* [PATCH] Add support for Xilinx MicroBlaze
@ 2009-09-23 18:33 Michael Eager
2009-09-25 9:15 ` Eli Zaretskii
2009-09-26 0:08 ` Joel Brobecker
0 siblings, 2 replies; 6+ messages in thread
From: Michael Eager @ 2009-09-23 18:33 UTC (permalink / raw)
To: gdb-patches
[-- Attachment #1: Type: text/plain, Size: 876 bytes --]
Hi --
Attached is a revised patch which adds support for the
Xilinx MicroBlaze. I believe that I've addressed all
of the previous comments.
This patch requires a small patch to binutils/opcodes
which is being submitted shortly.
The simulator for Xilinx MicroBlaze has been approved
by the maintainers and will be checked in.
2009-09-23 Michael Eager <eager@eagercon.com>
* configure.tgt: Add targets microblaze*-linux-*, microblaze*-xilinx-*.
* doc/gdb.texinfo: Add MicroBlaze.
* MAINTAINERS: Add self as maintainer for MicroBlaze.
* Makefile.in: Build microblaze-tdep.o, microblaze-linux-tdep.o.
* microblaze-linux-tdep.c: New.
* microblaze-rom.c: New.
* microblaze-tdep.c: New.
* microblaze-tdep.h: New.
* NEWS: Announce Xilinx MicroBlaze support.
--
Michael Eager eager@eagercon.com
1960 Park Blvd., Palo Alto, CA 94306 650-325-8077
[-- Attachment #2: mb-gdb.patch --]
[-- Type: text/plain, Size: 52014 bytes --]
diff -urNp --exclude '*.swp' --exclude DEV-PHASE --exclude .svn --exclude CVS gdb/gdb/configure.tgt mb-gdb/gdb/configure.tgt
--- gdb/gdb/configure.tgt 2009-08-06 03:28:38.000000000 -0700
+++ mb-gdb/gdb/configure.tgt 2009-09-21 12:12:32.000000000 -0700
@@ -313,6 +313,19 @@ mep-*-*)
# No sim needed. Target uses SID.
;;
+microblaze*-linux-*)
+ # Target: Xilinx MicroBlaze running Linux
+ gdb_target_obs="microblaze-tdep.o microblaze-linux-tdep.o microblaze-rom.o \
+ monitor.o dsrec.o solib.o solib-svr4.o corelow.o \
+ symfile-mem.o"
+ gdb_sim=../sim/microblaze/libsim.a
+ ;;
+microblaze*-xilinx-*)
+ # Target: Xilinx MicroBlaze running standalone
+ gdb_target_obs="microblaze-tdep.o microblaze-rom.o monitor.o dsrec.o"
+ gdb_sim=../sim/microblaze/libsim.a
+ ;;
+
mips*-sgi-irix5*)
# Target: MIPS SGI running Irix 5
gdb_target_obs="mips-tdep.o mips-irix-tdep.o solib.o solib-irix.o"
diff -urNp --exclude '*.swp' --exclude DEV-PHASE --exclude .svn --exclude CVS gdb/gdb/doc/gdb.texinfo mb-gdb/gdb/doc/gdb.texinfo
--- gdb/gdb/doc/gdb.texinfo 2009-09-06 10:14:43.000000000 -0700
+++ mb-gdb/gdb/doc/gdb.texinfo 2009-09-23 11:02:38.000000000 -0700
@@ -517,6 +517,9 @@ Tensilica, Inc.@: contributed support fo
who have worked on the Xtensa port of @value{GDBN} in the past include
Steve Tjiang, John Newlin, and Scott Foehner.
+Michael Eager and staff of Xilinx, Inc., contributed support for the
+Xilinx MicroBlaze architecture.
+
@node Sample Session
@chapter A Sample @value{GDBN} Session
@@ -16427,6 +16430,7 @@ acceptable commands.
* ARM:: ARM RDI
* M32R/D:: Renesas M32R/D
* M68K:: Motorola M68K
+* MicroBlaze:: Xilinx MicroBlaze
* MIPS Embedded:: MIPS Embedded
* OpenRISC 1000:: OpenRisc 1000
* PA:: HP PA Embedded
@@ -16678,6 +16682,38 @@ dBUG ROM monitor for Motorola ColdFire.
@end table
+@node MicroBlaze
+@subsection MicroBlaze
+@cindex Xilinx MicroBlaze
+@cindex XMD, Xilinx Microprocessor Debugger
+
+The MicroBlaze is a soft-core processor supported on various Xilinx
+FPGAs, such as Spartan or Virtex series. Boards with these processors
+usually have JTAG ports which connect to a host system running the Xilinx
+Embedded Development Kit (EDK) or Software Development Kit (SDK).
+This host system is used to download the configuration bitstream to
+the target FPGA. The Xilinx Microprocessor Debugger (XMD) program
+communicates with the target board using the JTAG interface and
+presents a @code{gdbserver} interface to the board. By default
+@code{xmd} uses port @code{1234}. (While it is possible to change
+this default port, it requires the use of undocumented @code{xmd}
+commands. Contact Xilinx support if you need to do this.)
+
+Use these GDB commands to connect to the MicroBlaze target processor.
+
+@table @code
+@item target remote :1234
+Use this command to connect to the target if you are running @value{GDBN}
+on the same system as @code{xmd}.
+
+@item target remote XMD-HOST:1234
+Use this command to connect to the target if it is connected to @code{xmd}
+running on a different system named @code{XMD-HOST}.
+
+@item load
+Use this command to download a program to the MicroBlaze target.
+@end table
+
@node MIPS Embedded
@subsection MIPS Embedded
diff -urNp --exclude '*.swp' --exclude DEV-PHASE --exclude .svn --exclude CVS gdb/gdb/MAINTAINERS mb-gdb/gdb/MAINTAINERS
--- gdb/gdb/MAINTAINERS 2009-08-06 03:28:37.000000000 -0700
+++ mb-gdb/gdb/MAINTAINERS 2009-09-06 12:26:05.000000000 -0700
@@ -290,6 +290,10 @@ the native maintainer when resolving ABI
mep --target=mep-elf ,-Werror
Kevin Buettner kevinb@redhat.com
+ microblaze --target=microblaze-xilinx-elf ,-Werror
+ --target=microblaze-linux-gnu ,-Werror
+ Michael Eager eager@eagercon.com
+
mips --target=mips-elf ,-Werror
mn10300 --target=mn10300-elf broken
diff -urNp --exclude '*.swp' --exclude DEV-PHASE --exclude .svn --exclude CVS gdb/gdb/Makefile.in mb-gdb/gdb/Makefile.in
--- gdb/gdb/Makefile.in 2009-08-27 09:08:34.000000000 -0700
+++ mb-gdb/gdb/Makefile.in 2009-08-27 12:20:01.000000000 -0700
@@ -498,6 +498,7 @@ ALL_TARGET_OBS = \
m68kbsd-tdep.o m68klinux-tdep.o m68k-tdep.o \
m88k-tdep.o \
mep-tdep.o \
+ microblaze-tdep.o microblaze-linux-tdep.o \
mips-irix-tdep.o mips-linux-tdep.o \
mipsnbsd-tdep.o mips-tdep.o \
mn10300-linux-tdep.o mn10300-tdep.o \
@@ -1390,6 +1391,7 @@ ALLDEPFILES = \
m68kbsd-nat.c m68kbsd-tdep.c \
m68klinux-nat.c m68klinux-tdep.c \
m88k-tdep.c m88kbsd-nat.c \
+ microblaze-tdep.c microblaze-linux-tdep.c \
mingw-hdep.c \
mips-linux-nat.c mips-linux-tdep.c \
mips-irix-tdep.c \
diff -urNp --exclude '*.swp' --exclude DEV-PHASE --exclude .svn --exclude CVS gdb/gdb/microblaze-linux-tdep.c mb-gdb/gdb/microblaze-linux-tdep.c
--- gdb/gdb/microblaze-linux-tdep.c 1969-12-31 16:00:00.000000000 -0800
+++ mb-gdb/gdb/microblaze-linux-tdep.c 2009-09-21 12:23:33.000000000 -0700
@@ -0,0 +1,139 @@
+/* Target-dependent code for Xilinx MicroBlaze.
+ Copyright 2009 Free Software Foundation, Inc.
+
+ 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 2 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, write to the Free Software
+ Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "defs.h"
+#include "frame.h"
+#include "inferior.h"
+#include "symtab.h"
+#include "target.h"
+#include "gdbcore.h"
+#include "gdbcmd.h"
+#include "symfile.h"
+#include "objfiles.h"
+#include "regcache.h"
+#include "value.h"
+#include "osabi.h"
+#include "regset.h"
+#include "solib-svr4.h"
+#include "microblaze-tdep.h"
+#include "trad-frame.h"
+#include "frame-unwind.h"
+#include "tramp-frame.h"
+
+
+static int
+microblaze_linux_memory_remove_breakpoint (struct bp_target_info *bp_tgt)
+{
+ CORE_ADDR addr = bp_tgt->placed_address;
+ const gdb_byte *bp;
+ int val;
+ int bplen;
+ gdb_byte old_contents[BREAKPOINT_MAX];
+
+ /* Determine appropriate breakpoint contents and size for this address. */
+ bp = gdbarch_breakpoint_from_pc (gdbarch, &addr, &bplen);
+ if (bp == NULL)
+ error (_("Software breakpoints not implemented for this target."));
+
+ val = target_read_memory (addr, old_contents, bplen);
+
+ /* If our breakpoint is no longer at the address, this means that the
+ program modified the code on us, so it is wrong to put back the
+ old value. */
+ if ((val == 0) && (memcmp (bp, old_contents, bplen) == 0))
+ val = target_write_memory (addr, bp_tgt->shadow_contents, bplen);
+
+ return val;
+}
+
+static void
+microblaze_linux_sigtramp_cache (struct frame_info *next_frame,
+ struct trad_frame_cache *this_cache,
+ CORE_ADDR func, LONGEST offset,
+ int bias)
+{
+ CORE_ADDR base;
+ CORE_ADDR gpregs;
+ int regnum;
+ struct gdbarch *gdbarch = get_frame_arch (next_frame);
+ struct gdbarch_tdep *tdep = gdbarch_tdep (gdbarch);
+
+ base = frame_unwind_register_unsigned (next_frame, SP_REGNUM);
+ if ((bias > 0) && (frame_pc_unwind (next_frame) != func))
+ /* See below, some signal trampolines increment the stack as their
+ first instruction, need to compensate for that. */
+ base -= bias;
+
+ /* Find the address of the register buffer. */
+ gpregs = base + offset;
+
+ /* Registers saved on stack. */
+ for (regnum = 0; regnum < MICROBLAZE_BTR_REGNUM; regnum++)
+ trad_frame_set_reg_addr (this_cache, regnum, gpregs + regnum * REGISTER_SIZE);
+ trad_frame_set_id (this_cache, frame_id_build (base, func));
+}
+
+
+static void
+microblaze_linux_sighandler_cache_init (const struct tramp_frame *self,
+ struct frame_info *next_frame,
+ struct trad_frame_cache *this_cache,
+ CORE_ADDR func)
+{
+ microblaze_linux_sigtramp_cache (next_frame, this_cache, func,
+ 0 /* Offset to ucontext_t. */
+ + 24 /* Offset to .reg. */,
+ 0);
+}
+
+static struct tramp_frame microblaze_linux_sighandler_tramp_frame = {
+ SIGTRAMP_FRAME,
+ 4,
+ {
+ { 0x31800077, -1 }, /* addik R12,R0,119. */
+ { 0xb9cc0008, -1 }, /* brki R14,8. */
+ { TRAMP_SENTINEL_INSN },
+ },
+ microblaze_linux_sighandler_cache_init
+};
+
+
+static void
+microblaze_linux_init_abi (struct gdbarch_info info,
+ struct gdbarch *gdbarch)
+{
+ struct gdbarch_tdep *tdep = gdbarch_tdep (gdbarch);
+
+ set_gdbarch_memory_remove_breakpoint (gdbarch,
+ microblaze_linux_memory_remove_breakpoint);
+
+ /* Shared library handling. */
+ set_solib_svr4_fetch_link_map_offsets (gdbarch,
+ svr4_ilp32_fetch_link_map_offsets);
+
+ /* Trampolines. */
+ tramp_frame_prepend_unwinder (gdbarch,
+ µblaze_linux_sighandler_tramp_frame);
+}
+
+void
+_initialize_microblaze_linux_tdep (void)
+{
+ gdbarch_register_osabi (bfd_arch_microblaze, 0,
+ GDB_OSABI_LINUX, microblaze_linux_init_abi);
+}
diff -urNp --exclude '*.swp' --exclude DEV-PHASE --exclude .svn --exclude CVS gdb/gdb/microblaze-rom.c mb-gdb/gdb/microblaze-rom.c
--- gdb/gdb/microblaze-rom.c 1969-12-31 16:00:00.000000000 -0800
+++ mb-gdb/gdb/microblaze-rom.c 2009-09-21 12:23:30.000000000 -0700
@@ -0,0 +1,190 @@
+/* Remote debugging interface to Xilinx MicroBlaze.
+ Copyright 2009 Free Software Foundation, Inc.
+
+ 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 2 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, write to the Free Software
+ Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "defs.h"
+#include "gdbcore.h"
+#include "target.h"
+#include "monitor.h"
+#include "gdb_string.h"
+#include "serial.h"
+#include "regcache.h"
+
+static char *picobug_inits[] =
+{"\r", NULL};
+
+static struct target_ops picobug_ops;
+static struct monitor_ops picobug_cmds;
+
+/* Picobug only supports a subset of registers from MCore. In reality,
+ it doesn't support ss1, either. */
+static char *picobug_regnames[] = {
+ "r0", "r1", "r2", "r3", "r4", "r5", "r6", "r7",
+ "r8", "r9", "r10", "r11", "r12", "r13", "r14", "r15",
+ 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0,
+ "psr", "vbr", "epsr", "fpsr", "epc", "fpc", 0, "ss1",
+ "ss2", "ss3", "ss4", 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0,
+ "pc" };
+
+
+
+void
+picobug_open (char *args, int from_tty)
+{
+ monitor_open (args, &picobug_cmds, from_tty);
+}
+/* We choose to write our own dumpregs routine, since the output of
+ the register dumping is rather difficult to encapsulate in a
+ regexp:
+
+picobug> rd
+ pc 2f00031e epc 2f00031e fpc 00000000
+ psr 80000101 epsr 80000101 fpsr 00000000
+ss0-ss4 bad0beef 00000000 00000000 00000000 00000000 vbr 30005c00
+ r0-r7 2f0fff4c 00000090 00000001 00000002 00000003 00000004 00000005 00000006
+ r8-r15 2f0fff64 00000000 00000000 00000000 00000000 00000000 00000000 2f00031e
+*/
+
+static int
+picobug_dumpregs (struct regcache *regcache)
+{
+ struct gdbarch *gdbarch = get_regcache_arch (regcache);
+ char buf[1024];
+ int resp_len;
+ char *p;
+
+ /* Send the dump register command to the monitor and
+ get the reply. */
+ monitor_printf (picobug_cmds.dump_registers);
+ resp_len = monitor_expect_prompt (buf, sizeof (buf));
+
+ p = strtok (buf, " \t\r\n");
+ while (p)
+ {
+ if (strchr (p, '-'))
+ {
+ /* got a range. either r0-r7, r8-r15 or ss0-ss4. */
+ if (strncmp (p, "r0", 2) == 0 || strncmp (p, "r8", 2) == 0)
+ {
+ int rn = (p[1] == '0' ? 0 : 8);
+ int i = 0;
+
+ /* Get the next 8 values and record them. */
+ while (i < 8)
+ {
+ p = strtok (NULL, " \t\r\n");
+ if (p)
+ monitor_supply_register (regcache, rn + i, p);
+ i++;
+ }
+ }
+ else if (strncmp (p, "ss", 2) == 0)
+ {
+ /* get the next five values, ignoring the first. */
+ int rn;
+ p = strtok (NULL, " \t\r\n");
+ for (rn = 39; rn < 43; rn++)
+ {
+ p = strtok (NULL, " \t\r\n");
+ if (p)
+ monitor_supply_register (regcache, rn, p);
+ }
+ }
+ else
+ {
+ break;
+ }
+ }
+ else
+ {
+ /* Simple register type, paired. */
+ char *name = p;
+ int i;
+
+ /* Get and record value. */
+ p = strtok (NULL, " \t\r\n");
+ if (p)
+ {
+ for (i = 0; i < gdbarch_num_regs (gdbarch); i++)
+ {
+ if (picobug_regnames[i]
+ && (strcmp (picobug_regnames[i], name) == 0))
+ break;
+ }
+
+ if (i <= gdbarch_num_regs (gdbarch))
+ monitor_supply_register (regcache, i, p);
+ }
+ }
+ p = strtok (NULL, " \t\r\n");
+ }
+
+ return 0;
+}
+
+static void
+init_picobug_cmds (void)
+{
+ picobug_cmds.flags = MO_GETMEM_NEEDS_RANGE | MO_CLR_BREAK_USES_ADDR
+ | MO_PRINT_PROGRAM_OUTPUT;
+
+ picobug_cmds.init = picobug_inits; /* Init strings */
+ picobug_cmds.cont = "g\n"; /* continue command */
+ picobug_cmds.step = "s\n"; /* single step */
+ picobug_cmds.set_break = "br %x\n"; /* set a breakpoint */
+ picobug_cmds.clr_break = "nobr %x\n"; /* clear a breakpoint */
+ picobug_cmds.clr_all_break = "nobr\n"; /* clear all breakpoints */
+ picobug_cmds.setmem.cmdb = "mm %x %x ;b\n"; /* setmem.cmdb (addr, value) */
+ picobug_cmds.setmem.cmdw = "mm %x %x ;h\n"; /* setmem.cmdw (addr, value) */
+ picobug_cmds.setmem.cmdl = "mm %x %x ;w\n"; /* setmem.cmdl (addr, value) */
+ picobug_cmds.getmem.cmdb = "md %x %x\n"; /* getmem.cmdb (start addr,
+ end addr) */
+ picobug_cmds.getmem.resp_delim = ":"; /* getmem.resp_delim */
+ picobug_cmds.setreg.cmd = "rm %s %x\n"; /* setreg.cmd (name, value) */
+ picobug_cmds.getreg.cmd = "rd %s\n"; /* getreg.cmd (name) */
+ picobug_cmds.getreg.resp_delim = ":"; /* getreg.resp_delim */
+ picobug_cmds.dump_registers = "rd\n"; /* dump_registers */
+ picobug_cmds.dumpregs = picobug_dumpregs; /* dump registers parser */
+ picobug_cmds.load = "lo\n"; /* download command */
+ picobug_cmds.prompt = "picobug> "; /* monitor command prompt */
+ picobug_cmds.line_term = "\n"; /* end-of-line terminator */
+ picobug_cmds.target = &picobug_ops; /* target operations */
+ picobug_cmds.stopbits = SERIAL_1_STOPBITS; /* number of stop bits */
+ picobug_cmds.regnames = picobug_regnames; /* registers names */
+ picobug_cmds.num_breakpoints = 20; /* number of breakpoints */
+ picobug_cmds.magic = MONITOR_OPS_MAGIC; /* magic */
+}
+
+void
+_initialize_picobug_rom ()
+{
+ int i;
+
+ /* Initialize m32r RevC monitor target. */
+ init_picobug_cmds ();
+ init_monitor_ops (&picobug_ops);
+ picobug_ops.to_shortname = "picobug";
+ picobug_ops.to_longname = "picobug monitor";
+ picobug_ops.to_doc = "Debug via the picobug monitor.\n\
+Specify the serial device it is connected to (e.g. /dev/ttya).";
+ picobug_ops.to_open = picobug_open;
+
+ add_target (&picobug_ops);
+}
diff -urNp --exclude '*.swp' --exclude DEV-PHASE --exclude .svn --exclude CVS gdb/gdb/microblaze-tdep.c mb-gdb/gdb/microblaze-tdep.c
--- gdb/gdb/microblaze-tdep.c 1969-12-31 16:00:00.000000000 -0800
+++ mb-gdb/gdb/microblaze-tdep.c 2009-09-23 10:27:08.000000000 -0700
@@ -0,0 +1,895 @@
+/* Target-dependent code for Xilinx MicroBlaze.
+ Copyright 2009 Free Software Foundation, Inc.
+
+ 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 2 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, write to the Free Software
+ Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "defs.h"
+#include "arch-utils.h"
+#include "dis-asm.h"
+#include "frame.h"
+#include "trad-frame.h"
+#include "symtab.h"
+#include "value.h"
+#include "gdbcmd.h"
+#include "breakpoint.h"
+#include "inferior.h"
+#include "regcache.h"
+#include "target.h"
+#include "frame.h"
+#include "frame-base.h"
+#include "frame-unwind.h"
+#include "dwarf2-frame.h"
+#include "osabi.h"
+
+#include "gdb_assert.h"
+#include "gdb_string.h"
+#include "target-descriptions.h"
+
+
+/* FIXME - this shouldn't be here */
+#include "../opcodes/microblaze-opcm.h"
+
+#include "microblaze-tdep.h"
+
+extern enum microblaze_instr get_insn_microblaze (long, bfd_boolean *,
+ enum microblaze_instr_type *, short *);
+extern unsigned long microblaze_get_target_address (long, bfd_boolean, int, long, long,
+ long, bfd_boolean *, bfd_boolean *);
+
+\f
+/* The registers of the Xilinx microblaze processor */
+
+static const char *microblaze_register_names[] =
+{
+ "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",
+ "rpc", "rmsr", "rear", "resr", "rfsr", "rbtr",
+ "rpvr0", "rpvr1", "rpvr2", "rpvr3", "rpvr4", "rpvr5", "rpvr6",
+ "rpvr7", "rpvr8", "rpvr9", "rpvr10", "rpvr11",
+ "redr", "rpid", "rzpr", "rtlbx", "rtlbsx", "rtlblo", "rtlbhi"
+};
+
+#define MICROBLAZE_NUM_REGS ARRAY_SIZE (microblaze_register_names)
+
+#define mb_error(msg) internal_error (__FILE__, __LINE__, _(msg))
+
+extern enum microblaze_instr microblaze_decode_insn (long, int *, int *,
+ int *, int *);
+\f
+static int microblaze_debug = 0;
+
+static void
+microblaze_dump_insn (char *commnt, CORE_ADDR pc, int insn)
+{
+ if (microblaze_debug)
+ {
+ printf_filtered ("MICROBLAZE: %s %08x %08x ",
+ commnt, (unsigned int) pc, (unsigned int) insn);
+ /* print_insn_microblaze (pc, &tm_print_insn_info); */
+ printf_filtered ("\n");
+ }
+}
+#define microblaze_insn_debug(args) \
+ { if (microblaze_debug) printf_filtered args; }
+#define mb_warn(msg) \
+ if (microblaze_debug) printf ("mb_warning: %s:%d %s\n", __FILE__, __LINE__, _(msg))
+
+\f
+/* Return the name of register REGNUM. */
+
+static const char *
+microblaze_register_name (struct gdbarch *gdbarch, int regnum)
+{
+ if ((regnum >= 0) && (regnum < MICROBLAZE_NUM_REGS))
+ return microblaze_register_names[regnum];
+ return NULL;
+}
+
+static struct type *
+microblaze_register_type (struct gdbarch *gdbarch, int regnum)
+{
+ if (regnum == MICROBLAZE_SP_REGNUM)
+ return builtin_type (gdbarch)->builtin_data_ptr;
+
+ if (regnum == MICROBLAZE_PC_REGNUM)
+ return builtin_type (gdbarch)->builtin_func_ptr;
+
+ return builtin_type (gdbarch)->builtin_int;
+}
+
+\f
+/* Fetch the instruction at PC. */
+
+unsigned long
+microblaze_fetch_instruction (CORE_ADDR pc)
+{
+ gdb_byte buf[4];
+ unsigned long insn;
+ int i;
+
+ /* If we can't read the instruction at PC, return zero. */
+ if (target_read_memory (pc, buf, sizeof (buf)))
+ return 0;
+
+ insn = 0;
+ for (i = 0; i < sizeof (buf); i++)
+ insn = (insn << 8) | buf[i];
+ return insn;
+}
+\f
+
+static CORE_ADDR
+microblaze_push_dummy_code (struct gdbarch *gdbarch, CORE_ADDR sp,
+ CORE_ADDR funcaddr,
+ struct value **args, int nargs,
+ struct type *value_type,
+ CORE_ADDR *real_pc, CORE_ADDR *bp_addr,
+ struct regcache *regcache)
+{
+ mb_warn ("push_dummy_code not implemented");
+ return sp;
+}
+
+
+static CORE_ADDR
+microblaze_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)
+{
+ mb_warn ("store_arguments not implemented");
+ return sp;
+}
+
+/* Return a pointer to a string of bytes that encode a breakpoint
+ instruction, store the length of the string in *LEN. */
+
+static const gdb_byte *
+microblaze_breakpoint_from_pc (struct gdbarch *gdbarch, CORE_ADDR *pc, int *len)
+{
+ static gdb_byte break_insn[] = BREAKPOINT;
+
+ *len = sizeof (break_insn);
+ return break_insn;
+}
+\f
+/* Allocate and initialize a frame cache. */
+
+static struct microblaze_frame_cache *
+microblaze_alloc_frame_cache (void)
+{
+ struct microblaze_frame_cache *cache;
+ int i;
+
+ cache = FRAME_OBSTACK_ZALLOC (struct microblaze_frame_cache);
+
+ /* Base address. */
+ cache->base = 0;
+ cache->pc = 0;
+
+ /* Frameless until proven otherwise. */
+ cache->frameless_p = 1;
+
+ return cache;
+}
+
+/* Function prologues on the Xilinx microblaze processors consist of:
+
+ - adjustments to the stack pointer (r1) (addi r1, r1, imm)
+ - making a copy of r1 into another register (a "frame" pointer)
+ (add r?, r1, r0)
+ - store word/multiples that use r1 or the frame pointer as the
+ base address (swi r?, r1, imm OR swi r?, fp, imm)
+
+ Note that microblaze really doesn't have a real frame pointer.
+ Instead, the compiler may copy the SP into a register (usually
+ r19) to act as an arg pointer. For our target-dependent purposes,
+ the frame info's "frame" member will be the beginning of the
+ frame. The SP could, in fact, point below this.
+
+ The prologue ends when an instruction fails to meet either of
+ these criteria. */
+
+/* Analyze the prologue to determine where registers are saved,
+ the end of the prologue, etc. Return the address of the first line
+ of "real" code (i.e., the end of the prologue). */
+
+CORE_ADDR
+microblaze_analyze_prologue (CORE_ADDR pc, CORE_ADDR current_pc,
+ struct microblaze_frame_cache *cache)
+{
+ char *name;
+ CORE_ADDR func_addr, func_end, addr, stop, prologue_end_addr = 0;
+ unsigned long insn;
+ int rn, rd, ra, rb, imm;
+ enum microblaze_instr op;
+ int flags = 0;
+ bfd_boolean save_hidden_pointer_found = FALSE;
+ bfd_boolean non_stack_instruction_found = FALSE;
+
+ /* printf("microblaze_analyze_prologue (pc = 0x%8.8x, "
+ "current_pc = 0x%8.8x, cache = 0x%8.8x)\n",
+ (int) pc, (int) current_pc, (int) cache); */
+
+ /* Find the start of this function. */
+ find_pc_partial_function (pc, &name, &func_addr, &func_end);
+ if (func_addr < pc)
+ pc = func_addr;
+
+ if (current_pc < pc)
+ return current_pc;
+
+ /* Initialize info about frame */
+ cache->framesize = 0;
+ cache->fp_regnum = MICROBLAZE_SP_REGNUM;
+ cache->frameless_p = 1;
+
+ /* Start decoding the prologue. We start by checking two special cases:
+
+ 1. We're about to return
+ 2. We're at the first insn of the prologue.
+
+ If we're about to return, our frame has already been deallocated.
+ If we are stopped at the first instruction of a prologue,
+ then our frame has not yet been set up. */
+
+ /* Get the first insn from memory */
+
+ insn = microblaze_fetch_instruction (pc);
+ op = microblaze_decode_insn (insn, &rd, &ra, &rb, &imm);
+
+ if (IS_RETURN(op))
+ return pc;
+
+ /* Start at beginning of function and analyze until we get to the
+ current pc, or the end of the function, whichever is first. */
+ stop = (current_pc < func_end ? current_pc : func_end);
+
+ microblaze_insn_debug(("MICROBLAZE: Scanning prologue: name=%s, "
+ "func_addr=0x%x, stop=0x%x\n",
+ name, (unsigned int) func_addr,
+ (unsigned int) stop));
+
+ for (addr = func_addr; addr < stop; addr += INST_WORD_SIZE)
+ {
+ insn = microblaze_fetch_instruction (addr);
+ op = microblaze_decode_insn (insn, &rd, &ra, &rb, &imm);
+ microblaze_dump_insn ("got 2: ", addr, insn);
+
+ /* This code is very sensitive to what functions are present in the
+ prologue. It assumes that the (addi, addik, swi, sw) can be the
+ only instructions in the prologue. */
+ if (IS_UPDATE_SP(op, rd, ra))
+ {
+ microblaze_insn_debug (
+ ("MICROBLAZE: got addi r1,r1,%d; contnuing\n", imm));
+ if (cache->framesize)
+ break; /* break if framesize already computed */
+ cache->framesize = -1 * imm; /* stack grows towards low memory */
+ cache->frameless_p = 0; /* Frame found. */
+ save_hidden_pointer_found = FALSE;
+ non_stack_instruction_found = FALSE;
+ continue;
+ }
+ else if (IS_SPILL_SP(op, rd, ra))
+ {
+ /* Spill stack pointer */
+ cache->register_offsets[rd] = imm; /* SP spilled before updating */
+
+ microblaze_insn_debug (
+ ("MICROBLAZE: swi r1 r1 %d, continuing\n", imm));
+ save_hidden_pointer_found = FALSE;
+ if (!cache->framesize)
+ non_stack_instruction_found = FALSE;
+ continue;
+ }
+ else if (IS_SPILL_REG(op, rd, ra))
+ {
+ /* Spill register */
+ cache->register_offsets[rd] = imm - cache->framesize;
+ /* reg spilled after updating */
+
+ microblaze_insn_debug (
+ ("MICROBLAZE: swi %d r1 %d, continuing\n", rd, imm));
+ save_hidden_pointer_found = FALSE;
+ if (!cache->framesize)
+ non_stack_instruction_found = FALSE;
+ continue;
+ }
+ else if (IS_ALSO_SPILL_REG(op, rd, ra, rb))
+ {
+ /* Spill register */
+ cache->register_offsets[rd] = 0 - cache->framesize;
+ /* reg spilled after updating */
+
+ microblaze_insn_debug (
+ ("MICROBLAZE: sw %d r0 r1, continuing\n", rd));
+ save_hidden_pointer_found = FALSE;
+ if (!cache->framesize)
+ non_stack_instruction_found = FALSE;
+ continue;
+ }
+ else if (IS_SETUP_FP(op, ra, rb))
+ {
+ /* We have a frame pointer. Note
+ the register which is acting as the frame pointer. */
+ flags |= MY_FRAME_IN_FP;
+ flags &= ~MY_FRAME_IN_SP;
+ cache->fp_regnum = rd;
+ microblaze_insn_debug (
+ ("MICROBLAZE: Found a frame pointer: r%d\n",
+ cache->fp_regnum));
+ save_hidden_pointer_found = FALSE;
+ if (!cache->framesize)
+ non_stack_instruction_found = FALSE;
+ continue;
+ }
+ else if (IS_SPILL_REG_FP(op, rd, ra, cache->fp_regnum))
+ {
+ cache->register_offsets[rd] = imm - cache->framesize;
+ /* reg spilled after updating */
+
+ microblaze_insn_debug (
+ ("MICROBLAZE: swi %d %d %d, continuing\n", rd, ra, imm));
+ save_hidden_pointer_found = FALSE;
+ if (!cache->framesize)
+ non_stack_instruction_found = FALSE;
+ continue;
+ }
+ else if (IS_SAVE_HIDDEN_PTR(op, rd, ra, rb))
+ {
+ /* If the first argument is a hidden pointer to the area where the
+ return structure is to be saved, then it is saved as part of the
+ prologue */
+
+ microblaze_insn_debug (
+ ("MICROBLAZE: add %d %d %d, continuing\n", rd, ra, rb));
+ save_hidden_pointer_found = TRUE;
+ if (!cache->framesize)
+ non_stack_instruction_found = FALSE;
+ continue;
+ }
+
+ /* As a result of the modification in the next step where we continue
+ to analyze the prologue till we reach a control flow instruction,
+ we need another variable to store when exactly a non-stack
+ instruction was encountered, which is the current definition
+ of a prologue. */
+ if (!non_stack_instruction_found)
+ prologue_end_addr = addr;
+ non_stack_instruction_found = TRUE;
+
+ /* When optimizations are enabled, it is not guaranteed that prologue
+ instructions are not mixed in with other instructions from the
+ program. Some programs show this behavior at -O2. This can be
+ avoided by adding -fno-schedule-insns2 switch as of now (edk 8.1)
+ In such cases, we scan the function until we see the first control
+ instruction. */
+
+ {
+ unsigned op = (unsigned)insn >> 26;
+
+ if (!(op == 0x26 || op == 0x27 || op == 0x2d || op == 0x2e
+ || op == 0x2f))
+ /* continue if not control flow (branch, return) */
+ continue;
+ else if (op == 0x2c)
+ continue; /* continue if imm */
+ }
+
+ /* This is not a prologue insn, so stop here. */
+ microblaze_insn_debug (
+ ("microblaze: insn is not a prologue insn -- ending scan\n"));
+ break;
+ }
+
+ microblaze_insn_debug (("microblaze: done analyzing prologue\n"));
+ microblaze_insn_debug (("microblaze: prologue end = 0x%x\n", (int) addr));
+
+ /* If the last instruction was an add rd, r5, r0 then don't count it as
+ part of the prologue */
+ if (save_hidden_pointer_found)
+ prologue_end_addr -= INST_WORD_SIZE;
+
+ return prologue_end_addr;
+}
+
+static CORE_ADDR
+microblaze_unwind_pc (struct gdbarch *gdbarch, struct frame_info *next_frame)
+{
+ gdb_byte buf[4];
+ CORE_ADDR pc;
+
+ frame_unwind_register (next_frame, MICROBLAZE_PC_REGNUM, buf);
+ pc = extract_typed_address (buf, builtin_type (gdbarch)->builtin_func_ptr);
+ /* For sentinel frame, return address is actual PC. For other frames,
+ return address is pc+8. This is a workaround because gcc does not
+ generate correct return address in CIE. */
+ if (frame_relative_level (next_frame) >= 0)
+ pc += 8;
+ return pc;
+}
+
+/* Return PC of first real instruction of the function starting at
+ START_PC. */
+
+CORE_ADDR
+microblaze_skip_prologue (struct gdbarch *gdbarch, CORE_ADDR start_pc)
+{
+ struct symtab_and_line sal;
+ CORE_ADDR func_start, func_end, ostart_pc;
+ struct microblaze_frame_cache cache;
+
+ /* This is the preferred method, find the end of the prologue by
+ using the debugging information. Debugging info does not always
+ give the right answer since parameters are stored on stack after this.
+ Always analyze the prologue. */
+ if (find_pc_partial_function (start_pc, NULL, &func_start, &func_end))
+ {
+ sal = find_pc_line (func_start, 0);
+
+ if (sal.end < func_end
+ && start_pc <= sal.end)
+ start_pc = sal.end;
+ }
+
+ ostart_pc = microblaze_analyze_prologue (func_start, 0xffffffffUL, &cache);
+
+ if (ostart_pc > start_pc)
+ return ostart_pc;
+ return start_pc;
+}
+
+/* Normal frames. */
+
+struct microblaze_frame_cache *
+microblaze_frame_cache (struct frame_info *next_frame, void **this_cache)
+{
+ struct microblaze_frame_cache *cache;
+ struct gdbarch *gdbarch = get_frame_arch (next_frame);
+ CORE_ADDR func, pc, fp;
+ int rn;
+
+ if (*this_cache)
+ return *this_cache;
+
+ cache = microblaze_alloc_frame_cache ();
+ *this_cache = cache;
+ cache->saved_regs = trad_frame_alloc_saved_regs (next_frame);
+
+ /* Clear offsets to saved regs in frame. */
+ for (rn = 0; rn < gdbarch_num_regs (gdbarch); rn++)
+ cache->register_offsets[rn] = -1;
+
+ func = get_frame_func (next_frame);
+
+ cache->pc = get_frame_address_in_block (next_frame);
+
+ return cache;
+}
+
+static void
+microblaze_frame_this_id (struct frame_info *next_frame, void **this_cache,
+ struct frame_id *this_id)
+{
+ struct microblaze_frame_cache *cache =
+ microblaze_frame_cache (next_frame, this_cache);
+
+ /* This marks the outermost frame. */
+ if (cache->base == 0)
+ return;
+
+ (*this_id) = frame_id_build (cache->base, cache->pc);
+}
+
+static struct value *
+microblaze_frame_prev_register (struct frame_info *this_frame,
+ void **this_cache, int regnum)
+{
+ struct microblaze_frame_cache *cache =
+ microblaze_frame_cache (this_frame, this_cache);
+
+ /* FIXME: Rework this code and add comments. */
+ if (cache->frameless_p)
+ {
+ if (regnum == MICROBLAZE_PC_REGNUM)
+ regnum = 15;
+ if (regnum == MICROBLAZE_SP_REGNUM)
+ regnum = 1;
+ return trad_frame_get_prev_register (this_frame, cache->saved_regs, regnum);
+ }
+ else
+ return trad_frame_get_prev_register (this_frame, cache->saved_regs,
+ regnum);
+
+}
+
+static const struct frame_unwind microblaze_frame_unwind =
+{
+ NORMAL_FRAME,
+ microblaze_frame_this_id,
+ microblaze_frame_prev_register,
+ NULL,
+ default_frame_sniffer
+};
+\f
+static CORE_ADDR
+microblaze_frame_base_address (struct frame_info *next_frame, void **this_cache)
+{
+ struct microblaze_frame_cache *cache =
+ microblaze_frame_cache (next_frame, this_cache);
+
+ return cache->base;
+}
+
+static const struct frame_base microblaze_frame_base =
+{
+ µblaze_frame_unwind,
+ microblaze_frame_base_address,
+ microblaze_frame_base_address,
+ microblaze_frame_base_address
+};
+\f
+/* Extract from an array REGBUF containing the (raw) register state, a
+ function return value of TYPE, and copy that into VALBUF. */
+static void
+microblaze_extract_return_value (struct type *type, struct regcache *regcache,
+ gdb_byte *valbuf)
+{
+ gdb_byte buf[8];
+
+ /* Copy the return value (starting) in RETVAL_REGNUM to VALBUF. */
+ switch (TYPE_LENGTH(type))
+ {
+ case 1: /* return last byte in the register */
+ regcache_cooked_read (regcache, MICROBLAZE_RETVAL_REGNUM, buf);
+ memcpy(valbuf, buf + REGISTER_SIZE - 1, 1);
+ return;
+ case 2: /* return last 2 bytes in register */
+ memcpy(valbuf, buf + REGISTER_SIZE - 2, 2);
+ return;
+ case 4: /* for sizes 4 or 8, copy the required length */
+ case 8:
+ regcache_cooked_read (regcache, MICROBLAZE_RETVAL_REGNUM, buf);
+ regcache_cooked_read (regcache, MICROBLAZE_RETVAL_REGNUM+1, buf+4);
+ memcpy (valbuf, buf, TYPE_LENGTH (type));
+ return;
+ default:
+ printf_filtered("Fatal error: unsupported return value size "
+ "requested (%s @ %d)\n", __FILE__, __LINE__);
+ }
+}
+
+/* Store the return value in VALBUF (of type TYPE) where the caller
+ expects to see it.
+
+ Integers are stored in r3.
+
+ Values less than 32 bits (short, boolean) are stored in r2, right
+ justified and sign or zero extended. FIXME
+
+ Longs are stored in r3 (most significant word) and r4 (least
+ significant word).
+
+ Small structures are always returned on stack.
+*/
+
+static void
+microblaze_store_return_value (struct type *type, struct regcache *regcache,
+ const gdb_byte *valbuf)
+{
+ int len = TYPE_LENGTH (type);
+ gdb_byte buf[8];
+
+ memset (buf, 0, sizeof(buf));
+
+ /* Integral and pointer return values. */
+
+ if (len > 4)
+ {
+ gdb_assert (len == 8);
+ memcpy (buf, valbuf, 8);
+ regcache_cooked_write (regcache, MICROBLAZE_RETVAL_REGNUM+1, buf + 4);
+ }
+ else
+ {
+ /* ??? Do we need to do any sign-extension here? */
+ memcpy (buf + 4 - len, valbuf, len);
+ }
+ regcache_cooked_write (regcache, MICROBLAZE_RETVAL_REGNUM, buf);
+}
+
+static enum return_value_convention
+microblaze_return_value (struct gdbarch *gdbarch, struct type *func_type,
+ struct type *type, struct regcache *regcache,
+ gdb_byte *readbuf, const gdb_byte *writebuf)
+{
+ if (readbuf)
+ microblaze_extract_return_value (type, regcache, readbuf);
+ if (writebuf)
+ microblaze_store_return_value (type, regcache, writebuf);
+
+ return RETURN_VALUE_REGISTER_CONVENTION;
+}
+
+static int
+microblaze_stabs_argument_has_addr (struct gdbarch *gdbarch, struct type *type)
+{
+ return (TYPE_LENGTH (type) == 16); /* FIXME */
+}
+\f
+int
+microblaze_can_use_hardware_watchpoints (enum bptype type, int len, int ot)
+{
+ int count = 0;
+
+ switch (type)
+ {
+ case bp_hardware_breakpoint:
+ count = 2;
+ break;
+ case bp_read_watchpoint:
+ count = 1;
+ break;
+ case bp_hardware_watchpoint:
+ count = 1;
+ break;
+ default:
+ count = 0;
+ }
+ return count;
+}
+
+/* Simulate single-step calls, since microblaze does not support
+ single-stepping */
+
+/* single_step() is called just before we want to resume the inferior,
+ if we want to single-step it but there is no hardware or kernel
+ single-step support. We find all the possible targets of the
+ coming instruction and breakpoint them.
+
+ single_step is also called just after the inferior stops. If we had
+ set up a simulated single-step, we undo our damage. */
+
+int
+microblaze_software_single_step (struct frame_info *frame)
+{
+ struct gdbarch *gdbarch = get_frame_arch (frame);
+ struct gdbarch_tdep *tdep = gdbarch_tdep (gdbarch);
+ struct regcache *regcache = get_current_regcache ();
+
+ /* Save the address and the values of the next_pc and the target */
+ static struct sstep_breaks
+ {
+ CORE_ADDR address;
+ bfd_boolean valid;
+ } stepbreaks[2];
+
+ CORE_ADDR pc;
+ long insn;
+ enum microblaze_instr minstr;
+ bfd_boolean isunsignednum;
+ enum microblaze_instr_type insn_type;
+ short delay_slots;
+ int imm;
+ bfd_boolean immfound = FALSE;
+
+ /* Set a breakpoint at the next instruction */
+ /* If the current instruction is an imm, set it at the inst after */
+ /* If the instruction has a delay slot, skip the delay slot */
+ pc = get_frame_register_unsigned (frame, MICROBLAZE_PC_REGNUM);
+ insn = microblaze_fetch_instruction (pc);
+ minstr = get_insn_microblaze (insn, &isunsignednum, &insn_type,
+ &delay_slots);
+ if (insn_type == immediate_inst)
+ {
+ int rd, ra, rb;
+ immfound = TRUE;
+ minstr = microblaze_decode_insn (insn, &rd, &ra, &rb, &imm);
+ pc = pc + INST_WORD_SIZE;
+ insn = microblaze_fetch_instruction (pc);
+ minstr = get_insn_microblaze (insn, &isunsignednum, &insn_type,
+ &delay_slots);
+ }
+ stepbreaks[0].address = pc + (delay_slots * INST_WORD_SIZE) + INST_WORD_SIZE;
+ if (insn_type != return_inst)
+ {
+ stepbreaks[0].valid = TRUE;
+ }
+ else
+ {
+ stepbreaks[0].valid = FALSE;
+ }
+
+ /* Now check for branch or return instructions */
+ if (insn_type == branch_inst || insn_type == return_inst)
+ {
+ int limm;
+ int lrd, lra, lrb;
+ ULONGEST ra, rb;
+ bfd_boolean targetvalid;
+ bfd_boolean unconditionalbranch;
+ microblaze_decode_insn(insn, &lrd, &lra, &lrb, &limm);
+ if (lra >= 0 && lra < gdbarch_num_regs (gdbarch))
+ regcache_cooked_read_unsigned (regcache, lra, &ra);
+ else
+ ra = 0;
+ if (lrb >= 0 && lrb < gdbarch_num_regs (gdbarch))
+ regcache_cooked_read_unsigned (regcache, lrb, &rb);
+ else
+ rb = 0;
+ stepbreaks[1].address =
+ microblaze_get_target_address (insn, immfound, imm, pc,
+ ra, rb, &targetvalid,
+ &unconditionalbranch);
+ if (unconditionalbranch)
+ /* This is a unconditional branch: will not come to the next address */
+ stepbreaks[0].valid = FALSE;
+ if (targetvalid
+ && (stepbreaks[0].valid == FALSE
+ || (stepbreaks[0].address != stepbreaks[1].address))
+ && (stepbreaks[1].address != pc))
+ {
+ stepbreaks[1].valid = TRUE;
+ }
+ else
+ {
+ stepbreaks[1].valid = FALSE;
+ }
+ }
+ else
+ {
+ stepbreaks[1].valid = FALSE;
+ }
+
+ /* Insert the breakpoints */
+ if (stepbreaks[0].valid)
+ insert_single_step_breakpoint (gdbarch, stepbreaks[0].address);
+ if (stepbreaks[1].valid)
+ insert_single_step_breakpoint (gdbarch, stepbreaks[1].address);
+
+ return 1;
+}
+
+static void
+microblaze_write_pc (struct regcache *regcache, CORE_ADDR pc)
+{
+ regcache_cooked_write_unsigned (regcache, MICROBLAZE_PC_REGNUM, pc);
+}
+\f
+static int dwarf2_to_reg_map[78] =
+{ 0 /* r0 */, 1 /* r1 */, 2 /* r2 */, 3 /* r3 */, /* 0- 3 */
+ 4 /* r4 */, 5 /* r5 */, 6 /* r6 */, 7 /* r7 */, /* 4- 7 */
+ 8 /* r8 */, 9 /* r9 */, 10 /* r10 */, 11 /* r11 */, /* 8-11 */
+ 12 /* r12 */, 13 /* r13 */, 14 /* r14 */, 15 /* r15 */, /* 12-15 */
+ 16 /* r16 */, 17 /* r17 */, 18 /* r18 */, 19 /* r19 */, /* 16-19 */
+ 20 /* r20 */, 21 /* r21 */, 22 /* r22 */, 23 /* r23 */, /* 20-23 */
+ 24 /* r24 */, 25 /* r25 */, 26 /* r26 */, 27 /* r27 */, /* 24-25 */
+ 28 /* r28 */, 29 /* r29 */, 30 /* r30 */, 31 /* r31 */, /* 28-31 */
+ -1 /* $f0 */, -1 /* $f1 */, -1 /* $f2 */, -1 /* $f3 */, /* 32-35 */
+ -1 /* $f4 */, -1 /* $f5 */, -1 /* $f6 */, -1 /* $f7 */, /* 36-39 */
+ -1 /* $f8 */, -1 /* $f9 */, -1 /* $f10 */, -1 /* $f11 */, /* 40-43 */
+ -1 /* $f12 */, -1 /* $f13 */, -1 /* $f14 */, -1 /* $f15 */, /* 44-47 */
+ -1 /* $f16 */, -1 /* $f17 */, -1 /* $f18 */, -1 /* $f19 */, /* 48-51 */
+ -1 /* $f20 */, -1 /* $f21 */, -1 /* $f22 */, -1 /* $f23 */, /* 52-55 */
+ -1 /* $f24 */, -1 /* $f25 */, -1 /* $f26 */, -1 /* $f27 */, /* 56-59 */
+ -1 /* $f28 */, -1 /* $f29 */, -1 /* $f30 */, -1 /* $f31 */, /* 60-63 */
+ -1 /* hi */, -1 /* lo */, -1 /* accum*/, 33 /* rmsr */, /* 64-67 */
+ -1 /* $fcc1*/, -1 /* $fcc2*/, -1 /* $fcc3*/, -1 /* $fcc4*/, /* 68-71 */
+ -1 /* $fcc5*/, -1 /* $fcc6*/, -1 /* $fcc7*/, -1 /* $ap */, /* 72-75 */
+ -1 /* $rap */, -1 /* $frp */ /* 76-77 */
+};
+
+static int
+microblaze_dwarf2_reg_to_regnum (struct gdbarch *gdbarch, int reg)
+{
+ return dwarf2_to_reg_map[reg];
+}
+
+static struct gdbarch *
+microblaze_gdbarch_init (struct gdbarch_info info, struct gdbarch_list *arches)
+{
+ struct gdbarch_tdep *tdep;
+ struct gdbarch *gdbarch;
+
+ /* If there is already a candidate, use it. */
+ arches = gdbarch_list_lookup_by_info (arches, &info);
+ if (arches != NULL)
+ return arches->gdbarch;
+
+ /* Allocate space for the new architecture. */
+ tdep = XMALLOC (struct gdbarch_tdep);
+ gdbarch = gdbarch_alloc (&info, tdep);
+
+ set_gdbarch_long_double_bit (gdbarch, 128);
+ /* set_gdbarch_long_double_format (gdbarch, &floatformat_microblaze_quad); */
+
+ set_gdbarch_num_regs (gdbarch, MICROBLAZE_NUM_REGS);
+ set_gdbarch_register_name (gdbarch, microblaze_register_name);
+ set_gdbarch_register_type (gdbarch, microblaze_register_type);
+
+ /* Register numbers of various important registers. */
+ set_gdbarch_sp_regnum (gdbarch, MICROBLAZE_SP_REGNUM); /* $sp */
+ set_gdbarch_pc_regnum (gdbarch, MICROBLAZE_PC_REGNUM); /* $pc */
+
+ /* Map Dwarf2 registers to GDB registers. */
+ set_gdbarch_dwarf2_reg_to_regnum (gdbarch, microblaze_dwarf2_reg_to_regnum);
+
+ /* Call dummy code. */
+ set_gdbarch_call_dummy_location (gdbarch, ON_STACK);
+ set_gdbarch_push_dummy_code (gdbarch, microblaze_push_dummy_code);
+ set_gdbarch_push_dummy_call (gdbarch, microblaze_push_dummy_call);
+
+ set_gdbarch_return_value (gdbarch, microblaze_return_value);
+ set_gdbarch_stabs_argument_has_addr
+ (gdbarch, microblaze_stabs_argument_has_addr);
+
+ set_gdbarch_skip_prologue (gdbarch, microblaze_skip_prologue);
+
+ /* Stack grows downward. */
+ set_gdbarch_inner_than (gdbarch, core_addr_lessthan);
+
+ set_gdbarch_breakpoint_from_pc (gdbarch, microblaze_breakpoint_from_pc);
+
+ set_gdbarch_frame_args_skip (gdbarch, 8);
+
+ set_gdbarch_print_insn (gdbarch, print_insn_microblaze);
+
+ set_gdbarch_software_single_step (gdbarch, microblaze_software_single_step);
+ set_gdbarch_write_pc (gdbarch, microblaze_write_pc);
+
+ set_gdbarch_unwind_pc (gdbarch, microblaze_unwind_pc);
+
+ frame_base_set_default (gdbarch, µblaze_frame_base);
+
+ /* Hook in ABI-specific overrides, if they have been registered. */
+ gdbarch_init_osabi (info, gdbarch);
+
+ /* Unwind the frame. */
+ dwarf2_append_unwinders (gdbarch);
+ frame_unwind_append_unwinder (gdbarch, µblaze_frame_unwind);
+ frame_base_append_sniffer (gdbarch, dwarf2_frame_base_sniffer);
+
+ return gdbarch;
+}
+
+int
+microblaze_frame_num_args_p (struct gdbarch *gdbarch)
+{
+ return 0;
+}
+
+/* Provide a prototype to silence -Wmissing-prototypes. */
+void _initialize_microblaze_tdep (void);
+
+void
+_initialize_microblaze_tdep (void)
+{
+ register_gdbarch_init (bfd_arch_microblaze, microblaze_gdbarch_init);
+
+ /* Debug this files internals. */
+ add_setshow_zinteger_cmd ("microblaze", class_maintenance,
+ µblaze_debug, _("\
+Set microblaze debugging."), _("\
+Show microblaze debugging."), _("\
+When non-zero, microblaze specific debugging is enabled."),
+ NULL,
+ NULL,
+ &setdebuglist, &showdebuglist);
+
+}
diff -urNp --exclude '*.swp' --exclude DEV-PHASE --exclude .svn --exclude CVS gdb/gdb/microblaze-tdep.h mb-gdb/gdb/microblaze-tdep.h
--- gdb/gdb/microblaze-tdep.h 1969-12-31 16:00:00.000000000 -0800
+++ mb-gdb/gdb/microblaze-tdep.h 2009-09-21 14:05:50.000000000 -0700
@@ -0,0 +1,197 @@
+/* Target-dependent code for Xilinx MicroBlaze.
+ Copyright 2009 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 2 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, write to the Free Software
+ Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#ifndef MICROBLAZE_TDEP_H
+#define MICROBLAZE_TDEP_H 1
+
+#undef MICROBLAZE_DEBUG
+#ifdef MICROBLAZE_DEBUG
+int microblaze_debug = 1;
+#endif
+
+/* Microblaze architecture-specific information. */
+
+struct gdbarch_tdep
+{
+};
+
+struct microblaze_frame_cache
+{
+ /* Base address. */
+ CORE_ADDR base;
+ CORE_ADDR pc;
+
+ /* Do we have a frame? */
+ int frameless_p;
+
+ /* Frame size. */
+ int framesize;
+
+ /* Frame register. */
+ int fp_regnum;
+
+ /* Offsets to saved registers. */
+ int register_offsets[57]; /* Must match MICROBLAZE_NUM_REGS. */
+
+ /* Table of saved registers. */
+ struct trad_frame_saved_reg *saved_regs;
+};
+
+/* Register numbers. */
+enum microblaze_regnum
+{
+ MICROBLAZE_R0_REGNUM,
+ MICROBLAZE_R1_REGNUM, MICROBLAZE_SP_REGNUM = MICROBLAZE_R1_REGNUM,
+ MICROBLAZE_R2_REGNUM,
+ MICROBLAZE_R3_REGNUM, MICROBLAZE_RETVAL_REGNUM = MICROBLAZE_R3_REGNUM,
+ MICROBLAZE_R4_REGNUM,
+ MICROBLAZE_R5_REGNUM,
+ MICROBLAZE_R6_REGNUM,
+ MICROBLAZE_R7_REGNUM,
+ MICROBLAZE_R8_REGNUM,
+ MICROBLAZE_R9_REGNUM,
+ MICROBLAZE_R10_REGNUM,
+ MICROBLAZE_R11_REGNUM,
+ MICROBLAZE_R12_REGNUM,
+ MICROBLAZE_R13_REGNUM,
+ MICROBLAZE_R14_REGNUM,
+ MICROBLAZE_R15_REGNUM,
+ MICROBLAZE_R16_REGNUM,
+ MICROBLAZE_R17_REGNUM,
+ MICROBLAZE_R18_REGNUM,
+ MICROBLAZE_R19_REGNUM,
+ MICROBLAZE_R20_REGNUM,
+ MICROBLAZE_R21_REGNUM,
+ MICROBLAZE_R22_REGNUM,
+ MICROBLAZE_R23_REGNUM,
+ MICROBLAZE_R24_REGNUM,
+ MICROBLAZE_R25_REGNUM,
+ MICROBLAZE_R26_REGNUM,
+ MICROBLAZE_R27_REGNUM,
+ MICROBLAZE_R28_REGNUM,
+ MICROBLAZE_R29_REGNUM,
+ MICROBLAZE_R30_REGNUM,
+ MICROBLAZE_R31_REGNUM,
+ MICROBLAZE_PC_REGNUM,
+ MICROBLAZE_MSR_REGNUM,
+ MICROBLAZE_EAR_REGNUM,
+ MICROBLAZE_ESR_REGNUM,
+ MICROBLAZE_FSR_REGNUM,
+ MICROBLAZE_BTR_REGNUM,
+ MICROBLAZE_PVR0_REGNUM,
+ MICROBLAZE_PVR1_REGNUM,
+ MICROBLAZE_PVR2_REGNUM,
+ MICROBLAZE_PVR3_REGNUM,
+ MICROBLAZE_PVR4_REGNUM,
+ MICROBLAZE_PVR5_REGNUM,
+ MICROBLAZE_PVR6_REGNUM,
+ MICROBLAZE_PVR7_REGNUM,
+ MICROBLAZE_PVR8_REGNUM,
+ MICROBLAZE_PVR9_REGNUM,
+ MICROBLAZE_PVR10_REGNUM,
+ MICROBLAZE_PVR11_REGNUM,
+ MICROBLAZE_REDR_REGNUM,
+ MICROBLAZE_RPID_REGNUM,
+ MICROBLAZE_RZPR_REGNUM,
+ MICROBLAZE_RTLBX_REGNUM,
+ MICROBLAZE_RTLBSX_REGNUM,
+ MICROBLAZE_RTLBLO_REGNUM,
+ MICROBLAZE_RTLBHI_REGNUM
+};
+
+#define RWSDP_REGNUM 13
+#define LINK_REGNUM 15
+#define PR_REGNUM 15
+#define FIRST_ARGREG 5
+#define LAST_ARGREG 10
+#define RETVAL_REGNUM 3
+
+/* All registers are 32 bits. */
+#define REGISTER_SIZE 4
+
+/* The base of the current frame is actually in the stack pointer.
+ This happens when there is no frame pointer (microblaze ABI does not
+ require a frame pointer) or when we're stopped in the prologue or
+ epilogue itself. In these cases, microblaze_analyze_prologue will need
+ to update fi->frame before returning or analyzing the register
+ save instructions. */
+#define MY_FRAME_IN_SP 0x1
+
+/* The base of the current frame is in a frame pointer register.
+ This register is noted in frame_extra_info->fp_regnum.
+
+ Note that the existance of an FP might also indicate that the
+ function has called alloca. */
+#define MY_FRAME_IN_FP 0x2
+
+/* This flag is set to indicate that this frame is the top-most
+ frame. This tells frame chain not to bother trying to unwind
+ beyond this frame. */
+#define NO_MORE_FRAMES 0x4
+
+
+/* Instruction macros used for analyzing the prologue. */
+/* This set of instruction macros need to be changed whenever the
+ prologue generated by the compiler could have more instructions or
+ different type of instructions.
+ This set also needs to be verified if it is complete. */
+#define IS_RETURN(op) (op == rtsd || op == rtid)
+#define IS_UPDATE_SP(op, rd, ra) \
+ ((op == addik || op == addi) && rd == REG_SP && ra == REG_SP)
+#define IS_SPILL_SP(op, rd, ra) \
+ ((op == swi || op == sw) && rd == REG_SP && ra == REG_SP)
+#define IS_SPILL_REG(op, rd, ra) \
+ ((op == swi || op == sw) && rd != REG_SP && ra == REG_SP)
+/* The base of the current frame is actually in the stack pointer.
+ This happens when there is no frame pointer (microblaze ABI does not
+ require a frame pointer) or when we're stopped in the prologue or
+ epilogue itself. In these cases, microblaze_analyze_prologue will need
+ to update fi->frame before returning or analyzing the register
+ save instructions. */
+#define MY_FRAME_IN_SP 0x1
+
+/* The base of the current frame is in a frame pointer register.
+ This register is noted in frame_extra_info->fp_regnum.
+
+ Note that the existance of an FP might also indicate that the
+ function has called alloca. */
+#define MY_FRAME_IN_FP 0x2
+
+/* This flag is set to indicate that this frame is the top-most
+ frame. This tells frame chain not to bother trying to unwind
+ beyond this frame. */
+#define NO_MORE_FRAMES 0x4
+#define IS_ALSO_SPILL_REG(op, rd, ra, rb) \
+ ((op == swi || op == sw) && rd != REG_SP && ra == 0 && rb == REG_SP)
+#define IS_SETUP_FP(op, ra, rb) \
+ ((op == add || op == addik || op == addk) && ra == REG_SP && rb == 0)
+#define IS_SPILL_REG_FP(op, rd, ra, fpregnum) \
+ ((op == swi || op == sw) && rd != REG_SP && ra == fpregnum && ra != 0)
+#define IS_SAVE_HIDDEN_PTR(op, rd, ra, rb) \
+ ((op == add || op == addik) && ra == FIRST_ARGREG && rb == 0)
+
+/* BREAKPOINT defines the breakpoint that should be used. */
+#ifdef __MICROBLAZE_UCLINUX__
+#define BREAKPOINT {0xb9, 0xcc, 0x00, 0x60}
+#else
+#define BREAKPOINT {0x98, 0x0c, 0x00, 0x00}
+#endif
+
+#endif /* microblaze-tdep.h */
diff -urNp --exclude '*.swp' --exclude DEV-PHASE --exclude .svn --exclude CVS gdb/gdb/NEWS mb-gdb/gdb/NEWS
--- gdb/gdb/NEWS 2009-09-06 10:14:42.000000000 -0700
+++ mb-gdb/gdb/NEWS 2009-09-23 10:41:55.000000000 -0700
@@ -442,6 +442,11 @@ Lattice Mico32 lm32-*
x86 DICOS i[34567]86-*-dicos*
x86_64 DICOS x86_64-*-dicos*
S+core 3 score-*-*
+Xilinx MicroBlaze microblaze-*-*
+
+* New Simulators
+
+Xilinx MicroBlaze microblaze
* The GDB remote stub, gdbserver, now supports x86 Windows CE
(mingw32ce) debugging.
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [PATCH] Add support for Xilinx MicroBlaze
2009-09-23 18:33 [PATCH] Add support for Xilinx MicroBlaze Michael Eager
@ 2009-09-25 9:15 ` Eli Zaretskii
2009-09-25 19:16 ` Michael Eager
2009-09-26 0:08 ` Joel Brobecker
1 sibling, 1 reply; 6+ messages in thread
From: Eli Zaretskii @ 2009-09-25 9:15 UTC (permalink / raw)
To: Michael Eager; +Cc: gdb-patches
> Date: Wed, 23 Sep 2009 11:32:46 -0700
> From: Michael Eager <eager@eagercon.com>
>
> Attached is a revised patch which adds support for the
> Xilinx MicroBlaze. I believe that I've addressed all
> of the previous comments.
Thanks. And sorry for the delay ion reviewing your patch.
> * microblaze-linux-tdep.c: New.
> * microblaze-rom.c: New.
> * microblaze-tdep.c: New.
> * microblaze-tdep.h: New.
Please add the necessary tweaks for these new files to
gdb/config/djgpp/fnchange.lst. They will clash after 8+3 truncation.
> +@code{xmd} uses port @code{1234}. (While it is possible to change
^^
Two spaces between sentences, please.
> +@item target remote XMD-HOST:1234
> +Use this command to connect to the target if it is connected to @code{xmd}
> +running on a different system named @code{XMD-HOST}.
Please use "@var{xmd-host}" instead of a literal upper-case XMD-HOST.
@var is what should be used in Texinfo for identifiers that stand for
something else, in this case an actual host name.
> + /* printf("microblaze_analyze_prologue (pc = 0x%8.8x, "
> + "current_pc = 0x%8.8x, cache = 0x%8.8x)\n",
> + (int) pc, (int) current_pc, (int) cache); */
I believe we don't like dead code in GDB sources.
> + /* Spill register */
Comments should have a period and 2 blanks at their end.
> + cache->register_offsets[rd] = imm - cache->framesize;
> + /* reg spilled after updating */
Strange formatting of a comment, and indentation seems wrong.
> + /* We have a frame pointer. Note
> + the register which is acting as the frame pointer. */
Please reformat whitespace here: the first line is broken too early.
> + switch (TYPE_LENGTH(type))
I believe GNU standards call for a space before the left paren.
> + default:
> + printf_filtered("Fatal error: unsupported return value size "
> + "requested (%s @ %d)\n", __FILE__, __LINE__);
Shouldn't you call `fatal' here?
In any case, messages presented to the user should be in _(), to be
translatable (the debug messages are exempt from this rule).
> + /* Debug this files internals. */
> + add_setshow_zinteger_cmd ("microblaze", class_maintenance,
> + µblaze_debug, _("\
> +Set microblaze debugging."), _("\
> +Show microblaze debugging."), _("\
> +When non-zero, microblaze specific debugging is enabled."),
> + NULL,
> + NULL,
> + &setdebuglist, &showdebuglist);
This command should be documented in the user manual.
> --- gdb/gdb/NEWS 2009-09-06 10:14:42.000000000 -0700
> +++ mb-gdb/gdb/NEWS 2009-09-23 10:41:55.000000000 -0700
> @@ -442,6 +442,11 @@ Lattice Mico32 lm32-*
> x86 DICOS i[34567]86-*-dicos*
> x86_64 DICOS x86_64-*-dicos*
> S+core 3 score-*-*
> +Xilinx MicroBlaze microblaze-*-*
> +
> +* New Simulators
> +
> +Xilinx MicroBlaze microblaze
The patch for NEWS is fine.
Thanks.
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [PATCH] Add support for Xilinx MicroBlaze
2009-09-25 9:15 ` Eli Zaretskii
@ 2009-09-25 19:16 ` Michael Eager
2009-09-25 21:47 ` Joel Brobecker
0 siblings, 1 reply; 6+ messages in thread
From: Michael Eager @ 2009-09-25 19:16 UTC (permalink / raw)
To: Eli Zaretskii; +Cc: gdb-patches
Eli Zaretskii wrote:
>> Date: Wed, 23 Sep 2009 11:32:46 -0700
>> From: Michael Eager <eager@eagercon.com>
>>
>> Attached is a revised patch which adds support for the
>> Xilinx MicroBlaze. I believe that I've addressed all
>> of the previous comments.
>
> Thanks. And sorry for the delay ion reviewing your patch.
Thanks for the review. I'll fix these issues and resubmit.
Sorry this is taking so much back-and-forth.
>
>> * microblaze-linux-tdep.c: New.
>> * microblaze-rom.c: New.
>> * microblaze-tdep.c: New.
>> * microblaze-tdep.h: New.
>
> Please add the necessary tweaks for these new files to
> gdb/config/djgpp/fnchange.lst. They will clash after 8+3 truncation.
>
>> +@code{xmd} uses port @code{1234}. (While it is possible to change
> ^^
> Two spaces between sentences, please.
>
>> +@item target remote XMD-HOST:1234
>> +Use this command to connect to the target if it is connected to @code{xmd}
>> +running on a different system named @code{XMD-HOST}.
>
> Please use "@var{xmd-host}" instead of a literal upper-case XMD-HOST.
> @var is what should be used in Texinfo for identifiers that stand for
> something else, in this case an actual host name.
>
>> + /* printf("microblaze_analyze_prologue (pc = 0x%8.8x, "
>> + "current_pc = 0x%8.8x, cache = 0x%8.8x)\n",
>> + (int) pc, (int) current_pc, (int) cache); */
>
> I believe we don't like dead code in GDB sources.
>
>> + /* Spill register */
>
> Comments should have a period and 2 blanks at their end.
>
>> + cache->register_offsets[rd] = imm - cache->framesize;
>> + /* reg spilled after updating */
>
> Strange formatting of a comment, and indentation seems wrong.
>
>> + /* We have a frame pointer. Note
>> + the register which is acting as the frame pointer. */
>
> Please reformat whitespace here: the first line is broken too early.
>
>> + switch (TYPE_LENGTH(type))
>
> I believe GNU standards call for a space before the left paren.
>
>> + default:
>> + printf_filtered("Fatal error: unsupported return value size "
>> + "requested (%s @ %d)\n", __FILE__, __LINE__);
>
> Shouldn't you call `fatal' here?
>
> In any case, messages presented to the user should be in _(), to be
> translatable (the debug messages are exempt from this rule).
>
>> + /* Debug this files internals. */
>> + add_setshow_zinteger_cmd ("microblaze", class_maintenance,
>> + µblaze_debug, _("\
>> +Set microblaze debugging."), _("\
>> +Show microblaze debugging."), _("\
>> +When non-zero, microblaze specific debugging is enabled."),
>> + NULL,
>> + NULL,
>> + &setdebuglist, &showdebuglist);
>
> This command should be documented in the user manual.
>
>> --- gdb/gdb/NEWS 2009-09-06 10:14:42.000000000 -0700
>> +++ mb-gdb/gdb/NEWS 2009-09-23 10:41:55.000000000 -0700
>> @@ -442,6 +442,11 @@ Lattice Mico32 lm32-*
>> x86 DICOS i[34567]86-*-dicos*
>> x86_64 DICOS x86_64-*-dicos*
>> S+core 3 score-*-*
>> +Xilinx MicroBlaze microblaze-*-*
>> +
>> +* New Simulators
>> +
>> +Xilinx MicroBlaze microblaze
>
> The patch for NEWS is fine.
>
> Thanks.
>
--
Michael Eager eager@eagercon.com
1960 Park Blvd., Palo Alto, CA 94306 650-325-8077
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [PATCH] Add support for Xilinx MicroBlaze
2009-09-25 19:16 ` Michael Eager
@ 2009-09-25 21:47 ` Joel Brobecker
0 siblings, 0 replies; 6+ messages in thread
From: Joel Brobecker @ 2009-09-25 21:47 UTC (permalink / raw)
To: Michael Eager; +Cc: Eli Zaretskii, gdb-patches
> Thanks for the review. I'll fix these issues and resubmit.
> Sorry this is taking so much back-and-forth.
No problem with the back-and-forth. But before you re-submit, give me
some time to look at your patch too. I remember having some minor
style comments as well.
--
Joel
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] Add support for Xilinx MicroBlaze
2009-09-23 18:33 [PATCH] Add support for Xilinx MicroBlaze Michael Eager
2009-09-25 9:15 ` Eli Zaretskii
@ 2009-09-26 0:08 ` Joel Brobecker
2009-09-26 1:30 ` Michael Eager
1 sibling, 1 reply; 6+ messages in thread
From: Joel Brobecker @ 2009-09-26 0:08 UTC (permalink / raw)
To: Michael Eager; +Cc: gdb-patches
> * configure.tgt: Add targets microblaze*-linux-*, microblaze*-xilinx-*.
> * doc/gdb.texinfo: Add MicroBlaze.
> * MAINTAINERS: Add self as maintainer for MicroBlaze.
> * Makefile.in: Build microblaze-tdep.o, microblaze-linux-tdep.o.
> * microblaze-linux-tdep.c: New.
> * microblaze-rom.c: New.
> * microblaze-tdep.c: New.
> * microblaze-tdep.h: New.
> * NEWS: Announce Xilinx MicroBlaze support.
In the makefile, you'll need to add microblaze-tdep.h to the
HFILES_NO_SRCDIR list. Not strictly needed in order to build
GDB, but this is used by the developers who use ctags/etags.
> + Copyright 2009 Free Software Foundation, Inc.
> +
> + 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 2 of the License, or
> + (at your option) any later version.
The copyright header is out of date. It should refer to version 3
of the GPL. Any C file in the GDB sources should have the correct
copyright header. Can you update all the new files that you are
submitting?
> + if ((val == 0) && (memcmp (bp, old_contents, bplen) == 0))
The extra parens should be removed. We discussed this with other
Global Maintainers, and we all felt that this difference in style
made the code harder to read, and thus should be avoided. So it
should be:
if (val == 0 && memcmp (bp, old_contents, bplen) == 0)
Can you fix the rest of the code accordingly?
> +static struct tramp_frame microblaze_linux_sighandler_tramp_frame = {
Just to be consistent, this curly brace should be on the next line.
> + gdbarch_register_osabi (bfd_arch_microblaze, 0,
> + GDB_OSABI_LINUX, microblaze_linux_init_abi);
Nit: formatting is wrong on the second line.
> +/* Picobug only supports a subset of registers from MCore. In reality,
> + it doesn't support ss1, either. */
Formatting nit: 2 spaces after the period.
> +void
> +picobug_open (char *args, int from_tty)
Can this be made static?
> +/* FIXME - this shouldn't be here */
> +#include "../opcodes/microblaze-opcm.h"
FIXME? If you remove the "../", I *think* it's fine to include
opcodes headers files from GDB code.
> +extern enum microblaze_instr get_insn_microblaze (long, bfd_boolean *,
> + enum microblaze_instr_type *, short *);
> +extern unsigned long microblaze_get_target_address (long, bfd_boolean, int, long, long,
> + long, bfd_boolean *, bfd_boolean *);
> +extern enum microblaze_instr microblaze_decode_insn (long, int *, int *,
> + int *, int *);
Can we avoid these extern declarations, though? It would be nice if
we could get a warning if the implementation of these functions changed.
This won't happen with these locally-defined externs.
> +#define mb_error(msg) internal_error (__FILE__, __LINE__, _(msg))
I don't like this define much, because an outsider inspecting part
of your code might not realize that mb_error is actually an internal
error. But if you prefer it that way...
> +static void
> +microblaze_dump_insn (char *commnt, CORE_ADDR pc, int insn)
> +{
> + if (microblaze_debug)
> + {
> + printf_filtered ("MICROBLAZE: %s %08x %08x ",
> + commnt, (unsigned int) pc, (unsigned int) insn);
> + /* print_insn_microblaze (pc, &tm_print_insn_info); */
> + printf_filtered ("\n");
> + }
> +}
Er, this function does nothing functional, whereas I was expecting
it to do something. But then, looking at the usage, it's just a debug
function. This reminds me that all new functions should be documented
with a comment at the beginning of their implementation. There are
exceptions, however; All the routines that implement gdbarch methods
or target methods are already documented as part of either gdbarch
or target. Same for solib vectors, etc. So you don't need to provide
the overal documentation for the functions used in your particular
target.
In this case, the general style is usually to conditionalize the call
to that function to microblaze_debug, rather than having the condition
inside the function. Some purists will say that this avoids a function
call if debug is off, but I'll just say that it makes the naming of that
function a littler easier :). If you prefer to keep the condition inside
the function, not a problem, but please consider choosing a name that
expresses this fact.
> +#define microblaze_insn_debug(args) \
> + { if (microblaze_debug) printf_filtered args; }
Perhaps if you added the braces around args in your call to
printf_filtered, you wouldn't need the extra parens when using
his macro?
> +#define mb_warn(msg) \
> + if (microblaze_debug) printf ("mb_warning: %s:%d %s\n", __FILE__, __LINE__, _(msg))
> +
Let's not use this home-made warning. See below.
> +unsigned long
> +microblaze_fetch_instruction (CORE_ADDR pc)
[...]
> + insn = 0;
> + for (i = 0; i < sizeof (buf); i++)
> + insn = (insn << 8) | buf[i];
> + return insn;
You can replace all this by a call to extract_unsigned_integer...
> + mb_warn ("push_dummy_code not implemented");
> + mb_warn ("store_arguments not implemented");
Please call "error" instead: The user will then be told that this
feature is not implemented, and the action that required this function
call will be automatically aborted as a result.
> +/* Return a pointer to a string of bytes that encode a breakpoint
> + instruction, store the length of the string in *LEN. */
As explained above, the comment is unnecessary; but if you decide
to keep it, there is just one nit: you need an extra space after
the period (I can reassure you that I make these mistakes myself,
so you're not the only one commented on these).
> +static const gdb_byte *
> +microblaze_breakpoint_from_pc (struct gdbarch *gdbarch, CORE_ADDR *pc, int *len)
Formatting: This line is a little too long...
> + The prologue ends when an instruction fails to meet either of
> + these criteria. */
> +/* Analyze the prologue to determine where registers are saved,
> + the end of the prologue, etc. Return the address of the first line
> + of "real" code (i.e., the end of the prologue). */
> + /* Find the start of this function. */
Missing spaces after periods. There are other instances of this,
but I stopped pasting them here. Can you make sure you take care
of them too?
> + bfd_boolean save_hidden_pointer_found = FALSE;
> + bfd_boolean non_stack_instruction_found = FALSE;
We tend to prefer using integers for predicates.
> + cache->framesize = -1 * imm; /* stack grows towards low memory */
Can't we just do -imm?
> + /* FIXME: Rework this code and add comments. */
Please consider doing this now :)
> + default:
> + printf_filtered("Fatal error: unsupported return value size "
> + "requested (%s @ %d)\n", __FILE__, __LINE__);
This should be replaced by a call to "error".
> + Values less than 32 bits (short, boolean) are stored in r2, right
> + justified and sign or zero extended. FIXME
It would be nice to have the FIXME addressed now.
> +static int
> +microblaze_stabs_argument_has_addr (struct gdbarch *gdbarch, struct type *type)
> +{
> + return (TYPE_LENGTH (type) == 16); /* FIXME */
Can you take care of this FIXME as well? If you don't know whether
you have anything to fix or not, just put a note saying that, as far
as you know, the only matching case is when TYPE_LENGTH is 16.
I'm surprised you even need this at all. Are you using stabs?
> +microblaze_can_use_hardware_watchpoints (enum bptype type, int len, int ot)
I assume you have unlimited number of hardware breakpoint/watchpoints?
How nice.
> + bfd_boolean isunsignednum;
Can you rename this variable to is_unsingned_num? It's so much more
readable that way...
> + bfd_boolean unconditionalbranch;
> + microblaze_decode_insn(insn, &lrd, &lra, &lrb, &limm);
Can you use an empty line between the last declaration and the first
line of code?
> + {
> + stepbreaks[1].valid = TRUE;
> + }
Another small nit. We prefer not to use the curly braces if the
if block contains only one statement.
> +static int
> +microblaze_dwarf2_reg_to_regnum (struct gdbarch *gdbarch, int reg)
> +{
> + return dwarf2_to_reg_map[reg];
Can you add a call to gdb_assert that verifies that REG is within
the bounds of dwarf2_to_reg_map?
> +#undef MICROBLAZE_DEBUG
> +#ifdef MICROBLAZE_DEBUG
> +int microblaze_debug = 1;
> +#endif
You probably meant to remove this part.
> +#define RWSDP_REGNUM 13
> +#define LINK_REGNUM 15
> +#define PR_REGNUM 15
> +#define FIRST_ARGREG 5
> +#define LAST_ARGREG 10
> +#define RETVAL_REGNUM 3
Just a thought: Why not use the register-name litterals that you just
defined above? Wouldn't that be clearer?
> +/* All registers are 32 bits. */
> +#define REGISTER_SIZE 4
This is slightly paranoin, but we used to have a REGISTER_SIZE macro
that corresponded to a call to gdbarch_register_size. Since you are
defining this macro in a header file, can you rename it to
MICROBLAZE_REGISTER_SIZE? I just want to avoid the confusion as well
as avoid a clash with other parts of the code.
> +#define MY_FRAME_IN_SP 0x1
> +#define MY_FRAME_IN_FP 0x2
Can you either move these macros to microblaze-tdep.c or prefix
them with MICROBLAZE_? Note that you have a second identical
definition of these macros just a few lines below. The compiler
usually emits a warning about that...
> +#define NO_MORE_FRAMES 0x4
This one does not appear to be used anywhere. It also appears to be
duplicated.
> +#define IS_RETURN(op) (op == rtsd || op == rtid)
> +#define IS_UPDATE_SP(op, rd, ra) \
> + ((op == addik || op == addi) && rd == REG_SP && ra == REG_SP)
> +#define IS_SPILL_SP(op, rd, ra) \
> + ((op == swi || op == sw) && rd == REG_SP && ra == REG_SP)
> +#define IS_SPILL_REG(op, rd, ra) \
> + ((op == swi || op == sw) && rd != REG_SP && ra == REG_SP)
> +#define IS_ALSO_SPILL_REG(op, rd, ra, rb) \
> + ((op == swi || op == sw) && rd != REG_SP && ra == 0 && rb == REG_SP)
> +#define IS_SETUP_FP(op, ra, rb) \
> + ((op == add || op == addik || op == addk) && ra == REG_SP && rb == 0)
> +#define IS_SPILL_REG_FP(op, rd, ra, fpregnum) \
> + ((op == swi || op == sw) && rd != REG_SP && ra == fpregnum && ra != 0)
> +#define IS_SAVE_HIDDEN_PTR(op, rd, ra, rb) \
> + ((op == add || op == addik) && ra == FIRST_ARGREG && rb == 0)
Same for these, I think they should be moved to a .c file or
prefixed by MICROBLAZE_
> +/* BREAKPOINT defines the breakpoint that should be used. */
> +#ifdef __MICROBLAZE_UCLINUX__
> +#define BREAKPOINT {0xb9, 0xcc, 0x00, 0x60}
> +#else
> +#define BREAKPOINT {0x98, 0x0c, 0x00, 0x00}
> +#endif
Same remark about the name of the macro. But a bigger concern is
the fact that the macro is statically defined based on what I am
guessing is a host-specific macro. This is a target-specific file.
This goes against what we call our "multi-arch" design.
Is there no way for you to determine that at run time? For instance,
is there any microblaze-linux other than uclinux? If the breakpoint
instruction on linux is always {0xb9, 0xcc, 0x00, 0x60} while it is
{0x98, 0x0c, 0x00, 0x00} on the rest of the targets, that's easy to fix.
Otherwise, we'll have to come up with another form of detection.
--
Joel
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [PATCH] Add support for Xilinx MicroBlaze
2009-09-26 0:08 ` Joel Brobecker
@ 2009-09-26 1:30 ` Michael Eager
0 siblings, 0 replies; 6+ messages in thread
From: Michael Eager @ 2009-09-26 1:30 UTC (permalink / raw)
To: Joel Brobecker; +Cc: gdb-patches
Joel Brobecker wrote:
>> +extern enum microblaze_instr get_insn_microblaze (long, bfd_boolean *,
>> + enum microblaze_instr_type *, short *);
>> +extern unsigned long microblaze_get_target_address (long, bfd_boolean, int, long, long,
>> + long, bfd_boolean *, bfd_boolean *);
>> +extern enum microblaze_instr microblaze_decode_insn (long, int *, int *,
>> + int *, int *);
>
> Can we avoid these extern declarations, though? It would be nice if
> we could get a warning if the implementation of these functions changed.
> This won't happen with these locally-defined externs.
Actually, I meant to move these to a header file and include it
here and where the functions are defined.
>> + /* FIXME: Rework this code and add comments. */
>
> Please consider doing this now :)
Maybe, if I can figure out why the note was added and
what changes are needed. :-\
>> + Values less than 32 bits (short, boolean) are stored in r2, right
>> + justified and sign or zero extended. FIXME
>
> It would be nice to have the FIXME addressed now.
Ditto.
>
>> +static int
>> +microblaze_stabs_argument_has_addr (struct gdbarch *gdbarch, struct type *type)
>> +{
>> + return (TYPE_LENGTH (type) == 16); /* FIXME */
>
> Can you take care of this FIXME as well? If you don't know whether
> you have anything to fix or not, just put a note saying that, as far
> as you know, the only matching case is when TYPE_LENGTH is 16.
> I'm surprised you even need this at all. Are you using stabs?
It depends on whether users are still using older compilers (gcc-2.95 era)
which generated stabs. I don't want to remove backward compatibility.
On the other hand, I don't want to spend any time and energy fixing
potential problems in code that has no future (or present).
>> +microblaze_can_use_hardware_watchpoints (enum bptype type, int len, int ot)
>
> I assume you have unlimited number of hardware breakpoint/watchpoints?
> How nice.
Infinite is always good. I'll find out how many hw bp's actually exist.
>> +/* BREAKPOINT defines the breakpoint that should be used. */
>> +#ifdef __MICROBLAZE_UCLINUX__
>> +#define BREAKPOINT {0xb9, 0xcc, 0x00, 0x60}
>> +#else
>> +#define BREAKPOINT {0x98, 0x0c, 0x00, 0x00}
>> +#endif
>
> Same remark about the name of the macro. But a bigger concern is
> the fact that the macro is statically defined based on what I am
> guessing is a host-specific macro. This is a target-specific file.
> This goes against what we call our "multi-arch" design.
>
> Is there no way for you to determine that at run time? For instance,
> is there any microblaze-linux other than uclinux? If the breakpoint
> instruction on linux is always {0xb9, 0xcc, 0x00, 0x60} while it is
> {0x98, 0x0c, 0x00, 0x00} on the rest of the targets, that's easy to fix.
> Otherwise, we'll have to come up with another form of detection.
Yes, there is MicroBlaze Linux with glibc, at least in development
if not currently available. I'll have to find out if it uses the
same breakpoint; it probably does. I'll see if I can come up with a
run-time test to distinguish embedded from hosted.
Thanks for your review and comments!
--
Michael Eager eager@eagercon.com
1960 Park Blvd., Palo Alto, CA 94306 650-325-8077
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2009-09-26 1:30 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2009-09-23 18:33 [PATCH] Add support for Xilinx MicroBlaze Michael Eager
2009-09-25 9:15 ` Eli Zaretskii
2009-09-25 19:16 ` Michael Eager
2009-09-25 21:47 ` Joel Brobecker
2009-09-26 0:08 ` Joel Brobecker
2009-09-26 1:30 ` Michael Eager
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox