[PATCH v2 0/2] target/m68k: Reject immediate as destination

[PATCH v2 0/2] target/m68k: Reject immediate as destination

[Richard Henderson]

The translate.c fix is extracted from a larger patch set.
I then add a test case and fix cpu_loop to raise the proper signal.


r~


Richard Henderson (2):
  target/m68k: Reject immediate as destination in gen_ea_mode
  linux-user/m68k: Handle EXCP_ADDRESS in cpu_loop

 linux-user/m68k/cpu_loop.c     |  5 ++++-
 target/m68k/translate.c        |  4 ++++
 tests/tcg/m68k/excp-address.c  | 32 ++++++++++++++++++++++++++++++++
 tests/tcg/m68k/Makefile.target |  1 +
 4 files changed, 41 insertions(+), 1 deletion(-)
 create mode 100644 tests/tcg/m68k/excp-address.c

-- 
2.34.1

[PATCH v2 1/2] target/m68k: Reject immediate as destination in gen_ea_mode

[Richard Henderson]

In theory this should never happen, as all such instructions
are illegal.  This is checked in e.g. gen_lea_mode and
gen_ea_mode_fp but not here.  In case something higher up
isn't checking modes properly, return NULL_QREG.  This will
result in an illegal instruction exception being raised.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 target/m68k/translate.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/target/m68k/translate.c b/target/m68k/translate.c
index 422f4652f1..e16c608ef8 100644
--- a/target/m68k/translate.c
+++ b/target/m68k/translate.c
@@ -894,6 +894,10 @@ static TCGv gen_ea_mode(CPUM68KState *env, DisasContext *s, int mode, int reg0,
         case 3: /* pc index+displacement.  */
             goto do_indirect;
         case 4: /* Immediate.  */
+            /* Should never be used for an output or RMW input. */
+            if (what == EA_STORE || addrp) {
+                return NULL_QREG;
+            }
             /* Sign extend values for consistency.  */
             switch (opsize) {
             case OS_BYTE:
-- 
2.34.1

Re: [PATCH v2 1/2] target/m68k: Reject immediate as destination in gen_ea_mode

[Laurent Vivier]

Le 09/03/2023 à 21:16, Richard Henderson a écrit :
> In theory this should never happen, as all such instructions
> are illegal.  This is checked in e.g. gen_lea_mode and
> gen_ea_mode_fp but not here.  In case something higher up
> isn't checking modes properly, return NULL_QREG.  This will
> result in an illegal instruction exception being raised.
> 
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> ---
>   target/m68k/translate.c | 4 ++++
>   1 file changed, 4 insertions(+)
> 
> diff --git a/target/m68k/translate.c b/target/m68k/translate.c
> index 422f4652f1..e16c608ef8 100644
> --- a/target/m68k/translate.c
> +++ b/target/m68k/translate.c
> @@ -894,6 +894,10 @@ static TCGv gen_ea_mode(CPUM68KState *env, DisasContext *s, int mode, int reg0,
>           case 3: /* pc index+displacement.  */
>               goto do_indirect;
>           case 4: /* Immediate.  */
> +            /* Should never be used for an output or RMW input. */
> +            if (what == EA_STORE || addrp) {
> +                return NULL_QREG;
> +            }
>               /* Sign extend values for consistency.  */
>               switch (opsize) {
>               case OS_BYTE:

Reviewed-by: Laurent Vivier <laurent@vivier.eu>

Re: [PATCH v2 1/2] target/m68k: Reject immediate as destination in gen_ea_mode

[Laurent Vivier]

Le 09/03/2023 à 21:16, Richard Henderson a écrit :
> In theory this should never happen, as all such instructions
> are illegal.  This is checked in e.g. gen_lea_mode and
> gen_ea_mode_fp but not here.  In case something higher up
> isn't checking modes properly, return NULL_QREG.  This will
> result in an illegal instruction exception being raised.
> 
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> ---
>   target/m68k/translate.c | 4 ++++
>   1 file changed, 4 insertions(+)
> 
> diff --git a/target/m68k/translate.c b/target/m68k/translate.c
> index 422f4652f1..e16c608ef8 100644
> --- a/target/m68k/translate.c
> +++ b/target/m68k/translate.c
> @@ -894,6 +894,10 @@ static TCGv gen_ea_mode(CPUM68KState *env, DisasContext *s, int mode, int reg0,
>           case 3: /* pc index+displacement.  */
>               goto do_indirect;
>           case 4: /* Immediate.  */
> +            /* Should never be used for an output or RMW input. */
> +            if (what == EA_STORE || addrp) {
> +                return NULL_QREG;
> +            }
>               /* Sign extend values for consistency.  */
>               switch (opsize) {
>               case OS_BYTE:

Reviewed-by: Laurent Vivier <laurent@vivier.eu>

[PATCH v2 2/2] linux-user/m68k: Handle EXCP_ADDRESS in cpu_loop

[Richard Henderson]

This exception can be raised by illegal instructions.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 linux-user/m68k/cpu_loop.c     |  5 ++++-
 tests/tcg/m68k/excp-address.c  | 32 ++++++++++++++++++++++++++++++++
 tests/tcg/m68k/Makefile.target |  1 +
 3 files changed, 37 insertions(+), 1 deletion(-)
 create mode 100644 tests/tcg/m68k/excp-address.c

diff --git a/linux-user/m68k/cpu_loop.c b/linux-user/m68k/cpu_loop.c
index caead1cb74..84b5d290c0 100644
--- a/linux-user/m68k/cpu_loop.c
+++ b/linux-user/m68k/cpu_loop.c
@@ -35,7 +35,10 @@ void cpu_loop(CPUM68KState *env)
         cpu_exec_end(cs);
         process_queued_cpu_work(cs);
 
-        switch(trapnr) {
+        switch (trapnr) {
+        case EXCP_ADDRESS:
+            force_sig_fault(TARGET_SIGBUS, TARGET_BUS_ADRALN, env->pc);
+            break;
         case EXCP_ILLEGAL:
         case EXCP_LINEA:
         case EXCP_LINEF:
diff --git a/tests/tcg/m68k/excp-address.c b/tests/tcg/m68k/excp-address.c
new file mode 100644
index 0000000000..1dbdddef58
--- /dev/null
+++ b/tests/tcg/m68k/excp-address.c
@@ -0,0 +1,32 @@
+/*
+ * Test m68k address exception
+ */
+
+#define _GNU_SOURCE 1
+#include <signal.h>
+#include <stdlib.h>
+
+static void sig_handler(int sig, siginfo_t *si, void *puc)
+{
+    exit(0);
+}
+
+int main(int argc, char **argv)
+{
+    struct sigaction act = {
+        .sa_sigaction = sig_handler,
+        .sa_flags = SA_SIGINFO
+    };
+
+    sigaction(SIGBUS, &act, NULL);
+
+    /*
+     * addl %d0,#0 -- with immediate as destination is illegal.
+     * Buggy qemu interpreted the insn as 5 words: 2 for immediate source
+     * and another 2 for immediate destination.  Provide all that padding
+     * so that abort gets called.
+     */
+    asm volatile(".word 0xd1bc,0,0,0,0");
+
+    abort();
+}
diff --git a/tests/tcg/m68k/Makefile.target b/tests/tcg/m68k/Makefile.target
index 1163c7ef03..d3b0bc48a3 100644
--- a/tests/tcg/m68k/Makefile.target
+++ b/tests/tcg/m68k/Makefile.target
@@ -4,6 +4,7 @@
 #
 
 VPATH += $(SRC_PATH)/tests/tcg/m68k
+TESTS += excp-address
 TESTS += trap
 
 # On m68k Linux supports 4k and 8k pages (but 8k is currently broken)
-- 
2.34.1

Re: [PATCH v2 2/2] linux-user/m68k: Handle EXCP_ADDRESS in cpu_loop

[Laurent Vivier]

Le 09/03/2023 à 21:16, Richard Henderson a écrit :
> This exception can be raised by illegal instructions.
> 
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> ---
>   linux-user/m68k/cpu_loop.c     |  5 ++++-
>   tests/tcg/m68k/excp-address.c  | 32 ++++++++++++++++++++++++++++++++
>   tests/tcg/m68k/Makefile.target |  1 +
>   3 files changed, 37 insertions(+), 1 deletion(-)
>   create mode 100644 tests/tcg/m68k/excp-address.c
> 
> diff --git a/linux-user/m68k/cpu_loop.c b/linux-user/m68k/cpu_loop.c
> index caead1cb74..84b5d290c0 100644
> --- a/linux-user/m68k/cpu_loop.c
> +++ b/linux-user/m68k/cpu_loop.c
> @@ -35,7 +35,10 @@ void cpu_loop(CPUM68KState *env)
>           cpu_exec_end(cs);
>           process_queued_cpu_work(cs);
>   
> -        switch(trapnr) {
> +        switch (trapnr) {
> +        case EXCP_ADDRESS:
> +            force_sig_fault(TARGET_SIGBUS, TARGET_BUS_ADRALN, env->pc);
> +            break;
>           case EXCP_ILLEGAL:
>           case EXCP_LINEA:
>           case EXCP_LINEF:
> diff --git a/tests/tcg/m68k/excp-address.c b/tests/tcg/m68k/excp-address.c
> new file mode 100644
> index 0000000000..1dbdddef58
> --- /dev/null
> +++ b/tests/tcg/m68k/excp-address.c
> @@ -0,0 +1,32 @@
> +/*
> + * Test m68k address exception
> + */
> +
> +#define _GNU_SOURCE 1
> +#include <signal.h>
> +#include <stdlib.h>
> +
> +static void sig_handler(int sig, siginfo_t *si, void *puc)
> +{
> +    exit(0);
> +}
> +
> +int main(int argc, char **argv)
> +{
> +    struct sigaction act = {
> +        .sa_sigaction = sig_handler,
> +        .sa_flags = SA_SIGINFO
> +    };
> +
> +    sigaction(SIGBUS, &act, NULL);
> +
> +    /*
> +     * addl %d0,#0 -- with immediate as destination is illegal.
> +     * Buggy qemu interpreted the insn as 5 words: 2 for immediate source
> +     * and another 2 for immediate destination.  Provide all that padding
> +     * so that abort gets called.
> +     */
> +    asm volatile(".word 0xd1bc,0,0,0,0");
> +
> +    abort();
> +}
> diff --git a/tests/tcg/m68k/Makefile.target b/tests/tcg/m68k/Makefile.target
> index 1163c7ef03..d3b0bc48a3 100644
> --- a/tests/tcg/m68k/Makefile.target
> +++ b/tests/tcg/m68k/Makefile.target
> @@ -4,6 +4,7 @@
>   #
>   
>   VPATH += $(SRC_PATH)/tests/tcg/m68k
> +TESTS += excp-address
>   TESTS += trap
>   
>   # On m68k Linux supports 4k and 8k pages (but 8k is currently broken)

Reviewed-by: Laurent Vivier <laurent@vivier.eu>

I've tested tests/tcg/m68k/excp-address.c on a real hardware (Q800), and the result differs from the 
one from QEMU:

On Q800 (etch m68k, kernel 5.14.0):

$ ./excp-address
Illegal instruction
$ strace ./excp-address
...
rt_sigaction(SIGBUS, {0x80000478, [], SA_SIGINFO}, NULL, 8) = 0
--- SIGILL (Illegal instruction) @ 0 (0) ---
+++ killed by SIGILL +++

With QEMU, we have:

# QEMU_STRACE= ./excp-address
...
677354 rt_sigaction(SIGBUS,0x40800454,NULL) = 0
--- SIGBUS {si_signo=SIGBUS, si_code=1, si_addr=0x800004ce} ---
677354 exit_group(0)

Thanks,
Laurent

Re: [PATCH v2 2/2] linux-user/m68k: Handle EXCP_ADDRESS in cpu_loop

[Richard Henderson]

On 3/15/23 11:16, Laurent Vivier wrote:
> I've tested tests/tcg/m68k/excp-address.c on a real hardware (Q800), and the result 
> differs from the one from QEMU:
> 
> On Q800 (etch m68k, kernel 5.14.0):
> 
> $ ./excp-address
> Illegal instruction
> $ strace ./excp-address
> ...
> rt_sigaction(SIGBUS, {0x80000478, [], SA_SIGINFO}, NULL, 8) = 0
> --- SIGILL (Illegal instruction) @ 0 (0) ---
> +++ killed by SIGILL +++

Ok, that suggests that we need to do something different in the translator in patch 1. 
I'll give it some thought.


r~