[Qemu-devel] [PATCH 0/2] target/arm: gdbstub system registers refinement

[Qemu-devel] [PATCH 0/2] target/arm: gdbstub system registers refinement

[Richard Henderson]

This was working toward fixing 

  https://bugs.linaro.org/show_bug.cgi?id=4274

although I don't think I'm entirely successful.  I would argue,
however, that gdb's behaviour in this is odd.  Why are registers
that are clearly marked "system" being added to "general" simply
because they have integer type?

In the meantime, we don't really need to present these registers
to the user for CONFIG_USER_ONLY.


r~


Richard Henderson (2):
  target/arm: Put system registers in "system" group
  target/arm: Don't add system-registers.xml for user-only

 target/arm/gdbstub.c | 2 +-
 target/arm/helper.c  | 4 ++++
 2 files changed, 5 insertions(+), 1 deletion(-)

-- 
2.17.2

[Qemu-devel] [PATCH 1/2] target/arm: Put system registers in "system" group

[Richard Henderson]

The system group is predefined within gdb.  While the spec
seems to allow arbitrary group names, that seems to mess up
the addition of all registers to any groups.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 target/arm/gdbstub.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/target/arm/gdbstub.c b/target/arm/gdbstub.c
index 0c64c0292e..67c8874b3b 100644
--- a/target/arm/gdbstub.c
+++ b/target/arm/gdbstub.c
@@ -113,7 +113,7 @@ static void arm_gen_one_xml_reg_tag(GString *s, DynamicGDBXMLInfo *dyn_xml,
 {
     g_string_append_printf(s, "<reg name=\"%s\"", ri->name);
     g_string_append_printf(s, " bitsize=\"%d\"", bitsize);
-    g_string_append_printf(s, " group=\"cp_regs\"/>");
+    g_string_append_printf(s, " group=\"system\"/>");
     dyn_xml->num_cpregs++;
     dyn_xml->cpregs_keys[dyn_xml->num_cpregs - 1] = ri_key;
 }
-- 
2.17.2

[Qemu-devel] [PATCH 2/2] target/arm: Don't add system-registers.xml for user-only

[Richard Henderson]

The system registers aren't that interesting for debugging
user binaries.  Avoid adding them in this case.

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

diff --git a/target/arm/helper.c b/target/arm/helper.c
index 2607d39ad1..ca5632d26f 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -221,6 +221,7 @@ static void write_raw_cp_reg(CPUARMState *env, const ARMCPRegInfo *ri,
     }
 }
 
+#ifndef CONFIG_USER_ONLY
 static int arm_gdb_get_sysreg(CPUARMState *env, uint8_t *buf, int reg)
 {
     ARMCPU *cpu = arm_env_get_cpu(env);
@@ -243,6 +244,7 @@ static int arm_gdb_set_sysreg(CPUARMState *env, uint8_t *buf, int reg)
 {
     return 0;
 }
+#endif
 
 static bool raw_accessors_invalid(const ARMCPRegInfo *ri)
 {
@@ -6694,9 +6696,11 @@ void arm_cpu_register_gdb_regs_for_features(ARMCPU *cpu)
         gdb_register_coprocessor(cs, vfp_gdb_get_reg, vfp_gdb_set_reg,
                                  19, "arm-vfp.xml", 0);
     }
+#ifndef CONFIG_USER_ONLY
     gdb_register_coprocessor(cs, arm_gdb_get_sysreg, arm_gdb_set_sysreg,
                              arm_gen_dynamic_xml(cs),
                              "system-registers.xml", 0);
+#endif
 }
 
 /* Sort alphabetically by type name, except for "any". */
-- 
2.17.2

Re: [Qemu-devel] [PATCH 0/2] target/arm: gdbstub system registers refinement

[Peter Maydell]

On Thu, 7 Mar 2019 at 20:12, Richard Henderson
<richard.henderson@linaro.org> wrote:
>
> This was working toward fixing
>
>   https://bugs.linaro.org/show_bug.cgi?id=4274
>
> although I don't think I'm entirely successful.  I would argue,
> however, that gdb's behaviour in this is odd.  Why are registers
> that are clearly marked "system" being added to "general" simply
> because they have integer type?

Yeah, this does seem to me like it might be a bug on
gdb's end -- are you following up with them?

> In the meantime, we don't really need to present these registers
> to the user for CONFIG_USER_ONLY.

Reasonable enough, I guess, though a few sysregs are
visible to EL0.

thanks
-- PMM

Re: [Qemu-devel] [PATCH 0/2] target/arm: gdbstub system registers refinement

[Richard Henderson]

On 3/8/19 7:59 AM, Peter Maydell wrote:
> On Thu, 7 Mar 2019 at 20:12, Richard Henderson
> <richard.henderson@linaro.org> wrote:
>>
>> This was working toward fixing
>>
>>   https://bugs.linaro.org/show_bug.cgi?id=4274
>>
>> although I don't think I'm entirely successful.  I would argue,
>> however, that gdb's behaviour in this is odd.  Why are registers
>> that are clearly marked "system" being added to "general" simply
>> because they have integer type?
> 
> Yeah, this does seem to me like it might be a bug on
> gdb's end -- are you following up with them?

Yes.


r~