[PATCH] [AArch64] Fix pseudo-register numbering in the presence of unexpected additional registers

[PATCH] [AArch64] Fix pseudo-register numbering in the presence of unexpected additional registers

[Luis Machado]

When using AArch64 GDB with the QEMU debugging stub (in user mode), we get
additional system registers that GDB doesn't particularly care about, so
it doesn't number those explicitly.

But given the pseudo-register numbers are above the number of real registers,
we need to setup/account for the real registers first before going ahead and
numbering the pseudo-registers.  This has to happen at the end of
aarch64_gdbarch_init, after the call to tdesc_use_registers, as that
updates the total number of real registers.

This is in preparation to supporting pointer authentication for bare metal
aarch64 (QEMU).
---
 gdb/aarch64-tdep.c | 15 +++++++++++++--
 1 file changed, 13 insertions(+), 2 deletions(-)

diff --git a/gdb/aarch64-tdep.c b/gdb/aarch64-tdep.c
index 7229b53838e..f94e810f904 100644
--- a/gdb/aarch64-tdep.c
+++ b/gdb/aarch64-tdep.c
@@ -3598,8 +3598,7 @@ aarch64_gdbarch_init (struct gdbarch_info info, struct gdbarch_list *arches)
   tdep->jb_elt_size = 8;
   tdep->vq = vq;
   tdep->pauth_reg_base = first_pauth_regnum;
-  tdep->ra_sign_state_regnum = (feature_pauth == NULL) ? -1
-				: ra_sign_state_offset + num_regs;
+  tdep->ra_sign_state_regnum = -1;
   tdep->mte_reg_base = first_mte_regnum;
   tdep->tls_regnum = tls_regnum;
 
@@ -3698,6 +3697,18 @@ aarch64_gdbarch_init (struct gdbarch_info info, struct gdbarch_list *arches)
 
   tdesc_use_registers (gdbarch, tdesc, std::move (tdesc_data));
 
+  /* Fetch the updated number of registers after we're done adding all
+     entries from features we don't explicitly care about.  This is the case
+     for bare metal debugging stubs that include a lot of system registers.  */
+  num_regs = gdbarch_num_regs (gdbarch);
+
+  /* With the number of real registers updated, setup the pseudo-registers and
+     record their numbers.  */
+
+  /* Pointer authentication pseudo-registers.  */
+  if (tdep->has_pauth ())
+    tdep->ra_sign_state_regnum = ra_sign_state_offset + num_regs;
+
   /* Add standard register aliases.  */
   for (i = 0; i < ARRAY_SIZE (aarch64_register_aliases); i++)
     user_reg_add (gdbarch, aarch64_register_aliases[i].name,
-- 
2.25.1

Re: [PATCH] [AArch64] Fix pseudo-register numbering in the presence of unexpected additional registers

[Luis Machado]

On 9/21/22 18:53, Luis Machado via Gdb-patches wrote:
> When using AArch64 GDB with the QEMU debugging stub (in user mode), we get
> additional system registers that GDB doesn't particularly care about, so
> it doesn't number those explicitly.
> 
> But given the pseudo-register numbers are above the number of real registers,
> we need to setup/account for the real registers first before going ahead and
> numbering the pseudo-registers.  This has to happen at the end of
> aarch64_gdbarch_init, after the call to tdesc_use_registers, as that
> updates the total number of real registers.
> 
> This is in preparation to supporting pointer authentication for bare metal
> aarch64 (QEMU).
> ---
>   gdb/aarch64-tdep.c | 15 +++++++++++++--
>   1 file changed, 13 insertions(+), 2 deletions(-)
> 
> diff --git a/gdb/aarch64-tdep.c b/gdb/aarch64-tdep.c
> index 7229b53838e..f94e810f904 100644
> --- a/gdb/aarch64-tdep.c
> +++ b/gdb/aarch64-tdep.c
> @@ -3598,8 +3598,7 @@ aarch64_gdbarch_init (struct gdbarch_info info, struct gdbarch_list *arches)
>     tdep->jb_elt_size = 8;
>     tdep->vq = vq;
>     tdep->pauth_reg_base = first_pauth_regnum;
> -  tdep->ra_sign_state_regnum = (feature_pauth == NULL) ? -1
> -				: ra_sign_state_offset + num_regs;
> +  tdep->ra_sign_state_regnum = -1;
>     tdep->mte_reg_base = first_mte_regnum;
>     tdep->tls_regnum = tls_regnum;
>   
> @@ -3698,6 +3697,18 @@ aarch64_gdbarch_init (struct gdbarch_info info, struct gdbarch_list *arches)
>   
>     tdesc_use_registers (gdbarch, tdesc, std::move (tdesc_data));
>   
> +  /* Fetch the updated number of registers after we're done adding all
> +     entries from features we don't explicitly care about.  This is the case
> +     for bare metal debugging stubs that include a lot of system registers.  */
> +  num_regs = gdbarch_num_regs (gdbarch);
> +
> +  /* With the number of real registers updated, setup the pseudo-registers and
> +     record their numbers.  */
> +
> +  /* Pointer authentication pseudo-registers.  */
> +  if (tdep->has_pauth ())
> +    tdep->ra_sign_state_regnum = ra_sign_state_offset + num_regs;
> +
>     /* Add standard register aliases.  */
>     for (i = 0; i < ARRAY_SIZE (aarch64_register_aliases); i++)
>       user_reg_add (gdbarch, aarch64_register_aliases[i].name,

Pushed now.

Re: [PATCH] [AArch64] Fix pseudo-register numbering in the presence of unexpected additional registers

[Luis Machado]

Hi,

On 10/3/22 14:16, Luis Machado wrote:
> On 9/21/22 18:53, Luis Machado via Gdb-patches wrote:
>> When using AArch64 GDB with the QEMU debugging stub (in user mode), we get
>> additional system registers that GDB doesn't particularly care about, so
>> it doesn't number those explicitly.
>>
>> But given the pseudo-register numbers are above the number of real registers,
>> we need to setup/account for the real registers first before going ahead and
>> numbering the pseudo-registers.  This has to happen at the end of
>> aarch64_gdbarch_init, after the call to tdesc_use_registers, as that
>> updates the total number of real registers.
>>
>> This is in preparation to supporting pointer authentication for bare metal
>> aarch64 (QEMU).
>> ---
>>   gdb/aarch64-tdep.c | 15 +++++++++++++--
>>   1 file changed, 13 insertions(+), 2 deletions(-)
>>
>> diff --git a/gdb/aarch64-tdep.c b/gdb/aarch64-tdep.c
>> index 7229b53838e..f94e810f904 100644
>> --- a/gdb/aarch64-tdep.c
>> +++ b/gdb/aarch64-tdep.c
>> @@ -3598,8 +3598,7 @@ aarch64_gdbarch_init (struct gdbarch_info info, struct gdbarch_list *arches)
>>     tdep->jb_elt_size = 8;
>>     tdep->vq = vq;
>>     tdep->pauth_reg_base = first_pauth_regnum;
>> -  tdep->ra_sign_state_regnum = (feature_pauth == NULL) ? -1
>> -                : ra_sign_state_offset + num_regs;
>> +  tdep->ra_sign_state_regnum = -1;
>>     tdep->mte_reg_base = first_mte_regnum;
>>     tdep->tls_regnum = tls_regnum;
>> @@ -3698,6 +3697,18 @@ aarch64_gdbarch_init (struct gdbarch_info info, struct gdbarch_list *arches)
>>     tdesc_use_registers (gdbarch, tdesc, std::move (tdesc_data));
>> +  /* Fetch the updated number of registers after we're done adding all
>> +     entries from features we don't explicitly care about.  This is the case
>> +     for bare metal debugging stubs that include a lot of system registers.  */
>> +  num_regs = gdbarch_num_regs (gdbarch);
>> +
>> +  /* With the number of real registers updated, setup the pseudo-registers and
>> +     record their numbers.  */
>> +
>> +  /* Pointer authentication pseudo-registers.  */
>> +  if (tdep->has_pauth ())
>> +    tdep->ra_sign_state_regnum = ra_sign_state_offset + num_regs;
>> +
>>     /* Add standard register aliases.  */
>>     for (i = 0; i < ARRAY_SIZE (aarch64_register_aliases); i++)
>>       user_reg_add (gdbarch, aarch64_register_aliases[i].name,
> 
> Pushed now.

I'd like to push this patch to both gdb 12 and gdb 11. Does that sound reasonable?