[v5] nptl: Add <thread_pointer.h> for RISC-V

This will be required by the rseq extensible ABI implementation on all
Linux architectures exposing the '__rseq_size' and '__rseq_offset'
symbols to set the initial value of the 'cpu_id' field which can be used
by applications to test if rseq is available and registered. As long as
the symbols are exposed it is valid for an application to perform this
test even if rseq is not yet implemented in libc for this architecture.

Both code paths tested on a Visionfive 2 with Debian sid.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
---
Changes since v4:
- Revert to GCC/Clang version check, __has_builtin can't reliably detect
  if __builtin_thread_pointer is implemented in the backend
Changes since v3:
- Revert to inline asm for the fallback code
Changes since v2:
- The 'register' variable has to be global to generate the correct asm
- Reuse '__thread_self' from 'tls.h'
Changes since v1:
- Use __glibc_has_builtin instead of GCC version check
- Use 'register' variable instead of inline asm
---
Cc: Florian Weimer <fweimer@redhat.com>
Cc: Palmer Dabbelt <palmer@rivosinc.com>
Cc: Darius Rad <darius@bluespec.com>
Cc: Jeff Law <jeffreyalaw@gmail.com>
---
 sysdeps/riscv/nptl/thread_pointer.h | 40 +++++++++++++++++++++++++++++
 1 file changed, 40 insertions(+)
 create mode 100644 sysdeps/riscv/nptl/thread_pointer.h

On 2024-11-01 12:41, Michael Jeanson wrote:
> This will be required by the rseq extensible ABI implementation on all
> Linux architectures exposing the '__rseq_size' and '__rseq_offset'
> symbols to set the initial value of the 'cpu_id' field which can be used
> by applications to test if rseq is available and registered. As long as
> the symbols are exposed it is valid for an application to perform this
> test even if rseq is not yet implemented in libc for this architecture.
> 
> Both code paths tested on a Visionfive 2 with Debian sid.
> 
> Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
> ---
> Changes since v4:
> - Revert to GCC/Clang version check, __has_builtin can't reliably detect
>   if __builtin_thread_pointer is implemented in the backend
> Changes since v3:
> - Revert to inline asm for the fallback code
> Changes since v2:
> - The 'register' variable has to be global to generate the correct asm
> - Reuse '__thread_self' from 'tls.h'
> Changes since v1:
> - Use __glibc_has_builtin instead of GCC version check
> - Use 'register' variable instead of inline asm
> ---
> Cc: Florian Weimer <fweimer@redhat.com>
> Cc: Palmer Dabbelt <palmer@rivosinc.com>
> Cc: Darius Rad <darius@bluespec.com>
> Cc: Jeff Law <jeffreyalaw@gmail.com>
> ---
>  sysdeps/riscv/nptl/thread_pointer.h | 40 +++++++++++++++++++++++++++++
>  1 file changed, 40 insertions(+)
>  create mode 100644 sysdeps/riscv/nptl/thread_pointer.h
> 
> diff --git a/sysdeps/riscv/nptl/thread_pointer.h b/sysdeps/riscv/nptl/thread_pointer.h
> new file mode 100644
> index 0000000000..9e7a84a78b
> --- /dev/null
> +++ b/sysdeps/riscv/nptl/thread_pointer.h
> @@ -0,0 +1,40 @@
> +/* __thread_pointer definition.  riscv version.
> +   Copyright (C) 2021-2024 Free Software Foundation, Inc.
> +   This file is part of the GNU C Library.
> +
> +   The GNU C Library is free software; you can redistribute it and/or
> +   modify it under the terms of the GNU Lesser General Public
> +   License as published by the Free Software Foundation; either
> +   version 2.1 of the License, or (at your option) any later version.
> +
> +   The GNU C Library 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
> +   Lesser General Public License for more details.
> +
> +   You should have received a copy of the GNU Lesser General Public
> +   License along with the GNU C Library.  If not, see
> +   <https://www.gnu.org/licenses/>.  */
> +
> +#ifndef _SYS_THREAD_POINTER_H
> +#define _SYS_THREAD_POINTER_H
> +
> +#include <features.h>
> +
> +#if __GNUC_PREREQ (10, 3) || __glibc_clang_prereq (11, 0)
> +static inline void *
> +__thread_pointer (void)
> +{
> +  return __builtin_thread_pointer ();
> +}
> +#else
> +static inline void *
> +__thread_pointer (void)
> +{
> +  void *__thread_register;
> +  __asm__ ("mv %0, tp" : "=r" (__thread_register));
> +  return __thread_register;
> +}
> +#endif
> +
> +#endif /* _SYS_THREAD_POINTER_H */

Ping? I think we were in agreement on v4 and this is just a cleanup
of the builtin check. With a reviewed-by I could push this.

Thanks,

Michael

On Mon, 09 Dec 2024 08:33:43 PST (-0800), mjeanson@efficios.com wrote:
> On 2024-11-01 12:41, Michael Jeanson wrote:
>> This will be required by the rseq extensible ABI implementation on all
>> Linux architectures exposing the '__rseq_size' and '__rseq_offset'
>> symbols to set the initial value of the 'cpu_id' field which can be used
>> by applications to test if rseq is available and registered. As long as
>> the symbols are exposed it is valid for an application to perform this
>> test even if rseq is not yet implemented in libc for this architecture.
>>
>> Both code paths tested on a Visionfive 2 with Debian sid.
>>
>> Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
>> ---
>> Changes since v4:
>> - Revert to GCC/Clang version check, __has_builtin can't reliably detect
>>   if __builtin_thread_pointer is implemented in the backend
>> Changes since v3:
>> - Revert to inline asm for the fallback code
>> Changes since v2:
>> - The 'register' variable has to be global to generate the correct asm
>> - Reuse '__thread_self' from 'tls.h'
>> Changes since v1:
>> - Use __glibc_has_builtin instead of GCC version check
>> - Use 'register' variable instead of inline asm
>> ---
>> Cc: Florian Weimer <fweimer@redhat.com>
>> Cc: Palmer Dabbelt <palmer@rivosinc.com>
>> Cc: Darius Rad <darius@bluespec.com>
>> Cc: Jeff Law <jeffreyalaw@gmail.com>
>> ---
>>  sysdeps/riscv/nptl/thread_pointer.h | 40 +++++++++++++++++++++++++++++
>>  1 file changed, 40 insertions(+)
>>  create mode 100644 sysdeps/riscv/nptl/thread_pointer.h
>>
>> diff --git a/sysdeps/riscv/nptl/thread_pointer.h b/sysdeps/riscv/nptl/thread_pointer.h
>> new file mode 100644
>> index 0000000000..9e7a84a78b
>> --- /dev/null
>> +++ b/sysdeps/riscv/nptl/thread_pointer.h
>> @@ -0,0 +1,40 @@
>> +/* __thread_pointer definition.  riscv version.
>> +   Copyright (C) 2021-2024 Free Software Foundation, Inc.
>> +   This file is part of the GNU C Library.
>> +
>> +   The GNU C Library is free software; you can redistribute it and/or
>> +   modify it under the terms of the GNU Lesser General Public
>> +   License as published by the Free Software Foundation; either
>> +   version 2.1 of the License, or (at your option) any later version.
>> +
>> +   The GNU C Library 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
>> +   Lesser General Public License for more details.
>> +
>> +   You should have received a copy of the GNU Lesser General Public
>> +   License along with the GNU C Library.  If not, see
>> +   <https://www.gnu.org/licenses/>.  */
>> +
>> +#ifndef _SYS_THREAD_POINTER_H
>> +#define _SYS_THREAD_POINTER_H
>> +
>> +#include <features.h>
>> +
>> +#if __GNUC_PREREQ (10, 3) || __glibc_clang_prereq (11, 0)
>> +static inline void *
>> +__thread_pointer (void)
>> +{
>> +  return __builtin_thread_pointer ();
>> +}
>> +#else
>> +static inline void *
>> +__thread_pointer (void)
>> +{
>> +  void *__thread_register;
>> +  __asm__ ("mv %0, tp" : "=r" (__thread_register));
>> +  return __thread_register;
>> +}
>> +#endif
>> +
>> +#endif /* _SYS_THREAD_POINTER_H */
>
> Ping? I think we were in agreement on v4 and this is just a cleanup
> of the builtin check. With a reviewed-by I could push this.

Sorry, I thought this was already reviewed/ack'd

Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
Acked-by: Palmer Dabbelt <palmer@rivosinc.com>

> Thanks,
>
> Michael

On 2024-12-09 11:40, Palmer Dabbelt wrote:
>>
>> Ping? I think we were in agreement on v4 and this is just a cleanup
>> of the builtin check. With a reviewed-by I could push this.
> 
> Sorry, I thought this was already reviewed/ack'd
> 
> Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
> Acked-by: Palmer Dabbelt <palmer@rivosinc.com>

I pushed both patches, thanks!

Michael