[PATCH] unwind: Add sframe_(un)register() system calls

[PATCH] unwind: Add sframe_(un)register() system calls

[Steven Rostedt]

From: Steven Rostedt <rostedt@goodmis.org>

Add system calls to register and unregister sframes that can be used by
dynamic linkers to tell the kernel where the sframe section is in memory
for libraries it loads.

Both system calls take a pointer to a new structure:

  struct sframe_setup {
	unsigned long		sframe_start;
	unsigned long		sframe_size;
	unsigned long		text_start;
	unsigned long		text_size;
  };

and a size of the passed in structure. If the system call needs to be
extended, then the structure could be changed and the size of that
structure will tell the kernel that it is the new version. If the kernel
does not recognize the structure size, it will return -EINVAL.

  sframe_start - The virtual address of the sframe section
  sframe_size  - The length of the sframe section
  text_start   - the text section the sframe represents
  test_size    - the length of the section

If other stack tracing functionality is added, it will require a new
system call.

The unregister only needs the sframe_start and requires all the rest of
the fields to be 0. In the future, if more can be done, then user space
can update the other values and check the return code to see if the kernel
supports it.

Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
---

Based on top of Jens patches here:

  https://lore.kernel.org/linux-trace-kernel/20260520154004.3845823-1-jremus@linux.ibm.com/

[ Note, I tested this with the same program from the RFC patch ]

Changes from RFC: https://patch.msgid.link/20260429114355.6c712e6a@gandalf.local.home

- Remove the ioctl() like system call for a unique system call for each
  functionality. Right now there's two functionalities:
   1. register sframe section
   2. unregister sframe sections

- Added taking a lock around the mtree logic in __sframe_remove_section()
  as Sashiko mentioned that there could be races from user space
  registering and unregistering sframe sections at the same time.

- Removed [RFC] from subject as I believe this is more likely the way
  this system call will be done.

 arch/alpha/kernel/syscalls/syscall.tbl      |  2 +
 arch/arm/tools/syscall.tbl                  |  2 +
 arch/arm64/tools/syscall_32.tbl             |  2 +
 arch/m68k/kernel/syscalls/syscall.tbl       |  2 +
 arch/microblaze/kernel/syscalls/syscall.tbl |  2 +
 arch/mips/kernel/syscalls/syscall_n32.tbl   |  2 +
 arch/mips/kernel/syscalls/syscall_n64.tbl   |  2 +
 arch/mips/kernel/syscalls/syscall_o32.tbl   |  2 +
 arch/parisc/kernel/syscalls/syscall.tbl     |  2 +
 arch/powerpc/kernel/syscalls/syscall.tbl    |  2 +
 arch/s390/kernel/syscalls/syscall.tbl       |  3 +
 arch/sh/kernel/syscalls/syscall.tbl         |  2 +
 arch/sparc/kernel/syscalls/syscall.tbl      |  2 +
 arch/x86/entry/syscalls/syscall_32.tbl      |  2 +
 arch/x86/entry/syscalls/syscall_64.tbl      |  2 +
 arch/xtensa/kernel/syscalls/syscall.tbl     |  2 +
 include/linux/syscalls.h                    |  2 +
 include/uapi/asm-generic/unistd.h           |  7 ++-
 include/uapi/linux/sframe.h                 | 12 ++++
 kernel/sys_ni.c                             |  3 +
 kernel/unwind/sframe.c                      | 63 ++++++++++++++++++++-
 scripts/syscall.tbl                         |  2 +
 22 files changed, 118 insertions(+), 4 deletions(-)
 create mode 100644 include/uapi/linux/sframe.h

diff --git a/arch/alpha/kernel/syscalls/syscall.tbl b/arch/alpha/kernel/syscalls/syscall.tbl
index f31b7afffc34..f0639b831f2a 100644
--- a/arch/alpha/kernel/syscalls/syscall.tbl
+++ b/arch/alpha/kernel/syscalls/syscall.tbl
@@ -511,3 +511,5 @@
 579	common	file_setattr			sys_file_setattr
 580	common	listns				sys_listns
 581	common	rseq_slice_yield		sys_rseq_slice_yield
+582	common	sframe_register			sys_sframe_register
+583	common	sframe_unregister		sys_sframe_unregister
diff --git a/arch/arm/tools/syscall.tbl b/arch/arm/tools/syscall.tbl
index 94351e22bfcf..887b242ffb25 100644
--- a/arch/arm/tools/syscall.tbl
+++ b/arch/arm/tools/syscall.tbl
@@ -486,3 +486,5 @@
 469	common	file_setattr			sys_file_setattr
 470	common	listns				sys_listns
 471	common	rseq_slice_yield		sys_rseq_slice_yield
+472	common	sframe_register			sys_sframe_register
+473	common	sframe_unregister		sys_sframe_unregister
diff --git a/arch/arm64/tools/syscall_32.tbl b/arch/arm64/tools/syscall_32.tbl
index 62d93d88e0fe..c820f1ff718c 100644
--- a/arch/arm64/tools/syscall_32.tbl
+++ b/arch/arm64/tools/syscall_32.tbl
@@ -483,3 +483,5 @@
 469	common	file_setattr			sys_file_setattr
 470	common	listns				sys_listns
 471	common	rseq_slice_yield		sys_rseq_slice_yield
+472	common	sframe_register			sys_sframe_register
+473	common	sframe_unregister		sys_sframe_unregister
diff --git a/arch/m68k/kernel/syscalls/syscall.tbl b/arch/m68k/kernel/syscalls/syscall.tbl
index 248934257101..4c7f17f0364b 100644
--- a/arch/m68k/kernel/syscalls/syscall.tbl
+++ b/arch/m68k/kernel/syscalls/syscall.tbl
@@ -471,3 +471,5 @@
 469	common	file_setattr			sys_file_setattr
 470	common	listns				sys_listns
 471	common	rseq_slice_yield		sys_rseq_slice_yield
+472	common	sframe_register			sys_sframe_register
+473	common	sframe_unregister		sys_sframe_unregister
diff --git a/arch/microblaze/kernel/syscalls/syscall.tbl b/arch/microblaze/kernel/syscalls/syscall.tbl
index 223d26303627..e8dc2cc149f4 100644
--- a/arch/microblaze/kernel/syscalls/syscall.tbl
+++ b/arch/microblaze/kernel/syscalls/syscall.tbl
@@ -477,3 +477,5 @@
 469	common	file_setattr			sys_file_setattr
 470	common	listns				sys_listns
 471	common	rseq_slice_yield		sys_rseq_slice_yield
+472	common	sframe_register			sys_sframe_register
+473	common	sframe_unregister		sys_sframe_unregister
diff --git a/arch/mips/kernel/syscalls/syscall_n32.tbl b/arch/mips/kernel/syscalls/syscall_n32.tbl
index 7430714e2b8f..d0bae05d16af 100644
--- a/arch/mips/kernel/syscalls/syscall_n32.tbl
+++ b/arch/mips/kernel/syscalls/syscall_n32.tbl
@@ -410,3 +410,5 @@
 469	n32	file_setattr			sys_file_setattr
 470	n32	listns				sys_listns
 471	n32	rseq_slice_yield		sys_rseq_slice_yield
+472	n32	sframe_register			sys_sframe_register
+473	n32	sframe_unregister		sys_sframe_unregister
diff --git a/arch/mips/kernel/syscalls/syscall_n64.tbl b/arch/mips/kernel/syscalls/syscall_n64.tbl
index 630aab9e5425..2e200de6a58c 100644
--- a/arch/mips/kernel/syscalls/syscall_n64.tbl
+++ b/arch/mips/kernel/syscalls/syscall_n64.tbl
@@ -386,3 +386,5 @@
 469	n64	file_setattr			sys_file_setattr
 470	n64	listns				sys_listns
 471	n64	rseq_slice_yield		sys_rseq_slice_yield
+472	n64	sframe_register			sys_sframe_register
+473	n64	sframe_unregister		sys_sframe_unregister
diff --git a/arch/mips/kernel/syscalls/syscall_o32.tbl b/arch/mips/kernel/syscalls/syscall_o32.tbl
index 128653112284..0e3b82011ae2 100644
--- a/arch/mips/kernel/syscalls/syscall_o32.tbl
+++ b/arch/mips/kernel/syscalls/syscall_o32.tbl
@@ -459,3 +459,5 @@
 469	o32	file_setattr			sys_file_setattr
 470	o32	listns				sys_listns
 471	o32	rseq_slice_yield		sys_rseq_slice_yield
+472	o32	sframe_register			sys_sframe_register
+473	o32	sframe_unregister		sys_sframe_unregister
diff --git a/arch/parisc/kernel/syscalls/syscall.tbl b/arch/parisc/kernel/syscalls/syscall.tbl
index c6331dad9461..e0758ef8667d 100644
--- a/arch/parisc/kernel/syscalls/syscall.tbl
+++ b/arch/parisc/kernel/syscalls/syscall.tbl
@@ -470,3 +470,5 @@
 469	common	file_setattr			sys_file_setattr
 470	common	listns				sys_listns
 471	common	rseq_slice_yield		sys_rseq_slice_yield
+472	common	sframe_register			sys_sframe_register
+473	common	sframe_unregister		sys_sframe_unregister
diff --git a/arch/powerpc/kernel/syscalls/syscall.tbl b/arch/powerpc/kernel/syscalls/syscall.tbl
index 4fcc7c58a105..eda40c4f4f2f 100644
--- a/arch/powerpc/kernel/syscalls/syscall.tbl
+++ b/arch/powerpc/kernel/syscalls/syscall.tbl
@@ -562,3 +562,5 @@
 469	common	file_setattr			sys_file_setattr
 470	common	listns				sys_listns
 471	nospu	rseq_slice_yield		sys_rseq_slice_yield
+472	nospu	sframe_register			sys_sframe_register
+473	nospu	sframe_unregister		sys_sframe_unregister
diff --git a/arch/s390/kernel/syscalls/syscall.tbl b/arch/s390/kernel/syscalls/syscall.tbl
index 09a7ef04d979..52519e2acdc8 100644
--- a/arch/s390/kernel/syscalls/syscall.tbl
+++ b/arch/s390/kernel/syscalls/syscall.tbl
@@ -398,3 +398,6 @@
 469	common	file_setattr			sys_file_setattr
 470	common	listns				sys_listns
 471	common	rseq_slice_yield		sys_rseq_slice_yield
+472	common	stacktrace_setup		sys_stacktrace_setup
+472	common	sframe_register			sys_sframe_register
+473	common	sframe_unregister		sys_sframe_unregister
diff --git a/arch/sh/kernel/syscalls/syscall.tbl b/arch/sh/kernel/syscalls/syscall.tbl
index 70b315cbe710..62ac7b1b4dd4 100644
--- a/arch/sh/kernel/syscalls/syscall.tbl
+++ b/arch/sh/kernel/syscalls/syscall.tbl
@@ -475,3 +475,5 @@
 469	common	file_setattr			sys_file_setattr
 470	common	listns				sys_listns
 471	common	rseq_slice_yield		sys_rseq_slice_yield
+472	common	sframe_register			sys_sframe_register
+473	common	sframe_unregister		sys_sframe_unregister
diff --git a/arch/sparc/kernel/syscalls/syscall.tbl b/arch/sparc/kernel/syscalls/syscall.tbl
index 7e71bf7fcd14..f92273ae608a 100644
--- a/arch/sparc/kernel/syscalls/syscall.tbl
+++ b/arch/sparc/kernel/syscalls/syscall.tbl
@@ -517,3 +517,5 @@
 469	common	file_setattr			sys_file_setattr
 470	common	listns				sys_listns
 471	common	rseq_slice_yield		sys_rseq_slice_yield
+472	common	sframe_register			sys_sframe_register
+473	common	sframe_unregister		sys_sframe_unregister
diff --git a/arch/x86/entry/syscalls/syscall_32.tbl b/arch/x86/entry/syscalls/syscall_32.tbl
index f832ebd2d79b..409a50df3b21 100644
--- a/arch/x86/entry/syscalls/syscall_32.tbl
+++ b/arch/x86/entry/syscalls/syscall_32.tbl
@@ -477,3 +477,5 @@
 469	i386	file_setattr		sys_file_setattr
 470	i386	listns			sys_listns
 471	i386	rseq_slice_yield	sys_rseq_slice_yield
+472	i386	sframe_register		sys_sframe_register
+473	i386	sframe_unregister	sys_sframe_unregister
diff --git a/arch/x86/entry/syscalls/syscall_64.tbl b/arch/x86/entry/syscalls/syscall_64.tbl
index 524155d655da..9b7c5a449751 100644
--- a/arch/x86/entry/syscalls/syscall_64.tbl
+++ b/arch/x86/entry/syscalls/syscall_64.tbl
@@ -396,6 +396,8 @@
 469	common	file_setattr		sys_file_setattr
 470	common	listns			sys_listns
 471	common	rseq_slice_yield	sys_rseq_slice_yield
+472	common	sframe_register		sys_sframe_register
+473	common	sframe_unregister	sys_sframe_unregister
 
 #
 # Due to a historical design error, certain syscalls are numbered differently
diff --git a/arch/xtensa/kernel/syscalls/syscall.tbl b/arch/xtensa/kernel/syscalls/syscall.tbl
index a9bca4e484de..037b8040f69d 100644
--- a/arch/xtensa/kernel/syscalls/syscall.tbl
+++ b/arch/xtensa/kernel/syscalls/syscall.tbl
@@ -442,3 +442,5 @@
 469	common	file_setattr			sys_file_setattr
 470	common	listns				sys_listns
 471	common	rseq_slice_yield		sys_rseq_slice_yield
+472	common	sframe_register'		sys_sframe_register
+473	common	sframe_unregister		sys_sframe_unregister
diff --git a/include/linux/syscalls.h b/include/linux/syscalls.h
index f5639d5ac331..992ccc401c5e 100644
--- a/include/linux/syscalls.h
+++ b/include/linux/syscalls.h
@@ -999,6 +999,8 @@ asmlinkage long sys_lsm_get_self_attr(unsigned int attr, struct lsm_ctx __user *
 asmlinkage long sys_lsm_set_self_attr(unsigned int attr, struct lsm_ctx __user *ctx,
 				      u32 size, u32 flags);
 asmlinkage long sys_lsm_list_modules(u64 __user *ids, u32 __user *size, u32 flags);
+asmlinkage long sys_sframe_register(void *data, unsigned int size);
+asmlinkage long sys_sframe_unregister(void *data, unsigned int size);
 
 /*
  * Architecture-specific system calls
diff --git a/include/uapi/asm-generic/unistd.h b/include/uapi/asm-generic/unistd.h
index a627acc8fb5f..17042d7e5e87 100644
--- a/include/uapi/asm-generic/unistd.h
+++ b/include/uapi/asm-generic/unistd.h
@@ -863,8 +863,13 @@ __SYSCALL(__NR_listns, sys_listns)
 #define __NR_rseq_slice_yield 471
 __SYSCALL(__NR_rseq_slice_yield, sys_rseq_slice_yield)
 
+#define __NR_sframe_register 472
+__SYSCALL(__NR_sframe_register, sys_sframe_register)
+#define __NR_sframe_unregister 473
+__SYSCALL(__NR_sframe_unregister, sys_sframe_unregister)
+
 #undef __NR_syscalls
-#define __NR_syscalls 472
+#define __NR_syscalls 474
 
 /*
  * 32 bit systems traditionally used different
diff --git a/include/uapi/linux/sframe.h b/include/uapi/linux/sframe.h
new file mode 100644
index 000000000000..137a2ebf91f4
--- /dev/null
+++ b/include/uapi/linux/sframe.h
@@ -0,0 +1,12 @@
+/* SPDX-License-Identifier: GPL-2.0+ WITH Linux-syscall-note */
+#ifndef _UAPI_LINUX_SFRAME_H
+#define _UAPI_LINUX_SFRAME_H
+
+struct sframe_setup {
+	unsigned long		sframe_start;
+	unsigned long		sframe_size;
+	unsigned long		text_start;
+	unsigned long		text_size;
+};
+
+#endif /* _UAPI_LINUX_SFRAME_H */
diff --git a/kernel/sys_ni.c b/kernel/sys_ni.c
index add3032da16f..eca5293f5d40 100644
--- a/kernel/sys_ni.c
+++ b/kernel/sys_ni.c
@@ -394,3 +394,6 @@ COND_SYSCALL(rseq_slice_yield);
 
 COND_SYSCALL(uretprobe);
 COND_SYSCALL(uprobe);
+
+COND_SYSCALL(sframe_register);
+COND_SYSCALL(sframe_unregister);
diff --git a/kernel/unwind/sframe.c b/kernel/unwind/sframe.c
index db88d993dff1..9956f1e3aba1 100644
--- a/kernel/unwind/sframe.c
+++ b/kernel/unwind/sframe.c
@@ -12,8 +12,10 @@
 #include <linux/mm.h>
 #include <linux/string_helpers.h>
 #include <linux/sframe.h>
+#include <linux/syscalls.h>
 #include <asm/unwind_user_sframe.h>
 #include <linux/unwind_user_types.h>
+#include <uapi/linux/sframe.h>
 
 #include "sframe.h"
 #include "sframe_debug.h"
@@ -842,9 +844,11 @@ static void sframe_free_srcu(struct rcu_head *rcu)
 static int __sframe_remove_section(struct mm_struct *mm,
 				   struct sframe_section *sec)
 {
-	if (!mtree_erase(&mm->sframe_mt, sec->text_start)) {
-		dbg_sec("mtree_erase failed: text=%lx\n", sec->text_start);
-		return -EINVAL;
+	scoped_guard(mmap_read_lock, mm) {
+		if (!mtree_erase(&mm->sframe_mt, sec->text_start)) {
+			dbg_sec("mtree_erase failed: text=%lx\n", sec->text_start);
+			return -EINVAL;
+		}
 	}
 
 	call_srcu(&sframe_srcu, &sec->rcu, sframe_free_srcu);
@@ -936,3 +940,56 @@ void sframe_free_mm(struct mm_struct *mm)
 
 	mtree_destroy(&mm->sframe_mt);
 }
+
+/**
+ * sys_sframe_register - register an address for user space stacktrace walking.
+ * @data: Structure of sframe data used to register the sframe section
+ * @size: The size of the given structure.
+ *
+ * This system call is used by dynamic library utilities to inform the kernel
+ * of meta data that it loaded that can be used by the kernel to know how
+ * to stack walk the given text locations.
+ *
+ * Return: 0 if successful, otherwise a negative error.
+ */
+SYSCALL_DEFINE2(sframe_register, __user struct sframe_setup *, data, unsigned int, size)
+{
+	struct sframe_setup sframe;
+
+	if (sizeof(sframe) != size)
+		return -EINVAL;
+
+	if (copy_from_user(&sframe, data, size))
+		return -EFAULT;
+
+	return sframe_add_section(sframe.sframe_start,
+				  sframe.sframe_start + sframe.sframe_size,
+				  sframe.text_start,
+				  sframe.text_start + sframe.text_size);
+}
+
+/**
+ * sys_sframe_unregister - unregister an sframe address
+ * @data: Structure of sframe data used to register the sframe section
+ * @size: The size of the given structure.
+ *
+ * The data->sframe_start is the only value that is used. The rest must
+ * be zero.
+ *
+ * Return: 0 if successful, otherwise a negative error.
+ */
+SYSCALL_DEFINE2(sframe_unregister, __user struct sframe_setup *, data, unsigned int, size)
+{
+	struct sframe_setup sframe;
+
+	if (sizeof(sframe) != size)
+		return -EINVAL;
+
+	if (copy_from_user(&sframe, data, size))
+		return -EFAULT;
+
+	if (sframe.sframe_size || sframe.text_start || sframe.text_size)
+		return -EINVAL;
+
+	return sframe_remove_section(sframe.sframe_start);
+}
diff --git a/scripts/syscall.tbl b/scripts/syscall.tbl
index 7a42b32b6577..46ec22b50042 100644
--- a/scripts/syscall.tbl
+++ b/scripts/syscall.tbl
@@ -412,3 +412,5 @@
 469	common	file_setattr			sys_file_setattr
 470	common	listns				sys_listns
 471	common	rseq_slice_yield		sys_rseq_slice_yield
+472	common	sframe_register			sys_sframe_register
+473	common	sframe_unregister		sys_sframe_unregister
-- 
2.53.0

Re: [PATCH] unwind: Add sframe_(un)register() system calls

[Jens Remus]

On 5/22/2026 12:35 AM, Steven Rostedt wrote:
> From: Steven Rostedt <rostedt@goodmis.org>
> 
> Add system calls to register and unregister sframes that can be used by
> dynamic linkers to tell the kernel where the sframe section is in memory
> for libraries it loads.

Why two separate system calls?  Can't that be one single stacktracectl?
Could they at least be non-sframe specific, e.g. stracktrace_register
and stracktrace_unregister, so that if one would implement e.g. unwind
user dwarf/eh_frame in the future one could pass ehframe_start and
ehframe_end in addition to sframe_start and sframe_end?

> 
> Both system calls take a pointer to a new structure:
> 
>   struct sframe_setup {
> 	unsigned long		sframe_start;
> 	unsigned long		sframe_size;
> 	unsigned long		text_start;
> 	unsigned long		text_size;
>   };
> 
> and a size of the passed in structure. If the system call needs to be
> extended, then the structure could be changed and the size of that
> structure will tell the kernel that it is the new version. If the kernel
> does not recognize the structure size, it will return -EINVAL.
> 
>   sframe_start - The virtual address of the sframe section
>   sframe_size  - The length of the sframe section
>   text_start   - the text section the sframe represents
>   test_size    - the length of the section
> 
> If other stack tracing functionality is added, it will require a new
> system call.
> 
> The unregister only needs the sframe_start and requires all the rest of
> the fields to be 0. In the future, if more can be done, then user space
> can update the other values and check the return code to see if the kernel
> supports it.
> 
> Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
> ---
> 
> Based on top of Jens patches here:
> 
>   https://lore.kernel.org/linux-trace-kernel/20260520154004.3845823-1-jremus@linux.ibm.com/
> 
> [ Note, I tested this with the same program from the RFC patch ]
> 
> Changes from RFC: https://patch.msgid.link/20260429114355.6c712e6a@gandalf.local.home
> 
> - Remove the ioctl() like system call for a unique system call for each
>   functionality. Right now there's two functionalities:
>    1. register sframe section
>    2. unregister sframe sections
> 
> - Added taking a lock around the mtree logic in __sframe_remove_section()
>   as Sashiko mentioned that there could be races from user space
>   registering and unregistering sframe sections at the same time.

Doesn't sframe_add_section() then also need likewise?

> 
> - Removed [RFC] from subject as I believe this is more likely the way
>   this system call will be done.

> diff --git a/include/linux/syscalls.h b/include/linux/syscalls.h

> @@ -999,6 +999,8 @@ asmlinkage long sys_lsm_get_self_attr(unsigned int attr, struct lsm_ctx __user *
>  asmlinkage long sys_lsm_set_self_attr(unsigned int attr, struct lsm_ctx __user *ctx,
>  				      u32 size, u32 flags);
>  asmlinkage long sys_lsm_list_modules(u64 __user *ids, u32 __user *size, u32 flags);
> +asmlinkage long sys_sframe_register(void *data, unsigned int size);
> +asmlinkage long sys_sframe_unregister(void *data, unsigned int size);
>  
>  /*
>   * Architecture-specific system calls


> diff --git a/include/uapi/linux/sframe.h b/include/uapi/linux/sframe.h

> @@ -0,0 +1,12 @@
> +/* SPDX-License-Identifier: GPL-2.0+ WITH Linux-syscall-note */
> +#ifndef _UAPI_LINUX_SFRAME_H
> +#define _UAPI_LINUX_SFRAME_H
> +
> +struct sframe_setup {
> +	unsigned long		sframe_start;
> +	unsigned long		sframe_size;
> +	unsigned long		text_start;
> +	unsigned long		text_size;
> +};
> +
> +#endif /* _UAPI_LINUX_SFRAME_H */

> diff --git a/kernel/unwind/sframe.c b/kernel/unwind/sframe.c

> @@ -842,9 +844,11 @@ static void sframe_free_srcu(struct rcu_head *rcu)
>  static int __sframe_remove_section(struct mm_struct *mm,
>  				   struct sframe_section *sec)
>  {
> -	if (!mtree_erase(&mm->sframe_mt, sec->text_start)) {
> -		dbg_sec("mtree_erase failed: text=%lx\n", sec->text_start);
> -		return -EINVAL;
> +	scoped_guard(mmap_read_lock, mm) {

Why is a read lock sufficient? Doesn't that allow multiple readers?
How does that prevent a concurrent modification of the mm->sframe_mt?

> +		if (!mtree_erase(&mm->sframe_mt, sec->text_start)) {
> +			dbg_sec("mtree_erase failed: text=%lx\n", sec->text_start);
> +			return -EINVAL;
> +		}

Is (or why not) likewise required in sframe_add_section() for the
mtree_insert_range()?

Wasn't the reported issue that while mt_for_each() in
sframe_remove_section() there could be concurrent mtree_erase() in
__sframe_remove_section() followed by mtree_insert_range() in
sframe_add_section(), so that the mt_for_each() could get confused?

>  	}
>  
>  	call_srcu(&sframe_srcu, &sec->rcu, sframe_free_srcu);
> @@ -936,3 +940,56 @@ void sframe_free_mm(struct mm_struct *mm)
>  
>  	mtree_destroy(&mm->sframe_mt);
>  }
> +
> +/**
> + * sys_sframe_register - register an address for user space stacktrace walking.
> + * @data: Structure of sframe data used to register the sframe section
> + * @size: The size of the given structure.
> + *
> + * This system call is used by dynamic library utilities to inform the kernel
> + * of meta data that it loaded that can be used by the kernel to know how
> + * to stack walk the given text locations.
> + *
> + * Return: 0 if successful, otherwise a negative error.
> + */
> +SYSCALL_DEFINE2(sframe_register, __user struct sframe_setup *, data, unsigned int, size)
> +{
> +	struct sframe_setup sframe;
> +
> +	if (sizeof(sframe) != size)
> +		return -EINVAL;
> +
> +	if (copy_from_user(&sframe, data, size))
> +		return -EFAULT;
> +
> +	return sframe_add_section(sframe.sframe_start,
> +				  sframe.sframe_start + sframe.sframe_size,
> +				  sframe.text_start,
> +				  sframe.text_start + sframe.text_size);
> +}
> +
> +/**
> + * sys_sframe_unregister - unregister an sframe address
> + * @data: Structure of sframe data used to register the sframe section
> + * @size: The size of the given structure.
> + *
> + * The data->sframe_start is the only value that is used. The rest must
> + * be zero.
> + *
> + * Return: 0 if successful, otherwise a negative error.
> + */
> +SYSCALL_DEFINE2(sframe_unregister, __user struct sframe_setup *, data, unsigned int, size)
> +{
> +	struct sframe_setup sframe;
> +
> +	if (sizeof(sframe) != size)
> +		return -EINVAL;
> +
> +	if (copy_from_user(&sframe, data, size))
> +		return -EFAULT;
> +
> +	if (sframe.sframe_size || sframe.text_start || sframe.text_size)
> +		return -EINVAL;
> +
> +	return sframe_remove_section(sframe.sframe_start);
> +}
Thanks and regards,
Jens
-- 
Jens Remus
Linux on Z Development (D3303)
jremus@de.ibm.com / jremus@linux.ibm.com

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

Re: [PATCH] unwind: Add sframe_(un)register() system calls

[Steven Rostedt]

On Fri, 22 May 2026 11:43:06 +0200
Jens Remus <jremus@linux.ibm.com> wrote:

> On 5/22/2026 12:35 AM, Steven Rostedt wrote:
> > From: Steven Rostedt <rostedt@goodmis.org>
> > 
> > Add system calls to register and unregister sframes that can be used by
> > dynamic linkers to tell the kernel where the sframe section is in memory
> > for libraries it loads.  
> 
> Why two separate system calls?  Can't that be one single stacktracectl?
> Could they at least be non-sframe specific, e.g. stracktrace_register
> and stracktrace_unregister, so that if one would implement e.g. unwind
> user dwarf/eh_frame in the future one could pass ehframe_start and
> ehframe_end in addition to sframe_start and sframe_end?

Talking with everyone at LSF/MM/BPF the consensus was to avoid an ioctl
like system call. Everyone hates them. They told me that a system call
should do one thing. They wanted a separate system call to register and to
unregister.

Note this also helps to see what the user is doing via monitoring via
ftrace, strace, and security wise via LSMs and seccomp.

> 
> > 
> > Both system calls take a pointer to a new structure:
> > 
> >   struct sframe_setup {
> > 	unsigned long		sframe_start;
> > 	unsigned long		sframe_size;
> > 	unsigned long		text_start;
> > 	unsigned long		text_size;
> >   };
> > 
> > and a size of the passed in structure. If the system call needs to be
> > extended, then the structure could be changed and the size of that
> > structure will tell the kernel that it is the new version. If the kernel
> > does not recognize the structure size, it will return -EINVAL.
> > 
> >   sframe_start - The virtual address of the sframe section
> >   sframe_size  - The length of the sframe section
> >   text_start   - the text section the sframe represents
> >   test_size    - the length of the section
> > 
> > If other stack tracing functionality is added, it will require a new
> > system call.
> > 
> > The unregister only needs the sframe_start and requires all the rest of
> > the fields to be 0. In the future, if more can be done, then user space
> > can update the other values and check the return code to see if the kernel
> > supports it.
> > 
> > Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
> > ---
> > 
> > Based on top of Jens patches here:
> > 
> >   https://lore.kernel.org/linux-trace-kernel/20260520154004.3845823-1-jremus@linux.ibm.com/
> > 
> > [ Note, I tested this with the same program from the RFC patch ]
> > 
> > Changes from RFC: https://patch.msgid.link/20260429114355.6c712e6a@gandalf.local.home
> > 
> > - Remove the ioctl() like system call for a unique system call for each
> >   functionality. Right now there's two functionalities:
> >    1. register sframe section
> >    2. unregister sframe sections
> > 
> > - Added taking a lock around the mtree logic in __sframe_remove_section()
> >   as Sashiko mentioned that there could be races from user space
> >   registering and unregistering sframe sections at the same time.  
> 
> Doesn't sframe_add_section() then also need likewise?

Ah, I saw the lock grabbed on the vma lookup. It should also be done for the
mtree_insert_range(). Thanks, will fix.

> 
> > 
> > - Removed [RFC] from subject as I believe this is more likely the way
> >   this system call will be done.  
> 
> > diff --git a/include/linux/syscalls.h b/include/linux/syscalls.h  
> 
> > @@ -999,6 +999,8 @@ asmlinkage long sys_lsm_get_self_attr(unsigned int attr, struct lsm_ctx __user *
> >  asmlinkage long sys_lsm_set_self_attr(unsigned int attr, struct lsm_ctx __user *ctx,
> >  				      u32 size, u32 flags);
> >  asmlinkage long sys_lsm_list_modules(u64 __user *ids, u32 __user *size, u32 flags);
> > +asmlinkage long sys_sframe_register(void *data, unsigned int size);
> > +asmlinkage long sys_sframe_unregister(void *data, unsigned int size);
> >  
> >  /*
> >   * Architecture-specific system calls  
> 
> 
> > diff --git a/include/uapi/linux/sframe.h b/include/uapi/linux/sframe.h  
> 
> > @@ -0,0 +1,12 @@
> > +/* SPDX-License-Identifier: GPL-2.0+ WITH Linux-syscall-note */
> > +#ifndef _UAPI_LINUX_SFRAME_H
> > +#define _UAPI_LINUX_SFRAME_H
> > +
> > +struct sframe_setup {
> > +	unsigned long		sframe_start;
> > +	unsigned long		sframe_size;
> > +	unsigned long		text_start;
> > +	unsigned long		text_size;
> > +};
> > +
> > +#endif /* _UAPI_LINUX_SFRAME_H */  
> 
> > diff --git a/kernel/unwind/sframe.c b/kernel/unwind/sframe.c  
> 
> > @@ -842,9 +844,11 @@ static void sframe_free_srcu(struct rcu_head *rcu)
> >  static int __sframe_remove_section(struct mm_struct *mm,
> >  				   struct sframe_section *sec)
> >  {
> > -	if (!mtree_erase(&mm->sframe_mt, sec->text_start)) {
> > -		dbg_sec("mtree_erase failed: text=%lx\n", sec->text_start);
> > -		return -EINVAL;
> > +	scoped_guard(mmap_read_lock, mm) {  
> 
> Why is a read lock sufficient? Doesn't that allow multiple readers?
> How does that prevent a concurrent modification of the mm->sframe_mt?

That was a cut and paste error. I meant to change it to a write lock, but
got distracted :-p Thanks, will fix.

> 
> > +		if (!mtree_erase(&mm->sframe_mt, sec->text_start)) {
> > +			dbg_sec("mtree_erase failed: text=%lx\n", sec->text_start);
> > +			return -EINVAL;
> > +		}  
> 
> Is (or why not) likewise required in sframe_add_section() for the
> mtree_insert_range()?
> 
> Wasn't the reported issue that while mt_for_each() in
> sframe_remove_section() there could be concurrent mtree_erase() in
> __sframe_remove_section() followed by mtree_insert_range() in
> sframe_add_section(), so that the mt_for_each() could get confused?

I'll take a closer look. But let me fix the obvious bugs first.

-- Steve

> 
> >  	}
> >  
> >  	call_srcu(&sframe_srcu, &sec->rcu, sframe_free_srcu);
> > @@ -936,3 +940,56 @@ void sframe_free_mm(struct mm_struct *mm)
> >  
> >  	mtree_destroy(&mm->sframe_mt);
> >  }

Re: [PATCH] unwind: Add sframe_(un)register() system calls

[Thomas Weißschuh]

On 2026-05-21 18:35:32-0400, Steven Rostedt wrote:
> From: Steven Rostedt <rostedt@goodmis.org>
> 
> Add system calls to register and unregister sframes that can be used by
> dynamic linkers to tell the kernel where the sframe section is in memory
> for libraries it loads.

How is this system call related to the prctl() with the same
functionality from Jens' series? I guess it will replace it,
but some explanation would be nice.

(...)

> diff --git a/include/linux/syscalls.h b/include/linux/syscalls.h
> index f5639d5ac331..992ccc401c5e 100644
> --- a/include/linux/syscalls.h
> +++ b/include/linux/syscalls.h
> @@ -999,6 +999,8 @@ asmlinkage long sys_lsm_get_self_attr(unsigned int attr, struct lsm_ctx __user *
>  asmlinkage long sys_lsm_set_self_attr(unsigned int attr, struct lsm_ctx __user *ctx,
>  				      u32 size, u32 flags);
>  asmlinkage long sys_lsm_list_modules(u64 __user *ids, u32 __user *size, u32 flags);
> +asmlinkage long sys_sframe_register(void *data, unsigned int size);
> +asmlinkage long sys_sframe_unregister(void *data, unsigned int size);

Why not use the actual structure here?

>  /*
>   * Architecture-specific system calls

(...)

> diff --git a/include/uapi/linux/sframe.h b/include/uapi/linux/sframe.h
> new file mode 100644
> index 000000000000..137a2ebf91f4
> --- /dev/null
> +++ b/include/uapi/linux/sframe.h
> @@ -0,0 +1,12 @@
> +/* SPDX-License-Identifier: GPL-2.0+ WITH Linux-syscall-note */
> +#ifndef _UAPI_LINUX_SFRAME_H
> +#define _UAPI_LINUX_SFRAME_H
> +
> +struct sframe_setup {
> +	unsigned long		sframe_start;
> +	unsigned long		sframe_size;
> +	unsigned long		text_start;
> +	unsigned long		text_size;
> +};

This will break for compat processes, as they use a different 'unsigned
long' than the host kernel. Maybe just use __u64.

> +
> +#endif /* _UAPI_LINUX_SFRAME_H */

(...)

> +/**
> + * sys_sframe_register - register an address for user space stacktrace walking.
> + * @data: Structure of sframe data used to register the sframe section
> + * @size: The size of the given structure.
> + *
> + * This system call is used by dynamic library utilities to inform the kernel
> + * of meta data that it loaded that can be used by the kernel to know how
> + * to stack walk the given text locations.
> + *
> + * Return: 0 if successful, otherwise a negative error.
> + */
> +SYSCALL_DEFINE2(sframe_register, __user struct sframe_setup *, data, unsigned int, size)

AFAIK the normal place for the '__user' is right before '*':

	struct sframe_setup __user *, data,

Use __kernel_size_t for 'size'?

> +{
> +	struct sframe_setup sframe;

(...)

Re: [PATCH] unwind: Add sframe_(un)register() system calls

[Steven Rostedt]

On Fri, 22 May 2026 16:36:56 +0200
Thomas Weißschuh <thomas@t-8ch.de> wrote:

> On 2026-05-21 18:35:32-0400, Steven Rostedt wrote:
> > From: Steven Rostedt <rostedt@goodmis.org>
> > 
> > Add system calls to register and unregister sframes that can be used by
> > dynamic linkers to tell the kernel where the sframe section is in memory
> > for libraries it loads.  
> 
> How is this system call related to the prctl() with the same
> functionality from Jens' series? I guess it will replace it,
> but some explanation would be nice.

I thought the patch with the prctl() stated it was for debug purposes only.
From the change log:

[
  This adds an interface for prctl() for testing loading of sframes for
  libraries. But this interface should really be a system call. This patch
  is for testing purposes only and should not be applied to mainline.
]

Hence I didn't think there needs to be any explanation. The prctl() patch
should never be applied upstream.

> 
> (...)
> 
> > diff --git a/include/linux/syscalls.h b/include/linux/syscalls.h
> > index f5639d5ac331..992ccc401c5e 100644
> > --- a/include/linux/syscalls.h
> > +++ b/include/linux/syscalls.h
> > @@ -999,6 +999,8 @@ asmlinkage long sys_lsm_get_self_attr(unsigned int attr, struct lsm_ctx __user *
> >  asmlinkage long sys_lsm_set_self_attr(unsigned int attr, struct lsm_ctx __user *ctx,
> >  				      u32 size, u32 flags);
> >  asmlinkage long sys_lsm_list_modules(u64 __user *ids, u32 __user *size, u32 flags);
> > +asmlinkage long sys_sframe_register(void *data, unsigned int size);
> > +asmlinkage long sys_sframe_unregister(void *data, unsigned int size);  
> 
> Why not use the actual structure here?

Yeah, I was somewhat lazy here to make sure that this was the direction we
want to go. I just need to add a structure pointer reference at the top of
that file.

Will update in v2.

> 
> >  /*
> >   * Architecture-specific system calls  
> 
> (...)
> 
> > diff --git a/include/uapi/linux/sframe.h b/include/uapi/linux/sframe.h
> > new file mode 100644
> > index 000000000000..137a2ebf91f4
> > --- /dev/null
> > +++ b/include/uapi/linux/sframe.h
> > @@ -0,0 +1,12 @@
> > +/* SPDX-License-Identifier: GPL-2.0+ WITH Linux-syscall-note */
> > +#ifndef _UAPI_LINUX_SFRAME_H
> > +#define _UAPI_LINUX_SFRAME_H
> > +
> > +struct sframe_setup {
> > +	unsigned long		sframe_start;
> > +	unsigned long		sframe_size;
> > +	unsigned long		text_start;
> > +	unsigned long		text_size;
> > +};  
> 
> This will break for compat processes, as they use a different 'unsigned
> long' than the host kernel. Maybe just use __u64.

I'll update it. I was thinking we wouldn't support compat, but in case we
decide we should forcing the size is better than being architecture
specific.

> 
> > +
> > +#endif /* _UAPI_LINUX_SFRAME_H */  
> 
> (...)
> 
> > +/**
> > + * sys_sframe_register - register an address for user space stacktrace walking.
> > + * @data: Structure of sframe data used to register the sframe section
> > + * @size: The size of the given structure.
> > + *
> > + * This system call is used by dynamic library utilities to inform the kernel
> > + * of meta data that it loaded that can be used by the kernel to know how
> > + * to stack walk the given text locations.
> > + *
> > + * Return: 0 if successful, otherwise a negative error.
> > + */
> > +SYSCALL_DEFINE2(sframe_register, __user struct sframe_setup *, data, unsigned int, size)  
> 
> AFAIK the normal place for the '__user' is right before '*':
> 
> 	struct sframe_setup __user *, data,

Will update.

> 
> Use __kernel_size_t for 'size'?

Looking at the history of the accept() system call that started with int
and then wanted size_t, then changed to socklen_t, I guess there's
precedence to use __kernel_size_t.

Will update.

Thanks!

-- Steve