This is the mail archive of the
gdb-cvs@sourceware.org
mailing list for the GDB project.
[binutils-gdb] gdb: Add OpenRISC or1k and or1knd target support
- From: Stafford Horne <shorne at sourceware dot org>
- To: gdb-cvs at sourceware dot org
- Date: 12 Dec 2017 14:57:51 -0000
- Subject: [binutils-gdb] gdb: Add OpenRISC or1k and or1knd target support
https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;h=a994fec4f8f7a886be61cfb3023c354cd2483b9d
commit a994fec4f8f7a886be61cfb3023c354cd2483b9d
Author: Franck Jullien <franck.jullien@gmail.com>
Date: Sat Dec 9 05:57:25 2017 +0900
gdb: Add OpenRISC or1k and or1knd target support
This patch prepares the current GDB port of the OpenRISC processor from
https://github.com/openrisc/binutils-gdb for upstream merging.
Testing has been done with a cgen sim provided in a separate patch. This
has been tested with 2 toolchains. GCC [1] 5.4.0 from the OpenRISC
project with Newlib [2] and GCC 5.4.0 with Musl [3] 1.1.4.
It supports or1knd (no delay slot target).
The default target is or1k (with delay slot).
You can change the target arch with:
(gdb) set architecture or1knd
The target architecture is assumed to be or1knd
[1] https://github.com/openrisc/or1k-gcc
[2] https://github.com/openrisc/newlib
[3] https://github.com/openrisc/musl-cross
gdb/doc/ChangeLog:
2017-12-12 Stafford Horne <shorne@gmail.com>
Stefan Wallentowitz <stefan@wallentowitz.de>
Franck Jullien <franck.jullien@gmail.com>
Jeremy Bennett <jeremy.bennett@embecosm.com>
* gdb.texinfo: Add OpenRISC documentation.
gdb/ChangeLog:
2017-12-12 Stafford Horne <shorne@gmail.com>
Stefan Wallentowitz <stefan@wallentowitz.de>
Stefan Kristiansson <stefan.kristiansson@saunalahti.fi>
Franck Jullien <franck.jullien@gmail.com>
Jeremy Bennett <jeremy.bennett@embecosm.com>
* configure.tgt: Add targets for or1k and or1knd.
* or1k-tdep.c: New file.
* or1k-tdep.h: New file.
* features/Makefile: Add or1k.xml to build.
* features/or1k.xml: New file.
* features/or1k-core.xml: New file.
* features/or1k.c: Generated.
Diff:
---
gdb/ChangeLog | 14 +
gdb/configure.tgt | 6 +
gdb/doc/ChangeLog | 7 +
gdb/doc/gdb.texinfo | 47 ++
gdb/features/Makefile | 2 +
gdb/features/or1k-core.xml | 65 +++
gdb/features/or1k.c | 77 +++
gdb/features/or1k.xml | 12 +
gdb/or1k-tdep.c | 1294 ++++++++++++++++++++++++++++++++++++++++++++
gdb/or1k-tdep.h | 56 ++
10 files changed, 1580 insertions(+)
diff --git a/gdb/ChangeLog b/gdb/ChangeLog
index 52ccbf5..201a9f3 100644
--- a/gdb/ChangeLog
+++ b/gdb/ChangeLog
@@ -1,3 +1,17 @@
+2017-12-12 Stafford Horne <shorne@gmail.com>
+ Stefan Wallentowitz <stefan@wallentowitz.de>
+ Stefan Kristiansson <stefan.kristiansson@saunalahti.fi>
+ Franck Jullien <franck.jullien@gmail.com>
+ Jeremy Bennett <jeremy.bennett@embecosm.com>
+
+ * configure.tgt: Add targets for or1k and or1knd.
+ * or1k-tdep.c: New file.
+ * or1k-tdep.h: New file.
+ * features/Makefile: Add or1k.xml to build.
+ * features/or1k.xml: New file.
+ * features/or1k-core.xml: New file.
+ * features/or1k.c: Generated.
+
2017-12-12 Alan Modra <amodra@gmail.com>
PR tdep/22576
diff --git a/gdb/configure.tgt b/gdb/configure.tgt
index f720154..fb8014a 100644
--- a/gdb/configure.tgt
+++ b/gdb/configure.tgt
@@ -463,6 +463,12 @@ nios2*-*-*)
gdb_target_obs="nios2-tdep.o"
;;
+or1k-*-* | or1knd-*-*)
+ # Target: OpenCores OpenRISC 1000 32-bit implementation bare metal
+ gdb_target_obs="or1k-tdep.o"
+ gdb_sim=../sim/or1k/libsim.a
+ ;;
+
powerpc*-*-freebsd*)
# Target: FreeBSD/powerpc
gdb_target_obs="rs6000-tdep.o ppc-sysv-tdep.o ppc64-tdep.o \
diff --git a/gdb/doc/ChangeLog b/gdb/doc/ChangeLog
index 7710640..4e032ec 100644
--- a/gdb/doc/ChangeLog
+++ b/gdb/doc/ChangeLog
@@ -1,3 +1,10 @@
+2017-12-12 Stafford Horne <shorne@gmail.com>
+ Stefan Wallentowitz <stefan@wallentowitz.de>
+ Franck Jullien <franck.jullien@gmail.com>
+ Jeremy Bennett <jeremy.bennett@embecosm.com>
+
+ * gdb.texinfo: Add OpenRISC documentation.
+
2017-12-08 Jan Kratochvil <jan.kratochvil@redhat.com>
* gdb.texinfo (Index Files): Document .debug_names and -dwarf-5.
diff --git a/gdb/doc/gdb.texinfo b/gdb/doc/gdb.texinfo
index 062f01d..da3ed28 100644
--- a/gdb/doc/gdb.texinfo
+++ b/gdb/doc/gdb.texinfo
@@ -546,6 +546,11 @@ was developed by SRI International and the University of Cambridge
Computer Laboratory under DARPA/AFRL contract FA8750-10-C-0237
("CTSRD"), as part of the DARPA CRASH research programme.
+The original port to the OpenRISC 1000 is believed to be due to
+Alessandro Forin and Per Bothner. More recent ports have been the work
+of Jeremy Bennett, Franck Jullien, Stefan Wallentowitz and
+Stafford Horne.
+
@node Sample Session
@chapter A Sample @value{GDBN} Session
@@ -22442,6 +22447,7 @@ acceptable commands.
* M68K:: Motorola M68K
* MicroBlaze:: Xilinx MicroBlaze
* MIPS Embedded:: MIPS Embedded
+* OpenRISC 1000:: OpenRISC 1000 (or1k)
* PowerPC Embedded:: PowerPC Embedded
* AVR:: Atmel AVR
* CRIS:: CRIS
@@ -22651,6 +22657,38 @@ As usual, you can inquire about the @code{mipsfpu} variable with
@samp{show mipsfpu}.
@end table
+@node OpenRISC 1000
+@subsection OpenRISC 1000
+@cindex OpenRISC 1000
+
+@noindent
+The OpenRISC 1000 provides a free RISC instruction set architecture. It is
+mainly provided as a soft-core which can run on Xilinx, Altera and other
+FPGA's.
+
+@value{GDBN} for OpenRISC supports the below commands when connecting to
+a target:
+
+@table @code
+
+@kindex target sim
+@item target sim
+
+Runs the builtin CPU simulator which can run very basic
+programs but does not support most hardware functions like MMU.
+For more complex use cases the user is advised to run an external
+target, and connect using @samp{target remote}.
+
+Example: @code{target sim}
+
+@item set debug or1k
+Toggle whether to display OpenRISC-specific debugging messages from the
+OpenRISC target support subsystem.
+
+@item show debug or1k
+Show whether OpenRISC-specific debugging messages are enabled.
+@end table
+
@node PowerPC Embedded
@subsection PowerPC Embedded
@@ -41739,6 +41777,7 @@ registers using the capitalization used in the description.
* M68K Features::
* NDS32 Features::
* Nios II Features::
+* OpenRISC 1000 Features::
* PowerPC Features::
* S/390 and System z Features::
* Sparc Features::
@@ -42025,6 +42064,14 @@ targets. It should contain the 32 core registers (@samp{zero},
@samp{pc}, and the 16 control registers (@samp{status} through
@samp{mpuacc}).
+@node OpenRISC 1000 Features
+@subsection Openrisc 1000 Features
+@cindex target descriptions, OpenRISC 1000 features
+
+The @samp{org.gnu.gdb.or1k.group0} feature is required for OpenRISC 1000
+targets. It should contain the 32 general purpose registers (@samp{r0}
+through @samp{r31}), @samp{ppc}, @samp{npc} and @samp{sr}.
+
@node PowerPC Features
@subsection PowerPC Features
@cindex target descriptions, PowerPC features
diff --git a/gdb/features/Makefile b/gdb/features/Makefile
index 73be34c..c78349e 100644
--- a/gdb/features/Makefile
+++ b/gdb/features/Makefile
@@ -93,6 +93,7 @@ mips64-expedite = r29,pc
mips64-dsp-expedite = r29,pc
microblaze-expedite = r1,rpc
nios2-linux-expedite = sp,pc
+or1k-expedite = r1,npc
powerpc-expedite = r1,pc
rs6000/powerpc-cell32l-expedite = r1,pc,r0,orig_r3,r4
rs6000/powerpc-cell64l-expedite = r1,pc,r0,orig_r3,r4
@@ -136,6 +137,7 @@ XMLTOC = \
mips64-linux.xml \
nds32.xml \
nios2.xml \
+ or1k.xml \
rs6000/powerpc-32.xml \
rs6000/powerpc-32l.xml \
rs6000/powerpc-403.xml \
diff --git a/gdb/features/or1k-core.xml b/gdb/features/or1k-core.xml
new file mode 100644
index 0000000..1a9673d
--- /dev/null
+++ b/gdb/features/or1k-core.xml
@@ -0,0 +1,65 @@
+<?xml version="1.0"?>
+<!-- Copyright (C) 2017 Free Software Foundation, Inc.
+
+ Copying and distribution of this file, with or without modification,
+ are permitted in any medium without royalty provided the copyright
+ notice and this notice are preserved. -->
+
+<!DOCTYPE feature SYSTEM "gdb-target.dtd">
+<feature name="org.gnu.gdb.or1k.group0">
+ <reg name="r0" bitsize="32" type="int"/>
+ <reg name="r1" bitsize="32" type="data_ptr"/>
+ <reg name="r2" bitsize="32" type="data_ptr"/>
+ <reg name="r3" bitsize="32" type="int"/>
+ <reg name="r4" bitsize="32" type="int"/>
+ <reg name="r5" bitsize="32" type="int"/>
+ <reg name="r6" bitsize="32" type="int"/>
+ <reg name="r7" bitsize="32" type="int"/>
+ <reg name="r8" bitsize="32" type="int"/>
+ <reg name="r9" bitsize="32" type="code_ptr"/>
+ <reg name="r10" bitsize="32" type="int"/>
+ <reg name="r11" bitsize="32" type="int"/>
+ <reg name="r12" bitsize="32" type="int"/>
+ <reg name="r13" bitsize="32" type="int"/>
+ <reg name="r14" bitsize="32" type="int"/>
+ <reg name="r15" bitsize="32" type="int"/>
+ <reg name="r16" bitsize="32" type="int"/>
+ <reg name="r17" bitsize="32" type="int"/>
+ <reg name="r18" bitsize="32" type="int"/>
+ <reg name="r19" bitsize="32" type="int"/>
+ <reg name="r20" bitsize="32" type="int"/>
+ <reg name="r21" bitsize="32" type="int"/>
+ <reg name="r22" bitsize="32" type="int"/>
+ <reg name="r23" bitsize="32" type="int"/>
+ <reg name="r24" bitsize="32" type="int"/>
+ <reg name="r25" bitsize="32" type="int"/>
+ <reg name="r26" bitsize="32" type="int"/>
+ <reg name="r27" bitsize="32" type="int"/>
+ <reg name="r28" bitsize="32" type="int"/>
+ <reg name="r29" bitsize="32" type="int"/>
+ <reg name="r30" bitsize="32" type="int"/>
+ <reg name="r31" bitsize="32" type="int"/>
+ <reg name="ppc" bitsize="32" type="code_ptr"/>
+ <reg name="npc" bitsize="32" type="code_ptr"/>
+ <flags id="sr_flags" size="4">
+ <field name="SM" start="0" end="0"/>
+ <field name="TEE" start="1" end="1"/>
+ <field name="IEE" start="2" end="2"/>
+ <field name="DCE" start="3" end="3"/>
+ <field name="ICE" start="4" end="4"/>
+ <field name="DME" start="5" end="5"/>
+ <field name="IME" start="6" end="6"/>
+ <field name="LEE" start="7" end="7"/>
+ <field name="CE" start="8" end="8"/>
+ <field name="F" start="9" end="9"/>
+ <field name="CY" start="10" end="10"/>
+ <field name="OV" start="11" end="11"/>
+ <field name="OVE" start="12" end="12"/>
+ <field name="DSX" start="13" end="13"/>
+ <field name="EPH" start="14" end="14"/>
+ <field name="FO" start="15" end="15"/>
+ <field name="SUMRA" start="16" end="16"/>
+ <field name="CID" start="28" end="31"/>
+ </flags>
+ <reg name="sr" bitsize="32" type="sr_flags"/>
+</feature>
diff --git a/gdb/features/or1k.c b/gdb/features/or1k.c
new file mode 100644
index 0000000..929a5f9
--- /dev/null
+++ b/gdb/features/or1k.c
@@ -0,0 +1,77 @@
+/* THIS FILE IS GENERATED. -*- buffer-read-only: t -*- vi:set ro:
+ Original: or1k.xml */
+
+#include "defs.h"
+#include "osabi.h"
+#include "target-descriptions.h"
+
+struct target_desc *tdesc_or1k;
+static void
+initialize_tdesc_or1k (void)
+{
+ struct target_desc *result = allocate_target_description ();
+ set_tdesc_architecture (result, bfd_scan_arch ("or1k"));
+
+ struct tdesc_feature *feature;
+
+ feature = tdesc_create_feature (result, "org.gnu.gdb.or1k.group0");
+ tdesc_type_with_fields *type_with_fields;
+ tdesc_type *field_type;
+ type_with_fields = tdesc_create_flags (feature, "sr_flags", 4);
+ tdesc_add_flag (type_with_fields, 0, "SM");
+ tdesc_add_flag (type_with_fields, 1, "TEE");
+ tdesc_add_flag (type_with_fields, 2, "IEE");
+ tdesc_add_flag (type_with_fields, 3, "DCE");
+ tdesc_add_flag (type_with_fields, 4, "ICE");
+ tdesc_add_flag (type_with_fields, 5, "DME");
+ tdesc_add_flag (type_with_fields, 6, "IME");
+ tdesc_add_flag (type_with_fields, 7, "LEE");
+ tdesc_add_flag (type_with_fields, 8, "CE");
+ tdesc_add_flag (type_with_fields, 9, "F");
+ tdesc_add_flag (type_with_fields, 10, "CY");
+ tdesc_add_flag (type_with_fields, 11, "OV");
+ tdesc_add_flag (type_with_fields, 12, "OVE");
+ tdesc_add_flag (type_with_fields, 13, "DSX");
+ tdesc_add_flag (type_with_fields, 14, "EPH");
+ tdesc_add_flag (type_with_fields, 15, "FO");
+ tdesc_add_flag (type_with_fields, 16, "SUMRA");
+ tdesc_add_bitfield (type_with_fields, "CID", 28, 31);
+
+ tdesc_create_reg (feature, "r0", 0, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r1", 1, 1, NULL, 32, "data_ptr");
+ tdesc_create_reg (feature, "r2", 2, 1, NULL, 32, "data_ptr");
+ tdesc_create_reg (feature, "r3", 3, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r4", 4, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r5", 5, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r6", 6, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r7", 7, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r8", 8, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r9", 9, 1, NULL, 32, "code_ptr");
+ tdesc_create_reg (feature, "r10", 10, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r11", 11, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r12", 12, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r13", 13, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r14", 14, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r15", 15, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r16", 16, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r17", 17, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r18", 18, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r19", 19, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r20", 20, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r21", 21, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r22", 22, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r23", 23, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r24", 24, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r25", 25, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r26", 26, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r27", 27, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r28", 28, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r29", 29, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r30", 30, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "r31", 31, 1, NULL, 32, "int");
+ tdesc_create_reg (feature, "ppc", 32, 1, NULL, 32, "code_ptr");
+ tdesc_create_reg (feature, "npc", 33, 1, NULL, 32, "code_ptr");
+ tdesc_create_reg (feature, "sr", 34, 1, NULL, 32, "sr_flags");
+
+ tdesc_or1k = result;
+}
diff --git a/gdb/features/or1k.xml b/gdb/features/or1k.xml
new file mode 100644
index 0000000..c138a02
--- /dev/null
+++ b/gdb/features/or1k.xml
@@ -0,0 +1,12 @@
+<?xml version="1.0"?>
+<!-- Copyright (C) 2016 Free Software Foundation, Inc.
+
+ Copying and distribution of this file, with or without modification,
+ are permitted in any medium without royalty provided the copyright
+ notice and this notice are preserved. -->
+
+<!DOCTYPE target SYSTEM "gdb-target.dtd">
+<target>
+ <architecture>or1k</architecture>
+ <xi:include href="or1k-core.xml"/>
+</target>
diff --git a/gdb/or1k-tdep.c b/gdb/or1k-tdep.c
new file mode 100644
index 0000000..54b365e
--- /dev/null
+++ b/gdb/or1k-tdep.c
@@ -0,0 +1,1294 @@
+/* Target-dependent code for the 32-bit OpenRISC 1000, for the GDB.
+ Copyright (C) 2008-2017 Free Software Foundation, Inc.
+
+ This file is part of GDB.
+
+ This program is free software; you can redistribute it and/or modify
+ it under the terms of the GNU General Public License as published by
+ the Free Software Foundation; either version 3 of the License, or
+ (at your option) any later version.
+
+ This program is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ GNU General Public License for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with this program. If not, see <http://www.gnu.org/licenses/>. */
+
+#include "defs.h"
+#include "frame.h"
+#include "inferior.h"
+#include "symtab.h"
+#include "value.h"
+#include "gdbcmd.h"
+#include "language.h"
+#include "gdbcore.h"
+#include "symfile.h"
+#include "objfiles.h"
+#include "gdbtypes.h"
+#include "target.h"
+#include "regcache.h"
+#include "safe-ctype.h"
+#include "block.h"
+#include "reggroups.h"
+#include "arch-utils.h"
+#include "frame.h"
+#include "frame-unwind.h"
+#include "frame-base.h"
+#include "dwarf2-frame.h"
+#include "trad-frame.h"
+#include "regset.h"
+#include "remote.h"
+#include "target-descriptions.h"
+#include <inttypes.h>
+#include "dis-asm.h"
+
+/* OpenRISC specific includes. */
+#include "or1k-tdep.h"
+#include "features/or1k.c"
+
+
+/* Global debug flag. */
+
+static int or1k_debug = 0;
+
+static void
+show_or1k_debug (struct ui_file *file, int from_tty,
+ struct cmd_list_element *c, const char *value)
+{
+ fprintf_filtered (file, _("OpenRISC debugging is %s.\n"), value);
+}
+
+
+/* The target-dependent structure for gdbarch. */
+
+struct gdbarch_tdep
+{
+ int bytes_per_word;
+ int bytes_per_address;
+ CGEN_CPU_DESC gdb_cgen_cpu_desc;
+};
+
+/* Support functions for the architecture definition. */
+
+/* Get an instruction from memory. */
+
+static ULONGEST
+or1k_fetch_instruction (struct gdbarch *gdbarch, CORE_ADDR addr)
+{
+ enum bfd_endian byte_order = gdbarch_byte_order (gdbarch);
+ gdb_byte buf[OR1K_INSTLEN];
+
+ if (target_read_code (addr, buf, OR1K_INSTLEN)) {
+ memory_error (TARGET_XFER_E_IO, addr);
+ }
+
+ return extract_unsigned_integer (buf, OR1K_INSTLEN, byte_order);
+}
+
+/* Generic function to read bits from an instruction. */
+
+static bool
+or1k_analyse_inst (uint32_t inst, const char *format, ...)
+{
+ /* Break out each field in turn, validating as we go. */
+ va_list ap;
+ int i;
+ int iptr = 0; /* Instruction pointer */
+
+ va_start (ap, format);
+
+ for (i = 0; 0 != format[i];)
+ {
+ const char *start_ptr;
+ char *end_ptr;
+
+ uint32_t bits; /* Bit substring of interest */
+ uint32_t width; /* Substring width */
+ uint32_t *arg_ptr;
+
+ switch (format[i])
+ {
+ case ' ':
+ i++;
+ break; /* Formatting: ignored */
+
+ case '0':
+ case '1': /* Constant bit field */
+ bits = (inst >> (OR1K_INSTBITLEN - iptr - 1)) & 0x1;
+
+ if ((format[i] - '0') != bits)
+ return false;
+
+ iptr++;
+ i++;
+ break;
+
+ case '%': /* Bit field */
+ i++;
+ start_ptr = &(format[i]);
+ width = strtoul (start_ptr, &end_ptr, 10);
+
+ /* Check we got something, and if so skip on. */
+ if (start_ptr == end_ptr)
+ error ("bitstring \"%s\" at offset %d has no length field.\n",
+ format, i);
+
+ i += end_ptr - start_ptr;
+
+ /* Look for and skip the terminating 'b'. If it's not there, we
+ still give a fatal error, because these are fixed strings that
+ just should not be wrong. */
+ if ('b' != format[i++])
+ error ("bitstring \"%s\" at offset %d has no terminating 'b'.\n",
+ format, i);
+
+ /* Break out the field. There is a special case with a bit width
+ of 32. */
+ if (32 == width)
+ bits = inst;
+ else
+ bits =
+ (inst >> (OR1K_INSTBITLEN - iptr - width)) & ((1 << width) - 1);
+
+ arg_ptr = va_arg (ap, uint32_t *);
+ *arg_ptr = bits;
+ iptr += width;
+ break;
+
+ default:
+ error ("invalid character in bitstring \"%s\" at offset %d.\n",
+ format, i);
+ break;
+ }
+ }
+
+ /* Is the length OK? */
+ gdb_assert (OR1K_INSTBITLEN == iptr);
+
+ return true; /* Success */
+}
+
+/* This is used to parse l.addi instructions during various prologue
+ analysis routines. The l.addi instruction has semantics:
+
+ assembly: l.addi rD,rA,I
+ implementation: rD = rA + sign_extend(Immediate)
+
+ The rd_ptr, ra_ptr and simm_ptr must be non NULL pointers and are used
+ to store the parse results. Upon successful parsing true is returned,
+ false on failure. */
+
+static bool
+or1k_analyse_l_addi (uint32_t inst, unsigned int *rd_ptr,
+ unsigned int *ra_ptr, int *simm_ptr)
+{
+ /* Instruction fields */
+ uint32_t rd, ra, i;
+
+ if (or1k_analyse_inst (inst, "10 0111 %5b %5b %16b", &rd, &ra, &i))
+ {
+ /* Found it. Construct the result fields. */
+ *rd_ptr = (unsigned int) rd;
+ *ra_ptr = (unsigned int) ra;
+ *simm_ptr = (int) (((i & 0x8000) == 0x8000) ? 0xffff0000 | i : i);
+
+ return true; /* Success */
+ }
+ else
+ return false; /* Failure */
+}
+
+/* This is used to to parse store instructions during various prologue
+ analysis routines. The l.sw instruction has semantics:
+
+ assembly: l.sw I(rA),rB
+ implementation: store rB contents to memory at effective address of
+ rA + sign_extend(Immediate)
+
+ The simm_ptr, ra_ptr and rb_ptr must be non NULL pointers and are used
+ to store the parse results. Upon successful parsing true is returned,
+ false on failure. */
+
+static bool
+or1k_analyse_l_sw (uint32_t inst, int *simm_ptr, unsigned int *ra_ptr,
+ unsigned int *rb_ptr)
+{
+ /* Instruction fields */
+ uint32_t ihi, ilo, ra, rb;
+
+ if (or1k_analyse_inst (inst, "11 0101 %5b %5b %5b %11b", &ihi, &ra, &rb,
+ &ilo))
+
+ {
+ /* Found it. Construct the result fields. */
+ *simm_ptr = (int) ((ihi << 11) | ilo);
+ *simm_ptr |= ((ihi & 0x10) == 0x10) ? 0xffff0000 : 0;
+
+ *ra_ptr = (unsigned int) ra;
+ *rb_ptr = (unsigned int) rb;
+
+ return true; /* Success */
+ }
+ else
+ return false; /* Failure */
+}
+
+
+/* Functions defining the architecture. */
+
+/* Implement the return_value gdbarch method. */
+
+static enum return_value_convention
+or1k_return_value (struct gdbarch *gdbarch, struct value *functype,
+ struct type *valtype, struct regcache *regcache,
+ gdb_byte *readbuf, const gdb_byte *writebuf)
+{
+ enum bfd_endian byte_order = gdbarch_byte_order (gdbarch);
+ enum type_code rv_type = TYPE_CODE (valtype);
+ unsigned int rv_size = TYPE_LENGTH (valtype);
+ int bpw = (gdbarch_tdep (gdbarch))->bytes_per_word;
+
+ /* Deal with struct/union as addresses. If an array won't fit in a
+ single register it is returned as address. Anything larger than 2
+ registers needs to also be passed as address (matches gcc
+ default_return_in_memory). */
+ if ((TYPE_CODE_STRUCT == rv_type) || (TYPE_CODE_UNION == rv_type)
+ || ((TYPE_CODE_ARRAY == rv_type) && (rv_size > bpw))
+ || (rv_size > 2 * bpw))
+ {
+ if (readbuf != NULL)
+ {
+ ULONGEST tmp;
+
+ regcache_cooked_read_unsigned (regcache, OR1K_RV_REGNUM, &tmp);
+ read_memory (tmp, readbuf, rv_size);
+ }
+ if (writebuf != NULL)
+ {
+ ULONGEST tmp;
+
+ regcache_cooked_read_unsigned (regcache, OR1K_RV_REGNUM, &tmp);
+ write_memory (tmp, writebuf, rv_size);
+ }
+
+ return RETURN_VALUE_ABI_RETURNS_ADDRESS;
+ }
+
+ if (rv_size <= bpw)
+ {
+ /* Up to one word scalars are returned in R11. */
+ if (readbuf != NULL)
+ {
+ ULONGEST tmp;
+
+ regcache_cooked_read_unsigned (regcache, OR1K_RV_REGNUM, &tmp);
+ store_unsigned_integer (readbuf, rv_size, byte_order, tmp);
+
+ }
+ if (writebuf != NULL)
+ {
+ gdb_byte *buf = XCNEWVEC(gdb_byte, bpw);
+
+ if (BFD_ENDIAN_BIG == byte_order)
+ memcpy (buf + (sizeof (gdb_byte) * bpw) - rv_size, writebuf,
+ rv_size);
+ else
+ memcpy (buf, writebuf, rv_size);
+
+ regcache_cooked_write (regcache, OR1K_RV_REGNUM, buf);
+
+ free (buf);
+ }
+ }
+ else
+ {
+ /* 2 word scalars are returned in r11/r12 (with the MS word in r11). */
+ if (readbuf != NULL)
+ {
+ ULONGEST tmp_lo;
+ ULONGEST tmp_hi;
+ ULONGEST tmp;
+
+ regcache_cooked_read_unsigned (regcache, OR1K_RV_REGNUM,
+ &tmp_hi);
+ regcache_cooked_read_unsigned (regcache, OR1K_RV_REGNUM + 1,
+ &tmp_lo);
+ tmp = (tmp_hi << (bpw * 8)) | tmp_lo;
+
+ store_unsigned_integer (readbuf, rv_size, byte_order, tmp);
+ }
+ if (writebuf != NULL)
+ {
+ gdb_byte *buf_lo = XCNEWVEC(gdb_byte, bpw);
+ gdb_byte *buf_hi = XCNEWVEC(gdb_byte, bpw);
+
+ /* This is cheating. We assume that we fit in 2 words exactly,
+ which wouldn't work if we had (say) a 6-byte scalar type on a
+ big endian architecture (with the OpenRISC 1000 usually is). */
+ memcpy (buf_hi, writebuf, rv_size - bpw);
+ memcpy (buf_lo, writebuf + bpw, bpw);
+
+ regcache_cooked_write (regcache, OR1K_RV_REGNUM, buf_hi);
+ regcache_cooked_write (regcache, OR1K_RV_REGNUM + 1, buf_lo);
+
+ free (buf_lo);
+ free (buf_hi);
+ }
+ }
+
+ return RETURN_VALUE_REGISTER_CONVENTION;
+}
+
+/* OR1K always uses a l.trap instruction for breakpoints. */
+
+constexpr gdb_byte or1k_break_insn[] = {0x21, 0x00, 0x00, 0x01};
+
+typedef BP_MANIPULATION (or1k_break_insn) or1k_breakpoint;
+
+/* Implement the single_step_through_delay gdbarch method. */
+
+static int
+or1k_single_step_through_delay (struct gdbarch *gdbarch,
+ struct frame_info *this_frame)
+{
+ ULONGEST val;
+ CORE_ADDR ppc;
+ CORE_ADDR npc;
+ CGEN_FIELDS tmp_fields;
+ const CGEN_INSN *insn;
+ struct regcache *regcache = get_current_regcache ();
+ struct gdbarch_tdep *tdep = gdbarch_tdep (gdbarch);
+
+ /* Get the previous and current instruction addresses. If they are not
+ adjacent, we cannot be in a delay slot. */
+ regcache_cooked_read_unsigned (regcache, OR1K_PPC_REGNUM, &val);
+ ppc = (CORE_ADDR) val;
+ regcache_cooked_read_unsigned (regcache, OR1K_NPC_REGNUM, &val);
+ npc = (CORE_ADDR) val;
+
+ if (0x4 != (npc - ppc))
+ return 0;
+
+ insn = cgen_lookup_insn (tdep->gdb_cgen_cpu_desc,
+ NULL,
+ or1k_fetch_instruction (gdbarch, ppc),
+ NULL, 32, &tmp_fields, 0);
+
+ /* NULL here would mean the last instruction was not understood by cgen.
+ This should not usually happen, but if does its not a delay slot. */
+ if (insn == NULL)
+ return 0;
+
+ /* TODO: we should add a delay slot flag to the CGEN_INSN and remove
+ this hard coded test. */
+ return ((CGEN_INSN_NUM (insn) == OR1K_INSN_L_J)
+ || (CGEN_INSN_NUM (insn) == OR1K_INSN_L_JAL)
+ || (CGEN_INSN_NUM (insn) == OR1K_INSN_L_JR)
+ || (CGEN_INSN_NUM (insn) == OR1K_INSN_L_JALR)
+ || (CGEN_INSN_NUM (insn) == OR1K_INSN_L_BNF)
+ || (CGEN_INSN_NUM (insn) == OR1K_INSN_L_BF));
+}
+
+/* Name for or1k general registers. */
+
+static const char *const or1k_reg_names[OR1K_NUM_REGS] = {
+ /* general purpose registers */
+ "r0", "r1", "r2", "r3", "r4", "r5", "r6", "r7",
+ "r8", "r9", "r10", "r11", "r12", "r13", "r14", "r15",
+ "r16", "r17", "r18", "r19", "r20", "r21", "r22", "r23",
+ "r24", "r25", "r26", "r27", "r28", "r29", "r30", "r31",
+
+ /* previous program counter, next program counter and status register */
+ "ppc", "npc", "sr"
+};
+
+static int
+or1k_is_arg_reg (unsigned int regnum)
+{
+ return (OR1K_FIRST_ARG_REGNUM <= regnum)
+ && (regnum <= OR1K_LAST_ARG_REGNUM);
+}
+
+static int
+or1k_is_callee_saved_reg (unsigned int regnum)
+{
+ return (OR1K_FIRST_SAVED_REGNUM <= regnum) && (0 == regnum % 2);
+}
+
+/* Implement the skip_prologue gdbarch method. */
+
+static CORE_ADDR
+or1k_skip_prologue (struct gdbarch *gdbarch, CORE_ADDR pc)
+{
+ CORE_ADDR start_pc;
+ CORE_ADDR addr;
+ uint32_t inst;
+
+ unsigned int ra, rb, rd; /* for instruction analysis */
+ int simm;
+
+ int frame_size = 0;
+
+ /* Try using SAL first if we have symbolic information available. This
+ only works for DWARF 2, not STABS. */
+
+ if (find_pc_partial_function (pc, NULL, &start_pc, NULL))
+ {
+ CORE_ADDR prologue_end = skip_prologue_using_sal (gdbarch, pc);
+
+ if (0 != prologue_end)
+ {
+ struct symtab_and_line prologue_sal = find_pc_line (start_pc, 0);
+ struct compunit_symtab *compunit
+ = SYMTAB_COMPUNIT (prologue_sal.symtab);
+ const char *debug_format = COMPUNIT_DEBUGFORMAT (compunit);
+
+ if ((NULL != debug_format)
+ && (strlen ("dwarf") <= strlen (debug_format))
+ && (0 == strncasecmp ("dwarf", debug_format, strlen ("dwarf"))))
+ return (prologue_end > pc) ? prologue_end : pc;
+ }
+ }
+
+ /* Look to see if we can find any of the standard prologue sequence. All
+ quite difficult, since any or all of it may be missing. So this is
+ just a best guess! */
+
+ addr = pc; /* Where we have got to */
+ inst = or1k_fetch_instruction (gdbarch, addr);
+
+ /* Look for the new stack pointer being set up. */
+ if (or1k_analyse_l_addi (inst, &rd, &ra, &simm)
+ && (OR1K_SP_REGNUM == rd) && (OR1K_SP_REGNUM == ra)
+ && (simm < 0) && (0 == (simm % 4)))
+ {
+ frame_size = -simm;
+ addr += OR1K_INSTLEN;
+ inst = or1k_fetch_instruction (gdbarch, addr);
+ }
+
+ /* Look for the frame pointer being manipulated. */
+ if (or1k_analyse_l_sw (inst, &simm, &ra, &rb)
+ && (OR1K_SP_REGNUM == ra) && (OR1K_FP_REGNUM == rb)
+ && (simm >= 0) && (0 == (simm % 4)))
+ {
+ addr += OR1K_INSTLEN;
+ inst = or1k_fetch_instruction (gdbarch, addr);
+
+ gdb_assert (or1k_analyse_l_addi (inst, &rd, &ra, &simm)
+ && (OR1K_FP_REGNUM == rd) && (OR1K_SP_REGNUM == ra)
+ && (simm == frame_size));
+
+ addr += OR1K_INSTLEN;
+ inst = or1k_fetch_instruction (gdbarch, addr);
+ }
+
+ /* Look for the link register being saved. */
+ if (or1k_analyse_l_sw (inst, &simm, &ra, &rb)
+ && (OR1K_SP_REGNUM == ra) && (OR1K_LR_REGNUM == rb)
+ && (simm >= 0) && (0 == (simm % 4)))
+ {
+ addr += OR1K_INSTLEN;
+ inst = or1k_fetch_instruction (gdbarch, addr);
+ }
+
+ /* Look for arguments or callee-saved register being saved. The register
+ must be one of the arguments (r3-r8) or the 10 callee saved registers
+ (r10, r12, r14, r16, r18, r20, r22, r24, r26, r28, r30). The base
+ register must be the FP (for the args) or the SP (for the callee_saved
+ registers). */
+ while (1)
+ {
+ if (or1k_analyse_l_sw (inst, &simm, &ra, &rb)
+ && (((OR1K_FP_REGNUM == ra) && or1k_is_arg_reg (rb))
+ || ((OR1K_SP_REGNUM == ra) && or1k_is_callee_saved_reg (rb)))
+ && (0 == (simm % 4)))
+ {
+ addr += OR1K_INSTLEN;
+ inst = or1k_fetch_instruction (gdbarch, addr);
+ }
+ else
+ {
+ /* Nothing else to look for. We have found the end of the
+ prologue. */
+ break;
+ }
+ }
+ return addr;
+}
+
+/* Implement the frame_align gdbarch method. */
+
+static CORE_ADDR
+or1k_frame_align (struct gdbarch *gdbarch, CORE_ADDR sp)
+{
+ return align_down (sp, OR1K_STACK_ALIGN);
+}
+
+/* Implement the unwind_pc gdbarch method. */
+
+static CORE_ADDR
+or1k_unwind_pc (struct gdbarch *gdbarch, struct frame_info *next_frame)
+{
+ CORE_ADDR pc;
+
+ if (or1k_debug)
+ fprintf_unfiltered (gdb_stdlog, "or1k_unwind_pc, next_frame=%d\n",
+ frame_relative_level (next_frame));
+
+ pc = frame_unwind_register_unsigned (next_frame, OR1K_NPC_REGNUM);
+
+ if (or1k_debug)
+ fprintf_unfiltered (gdb_stdlog, "or1k_unwind_pc, pc=0x%p\n",
+ (void *) pc);
+
+ return pc;
+}
+
+/* Implement the unwind_sp gdbarch method. */
+
+static CORE_ADDR
+or1k_unwind_sp (struct gdbarch *gdbarch, struct frame_info *next_frame)
+{
+ CORE_ADDR sp;
+
+ if (or1k_debug)
+ fprintf_unfiltered (gdb_stdlog, "or1k_unwind_sp, next_frame=%d\n",
+ frame_relative_level (next_frame));
+
+ sp = frame_unwind_register_unsigned (next_frame, OR1K_SP_REGNUM);
+
+ if (or1k_debug)
+ fprintf_unfiltered (gdb_stdlog, "or1k_unwind_sp, sp=0x%p\n",
+ (void *) sp);
+
+ return sp;
+}
+
+/* Implement the push_dummy_code gdbarch method. */
+
+static CORE_ADDR
+or1k_push_dummy_code (struct gdbarch *gdbarch, CORE_ADDR sp,
+ CORE_ADDR function, struct value **args, int nargs,
+ struct type *value_type, CORE_ADDR * real_pc,
+ CORE_ADDR * bp_addr, struct regcache *regcache)
+{
+ CORE_ADDR bp_slot;
+
+ /* Reserve enough room on the stack for our breakpoint instruction. */
+ bp_slot = sp - 4;
+ /* Store the address of that breakpoint. */
+ *bp_addr = bp_slot;
+ /* keeping the stack aligned. */
+ sp = or1k_frame_align (gdbarch, bp_slot);
+ /* The call starts at the callee's entry point. */
+ *real_pc = function;
+
+ return sp;
+}
+
+/* Implement the push_dummy_call gdbarch method. */
+
+static CORE_ADDR
+or1k_push_dummy_call (struct gdbarch *gdbarch, struct value *function,
+ struct regcache *regcache, CORE_ADDR bp_addr,
+ int nargs, struct value **args, CORE_ADDR sp,
+ int struct_return, CORE_ADDR struct_addr)
+{
+
+ int argreg;
+ int argnum;
+ int first_stack_arg;
+ int stack_offset = 0;
+ int heap_offset = 0;
+ CORE_ADDR heap_sp = sp - 128;
+ enum bfd_endian byte_order = gdbarch_byte_order (gdbarch);
+ int bpa = (gdbarch_tdep (gdbarch))->bytes_per_address;
+ int bpw = (gdbarch_tdep (gdbarch))->bytes_per_word;
+ struct type *func_type = value_type (function);
+
+ /* Return address */
+ regcache_cooked_write_unsigned (regcache, OR1K_LR_REGNUM, bp_addr);
+
+ /* Register for the next argument. */
+ argreg = OR1K_FIRST_ARG_REGNUM;
+
+ /* Location for a returned structure. This is passed as a silent first
+ argument. */
+ if (struct_return)
+ {
+ regcache_cooked_write_unsigned (regcache, OR1K_FIRST_ARG_REGNUM,
+ struct_addr);
+ argreg++;
+ }
+
+ /* Put as many args as possible in registers. */
+ for (argnum = 0; argnum < nargs; argnum++)
+ {
+ const gdb_byte *val;
+ gdb_byte valbuf[sizeof (ULONGEST)];
+
+ struct value *arg = args[argnum];
+ struct type *arg_type = check_typedef (value_type (arg));
+ int len = TYPE_LENGTH (arg_type);
+ enum type_code typecode = TYPE_CODE (arg_type);
+
+ if (TYPE_VARARGS (func_type) && argnum >= TYPE_NFIELDS (func_type))
+ break; /* end or regular args, varargs go to stack. */
+
+ /* Extract the value, either a reference or the data. */
+ if ((TYPE_CODE_STRUCT == typecode) || (TYPE_CODE_UNION == typecode)
+ || (len > bpw * 2))
+ {
+ CORE_ADDR valaddr = value_address (arg);
+
+ /* If the arg is fabricated (i.e. 3*i, instead of i) valaddr is
+ undefined. */
+ if (valaddr == 0)
+ {
+ /* The argument needs to be copied into the target space.
+ Since the bottom of the stack is reserved for function
+ arguments we store this at the these at the top growing
+ down. */
+ heap_offset += align_up (len, bpw);
+ valaddr = heap_sp + heap_offset;
+
+ write_memory (valaddr, value_contents (arg), len);
+ }
+
+ /* The ABI passes all structures by reference, so get its
+ address. */
+ store_unsigned_integer (valbuf, bpa, byte_order, valaddr);
+ len = bpa;
+ val = valbuf;
+ }
+ else
+ {
+ /* Everything else, we just get the value. */
+ val = value_contents (arg);
+ }
+
+ /* Stick the value in a register. */
+ if (len > bpw)
+ {
+ /* Big scalars use two registers, but need NOT be pair aligned. */
+
+ if (argreg <= (OR1K_LAST_ARG_REGNUM - 1))
+ {
+ ULONGEST regval = extract_unsigned_integer (val, len,
+ byte_order);
+
+ unsigned int bits_per_word = bpw * 8;
+ ULONGEST mask = (((ULONGEST) 1) << bits_per_word) - 1;
+ ULONGEST lo = regval & mask;
+ ULONGEST hi = regval >> bits_per_word;
+
+ regcache_cooked_write_unsigned (regcache, argreg, hi);
+ regcache_cooked_write_unsigned (regcache, argreg + 1, lo);
+ argreg += 2;
+ }
+ else
+ {
+ /* Run out of regs */
+ break;
+ }
+ }
+ else if (argreg <= OR1K_LAST_ARG_REGNUM)
+ {
+ /* Smaller scalars fit in a single register. */
+ regcache_cooked_write_unsigned
+ (regcache, argreg, extract_unsigned_integer (val, len,
+ byte_order));
+ argreg++;
+ }
+ else
+ {
+ /* Ran out of regs. */
+ break;
+ }
+ }
+
+ first_stack_arg = argnum;
+
+ /* If we get here with argnum < nargs, then arguments remain to be
+ placed on the stack. This is tricky, since they must be pushed in
+ reverse order and the stack in the end must be aligned. The only
+ solution is to do it in two stages, the first to compute the stack
+ size, the second to save the args. */
+
+ for (argnum = first_stack_arg; argnum < nargs; argnum++)
+ {
+ struct value *arg = args[argnum];
+ struct type *arg_type = check_typedef (value_type (arg));
+ int len = TYPE_LENGTH (arg_type);
+ enum type_code typecode = TYPE_CODE (arg_type);
+
+ if ((TYPE_CODE_STRUCT == typecode) || (TYPE_CODE_UNION == typecode)
+ || (len > bpw * 2))
+ {
+ /* Structures are passed as addresses. */
+ sp -= bpa;
+ }
+ else
+ {
+ /* Big scalars use more than one word. Code here allows for
+ future quad-word entities (e.g. long double.) */
+ sp -= align_up (len, bpw);
+ }
+
+ /* Ensure our dummy heap doesn't touch the stack, this could only
+ happen if we have many arguments including fabricated arguments. */
+ gdb_assert (heap_offset == 0 || ((heap_sp + heap_offset) < sp));
+ }
+
+ sp = gdbarch_frame_align (gdbarch, sp);
+ stack_offset = 0;
+
+ /* Push the remaining args on the stack. */
+ for (argnum = first_stack_arg; argnum < nargs; argnum++)
+ {
+ const gdb_byte *val;
+ gdb_byte valbuf[sizeof (ULONGEST)];
+
+ struct value *arg = args[argnum];
+ struct type *arg_type = check_typedef (value_type (arg));
+ int len = TYPE_LENGTH (arg_type);
+ enum type_code typecode = TYPE_CODE (arg_type);
+ /* The EABI passes structures that do not fit in a register by
+ reference. In all other cases, pass the structure by value. */
+ if ((TYPE_CODE_STRUCT == typecode) || (TYPE_CODE_UNION == typecode)
+ || (len > bpw * 2))
+ {
+ store_unsigned_integer (valbuf, bpa, byte_order,
+ value_address (arg));
+ len = bpa;
+ val = valbuf;
+ }
+ else
+ val = value_contents (arg);
+
+ while (len > 0)
+ {
+ int partial_len = (len < bpw ? len : bpw);
+
+ write_memory (sp + stack_offset, val, partial_len);
+ stack_offset += align_up (partial_len, bpw);
+ len -= partial_len;
+ val += partial_len;
+ }
+ }
+
+ /* Save the updated stack pointer. */
+ regcache_cooked_write_unsigned (regcache, OR1K_SP_REGNUM, sp);
+
+ if (heap_offset > 0)
+ sp = heap_sp;
+
+ return sp;
+}
+
+/* Implement the dummy_id gdbarch method. */
+
+static struct frame_id
+or1k_dummy_id (struct gdbarch *gdbarch, struct frame_info *this_frame)
+{
+ return frame_id_build (get_frame_sp (this_frame),
+ get_frame_pc (this_frame));
+}
+
+
+/* Support functions for frame handling. */
+
+/* Initialize a prologue cache
+
+ We build a cache, saying where registers of the prev frame can be found
+ from the data so far set up in this this.
+
+ We also compute a unique ID for this frame, based on the function start
+ address and the stack pointer (as it will be, even if it has yet to be
+ computed.
+
+ STACK FORMAT
+ ============
+
+ The OR1K has a falling stack frame and a simple prolog. The Stack
+ pointer is R1 and the frame pointer R2. The frame base is therefore the
+ address held in R2 and the stack pointer (R1) is the frame base of the
+ next frame.
+
+ l.addi r1,r1,-frame_size # SP now points to end of new stack frame
+
+ The stack pointer may not be set up in a frameless function (e.g. a
+ simple leaf function).
+
+ l.sw fp_loc(r1),r2 # old FP saved in new stack frame
+ l.addi r2,r1,frame_size # FP now points to base of new stack frame
+
+ The frame pointer is not necessarily saved right at the end of the stack
+ frame - OR1K saves enough space for any args to called functions right
+ at the end (this is a difference from the Architecture Manual).
+
+ l.sw lr_loc(r1),r9 # Link (return) address
+
+ The link register is usally saved at fp_loc - 4. It may not be saved at
+ all in a leaf function.
+
+ l.sw reg_loc(r1),ry # Save any callee saved regs
+
+ The offsets x for the callee saved registers generally (always?) rise in
+ increments of 4, starting at fp_loc + 4. If the frame pointer is
+ omitted (an option to GCC), then it may not be saved at all. There may
+ be no callee saved registers.
+
+ So in summary none of this may be present. However what is present
+ seems always to follow this fixed order, and occur before any
+ substantive code (it is possible for GCC to have more flexible
+ scheduling of the prologue, but this does not seem to occur for OR1K).
+
+ ANALYSIS
+ ========
+
+ This prolog is used, even for -O3 with GCC.
+
+ All this analysis must allow for the possibility that the PC is in the
+ middle of the prologue. Data in the cache should only be set up insofar
+ as it has been computed.
+
+ HOWEVER. The frame_id must be created with the SP *as it will be* at
+ the end of the Prologue. Otherwise a recursive call, checking the frame
+ with the PC at the start address will end up with the same frame_id as
+ the caller.
+
+ A suite of "helper" routines are used, allowing reuse for
+ or1k_skip_prologue().
+
+ Reportedly, this is only valid for frames less than 0x7fff in size. */
+
+static struct trad_frame_cache *
+or1k_frame_cache (struct frame_info *this_frame, void **prologue_cache)
+{
+ struct gdbarch *gdbarch;
+ struct trad_frame_cache *info;
+
+ CORE_ADDR this_pc;
+ CORE_ADDR this_sp;
+ CORE_ADDR this_sp_for_id;
+ int frame_size = 0;
+
+ CORE_ADDR start_addr;
+ CORE_ADDR end_addr;
+
+ if (or1k_debug)
+ fprintf_unfiltered (gdb_stdlog,
+ "or1k_frame_cache, prologue_cache = 0x%p\n",
+ *prologue_cache);
+
+ /* Nothing to do if we already have this info. */
+ if (NULL != *prologue_cache)
+ return (struct trad_frame_cache *) *prologue_cache;
+
+ /* Get a new prologue cache and populate it with default values. */
+ info = trad_frame_cache_zalloc (this_frame);
+ *prologue_cache = info;
+
+ /* Find the start address of this function (which is a normal frame, even
+ if the next frame is the sentinel frame) and the end of its prologue. */
+ this_pc = get_frame_pc (this_frame);
+ find_pc_partial_function (this_pc, NULL, &start_addr, NULL);
+
+ /* Get the stack pointer if we have one (if there's no process executing
+ yet we won't have a frame. */
+ this_sp = (NULL == this_frame) ? 0 :
+ get_frame_register_unsigned (this_frame, OR1K_SP_REGNUM);
+
+ /* Return early if GDB couldn't find the function. */
+ if (start_addr == 0)
+ {
+ if (or1k_debug)
+ fprintf_unfiltered (gdb_stdlog, " couldn't find function\n");
+
+ /* JPB: 28-Apr-11. This is a temporary patch, to get round GDB
+ crashing right at the beginning. Build the frame ID as best we
+ can. */
+ trad_frame_set_id (info, frame_id_build (this_sp, this_pc));
+
+ return info;
+ }
+
+ /* The default frame base of this frame (for ID purposes only - frame
+ base is an overloaded term) is its stack pointer. For now we use the
+ value of the SP register in this frame. However if the PC is in the
+ prologue of this frame, before the SP has been set up, then the value
+ will actually be that of the prev frame, and we'll need to adjust it
+ later. */
+ trad_frame_set_this_base (info, this_sp);
+ this_sp_for_id = this_sp;
+
+ /* The default is to find the PC of the previous frame in the link
+ register of this frame. This may be changed if we find the link
+ register was saved on the stack. */
+ trad_frame_set_reg_realreg (info, OR1K_NPC_REGNUM, OR1K_LR_REGNUM);
+
+ /* We should only examine code that is in the prologue. This is all code
+ up to (but not including) end_addr. We should only populate the cache
+ while the address is up to (but not including) the PC or end_addr,
+ whichever is first. */
+ gdbarch = get_frame_arch (this_frame);
+ end_addr = or1k_skip_prologue (gdbarch, start_addr);
+
+ /* All the following analysis only occurs if we are in the prologue and
+ have executed the code. Check we have a sane prologue size, and if
+ zero we are frameless and can give up here. */
+ if (end_addr < start_addr)
+ error ("end addr 0x%08x is less than start addr 0x%08x\n",
+ (unsigned int) end_addr, (unsigned int) start_addr);
+
+ if (end_addr == start_addr)
+ frame_size = 0;
+ else
+ {
+ /* We have a frame. Look for the various components. */
+ CORE_ADDR addr = start_addr; /* Where we have got to */
+ uint32_t inst = or1k_fetch_instruction (gdbarch, addr);
+
+ unsigned int ra, rb, rd; /* for instruction analysis */
+ int simm;
+
+ /* Look for the new stack pointer being set up. */
+ if (or1k_analyse_l_addi (inst, &rd, &ra, &simm)
+ && (OR1K_SP_REGNUM == rd) && (OR1K_SP_REGNUM == ra)
+ && (simm < 0) && (0 == (simm % 4)))
+ {
+ frame_size = -simm;
+ addr += OR1K_INSTLEN;
+ inst = or1k_fetch_instruction (gdbarch, addr);
+
+ /* If the PC has not actually got to this point, then the frame
+ base will be wrong, and we adjust it.
+
+ If we are past this point, then we need to populate the stack
+ accordingly. */
+ if (this_pc <= addr)
+ {
+ /* Only do if executing. */
+ if (0 != this_sp)
+ {
+ this_sp_for_id = this_sp + frame_size;
+ trad_frame_set_this_base (info, this_sp_for_id);
+ }
+ }
+ else
+ {
+ /* We are past this point, so the stack pointer of the prev
+ frame is frame_size greater than the stack pointer of this
+ frame. */
+ trad_frame_set_reg_value (info, OR1K_SP_REGNUM,
+ this_sp + frame_size);
+ }
+ }
+
+ /* From now on we are only populating the cache, so we stop once we
+ get to either the end OR the current PC. */
+ end_addr = (this_pc < end_addr) ? this_pc : end_addr;
+
+ /* Look for the frame pointer being manipulated. */
+ if ((addr < end_addr)
+ && or1k_analyse_l_sw (inst, &simm, &ra, &rb)
+ && (OR1K_SP_REGNUM == ra) && (OR1K_FP_REGNUM == rb)
+ && (simm >= 0) && (0 == (simm % 4)))
+ {
+ addr += OR1K_INSTLEN;
+ inst = or1k_fetch_instruction (gdbarch, addr);
+
+ /* At this stage, we can find the frame pointer of the previous
+ frame on the stack of the current frame. */
+ trad_frame_set_reg_addr (info, OR1K_FP_REGNUM, this_sp + simm);
+
+ /* Look for the new frame pointer being set up. */
+ if ((addr < end_addr)
+ && or1k_analyse_l_addi (inst, &rd, &ra, &simm)
+ && (OR1K_FP_REGNUM == rd) && (OR1K_SP_REGNUM == ra)
+ && (simm == frame_size))
+ {
+ addr += OR1K_INSTLEN;
+ inst = or1k_fetch_instruction (gdbarch, addr);
+
+ /* If we have got this far, the stack pointer of the previous
+ frame is the frame pointer of this frame. */
+ trad_frame_set_reg_realreg (info, OR1K_SP_REGNUM,
+ OR1K_FP_REGNUM);
+ }
+ }
+
+ /* Look for the link register being saved. */
+ if ((addr < end_addr)
+ && or1k_analyse_l_sw (inst, &simm, &ra, &rb)
+ && (OR1K_SP_REGNUM == ra) && (OR1K_LR_REGNUM == rb)
+ && (simm >= 0) && (0 == (simm % 4)))
+ {
+ addr += OR1K_INSTLEN;
+ inst = or1k_fetch_instruction (gdbarch, addr);
+
+ /* If the link register is saved in the this frame, it holds the
+ value of the PC in the previous frame. This overwrites the
+ previous information about finding the PC in the link
+ register. */
+ trad_frame_set_reg_addr (info, OR1K_NPC_REGNUM, this_sp + simm);
+ }
+
+ /* Look for arguments or callee-saved register being saved. The
+ register must be one of the arguments (r3-r8) or the 10 callee
+ saved registers (r10, r12, r14, r16, r18, r20, r22, r24, r26, r28,
+ r30). The base register must be the FP (for the args) or the SP
+ (for the callee_saved registers). */
+ while (addr < end_addr)
+ {
+ if (or1k_analyse_l_sw (inst, &simm, &ra, &rb)
+ && (((OR1K_FP_REGNUM == ra) && or1k_is_arg_reg (rb))
+ || ((OR1K_SP_REGNUM == ra)
+ && or1k_is_callee_saved_reg (rb)))
+ && (0 == (simm % 4)))
+ {
+ addr += OR1K_INSTLEN;
+ inst = or1k_fetch_instruction (gdbarch, addr);
+
+ /* The register in the previous frame can be found at this
+ location in this frame. */
+ trad_frame_set_reg_addr (info, rb, this_sp + simm);
+ }
+ else
+ break; /* Not a register save instruction. */
+ }
+ }
+
+ /* Build the frame ID */
+ trad_frame_set_id (info, frame_id_build (this_sp_for_id, start_addr));
+
+ if (or1k_debug)
+ {
+ fprintf_unfiltered (gdb_stdlog, " this_sp_for_id = 0x%p\n",
+ (void *) this_sp_for_id);
+ fprintf_unfiltered (gdb_stdlog, " start_addr = 0x%p\n",
+ (void *) start_addr);
+ }
+
+ return info;
+}
+
+/* Implement the this_id function for the stub unwinder. */
+
+static void
+or1k_frame_this_id (struct frame_info *this_frame,
+ void **prologue_cache, struct frame_id *this_id)
+{
+ struct trad_frame_cache *info = or1k_frame_cache (this_frame,
+ prologue_cache);
+
+ trad_frame_get_id (info, this_id);
+}
+
+/* Implement the prev_register function for the stub unwinder. */
+
+static struct value *
+or1k_frame_prev_register (struct frame_info *this_frame,
+ void **prologue_cache, int regnum)
+{
+ struct trad_frame_cache *info = or1k_frame_cache (this_frame,
+ prologue_cache);
+
+ return trad_frame_get_register (info, this_frame, regnum);
+}
+
+/* Data structures for the normal prologue-analysis-based unwinder. */
+
+static const struct frame_unwind or1k_frame_unwind = {
+ NORMAL_FRAME,
+ default_frame_unwind_stop_reason,
+ or1k_frame_this_id,
+ or1k_frame_prev_register,
+ NULL,
+ default_frame_sniffer,
+ NULL,
+};
+
+/* Architecture initialization for OpenRISC 1000. */
+
+static struct gdbarch *
+or1k_gdbarch_init (struct gdbarch_info info, struct gdbarch_list *arches)
+{
+ struct gdbarch *gdbarch;
+ struct gdbarch_tdep *tdep;
+ const struct bfd_arch_info *binfo;
+ struct tdesc_arch_data *tdesc_data = NULL;
+ const struct target_desc *tdesc = info.target_desc;
+
+ /* Find a candidate among the list of pre-declared architectures. */
+ arches = gdbarch_list_lookup_by_info (arches, &info);
+ if (NULL != arches)
+ return arches->gdbarch;
+
+ /* None found, create a new architecture from the information
+ provided. Can't initialize all the target dependencies until we
+ actually know which target we are talking to, but put in some defaults
+ for now. */
+ binfo = info.bfd_arch_info;
+ tdep = XCNEW (struct gdbarch_tdep);
+ tdep->bytes_per_word = binfo->bits_per_word / binfo->bits_per_byte;
+ tdep->bytes_per_address = binfo->bits_per_address / binfo->bits_per_byte;
+ gdbarch = gdbarch_alloc (&info, tdep);
+
+ /* Target data types */
+ set_gdbarch_short_bit (gdbarch, 16);
+ set_gdbarch_int_bit (gdbarch, 32);
+ set_gdbarch_long_bit (gdbarch, 32);
+ set_gdbarch_long_long_bit (gdbarch, 64);
+ set_gdbarch_float_bit (gdbarch, 32);
+ set_gdbarch_float_format (gdbarch, floatformats_ieee_single);
+ set_gdbarch_double_bit (gdbarch, 64);
+ set_gdbarch_double_format (gdbarch, floatformats_ieee_double);
+ set_gdbarch_long_double_bit (gdbarch, 64);
+ set_gdbarch_long_double_format (gdbarch, floatformats_ieee_double);
+ set_gdbarch_ptr_bit (gdbarch, binfo->bits_per_address);
+ set_gdbarch_addr_bit (gdbarch, binfo->bits_per_address);
+ set_gdbarch_char_signed (gdbarch, 1);
+
+ /* Information about the target architecture */
+ set_gdbarch_return_value (gdbarch, or1k_return_value);
+ set_gdbarch_breakpoint_kind_from_pc (gdbarch,
+ or1k_breakpoint::kind_from_pc);
+ set_gdbarch_sw_breakpoint_from_kind (gdbarch,
+ or1k_breakpoint::bp_from_kind);
+ set_gdbarch_have_nonsteppable_watchpoint (gdbarch, 1);
+
+ /* Register architecture */
+ set_gdbarch_num_regs (gdbarch, OR1K_NUM_REGS);
+ set_gdbarch_num_pseudo_regs (gdbarch, OR1K_NUM_PSEUDO_REGS);
+ set_gdbarch_sp_regnum (gdbarch, OR1K_SP_REGNUM);
+ set_gdbarch_pc_regnum (gdbarch, OR1K_NPC_REGNUM);
+ set_gdbarch_ps_regnum (gdbarch, OR1K_SR_REGNUM);
+ set_gdbarch_deprecated_fp_regnum (gdbarch, OR1K_FP_REGNUM);
+
+ /* Functions to analyse frames */
+ set_gdbarch_skip_prologue (gdbarch, or1k_skip_prologue);
+ set_gdbarch_inner_than (gdbarch, core_addr_lessthan);
+ set_gdbarch_frame_align (gdbarch, or1k_frame_align);
+ set_gdbarch_frame_red_zone_size (gdbarch, OR1K_FRAME_RED_ZONE_SIZE);
+
+ /* Functions to access frame data */
+ set_gdbarch_unwind_pc (gdbarch, or1k_unwind_pc);
+ set_gdbarch_unwind_sp (gdbarch, or1k_unwind_sp);
+
+ /* Functions handling dummy frames */
+ set_gdbarch_call_dummy_location (gdbarch, ON_STACK);
+ set_gdbarch_push_dummy_code (gdbarch, or1k_push_dummy_code);
+ set_gdbarch_push_dummy_call (gdbarch, or1k_push_dummy_call);
+ set_gdbarch_dummy_id (gdbarch, or1k_dummy_id);
+
+ /* Frame unwinders. Use DWARF debug info if available, otherwise use our
+ own unwinder. */
+ dwarf2_append_unwinders (gdbarch);
+ frame_unwind_append_unwinder (gdbarch, &or1k_frame_unwind);
+
+ /* Get a CGEN CPU descriptor for this architecture. */
+ {
+
+ const char *mach_name = binfo->printable_name;
+ enum cgen_endian endian = (info.byte_order == BFD_ENDIAN_BIG
+ ? CGEN_ENDIAN_BIG : CGEN_ENDIAN_LITTLE);
+
+ tdep->gdb_cgen_cpu_desc =
+ or1k_cgen_cpu_open (CGEN_CPU_OPEN_BFDMACH, mach_name,
+ CGEN_CPU_OPEN_ENDIAN, endian, CGEN_CPU_OPEN_END);
+
+ or1k_cgen_init_asm (tdep->gdb_cgen_cpu_desc);
+ }
+
+ /* If this mach has a delay slot. */
+ if (binfo->mach == bfd_mach_or1k)
+ set_gdbarch_single_step_through_delay (gdbarch,
+ or1k_single_step_through_delay);
+
+ if (!tdesc_has_registers (info.target_desc))
+ /* Pick a default target description. */
+ tdesc = tdesc_or1k;
+
+ /* Check any target description for validity. */
+ if (tdesc_has_registers (tdesc))
+ {
+ const struct tdesc_feature *feature;
+ int valid_p;
+ int i;
+
+ feature = tdesc_find_feature (tdesc, "org.gnu.gdb.or1k.group0");
+ if (feature == NULL)
+ return NULL;
+
+ tdesc_data = tdesc_data_alloc ();
+
+ valid_p = 1;
+
+ for (i = 0; i < OR1K_NUM_REGS; i++)
+ valid_p &= tdesc_numbered_register (feature, tdesc_data, i,
+ or1k_reg_names[i]);
+
+ if (!valid_p)
+ {
+ tdesc_data_cleanup (tdesc_data);
+ return NULL;
+ }
+ }
+
+ if (tdesc_data != NULL)
+ {
+ /* If we are using tdesc, register our own reggroups, otherwise we
+ will used the defaults. */
+ reggroup_add (gdbarch, general_reggroup);
+ reggroup_add (gdbarch, system_reggroup);
+ reggroup_add (gdbarch, float_reggroup);
+ reggroup_add (gdbarch, vector_reggroup);
+ reggroup_add (gdbarch, all_reggroup);
+ reggroup_add (gdbarch, save_reggroup);
+ reggroup_add (gdbarch, restore_reggroup);
+
+ tdesc_use_registers (gdbarch, tdesc, tdesc_data);
+ }
+
+ return gdbarch;
+}
+
+/* Dump the target specific data for this architecture. */
+
+static void
+or1k_dump_tdep (struct gdbarch *gdbarch, struct ui_file *file)
+{
+ struct gdbarch_tdep *tdep = gdbarch_tdep (gdbarch);
+
+ if (NULL == tdep)
+ return; /* Nothing to report */
+
+ fprintf_unfiltered (file, "or1k_dump_tdep: %d bytes per word\n",
+ tdep->bytes_per_word);
+ fprintf_unfiltered (file, "or1k_dump_tdep: %d bytes per address\n",
+ tdep->bytes_per_address);
+}
+
+
+void
+_initialize_or1k_tdep (void)
+{
+ /* Register this architecture. */
+ gdbarch_register (bfd_arch_or1k, or1k_gdbarch_init, or1k_dump_tdep);
+
+ initialize_tdesc_or1k ();
+
+ /* Debugging flag. */
+ add_setshow_boolean_cmd ("or1k", class_maintenance, &or1k_debug,
+ _("Set OpenRISC debugging."),
+ _("Show OpenRISC debugging."),
+ _("When on, OpenRISC specific debugging is enabled."),
+ NULL,
+ show_or1k_debug,
+ &setdebuglist, &showdebuglist);
+}
diff --git a/gdb/or1k-tdep.h b/gdb/or1k-tdep.h
new file mode 100644
index 0000000..edcad88
--- /dev/null
+++ b/gdb/or1k-tdep.h
@@ -0,0 +1,56 @@
+/* Definitions to target GDB to OpenRISC 1000 32-bit targets.
+ Copyright (C) 2008-2017 Free Software Foundation, Inc.
+
+ This file is part of GDB.
+
+ This program is free software; you can redistribute it and/or modify it
+ under the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 3 of the License, or (at your option)
+ any later version.
+
+ This program is distributed in the hope that it will be useful, but WITHOUT
+ ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
+ more details.
+
+ You should have received a copy of the GNU General Public License along
+ With this program. If not, see <http://www.gnu.org/licenses/>. */
+
+
+#ifndef OR1K_TDEP__H
+#define OR1K_TDEP__H
+
+#ifndef TARGET_OR1K
+#define TARGET_OR1K
+#endif
+
+#include "opcodes/or1k-desc.h"
+#include "opcodes/or1k-opc.h"
+
+/* General Purpose Registers */
+#define OR1K_ZERO_REGNUM 0
+#define OR1K_SP_REGNUM 1
+#define OR1K_FP_REGNUM 2
+#define OR1K_FIRST_ARG_REGNUM 3
+#define OR1K_LAST_ARG_REGNUM 8
+#define OR1K_LR_REGNUM 9
+#define OR1K_FIRST_SAVED_REGNUM 10
+#define OR1K_RV_REGNUM 11
+#define OR1K_PPC_REGNUM (OR1K_MAX_GPR_REGS + 0)
+#define OR1K_NPC_REGNUM (OR1K_MAX_GPR_REGS + 1)
+#define OR1K_SR_REGNUM (OR1K_MAX_GPR_REGS + 2)
+
+/* Properties of the architecture. GDB mapping of registers is all the GPRs
+ and SPRs followed by the PPC, NPC and SR at the end. Red zone is the area
+ past the end of the stack reserved for exception handlers etc. */
+
+#define OR1K_MAX_GPR_REGS 32
+#define OR1K_NUM_PSEUDO_REGS 0
+#define OR1K_NUM_REGS (OR1K_MAX_GPR_REGS + 3)
+#define OR1K_STACK_ALIGN 4
+#define OR1K_INSTLEN 4
+#define OR1K_INSTBITLEN (OR1K_INSTLEN * 8)
+#define OR1K_NUM_TAP_RECORDS 8
+#define OR1K_FRAME_RED_ZONE_SIZE 2536
+
+#endif /* OR1K_TDEP__H */