure that all the rest needs. Of the 14 patches, only 3 are x86 specific. The rest is simply the unwinding code that s390 can build against. I moved the 3 x86 specific to the end of the series too.

Since multiple tracers (like perf, ftrace, bpf, etc) can attach to the deferred unwinder and each of these tracers can attach to some or all of the tasks to trace, there is a many to many relationship. This relationship needs to be made in interrupt or NMI context so it can not rely on any allocation. To handle this, a bitmask is used. There's a global bitmask of size long which will allocate a single bit when a tracer registers for deferred stack traces. The task struct will also have a bitmask where a request comes in from one of the tracers to have a deferred stack trace, it will set the corresponding bit for that tracer it its mask. As one of the bits represents that a request has been made, this means at most 31 on 32 bit systems or 63 on 64 bit systems of tracers may be registered at a given time. This should not be an issue as only one perf application, or ftrace instance should request a bit. BPF should also use only one bit and handle any multiplexing for its users.

When the first request is made for a deferred stack trace from a task, it will take a timestamp. This timestamp will be used as the identifier for the user space stack trace. As the user space stack trace does not change while the task is in the kernel, requests that come in after the first request and before the task goes back to user space will get the same timestamp. This timestamp also serves the purpose of knowing how far back a given user space stack trace goes. If there's dropped events, and the events dropped miss a task entering user space and coming back to the kernel, the new stack trace taken when it goes back to user space should not be used with the events before the drop happened.

When a tracer makes a request, it gets this timestamp, and the tasks bitmask sets the bit for the requesting tracer. A task work is used to have the task do the callbacks before it goes back to user space. When it does, it will scan its bitmask and call all the callbacks for the tracers that have their representing bit set. The callback will receive the user space stack trace as well as the timestamp that was used.

That's the basic idea. Obviously there's more to it than the above explanation, but each patch explains what it is doing, and it is broken up step by step.

I run two SFrame meetings once a month (one in Asia friendly timezone and the other in Europe friendly). We have developers from Google, Oracle, Red Hat, IBM, EfficiOS, Meta, Microsoft, and more that attend. (If anyone is interested in attending let me know). I have been running this since December of 2024. Last year in GNU Cauldron, a few of us got together to discuss the design and such. We are pretty confident that the current design is sound. We have working code on top of this and have been testing it.

Since the s390 folks want to start working on this (they have patches to sframes already from working on the prototypes), I would like this series to be a separate branch based on top of v6.16-rc2. Then all the subsystems that want to work on top of this can as there's no real dependency between them.

I have more patches on top of this series that add perf support, ftrace support, sframe support and the x86 fix ups (for VDSO). But each of those patch series can be worked on independently, but they all depend on this series (although the x86 specific patches at the end isn't necessarily needed, at least for other architectures).

Please review, and if you are happy with them, lets get them in a branch that we all can use. I'm happy to take it in my tree if I can get acks on the x86 code. Or it can be in the tip tree as a separate branch on top of 6.16-rc4 and I'll just base my work on top of that. Doesn't matter either way.

This is based on top of v6.16-rc4 and the code is here:

git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace.git unwind/core

Head SHA1: 649fe8a37fbb8bc7eb1d420630523feb4f44d1d7

Changes since v11: https://lore.kernel.org/linux-trace-kernel/[email protected]/

- Add USED bit to the task's unwind_mask to know if the faultable user stack function was used or not. This allows for only having to check one value on the way back to user space to know if it has to do more work or not.

- Fix header macro protection name to include X86 (Ingo Molnar)

- Use insn_get_seg_base() to get segment registers instead of using the function perf uses and making it global. Also as that function doesn't look to have a requirement to disable interrupts, the scoped_guard(irqsave) is removed.

- Check return code of insn_get_seg_base() for the unlikely event that it returns invalid (-1).

- Moved arch_unwind_user_init() into stacktrace.c as to use insn_get_seg_base(), it must include insn-eval.h that defines pt_regs_offset(), but that is also used in the perf generic code as an array and if it is included in the header file, it causes a build conflict.

- Update the comments that explain arch_unwind_user_init/next that a macro needs to be defined with those names if they are going to be used.

Josh Poimboeuf (7):
unwind_user: Add user space unwinding API
unwind_user: Add frame pointer support
unwind_user: Add compat mode frame pointer support unwind_user/deferred: Add unwind cache unwind_user/deferred: Add deferred unwinding interface unwind_user/x86: Enable frame pointer unwinding on x86 unwind_user/x86: Enable compat mode frame pointer unwinding on x86

Steven Rostedt (7):
unwind_user/deferred: Add unwind_user_faultable() unwind_user/deferred: Make unwind deferral requests NMI-safe unwind deferred: Use bitmask to determine which callbacks to call unwind deferred: Use SRCU unwind_deferred_task_work()
unwind: Clear unwind_mask on exit back to user space
unwind: Add USED bit to only have one conditional on way back to user space
unwind: Finish up unwind when a task exits

----

 MAINTAINERS                              |   8 +
 arch/Kconfig                             |  11 +
 arch/x86/Kconfig                         |   2 +
 arch/x86/include/asm/unwind_user.h       |  42 ++++
 arch/x86/include/asm/unwind_user_types.h |  17 ++
 arch/x86/kernel/stacktrace.c             |  28 +++
 include/asm-generic/Kbuild               |   2 +
 include/asm-generic/unwind_user.h        |   5 +
 include/asm-generic/unwind_user_types.h  |   5 +
 include/linux/entry-common.h             |   2 +
 include/linux/sched.h                    |   5 +
 include/linux/unwind_deferred.h          |  79 +++++++
 include/linux/unwind_deferred_types.h    |  20 ++
 include/linux/unwind_user.h              |  45 ++++
 include/linux/unwind_user_types.h        |  39 ++++
 kernel/Makefile                          |   1 +
 kernel/exit.c                            |   2 +
 kernel/fork.c                            |   4 +
 kernel/unwind/Makefile                   |   1 +
 kernel/unwind/deferred.c                 | 357 +++++++++++++++++++++++++++++++
 kernel/unwind/user.c                     | 130 +++++++++++
 21 files changed, 805 insertions(+)
 create mode 100644 arch/x86/include/asm/unwind_user.h
 create mode 100644 arch/x86/include/asm/unwind_user_types.h
 create mode 100644 include/asm-generic/unwind_user.h
 create mode 100644 include/asm-generic/unwind_user_types.h
 create mode 100644 include/linux/unwind_deferred.h
 create mode 100644 include/linux/unwind_deferred_types.h
 create mode 100644 include/linux/unwind_user.h
 create mode 100644 include/linux/unwind_user_types.h
 create mode 100644 kernel/unwind/Makefile
 create mode 100644 kernel/unwind/deferred.c
 create mode 100644 kernel/unwind/user.c
author maintainer
#1 lore raw

PATCH v12 1/14 unwind_user: Add user space unwinding API

From: Josh Poimboeuf <[email protected]>

Introduce a generic API for unwinding user stacks.

In order to expand user space unwinding to be able to handle more complex scenarios, such as deferred unwinding and reading user space information, create a generic interface that all architectures can use that support the various unwinding methods.

This is an alternative method for handling user space stack traces from the simple stack_trace_save_user() API. This does not replace that interface, but this interface will be used to expand the functionality of user space stack walking.

None of the structures introduced will be exposed to user space tooling.

---
 MAINTAINERS                       |  8 +++++
 arch/Kconfig                      |  3 ++
 include/linux/unwind_user.h       | 15 +++++++++
 include/linux/unwind_user_types.h | 31 +++++++++++++++++
 kernel/Makefile                   |  1 +
 kernel/unwind/Makefile            |  1 +
 kernel/unwind/user.c              | 55 +++++++++++++++++++++++++++++++
 7 files changed, 114 insertions(+)
 create mode 100644 include/linux/unwind_user.h
 create mode 100644 include/linux/unwind_user_types.h
 create mode 100644 kernel/unwind/Makefile
 create mode 100644 kernel/unwind/user.c
MAINTAINERS +8 -0
index 4bac4ea21b64..ed5705c4f7d9 100644
@@ -25924,6 +25924,14 @@ F:	Documentation/driver-api/uio-howto.rst
 F:	drivers/uio/
 F:	include/linux/uio_driver.h
 
+USERSPACE STACK UNWINDING
+M:	Josh Poimboeuf <[email protected]>
+M:	Steven Rostedt <[email protected]>
+S:	Maintained
+F:	include/linux/unwind*.h
+F:	kernel/unwind/
+
+
 UTIL-LINUX PACKAGE
 M:	Karel Zak <[email protected]>
 L:	[email protected]
arch/Kconfig +3 -0
index a3308a220f86..ea59e5d7cc69 100644
@@ -435,6 +435,9 @@ config HAVE_HARDLOCKUP_DETECTOR_ARCH
 	  It uses the same command line parameters, and sysctl interface,
 	  as the generic hardlockup detectors.
 
+config UNWIND_USER
+	bool
+
 config HAVE_PERF_REGS
 	bool
 	help
include/linux/unwind_user.h new file +15 -0
new file mode 100644
index 000000000000..aa7923c1384f
@@ -0,0 +1,15 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _LINUX_UNWIND_USER_H
+#define _LINUX_UNWIND_USER_H
+
+#include <linux/unwind_user_types.h>
+
+int unwind_user_start(struct unwind_user_state *state);
+int unwind_user_next(struct unwind_user_state *state);
+
+int unwind_user(struct unwind_stacktrace *trace, unsigned int max_entries);
+
+#define for_each_user_frame(state) \
+	for (unwind_user_start((state)); !(state)->done; unwind_user_next((state)))
+
+#endif /* _LINUX_UNWIND_USER_H */
include/linux/unwind_user_types.h new file +31 -0
new file mode 100644
index 000000000000..6ed1b4ae74e1
@@ -0,0 +1,31 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _LINUX_UNWIND_USER_TYPES_H
+#define _LINUX_UNWIND_USER_TYPES_H
+
+#include <linux/types.h>
+
+enum unwind_user_type {
+	UNWIND_USER_TYPE_NONE,
+};
+
+struct unwind_stacktrace {
+	unsigned int	nr;
+	unsigned long	*entries;
+};
+
+struct unwind_user_frame {
+	s32 cfa_off;
+	s32 ra_off;
+	s32 fp_off;
+	bool use_fp;
+};
+
+struct unwind_user_state {
+	unsigned long ip;
+	unsigned long sp;
+	unsigned long fp;
+	enum unwind_user_type type;
+	bool done;
+};
+
+#endif /* _LINUX_UNWIND_USER_TYPES_H */
kernel/Makefile +1 -0
index 32e80dd626af..541186050251 100644
@@ -55,6 +55,7 @@ obj-y += rcu/
 obj-y += livepatch/
 obj-y += dma/
 obj-y += entry/
+obj-y += unwind/
 obj-$(CONFIG_MODULES) += module/
 
 obj-$(CONFIG_KCMP) += kcmp.o
kernel/unwind/Makefile new file +1 -0
new file mode 100644
index 000000000000..349ce3677526
@@ -0,0 +1 @@
+ obj-$(CONFIG_UNWIND_USER) += user.o
kernel/unwind/user.c new file +55 -0

55 lines not on this page · raw message

Signature · 1 line
2.47.2
#2 lore raw

Re: [PATCH v12 01/14] unwind_user: Add user space unwinding API

On 2025-06-30 20:53, Steven Rostedt wrote:
[...]

Quoted 5 lines
include/linux/unwind_user.h new file +0 -0
new file mode 100644
index 000000000000..aa7923c1384f

[...]

include/linux/unwind_user.h line 13

Quoted 5 lines
include/linux/unwind_user.h new file +5 -0
+
+int unwind_user(struct unwind_stacktrace *trace, unsigned int max_entries);
+
+#define for_each_user_frame(state) \
+	for (unwind_user_start((state)); !(state)->done; unwind_user_next((state)))

You can remove the () around "state" when it's already surrounded by parentheses:

+#define for_each_user_frame(state) \
+ for (unwind_user_start(state); !(state)->done; unwind_user_next(state))

Thanks,

Mathieu

Signature · 3 lines
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com
#3 lore raw

Re: [PATCH v12 01/14] unwind_user: Add user space unwinding API

In reply to Steven Rostedt
Quoted 9 lines · Steven Rostedt

From: Josh Poimboeuf <[email protected]>

Introduce a generic API for unwinding user stacks.

In order to expand user space unwinding to be able to handle more complex scenarios, such as deferred unwinding and reading user space information, create a generic interface that all architectures can use that support the various unwinding methods.

Quoted 6 lines

This is an alternative method for handling user space stack traces from the simple stack_trace_save_user() API. This does not replace that interface, but this interface will be used to expand the functionality of user space stack walking.

None of the structures introduced will be exposed to user space tooling.

Would it be possible to make those unwind APIs EXPORT_SYMBOL_GPL so they are available for GPL kernel modules ?

Thanks,

Mathieu

Quoted 71 lines
---
  MAINTAINERS                       |  8 +++++
  arch/Kconfig                      |  3 ++
  include/linux/unwind_user.h       | 15 +++++++++
  include/linux/unwind_user_types.h | 31 +++++++++++++++++
  kernel/Makefile                   |  1 +
  kernel/unwind/Makefile            |  1 +
  kernel/unwind/user.c              | 55 +++++++++++++++++++++++++++++++
  7 files changed, 114 insertions(+)

create mode 100644 include/linux/unwind_user.h create mode 100644 include/linux/unwind_user_types.h create mode 100644 kernel/unwind/Makefile create mode 100644 kernel/unwind/user.c

MAINTAINERS +8 -0
index 4bac4ea21b64..ed5705c4f7d9 100644
@@ -25924,6 +25924,14 @@ F:	Documentation/driver-api/uio-howto.rst
  F:	drivers/uio/
  F:	include/linux/uio_driver.h
 
+USERSPACE STACK UNWINDING
+M:	Josh Poimboeuf <[email protected]>
+M:	Steven Rostedt <[email protected]>
+S:	Maintained
+F:	include/linux/unwind*.h
+F:	kernel/unwind/
+
+
  UTIL-LINUX PACKAGE
  M:	Karel Zak <[email protected]>
  L:	[email protected]
arch/Kconfig +3 -0
index a3308a220f86..ea59e5d7cc69 100644
@@ -435,6 +435,9 @@ config HAVE_HARDLOCKUP_DETECTOR_ARCH
  	  It uses the same command line parameters, and sysctl interface,
  	  as the generic hardlockup detectors.
 
+config UNWIND_USER
+	bool
+
  config HAVE_PERF_REGS
  	bool
  	help
include/linux/unwind_user.h new file +15 -0
new file mode 100644
index 000000000000..aa7923c1384f
@@ -0,0 +1,15 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _LINUX_UNWIND_USER_H
+#define _LINUX_UNWIND_USER_H
+
+#include <linux/unwind_user_types.h>
+
+int unwind_user_start(struct unwind_user_state *state);
+int unwind_user_next(struct unwind_user_state *state);
+
+int unwind_user(struct unwind_stacktrace *trace, unsigned int max_entries);
+
+#define for_each_user_frame(state) \
+	for (unwind_user_start((state)); !(state)->done; unwind_user_next((state)))
+
+#endif /* _LINUX_UNWIND_USER_H */

[...]

Signature · 3 lines
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com
author maintainer
#4 lore raw

Re: [PATCH v12 01/14] unwind_user: Add user space unwinding API

Quoted 4 lines · On Fri, 4 Jul 2025 14:20:54 -0400 Mathieu Desnoyers…
Quoted 1 line

None of the structures introduced will be exposed to user space tooling.

Would it be possible to make those unwind APIs EXPORT_SYMBOL_GPL so they are available for GPL kernel modules ?

I'm OK with that, but others tend to complain about EXPORT_SYMBOL_GPL for functions not used by modules in the kernel. But I personally feel that LTTng should get an exception for that rule ;-)

-- Steve

author maintainer
#5 lore raw

Re: [PATCH v12 01/14] unwind_user: Add user space unwinding API

Quoted 11 lines · On Mon, 7 Jul 2025 15:42:45 -0400 Steven Rostedt <[email protected]> wrote:
Quoted 4 lines · On Fri, 4 Jul 2025 14:20:54 -0400 Mathieu Desnoyers…
Quoted 1 line

None of the structures introduced will be exposed to user space tooling.

Would it be possible to make those unwind APIs EXPORT_SYMBOL_GPL so they are available for GPL kernel modules ?

I'm OK with that, but others tend to complain about EXPORT_SYMBOL_GPL for functions not used by modules in the kernel. But I personally feel that LTTng should get an exception for that rule ;-)

I just noticed that this was to patch 1. The functions here probably shouldn't be exported as they are more internal to the infrastructure.

In fact, I think I'll move that macro into the user.c code. I don't think it should be used outside that function. And the unwind_user_start/next() could also be static functions.

-- Steve

author maintainer
#6 lore raw

PATCH v12 2/14 unwind_user: Add frame pointer support

In reply to Steven Rostedt

From: Josh Poimboeuf <[email protected]>

Add optional support for user space frame pointer unwinding. If supported, the arch needs to enable CONFIG_HAVE_UNWIND_USER_FP and define ARCH_INIT_USER_FP_FRAME.

By encoding the frame offsets in struct unwind_user_frame, much of this code can also be reused for future unwinder implementations like sframe.

---
 arch/Kconfig                      |  4 +++
 include/asm-generic/Kbuild        |  1 +
 include/asm-generic/unwind_user.h |  5 +++
 include/linux/unwind_user.h       |  5 +++
 include/linux/unwind_user_types.h |  1 +
 kernel/unwind/user.c              | 51 +++++++++++++++++++++++++++++--
 6 files changed, 65 insertions(+), 2 deletions(-)
 create mode 100644 include/asm-generic/unwind_user.h
arch/Kconfig +4 -0
index ea59e5d7cc69..8e3fd723bd74 100644
@@ -438,6 +438,10 @@ config HAVE_HARDLOCKUP_DETECTOR_ARCH
 config UNWIND_USER
 	bool
 
+config HAVE_UNWIND_USER_FP
+	bool
+	select UNWIND_USER
+
 config HAVE_PERF_REGS
 	bool
 	help
include/asm-generic/Kbuild +1 -0
index 8675b7b4ad23..295c94a3ccc1 100644
@@ -59,6 +59,7 @@ mandatory-y += tlbflush.h
 mandatory-y += topology.h
 mandatory-y += trace_clock.h
 mandatory-y += uaccess.h
+mandatory-y += unwind_user.h
 mandatory-y += vermagic.h
 mandatory-y += vga.h
 mandatory-y += video.h
include/asm-generic/unwind_user.h new file +5 -0
new file mode 100644
index 000000000000..b8882b909944
@@ -0,0 +1,5 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _ASM_GENERIC_UNWIND_USER_H
+#define _ASM_GENERIC_UNWIND_USER_H
+
+#endif /* _ASM_GENERIC_UNWIND_USER_H */
include/linux/unwind_user.h +5 -0
index aa7923c1384f..a405111c41b0 100644
@@ -3,6 +3,11 @@
 #define _LINUX_UNWIND_USER_H
 
 #include <linux/unwind_user_types.h>
+#include <asm/unwind_user.h>
+
+#ifndef ARCH_INIT_USER_FP_FRAME
+ #define ARCH_INIT_USER_FP_FRAME
+#endif
 
 int unwind_user_start(struct unwind_user_state *state);
 int unwind_user_next(struct unwind_user_state *state);
include/linux/unwind_user_types.h +1 -0
index 6ed1b4ae74e1..65bd070eb6b0 100644
@@ -6,6 +6,7 @@
 
 enum unwind_user_type {
 	UNWIND_USER_TYPE_NONE,
+	UNWIND_USER_TYPE_FP,
 };
 
 struct unwind_stacktrace {
kernel/unwind/user.c +49 -2

66 lines not on this page · raw message

Signature · 1 line
2.47.2
#7 lore raw

Re: [PATCH v12 02/14] unwind_user: Add frame pointer support

kernel/unwind/user.c line 39

Quoted 4 lines · Steven Rostedt
kernel/unwind/user.c +3 -0
+       /* stack going in wrong direction? */
+       if (cfa <= state->sp)
+               goto done;

I suspect this should do a lot more testing.

kernel/unwind/user.c line 46

Quoted 6 lines
kernel/unwind/user.c +6 -0
+       /* Find the Return Address (RA) */
+       if (get_user(ra, (unsigned long *)(cfa + frame->ra_off)))
+               goto done;
+
+       if (frame->fp_off && get_user(fp, (unsigned long __user *)(cfa + frame->fp_off)))
+               goto done;

.. and this should check the frame for validity too. At a minimum it should be properly aligned, but things like "it had better be above the current frame" to avoid having some loop would seem to be a good idea.

Maybe even check that it's the same vma?

Linus

author maintainer
#8 lore raw

Re: [PATCH v12 02/14] unwind_user: Add frame pointer support

Quoted 7 lines · On Mon, 30 Jun 2025 19:10:09 -0700 Linus Torvalds…

kernel/unwind/user.c line 39

Quoted 4 lines · Steven Rostedt
kernel/unwind/user.c +3 -0
+       /* stack going in wrong direction? */
+       if (cfa <= state->sp)
+               goto done;  

I suspect this should do a lot more testing.

Sure.

Quoted 8 lines

kernel/unwind/user.c line 46

Quoted 7 lines
kernel/unwind/user.c +6 -0
+       /* Find the Return Address (RA) */
+       if (get_user(ra, (unsigned long *)(cfa + frame->ra_off)))
+               goto done;
+
+       if (frame->fp_off && get_user(fp, (unsigned long __user *)(cfa + frame->fp_off)))
+               goto done;  
Quoted 4 lines

.. and this should check the frame for validity too. At a minimum it should be properly aligned, but things like "it had better be above the current frame" to avoid having some loop would seem to be a good idea.

Makes sense.

Quoted 2 lines

Maybe even check that it's the same vma?

Hmm, I call on to Jens Remus and ask if s390 can do anything whacky here? Where something that isn't allowed on other architectures may be allowed there? I know s390 has some strange type of stack usage.

-- Steve

author maintainer
#9 lore raw

Re: [PATCH v12 02/14] unwind_user: Add frame pointer support

Quoted 5 lines · On Mon, 30 Jun 2025 22:56:03 -0400 Steven Rostedt <[email protected]> wrote:
Quoted 2 lines · On Mon, 30 Jun 2025 19:10:09 -0700 Linus Torvalds…
Quoted 1 line · Steven Rostedt
Quoted 7 lines
Quoted 5 lines

kernel/unwind/user.c line 39

Quoted 3 lines
kernel/unwind/user.c +3 -0
+       /* stack going in wrong direction? */
+       if (cfa <= state->sp)
+               goto done;    

I suspect this should do a lot more testing.

Sure.

Adding Kees too.

Kees,

I'd like to get some security eyes on this code to take a look at it. As it is making decisions on input from user space, I'd like to have more security folks looking at this to make sure that some malicious task can't set up its stack in such a way that it can exploit something here.

The parsing of the sframe code (latest version net yet posted) will need a similar audit.

Thanks,

-- Steve

Quoted 24 lines
Quoted 9 lines

kernel/unwind/user.c line 46

Quoted 8 lines
kernel/unwind/user.c +6 -0
+       /* Find the Return Address (RA) */
+       if (get_user(ra, (unsigned long *)(cfa + frame->ra_off)))
+               goto done;
+
+       if (frame->fp_off && get_user(fp, (unsigned long __user *)(cfa + frame->fp_off)))
+               goto done;    
Quoted 4 lines

.. and this should check the frame for validity too. At a minimum it should be properly aligned, but things like "it had better be above the current frame" to avoid having some loop would seem to be a good idea.

Makes sense.

Quoted 2 lines

Maybe even check that it's the same vma?

Hmm, I call on to Jens Remus and ask if s390 can do anything whacky here? Where something that isn't allowed on other architectures may be allowed there? I know s390 has some strange type of stack usage.

-- Steve

#10 lore raw

Re: [PATCH v12 02/14] unwind_user: Add frame pointer support

In reply to Steven Rostedt
Quoted 11 lines · Steven Rostedt
Quoted 7 lines · On Mon, 30 Jun 2025 19:10:09 -0700 Linus Torvalds…

kernel/unwind/user.c line 39

Quoted 4 lines · Steven Rostedt
kernel/unwind/user.c +3 -0
+       /* stack going in wrong direction? */
+       if (cfa <= state->sp)
+               goto done;  

I suspect this should do a lot more testing.

Sure.

The above assumes:

	curr_frame_sp = state->sp
	prev_frame_sp = cfa   // for arches that define CFA as SP at call site

On s390 the prev_frame_sp may be equal to curr_frame_sp for the topmost frame, as long as the topmost function did not allocate any stack. For instance when early in the prologue or when in a leaf function that does not require any stack space. My s390 sframe support patches would therefore currently change above check to:

	/* stack going in wrong direction? */
	if (sp <= state->sp - topmost)
		goto done;

I assume that should be the case for all architectures whose function call instruction does not modify the SP, unlike x86-64's CALL does.

Quoted 9 lines
Quoted 7 lines

kernel/unwind/user.c line 46

Quoted 2 lines
kernel/unwind/user.c +2 -0
+       if (frame->fp_off && get_user(fp, (unsigned long __user *)(cfa + frame->fp_off)))
+               goto done;  

.. and this should check the frame for validity too. At a minimum it should be properly aligned, but things like "it had better be above the current frame" to avoid having some loop would seem to be a good idea.

Makes sense.

On s390 the FP (register) value does not necessarily need to be above the SP value, as the s390x ELF ABI does only designate a "preferred" FP register, so that the FP register may be used for other purposes, when a FP is not required in a function.

So the output FP value cannot be checked on s390. But the input FP value could probably be checked as follows before computing the CFA:

	if (frame->use_fp && state->fp < state->sp)
		goto done;

	/* Get the Canonical Frame Address (CFA) */
	cfa = (frame->use_fp ? state->fp : state->sp) + frame->cfa_off;
Quoted 5 lines
Quoted 1 line

Maybe even check that it's the same vma?

Hmm, I call on to Jens Remus and ask if s390 can do anything whacky here? Where something that isn't allowed on other architectures may be allowed there? I know s390 has some strange type of stack usage.

On s390, if a FP is required for dynamic stack allocation, it is only initialized as late as possible, that is usually after static stack allocation. Therefore SP == FP is rather seldom the case.

Regards,
Jens

Signature · 9 lines
Jens Remus
Linux on Z Development (D3303)
+49-7031-16-1128 Office
[email protected]

IBM

IBM Deutschland Research & Development GmbH; Vorsitzender des Aufsichtsrats: Wolfgang Wendt; Geschäftsführung: David Faller; Sitz der Gesellschaft: Böblingen; Registergericht: Amtsgericht Stuttgart, HRB 243294
IBM Data Privacy Statement: https://www.ibm.com/privacy/
author maintainer
#11 lore raw

Re: [PATCH v12 02/14] unwind_user: Add frame pointer support

Quoted 9 lines · On Tue, 1 Jul 2025 17:36:55 +0200 Jens Remus <[email protected]> wrote:

On s390 the prev_frame_sp may be equal to curr_frame_sp for the topmost frame, as long as the topmost function did not allocate any stack. For instance when early in the prologue or when in a leaf function that does not require any stack space. My s390 sframe support patches would therefore currently change above check to:

	/* stack going in wrong direction? */
	if (sp <= state->sp - topmost)
		goto done;

How do you calculate "topmost" then?

Is it another field you add to "state"?

-- Steve

#12 lore raw

Re: [PATCH v12 02/14] unwind_user: Add frame pointer support

Quoted 9 lines · Steven Rostedt
Quoted 6 lines · On Tue, 1 Jul 2025 17:36:55 +0200 Jens Remus <[email protected]> wrote:

On s390 the prev_frame_sp may be equal to curr_frame_sp for the topmost frame, as long as the topmost function did not allocate any stack. For instance when early in the prologue or when in a leaf function that does not require any stack space. My s390 sframe support patches would therefore currently change above check to:

Quoted 7 lines
Quoted 3 lines
	/* stack going in wrong direction? */
	if (sp <= state->sp - topmost)
		goto done;

How do you calculate "topmost" then?

Is it another field you add to "state"?

Correct. It is a boolean set to true in unwind_user_start() and set to false in unwind_user_next() when updating the state.

I assume most architectures need above change, as their SP at function entry should be equal to the SP at call site (unlike x86-64 due to CALL).

s390 also needs this information to allow restoring of FP/RA saved in other registers (instead of on the stack) only for the topmost frame. For any other frame arbitrary register contents would not be available, as user unwind only unwinds SP, FP, and RA.

I would post my s390 sframe support patches as RFC once you have provided a merged sframe branch as discussed in:
https://lore.kernel.org/all/[email protected]/

Regards,
Jens

Signature · 9 lines
Jens Remus
Linux on Z Development (D3303)
+49-7031-16-1128 Office
[email protected]

IBM

IBM Deutschland Research & Development GmbH; Vorsitzender des Aufsichtsrats: Wolfgang Wendt; Geschäftsführung: David Faller; Sitz der Gesellschaft: Böblingen; Registergericht: Amtsgericht Stuttgart, HRB 243294
IBM Data Privacy Statement: https://www.ibm.com/privacy/
author maintainer
#13 lore raw

Re: [PATCH v12 02/14] unwind_user: Add frame pointer support

Quoted 10 lines · On Thu, 3 Jul 2025 18:21:10 +0200 Jens Remus <[email protected]> wrote:
Quoted 7 lines
Quoted 3 lines
	/* stack going in wrong direction? */
	if (sp <= state->sp - topmost)
		goto done;  

How do you calculate "topmost" then?

Is it another field you add to "state"?

Correct. It is a boolean set to true in unwind_user_start() and set to false in unwind_user_next() when updating the state.

So it's subtracting 1 or zero? So that the topmost can be equal. Well, that would need a bit of commenting.

Quoted 4 lines

I assume most architectures need above change, as their SP at function entry should be equal to the SP at call site (unlike x86-64 due to CALL).

Quoted 8 lines

s390 also needs this information to allow restoring of FP/RA saved in other registers (instead of on the stack) only for the topmost frame. For any other frame arbitrary register contents would not be available, as user unwind only unwinds SP, FP, and RA.

I would post my s390 sframe support patches as RFC once you have provided a merged sframe branch as discussed in:
https://lore.kernel.org/all/[email protected]/

I did have a merge branch on my repo. But I was hoping to see your code so that I can add this to this patch before having to post again. But now I'm posting without this change, as I don't want to screw it up. I think I know what it it looks like, but it would be better to see what you did to make sure what I envision is correct.

Oh well. I'll post v13 without it.

-- Steve

#14 lore raw

Re: [PATCH v12 02/14] unwind_user: Add frame pointer support

In reply to Linus Torvalds

* Linus Torvalds:

Quoted 15 lines · Steven Rostedt

kernel/unwind/user.c line 39

Quoted 4 lines
kernel/unwind/user.c +3 -0
+       /* stack going in wrong direction? */
+       if (cfa <= state->sp)
+               goto done;

I suspect this should do a lot more testing.

kernel/unwind/user.c line 46

Quoted 6 lines
kernel/unwind/user.c +6 -0
+       /* Find the Return Address (RA) */
+       if (get_user(ra, (unsigned long *)(cfa + frame->ra_off)))
+               goto done;
+
+       if (frame->fp_off && get_user(fp, (unsigned long __user *)(cfa + frame->fp_off)))
+               goto done;
Quoted 4 lines

.. and this should check the frame for validity too. At a minimum it should be properly aligned, but things like "it had better be above the current frame" to avoid having some loop would seem to be a good idea.

I don't think SFrame as-is requires stacks to be contiguous. Maybe there could be a per-frame flag that indicates whether a stack switch is expected?

Thanks,
Florian

author maintainer
#15 lore raw

Re: [PATCH v12 02/14] unwind_user: Add frame pointer support

Quoted 17 lines · On Tue, 01 Jul 2025 06:46:14 +0200 Florian Weimer <[email protected]> wrote:

* Linus Torvalds:

Quoted 15 lines · Steven Rostedt

kernel/unwind/user.c line 39

Quoted 4 lines
kernel/unwind/user.c +3 -0
+       /* stack going in wrong direction? */
+       if (cfa <= state->sp)
+               goto done;  

I suspect this should do a lot more testing.

Quoted 6 lines
kernel/unwind/user.c +6 -0
+       /* Find the Return Address (RA) */
+       if (get_user(ra, (unsigned long *)(cfa + frame->ra_off)))
+               goto done;
+
+       if (frame->fp_off && get_user(fp, (unsigned long __user *)(cfa + frame->fp_off)))
+               goto done;  
Quoted 8 lines
Quoted 4 lines

.. and this should check the frame for validity too. At a minimum it should be properly aligned, but things like "it had better be above the current frame" to avoid having some loop would seem to be a good idea.

I don't think SFrame as-is requires stacks to be contiguous. Maybe there could be a per-frame flag that indicates whether a stack switch is expected?

Looking at the current code of perf, it appears to only check that the address is valid to read from user space. Perhaps that's the only check needed here too?

Now this loop will not go into an infinite loop as the code has:

	for_each_user_frame(&state) {
		trace->entries[trace->nr++] = state.ip;
		if (trace->nr >= max_entries)
			break;
	}

Where

#define for_each_user_frame(state) \
	for (unwind_user_start((state)); !(state)->done; unwind_user_next((state)))

It will stop at "max_entries" even if the user space tries to make it go forever. max_entries is either 511 (on 64 bit) or 1023 (on 32 bit), as it is defined as:

/* Make the cache fit in a 4K page */
#define UNWIND_MAX_ENTRIES \
((SZ_4K - sizeof(struct unwind_cache)) / sizeof(long))

Now, perhaps we need to verify that the cfa is indeed canonical, but what other test do we have perform?

In the future, this code will also be handling JIT and possibly interpreted code. How much do we really want to limit the stack walking due to checks?

-- Steve

author maintainer
#16 lore raw

PATCH v12 3/14 unwind_user: Add compat mode frame pointer support

In reply to Steven Rostedt

From: Josh Poimboeuf <[email protected]>

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.

---
 arch/Kconfig                            |  4 ++++
 include/asm-generic/Kbuild              |  1 +
 include/asm-generic/unwind_user_types.h |  5 ++++
 include/linux/unwind_user.h             |  5 ++++
 include/linux/unwind_user_types.h       |  7 ++++++
 kernel/unwind/user.c                    | 32 +++++++++++++++++++++----
 6 files changed, 50 insertions(+), 4 deletions(-)
 create mode 100644 include/asm-generic/unwind_user_types.h
arch/Kconfig +4 -0
index 8e3fd723bd74..2c41d3072910 100644
@@ -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
include/asm-generic/Kbuild +1 -0
index 295c94a3ccc1..b797a2434396 100644
@@ -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
include/asm-generic/unwind_user_types.h new file +5 -0
new file mode 100644
index 000000000000..f568b82e52cd
@@ -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 */
include/linux/unwind_user.h +5 -0
index a405111c41b0..ac007363820a 100644
@@ -9,6 +9,11 @@
  #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
+
 int unwind_user_start(struct unwind_user_state *state);
 int unwind_user_next(struct unwind_user_state *state);
 
include/linux/unwind_user_types.h +7 -0
index 65bd070eb6b0..0b6563951ca4 100644
@@ -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;
 };
kernel/unwind/user.c +28 -4
index 1201d655654a..3a0ac4346f5b 100644
@@ -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_fp_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_fp_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_fp_state(state))
+		frame = &compat_fp_frame;
+	else if (fp_state(state))
 		frame = &fp_frame;
 	else
 		goto done;
@@ -39,10 +61,10 @@ int unwind_user_next(struct unwind_user_state *state)
 		goto done;
 
 	/* Find the Return Address (RA) */
-	if (get_user(ra, (unsigned long *)(cfa + frame->ra_off)))
+	if (unwind_get_user_long(ra, cfa + frame->ra_off, state))
 		goto done;
 
-	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 done;
 
 	state->ip = ra;
@@ -68,7 +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;
Signature · 1 line
2.47.2
author maintainer
#17 lore raw

PATCH v12 4/14 unwind_user/deferred: Add unwind_user_faultable()

In reply to Steven Rostedt

From: Steven Rostedt <[email protected]>

Add a new API to retrieve a user space callstack called unwind_user_faultable(). The difference between this user space stack tracer from the current user space stack tracer is that this must be called from faultable context as it may use routines to access user space data that needs to be faulted in.

It can be safely called from entering or exiting a system call as the code can still be faulted in there.

This code is based on work by Josh Poimboeuf's deferred unwinding code:

---
 include/linux/sched.h                 |  5 +++
 include/linux/unwind_deferred.h       | 24 +++++++++++
 include/linux/unwind_deferred_types.h |  9 ++++
 kernel/fork.c                         |  4 ++
 kernel/unwind/Makefile                |  2 +-
 kernel/unwind/deferred.c              | 60 +++++++++++++++++++++++++++
 6 files changed, 103 insertions(+), 1 deletion(-)
 create mode 100644 include/linux/unwind_deferred.h
 create mode 100644 include/linux/unwind_deferred_types.h
 create mode 100644 kernel/unwind/deferred.c
include/linux/sched.h +5 -0
index 4f78a64beb52..59fdf7d9bb1e 100644
@@ -46,6 +46,7 @@
 #include <linux/rv.h>
 #include <linux/uidgid_types.h>
 #include <linux/tracepoint-defs.h>
+#include <linux/unwind_deferred_types.h>
 #include <asm/kmap_size.h>
 
 /* task_struct member predeclarations (sorted alphabetically): */
@@ -1654,6 +1655,10 @@ struct task_struct {
 	struct user_event_mm		*user_event_mm;
 #endif
 
+#ifdef CONFIG_UNWIND_USER
+	struct unwind_task_info		unwind_info;
+#endif
+
 	/* CPU-specific state of this task: */
 	struct thread_struct		thread;
 
include/linux/unwind_deferred.h new file +24 -0
new file mode 100644
index 000000000000..a5f6e8f8a1a2
@@ -0,0 +1,24 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _LINUX_UNWIND_USER_DEFERRED_H
+#define _LINUX_UNWIND_USER_DEFERRED_H
+
+#include <linux/unwind_user.h>
+#include <linux/unwind_deferred_types.h>
+
+#ifdef CONFIG_UNWIND_USER
+
+void unwind_task_init(struct task_struct *task);
+void unwind_task_free(struct task_struct *task);
+
+int unwind_user_faultable(struct unwind_stacktrace *trace);
+
+#else /* !CONFIG_UNWIND_USER */
+
+static inline void unwind_task_init(struct task_struct *task) {}
+static inline void unwind_task_free(struct task_struct *task) {}
+
+static inline int unwind_user_faultable(struct unwind_stacktrace *trace) { return -ENOSYS; }
+
+#endif /* !CONFIG_UNWIND_USER */
+
+#endif /* _LINUX_UNWIND_USER_DEFERRED_H */
include/linux/unwind_deferred_types.h new file +9 -0
new file mode 100644
index 000000000000..aa32db574e43
@@ -0,0 +1,9 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _LINUX_UNWIND_USER_DEFERRED_TYPES_H
+#define _LINUX_UNWIND_USER_DEFERRED_TYPES_H
+
+struct unwind_task_info {
+	unsigned long		*entries;
+};
+
+#endif /* _LINUX_UNWIND_USER_DEFERRED_TYPES_H */
kernel/fork.c +4 -0
index 1ee8eb11f38b..3341d50c61f2 100644
@@ -105,6 +105,7 @@
 #include <uapi/linux/pidfd.h>
 #include <linux/pidfs.h>
 #include <linux/tick.h>
+#include <linux/unwind_deferred.h>
 
 #include <asm/pgalloc.h>
 #include <linux/uaccess.h>
@@ -732,6 +733,7 @@ void __put_task_struct(struct task_struct *tsk)
 	WARN_ON(refcount_read(&tsk->usage));
 	WARN_ON(tsk == current);
 
+	unwind_task_free(tsk);
 	sched_ext_free(tsk);
 	io_uring_free(tsk);
 	cgroup_free(tsk);
@@ -2135,6 +2137,8 @@ __latent_entropy struct task_struct *copy_process(
 	p->bpf_ctx = NULL;
 #endif
 
+	unwind_task_init(p);
+
 	/* Perform scheduler related setup. Assign this task to a CPU. */
 	retval = sched_fork(clone_flags, p);
 	if (retval)
kernel/unwind/Makefile +1 -1
index 349ce3677526..eae37bea54fd 100644
@@ -1 +1 @@
- obj-$(CONFIG_UNWIND_USER) += user.o
+ obj-$(CONFIG_UNWIND_USER)	+= user.o deferred.o
kernel/unwind/deferred.c new file +60 -0

60 lines not on this page · raw message

Signature · 1 line
2.47.2
author maintainer
#18 lore raw

PATCH v12 5/14 unwind_user/deferred: Add unwind cache

In reply to Steven Rostedt

From: Josh Poimboeuf <[email protected]>

Cache the results of the unwind to ensure the unwind is only performed once, even when called by multiple tracers.

The cache nr_entries gets cleared every time the task exits the kernel. When a stacktrace is requested, nr_entries gets set to the number of entries in the stacktrace. If another stacktrace is requested, if nr_entries is not zero, then it contains the same stacktrace that would be retrieved so it is not processed again and the entries is given to the caller.

---
 include/linux/entry-common.h          |  2 ++
 include/linux/unwind_deferred.h       |  8 +++++++
 include/linux/unwind_deferred_types.h |  7 +++++-
 kernel/unwind/deferred.c              | 31 +++++++++++++++++++++------
 4 files changed, 40 insertions(+), 8 deletions(-)
include/linux/entry-common.h +2 -0
index f94f3fdf15fc..8908b8eeb99b 100644
@@ -12,6 +12,7 @@
 #include <linux/resume_user_mode.h>
 #include <linux/tick.h>
 #include <linux/kmsan.h>
+#include <linux/unwind_deferred.h>
 
 #include <asm/entry-common.h>
 #include <asm/syscall.h>
@@ -362,6 +363,7 @@ static __always_inline void exit_to_user_mode(void)
 	lockdep_hardirqs_on_prepare();
 	instrumentation_end();
 
+	unwind_reset_info();
 	user_enter_irqoff();
 	arch_exit_to_user_mode();
 	lockdep_hardirqs_on(CALLER_ADDR0);
include/linux/unwind_deferred.h +8 -0
index a5f6e8f8a1a2..baacf4a1eb4c 100644
@@ -12,6 +12,12 @@ void unwind_task_free(struct task_struct *task);
 
 int unwind_user_faultable(struct unwind_stacktrace *trace);
 
+static __always_inline void unwind_reset_info(void)
+{
+	if (unlikely(current->unwind_info.cache))
+		current->unwind_info.cache->nr_entries = 0;
+}
+
 #else /* !CONFIG_UNWIND_USER */
 
 static inline void unwind_task_init(struct task_struct *task) {}
@@ -19,6 +25,8 @@ static inline void unwind_task_free(struct task_struct *task) {}
 
 static inline int unwind_user_faultable(struct unwind_stacktrace *trace) { return -ENOSYS; }
 
+static inline void unwind_reset_info(void) {}
+
 #endif /* !CONFIG_UNWIND_USER */
 
 #endif /* _LINUX_UNWIND_USER_DEFERRED_H */
include/linux/unwind_deferred_types.h +6 -1
index aa32db574e43..db5b54b18828 100644
@@ -2,8 +2,13 @@
 #ifndef _LINUX_UNWIND_USER_DEFERRED_TYPES_H
 #define _LINUX_UNWIND_USER_DEFERRED_TYPES_H
 
+struct unwind_cache {
+	unsigned int		nr_entries;
+	unsigned long		entries[];
+};
+
 struct unwind_task_info {
-	unsigned long		*entries;
+	struct unwind_cache	*cache;
 };
 
 #endif /* _LINUX_UNWIND_USER_DEFERRED_TYPES_H */
kernel/unwind/deferred.c +24 -7
index a0badbeb3cc1..96368a5aa522 100644
@@ -4,10 +4,13 @@
  */
 #include <linux/kernel.h>
 #include <linux/sched.h>
+#include <linux/sizes.h>
 #include <linux/slab.h>
 #include <linux/unwind_deferred.h>
 
-#define UNWIND_MAX_ENTRIES 512
+/* Make the cache fit in a 4K page */
+#define UNWIND_MAX_ENTRIES					\
+	((SZ_4K - sizeof(struct unwind_cache)) / sizeof(long))
 
 /**
  * unwind_user_faultable - Produce a user stacktrace in faultable context
@@ -24,6 +27,7 @@
 int unwind_user_faultable(struct unwind_stacktrace *trace)
 {
 	struct unwind_task_info *info = &current->unwind_info;
+	struct unwind_cache *cache;
 
 	/* Should always be called from faultable context */
 	might_fault();
@@ -31,17 +35,30 @@ int unwind_user_faultable(struct unwind_stacktrace *trace)
 	if (current->flags & PF_EXITING)
 		return -EINVAL;
 
-	if (!info->entries) {
-		info->entries = kmalloc_array(UNWIND_MAX_ENTRIES, sizeof(long),
-					      GFP_KERNEL);
-		if (!info->entries)
+	if (!info->cache) {
+		info->cache = kzalloc(struct_size(cache, entries, UNWIND_MAX_ENTRIES),
+				      GFP_KERNEL);
+		if (!info->cache)
 			return -ENOMEM;
 	}
 
+	cache = info->cache;
+	trace->entries = cache->entries;
+
+	if (cache->nr_entries) {
+		/*
+		 * The user stack has already been previously unwound in this
+		 * entry context.  Skip the unwind and use the cache.
+		 */
+		trace->nr = cache->nr_entries;
+		return 0;
+	}
+
 	trace->nr = 0;
-	trace->entries = info->entries;
 	unwind_user(trace, UNWIND_MAX_ENTRIES);
 
+	cache->nr_entries = trace->nr;
+
 	return 0;
 }
 
@@ -56,5 +73,5 @@ void unwind_task_free(struct task_struct *task)
 {
 	struct unwind_task_info *info = &task->unwind_info;
 
-	kfree(info->entries);
+	kfree(info->cache);
 }
Signature · 1 line
2.47.2
author maintainer
#19 lore raw

PATCH v12 6/14 unwind_user/deferred: Add deferred unwinding interface

In reply to Steven Rostedt

From: Josh Poimboeuf <[email protected]>

Add an interface for scheduling task work to unwind the user space stack before returning to user space. This solves several problems for its callers:

- Ensure the unwind happens in task context even if the caller may be running in interrupt context.

- Avoid duplicate unwinds, whether called multiple times by the same caller or by different callers.

- Take a timestamp when the first request comes in since the task entered the kernel. This will be returned to the calling function along with the stack trace when the task leaves the kernel. This timestamp can be used to correlate kernel unwinds/traces with the user unwind. For this to work properly, the architecture must have a local_clock() resolution that guarantees a different timestamp per a task systemcall.

The timestamp is created to detect when the stacktrace is the same. It is generated the first time a user space stacktrace is requested after the task enters the kernel.

The timestamp is passed to the caller on request, and when the stacktrace is generated upon returning to user space, it will call the requester's callback with the timestamp as well as the stacktrace. The timestamp is cleared when it goes back to user space. Note, this currently adds another conditional to the unwind_reset_info() path that is always called returning to user space, but future changes will put this back to a single conditional.

A global list is created and protected by a global mutex that holds tracers that register with the unwind infrastructure. The number of registered tracers will be limited in future changes. Each perf program or ftrace instance will register its own descriptor to use for deferred unwind stack traces.

Note, in the function unwind_deferred_task_work() that gets called when returning to user space, it uses a global mutex for synchronization which will cause a big bottleneck. This will be replaced by SRCU, but that change adds some complex synchronization that deservers its own commit.

---
Changes since v11: https://lore.kernel.org/[email protected]

- Still need to clear cache->nr_entries in unwind_reset_info even if timestamp is zero. This is because unwind_user_faultable() can be called directly and it requires nr_entries to be zeroed but it does not touch the timestamp.

 include/linux/unwind_deferred.h       |  24 +++++
 include/linux/unwind_deferred_types.h |   3 +
 kernel/unwind/deferred.c              | 139 +++++++++++++++++++++++++-
 3 files changed, 165 insertions(+), 1 deletion(-)
include/linux/unwind_deferred.h +24 -0
index baacf4a1eb4c..c6548e8d64d1 100644
@@ -2,9 +2,19 @@
 #ifndef _LINUX_UNWIND_USER_DEFERRED_H
 #define _LINUX_UNWIND_USER_DEFERRED_H
 
+#include <linux/task_work.h>
 #include <linux/unwind_user.h>
 #include <linux/unwind_deferred_types.h>
 
+struct unwind_work;
+
+typedef void (*unwind_callback_t)(struct unwind_work *work, struct unwind_stacktrace *trace, u64 timestamp);
+
+struct unwind_work {
+	struct list_head		list;
+	unwind_callback_t		func;
+};
+
 #ifdef CONFIG_UNWIND_USER
 
 void unwind_task_init(struct task_struct *task);
@@ -12,8 +22,19 @@ void unwind_task_free(struct task_struct *task);
 
 int unwind_user_faultable(struct unwind_stacktrace *trace);
 
+int unwind_deferred_init(struct unwind_work *work, unwind_callback_t func);
+int unwind_deferred_request(struct unwind_work *work, u64 *timestamp);
+void unwind_deferred_cancel(struct unwind_work *work);
+
 static __always_inline void unwind_reset_info(void)
 {
+	if (unlikely(current->unwind_info.timestamp))
+		current->unwind_info.timestamp = 0;
+	/*
+	 * As unwind_user_faultable() can be called directly and
+	 * depends on nr_entries being cleared on exit to user,
+	 * this needs to be a separate conditional.
+	 */
 	if (unlikely(current->unwind_info.cache))
 		current->unwind_info.cache->nr_entries = 0;
 }
@@ -24,6 +45,9 @@ static inline void unwind_task_init(struct task_struct *task) {}
 static inline void unwind_task_free(struct task_struct *task) {}
 
 static inline int unwind_user_faultable(struct unwind_stacktrace *trace) { return -ENOSYS; }
+static inline int unwind_deferred_init(struct unwind_work *work, unwind_callback_t func) { return -ENOSYS; }
+static inline int unwind_deferred_request(struct unwind_work *work, u64 *timestamp) { return -ENOSYS; }
+static inline void unwind_deferred_cancel(struct unwind_work *work) {}
 
 static inline void unwind_reset_info(void) {}
 
include/linux/unwind_deferred_types.h +3 -0
index db5b54b18828..5df264cf81ad 100644
@@ -9,6 +9,9 @@ struct unwind_cache {
 
 struct unwind_task_info {
 	struct unwind_cache	*cache;
+	struct callback_head	work;
+	u64			timestamp;
+	int			pending;
 };
 
 #endif /* _LINUX_UNWIND_USER_DEFERRED_TYPES_H */
kernel/unwind/deferred.c +138 -1

169 lines not on this page · raw message

Signature · 1 line
2.47.2
maintainer
#20 lore raw

Re: [PATCH v12 06/14] unwind_user/deferred: Add deferred unwinding interface

Quoted 8 lines · Steven Rostedt
kernel/unwind/deferred.c +8 -0
+/*
+ * Read the task context timestamp, if this is the first caller then
+ * it will set the timestamp.
+ *
+ * For this to work properly, the timestamp (local_clock()) must
+ * have a resolution that will guarantee a different timestamp
+ * everytime a task makes a system call. That is, two short
+ * system calls back to back must have a different timestamp.

kernel/unwind/deferred.c line 40

Quoted 10 lines
kernel/unwind/deferred.c +10 -0
+ */
+static u64 get_timestamp(struct unwind_task_info *info)
+{
+	lockdep_assert_irqs_disabled();
+
+	if (!info->timestamp)
+		info->timestamp = local_clock();
+
+	return info->timestamp;
+}

I'm very hesitant about this. Modern hardware can do this, but older hardware (think Intel Core and AMD Bulldozer etc hardware) might struggle with this. They don't have stable TSC and as such will use the magic in kernel/sched/clock.c; which can get stuck on a window edge for a little bit and re-use timestamps.

author maintainer
#21 lore raw

Re: [PATCH v12 06/14] unwind_user/deferred: Add deferred unwinding interface

Quoted 7 lines · On Wed, 2 Jul 2025 18:36:09 +0200 Peter Zijlstra <[email protected]> wrote:
Quoted 7 lines
kernel/unwind/deferred.c +7 -0
+static u64 get_timestamp(struct unwind_task_info *info)
+{
+	lockdep_assert_irqs_disabled();
+
+	if (!info->timestamp)
+		info->timestamp = local_clock();
+
Quoted 8 lines

kernel/unwind/deferred.c line 40

Quoted 2 lines
kernel/unwind/deferred.c +2 -0
+	return info->timestamp;
+}  

I'm very hesitant about this. Modern hardware can do this, but older hardware (think Intel Core and AMD Bulldozer etc hardware) might struggle with this. They don't have stable TSC and as such will use the magic in kernel/sched/clock.c; which can get stuck on a window edge for a little bit and re-use timestamps.

Well, the idea of using timestamps came from Microsoft as that's what they do for a similar feature. It just seemed like an easier approach. But I could definitely go back to the "cookie" idea that is just a per cpu counter (with the CPU number as part of the cookie).

As the timestamp is likely not going to be as useful as it is with Microsoft as there's no guarantee that the timestamp counter used is the same as the timestamp used by the tracer asking for this, the cookie approach may indeed be better.

-- Steve

#22 lore raw

Re: [PATCH v12 06/14] unwind_user/deferred: Add deferred unwinding interface

Quoted 5 lines · Steven Rostedt

As the timestamp is likely not going to be as useful as it is with Microsoft as there's no guarantee that the timestamp counter used is the same as the timestamp used by the tracer asking for this, the cookie approach may indeed be better.

I think having just a percpu counter is probably the safest thing to do, since the main reason just seems to be "correlate with the user event". Using some kind of "real time" for correlation purposes seems like a bad idea from any portability standpoint, considering just how many broken timers we've seen across pretty much every architecture out there.

Also, does it actually have to be entirely unique? IOW, a 32-bit counter (or even less) might be sufficient if there's some guarantee that processing happens before the counter wraps around? Again - for correlation purposes, just *how* many outstanding events can you have that aren't ordered by other things too?

I'm sure people want to also get some kind of rough time idea, but don't most perf events have them simply because people want time information for _informatioal_ reasons, rather than to correlate two events?

Linus

author maintainer
#23 lore raw

Re: [PATCH v12 06/14] unwind_user/deferred: Add deferred unwinding interface

Quoted 10 lines · On Wed, 2 Jul 2025 09:56:39 -0700 Linus Torvalds…

Also, does it actually have to be entirely unique? IOW, a 32-bit counter (or even less) might be sufficient if there's some guarantee that processing happens before the counter wraps around? Again - for correlation purposes, just *how* many outstanding events can you have that aren't ordered by other things too?

I'm sure people want to also get some kind of rough time idea, but don't most perf events have them simply because people want time information for _informatioal_ reasons, rather than to correlate two events?

And it only needs to be unique per thread per system call. The real reason for this identifier is for lost events. As I explained in the perf patchset, the issues is this:

In case of dropped events, we could have the case of:

  system_call() {
    <nmi> {
      take kernel stack trace
      ask for deferred trace.
   [EVENTS START DROPPING HERE]
    }
    Call deferred callback to record trace [ BUT IS DROPPED ]
  }
  system_call() {
    <nmi> {
      take kernel stack trace
      ask for deferred trace [ STILL DROPPING ]
    }
    [ READER CATCHES UP AND STARTS READING EVENTS AGAIN]

Call deferred callback to record trace
}

The user space tool will see that kernel stack traces of the first system call, then it will see events dropped, and then it will see the deferred user space stack trace of the second call.

The identifier is only there for uniqueness for that one thread to let the tracer know if the deferred trace can be tied to events before it lost them.

We figured a single 32 bit counter would be good enough when we first discussed this idea, but we wanted per cpu counters to not have cache contention every time a CPU wanted to increment the counter. But each CPU would need an identifier so that a task migrating will not get the same identifier for a different system call just because it migrated.

We used 16 bits for the CPU counter thinking that 32K of CPUs would last some time in the future. We then chose to use a 64 bit number to allow us to have 48 bits left for uniqueness which is plenty.

If we use 32 bits, that would give us 32K of unique systemcalls, and it does seem possible that on a busy system, a tracer could lose 32K of system calls before it gets going again. But we could still use it anyway as the likelihood of losing exactly 32K of system calls and starting tracing back up again will probably never happen. And if it does, the worse thing that it will do is have the tracer mistake which user space stack trace goes to which event. If your are tracing that many events, this will likely be in the noise.

So I'm fine with making this a 32 bit counter using 16 bits for the CPU and 16 bits for per thread uniqueness.

-- Steve

author maintainer
#24 lore raw

Re: [PATCH v12 06/14] unwind_user/deferred: Add deferred unwinding interface

Quoted 2 lines · On Wed, 2 Jul 2025 13:26:05 -0400 Steven Rostedt <[email protected]> wrote:

So I'm fine with making this a 32 bit counter using 16 bits for the CPU and 16 bits for per thread uniqueness.

To still be able to use a 32 bit cmpxchg (for racing with an NMI), we could break this number up into two 32 bit words. One with the CPU that it was created on, and one with the per_cpu counter:

union unwind_task_id {
	struct {
		u32		cpu;
		u32		cnt;
	}
	u64 id;
};

static DEFINE_PER_CPU(u32, unwind_ctx_ctr);

static u64 get_cookie(struct unwind_task_info *info)

{
	u32 cpu_cnt;
	u32 cnt;
	u32 old = 0;

	if (info->id.cpu)
		return info->id.id;

	cpu_cnt = __this_cpu_read(unwind_ctx_ctr);
	cpu_cnt += 2;
	cnt = cpu_cnt | 1; /* Always make non zero */

	if (try_cmpxchg(&info->id.cnt, &old, cnt)) {
		/* Update the per cpu counter */
		__this_cpu_write(unwind_ctx_ctr, cpu_cnt);
	}
	/* Interrupts are disabled, the CPU will always be same */
	info->id.cpu = smp_processor_id() + 1; /* Must be non zero */

	return info->id.id;
}

When leaving the kernel it does:

	info->id.id = 0;

-- Steve

#25 lore raw

Re: [PATCH v12 06/14] unwind_user/deferred: Add deferred unwinding interface

Quoted 4 lines · Steven Rostedt

To still be able to use a 32 bit cmpxchg (for racing with an NMI), we could break this number up into two 32 bit words. One with the CPU that it was created on, and one with the per_cpu counter:

Do you actually even need a cpu number at all?

If this is per-thread, maybe just a per-thread counter would be good? And you already *have* that per-thread thing, in 'current->unwind_info'.

And the work is using task_work, so the worker callback is also per-thread.

Also, is racing with NMI even a thing for the sequence number? I would expect that the only thing that NMI would race with is the 'pending' field, not the sequence number.

IOW, I think the logic could be

- check 'pending' non-atomically, just because it's cheap

- do a try_cmpxchg() on pending to actually deal with nmi races

Actually, there are no SMP issues, just instruction atomicity - so a 'local_try_cmpxchg() would be sufficient, but that's a 'long' not a 'u32' ;^(

- now you are exclusive for that thread, you're done, no more need
for any atomic counter or percpu things

And then the next step is to just say "pending is the low bit of the id word" and having a separate 31-bit counter that gets incremented by "get_cookie()".

So then you end up with something like

// New name for 'get_timestamp()' get_current_cookie() { return current->unwind_info.cookie; } // New name for 'get_cookie()':
// 31-bit counter by just leaving bit #0 alone get_new_cookie() { current->unwind_info.cookie += 2; }

and then unwind_deferred_request() would do something like

  unwind_deferred_request()
  {
        int old, new;
        if (current->unwind_info.id)
                return 1;

        guard(irqsave)();
        // For NMI, if we race with 'get_new_cookie()'
        // we don't care if we get the old or the new one
        old = 0; new = get_current_cookie() | 1;
        if (!try_cmpxchg(&current->unwind_info.id, &old, new))
                return 1;
        .. now schedule the thing with that cookie set.

Hmm?

But I didn't actually look at the users, so maybe I'm missing some reason why you want to have a separate per-cpu value.

Or maybe I missed something else entirely, and the above is complete garbage and the ramblings of a insane mind.

It happens.

Off to get more coffee.

Linus

#26 lore raw

Re: [PATCH v12 06/14] unwind_user/deferred: Add deferred unwinding interface

Quoted 11 lines · Linus Torvalds
Quoted 4 lines · Steven Rostedt

To still be able to use a 32 bit cmpxchg (for racing with an NMI), we could break this number up into two 32 bit words. One with the CPU that it was created on, and one with the per_cpu counter:

Do you actually even need a cpu number at all?

If this is per-thread, maybe just a per-thread counter would be good? And you already *have* that per-thread thing, in 'current->unwind_info'.

AFAIR, one of the goals here is to save the cookie into the trace to allow trace post-processing to link the event triggering the unwinding with the deferred unwinding data.

In order to make the trace analysis results reliable, we'd like to avoid the following causes of uncertainty, which would mistakenly cause the post-processing analysis to associate a stack trace with the wrong event:

- Thread ID re-use (exit + clone/fork),
- Thread migration,
- Events discarded (e.g. buffer full) causing missing thread lifetime events or missing unwind-related events.

Unless I'm missing something, the per-thread counter would have issues with thread ID re-use during the trace lifetime.

One possibility to solve this would be to introduce a thread identifier (e.g. 64-bit thread ID value) which is unique across the entire kernel lifetime. This approach would actually be useful for other use-cases as well, but a 64-bit ID is not as compact as the CPU number, so it is somewhat wasteful in terms of trace bandwidth.

Hence the alternative we came up with, which is to combine the CPU number and a per-CPU counter to have a cheap way to keep track of a globally unique counter using per-CPU partitioning.

Thanks,

Mathieu

Signature · 3 lines
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com
author maintainer
#27 lore raw

Re: [PATCH v12 06/14] unwind_user/deferred: Add deferred unwinding interface

Quoted 10 lines · On Wed, 2 Jul 2025 14:47:10 -0400 Mathieu Desnoyers…

AFAIR, one of the goals here is to save the cookie into the trace to allow trace post-processing to link the event triggering the unwinding with the deferred unwinding data.

In order to make the trace analysis results reliable, we'd like to avoid the following causes of uncertainty, which would mistakenly cause the post-processing analysis to associate a stack trace with the wrong event:

Quoted 7 lines

- Thread ID re-use (exit + clone/fork),
- Thread migration,
- Events discarded (e.g. buffer full) causing missing thread lifetime events or missing unwind-related events.

Unless I'm missing something, the per-thread counter would have issues with thread ID re-use during the trace lifetime.

But you are missing one more thing that the trace can use, and that's the time sequence. As soon as the same thread has a new id you can assume all the older user space traces are not applicable for any new events for that thread, or any other thread with the same thread ID.

Thus the only issue that can truly be a problem is if you have missed events where thread id wraps around. I guess that could be possible if a long running task finally exits and it's thread id is reused immediately. Is that a common occurrence?

-- Steve.

#28 lore raw

Re: [PATCH v12 06/14] unwind_user/deferred: Add deferred unwinding interface

Quoted 17 lines · Steven Rostedt
Quoted 15 lines · On Wed, 2 Jul 2025 14:47:10 -0400 Mathieu Desnoyers…

AFAIR, one of the goals here is to save the cookie into the trace to allow trace post-processing to link the event triggering the unwinding with the deferred unwinding data.

In order to make the trace analysis results reliable, we'd like to avoid the following causes of uncertainty, which would mistakenly cause the post-processing analysis to associate a stack trace with the wrong event:

- Thread ID re-use (exit + clone/fork),
- Thread migration,
- Events discarded (e.g. buffer full) causing missing thread lifetime events or missing unwind-related events.

Quoted 7 lines
Quoted 2 lines

Unless I'm missing something, the per-thread counter would have issues with thread ID re-use during the trace lifetime.

But you are missing one more thing that the trace can use, and that's the time sequence. As soon as the same thread has a new id you can assume all the older user space traces are not applicable for any new events for that thread, or any other thread with the same thread ID.

In order for the scheme you describe to work, you need:

- instrumentation of task lifetime (exit/fork+clone),
- be sure that the events related to that instrumentation were not dropped.

I'm not sure about ftrace, but in LTTng enabling instrumentation of task lifetime is entirely up to the user.

And even if it's enabled, events can be discarded (e.g. buffer full).

Quoted 5 lines

Thus the only issue that can truly be a problem is if you have missed events where thread id wraps around. I guess that could be possible if a long running task finally exits and it's thread id is reused immediately. Is that a common occurrence?

You just need a combination of thread ID re-use and either no instrumentation of task lifetime or events discarded to trigger this. Even if it's not so frequent, at large scale and in production, I suspect that this will happen quite often.

You don't even need the task IDs to be re-used very quickly for this to be an issue.

Thanks,

Mathieu

Quoted 3 lines

-- Steve.

Signature · 3 lines
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com
author maintainer
#29 lore raw

Re: [PATCH v12 06/14] unwind_user/deferred: Add deferred unwinding interface

Quoted 5 lines · On Wed, 2 Jul 2025 15:12:45 -0400 Mathieu Desnoyers…
Quoted 4 lines

But you are missing one more thing that the trace can use, and that's the time sequence. As soon as the same thread has a new id you can assume all the older user space traces are not applicable for any new events for that thread, or any other thread with the same thread ID.

Quoted 8 lines

In order for the scheme you describe to work, you need:

- instrumentation of task lifetime (exit/fork+clone),
- be sure that the events related to that instrumentation were not dropped.

I'm not sure about ftrace, but in LTTng enabling instrumentation of task lifetime is entirely up to the user.

Has nothing to do with task lifetime. If you see a deferred request with id of 1 from task 8888, and then later you see either a deferred request or a stack trace with an id other than 1 for task 8888, you can then say all events before now are no longer eligible for new deferred stack traces.

Quoted 2 lines

And even if it's enabled, events can be discarded (e.g. buffer full).

The only case is if you see a deferred request with id 1 for task 8888, then you start dropping all events and that task 8888 exits and a new one appears with task id 8888 where it too has a deferred request with id 1 then you start picking up events again and see a deferred stack trace for the new task 8888 where it's id is 1, you lose.

But other than that exact scenario, it should not get confused.

Quoted 9 lines
Quoted 6 lines

Thus the only issue that can truly be a problem is if you have missed events where thread id wraps around. I guess that could be possible if a long running task finally exits and it's thread id is reused immediately. Is that a common occurrence?

You just need a combination of thread ID re-use and either no instrumentation of task lifetime or events discarded to trigger this.

Again, it's seeing a new request with another id for the same task, you don't need to worry about it. You don't even need to look at fork and exit events.

Quoted 2 lines

Even if it's not so frequent, at large scale and in production, I suspect that this will happen quite often.

Really? As I explained above?

-- Steve

author maintainer
#30 lore raw

Re: [PATCH v12 06/14] unwind_user/deferred: Add deferred unwinding interface

Quoted 5 lines · On Wed, 2 Jul 2025 15:21:11 -0400 Steven Rostedt <[email protected]> wrote:

The only case is if you see a deferred request with id 1 for task 8888, then you start dropping all events and that task 8888 exits and a new one appears with task id 8888 where it too has a deferred request with id 1 then you start picking up events again and see a deferred stack trace for the new task 8888 where it's id is 1, you lose.

And if we want to fix that, we could make the cookie 64 bit again, and set the timestamp on the first time it is used for the trace.

union unwind_task_id {
	struct {
		u32		task_id;
		u32		cnt;
	}
	u64 id;
};

static u64 get_cookie(struct unwind_task_info *info)

{
	u32 cnt = READ_ONCE(info->id.cnt);
	u32 new_cnt;

	if (cnt & 1)
		return info->id;

	if (unlikely(!info->id.task_id)) {
		u32 task_id = local_clock();

		cnt = 0;
		if (try_cmpxchg(&info->id.task_id, &cnt, task_id))
			task_id = cnt;
	}

	new_cnt = cnt + 3;
	if (try_cmpxchg(&info->id, &cnt, new_cnt))
		new_cnt = cnt; // try_cmpxchg() expects something

	return info->id;
}

So now each task will have its own id and even if we have a task wrap around, the cookie will never be the same, as fork sets the info->id to zero.

Yes, the local_clock() can wrap around, but now making all those the same to cause an issue is extremely unlikely, and still, if it happens, the worse thing that it causes is that the user space stack trace will be associated to the wrong events.

-- Steve

author maintainer
#31 lore raw

Re: [PATCH v12 06/14] unwind_user/deferred: Add deferred unwinding interface

Quoted 22 lines · On Wed, 2 Jul 2025 15:36:00 -0400 Steven Rostedt <[email protected]> wrote:
union unwind_task_id {
	struct {
		u32		task_id;
		u32		cnt;
	}
	u64 id;
};

static u64 get_cookie(struct unwind_task_info *info)

{
	u32 cnt = READ_ONCE(info->id.cnt);
	u32 new_cnt;

	if (cnt & 1)
		return info->id;

	if (unlikely(!info->id.task_id)) {
		u32 task_id = local_clock();

		cnt = 0;
		if (try_cmpxchg(&info->id.task_id, &cnt, task_id))
			task_id = cnt;
Quoted 8 lines
	}

	new_cnt = cnt + 3;
	if (try_cmpxchg(&info->id, &cnt, new_cnt))
		new_cnt = cnt; // try_cmpxchg() expects something

	return info->id;
}

Honestly I think this is way overkill. What I would do, is to have the cookie saved in the event be 64 bit, but we can start with the simple 32 bit solution keeping the top 32 bits zeros. If this does indeed become an issue in the future, we could fix it with a 64 bit number. By making sure all the exposed "cookies" are 64 bit, it should not break anything. The cookie is just supposed to be a random unique number that associates a request with its deferred user space stack trace.

With any exposed cookies to user space being 64 bits, this should not be an issue to address in the future.

-- Steve

#32 lore raw

Re: [PATCH v12 06/14] unwind_user/deferred: Add deferred unwinding interface

Quoted 43 lines · Steven Rostedt
Quoted 22 lines · On Wed, 2 Jul 2025 15:36:00 -0400 Steven Rostedt <[email protected]> wrote:
union unwind_task_id {
	struct {
		u32		task_id;
		u32		cnt;
	}
	u64 id;
};

static u64 get_cookie(struct unwind_task_info *info)

{
	u32 cnt = READ_ONCE(info->id.cnt);
	u32 new_cnt;

	if (cnt & 1)
		return info->id;

	if (unlikely(!info->id.task_id)) {
		u32 task_id = local_clock();

		cnt = 0;
		if (try_cmpxchg(&info->id.task_id, &cnt, task_id))
			task_id = cnt;
Quoted 8 lines
	}

	new_cnt = cnt + 3;
	if (try_cmpxchg(&info->id, &cnt, new_cnt))
		new_cnt = cnt; // try_cmpxchg() expects something

	return info->id;
}

Honestly I think this is way overkill. What I would do, is to have the cookie saved in the event be 64 bit, but we can start with the simple 32 bit solution keeping the top 32 bits zeros. If this does indeed become an issue in the future, we could fix it with a 64 bit number. By making sure all the exposed "cookies" are 64 bit, it should not break anything. The cookie is just supposed to be a random unique number that associates a request with its deferred user space stack trace.

Quoted 2 lines

With any exposed cookies to user space being 64 bits, this should not be an issue to address in the future.

FWIW, I liked your idea of making the cookie 64-bit with:

- 32-bit cpu number,
- 32-bit per-CPU free running counter.

This is simple, works on 32-bit archs, and it does not overflow as often as time LSB because it counts execution contexts.

Thanks,

Mathieu

Signature · 3 lines
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com
author maintainer
#33 lore raw

Re: [PATCH v12 06/14] unwind_user/deferred: Add deferred unwinding interface

Quoted 7 lines · On Wed, 2 Jul 2025 15:48:16 -0400 Mathieu Desnoyers…

FWIW, I liked your idea of making the cookie 64-bit with:

- 32-bit cpu number,
- 32-bit per-CPU free running counter.

This is simple, works on 32-bit archs, and it does not overflow as often as time LSB because it counts execution contexts.

I have no problem with implementing that. But Linus had his concerns. If he's OK with that version, then I'll go with that.

-- Steve

#34 lore raw

Re: [PATCH v12 06/14] unwind_user/deferred: Add deferred unwinding interface

In reply to Steven Rostedt
Quoted 26 lines · Steven Rostedt
Quoted 5 lines · On Wed, 2 Jul 2025 15:12:45 -0400 Mathieu Desnoyers…
Quoted 4 lines

But you are missing one more thing that the trace can use, and that's the time sequence. As soon as the same thread has a new id you can assume all the older user space traces are not applicable for any new events for that thread, or any other thread with the same thread ID.

Quoted 8 lines

In order for the scheme you describe to work, you need:

- instrumentation of task lifetime (exit/fork+clone),
- be sure that the events related to that instrumentation were not dropped.

I'm not sure about ftrace, but in LTTng enabling instrumentation of task lifetime is entirely up to the user.

Has nothing to do with task lifetime. If you see a deferred request with id of 1 from task 8888, and then later you see either a deferred request or a stack trace with an id other than 1 for task 8888, you can then say all events before now are no longer eligible for new deferred stack traces.

Quoted 2 lines

And even if it's enabled, events can be discarded (e.g. buffer full).

Quoted 7 lines

The only case is if you see a deferred request with id 1 for task 8888, then you start dropping all events and that task 8888 exits and a new one appears with task id 8888 where it too has a deferred request with id 1 then you start picking up events again and see a deferred stack trace for the new task 8888 where it's id is 1, you lose.

But other than that exact scenario, it should not get confused.

Correct.

Quoted 8 lines
Quoted 8 lines
Quoted 7 lines

Thus the only issue that can truly be a problem is if you have missed events where thread id wraps around. I guess that could be possible if a long running task finally exits and it's thread id is reused immediately. Is that a common occurrence?

Quoted 6 lines
Quoted 2 lines

You just need a combination of thread ID re-use and either no instrumentation of task lifetime or events discarded to trigger this.

Again, it's seeing a new request with another id for the same task, you don't need to worry about it. You don't even need to look at fork and exit events.

The reason why instrumentation of exit/{fork,clone} is useful is to know when a thread ID is re-used.

Quoted 5 lines
Quoted 3 lines

Even if it's not so frequent, at large scale and in production, I suspect that this will happen quite often.

Really? As I explained above?

Note that all newly forked threads will likely start counting near 0. So chances are that for short-lived threads most of the counter values will be in a low range.

So all you need is thread ID re-use for two threads which happen to use the deferred cookies within low-value ranges to hit this.

From my perspective, making trace analysis results reliable is the most basic guarantee tooling should provide in order to make it trusted by users. So I am tempted to err towards robustness rather than take shortcuts because "it does not happen often".

Thanks,

Mathieu

Signature · 3 lines
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com
author maintainer
#35 lore raw

Re: [PATCH v12 06/14] unwind_user/deferred: Add deferred unwinding interface

Quoted 4 lines · On Wed, 2 Jul 2025 15:43:08 -0400 Mathieu Desnoyers…

From my perspective, making trace analysis results reliable is the most
basic guarantee tooling should provide in order to make it trusted by users. So I am tempted to err towards robustness rather than take shortcuts because "it does not happen often".

Another solution which I'm thinking of doing for perf is to simply
state: a deferred stack trace does not go to any event before events were dropped.

That is, even if the stack trace is associated, if events are dropped before getting out of the kernel, just say, "Sorry, the events before the dropped events lost its user stack trace".

That may better, as having no data is better than incorrect data.

-- Steve

author maintainer
#36 lore raw

Re: [PATCH v12 06/14] unwind_user/deferred: Add deferred unwinding interface

In reply to Linus Torvalds
Quoted 11 lines · On Wed, 2 Jul 2025 11:21:34 -0700 Linus Torvalds…
Quoted 4 lines · Steven Rostedt

To still be able to use a 32 bit cmpxchg (for racing with an NMI), we could break this number up into two 32 bit words. One with the CPU that it was created on, and one with the per_cpu counter:

Do you actually even need a cpu number at all?

If this is per-thread, maybe just a per-thread counter would be good? And you already *have* that per-thread thing, in 'current->unwind_info'.

I just hate to increase the task struct even more. I'm trying to keep only the data I can't put elsewhere in the task struct.

Quoted 6 lines

And the work is using task_work, so the worker callback is also per-thread.

Also, is racing with NMI even a thing for the sequence number? I would expect that the only thing that NMI would race with is the 'pending' field, not the sequence number.

The sequence number is updated on the first time it's called after a task enters the kernel. Then it should return the same number every time after that. That's because this unique number is the identifier for the user space stack trace which doesn't change while the task is in the kernel, hence the id getting updated every time the task enters the kernel and not after that.

Quoted 4 lines

IOW, I think the logic could be

- check 'pending' non-atomically, just because it's cheap

Note, later patches move the "pending" bit into a mask that is used to figure out what callbacks to call.

Quoted 6 lines

- do a try_cmpxchg() on pending to actually deal with nmi races

Actually, there are no SMP issues, just instruction atomicity - so a 'local_try_cmpxchg() would be sufficient, but that's a 'long' not a 'u32' ;^(

Yeah, later patches do try to use more local_try_cmpxchg() at different parts. Even for the timestamp.

Quoted 3 lines

- now you are exclusive for that thread, you're done, no more need
for any atomic counter or percpu things

The trick and race with NMIs is, this needs to return that cookie for both callers, where the cookie is the same number.

Quoted 30 lines

And then the next step is to just say "pending is the low bit of the id word" and having a separate 31-bit counter that gets incremented by "get_cookie()".

So then you end up with something like

// New name for 'get_timestamp()' get_current_cookie() { return current->unwind_info.cookie; } // New name for 'get_cookie()':
// 31-bit counter by just leaving bit #0 alone get_new_cookie() { current->unwind_info.cookie += 2; }

and then unwind_deferred_request() would do something like

  unwind_deferred_request()
  {
        int old, new;
        if (current->unwind_info.id)
                return 1;

        guard(irqsave)();
        // For NMI, if we race with 'get_new_cookie()'
        // we don't care if we get the old or the new one
        old = 0; new = get_current_cookie() | 1;
        if (!try_cmpxchg(&current->unwind_info.id, &old, new))
                return 1;
        .. now schedule the thing with that cookie set.
Quoted 4 lines

Hmm?

But I didn't actually look at the users, so maybe I'm missing some reason why you want to have a separate per-cpu value.

Now we could just have the counter be the 32 bit cookie (old timestamp) in the task struct, and have bit 0 be if it is valued or not.

static u32 get_cookie(struct unwind_task_info *info)

{
	u32 cnt = READ_ONCE(info->id);
	u32 new_cnt;

	if (cnt & 1)
		return cnt;

	new_cnt = cnt + 3;
	if (try_cmpxchg(&info->id, &cnt, new_cnt))
		return new_cnt;

	return cnt; // return info->id; would work too
}

Then when going back to user space:

	info->id &= ~1;

Now the counter is stored in the task struct and no other info is needed.

Quoted 7 lines

Or maybe I missed something else entirely, and the above is complete garbage and the ramblings of a insane mind.

It happens.

Off to get more coffee.

Enjoy, I just grabbed my last cup of the day.

-- Steve

author maintainer
#37 lore raw

PATCH v12 7/14 unwind_user/deferred: Make unwind deferral requests NMI-safe

In reply to Steven Rostedt

From: Steven Rostedt <[email protected]>

Make unwind_deferred_request() NMI-safe so tracers in NMI context can call it and safely request a user space stacktrace when the task exits.

Note, this is only allowed for architectures that implement a safe 64 bit cmpxchg. Which rules out some 32bit architectures and even some 64 bit ones. If an architecture requests a deferred stack trace from NMI context that does not support a safe NMI 64 bit cmpxchg, it will get an -EINVAL. For those architectures, they would need another method (perhaps an irqwork), to request a deferred user space stack trace. That can be dealt with later if one of theses architectures require this feature.

---
 include/linux/unwind_deferred.h       |  4 +-
 include/linux/unwind_deferred_types.h |  7 ++-
 kernel/unwind/deferred.c              | 74 ++++++++++++++++++++++-----
 3 files changed, 69 insertions(+), 16 deletions(-)
include/linux/unwind_deferred.h +2 -2
index c6548e8d64d1..73f6cac53530 100644
@@ -28,8 +28,8 @@ void unwind_deferred_cancel(struct unwind_work *work);
 
 static __always_inline void unwind_reset_info(void)
 {
-	if (unlikely(current->unwind_info.timestamp))
-		current->unwind_info.timestamp = 0;
+	if (unlikely(local64_read(&current->unwind_info.timestamp)))
+		local64_set(&current->unwind_info.timestamp, 0);
 	/*
 	 * As unwind_user_faultable() can be called directly and
 	 * depends on nr_entries being cleared on exit to user,
include/linux/unwind_deferred_types.h +5 -2
index 5df264cf81ad..0d722e877473 100644
@@ -2,6 +2,9 @@
 #ifndef _LINUX_UNWIND_USER_DEFERRED_TYPES_H
 #define _LINUX_UNWIND_USER_DEFERRED_TYPES_H
 
+#include <asm/local64.h>
+#include <asm/local.h>
+
 struct unwind_cache {
 	unsigned int		nr_entries;
 	unsigned long		entries[];
@@ -10,8 +13,8 @@ struct unwind_cache {
 struct unwind_task_info {
 	struct unwind_cache	*cache;
 	struct callback_head	work;
-	u64			timestamp;
-	int			pending;
+	local64_t		timestamp;
+	local_t			pending;
 };
 
 #endif /* _LINUX_UNWIND_USER_DEFERRED_TYPES_H */
kernel/unwind/deferred.c +62 -12

129 lines not on this page · raw message

Signature · 1 line
2.47.2
#38 lore raw

Re: [PATCH v12 07/14] unwind_user/deferred: Make unwind deferral requests NMI-safe

Hello Steve!

include/linux/unwind_deferred_types.h line 6

Quoted 11 lines · Steven Rostedt

Make unwind_deferred_request() NMI-safe so tracers in NMI context can call it and safely request a user space stacktrace when the task exits.

include/linux/unwind_deferred_types.h +2 -0
@@ -2,6 +2,9 @@
 #ifndef _LINUX_UNWIND_USER_DEFERRED_TYPES_H
 #define _LINUX_UNWIND_USER_DEFERRED_TYPES_H
 
+#include <asm/local64.h>
+#include <asm/local.h>

This creates the following circular dependency, that breaks the build on s390 as follows, whenever local64.h is included first, so that local64_t is not yet defined when unwind_deferred_types.h gets included down the line:

Circular dependency:

include/linux/unwind_deferred_types.h
arch/<arch>/include/generated/asm/local64.h
include/asm-generic/local64.h
include/linux/percpu.h
include/linux/sched.h
include/linux/unwind_deferred_types.h

Compile error on s390:

CC net/ipv4/proc.o
In file included from ./include/linux/sched.h:49,

                 from ./include/linux/percpu.h:12,
                 from ./include/asm-generic/local64.h:5,
                 from ./arch/s390/include/generated/asm/local64.h:1,
                 from ./include/linux/u64_stats_sync.h:71,
                 from ./include/net/snmp.h:47,
                 from ./include/net/netns/mib.h:5,
                 from ./include/net/net_namespace.h:17,
                 from net/ipv4/proc.c:31:
./include/linux/unwind_deferred_types.h:17:9: error: unknown type name ‘local64_t’; did you mean ‘local_t’?
   17 |         local64_t               timestamp;
      |         ^~~~~~~~~
      |         local_t

Reason it fails on s390:

net/ipv4/proc.c

+- include/net/net_namespace.h
   +- include/net/netns/mib.h
      +- include/net/snmp.h
         +- include/linux/u64_stats_sync.h
            +- arch/s390/include/generated/asm/local64.h   <-- ISSUE: local64.h gets included before unwind_deferred_types.h
               +- include/asm-generic/local64.h
                  +- include/linux/percpu.h
                     +- include/linux/sched.h
                        +- include/linux/unwind_deferred_types.h   <-- ERROR: local64_t not yet defined

Reason it does not fail on x86:

net/ipv4/proc.c

+- include/net/net_namespace.h
   +- include/linux/workqueue.h
   |  +- include/linux/timer.h
   |     +- include/linux/ktime.h
   |        +- include/linux/jiffies.h
   |           +- include/linux/time.h
   |               +- include/linux/time32.h
   |                  +- include/linux/timex.h
   |                     +- arch/x86/include/asm/timex.h
   |                        +- arch/x86/include/asm/tsc.h
   |                           +- arch/x86/include/asm/msr.h
   |                              +- include/linux/percpu.h
   |                                 +- include/linux/sched.h
   |                                    +- include/linux/unwind_deferred_types.h   <-- OK: unwind_deferred_types.h gets included before local64.h
   |                                       +- arch/x86/include/generated/asm/local64.h
   |                                          +- include/asm-generic/local64.h
   |                                           [ +- include/linux/percpu.h ]
   +- include/net/netns/mib.h
      +- include/net/snmp.h
         +- include/linux/u64_stats_sync.h
            +- arch/x86/include/generated/asm/local64.h   <-- OK: local64.h comes after unwind_deferred_types.h

[ +- include/asm-generic/local64.h ]
[ +- include/linux/percpu.h ]
[ +- include/linux/sched.h ]
[ +- include/linux/unwind_deferred_types.h ]

My colleague Heiko Carstens (on Cc) suggested the following potential fix that breaks the circular dependency, provided local[64].h does not ever need percpu.h. At least I verified that defconfig x86 and s390 builds still work.

include/asm-generic/local.h +0 -1
index 7f97018df66f..3e7ce6a9e18e 100644
@@ -2,7 +2,6 @@
 #ifndef _ASM_GENERIC_LOCAL_H
 #define _ASM_GENERIC_LOCAL_H
 
-#include <linux/percpu.h>
 #include <linux/atomic.h>
 #include <asm/types.h>
 
include/asm-generic/local64.h +0 -1
index 14963a7a6253..1f9af89916cb 100644
@@ -2,7 +2,6 @@
 #ifndef _ASM_GENERIC_LOCAL64_H
 #define _ASM_GENERIC_LOCAL64_H
 
-#include <linux/percpu.h>
 #include <asm/types.h>
 
Quoted 5 lines
include/linux/unwind_deferred_types.h +1 -0
+
 struct unwind_cache {
 	unsigned int		nr_entries;
 	unsigned long		entries[];

include/linux/unwind_deferred_types.h struct unwind_cache { · line 20

Quoted 10 lines
include/linux/unwind_deferred_types.h +2 -2
@@ -10,8 +13,8 @@ struct unwind_cache {
 struct unwind_task_info {
 	struct unwind_cache	*cache;
 	struct callback_head	work;
-	u64			timestamp;
-	int			pending;
+	local64_t		timestamp;
+	local_t			pending;
 };
 
 #endif /* _LINUX_UNWIND_USER_DEFERRED_TYPES_H */

Regards,
Jens

Signature · 9 lines
Jens Remus
Linux on Z Development (D3303)
+49-7031-16-1128 Office
[email protected]

IBM

IBM Deutschland Research & Development GmbH; Vorsitzender des Aufsichtsrats: Wolfgang Wendt; Geschäftsführung: David Faller; Sitz der Gesellschaft: Böblingen; Registergericht: Amtsgericht Stuttgart, HRB 243294
IBM Data Privacy Statement: https://www.ibm.com/privacy/
author maintainer
#39 lore raw

Re: [PATCH v12 07/14] unwind_user/deferred: Make unwind deferral requests NMI-safe

Quoted 11 lines · On Wed, 2 Jul 2025 17:53:05 +0200 Jens Remus <[email protected]> wrote:

Review line 6

Quoted 6 lines
(quoted diff) +2 -0
@@ -2,6 +2,9 @@
 #ifndef _LINUX_UNWIND_USER_DEFERRED_TYPES_H
 #define _LINUX_UNWIND_USER_DEFERRED_TYPES_H
 
+#include <asm/local64.h>
+#include <asm/local.h>  

This creates the following circular dependency, that breaks the build on s390 as follows, whenever local64.h is included first, so that local64_t is not yet defined when unwind_deferred_types.h gets included down the line:

As per the discussion on patch 6, this may not be an issue in the next version. I'm looking to get rid of 64 bit cmpxchg.

-- Steve

author maintainer
#40 lore raw

PATCH v12 8/14 unwind deferred: Use bitmask to determine which callbacks to call

In reply to Steven Rostedt

From: Steven Rostedt <[email protected]>

In order to know which registered callback requested a stacktrace for when the task goes back to user space, add a bitmask to keep track of all registered tracers. The bitmask is the size of long, which means that on a 32 bit machine, it can have at most 32 registered tracers, and on 64 bit, it can have at most 64 registered tracers. This should not be an issue as there should not be more than 10 (unless BPF can abuse this?).

When a tracer registers with unwind_deferred_init() it will get a bit number assigned to it. When a tracer requests a stacktrace, it will have its bit set within the task_struct. When the task returns back to user space, it will call the callbacks for all the registered tracers where their bits are set in the task's mask.

When a tracer is removed by the unwind_deferred_cancel() all current tasks will clear the associated bit, just in case another tracer gets registered immediately afterward and then gets their callback called unexpectedly.

---
 include/linux/unwind_deferred.h       |  1 +
 include/linux/unwind_deferred_types.h |  1 +
 kernel/unwind/deferred.c              | 36 ++++++++++++++++++++++++---
 3 files changed, 34 insertions(+), 4 deletions(-)
include/linux/unwind_deferred.h +1 -0
index 73f6cac53530..538b4b7968dc 100644
@@ -13,6 +13,7 @@ typedef void (*unwind_callback_t)(struct unwind_work *work, struct unwind_stackt
 struct unwind_work {
 	struct list_head		list;
 	unwind_callback_t		func;
+	int				bit;
 };
 
 #ifdef CONFIG_UNWIND_USER
include/linux/unwind_deferred_types.h +1 -0
index 0d722e877473..5863bf4eb436 100644
@@ -13,6 +13,7 @@ struct unwind_cache {
 struct unwind_task_info {
 	struct unwind_cache	*cache;
 	struct callback_head	work;
+	unsigned long		unwind_mask;
 	local64_t		timestamp;
 	local_t			pending;
 };
kernel/unwind/deferred.c +32 -4
index dd36e58c8cad..6c558d00ff41 100644
@@ -48,6 +48,7 @@ static inline u64 assign_timestamp(struct unwind_task_info *info,
 /* Guards adding to and reading the list of callbacks */
 static DEFINE_MUTEX(callback_mutex);
 static LIST_HEAD(callbacks);
+static unsigned long unwind_mask;
 
 /*
  * Read the task context timestamp, if this is the first caller then
@@ -153,7 +154,10 @@ static void unwind_deferred_task_work(struct callback_head *head)
 
 	guard(mutex)(&callback_mutex);
 	list_for_each_entry(work, &callbacks, list) {
-		work->func(work, &trace, timestamp);
+		if (test_bit(work->bit, &info->unwind_mask)) {
+			work->func(work, &trace, timestamp);
+			clear_bit(work->bit, &info->unwind_mask);
+		}
 	}
 }
 
@@ -205,15 +209,19 @@ int unwind_deferred_request(struct unwind_work *work, u64 *timestamp)
 
 	*timestamp = get_timestamp(info);
 
+	/* This is already queued */
+	if (test_bit(work->bit, &info->unwind_mask))
+		return 1;
+
 	/* callback already pending? */
 	pending = local_read(&info->pending);
 	if (pending)
-		return 1;
+		goto out;
 
 	if (CAN_USE_IN_NMI) {
 		/* Claim the work unless an NMI just now swooped in to do so. */
 		if (!local_try_cmpxchg(&info->pending, &pending, 1))
-			return 1;
+			goto out;
 	} else {
 		local_set(&info->pending, 1);
 	}
@@ -225,16 +233,27 @@ int unwind_deferred_request(struct unwind_work *work, u64 *timestamp)
 		return ret;
 	}
 
-	return 0;
+ out:
+	return test_and_set_bit(work->bit, &info->unwind_mask);
 }
 
 void unwind_deferred_cancel(struct unwind_work *work)
 {
+	struct task_struct *g, *t;
+
 	if (!work)
 		return;
 
 	guard(mutex)(&callback_mutex);
 	list_del(&work->list);
+
+	__clear_bit(work->bit, &unwind_mask);
+
+	guard(rcu)();
+	/* Clear this bit from all threads */
+	for_each_process_thread(g, t) {
+		clear_bit(work->bit, &t->unwind_info.unwind_mask);
+	}
 }
 
 int unwind_deferred_init(struct unwind_work *work, unwind_callback_t func)
@@ -242,6 +261,14 @@ int unwind_deferred_init(struct unwind_work *work, unwind_callback_t func)
 	memset(work, 0, sizeof(*work));
 
 	guard(mutex)(&callback_mutex);
+
+	/* See if there's a bit in the mask available */
+	if (unwind_mask == ~0UL)
+		return -EBUSY;
+
+	work->bit = ffz(unwind_mask);
+	__set_bit(work->bit, &unwind_mask);
+
 	list_add(&work->list, &callbacks);
 	work->func = func;
 	return 0;
@@ -253,6 +280,7 @@ void unwind_task_init(struct task_struct *task)
 
 	memset(info, 0, sizeof(*info));
 	init_task_work(&info->work, unwind_deferred_task_work);
+	info->unwind_mask = 0;
 }
 
 void unwind_task_free(struct task_struct *task)
Signature · 1 line
2.47.2
author maintainer
#41 lore raw

PATCH v12 9/14 unwind deferred: Use SRCU unwind_deferred_task_work()

In reply to Steven Rostedt

From: Steven Rostedt <[email protected]>

Instead of using the callback_mutex to protect the link list of callbacks in unwind_deferred_task_work(), use SRCU instead. This gets called every time a task exits that has to record a stack trace that was requested. This can happen for many tasks on several CPUs at the same time. A mutex is a bottleneck and can cause a bit of contention and slow down performance.

As the callbacks themselves are allowed to sleep, regular RCU cannot be used to protect the list. Instead use SRCU, as that still allows the callbacks to sleep and the list can be read without needing to hold the callback_mutex.

---
 kernel/unwind/deferred.c | 35 ++++++++++++++++++++++++++---------
 1 file changed, 26 insertions(+), 9 deletions(-)
kernel/unwind/deferred.c +26 -9
index 6c558d00ff41..7309c9e0e57a 100644
@@ -45,10 +45,11 @@ static inline u64 assign_timestamp(struct unwind_task_info *info,
 #define UNWIND_MAX_ENTRIES					\
 	((SZ_4K - sizeof(struct unwind_cache)) / sizeof(long))
 
-/* Guards adding to and reading the list of callbacks */
+/* Guards adding to or removing from the list of callbacks */
 static DEFINE_MUTEX(callback_mutex);
 static LIST_HEAD(callbacks);
 static unsigned long unwind_mask;
+DEFINE_STATIC_SRCU(unwind_srcu);
 
 /*
  * Read the task context timestamp, if this is the first caller then
@@ -134,6 +135,7 @@ static void unwind_deferred_task_work(struct callback_head *head)
 	struct unwind_stacktrace trace;
 	struct unwind_work *work;
 	u64 timestamp;
+	int idx;
 
 	if (WARN_ON_ONCE(!local_read(&info->pending)))
 		return;
@@ -152,13 +154,15 @@ static void unwind_deferred_task_work(struct callback_head *head)
 
 	timestamp = local64_read(&info->timestamp);
 
-	guard(mutex)(&callback_mutex);
-	list_for_each_entry(work, &callbacks, list) {
+	idx = srcu_read_lock(&unwind_srcu);
+	list_for_each_entry_srcu(work, &callbacks, list,
+				 srcu_read_lock_held(&unwind_srcu)) {
 		if (test_bit(work->bit, &info->unwind_mask)) {
 			work->func(work, &trace, timestamp);
 			clear_bit(work->bit, &info->unwind_mask);
 		}
 	}
+	srcu_read_unlock(&unwind_srcu, idx);
 }
 
 /**
@@ -193,6 +197,7 @@ int unwind_deferred_request(struct unwind_work *work, u64 *timestamp)
 {
 	struct unwind_task_info *info = &current->unwind_info;
 	long pending;
+	int bit;
 	int ret;
 
 	*timestamp = 0;
@@ -205,12 +210,17 @@ int unwind_deferred_request(struct unwind_work *work, u64 *timestamp)
 	if (!CAN_USE_IN_NMI && in_nmi())
 		return -EINVAL;
 
+	/* Do not allow cancelled works to request again */
+	bit = READ_ONCE(work->bit);
+	if (WARN_ON_ONCE(bit < 0))
+		return -EINVAL;
+
 	guard(irqsave)();
 
 	*timestamp = get_timestamp(info);
 
 	/* This is already queued */
-	if (test_bit(work->bit, &info->unwind_mask))
+	if (test_bit(bit, &info->unwind_mask))
 		return 1;
 
 	/* callback already pending? */
@@ -234,25 +244,32 @@ int unwind_deferred_request(struct unwind_work *work, u64 *timestamp)
 	}
 
  out:
-	return test_and_set_bit(work->bit, &info->unwind_mask);
+	return test_and_set_bit(bit, &info->unwind_mask);
 }
 
 void unwind_deferred_cancel(struct unwind_work *work)
 {
 	struct task_struct *g, *t;
+	int bit;
 
 	if (!work)
 		return;
 
 	guard(mutex)(&callback_mutex);
-	list_del(&work->list);
+	list_del_rcu(&work->list);
+	bit = work->bit;
+
+	/* Do not allow any more requests and prevent callbacks */
+	work->bit = -1;
+
+	__clear_bit(bit, &unwind_mask);
 
-	__clear_bit(work->bit, &unwind_mask);
+	synchronize_srcu(&unwind_srcu);
 
 	guard(rcu)();
 	/* Clear this bit from all threads */
 	for_each_process_thread(g, t) {
-		clear_bit(work->bit, &t->unwind_info.unwind_mask);
+		clear_bit(bit, &t->unwind_info.unwind_mask);
 	}
 }
 
@@ -269,7 +286,7 @@ int unwind_deferred_init(struct unwind_work *work, unwind_callback_t func)
 	work->bit = ffz(unwind_mask);
 	__set_bit(work->bit, &unwind_mask);
 
-	list_add(&work->list, &callbacks);
+	list_add_rcu(&work->list, &callbacks);
 	work->func = func;
 	return 0;
 }
Signature · 1 line
2.47.2
author maintainer
#42 lore raw

PATCH v12 10/14 unwind: Clear unwind_mask on exit back to user space

In reply to Steven Rostedt

From: Steven Rostedt <[email protected]>

When testing the deferred unwinder by attaching deferred user space stacktraces to events, a live lock happened. This was when the deferred unwinding was added to the irqs_disabled event, which happens after the task_work callbacks are called and before the task goes back to user space.

The event callback would be registered when irqs were disabled, the task_work would trigger, call the callback for this work and clear the work's bit. Then before getting back to user space, irqs would be disabled again, the event triggered again, and a new task_work registered. This caused an infinite loop and the system hung.

To prevent this, clear the bits at the very last moment before going back to user space and when instrumentation is disabled. That is in unwind_exit_to_user_mode().

Move the pending bit from a value on the task_struct to the most significant bit of the unwind_mask (saves space on the task_struct). This will allow modifying the pending bit along with the work bits atomically.

Instead of clearing a work's bit after its callback is called, it is delayed until exit. If the work is requested again, the task_work is not queued again and the work will be notified that the task has already been called (via UNWIND_ALREADY_EXECUTED return value).

The pending bit is cleared before calling the callback functions but the current work bits remain. If one of the called works registers again, it will not trigger a task_work if its bit is still present in the task's unwind_mask.

If a new work registers, then it will set both the pending bit and its own bit but clear the other work bits so that their callbacks do not get called again.

---
Changes since v11: https://lore.kernel.org/[email protected]

- Still clear info->cache->nr_entries in unwind_reset_info() If unwind_user_faultable() is called directly (like perf will do) it still needs to clear the nr_entries as calling the function directly does not set the bit.

Later code will change this so it has only one conditional to check when not enabled.

 include/linux/unwind_deferred.h       | 25 +++++++--
 include/linux/unwind_deferred_types.h |  1 -
 kernel/unwind/deferred.c              | 76 ++++++++++++++++++---------
 3 files changed, 74 insertions(+), 28 deletions(-)
include/linux/unwind_deferred.h +22 -3
index 538b4b7968dc..d25a72fb21ef 100644
@@ -18,6 +18,14 @@ struct unwind_work {
 
 #ifdef CONFIG_UNWIND_USER
 
+#define UNWIND_PENDING_BIT	(BITS_PER_LONG - 1)
+#define UNWIND_PENDING		BIT(UNWIND_PENDING_BIT)
+
+enum {
+	UNWIND_ALREADY_PENDING	= 1,
+	UNWIND_ALREADY_EXECUTED	= 2,
+};
+
 void unwind_task_init(struct task_struct *task);
 void unwind_task_free(struct task_struct *task);
 
@@ -29,15 +37,26 @@ void unwind_deferred_cancel(struct unwind_work *work);
 
 static __always_inline void unwind_reset_info(void)
 {
-	if (unlikely(local64_read(&current->unwind_info.timestamp)))
+	struct unwind_task_info *info = &current->unwind_info;
+	unsigned long bits;
+
+	/* Was there any unwinding? */
+	if (unlikely(info->unwind_mask)) {
+		bits = info->unwind_mask;
+		do {
+			/* Is a task_work going to run again before going back */
+			if (bits & UNWIND_PENDING)
+				return;
+		} while (!try_cmpxchg(&info->unwind_mask, &bits, 0UL));
 		local64_set(&current->unwind_info.timestamp, 0);
+	}
 	/*
 	 * As unwind_user_faultable() can be called directly and
 	 * depends on nr_entries being cleared on exit to user,
 	 * this needs to be a separate conditional.
 	 */
-	if (unlikely(current->unwind_info.cache))
-		current->unwind_info.cache->nr_entries = 0;
+	if (unlikely(info->cache))
+		info->cache->nr_entries = 0;
 }
 
 #else /* !CONFIG_UNWIND_USER */
include/linux/unwind_deferred_types.h +0 -1
index 5863bf4eb436..4308367f1887 100644
@@ -15,7 +15,6 @@ struct unwind_task_info {
 	struct callback_head	work;
 	unsigned long		unwind_mask;
 	local64_t		timestamp;
-	local_t			pending;
 };
 
 #endif /* _LINUX_UNWIND_USER_DEFERRED_TYPES_H */
kernel/unwind/deferred.c +52 -24

136 lines not on this page · raw message

Signature · 1 line
2.47.2
author maintainer
#43 lore raw

PATCH v12 11/14 unwind: Add USED bit to only have one conditional on way back to user space

In reply to Steven Rostedt

From: Steven Rostedt <[email protected]>

On the way back to user space, the function unwind_reset_info() is called unconditionally (but always inlined). It currently has two conditionals. One that checks the unwind_mask which is set whenever a deferred trace is called and is used to know that the mask needs to be cleared. The other checks if the cache has been allocated, and if so, it resets the nr_entries so that the unwinder knows it needs to do the work to get a new user space stack trace again (it only does it once per entering the kernel).

Use one of the bits in the unwind mask as a "USED" bit that gets set whenever a trace is created. This will make it possible to only check the unwind_mask in the unwind_reset_info() to know if it needs to do work or not and eliminates a conditional that happens every time the task goes back to user space.

---
 include/linux/unwind_deferred.h | 14 +++++++-------
 kernel/unwind/deferred.c        |  5 ++++-
 2 files changed, 11 insertions(+), 8 deletions(-)
include/linux/unwind_deferred.h +7 -7
index d25a72fb21ef..a1c62097f142 100644
@@ -21,6 +21,10 @@ struct unwind_work {
 #define UNWIND_PENDING_BIT	(BITS_PER_LONG - 1)
 #define UNWIND_PENDING		BIT(UNWIND_PENDING_BIT)
 
+/* Set if the unwinding was used (directly or deferred) */
+#define UNWIND_USED_BIT		(UNWIND_PENDING_BIT - 1)
+#define UNWIND_USED		BIT(UNWIND_USED_BIT)
+
 enum {
 	UNWIND_ALREADY_PENDING	= 1,
 	UNWIND_ALREADY_EXECUTED	= 2,
@@ -49,14 +53,10 @@ static __always_inline void unwind_reset_info(void)
 				return;
 		} while (!try_cmpxchg(&info->unwind_mask, &bits, 0UL));
 		local64_set(&current->unwind_info.timestamp, 0);
+
+		if (unlikely(info->cache))
+			info->cache->nr_entries = 0;
 	}
-	/*
-	 * As unwind_user_faultable() can be called directly and
-	 * depends on nr_entries being cleared on exit to user,
-	 * this needs to be a separate conditional.
-	 */
-	if (unlikely(info->cache))
-		info->cache->nr_entries = 0;
 }
 
 #else /* !CONFIG_UNWIND_USER */
kernel/unwind/deferred.c +4 -1
index e7e4442926d3..5ab9b9045ae5 100644
@@ -131,6 +131,9 @@ int unwind_user_faultable(struct unwind_stacktrace *trace)
 
 	cache->nr_entries = trace->nr;
 
+	/* Clear nr_entries on way back to user space */
+	set_bit(UNWIND_USED_BIT, &info->unwind_mask);
+
 	return 0;
 }
 
@@ -308,7 +311,7 @@ int unwind_deferred_init(struct unwind_work *work, unwind_callback_t func)
 	guard(mutex)(&callback_mutex);
 
 	/* See if there's a bit in the mask available */
-	if (unwind_mask == ~(UNWIND_PENDING))
+	if (unwind_mask == ~(UNWIND_PENDING|UNWIND_USED))
 		return -EBUSY;
 
 	work->bit = ffz(unwind_mask);
Signature · 1 line
2.47.2
author maintainer
#44 lore raw

PATCH v12 12/14 unwind: Finish up unwind when a task exits

In reply to Steven Rostedt

From: Steven Rostedt <[email protected]>

On do_exit() when a task is exiting, if a unwind is requested and the deferred user stacktrace is deferred via the task_work, the task_work callback is called after exit_mm() is called in do_exit(). This means that the user stack trace will not be retrieved and an empty stack is created.

Instead, add a function unwind_deferred_task_exit() and call it just before exit_mm() so that the unwinder can call the requested callbacks with the user space stack.

---
 include/linux/unwind_deferred.h |  3 +++
 kernel/exit.c                   |  2 ++
 kernel/unwind/deferred.c        | 23 ++++++++++++++++++++---
 3 files changed, 25 insertions(+), 3 deletions(-)
include/linux/unwind_deferred.h +3 -0
index a1c62097f142..44654e6149ec 100644
@@ -39,6 +39,8 @@ int unwind_deferred_init(struct unwind_work *work, unwind_callback_t func);
 int unwind_deferred_request(struct unwind_work *work, u64 *timestamp);
 void unwind_deferred_cancel(struct unwind_work *work);
 
+void unwind_deferred_task_exit(struct task_struct *task);
+
 static __always_inline void unwind_reset_info(void)
 {
 	struct unwind_task_info *info = &current->unwind_info;
@@ -69,6 +71,7 @@ static inline int unwind_deferred_init(struct unwind_work *work, unwind_callback
 static inline int unwind_deferred_request(struct unwind_work *work, u64 *timestamp) { return -ENOSYS; }
 static inline void unwind_deferred_cancel(struct unwind_work *work) {}
 
+static inline void unwind_deferred_task_exit(struct task_struct *task) {}
 static inline void unwind_reset_info(void) {}
 
 #endif /* !CONFIG_UNWIND_USER */
kernel/exit.c +2 -0
index bb184a67ac73..1d8c8ac33c4f 100644
@@ -68,6 +68,7 @@
 #include <linux/rethook.h>
 #include <linux/sysfs.h>
 #include <linux/user_events.h>
+#include <linux/unwind_deferred.h>
 #include <linux/uaccess.h>
 #include <linux/pidfs.h>
 
@@ -938,6 +939,7 @@ void __noreturn do_exit(long code)
 
 	tsk->exit_code = code;
 	taskstats_exit(tsk, group_dead);
+	unwind_deferred_task_exit(tsk);
 	trace_sched_process_exit(tsk, group_dead);
 
 	/*
kernel/unwind/deferred.c +20 -3
index 5ab9b9045ae5..9ec1e74c6469 100644
@@ -104,7 +104,7 @@ int unwind_user_faultable(struct unwind_stacktrace *trace)
 	/* Should always be called from faultable context */
 	might_fault();
 
-	if (current->flags & PF_EXITING)
+	if (!current->mm)
 		return -EINVAL;
 
 	if (!info->cache) {
@@ -137,9 +137,9 @@ int unwind_user_faultable(struct unwind_stacktrace *trace)
 	return 0;
 }
 
-static void unwind_deferred_task_work(struct callback_head *head)
+static void process_unwind_deferred(struct task_struct *task)
 {
-	struct unwind_task_info *info = container_of(head, struct unwind_task_info, work);
+	struct unwind_task_info *info = &task->unwind_info;
 	struct unwind_stacktrace trace;
 	struct unwind_work *work;
 	unsigned long bits;
@@ -174,6 +174,23 @@ static void unwind_deferred_task_work(struct callback_head *head)
 	srcu_read_unlock(&unwind_srcu, idx);
 }
 
+static void unwind_deferred_task_work(struct callback_head *head)
+{
+	process_unwind_deferred(current);
+}
+
+void unwind_deferred_task_exit(struct task_struct *task)
+{
+	struct unwind_task_info *info = &current->unwind_info;
+
+	if (!unwind_pending(info))
+		return;
+
+	process_unwind_deferred(task);
+
+	task_work_cancel(task, &info->work);
+}
+
 /**
  * unwind_deferred_request - Request a user stacktrace on task exit
  * @work: Unwind descriptor requesting the trace
Signature · 1 line
2.47.2
author maintainer
#45 lore raw

PATCH v12 13/14 unwind_user/x86: Enable frame pointer unwinding on x86

In reply to Steven Rostedt

From: Josh Poimboeuf <[email protected]>

Use ARCH_INIT_USER_FP_FRAME to describe how frame pointers are unwound on x86, and enable CONFIG_HAVE_UNWIND_USER_FP accordingly so the unwind_user interfaces can be used.

---
 arch/x86/Kconfig                   |  1 +
 arch/x86/include/asm/unwind_user.h | 11 +++++++++++
 2 files changed, 12 insertions(+)
 create mode 100644 arch/x86/include/asm/unwind_user.h
arch/x86/Kconfig +1 -0
index 71019b3b54ea..5862433c81e1 100644
@@ -302,6 +302,7 @@ config X86
 	select HAVE_SYSCALL_TRACEPOINTS
 	select HAVE_UACCESS_VALIDATION		if HAVE_OBJTOOL
 	select HAVE_UNSTABLE_SCHED_CLOCK
+	select HAVE_UNWIND_USER_FP		if X86_64
 	select HAVE_USER_RETURN_NOTIFIER
 	select HAVE_GENERIC_VDSO
 	select VDSO_GETRANDOM			if X86_64
arch/x86/include/asm/unwind_user.h new file +11 -0
new file mode 100644
index 000000000000..8597857bf896
@@ -0,0 +1,11 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _ASM_X86_UNWIND_USER_H
+#define _ASM_X86_UNWIND_USER_H
+
+#define ARCH_INIT_USER_FP_FRAME							\
+	.cfa_off	= (s32)sizeof(long) *  2,				\
+	.ra_off		= (s32)sizeof(long) * -1,				\
+	.fp_off		= (s32)sizeof(long) * -2,				\
+	.use_fp		= true,
+
+#endif /* _ASM_X86_UNWIND_USER_H */
Signature · 1 line
2.47.2
author maintainer
#46 lore raw

PATCH v12 14/14 unwind_user/x86: Enable compat mode frame pointer unwinding on x86

In reply to Steven Rostedt

From: Josh Poimboeuf <[email protected]>

Use ARCH_INIT_USER_COMPAT_FP_FRAME to describe how frame pointers are unwound on x86, and implement the hooks needed to add the segment base addresses. Enable HAVE_UNWIND_USER_COMPAT_FP if the system has compat mode compiled in.

---
Changes since v11: https://lore.kernel.org/[email protected]

- Fix header macro protection name to include X86 (Ingo Molnar)

- Use insn_get_seg_base() to get segment registers instead of using the function perf uses and making it global. Also as that function doesn't look to have a requirement to disable interrupts, the scoped_guard(irqsave) is removed.

- Check return code of insn_get_seg_base() for the unlikely event that it returns invalid (-1).

- Moved arch_unwind_user_init() into stacktrace.c as to use insn_get_seg_base(), it must include insn-eval.h that defines pt_regs_offset(), but that is also used in the perf generic code as an array and if it is included in the header file, it causes a build conflict.

- Update the comments that explain arch_unwind_user_init/next that a macro needs to be defined with those names if they are going to be used.

 arch/x86/Kconfig                         |  1 +
 arch/x86/include/asm/unwind_user.h       | 31 ++++++++++++++++++++++++
 arch/x86/include/asm/unwind_user_types.h | 17 +++++++++++++
 arch/x86/kernel/stacktrace.c             | 28 +++++++++++++++++++++
 include/linux/unwind_user.h              | 20 +++++++++++++++
 kernel/unwind/user.c                     |  4 +++
 6 files changed, 101 insertions(+)
 create mode 100644 arch/x86/include/asm/unwind_user_types.h
arch/x86/Kconfig +1 -0
index 5862433c81e1..17d4094c821b 100644
@@ -302,6 +302,7 @@ config X86
 	select HAVE_SYSCALL_TRACEPOINTS
 	select HAVE_UACCESS_VALIDATION		if HAVE_OBJTOOL
 	select HAVE_UNSTABLE_SCHED_CLOCK
+	select HAVE_UNWIND_USER_COMPAT_FP	if IA32_EMULATION
 	select HAVE_UNWIND_USER_FP		if X86_64
 	select HAVE_USER_RETURN_NOTIFIER
 	select HAVE_GENERIC_VDSO
arch/x86/include/asm/unwind_user.h +31 -0
index 8597857bf896..19634a73612d 100644
@@ -2,10 +2,41 @@
 #ifndef _ASM_X86_UNWIND_USER_H
 #define _ASM_X86_UNWIND_USER_H
 
+#include <linux/unwind_user_types.h>
+
 #define ARCH_INIT_USER_FP_FRAME							\
 	.cfa_off	= (s32)sizeof(long) *  2,				\
 	.ra_off		= (s32)sizeof(long) * -1,				\
 	.fp_off		= (s32)sizeof(long) * -2,				\
 	.use_fp		= true,
 
+#ifdef CONFIG_IA32_EMULATION
+
+#define ARCH_INIT_USER_COMPAT_FP_FRAME						\
+	.cfa_off	= (s32)sizeof(u32)  *  2,				\
+	.ra_off		= (s32)sizeof(u32)  * -1,				\
+	.fp_off		= (s32)sizeof(u32)  * -2,				\
+	.use_fp		= true,
+
+#define in_compat_mode(regs) !user_64bit_mode(regs)
+
+void arch_unwind_user_init(struct unwind_user_state *state,
+			   struct pt_regs *regs);
+
+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_init arch_unwind_user_init
+#define arch_unwind_user_next arch_unwind_user_next
+
+#endif /* CONFIG_IA32_EMULATION */
+
+#include <asm-generic/unwind_user.h>
+
 #endif /* _ASM_X86_UNWIND_USER_H */
arch/x86/include/asm/unwind_user_types.h new file +17 -0
new file mode 100644
index 000000000000..f93d535f900e
@@ -0,0 +1,17 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _ASM_X86_UNWIND_USER_TYPES_H
+#define _ASM_X86_UNWIND_USER_TYPES_H
+
+#ifdef CONFIG_IA32_EMULATION
+
+struct arch_unwind_user_state {
+	unsigned long ss_base;
+	unsigned long cs_base;
+};
+#define arch_unwind_user_state arch_unwind_user_state
+
+#endif /* CONFIG_IA32_EMULATION */
+
+#include <asm-generic/unwind_user_types.h>
+
+#endif /* _ASM_UNWIND_USER_TYPES_H */
arch/x86/kernel/stacktrace.c +28 -0
index ee117fcf46ed..8ef9d8c71df9 100644
@@ -9,7 +9,10 @@
 #include <linux/stacktrace.h>
 #include <linux/export.h>
 #include <linux/uaccess.h>
+#include <asm/unwind_user.h>
 #include <asm/stacktrace.h>
+#include <asm/insn.h>
+#include <asm/insn-eval.h>
 #include <asm/unwind.h>
 
 void arch_stack_walk(stack_trace_consume_fn consume_entry, void *cookie,
@@ -128,3 +131,28 @@ void arch_stack_walk_user(stack_trace_consume_fn consume_entry, void *cookie,
 	}
 }
 
+#ifdef CONFIG_IA32_EMULATION
+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;
+
+	cs_base = insn_get_seg_base(regs, INAT_SEG_REG_CS);
+	ss_base = insn_get_seg_base(regs, INAT_SEG_REG_SS);
+
+	if (cs_base == -1)
+		cs_base = 0;
+	if (ss_base == -1)
+		ss_base = 0;
+
+	state->arch.cs_base = cs_base;
+	state->arch.ss_base = ss_base;
+
+	state->ip += cs_base;
+	state->sp += ss_base;
+	state->fp += ss_base;
+}
+#endif /* CONFIG_IA32_EMULATION */
include/linux/unwind_user.h +20 -0
index ac007363820a..b57b68215c6f 100644
@@ -14,6 +14,26 @@
  #define in_compat_mode(regs) false
 #endif
 
+/*
+ * If an architecture needs to initialize the state for a specific
+ * reason, for example, it may need to do something different
+ * in compat mode, it can define a macro named arch_unwind_user_init
+ * with the name of the function that will perform this initialization.
+ */
+#ifndef arch_unwind_user_init
+static inline void arch_unwind_user_init(struct unwind_user_state *state, struct pt_regs *reg) {}
+#endif
+
+/*
+ * If an architecture requires some more updates to the state between
+ * stack frames, it can define a macro named arch_unwind_user_next
+ * with the name of the function that will update the state between
+ * reading stack frames during the user space stack walk.
+ */
+#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);
 
kernel/unwind/user.c +4 -0
index 3a0ac4346f5b..2bb7995c3f23 100644
@@ -72,6 +72,8 @@ int unwind_user_next(struct unwind_user_state *state)
 	if (frame->fp_off)
 		state->fp = fp;
 
+	arch_unwind_user_next(state);
+
 	return 0;
 
 done:
@@ -101,6 +103,8 @@ 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;
 }
 
Signature · 1 line
2.47.2
#47 lore raw
In reply to Steven Rostedt
Quoted 4 lines · Steven Rostedt

This is the first patch series of a set that will make it possible to be able to use SFrames[1] in the Linux kernel. A quick recap of the motivation for doing this.

You have a '[1]' to indicate there's a link to what SFrames are.

But no such link actually exists in this email. Hmm?

Linus

author maintainer
#48 lore raw
Quoted 10 lines · On Mon, 30 Jun 2025 19:06:12 -0700 Linus Torvalds…
Quoted 4 lines · Steven Rostedt

This is the first patch series of a set that will make it possible to be able to use SFrames[1] in the Linux kernel. A quick recap of the motivation for doing this.

You have a '[1]' to indicate there's a link to what SFrames are.

But no such link actually exists in this email. Hmm?

Oops. I cut and pasted from v11:

https://lore.kernel.org/linux-trace-kernel/[email protected]/

But ended the cut too early. I stopped at my sig:

---
Thanks!

-- Steve

[1] https://sourceware.org/binutils/wiki/sframe

-- Steve

maintainer
#49 lore raw
Quoted 5 lines · Steven Rostedt
Quoted 2 lines · On Mon, 30 Jun 2025 19:06:12 -0700 Linus Torvalds…
Quoted 1 line · Steven Rostedt
Quoted 7 lines
Quoted 5 lines
Quoted 3 lines

This is the first patch series of a set that will make it possible to be able to use SFrames[1] in the Linux kernel. A quick recap of the motivation for doing this.

You have a '[1]' to indicate there's a link to what SFrames are.

[...]
[1] https://sourceware.org/binutils/wiki/sframe

Okay, I've read the cover letter and this wiki page, but I am dense: why does the _kernel_ want to do this? Shouldn't it only be userspace that cares about userspace unwinding? I don't use perf, ftrace, and ebpf enough to make this obvious to me, I guess. ;)

Signature · 1 line
Kees Cook
author maintainer
#50 lore raw
Quoted 10 lines · On Tue, 1 Jul 2025 15:49:23 -0700 Kees Cook <[email protected]> wrote:
Quoted 9 lines · Steven Rostedt
Quoted 6 lines · On Mon, 30 Jun 2025 19:06:12 -0700 Linus Torvalds…
Quoted 4 lines · Steven Rostedt

This is the first patch series of a set that will make it possible to be able to use SFrames[1] in the Linux kernel. A quick recap of the motivation for doing this.

Quoted 9 lines
Quoted 3 lines
Quoted 1 line

You have a '[1]' to indicate there's a link to what SFrames are.

[...]
[1] https://sourceware.org/binutils/wiki/sframe

Okay, I've read the cover letter and this wiki page, but I am dense: why does the _kernel_ want to do this? Shouldn't it only be userspace that cares about userspace unwinding? I don't use perf, ftrace, and ebpf enough to make this obvious to me, I guess. ;)

It's how perf does profiling. It needs to walk the user space stack to see what functions are being called. Ftrace can do the same thing, but is not as used because it doesn't have the tooling (yet) to figure out what the user space addresses mean (but I'm working on fixing that).

And BPF has commands that it can do, but I don't know BPF enough to comment.

The big user is perf with profiling. It currently uses frame pointers, but because of the way frame pointers are set up, it misses a lot of the leaf functions when the interrupt triggers (which sframes does not have that problem). Also, if frame pointers is not configured, perf may just copy thousands of bytes of the user space stack into the kernel ring buffer and then parse it later (this isn't used often due to the overhead).

Then there's s390 that doesn't have frame pointers and only has the copy of thousands of bytes to do any meaningful user space profiling.

Note, this has been a long standing issue where in 2022, we had a BOF on this, looking for something like ORC in user space as it would solve lots of our issues. Then December of that same year, we heard about SFrames.

At Kernel Recipes in 2023, Brendan Gregg during his talk was saying that there needs to be a better way to do profiling of user space from the kernel without frame pointers. I mentioned SFrames and he was quite excited to hear about it. That's also when Josh, who was in the attendance, asked if he could do the implementation of it in the kernel!

Anyway, yeah, it's something that has a ton of interest, as it's the way for tools like perf to give nice graphs of where user space bottlenecks exist.

-- Steve

maintainer
#51 lore raw
Quoted 9 lines · Steven Rostedt
Quoted 4 lines · On Tue, 1 Jul 2025 15:49:23 -0700 Kees Cook <[email protected]> wrote:

Okay, I've read the cover letter and this wiki page, but I am dense: why does the _kernel_ want to do this? Shouldn't it only be userspace that cares about userspace unwinding? I don't use perf, ftrace, and ebpf enough to make this obvious to me, I guess. ;)

[...]
Anyway, yeah, it's something that has a ton of interest, as it's the way for tools like perf to give nice graphs of where user space bottlenecks exist.

Right! Yeah, I know it's very wanted -- I wasn't saying "don't this in the kernel", but quite literally, "*I* am missing something about why this is so important." :) And thank you, now I see: the sampling-based profiling of userspace must happen via the kernel.

Signature · 1 line
Kees Cook
#52 lore raw
Quoted 11 lines · Kees Cook
Quoted 9 lines · Steven Rostedt
Quoted 4 lines · On Tue, 1 Jul 2025 15:49:23 -0700 Kees Cook <[email protected]> wrote:

Okay, I've read the cover letter and this wiki page, but I am dense: why does the _kernel_ want to do this? Shouldn't it only be userspace that cares about userspace unwinding? I don't use perf, ftrace, and ebpf enough to make this obvious to me, I guess. ;)

[...]
Anyway, yeah, it's something that has a ton of interest, as it's the way for tools like perf to give nice graphs of where user space bottlenecks exist.

Quoted 5 lines

Right! Yeah, I know it's very wanted -- I wasn't saying "don't this in the kernel", but quite literally, "*I* am missing something about why this is so important." :) And thank you, now I see: the sampling-based profiling of userspace must happen via the kernel.

I should add that once we have this in place for perf sampling, it enables the following additional use-cases:

- Sample stack traces from specific tracepoints, e.g. system call entry. This allows associating the kernel tracepoints with their userspace call chain (causality) without requiring tracing of userspace.

- Implement a system call that allows a userspace thread to use the kernel stack sampling facilities on itself rather than reimplement the stack walk and sframe registry handling + decoding on the userspace side.

For this last point, it's only relevant because the infrastructure will already be in place for stack sampling from the kernel. So we'd eliminate duplication by allowing its use from userspace as well.

Thanks,

Mathieu

Signature · 3 lines
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com