[PATCH v2 0/4] seccomp: remove the 'sd' argument from __secure_computing()

[PATCH v2 0/4] seccomp: remove the 'sd' argument from __secure_computing()

[Oleg Nesterov]

Hello,

Link to v1: https://lore.kernel.org/all/20250120134409.GA21241@redhat.com/
Only 2/4 was changed, please see interdiff at the end.

I've included the acks I got on 1/4, 3/4, and 4/4 (thanks!).

Oleg.
---

 arch/mips/kernel/ptrace.c           | 20 ++-----------------
 arch/powerpc/kernel/ptrace/ptrace.c |  2 +-
 include/linux/seccomp.h             | 12 ++++--------
 kernel/entry/common.c               |  2 +-
 kernel/seccomp.c                    | 39 ++++++++++++++++++-------------------
 5 files changed, 27 insertions(+), 48 deletions(-)

-------------------------------------------------------------------------------
diff --git a/include/linux/seccomp.h b/include/linux/seccomp.h
index 6125baa96b76..9b959972bf4a 100644
--- a/include/linux/seccomp.h
+++ b/include/linux/seccomp.h
@@ -22,8 +22,9 @@
 #include <linux/atomic.h>
 #include <asm/seccomp.h>
 
-#ifdef CONFIG_HAVE_ARCH_SECCOMP_FILTER
 extern int __secure_computing(void);
+
+#ifdef CONFIG_HAVE_ARCH_SECCOMP_FILTER
 static inline int secure_computing(void)
 {
 	if (unlikely(test_syscall_work(SECCOMP)))
@@ -32,7 +33,6 @@ static inline int secure_computing(void)
 }
 #else
 extern void secure_computing_strict(int this_syscall);
-static inline int __secure_computing(void) { return 0; }
 #endif
 
 extern long prctl_get_seccomp(void);
diff --git a/kernel/seccomp.c b/kernel/seccomp.c
index 75e293d3c1a1..4bd2eb50f77b 100644
--- a/kernel/seccomp.c
+++ b/kernel/seccomp.c
@@ -29,13 +29,11 @@
 #include <linux/syscalls.h>
 #include <linux/sysctl.h>
 
+#include <asm/syscall.h>
+
 /* Not exposed in headers: strictly internal use only. */
 #define SECCOMP_MODE_DEAD	(SECCOMP_MODE_FILTER + 1)
 
-#ifdef CONFIG_HAVE_ARCH_SECCOMP_FILTER
-#include <asm/syscall.h>
-#endif
-
 #ifdef CONFIG_SECCOMP_FILTER
 #include <linux/file.h>
 #include <linux/filter.h>
@@ -1062,6 +1060,13 @@ void secure_computing_strict(int this_syscall)
 	else
 		BUG();
 }
+int __secure_computing(void)
+{
+	int this_syscall = syscall_get_nr(current, current_pt_regs());
+
+	secure_computing_strict(this_syscall);
+	return 0;
+}
 #else
 
 #ifdef CONFIG_SECCOMP_FILTER

[PATCH v2 1/4] seccomp/mips: change syscall_trace_enter() to use secure_computing()

[Oleg Nesterov]

arch/mips/Kconfig selects HAVE_ARCH_SECCOMP_FILTER so syscall_trace_enter()
can just use __secure_computing(NULL) and rely on populate_seccomp_data(sd)
and "sd == NULL" checks in __secure_computing(sd) paths.

With the change above syscall_trace_enter() can just use secure_computing()
and avoid #ifdef + test_thread_flag(TIF_SECCOMP). CONFIG_GENERIC_ENTRY is
not defined, so test_syscall_work(SECCOMP) will check TIF_SECCOMP.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Reviewed-by: Kees Cook <kees@kernel.org>
---
 arch/mips/kernel/ptrace.c | 20 ++------------------
 1 file changed, 2 insertions(+), 18 deletions(-)

diff --git a/arch/mips/kernel/ptrace.c b/arch/mips/kernel/ptrace.c
index 61503a36067e..f7107479c7fa 100644
--- a/arch/mips/kernel/ptrace.c
+++ b/arch/mips/kernel/ptrace.c
@@ -1326,24 +1326,8 @@ asmlinkage long syscall_trace_enter(struct pt_regs *regs)
 			return -1;
 	}
 
-#ifdef CONFIG_SECCOMP
-	if (unlikely(test_thread_flag(TIF_SECCOMP))) {
-		int ret, i;
-		struct seccomp_data sd;
-		unsigned long args[6];
-
-		sd.nr = current_thread_info()->syscall;
-		sd.arch = syscall_get_arch(current);
-		syscall_get_arguments(current, regs, args);
-		for (i = 0; i < 6; i++)
-			sd.args[i] = args[i];
-		sd.instruction_pointer = KSTK_EIP(current);
-
-		ret = __secure_computing(&sd);
-		if (ret == -1)
-			return ret;
-	}
-#endif
+	if (secure_computing())
+		return -1;
 
 	if (unlikely(test_thread_flag(TIF_SYSCALL_TRACEPOINT)))
 		trace_sys_enter(regs, regs->regs[2]);
-- 
2.25.1.362.g51ebf55

[PATCH v2 2/4] seccomp: fix the __secure_computing() stub for !HAVE_ARCH_SECCOMP_FILTER

[Oleg Nesterov]

Depending on CONFIG_HAVE_ARCH_SECCOMP_FILTER, __secure_computing(NULL)
will crash or not. This is not consistent/safe, especially considering
that after the previous change __secure_computing(sd) is always called
with sd == NULL.

Fortunately, if CONFIG_HAVE_ARCH_SECCOMP_FILTER=n, __secure_computing()
has no callers, these architectures use secure_computing_strict(). Yet
it make sense make __secure_computing(NULL) safe in this case.

Note also that with this change we can unexport secure_computing_strict()
and change the current callers to use __secure_computing(NULL).

Fixes: 8cf8dfceebda ("seccomp: Stub for !HAVE_ARCH_SECCOMP_FILTER")
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
---
 include/linux/seccomp.h |  8 ++------
 kernel/seccomp.c        | 14 ++++++++++----
 2 files changed, 12 insertions(+), 10 deletions(-)

diff --git a/include/linux/seccomp.h b/include/linux/seccomp.h
index e45531455d3b..d55949071c30 100644
--- a/include/linux/seccomp.h
+++ b/include/linux/seccomp.h
@@ -22,8 +22,9 @@
 #include <linux/atomic.h>
 #include <asm/seccomp.h>
 
-#ifdef CONFIG_HAVE_ARCH_SECCOMP_FILTER
 extern int __secure_computing(const struct seccomp_data *sd);
+
+#ifdef CONFIG_HAVE_ARCH_SECCOMP_FILTER
 static inline int secure_computing(void)
 {
 	if (unlikely(test_syscall_work(SECCOMP)))
@@ -32,11 +33,6 @@ static inline int secure_computing(void)
 }
 #else
 extern void secure_computing_strict(int this_syscall);
-static inline int __secure_computing(const struct seccomp_data *sd)
-{
-	secure_computing_strict(sd->nr);
-	return 0;
-}
 #endif
 
 extern long prctl_get_seccomp(void);
diff --git a/kernel/seccomp.c b/kernel/seccomp.c
index 385d48293a5f..327b7b486f1c 100644
--- a/kernel/seccomp.c
+++ b/kernel/seccomp.c
@@ -29,13 +29,11 @@
 #include <linux/syscalls.h>
 #include <linux/sysctl.h>
 
+#include <asm/syscall.h>
+
 /* Not exposed in headers: strictly internal use only. */
 #define SECCOMP_MODE_DEAD	(SECCOMP_MODE_FILTER + 1)
 
-#ifdef CONFIG_HAVE_ARCH_SECCOMP_FILTER
-#include <asm/syscall.h>
-#endif
-
 #ifdef CONFIG_SECCOMP_FILTER
 #include <linux/file.h>
 #include <linux/filter.h>
@@ -1062,6 +1060,14 @@ void secure_computing_strict(int this_syscall)
 	else
 		BUG();
 }
+int __secure_computing(const struct seccomp_data *sd)
+{
+	int this_syscall = sd ? sd->nr :
+		syscall_get_nr(current, current_pt_regs());
+
+	secure_computing_strict(this_syscall);
+	return 0;
+}
 #else
 
 #ifdef CONFIG_SECCOMP_FILTER
-- 
2.25.1.362.g51ebf55

Re: [PATCH v2 2/4] seccomp: fix the __secure_computing() stub for !HAVE_ARCH_SECCOMP_FILTER

[Linus Walleij]

On Tue, Jan 28, 2025 at 4:03 PM Oleg Nesterov <oleg@redhat.com> wrote:

> Depending on CONFIG_HAVE_ARCH_SECCOMP_FILTER, __secure_computing(NULL)
> will crash or not. This is not consistent/safe, especially considering
> that after the previous change __secure_computing(sd) is always called
> with sd == NULL.
>
> Fortunately, if CONFIG_HAVE_ARCH_SECCOMP_FILTER=n, __secure_computing()
> has no callers, these architectures use secure_computing_strict(). Yet
> it make sense make __secure_computing(NULL) safe in this case.
>
> Note also that with this change we can unexport secure_computing_strict()
> and change the current callers to use __secure_computing(NULL).
>
> Fixes: 8cf8dfceebda ("seccomp: Stub for !HAVE_ARCH_SECCOMP_FILTER")
> Signed-off-by: Oleg Nesterov <oleg@redhat.com>

I had no idea it was this complex, thanks a lot for looking into this Oleg!
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>

Yours,
Linus Walleij

[PATCH v2 3/4] seccomp: remove the 'sd' argument from __secure_computing()

[Oleg Nesterov]

After the previous changes 'sd' is always NULL.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Reviewed-by: Kees Cook <kees@kernel.org>
---
 arch/powerpc/kernel/ptrace/ptrace.c |  2 +-
 include/linux/seccomp.h             |  6 +++---
 kernel/entry/common.c               |  2 +-
 kernel/seccomp.c                    | 12 +++++-------
 4 files changed, 10 insertions(+), 12 deletions(-)

diff --git a/arch/powerpc/kernel/ptrace/ptrace.c b/arch/powerpc/kernel/ptrace/ptrace.c
index 727ed4a14545..c6997df63287 100644
--- a/arch/powerpc/kernel/ptrace/ptrace.c
+++ b/arch/powerpc/kernel/ptrace/ptrace.c
@@ -215,7 +215,7 @@ static int do_seccomp(struct pt_regs *regs)
 	 * have already loaded -ENOSYS into r3, or seccomp has put
 	 * something else in r3 (via SECCOMP_RET_ERRNO/TRACE).
 	 */
-	if (__secure_computing(NULL))
+	if (__secure_computing())
 		return -1;
 
 	/*
diff --git a/include/linux/seccomp.h b/include/linux/seccomp.h
index d55949071c30..9b959972bf4a 100644
--- a/include/linux/seccomp.h
+++ b/include/linux/seccomp.h
@@ -22,13 +22,13 @@
 #include <linux/atomic.h>
 #include <asm/seccomp.h>
 
-extern int __secure_computing(const struct seccomp_data *sd);
+extern int __secure_computing(void);
 
 #ifdef CONFIG_HAVE_ARCH_SECCOMP_FILTER
 static inline int secure_computing(void)
 {
 	if (unlikely(test_syscall_work(SECCOMP)))
-		return  __secure_computing(NULL);
+		return  __secure_computing();
 	return 0;
 }
 #else
@@ -54,7 +54,7 @@ static inline int secure_computing(void) { return 0; }
 #else
 static inline void secure_computing_strict(int this_syscall) { return; }
 #endif
-static inline int __secure_computing(const struct seccomp_data *sd) { return 0; }
+static inline int __secure_computing(void) { return 0; }
 
 static inline long prctl_get_seccomp(void)
 {
diff --git a/kernel/entry/common.c b/kernel/entry/common.c
index e33691d5adf7..20154572ede9 100644
--- a/kernel/entry/common.c
+++ b/kernel/entry/common.c
@@ -49,7 +49,7 @@ long syscall_trace_enter(struct pt_regs *regs, long syscall,
 
 	/* Do seccomp after ptrace, to catch any tracer changes. */
 	if (work & SYSCALL_WORK_SECCOMP) {
-		ret = __secure_computing(NULL);
+		ret = __secure_computing();
 		if (ret == -1L)
 			return ret;
 	}
diff --git a/kernel/seccomp.c b/kernel/seccomp.c
index 327b7b486f1c..281e853bae8c 100644
--- a/kernel/seccomp.c
+++ b/kernel/seccomp.c
@@ -1060,10 +1060,9 @@ void secure_computing_strict(int this_syscall)
 	else
 		BUG();
 }
-int __secure_computing(const struct seccomp_data *sd)
+int __secure_computing(void)
 {
-	int this_syscall = sd ? sd->nr :
-		syscall_get_nr(current, current_pt_regs());
+	int this_syscall = syscall_get_nr(current, current_pt_regs());
 
 	secure_computing_strict(this_syscall);
 	return 0;
@@ -1353,7 +1352,7 @@ static int __seccomp_filter(int this_syscall, const struct seccomp_data *sd,
 }
 #endif
 
-int __secure_computing(const struct seccomp_data *sd)
+int __secure_computing(void)
 {
 	int mode = current->seccomp.mode;
 	int this_syscall;
@@ -1362,15 +1361,14 @@ int __secure_computing(const struct seccomp_data *sd)
 	    unlikely(current->ptrace & PT_SUSPEND_SECCOMP))
 		return 0;
 
-	this_syscall = sd ? sd->nr :
-		syscall_get_nr(current, current_pt_regs());
+	this_syscall = syscall_get_nr(current, current_pt_regs());
 
 	switch (mode) {
 	case SECCOMP_MODE_STRICT:
 		__secure_computing_strict(this_syscall);  /* may call do_exit */
 		return 0;
 	case SECCOMP_MODE_FILTER:
-		return __seccomp_filter(this_syscall, sd, false);
+		return __seccomp_filter(this_syscall, NULL, false);
 	/* Surviving SECCOMP_RET_KILL_* must be proactively impossible. */
 	case SECCOMP_MODE_DEAD:
 		WARN_ON_ONCE(1);
-- 
2.25.1.362.g51ebf55

[PATCH v2 4/4] seccomp: remove the 'sd' argument from __seccomp_filter()

[Oleg Nesterov]

After the previous change 'sd' is always NULL.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Reviewed-by: Kees Cook <kees@kernel.org>
---
 kernel/seccomp.c | 21 ++++++++-------------
 1 file changed, 8 insertions(+), 13 deletions(-)

diff --git a/kernel/seccomp.c b/kernel/seccomp.c
index 281e853bae8c..4bd2eb50f77b 100644
--- a/kernel/seccomp.c
+++ b/kernel/seccomp.c
@@ -1218,13 +1218,12 @@ static int seccomp_do_user_notification(int this_syscall,
 	return -1;
 }
 
-static int __seccomp_filter(int this_syscall, const struct seccomp_data *sd,
-			    const bool recheck_after_trace)
+static int __seccomp_filter(int this_syscall, const bool recheck_after_trace)
 {
 	u32 filter_ret, action;
+	struct seccomp_data sd;
 	struct seccomp_filter *match = NULL;
 	int data;
-	struct seccomp_data sd_local;
 
 	/*
 	 * Make sure that any changes to mode from another thread have
@@ -1232,12 +1231,9 @@ static int __seccomp_filter(int this_syscall, const struct seccomp_data *sd,
 	 */
 	smp_rmb();
 
-	if (!sd) {
-		populate_seccomp_data(&sd_local);
-		sd = &sd_local;
-	}
+	populate_seccomp_data(&sd);
 
-	filter_ret = seccomp_run_filters(sd, &match);
+	filter_ret = seccomp_run_filters(&sd, &match);
 	data = filter_ret & SECCOMP_RET_DATA;
 	action = filter_ret & SECCOMP_RET_ACTION_FULL;
 
@@ -1295,13 +1291,13 @@ static int __seccomp_filter(int this_syscall, const struct seccomp_data *sd,
 		 * a reload of all registers. This does not goto skip since
 		 * a skip would have already been reported.
 		 */
-		if (__seccomp_filter(this_syscall, NULL, true))
+		if (__seccomp_filter(this_syscall, true))
 			return -1;
 
 		return 0;
 
 	case SECCOMP_RET_USER_NOTIF:
-		if (seccomp_do_user_notification(this_syscall, match, sd))
+		if (seccomp_do_user_notification(this_syscall, match, &sd))
 			goto skip;
 
 		return 0;
@@ -1343,8 +1339,7 @@ static int __seccomp_filter(int this_syscall, const struct seccomp_data *sd,
 	return -1;
 }
 #else
-static int __seccomp_filter(int this_syscall, const struct seccomp_data *sd,
-			    const bool recheck_after_trace)
+static int __seccomp_filter(int this_syscall, const bool recheck_after_trace)
 {
 	BUG();
 
@@ -1368,7 +1363,7 @@ int __secure_computing(void)
 		__secure_computing_strict(this_syscall);  /* may call do_exit */
 		return 0;
 	case SECCOMP_MODE_FILTER:
-		return __seccomp_filter(this_syscall, NULL, false);
+		return __seccomp_filter(this_syscall, false);
 	/* Surviving SECCOMP_RET_KILL_* must be proactively impossible. */
 	case SECCOMP_MODE_DEAD:
 		WARN_ON_ONCE(1);
-- 
2.25.1.362.g51ebf55

Re: [PATCH v2 0/4] seccomp: remove the 'sd' argument from __secure_computing()

[Kees Cook]

On Tue, 28 Jan 2025 16:02:28 +0100, Oleg Nesterov wrote:
> Link to v1: https://lore.kernel.org/all/20250120134409.GA21241@redhat.com/
> Only 2/4 was changed, please see interdiff at the end.
> 
> I've included the acks I got on 1/4, 3/4, and 4/4 (thanks!).
> 
> Oleg.
> 
> [...]

Applied to for-next/seccomp, thanks!

[0/4] seccomp: remove the 'sd' argument from __secure_computing()
      https://git.kernel.org/kees/c/1027cd8084bb
[1/4] seccomp/mips: change syscall_trace_enter() to use secure_computing()
      https://git.kernel.org/kees/c/0fe1ebf3f056
[2/4] seccomp: fix the __secure_computing() stub for !HAVE_ARCH_SECCOMP_FILTER
      https://git.kernel.org/kees/c/b37778bec82b
[3/4] seccomp: remove the 'sd' argument from __secure_computing()
      https://git.kernel.org/kees/c/1027cd8084bb
[4/4] seccomp: remove the 'sd' argument from __seccomp_filter()
      https://git.kernel.org/kees/c/e1cec5107c39

Take care,

-- 
Kees Cook