This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH] gdb/riscv: Don't error when decoding a 6 or 8 byte instruction
- From: Andrew Burgess <andrew dot burgess at embecosm dot com>
- To: gdb-patches at sourceware dot org
- Cc: Jim Wilson <jimw at sifive dot com>, Palmer Dabbelt <palmer at sifive dot com>
- Date: Wed, 5 Jun 2019 10:08:16 +0100
- Subject: Re: [PATCH] gdb/riscv: Don't error when decoding a 6 or 8 byte instruction
- References: <20190604115054.25306-1-andrew.burgess@embecosm.com>
* Andrew Burgess <andrew.burgess@embecosm.com> [2019-06-04 12:50:54 +0100]:
> If the RISC-V prologue scanner finds a 6 or 8 byte instruction we
> currently throw an internal error, which is not great for the user.
>
> A mechanism already exists in the prologue scanner to leave
> instructions marked as unknown so that we can stop the prologue scan
> without raising an error, this is used for all 2 and 4 byte
> instructions that are not part of the small set the prologue scanner
> actually understands.
>
> This commit changes GDB so that all 6 and 8 byte instructions are
> marked as unknown, rather than causing an error.
>
> gdb/ChangeLog:
>
> * riscv-tdep.c (riscv_insn::decode): Gracefully ignore
> instructions of lengths 6 or 8 bytes.
>
> gdb/testsuite/ChangeLog:
>
> * gdb.arch/riscv-unwind-long-insn-6.s: New file.
> * gdb.arch/riscv-unwind-long-insn-8.s: New file.
> * gdb.arch/riscv-unwind-long-insn.c: New file.
> * gdb.arch/riscv-unwind-long-insn.exp: New file.
I have now pushed the patch version below. The only change from the
previous version is in riscv-unwind-long-insn.exp where I make the
test binary name unique over the two iterations of the test.
Thanks,
Andrew
--
commit 312617a3d06b8df67b9f4f63f92ebfaa6b591921
Author: Andrew Burgess <andrew.burgess@embecosm.com>
Date: Tue Jun 4 10:30:41 2019 +0100
gdb/riscv: Don't error when decoding a 6 or 8 byte instruction
If the RISC-V prologue scanner finds a 6 or 8 byte instruction we
currently throw an internal error, which is not great for the user.
A mechanism already exists in the prologue scanner to leave
instructions marked as unknown so that we can stop the prologue scan
without raising an error, this is used for all 2 and 4 byte
instructions that are not part of the small set the prologue scanner
actually understands.
This commit changes GDB so that all 6 and 8 byte instructions are
marked as unknown, rather than causing an error.
gdb/ChangeLog:
* riscv-tdep.c (riscv_insn::decode): Gracefully ignore
instructions of lengths 6 or 8 bytes.
gdb/testsuite/ChangeLog:
* gdb.arch/riscv-unwind-long-insn-6.s: New file.
* gdb.arch/riscv-unwind-long-insn-8.s: New file.
* gdb.arch/riscv-unwind-long-insn.c: New file.
* gdb.arch/riscv-unwind-long-insn.exp: New file.
diff --git a/gdb/riscv-tdep.c b/gdb/riscv-tdep.c
index 3fc86ab825..bae987cf66 100644
--- a/gdb/riscv-tdep.c
+++ b/gdb/riscv-tdep.c
@@ -1385,10 +1385,12 @@ riscv_insn::decode (struct gdbarch *gdbarch, CORE_ADDR pc)
m_opcode = OTHER;
}
else
- internal_error (__FILE__, __LINE__,
- _("unable to decode %d byte instructions in "
- "prologue at %s"), m_length,
- core_addr_to_string (pc));
+ {
+ /* This must be a 6 or 8 byte instruction, we don't currently decode
+ any of these, so just ignore it. */
+ gdb_assert (m_length == 6 || m_length == 8);
+ m_opcode = OTHER;
+ }
}
/* The prologue scanner. This is currently only used for skipping the
diff --git a/gdb/testsuite/gdb.arch/riscv-unwind-long-insn-6.s b/gdb/testsuite/gdb.arch/riscv-unwind-long-insn-6.s
new file mode 100644
index 0000000000..b21b1e10f3
--- /dev/null
+++ b/gdb/testsuite/gdb.arch/riscv-unwind-long-insn-6.s
@@ -0,0 +1,45 @@
+/* Copyright 2019 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 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/>. */
+
+ .option nopic
+ .text
+
+ .align 1
+ .globl bar
+ .type bar, @function
+bar:
+ tail 1f
+ .size bar, .-func
+
+ .align 1
+ .globl func
+ .type func, @function
+func:
+ /* A fake 6 byte instruction. This is never executed, but the
+ prologue scanner will try to decode it. These long
+ instructions are ISA extensions, I use .byte rather than an
+ actual instruction mnemonic so that the test can be compiled
+ with a toolchain that doesn't include any long instruction
+ extensions. */
+ .byte 0x1f, 0x00, 0x00, 0x00, 0x00, 0x00
+1:
+ addi sp,sp,-16
+ sw s0,12(sp)
+ addi s0,sp,16
+ nop
+ lw s0,12(sp)
+ addi sp,sp,16
+ jr ra
+ .size func, .-func
diff --git a/gdb/testsuite/gdb.arch/riscv-unwind-long-insn-8.s b/gdb/testsuite/gdb.arch/riscv-unwind-long-insn-8.s
new file mode 100644
index 0000000000..3fad07b59d
--- /dev/null
+++ b/gdb/testsuite/gdb.arch/riscv-unwind-long-insn-8.s
@@ -0,0 +1,45 @@
+/* Copyright 2019 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 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/>. */
+
+ .option nopic
+ .text
+
+ .align 1
+ .globl bar
+ .type bar, @function
+bar:
+ tail 1f
+ .size bar, .-func
+
+ .align 1
+ .globl func
+ .type func, @function
+func:
+ /* A fake 8 byte instruction. This is never executed, but the
+ prologue scanner will try to decode it. These long
+ instructions are ISA extensions, I use .byte rather than an
+ actual instruction mnemonic so that the test can be compiled
+ with a toolchain that doesn't include any long instruction
+ extensions. */
+ .byte 0x3f, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00
+1:
+ addi sp,sp,-16
+ sw s0,12(sp)
+ addi s0,sp,16
+ nop
+ lw s0,12(sp)
+ addi sp,sp,16
+ jr ra
+ .size func, .-func
diff --git a/gdb/testsuite/gdb.arch/riscv-unwind-long-insn.c b/gdb/testsuite/gdb.arch/riscv-unwind-long-insn.c
new file mode 100644
index 0000000000..d601e2d3d8
--- /dev/null
+++ b/gdb/testsuite/gdb.arch/riscv-unwind-long-insn.c
@@ -0,0 +1,25 @@
+/* Copyright 2019 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 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/>. */
+
+extern void func (void);
+extern void bar (void);
+
+int
+main ()
+{
+ bar ();
+ func ();
+ return 0;
+}
diff --git a/gdb/testsuite/gdb.arch/riscv-unwind-long-insn.exp b/gdb/testsuite/gdb.arch/riscv-unwind-long-insn.exp
new file mode 100644
index 0000000000..ddb914295c
--- /dev/null
+++ b/gdb/testsuite/gdb.arch/riscv-unwind-long-insn.exp
@@ -0,0 +1,61 @@
+# Copyright 2019 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 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/>.
+
+# This tests GDB's ability to handle 6 and 8 byte instructions in the
+# RISC-V prologue scanner. These instruction should be ignored, but
+# should not result in an error that interrupts the debug session.
+#
+# Each of the files riscv-unwind-long-insn-*.s include a function
+# (func) that contains a fake long instruction (6 or 8 bytes) in the
+# prologue. We trick GDB into parsing the fake instruction by tail
+# calling from a different function, 'bar' to the middle of 'func'.
+
+if {![istarget "riscv*-*-*"]} {
+ verbose "Skipping ${gdb_test_file_name}."
+ return
+}
+
+foreach_with_prefix {insn_size} {6 8} {
+ standard_testfile riscv-unwind-long-insn.c \
+ riscv-unwind-long-insn-${insn_size}.s
+
+ set testfile "${testfile}-${insn_size}"
+ if {[prepare_for_testing "failed to prepare" $testfile \
+ "$srcfile $srcfile2" debug]} {
+ return -1
+ }
+
+ if ![runto_main] then {
+ fail "can't run to main"
+ return 0
+ }
+
+ gdb_breakpoint "bar"
+ gdb_continue_to_breakpoint "bar"
+
+ # This next single instruction step takes us through a tail-call
+ # from 'bar' into 'func'.
+ gdb_test "si" "func \(\).*"
+
+ # Now check that we have a sane backtrace.
+ gdb_test "bt" \
+ [multi_line \
+ "#0\[ \t\]*func \\\(\\\) at .*$srcfile2:\[0-9\]+" \
+ "#1\[ \t\]*$hex in main \\\(\\\) at .*$srcfile:\[0-9\]+"] \
+ "Backtrace to the main frame"
+
+ # Finally finish, and we should end up back in main.
+ gdb_test "finish" "main \\\(\\\) at .*$srcfile:.*"
+}