diff mbox series

riscv: Add semihosting support [v4]

Message ID 20200128233224.515233-1-keithp@keithp.com (mailing list archive)
State New, archived
Headers show
Series riscv: Add semihosting support [v4] | expand

Commit Message

Denis V. Lunev" via Jan. 28, 2020, 11:32 p.m. UTC
Adapt the arm semihosting support code for RISCV. This implementation
is based on the standard for RISC-V semihosting as documented in

	https://riscv.org/specifications/

Signed-off-by: Keith Packard <keithp@keithp.com>

---

v2:
	Update PC after exception is handled to follow
	change in the ARM version for SYS_READC

v3:
	Disallow semihosting in user mode; report a regular
	breakpoint in that case.

v4:
	Fix errors reported by checkpatch
---
 default-configs/riscv32-softmmu.mak           |    1 +
 linux-user/qemu.h                             |    2 +
 qemu-options.hx                               |    4 +-
 target/riscv/Makefile.objs                    |    2 +-
 target/riscv/cpu.h                            |    2 +
 target/riscv/cpu_bits.h                       |    1 +
 target/riscv/cpu_helper.c                     |    9 +
 .../riscv/insn_trans/trans_privileged.inc.c   |   24 +-
 target/riscv/riscv-semi.c                     | 1084 +++++++++++++++++
 target/riscv/translate.c                      |   11 +
 10 files changed, 1136 insertions(+), 4 deletions(-)
 create mode 100644 target/riscv/riscv-semi.c

Comments

Peter Maydell Jan. 29, 2020, 3:58 p.m. UTC | #1
On Tue, 28 Jan 2020 at 23:34, Keith Packard via <qemu-devel@nongnu.org> wrote:
>
> Adapt the arm semihosting support code for RISCV. This implementation
> is based on the standard for RISC-V semihosting as documented in
>
>         https://riscv.org/specifications/
>
> Signed-off-by: Keith Packard <keithp@keithp.com>
>
> ---

> + *  ARM Semihosting is documented in:
> + *     Semihosting for AArch32 and AArch64 Release 2.0
> + *     https://static.docs.arm.com/100863/0200/semihosting.pdf

True but irrelevant. You need to refer to a proper
risc-v specification for your semihosting.

> +    case TARGET_SYS_EXIT:
> +    case TARGET_SYS_EXIT_EXTENDED:
> +        if (nr == TARGET_SYS_EXIT_EXTENDED || sizeof(target_ulong) == 0) {
> +            /*
> +             * The A64 version of SYS_EXIT takes a parameter block,
> +             * so the application-exit type can return a subcode which
> +             * is the exit status code from the application.
> +             * SYS_EXIT_EXTENDED is an a new-in-v2.0 optional function
> +             * which allows A32/T32 guests to also provide a status code.
> +             */

This code and comment describe Arm semihosting, where we
made this bad decision about the API for 32-bit Arm, fixed
it for 64-bit Arm and then put in an extension to add the more
sensible API to 32-bit as a backwards-compatible new function.
Nobody else should make this API choice from the start.
What does RISC-V want to do here? This should be in
your specification.

> +            GET_ARG(0);
> +            GET_ARG(1);
> +
> +            if (arg0 == ADP_Stopped_ApplicationExit) {
> +                ret = arg1;
> +            } else {
> +                ret = 1;
> +            }
> +        } else {
> +            /*
> +             * The A32/T32 version of SYS_EXIT specifies only
> +             * Stopped_ApplicationExit as normal exit, but does not
> +             * allow the guest to specify the exit status code.
> +             * Everything else is considered an error.
> +             */
> +            ret = (args == ADP_Stopped_ApplicationExit) ? 0 : 1;
> +        }
> +        gdb_exit(env, ret);
> +        exit(ret);
> +    case TARGET_SYS_SYNCCACHE:
> +        /*
> +         * Clean the D-cache and invalidate the I-cache for the specified
> +         * virtual address range. This is a nop for us since we don't
> +         * implement caches. This is only present on A64.
> +         */
> +        if (sizeof(target_ulong) == 8) {
> +            return 0;
> +        }
> +        /* fall through -- invalid for A32/T32 */

Again, this is an Arm-ism, where the old A32 ABI
doesn't have this function but the new A64 one does. What
does RISC-V want to specify here?

> +    default:
> +        fprintf(stderr, "qemu: Unsupported SemiHosting SWI 0x%02x\n", nr);
> +        cpu_dump_state(cs, stderr, 0);
> +        abort();

This is repeating the Arm ABI mistake of allowing implementations
to just crash and burn if they're handed a function call they don't
recognize. Ideally it should be avoided in a new ABI.

> +    }
> +}

thanks
-- PMM
Denis V. Lunev" via Jan. 29, 2020, 4:45 p.m. UTC | #2
Peter Maydell <peter.maydell@linaro.org> writes:

> True but irrelevant. You need to refer to a proper
> risc-v specification for your semihosting.

The RISC-V Foundation defined semihosting as relative to the existing
ARM specification, so using a link to that is appropriate here.

Here's the current specification of the unprivileged ISA, which includes
the definition of semihosting

        https://riscv.org/specifications/

While it may be nice in some abstract sense to create a "better"
semihosting spec, that's not what the RISC-V foundation has decided to
do.
Jonathan Behrens Jan. 29, 2020, 6:06 p.m. UTC | #3
The text you are referencing (the couple italic paragraphs below section
2.8 in the unprivileged ISA) is non-normative and "can be skipped if the
reader is only interested in the specification itself". This convention of
making indented italic text non-normative is described at the bottom of
page 1 of the linked document.

Jonathan

On Wed, Jan 29, 2020 at 11:45 AM Keith Packard via <qemu-riscv@nongnu.org>
wrote:

> Peter Maydell <peter.maydell@linaro.org> writes:
>
> > True but irrelevant. You need to refer to a proper
> > risc-v specification for your semihosting.
>
> The RISC-V Foundation defined semihosting as relative to the existing
> ARM specification, so using a link to that is appropriate here.
>
> Here's the current specification of the unprivileged ISA, which includes
> the definition of semihosting
>
>         https://riscv.org/specifications/
>
> While it may be nice in some abstract sense to create a "better"
> semihosting spec, that's not what the RISC-V foundation has decided to
> do.
>
> --
> -keith
>
Peter Maydell Jan. 30, 2020, 10:54 a.m. UTC | #4
On Wed, 29 Jan 2020 at 16:45, Keith Packard <keithp@keithp.com> wrote:
>
> Peter Maydell <peter.maydell@linaro.org> writes:
>
> > True but irrelevant. You need to refer to a proper
> > risc-v specification for your semihosting.
>
> The RISC-V Foundation defined semihosting as relative to the existing
> ARM specification, so using a link to that is appropriate here.
>
> Here's the current specification of the unprivileged ISA, which includes
> the definition of semihosting
>
>         https://riscv.org/specifications/
>
> While it may be nice in some abstract sense to create a "better"
> semihosting spec, that's not what the RISC-V foundation has decided to
> do.

We've gone round this several times. You can't just
say "it's the arm spec", because you're not arm and
your architecture is different. You need to actually
document what the architecture-specific parts are,
even if you want to mostly say "and we follow the
Arm spec most of the time". If you really really want
to say "32 bit RISC-V is gratuitously different from
64-bit RISC-V in these areas because we just blindly
copied the way Arm happened to have historically
developed" then you can do that if you like, but you
need to actually write it down in a document somewhere.

You're trying to implement an ABI which has multiple
different implementations of both consumers and
producers; it's not a good idea to shortcut the
process of actually writing down what the requirements
between the two ends are.

thanks
-- PMM
Palmer Dabbelt Jan. 30, 2020, 11:38 a.m. UTC | #5
On Thu, 30 Jan 2020 10:54:37 GMT (+0000), Peter Maydell wrote:
> On Wed, 29 Jan 2020 at 16:45, Keith Packard <keithp@keithp.com> wrote:
>>
>> Peter Maydell <peter.maydell@linaro.org> writes:
>>
>> > True but irrelevant. You need to refer to a proper
>> > risc-v specification for your semihosting.
>>
>> The RISC-V Foundation defined semihosting as relative to the existing
>> ARM specification, so using a link to that is appropriate here.
>>
>> Here's the current specification of the unprivileged ISA, which includes
>> the definition of semihosting
>>
>>         https://riscv.org/specifications/
>>
>> While it may be nice in some abstract sense to create a "better"
>> semihosting spec, that's not what the RISC-V foundation has decided to
>> do.
>
> We've gone round this several times. You can't just
> say "it's the arm spec", because you're not arm and
> your architecture is different. You need to actually
> document what the architecture-specific parts are,
> even if you want to mostly say "and we follow the
> Arm spec most of the time". If you really really want
> to say "32 bit RISC-V is gratuitously different from
> 64-bit RISC-V in these areas because we just blindly
> copied the way Arm happened to have historically
> developed" then you can do that if you like, but you
> need to actually write it down in a document somewhere.
>
> You're trying to implement an ABI which has multiple
> different implementations of both consumers and
> producers; it's not a good idea to shortcut the
> process of actually writing down what the requirements
> between the two ends are.

I think, more concretely, the issues here are:

* The only mention of "semihosting" in the RISC-V specification is in the
  commentary, which is not considered a normative part of the specification.
* The interface that commentary contains is defined as maybe being obsolete
  already, but no new concrete interface is defined.
* The semihosting comment doesn't define the semihosting call numbers, just the
  sequence to get to a call.  That said, we haven't written down the Linux ABI
  either -- though there's a much larger breadth of software out there that
  implements it and won't break ABI compatibility, so maybe that's considered
  sufficient in Linux land where it's not for semihosting.

It seems to me like the right way to go forward is to put together a
semihosting ABI specification as a RISC-V specification.  That would give
everyone who cares about their pre-specification implementations a chance to
get together to figure out which parts they want to keep, and give us a path to
move forwards.  There's more work that needs to be done here to put together a
system that will actually work (for example, how to we insert breakpoints if
we're using ebreak to indicate semihosting and haven't reserved NOP
sequences?), so this does warrant a bit of thought.

I know it's a bit of a headache in the short term to get a specification
together, but we've had so many headaches of the "there's no spec" sort that at
least it'll be a change of pace :)

>
> thanks
> -- PMM
Peter Maydell Jan. 30, 2020, 12:01 p.m. UTC | #6
On Thu, 30 Jan 2020 at 11:38, Palmer Dabbelt <palmerdabbelt@google.com> wrote:
> * The semihosting comment doesn't define the semihosting call numbers, just the
>   sequence to get to a call.  That said, we haven't written down the Linux ABI
>   either -- though there's a much larger breadth of software out there that
>   implements it and won't break ABI compatibility, so maybe that's considered
>   sufficient in Linux land where it's not for semihosting.

I think the difference with Linux is that there's a clear
single authoritative source for what the ABI is -- Linus's
'mainline' kernel (so for instance random out-of-tree forks,
syscall patch proposals, etc that have not yet hit mainline
don't count as being fixed-in-stone ABI). Semihosting doesn't
have the same single clear "owning project", at least on
the Arm side of things.

I don't think a semihosting specification for RISC-V needs
necessarily to be a very heavyweight thing -- it is, after
all, a debug API at heart -- but I do agree that you should
have one.

thanks
-- PMM
Denis V. Lunev" via March 7, 2020, 7:44 p.m. UTC | #7
Palmer Dabbelt <palmerdabbelt@google.com> writes:

> I think, more concretely, the issues here are:
>
> * The only mention of "semihosting" in the RISC-V specification is in the
>   commentary, which is not considered a normative part of the
> specification.

Ok, we're starting the process of creating a 'real' RISC-V semihosting
spec. This references the existing 2.0 version of the ARM semihosting
specification, defining only the RISC-V specific details necessary.
I've created a public github repository containing this work; that is
likely to eventually move over to the RISC-V github repository.

        https://github.com/keith-packard/riscv-semihosting-spec/

The goal is to enable RISC-V semihosting implementations that share most
of the underlying details with existing ARM semihosting implemenations.
Denis V. Lunev" via April 6, 2020, 5:24 a.m. UTC | #8
Peter Maydell <peter.maydell@linaro.org> writes:

> I don't think a semihosting specification for RISC-V needs
> necessarily to be a very heavyweight thing -- it is, after
> all, a debug API at heart -- but I do agree that you should
> have one.

We've had a couple of weeks of discussion in the RISC-V sw-dev group and
have created a new specification for RISC-V semihosting. It is located
within the RISC-V consortium github account here:

        https://github.com/riscv/riscv-semihosting-spec

The current version is 0.2, and this patch to QEMU implements that
version.
diff mbox series

Patch

diff --git a/default-configs/riscv32-softmmu.mak b/default-configs/riscv32-softmmu.mak
index 1ae077ed87..21b7bedf19 100644
--- a/default-configs/riscv32-softmmu.mak
+++ b/default-configs/riscv32-softmmu.mak
@@ -3,6 +3,7 @@ 
 # Uncomment the following lines to disable these optional devices:
 #
 #CONFIG_PCI_DEVICES=n
+CONFIG_SEMIHOSTING=y
 
 # Boards:
 #
diff --git a/linux-user/qemu.h b/linux-user/qemu.h
index 560a68090e..122d95d6b6 100644
--- a/linux-user/qemu.h
+++ b/linux-user/qemu.h
@@ -105,6 +105,8 @@  typedef struct TaskState {
     /* FPA state */
     FPA11 fpa;
 # endif
+#endif
+#if defined(TARGET_ARM) || defined(TARGET_RISCV)
     int swi_errno;
 #endif
 #if defined(TARGET_I386) && !defined(TARGET_X86_64)
diff --git a/qemu-options.hx b/qemu-options.hx
index 224a8e8712..4892e6b12c 100644
--- a/qemu-options.hx
+++ b/qemu-options.hx
@@ -4109,7 +4109,7 @@  ETEXI
 DEF("semihosting", 0, QEMU_OPTION_semihosting,
     "-semihosting    semihosting mode\n",
     QEMU_ARCH_ARM | QEMU_ARCH_M68K | QEMU_ARCH_XTENSA | QEMU_ARCH_LM32 |
-    QEMU_ARCH_MIPS | QEMU_ARCH_NIOS2)
+    QEMU_ARCH_MIPS | QEMU_ARCH_NIOS2 | QEMU_ARCH_RISCV)
 STEXI
 @item -semihosting
 @findex -semihosting
@@ -4119,7 +4119,7 @@  DEF("semihosting-config", HAS_ARG, QEMU_OPTION_semihosting_config,
     "-semihosting-config [enable=on|off][,target=native|gdb|auto][,chardev=id][,arg=str[,...]]\n" \
     "                semihosting configuration\n",
 QEMU_ARCH_ARM | QEMU_ARCH_M68K | QEMU_ARCH_XTENSA | QEMU_ARCH_LM32 |
-QEMU_ARCH_MIPS | QEMU_ARCH_NIOS2)
+QEMU_ARCH_MIPS | QEMU_ARCH_NIOS2 | QEMU_ARCH_RISCV)
 STEXI
 @item -semihosting-config [enable=on|off][,target=native|gdb|auto][,chardev=id][,arg=str[,...]]
 @findex -semihosting-config
diff --git a/target/riscv/Makefile.objs b/target/riscv/Makefile.objs
index ff651f69f6..6fd7f40e29 100644
--- a/target/riscv/Makefile.objs
+++ b/target/riscv/Makefile.objs
@@ -1,4 +1,4 @@ 
-obj-y += translate.o op_helper.o cpu_helper.o cpu.o csr.o fpu_helper.o gdbstub.o
+obj-y += translate.o op_helper.o cpu_helper.o cpu.o csr.o fpu_helper.o gdbstub.o riscv-semi.o
 obj-$(CONFIG_SOFTMMU) += pmp.o
 
 ifeq ($(CONFIG_SOFTMMU),y)
diff --git a/target/riscv/cpu.h b/target/riscv/cpu.h
index de0a8d893a..310ec6ab40 100644
--- a/target/riscv/cpu.h
+++ b/target/riscv/cpu.h
@@ -338,6 +338,8 @@  void riscv_cpu_register_gdb_regs_for_features(CPUState *cs);
 typedef CPURISCVState CPUArchState;
 typedef RISCVCPU ArchCPU;
 
+target_ulong do_riscv_semihosting(CPURISCVState *env);
+
 #include "exec/cpu-all.h"
 
 #endif /* RISCV_CPU_H */
diff --git a/target/riscv/cpu_bits.h b/target/riscv/cpu_bits.h
index e99834856c..c45745d73b 100644
--- a/target/riscv/cpu_bits.h
+++ b/target/riscv/cpu_bits.h
@@ -496,6 +496,7 @@ 
 #define RISCV_EXCP_INST_PAGE_FAULT         0xc /* since: priv-1.10.0 */
 #define RISCV_EXCP_LOAD_PAGE_FAULT         0xd /* since: priv-1.10.0 */
 #define RISCV_EXCP_STORE_PAGE_FAULT        0xf /* since: priv-1.10.0 */
+#define RISCV_EXCP_SEMIHOST                0x10
 
 #define RISCV_EXCP_INT_FLAG                0x80000000
 #define RISCV_EXCP_INT_MASK                0x7fffffff
diff --git a/target/riscv/cpu_helper.c b/target/riscv/cpu_helper.c
index 85403da9c8..6a04469936 100644
--- a/target/riscv/cpu_helper.c
+++ b/target/riscv/cpu_helper.c
@@ -532,6 +532,15 @@  void riscv_cpu_do_interrupt(CPUState *cs)
         [PRV_M] = RISCV_EXCP_M_ECALL
     };
 
+    if  (cause == RISCV_EXCP_SEMIHOST) {
+        if (env->priv >= PRV_S) {
+            env->gpr[xA0] = do_riscv_semihosting(env);
+            env->pc += 4;
+            return;
+        }
+        cause = RISCV_EXCP_BREAKPOINT;
+    }
+
     if (!async) {
         /* set tval to badaddr for traps with address information */
         switch (cause) {
diff --git a/target/riscv/insn_trans/trans_privileged.inc.c b/target/riscv/insn_trans/trans_privileged.inc.c
index c5e4b3e49a..decd59bbf2 100644
--- a/target/riscv/insn_trans/trans_privileged.inc.c
+++ b/target/riscv/insn_trans/trans_privileged.inc.c
@@ -29,7 +29,29 @@  static bool trans_ecall(DisasContext *ctx, arg_ecall *a)
 
 static bool trans_ebreak(DisasContext *ctx, arg_ebreak *a)
 {
-    generate_exception(ctx, RISCV_EXCP_BREAKPOINT);
+    uint32_t pre    = opcode_at(&ctx->base, ctx->base.pc_next - 4);
+    uint32_t ebreak = opcode_at(&ctx->base, ctx->base.pc_next);
+    uint32_t post   = opcode_at(&ctx->base, ctx->base.pc_next + 4);
+
+    /*
+     * The RISC-V semihosting spec specifies the following
+     * three-instruction sequence to flag a semihosting call:
+     *
+     *      slli zero, zero, 0x1f       0x01f01013
+     *      ebreak                      0x00100073
+     *      srai zero, zero, 0x7        0x40705013
+     *
+     * The two shift operations on the zero register are no-ops, used
+     * here to signify a semihosting exception, rather than a breakpoint.
+     *
+     * Uncompressed instructions are used so that the sequence is easy
+     * to validate.
+     */
+    if  (pre == 0x01f01013 && ebreak == 0x00100073 && post == 0x40705013) {
+        generate_exception(ctx, RISCV_EXCP_SEMIHOST);
+    } else {
+        generate_exception(ctx, RISCV_EXCP_BREAKPOINT);
+    }
     exit_tb(ctx); /* no chaining */
     ctx->base.is_jmp = DISAS_NORETURN;
     return true;
diff --git a/target/riscv/riscv-semi.c b/target/riscv/riscv-semi.c
new file mode 100644
index 0000000000..18707f32c5
--- /dev/null
+++ b/target/riscv/riscv-semi.c
@@ -0,0 +1,1084 @@ 
+/*
+ *  RISC-V semihosting syscalls
+ *
+ *  Copyright (c) 2005, 2007 CodeSourcery.
+ *  Copyright (c) 2019 Linaro
+ *  Copyright © 2019 Keith Packard
+ *  Written by Paul Brook.
+ *  Adapted for RISC-V by Keith Packard
+ *
+ *  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, see <http://www.gnu.org/licenses/>.
+ *
+ *  ARM Semihosting is documented in:
+ *     Semihosting for AArch32 and AArch64 Release 2.0
+ *     https://static.docs.arm.com/100863/0200/semihosting.pdf
+ */
+
+#include "qemu/osdep.h"
+
+#include "cpu.h"
+#include "hw/semihosting/semihost.h"
+#include "hw/semihosting/console.h"
+#include "qemu/log.h"
+#ifdef CONFIG_USER_ONLY
+#include "qemu.h"
+
+#define RISCV_HEAP_SIZE (128 * 1024 * 1024)
+#else
+#include "exec/gdbstub.h"
+#include "qemu/cutils.h"
+#endif
+
+#define TARGET_SYS_OPEN        0x01
+#define TARGET_SYS_CLOSE       0x02
+#define TARGET_SYS_WRITEC      0x03
+#define TARGET_SYS_WRITE0      0x04
+#define TARGET_SYS_WRITE       0x05
+#define TARGET_SYS_READ        0x06
+#define TARGET_SYS_READC       0x07
+#define TARGET_SYS_ISTTY       0x09
+#define TARGET_SYS_SEEK        0x0a
+#define TARGET_SYS_FLEN        0x0c
+#define TARGET_SYS_TMPNAM      0x0d
+#define TARGET_SYS_REMOVE      0x0e
+#define TARGET_SYS_RENAME      0x0f
+#define TARGET_SYS_CLOCK       0x10
+#define TARGET_SYS_TIME        0x11
+#define TARGET_SYS_SYSTEM      0x12
+#define TARGET_SYS_ERRNO       0x13
+#define TARGET_SYS_GET_CMDLINE 0x15
+#define TARGET_SYS_HEAPINFO    0x16
+#define TARGET_SYS_EXIT        0x18
+#define TARGET_SYS_SYNCCACHE   0x19
+#define TARGET_SYS_EXIT_EXTENDED 0x20
+
+/*
+ * ADP_Stopped_ApplicationExit is used for exit(0),
+ * anything else is implemented as exit(1)
+ */
+#define ADP_Stopped_ApplicationExit     (0x20026)
+
+#ifndef O_BINARY
+#define O_BINARY 0
+#endif
+
+#define GDB_O_RDONLY  0x000
+#define GDB_O_WRONLY  0x001
+#define GDB_O_RDWR    0x002
+#define GDB_O_APPEND  0x008
+#define GDB_O_CREAT   0x200
+#define GDB_O_TRUNC   0x400
+#define GDB_O_BINARY  0
+
+static int gdb_open_modeflags[12] = {
+    GDB_O_RDONLY,
+    GDB_O_RDONLY | GDB_O_BINARY,
+    GDB_O_RDWR,
+    GDB_O_RDWR | GDB_O_BINARY,
+    GDB_O_WRONLY | GDB_O_CREAT | GDB_O_TRUNC,
+    GDB_O_WRONLY | GDB_O_CREAT | GDB_O_TRUNC | GDB_O_BINARY,
+    GDB_O_RDWR | GDB_O_CREAT | GDB_O_TRUNC,
+    GDB_O_RDWR | GDB_O_CREAT | GDB_O_TRUNC | GDB_O_BINARY,
+    GDB_O_WRONLY | GDB_O_CREAT | GDB_O_APPEND,
+    GDB_O_WRONLY | GDB_O_CREAT | GDB_O_APPEND | GDB_O_BINARY,
+    GDB_O_RDWR | GDB_O_CREAT | GDB_O_APPEND,
+    GDB_O_RDWR | GDB_O_CREAT | GDB_O_APPEND | GDB_O_BINARY
+};
+
+static int open_modeflags[12] = {
+    O_RDONLY,
+    O_RDONLY | O_BINARY,
+    O_RDWR,
+    O_RDWR | O_BINARY,
+    O_WRONLY | O_CREAT | O_TRUNC,
+    O_WRONLY | O_CREAT | O_TRUNC | O_BINARY,
+    O_RDWR | O_CREAT | O_TRUNC,
+    O_RDWR | O_CREAT | O_TRUNC | O_BINARY,
+    O_WRONLY | O_CREAT | O_APPEND,
+    O_WRONLY | O_CREAT | O_APPEND | O_BINARY,
+    O_RDWR | O_CREAT | O_APPEND,
+    O_RDWR | O_CREAT | O_APPEND | O_BINARY
+};
+
+typedef enum GuestFDType {
+    GuestFDUnused = 0,
+    GuestFDHost = 1,
+    GuestFDGDB = 2,
+    GuestFDFeatureFile = 3,
+} GuestFDType;
+
+/*
+ * Guest file descriptors are integer indexes into an array of
+ * these structures (we will dynamically resize as necessary).
+ */
+typedef struct GuestFD {
+    GuestFDType type;
+    union {
+        int hostfd;
+        target_ulong featurefile_offset;
+    };
+} GuestFD;
+
+static GArray *guestfd_array;
+
+/*
+ * Allocate a new guest file descriptor and return it; if we
+ * couldn't allocate a new fd then return -1.
+ * This is a fairly simplistic implementation because we don't
+ * expect that most semihosting guest programs will make very
+ * heavy use of opening and closing fds.
+ */
+static int alloc_guestfd(void)
+{
+    guint i;
+
+    if (!guestfd_array) {
+        /* New entries zero-initialized, i.e. type GuestFDUnused */
+        guestfd_array = g_array_new(FALSE, TRUE, sizeof(GuestFD));
+    }
+
+    for (i = 0; i < guestfd_array->len; i++) {
+        GuestFD *gf = &g_array_index(guestfd_array, GuestFD, i);
+
+        if (gf->type == GuestFDUnused) {
+            return i;
+        }
+    }
+
+    /* All elements already in use: expand the array */
+    g_array_set_size(guestfd_array, i + 1);
+    return i;
+}
+
+/*
+ * Look up the guestfd in the data structure; return NULL
+ * for out of bounds, but don't check whether the slot is unused.
+ * This is used internally by the other guestfd functions.
+ */
+static GuestFD *do_get_guestfd(int guestfd)
+{
+    if (!guestfd_array) {
+        return NULL;
+    }
+
+    if (guestfd < 0 || guestfd >= guestfd_array->len) {
+        return NULL;
+    }
+
+    return &g_array_index(guestfd_array, GuestFD, guestfd);
+}
+
+/*
+ * Associate the specified guest fd (which must have been
+ * allocated via alloc_fd() and not previously used) with
+ * the specified host/gdb fd.
+ */
+static void associate_guestfd(int guestfd, int hostfd)
+{
+    GuestFD *gf = do_get_guestfd(guestfd);
+
+    assert(gf);
+    gf->type = use_gdb_syscalls() ? GuestFDGDB : GuestFDHost;
+    gf->hostfd = hostfd;
+}
+
+/*
+ * Deallocate the specified guest file descriptor. This doesn't
+ * close the host fd, it merely undoes the work of alloc_fd().
+ */
+static void dealloc_guestfd(int guestfd)
+{
+    GuestFD *gf = do_get_guestfd(guestfd);
+
+    assert(gf);
+    gf->type = GuestFDUnused;
+}
+
+/*
+ * Given a guest file descriptor, get the associated struct.
+ * If the fd is not valid, return NULL. This is the function
+ * used by the various semihosting calls to validate a handle
+ * from the guest.
+ * Note: calling alloc_guestfd() or dealloc_guestfd() will
+ * invalidate any GuestFD* obtained by calling this function.
+ */
+static GuestFD *get_guestfd(int guestfd)
+{
+    GuestFD *gf = do_get_guestfd(guestfd);
+
+    if (!gf || gf->type == GuestFDUnused) {
+        return NULL;
+    }
+    return gf;
+}
+
+/*
+ * The semihosting API has no concept of its errno being thread-safe,
+ * as the API design predates SMP CPUs and was intended as a simple
+ * real-hardware set of debug functionality. For QEMU, we make the
+ * errno be per-thread in linux-user mode; in softmmu it is a simple
+ * global, and we assume that the guest takes care of avoiding any races.
+ */
+#ifndef CONFIG_USER_ONLY
+static target_ulong syscall_err;
+
+#include "exec/softmmu-semi.h"
+#endif
+
+static inline uint32_t set_swi_errno(CPURISCVState *env, uint32_t code)
+{
+    if (code == (uint32_t)-1) {
+#ifdef CONFIG_USER_ONLY
+        CPUState *cs = env_cpu(env);
+        TaskState *ts = cs->opaque;
+
+        ts->swi_errno = errno;
+#else
+        syscall_err = errno;
+#endif
+    }
+    return code;
+}
+
+static inline uint32_t get_swi_errno(CPURISCVState *env)
+{
+#ifdef CONFIG_USER_ONLY
+    CPUState *cs = env_cpu(env);
+    TaskState *ts = cs->opaque;
+
+    return ts->swi_errno;
+#else
+    return syscall_err;
+#endif
+}
+
+static target_ulong riscv_semi_syscall_len;
+
+static void riscv_semi_cb(CPUState *cs, target_ulong ret, target_ulong err)
+{
+    RISCVCPU *cpu = RISCV_CPU(cs);
+    CPURISCVState *env = &cpu->env;
+    target_ulong reg0 = env->gpr[xA0];
+
+    if (ret == (target_ulong)-1) {
+        errno = err;
+        set_swi_errno(env, -1);
+        reg0 = ret;
+    } else {
+        /* Fixup syscalls that use nonstardard return conventions.  */
+        switch (reg0) {
+        case TARGET_SYS_WRITE:
+        case TARGET_SYS_READ:
+            reg0 = riscv_semi_syscall_len - ret;
+            break;
+        case TARGET_SYS_SEEK:
+            reg0 = 0;
+            break;
+        default:
+            reg0 = ret;
+            break;
+        }
+    }
+    env->gpr[xA0] = reg0;
+}
+
+static target_ulong riscv_flen_buf(RISCVCPU *cpu)
+{
+    /*
+     * Return an address in target memory of 64 bytes where the remote
+     * gdb should write its stat struct. (The format of this structure
+     * is defined by GDB's remote protocol and is not target-specific.)
+     * We put this on the guest's stack just below SP.
+     */
+    CPURISCVState *env = &cpu->env;
+    target_ulong sp;
+
+    sp = env->gpr[xSP];
+
+    return sp - 64;
+}
+
+static void riscv_semi_flen_cb(CPUState *cs, target_ulong ret, target_ulong err)
+{
+    RISCVCPU *cpu = RISCV_CPU(cs);
+    CPURISCVState *env = &cpu->env;
+    /*
+     * The size is always stored in big-endian order, extract
+     *  the value. We assume the size always fit in 32 bits.
+     */
+    uint32_t size;
+    cpu_memory_rw_debug(cs, riscv_flen_buf(cpu) + 32, (uint8_t *)&size, 4, 0);
+    size = be32_to_cpu(size);
+    env->gpr[xA0] = size;
+    errno = err;
+    set_swi_errno(env, -1);
+}
+
+static int riscv_semi_open_guestfd;
+
+static void riscv_semi_open_cb(CPUState *cs, target_ulong ret, target_ulong err)
+{
+    RISCVCPU *cpu = RISCV_CPU(cs);
+    CPURISCVState *env = &cpu->env;
+    if (ret == (target_ulong)-1) {
+        errno = err;
+        set_swi_errno(env, -1);
+        dealloc_guestfd(riscv_semi_open_guestfd);
+    } else {
+        associate_guestfd(riscv_semi_open_guestfd, ret);
+        ret = riscv_semi_open_guestfd;
+    }
+
+    env->gpr[xA0] = ret;
+}
+
+static target_ulong riscv_gdb_syscall(RISCVCPU *cpu, gdb_syscall_complete_cb cb,
+                                    const char *fmt, ...)
+{
+    va_list va;
+    CPURISCVState *env = &cpu->env;
+
+    va_start(va, fmt);
+    gdb_do_syscallv(cb, fmt, va);
+    va_end(va);
+
+    /*
+     * FIXME: in softmmu mode, the gdbstub will schedule our callback
+     * to occur, but will not actually call it to complete the syscall
+     * until after this function has returned and we are back in the
+     * CPU main loop. Therefore callers to this function must not
+     * do anything with its return value, because it is not necessarily
+     * the result of the syscall, but could just be the old value of X0.
+     * The only thing safe to do with this is that the callers of
+     * do_riscv_semihosting() will write it straight back into X0.
+     * (In linux-user mode, the callback will have happened before
+     * gdb_do_syscallv() returns.)
+     *
+     * We should tidy this up so neither this function nor
+     * do_riscv_semihosting() return a value, so the mistake of
+     * doing something with the return value is not possible to make.
+     */
+
+    return env->gpr[xA0];
+}
+
+/*
+ * Types for functions implementing various semihosting calls
+ * for specific types of guest file descriptor. These must all
+ * do the work and return the required return value for the guest,
+ * setting the guest errno if appropriate.
+ */
+typedef uint32_t sys_closefn(RISCVCPU *cpu, GuestFD *gf);
+typedef uint32_t sys_writefn(RISCVCPU *cpu, GuestFD *gf,
+                             target_ulong buf, uint32_t len);
+typedef uint32_t sys_readfn(RISCVCPU *cpu, GuestFD *gf,
+                            target_ulong buf, uint32_t len);
+typedef uint32_t sys_isattyfn(RISCVCPU *cpu, GuestFD *gf);
+typedef uint32_t sys_seekfn(RISCVCPU *cpu, GuestFD *gf,
+                            target_ulong offset);
+typedef uint32_t sys_flenfn(RISCVCPU *cpu, GuestFD *gf);
+
+static uint32_t host_closefn(RISCVCPU *cpu, GuestFD *gf)
+{
+    CPURISCVState *env = &cpu->env;
+
+    return set_swi_errno(env, close(gf->hostfd));
+}
+
+static uint32_t host_writefn(RISCVCPU *cpu, GuestFD *gf,
+                             target_ulong buf, uint32_t len)
+{
+    uint32_t ret;
+    CPURISCVState *env = &cpu->env;
+    char *s = lock_user(VERIFY_READ, buf, len, 1);
+    if (!s) {
+        /* Return bytes not written on error */
+        return len;
+    }
+    ret = set_swi_errno(env, write(gf->hostfd, s, len));
+    unlock_user(s, buf, 0);
+    if (ret == (uint32_t)-1) {
+        ret = 0;
+    }
+    /* Return bytes not written */
+    return len - ret;
+}
+
+static uint32_t host_readfn(RISCVCPU *cpu, GuestFD *gf,
+                            target_ulong buf, uint32_t len)
+{
+    uint32_t ret;
+    CPURISCVState *env = &cpu->env;
+    char *s = lock_user(VERIFY_WRITE, buf, len, 0);
+    if (!s) {
+        /* return bytes not read */
+        return len;
+    }
+    do {
+        ret = set_swi_errno(env, read(gf->hostfd, s, len));
+    } while (ret == -1 && errno == EINTR);
+    unlock_user(s, buf, len);
+    if (ret == (uint32_t)-1) {
+        ret = 0;
+    }
+    /* Return bytes not read */
+    return len - ret;
+}
+
+static uint32_t host_isattyfn(RISCVCPU *cpu, GuestFD *gf)
+{
+    return isatty(gf->hostfd);
+}
+
+static uint32_t host_seekfn(RISCVCPU *cpu, GuestFD *gf, target_ulong offset)
+{
+    CPURISCVState *env = &cpu->env;
+    uint32_t ret = set_swi_errno(env, lseek(gf->hostfd, offset, SEEK_SET));
+    if (ret == (uint32_t)-1) {
+        return -1;
+    }
+    return 0;
+}
+
+static uint32_t host_flenfn(RISCVCPU *cpu, GuestFD *gf)
+{
+    CPURISCVState *env = &cpu->env;
+    struct stat buf;
+    uint32_t ret = set_swi_errno(env, fstat(gf->hostfd, &buf));
+    if (ret == (uint32_t)-1) {
+        return -1;
+    }
+    return buf.st_size;
+}
+
+static uint32_t gdb_closefn(RISCVCPU *cpu, GuestFD *gf)
+{
+    return riscv_gdb_syscall(cpu, riscv_semi_cb, "close,%x", gf->hostfd);
+}
+
+static uint32_t gdb_writefn(RISCVCPU *cpu, GuestFD *gf,
+                            target_ulong buf, uint32_t len)
+{
+    riscv_semi_syscall_len = len;
+    return riscv_gdb_syscall(cpu, riscv_semi_cb, "write,%x,%x,%x",
+                           gf->hostfd, buf, len);
+}
+
+static uint32_t gdb_readfn(RISCVCPU *cpu, GuestFD *gf,
+                           target_ulong buf, uint32_t len)
+{
+    riscv_semi_syscall_len = len;
+    return riscv_gdb_syscall(cpu, riscv_semi_cb, "read,%x,%x,%x",
+                           gf->hostfd, buf, len);
+}
+
+static uint32_t gdb_isattyfn(RISCVCPU *cpu, GuestFD *gf)
+{
+    return riscv_gdb_syscall(cpu, riscv_semi_cb, "isatty,%x", gf->hostfd);
+}
+
+static uint32_t gdb_seekfn(RISCVCPU *cpu, GuestFD *gf, target_ulong offset)
+{
+    return riscv_gdb_syscall(cpu, riscv_semi_cb, "lseek,%x,%x,0",
+                           gf->hostfd, offset);
+}
+
+static uint32_t gdb_flenfn(RISCVCPU *cpu, GuestFD *gf)
+{
+    return riscv_gdb_syscall(cpu, riscv_semi_flen_cb, "fstat,%x,%x",
+                           gf->hostfd, riscv_flen_buf(cpu));
+}
+
+#define SHFB_MAGIC_0 0x53
+#define SHFB_MAGIC_1 0x48
+#define SHFB_MAGIC_2 0x46
+#define SHFB_MAGIC_3 0x42
+
+/* Feature bits reportable in feature byte 0 */
+#define SH_EXT_EXIT_EXTENDED (1 << 0)
+#define SH_EXT_STDOUT_STDERR (1 << 1)
+
+static const uint8_t featurefile_data[] = {
+    SHFB_MAGIC_0,
+    SHFB_MAGIC_1,
+    SHFB_MAGIC_2,
+    SHFB_MAGIC_3,
+    SH_EXT_EXIT_EXTENDED | SH_EXT_STDOUT_STDERR, /* Feature byte 0 */
+};
+
+static void init_featurefile_guestfd(int guestfd)
+{
+    GuestFD *gf = do_get_guestfd(guestfd);
+
+    assert(gf);
+    gf->type = GuestFDFeatureFile;
+    gf->featurefile_offset = 0;
+}
+
+static uint32_t featurefile_closefn(RISCVCPU *cpu, GuestFD *gf)
+{
+    /* Nothing to do */
+    return 0;
+}
+
+static uint32_t featurefile_writefn(RISCVCPU *cpu, GuestFD *gf,
+                                    target_ulong buf, uint32_t len)
+{
+    /* This fd can never be open for writing */
+    CPURISCVState *env = &cpu->env;
+
+    errno = EBADF;
+    return set_swi_errno(env, -1);
+}
+
+static uint32_t featurefile_readfn(RISCVCPU *cpu, GuestFD *gf,
+                                   target_ulong buf, uint32_t len)
+{
+    uint32_t i;
+#ifndef CONFIG_USER_ONLY
+    CPURISCVState *env = &cpu->env;
+#endif
+    char *s;
+
+    s = lock_user(VERIFY_WRITE, buf, len, 0);
+    if (!s) {
+        return len;
+    }
+
+    for (i = 0; i < len; i++) {
+        if (gf->featurefile_offset >= sizeof(featurefile_data)) {
+            break;
+        }
+        s[i] = featurefile_data[gf->featurefile_offset];
+        gf->featurefile_offset++;
+    }
+
+    unlock_user(s, buf, len);
+
+    /* Return number of bytes not read */
+    return len - i;
+}
+
+static uint32_t featurefile_isattyfn(RISCVCPU *cpu, GuestFD *gf)
+{
+    return 0;
+}
+
+static uint32_t featurefile_seekfn(RISCVCPU *cpu, GuestFD *gf,
+                                   target_ulong offset)
+{
+    gf->featurefile_offset = offset;
+    return 0;
+}
+
+static uint32_t featurefile_flenfn(RISCVCPU *cpu, GuestFD *gf)
+{
+    return sizeof(featurefile_data);
+}
+
+typedef struct GuestFDFunctions {
+    sys_closefn *closefn;
+    sys_writefn *writefn;
+    sys_readfn *readfn;
+    sys_isattyfn *isattyfn;
+    sys_seekfn *seekfn;
+    sys_flenfn *flenfn;
+} GuestFDFunctions;
+
+static const GuestFDFunctions guestfd_fns[] = {
+    [GuestFDHost] = {
+        .closefn = host_closefn,
+        .writefn = host_writefn,
+        .readfn = host_readfn,
+        .isattyfn = host_isattyfn,
+        .seekfn = host_seekfn,
+        .flenfn = host_flenfn,
+    },
+    [GuestFDGDB] = {
+        .closefn = gdb_closefn,
+        .writefn = gdb_writefn,
+        .readfn = gdb_readfn,
+        .isattyfn = gdb_isattyfn,
+        .seekfn = gdb_seekfn,
+        .flenfn = gdb_flenfn,
+    },
+    [GuestFDFeatureFile] = {
+        .closefn = featurefile_closefn,
+        .writefn = featurefile_writefn,
+        .readfn = featurefile_readfn,
+        .isattyfn = featurefile_isattyfn,
+        .seekfn = featurefile_seekfn,
+        .flenfn = featurefile_flenfn,
+    },
+};
+
+/*
+ * Read the input value from the argument block; fail the semihosting
+ * call if the memory read fails.
+ */
+#define GET_ARG(n) do {                                                 \
+        if (get_user_ual(arg ## n, args + (n) * sizeof(target_ulong))) { \
+            errno = EFAULT;                                             \
+            return set_swi_errno(env, -1);                              \
+        }                                                               \
+    } while (0)
+
+#define SET_ARG(n, val)                                 \
+    put_user_ual(val, args + (n) * sizeof(target_ulong))
+
+/*
+ * Do a semihosting call.
+ *
+ * The specification always says that the "return register" either
+ * returns a specific value or is corrupted, so we don't need to
+ * report to our caller whether we are returning a value or trying to
+ * leave the register unchanged. We use 0xdeadbeef as the return value
+ * when there isn't a defined return value for the call.
+ */
+target_ulong do_riscv_semihosting(CPURISCVState *env)
+{
+    RISCVCPU *cpu = env_archcpu(env);
+    CPUState *cs = env_cpu(env);
+    target_ulong args;
+    target_ulong arg0, arg1, arg2, arg3;
+    char *s;
+    int nr;
+    uint32_t ret;
+    uint32_t len;
+    GuestFD *gf;
+
+    nr = env->gpr[xA0] & 0xffffffffU;
+    args = env->gpr[xA1];
+
+    switch (nr) {
+    case TARGET_SYS_OPEN:
+    {
+        int guestfd;
+
+        GET_ARG(0);
+        GET_ARG(1);
+        GET_ARG(2);
+        s = lock_user_string(arg0);
+        if (!s) {
+            errno = EFAULT;
+            return set_swi_errno(env, -1);
+        }
+        if (arg1 >= 12) {
+            unlock_user(s, arg0, 0);
+            errno = EINVAL;
+            return set_swi_errno(env, -1);
+        }
+
+        guestfd = alloc_guestfd();
+        if (guestfd < 0) {
+            unlock_user(s, arg0, 0);
+            errno = EMFILE;
+            return set_swi_errno(env, -1);
+        }
+
+        if (strcmp(s, ":tt") == 0) {
+            int result_fileno;
+
+            /*
+             * We implement SH_EXT_STDOUT_STDERR, so:
+             *  open for read == stdin
+             *  open for write == stdout
+             *  open for append == stderr
+             */
+            if (arg1 < 4) {
+                result_fileno = STDIN_FILENO;
+            } else if (arg1 < 8) {
+                result_fileno = STDOUT_FILENO;
+            } else {
+                result_fileno = STDERR_FILENO;
+            }
+            associate_guestfd(guestfd, result_fileno);
+            unlock_user(s, arg0, 0);
+            return guestfd;
+        }
+        if (strcmp(s, ":semihosting-features") == 0) {
+            unlock_user(s, arg0, 0);
+            /* We must fail opens for modes other than 0 ('r') or 1 ('rb') */
+            if (arg1 != 0 && arg1 != 1) {
+                dealloc_guestfd(guestfd);
+                errno = EACCES;
+                return set_swi_errno(env, -1);
+            }
+            init_featurefile_guestfd(guestfd);
+            return guestfd;
+        }
+
+        if (use_gdb_syscalls()) {
+            riscv_semi_open_guestfd = guestfd;
+            ret = riscv_gdb_syscall(cpu, riscv_semi_open_cb,
+                                    "open,%s,%x,1a4", arg0,
+                                    (int)arg2 + 1, gdb_open_modeflags[arg1]);
+        } else {
+            ret = set_swi_errno(env, open(s, open_modeflags[arg1], 0644));
+            if (ret == (uint32_t)-1) {
+                dealloc_guestfd(guestfd);
+            } else {
+                associate_guestfd(guestfd, ret);
+                ret = guestfd;
+            }
+        }
+        unlock_user(s, arg0, 0);
+        return ret;
+    }
+    case TARGET_SYS_CLOSE:
+        GET_ARG(0);
+
+        gf = get_guestfd(arg0);
+        if (!gf) {
+            errno = EBADF;
+            return set_swi_errno(env, -1);
+        }
+
+        ret = guestfd_fns[gf->type].closefn(cpu, gf);
+        dealloc_guestfd(arg0);
+        return ret;
+    case TARGET_SYS_WRITEC:
+        qemu_semihosting_console_outc(env, args);
+        return 0xdeadbeef;
+    case TARGET_SYS_WRITE0:
+        return qemu_semihosting_console_outs(env, args);
+    case TARGET_SYS_WRITE:
+        GET_ARG(0);
+        GET_ARG(1);
+        GET_ARG(2);
+        len = arg2;
+
+        gf = get_guestfd(arg0);
+        if (!gf) {
+            errno = EBADF;
+            return set_swi_errno(env, -1);
+        }
+
+        return guestfd_fns[gf->type].writefn(cpu, gf, arg1, len);
+    case TARGET_SYS_READ:
+        GET_ARG(0);
+        GET_ARG(1);
+        GET_ARG(2);
+        len = arg2;
+
+        gf = get_guestfd(arg0);
+        if (!gf) {
+            errno = EBADF;
+            return set_swi_errno(env, -1);
+        }
+
+        return guestfd_fns[gf->type].readfn(cpu, gf, arg1, len);
+    case TARGET_SYS_READC:
+        return qemu_semihosting_console_inc(env);
+    case TARGET_SYS_ISTTY:
+        GET_ARG(0);
+
+        gf = get_guestfd(arg0);
+        if (!gf) {
+            errno = EBADF;
+            return set_swi_errno(env, -1);
+        }
+
+        return guestfd_fns[gf->type].isattyfn(cpu, gf);
+    case TARGET_SYS_SEEK:
+        GET_ARG(0);
+        GET_ARG(1);
+
+        gf = get_guestfd(arg0);
+        if (!gf) {
+            errno = EBADF;
+            return set_swi_errno(env, -1);
+        }
+
+        return guestfd_fns[gf->type].seekfn(cpu, gf, arg1);
+    case TARGET_SYS_FLEN:
+        GET_ARG(0);
+
+        gf = get_guestfd(arg0);
+        if (!gf) {
+            errno = EBADF;
+            return set_swi_errno(env, -1);
+        }
+
+        return guestfd_fns[gf->type].flenfn(cpu, gf);
+    case TARGET_SYS_TMPNAM:
+        qemu_log_mask(LOG_UNIMP, "%s: SYS_TMPNAM not implemented", __func__);
+        return -1;
+    case TARGET_SYS_REMOVE:
+        GET_ARG(0);
+        GET_ARG(1);
+        if (use_gdb_syscalls()) {
+            ret = riscv_gdb_syscall(cpu, riscv_semi_cb, "unlink,%s",
+                                  arg0, (int)arg1 + 1);
+        } else {
+            s = lock_user_string(arg0);
+            if (!s) {
+                errno = EFAULT;
+                return set_swi_errno(env, -1);
+            }
+            ret =  set_swi_errno(env, remove(s));
+            unlock_user(s, arg0, 0);
+        }
+        return ret;
+    case TARGET_SYS_RENAME:
+        GET_ARG(0);
+        GET_ARG(1);
+        GET_ARG(2);
+        GET_ARG(3);
+        if (use_gdb_syscalls()) {
+            return riscv_gdb_syscall(cpu, riscv_semi_cb, "rename,%s,%s",
+                                   arg0, (int)arg1 + 1, arg2, (int)arg3 + 1);
+        } else {
+            char *s2;
+            s = lock_user_string(arg0);
+            s2 = lock_user_string(arg2);
+            if (!s || !s2) {
+                errno = EFAULT;
+                ret = set_swi_errno(env, -1);
+            } else {
+                ret = set_swi_errno(env, rename(s, s2));
+            }
+            if (s2) {
+                unlock_user(s2, arg2, 0);
+            }
+            if (s) {
+                unlock_user(s, arg0, 0);
+            }
+            return ret;
+        }
+    case TARGET_SYS_CLOCK:
+        return clock() / (CLOCKS_PER_SEC / 100);
+    case TARGET_SYS_TIME:
+        return set_swi_errno(env, time(NULL));
+    case TARGET_SYS_SYSTEM:
+        GET_ARG(0);
+        GET_ARG(1);
+        if (use_gdb_syscalls()) {
+            return riscv_gdb_syscall(cpu, riscv_semi_cb, "system,%s",
+                                   arg0, (int)arg1 + 1);
+        } else {
+            s = lock_user_string(arg0);
+            if (!s) {
+                errno = EFAULT;
+                return set_swi_errno(env, -1);
+            }
+            ret = set_swi_errno(env, system(s));
+            unlock_user(s, arg0, 0);
+            return ret;
+        }
+    case TARGET_SYS_ERRNO:
+        return get_swi_errno(env);
+    case TARGET_SYS_GET_CMDLINE:
+        {
+            /*
+             * Build a command-line from the original argv.
+             *
+             * The inputs are:
+             *     * arg0, pointer to a buffer of at least the size
+             *               specified in arg1.
+             *     * arg1, size of the buffer pointed to by arg0 in
+             *               bytes.
+             *
+             * The outputs are:
+             *     * arg0, pointer to null-terminated string of the
+             *               command line.
+             *     * arg1, length of the string pointed to by arg0.
+             */
+
+            char *output_buffer;
+            size_t input_size;
+            size_t output_size;
+            int status = 0;
+#if !defined(CONFIG_USER_ONLY)
+            const char *cmdline;
+#else
+            TaskState *ts = cs->opaque;
+#endif
+            GET_ARG(0);
+            GET_ARG(1);
+            input_size = arg1;
+            /* Compute the size of the output string.  */
+#if !defined(CONFIG_USER_ONLY)
+            cmdline = semihosting_get_cmdline();
+            if (cmdline == NULL) {
+                cmdline = ""; /* Default to an empty line. */
+            }
+            output_size = strlen(cmdline) + 1; /* Count terminating 0. */
+#else
+            unsigned int i;
+
+            output_size = ts->info->arg_end - ts->info->arg_start;
+            if (!output_size) {
+                /*
+                 * We special-case the "empty command line" case (argc==0).
+                 * Just provide the terminating 0.
+                 */
+                output_size = 1;
+            }
+#endif
+
+            if (output_size > input_size) {
+                /* Not enough space to store command-line arguments.  */
+                errno = E2BIG;
+                return set_swi_errno(env, -1);
+            }
+
+            /* Adjust the command-line length.  */
+            if (SET_ARG(1, output_size - 1)) {
+                /* Couldn't write back to argument block */
+                errno = EFAULT;
+                return set_swi_errno(env, -1);
+            }
+
+            /* Lock the buffer on the RISC-V side.  */
+            output_buffer = lock_user(VERIFY_WRITE, arg0, output_size, 0);
+            if (!output_buffer) {
+                errno = EFAULT;
+                return set_swi_errno(env, -1);
+            }
+
+            /* Copy the command-line arguments.  */
+#if !defined(CONFIG_USER_ONLY)
+            pstrcpy(output_buffer, output_size, cmdline);
+#else
+            if (output_size == 1) {
+                /* Empty command-line.  */
+                output_buffer[0] = '\0';
+                goto out;
+            }
+
+            if (copy_from_user(output_buffer, ts->info->arg_start,
+                               output_size)) {
+                errno = EFAULT;
+                status = set_swi_errno(env, -1);
+                goto out;
+            }
+
+            /* Separate arguments by white spaces.  */
+            for (i = 0; i < output_size - 1; i++) {
+                if (output_buffer[i] == 0) {
+                    output_buffer[i] = ' ';
+                }
+            }
+        out:
+#endif
+            /* Unlock the buffer on the RISC-V side.  */
+            unlock_user(output_buffer, arg0, output_size);
+
+            return status;
+        }
+    case TARGET_SYS_HEAPINFO:
+        {
+            target_ulong retvals[4];
+            target_ulong limit;
+            int i;
+#ifdef CONFIG_USER_ONLY
+            TaskState *ts = cs->opaque;
+#endif
+
+            GET_ARG(0);
+
+#ifdef CONFIG_USER_ONLY
+            /*
+             * Some C libraries assume the heap immediately follows .bss, so
+             * allocate it using sbrk.
+             */
+            if (!ts->heap_limit) {
+                abi_ulong ret;
+
+                ts->heap_base = do_brk(0);
+                limit = ts->heap_base + RISCV_HEAP_SIZE;
+                /* Try a big heap, and reduce the size if that fails.  */
+                for (;;) {
+                    ret = do_brk(limit);
+                    if (ret >= limit) {
+                        break;
+                    }
+                    limit = (ts->heap_base >> 1) + (limit >> 1);
+                }
+                ts->heap_limit = limit;
+            }
+
+            retvals[0] = ts->heap_base;
+            retvals[1] = ts->heap_limit;
+            retvals[2] = ts->stack_base;
+            retvals[3] = 0; /* Stack limit.  */
+#else
+            limit = ram_size;
+            /* TODO: Make this use the limit of the loaded application.  */
+            retvals[0] = limit / 2;
+            retvals[1] = limit;
+            retvals[2] = limit; /* Stack base */
+            retvals[3] = 0; /* Stack limit.  */
+#endif
+
+            for (i = 0; i < ARRAY_SIZE(retvals); i++) {
+                bool fail;
+
+                fail = put_user_ual(retvals[i],
+                                    arg0 + i * sizeof(target_ulong));
+
+                if (fail) {
+                    /* Couldn't write back to argument block */
+                    errno = EFAULT;
+                    return set_swi_errno(env, -1);
+                }
+            }
+            return 0;
+        }
+    case TARGET_SYS_EXIT:
+    case TARGET_SYS_EXIT_EXTENDED:
+        if (nr == TARGET_SYS_EXIT_EXTENDED || sizeof(target_ulong) == 0) {
+            /*
+             * The A64 version of SYS_EXIT takes a parameter block,
+             * so the application-exit type can return a subcode which
+             * is the exit status code from the application.
+             * SYS_EXIT_EXTENDED is an a new-in-v2.0 optional function
+             * which allows A32/T32 guests to also provide a status code.
+             */
+            GET_ARG(0);
+            GET_ARG(1);
+
+            if (arg0 == ADP_Stopped_ApplicationExit) {
+                ret = arg1;
+            } else {
+                ret = 1;
+            }
+        } else {
+            /*
+             * The A32/T32 version of SYS_EXIT specifies only
+             * Stopped_ApplicationExit as normal exit, but does not
+             * allow the guest to specify the exit status code.
+             * Everything else is considered an error.
+             */
+            ret = (args == ADP_Stopped_ApplicationExit) ? 0 : 1;
+        }
+        gdb_exit(env, ret);
+        exit(ret);
+    case TARGET_SYS_SYNCCACHE:
+        /*
+         * Clean the D-cache and invalidate the I-cache for the specified
+         * virtual address range. This is a nop for us since we don't
+         * implement caches. This is only present on A64.
+         */
+        if (sizeof(target_ulong) == 8) {
+            return 0;
+        }
+        /* fall through -- invalid for A32/T32 */
+    default:
+        fprintf(stderr, "qemu: Unsupported SemiHosting SWI 0x%02x\n", nr);
+        cpu_dump_state(cs, stderr, 0);
+        abort();
+    }
+}
diff --git a/target/riscv/translate.c b/target/riscv/translate.c
index 14dc71156b..4ee0d75f86 100644
--- a/target/riscv/translate.c
+++ b/target/riscv/translate.c
@@ -55,6 +55,7 @@  typedef struct DisasContext {
        to reset this known value.  */
     int frm;
     bool ext_ifencei;
+    CPUState *cs;
 } DisasContext;
 
 #ifdef TARGET_RISCV64
@@ -698,6 +699,15 @@  static bool gen_shift(DisasContext *ctx, arg_r *a,
     return true;
 }
 
+static uint32_t opcode_at(DisasContextBase *dcbase, target_ulong pc)
+{
+    DisasContext *ctx = container_of(dcbase, DisasContext, base);
+    CPUState *cpu = ctx->cs;
+    CPURISCVState *env = cpu->env_ptr;
+
+    return cpu_ldl_code(env, pc);
+}
+
 /* Include insn module translation function */
 #include "insn_trans/trans_rvi.inc.c"
 #include "insn_trans/trans_rvm.inc.c"
@@ -743,6 +753,7 @@  static void riscv_tr_init_disas_context(DisasContextBase *dcbase, CPUState *cs)
     ctx->misa = env->misa;
     ctx->frm = -1;  /* unknown rounding mode */
     ctx->ext_ifencei = cpu->cfg.ext_ifencei;
+    ctx->cs = cs;
 }
 
 static void riscv_tr_tb_start(DisasContextBase *db, CPUState *cpu)