public inbox for linux-trace-kernel@vger.kernel.org
 help / color / mirror / Atom feed
From: Jens Remus <jremus@linux.ibm.com>
To: linux-kernel@vger.kernel.org, linux-trace-kernel@vger.kernel.org,
	linux-s390@vger.kernel.org, bpf@vger.kernel.org, x86@kernel.org,
	Steven Rostedt <rostedt@kernel.org>
Cc: Jens Remus <jremus@linux.ibm.com>,
	Heiko Carstens <hca@linux.ibm.com>,
	Vasily Gorbik <gor@linux.ibm.com>,
	Ilya Leoshkevich <iii@linux.ibm.com>,
	Josh Poimboeuf <jpoimboe@kernel.org>,
	Masami Hiramatsu <mhiramat@kernel.org>,
	Mathieu Desnoyers <mathieu.desnoyers@efficios.com>,
	Peter Zijlstra <peterz@infradead.org>,
	Ingo Molnar <mingo@kernel.org>, Jiri Olsa <jolsa@kernel.org>,
	Arnaldo Carvalho de Melo <acme@kernel.org>,
	Namhyung Kim <namhyung@kernel.org>,
	Thomas Gleixner <tglx@linutronix.de>,
	Andrii Nakryiko <andrii@kernel.org>,
	Indu Bhagat <indu.bhagat@oracle.com>,
	"Jose E. Marchesi" <jemarch@gnu.org>,
	Beau Belgrave <beaub@linux.microsoft.com>,
	Linus Torvalds <torvalds@linux-foundation.org>,
	Andrew Morton <akpm@linux-foundation.org>,
	Florian Weimer <fweimer@redhat.com>, Kees Cook <kees@kernel.org>,
	"Carlos O'Donell" <codonell@redhat.com>,
	Sam James <sam@gentoo.org>, Dylan Hatch <dylanbhatch@google.com>
Subject: [PATCH v4 12/12] s390/unwind_user/fp: Enable back chain unwinding of user space
Date: Tue, 27 Jan 2026 16:19:25 +0100	[thread overview]
Message-ID: <20260127151926.2805123-13-jremus@linux.ibm.com> (raw)
In-Reply-To: <20260127151926.2805123-1-jremus@linux.ibm.com>

Unwinding of user space using frame pointer (FP) is virtually impossible
on s390 for the following reasons:  The s390 64-bit (s390x) ELF ABI [1]
does only designate a "preferred" FP register and does not mandate fixed
FP and return address (RA) stack save slots.  Therefore neither the FP
register nor the FP/RA stack save slot offsets from CFA are known.
Compilers, such as GCC and Clang, do not necessarily setup a FP register
early in the function prologue, even not with compiler option
-fno-omit-frame-pointer.  Therefore the CFA offset from FP register is
not known.

This could be resolved by having compiler option -no-omit-frame-pointer
enforce all of the following:  Use the preferred FP register 11 as frame
pointer, use fixed FP/RA stack slot offsets from CFA (e.g. -72 for FP
and -48 for RA), and setup the FP register immediately after saving the
call saved registers.

Fortunately s390 provides an alternative to frame pointer:  back chain,
which can be enabled using s390-specific compiler option -mbackchain.
The back chain is very similar to a frame pointer on the stack.

Leverage the unwind user fp infrastructure to enable unwinding of user
space using back chain.  Enable HAVE_UNWIND_USER_FP and provide a s390-
specific implementation of unwind_user_fp_get_frame(), which uses the
back chain.

Signed-off-by: Jens Remus <jremus@linux.ibm.com>
---

Notes (jremus):
    Changes in v4:
    - Remove flawed heuristic to detect if topmost IP in early prologue.
      While it may resolve the caller getting skipped it may erroneously
      inject a callee as caller.
    - Fix outermost frame indication.
    - Adjust to flexible CFA and FP/RA rules.
    - Remove superfluous unwind_user_fp_get_frame define.
    
    Changes in RFC v3:
    - New patch.  Implement unwind user fp using back chain on s390. Reuses
      logic from RFC v2 patch "unwind_user/backchain: Introduce back chain
      user space unwinding". (Josh)

 arch/s390/Kconfig                   |  1 +
 arch/s390/include/asm/unwind_user.h | 62 +++++++++++++++++++++++++++++
 2 files changed, 63 insertions(+)

diff --git a/arch/s390/Kconfig b/arch/s390/Kconfig
index 063f0c857600..5f7e83ba54b2 100644
--- a/arch/s390/Kconfig
+++ b/arch/s390/Kconfig
@@ -250,6 +250,7 @@ config S390
 	select HAVE_SOFTIRQ_ON_OWN_STACK
 	select HAVE_STACKPROTECTOR if CC_HAS_STACKPROTECTOR_GLOBAL
 	select HAVE_SYSCALL_TRACEPOINTS
+	select HAVE_UNWIND_USER_FP
 	select HAVE_UNWIND_USER_SFRAME
 	select HAVE_VIRT_CPU_ACCOUNTING
 	select HAVE_VIRT_CPU_ACCOUNTING_IDLE
diff --git a/arch/s390/include/asm/unwind_user.h b/arch/s390/include/asm/unwind_user.h
index 941aa3f0f70f..a7b97ea01c26 100644
--- a/arch/s390/include/asm/unwind_user.h
+++ b/arch/s390/include/asm/unwind_user.h
@@ -3,8 +3,12 @@
 #define _ASM_S390_UNWIND_USER_H
 
 #include <linux/sched/task_stack.h>
+#include <linux/security.h>
 #include <linux/types.h>
+#include <asm/asm-offsets.h>
 #include <asm/fpu.h>
+#include <asm/stacktrace.h>
+#include <linux/unwind_user_types.h>
 
 #ifdef CONFIG_UNWIND_USER
 
@@ -65,6 +69,64 @@ static inline int arch_unwind_user_get_reg(unsigned long *val,
 
 #endif /* CONFIG_UNWIND_USER */
 
+#ifdef CONFIG_HAVE_UNWIND_USER_FP
+
+static inline bool ip_within_vdso(unsigned long ip)
+{
+	return in_range(ip, current->mm->context.vdso_base, vdso_text_size());
+}
+
+static inline int unwind_user_fp_get_frame(struct unwind_user_state *state,
+					   struct unwind_user_frame *frame)
+{
+	struct stack_frame_user __user *sf;
+	unsigned long __user *ra_addr;
+	unsigned long sp;
+
+	sf = (void __user *)state->sp;
+	if (__get_user(sp, (unsigned long __user *)&sf->back_chain))
+		return -EINVAL;
+	if (!sp && ip_within_vdso(state->ip)) {
+		/*
+		 * Assume non-standard vDSO user wrapper stack frame.
+		 * See vDSO user wrapper code for details.
+		 */
+		struct stack_frame_vdso_wrapper *sf_vdso = (void __user *)sf;
+
+		ra_addr = (unsigned long __user *)&sf_vdso->return_address;
+		sf = (void __user *)((unsigned long)sf + STACK_FRAME_VDSO_OVERHEAD);
+		if (__get_user(sp, (unsigned long __user *)&sf->back_chain))
+			return -EINVAL;
+	} else if (!sp) {
+		/*
+		 * Assume outermost frame reached. unwind_user_next_common()
+		 * disregards all other fields in outermost frame.
+		 */
+		frame->outermost = true;
+		return 0;
+	} else {
+		/*
+		 * Assume IP past prologue and new stack frame allocated.
+		 * Follow back chain, which then equals the SP at entry.
+		 * Skips caller if wrong in topmost frame.
+		 */
+		sf = (void __user *)sp;
+		ra_addr = (unsigned long __user *)&sf->gprs[8];
+	}
+
+	frame->cfa.rule = UNWIND_USER_CFA_RULE_SP_OFFSET;
+	frame->cfa.offset = sp - state->sp + 160;
+	frame->sp_off = -160;
+	frame->fp.rule = UNWIND_USER_RULE_ZERO;	/* Cannot unwind FP. */
+	frame->ra.rule = UNWIND_USER_RULE_CFA_OFFSET_DEREF;
+	frame->ra.offset = (unsigned long)ra_addr - (state->sp + frame->cfa.offset);
+	frame->outermost = false;
+
+	return 0;
+}
+
+#endif /* CONFIG_HAVE_UNWIND_USER_FP */
+
 #include <asm-generic/unwind_user.h>
 
 #endif /* _ASM_S390_UNWIND_USER_H */
-- 
2.51.0


      parent reply	other threads:[~2026-01-27 15:20 UTC|newest]

Thread overview: 13+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-01-27 15:19 [PATCH v4 00/12] s390: SFrame user space unwinding Jens Remus
2026-01-27 15:19 ` [PATCH v4 01/12] s390: asm/dwarf.h should only be included in assembly files Jens Remus
2026-01-27 15:19 ` [PATCH v4 02/12] s390/vdso: Avoid emitting DWARF CFI for non-vDSO Jens Remus
2026-01-27 15:19 ` [PATCH v4 03/12] s390/vdso: Keep function symbols in vDSO Jens Remus
2026-01-27 15:19 ` [PATCH v4 04/12] s390/vdso: Enable SFrame V3 generation " Jens Remus
2026-01-27 15:19 ` [PATCH v4 05/12] unwind_user: Enable archs that define CFA = SP_callsite + offset Jens Remus
2026-01-27 15:19 ` [PATCH v4 06/12] unwind_user/sframe: Enable archs with encoded SFrame CFA offsets Jens Remus
2026-01-27 15:19 ` [PATCH v4 07/12] s390/ptrace: Convert function macros to inline functions Jens Remus
2026-01-27 15:19 ` [PATCH v4 08/12] s390/ptrace: Provide frame_pointer() Jens Remus
2026-01-27 15:19 ` [PATCH v4 09/12] s390/unwind_user/sframe: Enable sframe unwinding on s390 Jens Remus
2026-01-27 15:19 ` [PATCH v4 10/12] unwind_user: Introduce FP/RA recovery rule unknown Jens Remus
2026-01-27 15:19 ` [PATCH v4 11/12] unwind_user/fp: Use arch-specific helper to initialize FP frame Jens Remus
2026-01-27 15:19 ` Jens Remus [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20260127151926.2805123-13-jremus@linux.ibm.com \
    --to=jremus@linux.ibm.com \
    --cc=acme@kernel.org \
    --cc=akpm@linux-foundation.org \
    --cc=andrii@kernel.org \
    --cc=beaub@linux.microsoft.com \
    --cc=bpf@vger.kernel.org \
    --cc=codonell@redhat.com \
    --cc=dylanbhatch@google.com \
    --cc=fweimer@redhat.com \
    --cc=gor@linux.ibm.com \
    --cc=hca@linux.ibm.com \
    --cc=iii@linux.ibm.com \
    --cc=indu.bhagat@oracle.com \
    --cc=jemarch@gnu.org \
    --cc=jolsa@kernel.org \
    --cc=jpoimboe@kernel.org \
    --cc=kees@kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-s390@vger.kernel.org \
    --cc=linux-trace-kernel@vger.kernel.org \
    --cc=mathieu.desnoyers@efficios.com \
    --cc=mhiramat@kernel.org \
    --cc=mingo@kernel.org \
    --cc=namhyung@kernel.org \
    --cc=peterz@infradead.org \
    --cc=rostedt@kernel.org \
    --cc=sam@gentoo.org \
    --cc=tglx@linutronix.de \
    --cc=torvalds@linux-foundation.org \
    --cc=x86@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox