Message-ID: <20260127150554.2760964-16-jremus@linux.ibm.com>
Date: Tue, 27 Jan 2026 16:05:50 +0100
From: Jens Remus <jremus@...ux.ibm.com>
To: linux-kernel@...r.kernel.org, linux-trace-kernel@...r.kernel.org,
        bpf@...r.kernel.org, x86@...nel.org, linux-mm@...ck.org,
        Steven Rostedt <rostedt@...nel.org>
Cc: Jens Remus <jremus@...ux.ibm.com>, Josh Poimboeuf <jpoimboe@...nel.org>,
        Masami Hiramatsu <mhiramat@...nel.org>,
        Mathieu Desnoyers <mathieu.desnoyers@...icios.com>,
        Peter Zijlstra <peterz@...radead.org>, Ingo Molnar <mingo@...nel.org>,
        Jiri Olsa <jolsa@...nel.org>,
        Arnaldo Carvalho de Melo <acme@...nel.org>,
        Namhyung Kim <namhyung@...nel.org>,
        Thomas Gleixner <tglx@...utronix.de>,
        Andrii Nakryiko <andrii@...nel.org>,
        Indu Bhagat <indu.bhagat@...cle.com>,
        "Jose E. Marchesi" <jemarch@....org>,
        Beau Belgrave <beaub@...ux.microsoft.com>,
        Linus Torvalds <torvalds@...ux-foundation.org>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Florian Weimer <fweimer@...hat.com>, Kees Cook <kees@...nel.org>,
        "Carlos O'Donell" <codonell@...hat.com>, Sam James <sam@...too.org>,
        Dylan Hatch <dylanbhatch@...gle.com>, Borislav Petkov <bp@...en8.de>,
        Dave Hansen <dave.hansen@...ux.intel.com>,
        David Hildenbrand <david@...hat.com>, "H. Peter Anvin" <hpa@...or.com>,
        "Liam R. Howlett" <Liam.Howlett@...cle.com>,
        Lorenzo Stoakes <lorenzo.stoakes@...cle.com>,
        Michal Hocko <mhocko@...e.com>, Mike Rapoport <rppt@...nel.org>,
        Suren Baghdasaryan <surenb@...gle.com>,
        Vlastimil Babka <vbabka@...e.cz>, Heiko Carstens <hca@...ux.ibm.com>,
        Vasily Gorbik <gor@...ux.ibm.com>
Subject: [PATCH v13 15/18] unwind_user/sframe: Add support for SFrame V3 flexible FDEs

SFrame V3 introduces flexible FDEs in addition to the regular FDEs.
The key difference is that flexible FDEs encode the CFA, RA, and FP
tracking information using two FRE data words, a control word and an
offset, or a single padding data word of zero (e.g. to represent FP
without RA tracking information).

The control word contains the following information:
- reg_p: Whether to use the register contents (reg_p=1) specified
  by regnum or the CFA (reg_p=0) as base.
- deref_p: Whether to dereference.
- regnum: A DWARF register number.

The offset is added to the base (i.e. CFA or register contents).  Then
the resulting address may optionally be dereferenced.

This enables the following flexible CFA and FP/RA recovery rules:
- CFA = register + offset	// reg_p=1, deref_p=0
- CFA = *(register + offset)	// reg_p=1, deref_p=1
- FP/RA = *(CFA + offset)	// reg_p=0, deref_p=0
- FP/RA = register + offset	// reg_p=1, deref_p=0
- FP/RA = *(register + offset)	// reg_p=1, deref_p=1

Note that for the CFA a rule with reg_p=0 is invalid, as the value of
the CFA cannot be described using itself as base.  For FP/RA a rule with
reg_p=0 and deref_p=0 and regnum=0 is invalid, as it that is equal to
the padding data word of zero.

Cc: Steven Rostedt <rostedt@...nel.org>
Cc: Josh Poimboeuf <jpoimboe@...nel.org>
Cc: Masami Hiramatsu <mhiramat@...nel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@...icios.com>
Cc: Peter Zijlstra <peterz@...radead.org>
Cc: Ingo Molnar <mingo@...nel.org>
Cc: Jiri Olsa <jolsa@...nel.org>
Cc: Arnaldo Carvalho de Melo <acme@...nel.org>
Cc: Namhyung Kim <namhyung@...nel.org>
Cc: Thomas Gleixner <tglx@...utronix.de>
Cc: Andrii Nakryiko <andrii@...nel.org>
Cc: Indu Bhagat <indu.bhagat@...cle.com>
Cc: "Jose E. Marchesi" <jemarch@....org>
Cc: Beau Belgrave <beaub@...ux.microsoft.com>
Cc: Jens Remus <jremus@...ux.ibm.com>
Cc: Linus Torvalds <torvalds@...ux-foundation.org>
Cc: Andrew Morton <akpm@...ux-foundation.org>
Cc: Florian Weimer <fweimer@...hat.com>
Cc: Sam James <sam@...too.org>
Cc: Kees Cook <kees@...nel.org>
Cc: "Carlos O'Donell" <codonell@...hat.com>
Signed-off-by: Jens Remus <jremus@...ux.ibm.com>
---

Notes (jremus):
    Changes in v13:
    - New patch.

 kernel/unwind/sframe.c | 249 ++++++++++++++++++++++++++++++++---------
 kernel/unwind/sframe.h |   5 +
 2 files changed, 204 insertions(+), 50 deletions(-)

diff --git a/kernel/unwind/sframe.c b/kernel/unwind/sframe.c
index 4dfc8cf2075e..ebf2a2905c5c 100644
--- a/kernel/unwind/sframe.c
+++ b/kernel/unwind/sframe.c
@@ -12,6 +12,7 @@
 #include <linux/mm.h>
 #include <linux/string_helpers.h>
 #include <linux/sframe.h>
+#include <asm/unwind_user_sframe.h>
 #include <linux/unwind_user_types.h>
 
 #include "sframe.h"
@@ -31,8 +32,11 @@ struct sframe_fde_internal {
 struct sframe_fre_internal {
 	unsigned int	size;
 	u32		ip_off;
+	u32		cfa_ctl;
 	s32		cfa_off;
+	u32		ra_ctl;
 	s32		ra_off;
+	u32		fp_ctl;
 	s32		fp_off;
 	u8		info;
 };
@@ -189,16 +193,147 @@ static __always_inline int __find_fde(struct sframe_section *sec,
 		 s32 :	UNSAFE_GET_USER_SIGNED_INC(to, from, size, label),	\
 		 s64 :	UNSAFE_GET_USER_SIGNED_INC(to, from, size, label))
 
+static __always_inline int
+__read_regular_fre_datawords(struct sframe_section *sec,
+			     struct sframe_fde_internal *fde,
+			     unsigned long cur,
+			     unsigned char dataword_count,
+			     unsigned char dataword_size,
+			     struct sframe_fre_internal *fre)
+{
+	s32 cfa_off, ra_off, fp_off;
+	unsigned int cfa_regnum;
+
+	UNSAFE_GET_USER_INC(cfa_off, cur, dataword_size, Efault);
+	dataword_count--;
+
+	ra_off = sec->ra_off;
+	if (!ra_off && dataword_count) {
+		dataword_count--;
+		UNSAFE_GET_USER_INC(ra_off, cur, dataword_size, Efault);
+	}
+
+	fp_off = sec->fp_off;
+	if (!fp_off && dataword_count) {
+		dataword_count--;
+		UNSAFE_GET_USER_INC(fp_off, cur, dataword_size, Efault);
+	}
+
+	if (dataword_count)
+		return -EFAULT;
+
+	cfa_regnum =
+		(SFRAME_V3_FRE_CFA_BASE_REG_ID(fre->info) == SFRAME_BASE_REG_FP) ?
+			SFRAME_REG_FP : SFRAME_REG_SP;
+
+	fre->cfa_ctl	= (cfa_regnum << 3) | 1; /* regnum, deref_p=0, reg_p=1 */
+	fre->cfa_off	= cfa_off;
+	fre->ra_ctl	= ra_off ? 2 : 0; /* regnum=0, deref_p=(ra_off != 0), reg_p=0 */
+	fre->ra_off	= ra_off;
+	fre->fp_ctl	= fp_off ? 2 : 0; /* regnum=0, deref_p=(fp_off != 0), reg_p=0 */
+	fre->fp_off	= fp_off;
+
+	return 0;
+
+Efault:
+	return -EFAULT;
+}
+
+static __always_inline int
+__read_flex_fde_fre_datawords(struct sframe_section *sec,
+			      struct sframe_fde_internal *fde,
+			      unsigned long cur,
+			      unsigned char dataword_count,
+			      unsigned char dataword_size,
+			      struct sframe_fre_internal *fre)
+{
+	u32 cfa_ctl, ra_ctl, fp_ctl;
+	s32 cfa_off, ra_off, fp_off;
+
+	if (dataword_count < 2)
+		return -EFAULT;
+	UNSAFE_GET_USER_INC(cfa_ctl, cur, dataword_size, Efault);
+	UNSAFE_GET_USER_INC(cfa_off, cur, dataword_size, Efault);
+	dataword_count -= 2;
+
+	ra_off = sec->ra_off;
+	ra_ctl = ra_off ? 2 : 0; /* regnum=0, deref_p=(ra_off != 0), reg_p=0 */
+	if (dataword_count >= 2) {
+		UNSAFE_GET_USER_INC(ra_ctl, cur, dataword_size, Efault);
+		dataword_count--;
+		if (ra_ctl) {
+			UNSAFE_GET_USER_INC(ra_off, cur, dataword_size, Efault);
+			dataword_count--;
+		} else {
+			/* Padding RA location info */
+			ra_ctl = ra_off ? 2 : 0; /* re-deduce (see above) */
+		}
+	}
+
+	fp_off = sec->fp_off;
+	fp_ctl = fp_off ? 2 : 0; /* regnum=0, deref_p=(fp_off != 0), reg_p=0 */
+	if (dataword_count >= 2) {
+		UNSAFE_GET_USER_INC(fp_ctl, cur, dataword_size, Efault);
+		dataword_count--;
+		if (fp_ctl) {
+			UNSAFE_GET_USER_INC(fp_off, cur, dataword_size, Efault);
+			dataword_count--;
+		} else {
+			/* Padding FP location info */
+			fp_ctl = fp_off ? 2 : 0; /* re-deduce (see above) */
+		}
+	}
+
+	if (dataword_count)
+		return -EFAULT;
+
+	fre->cfa_ctl	= cfa_ctl;
+	fre->cfa_off	= cfa_off;
+	fre->ra_ctl	= ra_ctl;
+	fre->ra_off	= ra_off;
+	fre->fp_ctl	= fp_ctl;
+	fre->fp_off	= fp_off;
+
+	return 0;
+
+Efault:
+	return -EFAULT;
+}
+
+static __always_inline int
+__read_fre_datawords(struct sframe_section *sec,
+		     struct sframe_fde_internal *fde,
+		     unsigned long cur,
+		     unsigned char dataword_count,
+		     unsigned char dataword_size,
+		     struct sframe_fre_internal *fre)
+{
+	unsigned char fde_type = SFRAME_V3_FDE_TYPE(fde->info2);
+
+	switch (fde_type) {
+	case SFRAME_FDE_TYPE_REGULAR:
+		return __read_regular_fre_datawords(sec, fde, cur,
+						    dataword_count,
+						    dataword_size,
+						    fre);
+	case SFRAME_FDE_TYPE_FLEXIBLE:
+		return __read_flex_fde_fre_datawords(sec, fde, cur,
+						     dataword_count,
+						     dataword_size,
+						     fre);
+	default:
+		return -EFAULT;
+	}
+}
+
 static __always_inline int __read_fre(struct sframe_section *sec,
 				      struct sframe_fde_internal *fde,
 				      unsigned long fre_addr,
 				      struct sframe_fre_internal *fre)
 {
-	unsigned char fde_type = SFRAME_V3_FDE_TYPE(fde->info2);
 	unsigned char fde_pctype = SFRAME_V3_FDE_PCTYPE(fde->info);
 	unsigned char fre_type = SFRAME_V3_FDE_FRE_TYPE(fde->info);
 	unsigned char dataword_count, dataword_size;
-	s32 cfa_off, ra_off, fp_off;
 	unsigned long cur = fre_addr;
 	unsigned char addr_size;
 	u32 ip_off;
@@ -224,75 +359,88 @@ static __always_inline int __read_fre(struct sframe_section *sec,
 	if (cur + (dataword_count * dataword_size) > sec->fres_end)
 		return -EFAULT;
 
-	/* TODO: Support for flexible FDEs not implemented yet. */
-	if (fde_type != SFRAME_FDE_TYPE_REGULAR)
-		return -EFAULT;
+	fre->size	= addr_size + 1 + (dataword_count * dataword_size);
+	fre->ip_off	= ip_off;
+	fre->info	= info;
 
 	if (!dataword_count) {
 		/*
 		 * A FRE without data words indicates RA undefined /
 		 * outermost frame.
 		 */
-		cfa_off	= 0;
-		ra_off	= 0;
-		fp_off	= 0;
-		goto done;
-	}
+		fre->cfa_ctl	= 0;
+		fre->cfa_off	= 0;
+		fre->ra_ctl	= 0;
+		fre->ra_off	= 0;
+		fre->fp_ctl	= 0;
+		fre->fp_off	= 0;
 
-	UNSAFE_GET_USER_INC(cfa_off, cur, dataword_size, Efault);
-	dataword_count--;
-
-	ra_off = sec->ra_off;
-	if (!ra_off && dataword_count) {
-		dataword_count--;
-		UNSAFE_GET_USER_INC(ra_off, cur, dataword_size, Efault);
-	}
-
-	fp_off = sec->fp_off;
-	if (!fp_off && dataword_count) {
-		dataword_count--;
-		UNSAFE_GET_USER_INC(fp_off, cur, dataword_size, Efault);
+		return 0;
 	}
 
-	if (dataword_count)
-		return -EFAULT;
-
-done:
-	fre->size	= addr_size + 1 + (dataword_count * dataword_size);
-	fre->ip_off	= ip_off;
-	fre->cfa_off	= cfa_off;
-	fre->ra_off	= ra_off;
-	fre->fp_off	= fp_off;
-	fre->info	= info;
-
-	return 0;
+	return __read_fre_datawords(sec, fde, cur, dataword_count, dataword_size, fre);
 
 Efault:
 	return -EFAULT;
 }
 
-static __always_inline void
+static __always_inline int
 sframe_init_cfa_rule_data(struct unwind_user_cfa_rule_data *cfa_rule_data,
-			  unsigned char fre_info,
-			  s32 offset)
+			  u32 ctlword, s32 offset)
 {
-	if (SFRAME_V3_FRE_CFA_BASE_REG_ID(fre_info) == SFRAME_BASE_REG_FP)
-		cfa_rule_data->rule = UNWIND_USER_CFA_RULE_FP_OFFSET;
-	else
-		cfa_rule_data->rule = UNWIND_USER_CFA_RULE_SP_OFFSET;
+	bool deref_p = SFRAME_V3_FLEX_FDE_CTLWORD_DEREF_P(ctlword);
+	bool reg_p = SFRAME_V3_FLEX_FDE_CTLWORD_REG_P(ctlword);
+
+	if (reg_p) {
+		unsigned int regnum = SFRAME_V3_FLEX_FDE_CTLWORD_REGNUM(ctlword);
+
+		switch (regnum) {
+		case SFRAME_REG_SP:
+			cfa_rule_data->rule = UNWIND_USER_CFA_RULE_SP_OFFSET;
+			break;
+		case SFRAME_REG_FP:
+			cfa_rule_data->rule = UNWIND_USER_CFA_RULE_FP_OFFSET;
+			break;
+		default:
+			cfa_rule_data->rule = UNWIND_USER_CFA_RULE_REG_OFFSET;
+			cfa_rule_data->regnum = regnum;
+		}
+	} else {
+		return -EINVAL;
+	}
+
+	if (deref_p)
+		cfa_rule_data->rule |= UNWIND_USER_RULE_DEREF;
+
 	cfa_rule_data->offset = offset;
+
+	return 0;
 }
 
 static __always_inline void
 sframe_init_rule_data(struct unwind_user_rule_data *rule_data,
-		      s32 offset)
+		      u32 ctlword, s32 offset)
 {
-	if (offset) {
-		rule_data->rule = UNWIND_USER_RULE_CFA_OFFSET_DEREF;
-		rule_data->offset = offset;
-	} else {
+	bool deref_p = SFRAME_V3_FLEX_FDE_CTLWORD_DEREF_P(ctlword);
+	bool reg_p = SFRAME_V3_FLEX_FDE_CTLWORD_REG_P(ctlword);
+
+	if (!ctlword && !offset) {
 		rule_data->rule = UNWIND_USER_RULE_RETAIN;
+		return;
+	}
+	if (reg_p) {
+		unsigned int regnum = SFRAME_V3_FLEX_FDE_CTLWORD_REGNUM(ctlword);
+
+		rule_data->rule = UNWIND_USER_RULE_REG_OFFSET;
+		rule_data->regnum = regnum;
+	} else {
+		rule_data->rule = UNWIND_USER_RULE_CFA_OFFSET;
 	}
+
+	if (deref_p)
+		rule_data->rule |= UNWIND_USER_RULE_DEREF;
+
+	rule_data->offset = offset;
 }
 
 static __always_inline int __find_fre(struct sframe_section *sec,
@@ -344,9 +492,10 @@ static __always_inline int __find_fre(struct sframe_section *sec,
 		return -EINVAL;
 	fre = prev_fre;
 
-	sframe_init_cfa_rule_data(&frame->cfa, fre->info, fre->cfa_off);
-	sframe_init_rule_data(&frame->ra, fre->ra_off);
-	sframe_init_rule_data(&frame->fp, fre->fp_off);
+	if (sframe_init_cfa_rule_data(&frame->cfa, fre->cfa_ctl, fre->cfa_off))
+		return -EINVAL;
+	sframe_init_rule_data(&frame->ra, fre->ra_ctl, fre->ra_off);
+	sframe_init_rule_data(&frame->fp, fre->fp_ctl, fre->fp_off);
 	frame->outermost = SFRAME_V3_FRE_RA_UNDEFINED_P(fre->info);
 
 	return 0;
diff --git a/kernel/unwind/sframe.h b/kernel/unwind/sframe.h
index 3fcc15534e5a..5b6112945b6c 100644
--- a/kernel/unwind/sframe.h
+++ b/kernel/unwind/sframe.h
@@ -66,6 +66,7 @@ struct sframe_fda_v3 {
 #define SFRAME_V3_AARCH64_FDE_PAUTH_KEY(info)	(((info) >> 5) & 0x1)
 
 #define SFRAME_FDE_TYPE_REGULAR			0
+#define SFRAME_FDE_TYPE_FLEXIBLE		1
 
 #define SFRAME_V3_FDE_TYPE_MASK			0x0f
 #define SFRAME_V3_FDE_TYPE(info2)		((info2) & SFRAME_V3_FDE_TYPE_MASK)
@@ -79,4 +80,8 @@ struct sframe_fda_v3 {
 #define SFRAME_V3_AARCH64_FRE_MANGLED_RA_P(info)	(((info) >> 7) & 0x1)
 #define SFRAME_V3_FRE_RA_UNDEFINED_P(info)		(SFRAME_V3_FRE_DATAWORD_COUNT(info) == 0)
 
+#define SFRAME_V3_FLEX_FDE_CTLWORD_REGNUM(data)		(((data) >> 3) & 0x1f)
+#define SFRAME_V3_FLEX_FDE_CTLWORD_DEREF_P(data)	(((data) >> 1) & 0x1)
+#define SFRAME_V3_FLEX_FDE_CTLWORD_REG_P(data)		((data) & 0x1)
+
 #endif /* _SFRAME_H */
-- 
2.51.0

Powered by blists - more mailing lists