[PATCH] gdbserver: Remove srv_linux_usrregs setting in RISCV

[PATCH] gdbserver: Remove srv_linux_usrregs setting in RISCV

[Youling Tang]

In the linux kernel code of RISCV, ptrace does not have the implementation
of PTRACE_POKEUSR/PTRACE_PEEKUSR in the arch_ptrace function (like aarch64),
If srv_linux_usrregs is set to yes, HAVE_LINUX_USRREGS will be defined, so
remove the srv_linux_usrregs setting.

Signed-off-by: Youling Tang <tangyouling@loongson.cn>
---
 gdbserver/configure.srv | 1 -
 1 file changed, 1 deletion(-)

diff --git a/gdbserver/configure.srv b/gdbserver/configure.srv
index d37053628fc..6840c3964cf 100644
--- a/gdbserver/configure.srv
+++ b/gdbserver/configure.srv
@@ -257,7 +257,6 @@ case "${gdbserver_host}" in
 			srv_tgtobj="${srv_tgtobj} linux-riscv-low.o"
 			srv_tgtobj="${srv_tgtobj} ${srv_linux_obj}"
 			srv_linux_regsets=yes
-			srv_linux_usrregs=yes
 			srv_linux_thread_db=yes
 			;;
   s390*-*-linux*)	srv_regobj="s390-linux32.o"
-- 
2.20.1

Re: [PATCH] gdbserver: Remove srv_linux_usrregs setting in RISCV

[Tom Tromey]

>>>>> Youling Tang <tangyouling@loongson.cn> writes:

> In the linux kernel code of RISCV, ptrace does not have the implementation
> of PTRACE_POKEUSR/PTRACE_PEEKUSR in the arch_ptrace function (like aarch64),
> If srv_linux_usrregs is set to yes, HAVE_LINUX_USRREGS will be defined, so
> remove the srv_linux_usrregs setting.

Looks reasonable to me.
Thank you.

Tom

Re: [PATCH] gdbserver: Remove srv_linux_usrregs setting in RISCV

[Maciej W. Rozycki]

On Mon, 9 May 2022, Tom Tromey wrote:

> > In the linux kernel code of RISCV, ptrace does not have the implementation
> > of PTRACE_POKEUSR/PTRACE_PEEKUSR in the arch_ptrace function (like aarch64),
> > If srv_linux_usrregs is set to yes, HAVE_LINUX_USRREGS will be defined, so
> > remove the srv_linux_usrregs setting.
> 
> Looks reasonable to me.

 Yeah, it must have been an oversight.  Thanks for the fix!

  Maciej

Re: [PATCH] gdbserver: Remove srv_linux_usrregs setting in RISCV

[Palmer Dabbelt]

On Mon, 09 May 2022 07:17:18 PDT (-0700), tom@tromey.com wrote:
>>>>>> Youling Tang <tangyouling@loongson.cn> writes:
>
>> In the linux kernel code of RISCV, ptrace does not have the implementation
>> of PTRACE_POKEUSR/PTRACE_PEEKUSR in the arch_ptrace function (like aarch64),
>> If srv_linux_usrregs is set to yes, HAVE_LINUX_USRREGS will be defined, so
>> remove the srv_linux_usrregs setting.
>
> Looks reasonable to me.

For some reason Maciej's email is only showing up in my gmail, so I'm 
just pasting in

>  Yeah, it must have been an oversight.  Thanks for the fix!

IIRC we had some out-of-tree ptrace() implementations that were to the 
old style, but they never made it upstream.  I bet the GDB support just 
stuck around, but I don't see a reason to keep it.