[PATCH v4 05/11] s390: Represent FP without RA saved in SFrame
Jens Remus
jremus@linux.ibm.com
Mon Jul 7 16:21:46 GMT 2025
If an architecture uses both SFrame RA and FP tracking SFrame assumes
that the RA offset is the 2nd offset and the FP offset is the 3rd offset
following a SFrame FRE. An architecture does not necessarily need to
save both on the stack (or in register) at the same time or even at all.
SFrame cannot represent FP without RA saved on stack (or in a register),
since it cannot distinguish whether the 2nd offset is the RA or FP
offset.
For s390x use an invalid SFrame RA offset from CFA value of zero as
padding to represent the FP being saved when the RA is not saved. This
aligns with the existing invalid SFrame fixed RA offset from CFA value
of zero. In a stack tracer this then also naturally falls into place,
as it can skip restoring the RA in the topmost frame, if both the fixed
RA offset (from SFrame header) and the RA offset (from FDE) are zero,
without any need to test architecture-specific flags.
include/
* sframe.h (SFRAME_FRE_RA_OFFSET_INVALID): New define. Used as
padding offset.
* sframe-api.h (sframe_fre_get_ra_offset): Add comment that for
s390x an offset value of SFRAME_FRE_RA_OFFSET_INVALID indicates
that the RA is not saved.
gas/
* gen-sframe.c (get_fre_num_offsets): For s390x account padding
RA offset, if FP without RA saved.
(sframe_get_fre_offset_size): Likewise.
(output_sframe_row_entry): For s390x write a padding RA offset,
if FP without RA needs to be represented.
(sframe_do_fde): Enable FP without RA saved to be represented
on s390x.
libsframe/
* sframe.c (sframe_fre_get_ra_offset): Add comment that for
s390x an offset value of SFRAME_FRE_RA_OFFSET_INVALID indicates
that the RA is not saved.
* sframe-dump.c (dump_sframe_func_with_fres): Treat invalid
RA offsets as if they were undefined. Display them as "U"
to distinguish them.
* doc/sframe-spec.texi (s390x): Document s390x-specific use of
SFRAME_FRE_RA_OFFSET_INVALID to represent FP without RA saved.
gas/testsuite/
* gas/cfi-sframe/cfi-sframe.exp: Rename s390x-specific tests.
* gas/cfi-sframe/cfi-sframe-s390x-fpra-offset-err-1.s: Rename
to ...
* cfi-sframe/cfi-sframe-s390x-fpra-offset-err-1.d: Likewise.
* gas/cfi-sframe/cfi-sframe-s390x-fpra-offset-2.s: This.
* gas/cfi-sframe/cfi-sframe-s390x-fpra-offset-2.d: Likewise.
Update test verification pattern accordingly.
* cfi-sframe/cfi-sframe-s390x-fpra-register-err-1.s: Rename
to ...
* cfi-sframe/cfi-sframe-s390x-fpra-register-err-1.d: Likewise.
* gas/cfi-sframe/cfi-sframe-s390x-fpra-register-2.s: This.
* gas/cfi-sframe/cfi-sframe-s390x-fpra-register-2.d: Likewise.
Update test verification pattern accordingly.
Signed-off-by: Jens Remus <jremus@linux.ibm.com>
---
Notes (jremus):
Changes in V4:
- Adapt for recent fix of relocatable SFrame links and expect
SFRAME_F_FDE_FUNC_START_PCREL in tests.
- Indent list of s390x-specific changes in SFrame spec.
Changes in V3:
- Update "Changes from Version 1 to Version 2" in SFrame spec. (Indu)
- Add comment on sframe_fre_get_ra_offset that for s390x the offset may
be SFRAME_FRE_RA_OFFSET_INVALID when FP without RA saved.
- Correct "a RA offset" to "an RA offset". (Indu)
- Reword comment on "SFRAME_FRE_RA_OFFSET_INVALID". (Indu)
Changes in V2:
- Reword commit message (s390 to s390x). (Indu)
- Rename s390x-specific macros from *_S390_* to *_S390X_*. (Indu)
- Use "s390x" instead of "S390" in comments and SFrame spec. (Indu)
- Rename CFI SFrame test file names from s390 to s390x. (Indu)
- Reword test names.
- Display padding RA offsets as "U" instead of "u*". (Indu)
An alternative approach would have been to redefine the 4-bit offset
count field in the FDE info word into a bit field as proposed with
"[RFC PATCH 0/1] sframe: Represent FP without RA on stack (bitmap)" [1].
With the current 4-bit count up to 15+1 offsets can be represented (+1
if the CFA offset would be considered mandatory, so that it does not
need to be explicitly accounted). With a 4 bit bit field only up to
4+1 offsets could be represented.
[1]: Path series "[RFC PATCH 0/1] sframe: Represent FP without RA on
stack (bitmap)",
https://inbox.sourceware.org/binutils/20240422155905.2497883-1-jremus@linux.ibm.com/
A test build of Glibc tag 2.41 on s390x libc.so shows 342 padding
RA offsets. With a total of 29792 offsets the padding RA offsets
account for only ~1.15% of the offsets, which seems acceptable.
Statistics for libc.so:
VALUE TOTAL MIN MAX AVG
FDEs: 3652 - - -
FREs/FDE: 15236 1 20 4
Offsets/FDE: 29792 1 38 8
8-bit: 0 0 0 0
16-bit: 29792 1 38 8
32-bit: 0 0 0 0
Offsets/FRE: 29792 1 3 1
8-bit: - 0 0 0
16-bit: - 1 3 1
32-bit: - 0 0 0
O_Padd/FDE: 342 - - 0
8-bit: 0
16-bit: 342
32-bit: 0
Note that on s390x the offsets are at minimum 16-bits in size, due to
the mandatory CFA offset being at least 160, which does not fit into
an 8-bit signed integer. This is addressed by a following patch.
gas/gen-sframe.c | 38 ++++++++++++++-----
.../cfi-sframe-s390x-fpra-offset-2.d | 22 +++++++++++
...r-1.s => cfi-sframe-s390x-fpra-offset-2.s} | 0
.../cfi-sframe-s390x-fpra-offset-err-1.d | 15 --------
.../cfi-sframe-s390x-fpra-register-2.d | 22 +++++++++++
...1.s => cfi-sframe-s390x-fpra-register-2.s} | 0
.../cfi-sframe-s390x-fpra-register-err-1.d | 15 --------
gas/testsuite/gas/cfi-sframe/cfi-sframe.exp | 4 +-
include/sframe-api.h | 2 +
include/sframe.h | 22 +++++++++++
libsframe/doc/sframe-spec.texi | 17 ++++++---
libsframe/sframe-dump.c | 4 ++
libsframe/sframe.c | 2 +
13 files changed, 116 insertions(+), 47 deletions(-)
create mode 100644 gas/testsuite/gas/cfi-sframe/cfi-sframe-s390x-fpra-offset-2.d
rename gas/testsuite/gas/cfi-sframe/{cfi-sframe-s390x-fpra-offset-err-1.s => cfi-sframe-s390x-fpra-offset-2.s} (100%)
delete mode 100644 gas/testsuite/gas/cfi-sframe/cfi-sframe-s390x-fpra-offset-err-1.d
create mode 100644 gas/testsuite/gas/cfi-sframe/cfi-sframe-s390x-fpra-register-2.d
rename gas/testsuite/gas/cfi-sframe/{cfi-sframe-s390x-fpra-register-err-1.s => cfi-sframe-s390x-fpra-register-2.s} (100%)
delete mode 100644 gas/testsuite/gas/cfi-sframe/cfi-sframe-s390x-fpra-register-err-1.d
diff --git a/gas/gen-sframe.c b/gas/gen-sframe.c
index c193c6e750a3..ab1dfcb49dd1 100644
--- a/gas/gen-sframe.c
+++ b/gas/gen-sframe.c
@@ -340,7 +340,10 @@ get_fre_num_offsets (struct sframe_row_entry *sframe_fre)
if (sframe_fre->bp_loc == SFRAME_FRE_ELEM_LOC_STACK)
fre_num_offsets++;
if (sframe_ra_tracking_p ()
- && sframe_fre->ra_loc == SFRAME_FRE_ELEM_LOC_STACK)
+ && (sframe_fre->ra_loc == SFRAME_FRE_ELEM_LOC_STACK
+ /* For s390x account padding RA offset, if FP without RA saved. */
+ || (sframe_get_abi_arch () == SFRAME_ABI_S390X_ENDIAN_BIG
+ && sframe_fre->bp_loc == SFRAME_FRE_ELEM_LOC_STACK)))
fre_num_offsets++;
return fre_num_offsets;
}
@@ -362,9 +365,15 @@ sframe_get_fre_offset_size (struct sframe_row_entry *sframe_fre)
cfa_offset_size = get_offset_size_in_bytes (sframe_fre->cfa_offset);
if (sframe_fre->bp_loc == SFRAME_FRE_ELEM_LOC_STACK)
bp_offset_size = get_offset_size_in_bytes (sframe_fre->bp_offset);
- if (sframe_ra_tracking_p ()
- && sframe_fre->ra_loc == SFRAME_FRE_ELEM_LOC_STACK)
- ra_offset_size = get_offset_size_in_bytes (sframe_fre->ra_offset);
+ if (sframe_ra_tracking_p ())
+ {
+ if (sframe_fre->ra_loc == SFRAME_FRE_ELEM_LOC_STACK)
+ ra_offset_size = get_offset_size_in_bytes (sframe_fre->ra_offset);
+ /* For s390x account padding RA offset, if FP without RA saved. */
+ else if (sframe_get_abi_arch () == SFRAME_ABI_S390X_ENDIAN_BIG
+ && sframe_fre->bp_loc == SFRAME_FRE_ELEM_LOC_STACK)
+ ra_offset_size = get_offset_size_in_bytes (SFRAME_FRE_RA_OFFSET_INVALID);
+ }
/* Get the maximum size needed to represent the offsets. */
max_offset_size = cfa_offset_size;
@@ -570,11 +579,20 @@ output_sframe_row_entry (symbolS *fde_start_addr,
fre_offset_func_map[idx].out_func (sframe_fre->cfa_offset);
fre_write_offsets++;
- if (sframe_ra_tracking_p ()
- && sframe_fre->ra_loc == SFRAME_FRE_ELEM_LOC_STACK)
+ if (sframe_ra_tracking_p ())
{
- fre_offset_func_map[idx].out_func (sframe_fre->ra_offset);
- fre_write_offsets++;
+ if (sframe_fre->ra_loc == SFRAME_FRE_ELEM_LOC_STACK)
+ {
+ fre_offset_func_map[idx].out_func (sframe_fre->ra_offset);
+ fre_write_offsets++;
+ }
+ /* For s390x write padding RA offset, if FP without RA saved. */
+ else if (sframe_get_abi_arch () == SFRAME_ABI_S390X_ENDIAN_BIG
+ && sframe_fre->bp_loc == SFRAME_FRE_ELEM_LOC_STACK)
+ {
+ fre_offset_func_map[idx].out_func (SFRAME_FRE_RA_OFFSET_INVALID);
+ fre_write_offsets++;
+ }
}
if (sframe_fre->bp_loc == SFRAME_FRE_ELEM_LOC_STACK)
{
@@ -1804,7 +1822,9 @@ sframe_do_fde (struct sframe_xlate_ctx *xlate_ctx,
= get_dw_fde_end_addrS (xlate_ctx->dw_fde);
}
- if (sframe_ra_tracking_p ())
+ /* ABI/arch except s390x cannot represent FP without RA saved. */
+ if (sframe_ra_tracking_p ()
+ && sframe_get_abi_arch () != SFRAME_ABI_S390X_ENDIAN_BIG)
{
struct sframe_row_entry *fre;
diff --git a/gas/testsuite/gas/cfi-sframe/cfi-sframe-s390x-fpra-offset-2.d b/gas/testsuite/gas/cfi-sframe/cfi-sframe-s390x-fpra-offset-2.d
new file mode 100644
index 000000000000..305a917e3255
--- /dev/null
+++ b/gas/testsuite/gas/cfi-sframe/cfi-sframe-s390x-fpra-offset-2.d
@@ -0,0 +1,22 @@
+#name: SFrame generation on s390x - FP and then RA saved on stack
+#objdump: --sframe=.sframe
+#...
+Contents of the SFrame section .sframe:
+
+ Header :
+
+ Version: SFRAME_VERSION_2
+ Flags: SFRAME_F_FDE_FUNC_START_PCREL
+ Num FDEs: 1
+ Num FREs: 5
+
+ Function Index :
+
+ func idx \[0\]: pc = 0x0, size = 34 bytes
+ STARTPC +CFA +FP +RA +
+ 0+0000 +sp\+160 +u +u +
+ 0+0006 +sp\+160 +c\-72 +U +
+ 0+000c +sp\+160 +c\-72 +c\-48 +
+ 0+001a +sp\+160 +c-72 +U +
+ 0+0020 +sp\+160 +u +u +
+#pass
diff --git a/gas/testsuite/gas/cfi-sframe/cfi-sframe-s390x-fpra-offset-err-1.s b/gas/testsuite/gas/cfi-sframe/cfi-sframe-s390x-fpra-offset-2.s
similarity index 100%
rename from gas/testsuite/gas/cfi-sframe/cfi-sframe-s390x-fpra-offset-err-1.s
rename to gas/testsuite/gas/cfi-sframe/cfi-sframe-s390x-fpra-offset-2.s
diff --git a/gas/testsuite/gas/cfi-sframe/cfi-sframe-s390x-fpra-offset-err-1.d b/gas/testsuite/gas/cfi-sframe/cfi-sframe-s390x-fpra-offset-err-1.d
deleted file mode 100644
index 7d71874111ba..000000000000
--- a/gas/testsuite/gas/cfi-sframe/cfi-sframe-s390x-fpra-offset-err-1.d
+++ /dev/null
@@ -1,15 +0,0 @@
-#name: SFrame generation on s390x - FP without RA saved on stack
-#as: --gsframe
-#warning: FP without RA on stack
-#objdump: --sframe=.sframe
-#...
-Contents of the SFrame section .sframe:
-
- Header :
-
- Version: SFRAME_VERSION_2
- Flags: SFRAME_F_FDE_FUNC_START_PCREL
- Num FDEs: 0
- Num FREs: 0
-
-#pass
diff --git a/gas/testsuite/gas/cfi-sframe/cfi-sframe-s390x-fpra-register-2.d b/gas/testsuite/gas/cfi-sframe/cfi-sframe-s390x-fpra-register-2.d
new file mode 100644
index 000000000000..7b719fd979fb
--- /dev/null
+++ b/gas/testsuite/gas/cfi-sframe/cfi-sframe-s390x-fpra-register-2.d
@@ -0,0 +1,22 @@
+#name: SFrame generation on s390x - FP and then RA saved in FPR registers
+#objdump: --sframe=.sframe
+#...
+Contents of the SFrame section .sframe:
+
+ Header :
+
+ Version: SFRAME_VERSION_2
+ Flags: SFRAME_F_FDE_FUNC_START_PCREL
+ Num FDEs: 1
+ Num FREs: 5
+
+ Function Index :
+
+ func idx \[0\]: pc = 0x0, size = 26 bytes
+ STARTPC +CFA +FP +RA +
+ 0+0000 +sp\+160 +u +u +
+ 0+0004 +sp\+160 +r17 +U +
+ 0+0008 +sp\+160 +r17 +r16 +
+ 0+0014 +sp\+160 +r17 +U +
+ 0+0018 +sp\+160 +u +u +
+#pass
diff --git a/gas/testsuite/gas/cfi-sframe/cfi-sframe-s390x-fpra-register-err-1.s b/gas/testsuite/gas/cfi-sframe/cfi-sframe-s390x-fpra-register-2.s
similarity index 100%
rename from gas/testsuite/gas/cfi-sframe/cfi-sframe-s390x-fpra-register-err-1.s
rename to gas/testsuite/gas/cfi-sframe/cfi-sframe-s390x-fpra-register-2.s
diff --git a/gas/testsuite/gas/cfi-sframe/cfi-sframe-s390x-fpra-register-err-1.d b/gas/testsuite/gas/cfi-sframe/cfi-sframe-s390x-fpra-register-err-1.d
deleted file mode 100644
index f6854c26814f..000000000000
--- a/gas/testsuite/gas/cfi-sframe/cfi-sframe-s390x-fpra-register-err-1.d
+++ /dev/null
@@ -1,15 +0,0 @@
-#name: SFrame generation on s390x - FP without RA saved in registers
-#as: --gsframe
-#warning: FP without RA on stack
-#objdump: --sframe=.sframe
-#...
-Contents of the SFrame section .sframe:
-
- Header :
-
- Version: SFRAME_VERSION_2
- Flags: SFRAME_F_FDE_FUNC_START_PCREL
- Num FDEs: 0
- Num FREs: 0
-
-#pass
diff --git a/gas/testsuite/gas/cfi-sframe/cfi-sframe.exp b/gas/testsuite/gas/cfi-sframe/cfi-sframe.exp
index 1e7fc9e03cad..b26ce60c6e59 100644
--- a/gas/testsuite/gas/cfi-sframe/cfi-sframe.exp
+++ b/gas/testsuite/gas/cfi-sframe/cfi-sframe.exp
@@ -119,7 +119,7 @@ if { [istarget "s390x*-*-*"] && [gas_sframe_check] } then {
run_dump_test "cfi-sframe-s390x-err-2"
run_dump_test "cfi-sframe-s390x-err-3"
run_dump_test "cfi-sframe-s390x-fpra-offset-1"
- run_dump_test "cfi-sframe-s390x-fpra-offset-err-1"
+ run_dump_test "cfi-sframe-s390x-fpra-offset-2"
run_dump_test "cfi-sframe-s390x-fpra-register-1"
- run_dump_test "cfi-sframe-s390x-fpra-register-err-1"
+ run_dump_test "cfi-sframe-s390x-fpra-register-2"
}
diff --git a/include/sframe-api.h b/include/sframe-api.h
index e27f78150e2b..239007a3b282 100644
--- a/include/sframe-api.h
+++ b/include/sframe-api.h
@@ -229,6 +229,8 @@ sframe_fre_get_fp_offset (sframe_decoder_ctx *dctx,
/* Get the RA offset from the FRE. If the offset is invalid, sets errp.
+ For s390x an RA offset value of SFRAME_FRE_RA_OFFSET_INVALID indicates
+ that the RA is not saved, which is only valid in the topmost frame.
For s390x the offset may be an encoded register number, indicated by
LSB set to one, which is only valid in the topmost frame. */
extern int32_t
diff --git a/include/sframe.h b/include/sframe.h
index 7bbdf04db171..28b625b3258f 100644
--- a/include/sframe.h
+++ b/include/sframe.h
@@ -248,6 +248,10 @@ typedef struct sframe_func_desc_entry
may or may not be tracked. */
#define SFRAME_FRE_FP_OFFSET_IDX 2
+/* Invalid RA offset. Currently used for s390x as padding to represent FP
+ without RA saved. */
+#define SFRAME_FRE_RA_OFFSET_INVALID 0
+
typedef struct sframe_fre_info
{
/* Information about
@@ -310,6 +314,24 @@ typedef struct sframe_fre_info
fi
Note that in AAPCS64, a frame record, if created, will save both FP and
LR on stack.
+
+ s390x:
+ offset1 (interpreted as CFA = BASE_REG + offset1)
+ if RA is being tracked
+ offset2 (interpreted as RA = CFA + offset2; an offset value of
+ SFRAME_FRE_RA_OFFSET_INVALID indicates a dummy padding RA offset
+ to represent FP without RA saved on stack)
+ if FP is being tracked
+ offset3 (intrepreted as FP = CFA + offset3)
+ fi
+ else
+ if FP is being tracked
+ offset2 (intrepreted as FP = CFA + offset2)
+ fi
+ fi
+ Note that in s390x, if a FP/RA offset2/offset3 value has the least-
+ significant bit set it represents a DWARF register number shifted to the
+ left by 1 to restore the FP/RA value from.
*/
/* Used when SFRAME_FRE_TYPE_ADDR1 is specified as FRE type. */
diff --git a/libsframe/doc/sframe-spec.texi b/libsframe/doc/sframe-spec.texi
index 1938207233cd..6ba6c131edb6 100644
--- a/libsframe/doc/sframe-spec.texi
+++ b/libsframe/doc/sframe-spec.texi
@@ -152,6 +152,9 @@ DWARF register number.
FP/RA offset.
@item SFRAME_V2_S390X_OFFSET_DECODE_REGNUM: Decode a DWARF register number from
an FP/RA offset.
+ @item SFRAME_FRE_RA_OFFSET_INVALID: Invalid RA offset value (like
+SFRAME_CFA_FIXED_RA_INVALID). Used on s390x as padding offset to represent
+FP without RA saved.
@end itemize
@end itemize
@@ -890,12 +893,13 @@ The (64-bit) s390x ELF ABI does not mandate the precise location in a function
where the return address (RA) and frame pointer (FP) are saved, if at all.
Hence the need to track RA in the SFrame stack trace format. As RA is being
tracked in this ABI, the second stack offset is always used to locate the RA
-stack slot, by interpreting it as: RA = CFA + offset2. RA remains unchanged,
-if the offset is not available. Stack tracers are recommended to validate that
-the "unchanged RA" pattern, when present, is seen only for the topmost stack
-frame. The third stack offset is used to locate the FP stack slot, by
-interpreting it as: FP = CFA + offset3. FP remains unchanged, if the offset is
-not available.
+stack slot, by interpreting it as: RA = CFA + offset2, unless the offset has a
+value of @code{SFRAME_FRE_RA_OFFSET_INVALID}. RA remains unchanged, if the
+offset is not available or has a value of @code{SFRAME_FRE_RA_OFFSET_INVALID}.
+Stack tracers are recommended to validate that the "unchanged RA" pattern, when
+present, is seen only for the topmost stack frame. The third stack offset is
+used to locate the FP stack slot, by interpreting it as: FP = CFA + offset3.
+FP remains unchanged, if the offset is not available.
In leaf functions the RA and FP may be saved in other registers, such as
floating-point registers (FPRs), instead of on the stack. To represent this
@@ -919,6 +923,7 @@ Hence, in summary:
@item 1 @tab CFA = @code{BASE_REG} + offset1
@item 2 @tab RA stack slot = CFA + offset2, if (offset2 & 1 == 0)
@*RA register number = offset2 >> 1, if (offset2 & 1 == 1)
+ @*RA not saved if (offset2 == @code{SFRAME_FRE_RA_OFFSET_INVALID})
@item 3 @tab FP stack slot = CFA + offset3, if (offset3 & 1 == 0)
@*FP register number = offset3 >> 1, if (offset3 & 1 == 1)
@end multitable
diff --git a/libsframe/sframe-dump.c b/libsframe/sframe-dump.c
index f17ff64e1224..d55d3847194e 100644
--- a/libsframe/sframe-dump.c
+++ b/libsframe/sframe-dump.c
@@ -211,6 +211,10 @@ dump_sframe_func_with_fres (sframe_decoder_ctx *sfd_ctx,
if (sframe_decoder_get_fixed_ra_offset (sfd_ctx)
!= SFRAME_CFA_FIXED_RA_INVALID)
strcpy (temp, "f");
+ /* If an ABI does track RA offset, e.g. s390x, it can be a padding
+ to represent FP without RA being saved on stack. */
+ else if (err[2] == 0 && ra_offset == SFRAME_FRE_RA_OFFSET_INVALID)
+ sprintf (temp, "U");
else if (err[2] == 0)
{
if (is_sframe_abi_arch_s390x (sfd_ctx)
diff --git a/libsframe/sframe.c b/libsframe/sframe.c
index af40e5b4fadf..d8d370da90dd 100644
--- a/libsframe/sframe.c
+++ b/libsframe/sframe.c
@@ -733,6 +733,8 @@ sframe_fre_get_fp_offset (sframe_decoder_ctx *dctx,
/* Get the RA offset from the FRE. If the offset is invalid, sets errp.
+ For s390x an RA offset value of SFRAME_FRE_RA_OFFSET_INVALID indicates
+ that the RA is not saved, which is only valid in the topmost frame.
For s390x the offset may be an encoded register number, indicated by
LSB set to one, which is only valid in the topmost frame. */
--
2.48.1
More information about the Binutils
mailing list