[GIT PULL] RISC-V Fixes for 6.16-rc1

[GIT PULL] RISC-V Fixes for 6.16-rc1

[Alexandre Ghiti]

The following changes since commit 19272b37aa4f83ca52bdf9c16d5d81bdd1354494:

   Linux 6.16-rc1 (2025-06-08 13:44:43 -0700)

are available in the Git repository at:

   git://git.kernel.org/pub/scm/linux/kernel/git/alexghiti/linux 
tags/riscv-fixes-6.16-rc1

for you to fetch changes up to 0e7815e464847acbe5d9adfb351c68f822bbaa7e:

   RISC-V: vDSO: Correct inline assembly constraints in the getrandom 
syscall wrapper (2025-06-10 08:43:46 +0000)

----------------------------------------------------------------
riscv fixes for 6.16-rc1

- A fix for the newly introduced getrandom vdso where clang optimizes 
away a register variable which is both an input and an output parameter
- Another fix for vdso to prevent an error from llvm-readelf
- A fix for the runtime constant support which did not handle the zero 
case correctly
- A fix for theadvector where we did not save all the vector registers, 
only a few of them

----------------------------------------------------------------

Note that there are still important pending fixes that need reviews and 
tests:

- raid6 
(https://lore.kernel.org/linux-riscv/20250610101234.1100660-1-zhangchunyan@iscas.ac.cn/T/#t)
- initialization of vdso static data 
(https://lore.kernel.org/linux-riscv/20250522073327.246668-1-wangjingwei@iscas.ac.cn/)

So I may come with another PR this week.

----------------------------------------------------------------

Alexandre Ghiti (1):
       Merge patch "fix riscv runtime constant support"

Charles Mirabile (1):
       riscv: fix runtime constant support for nommu kernels

Fangrui Song (1):
       riscv: vdso: Exclude .rodata from the PT_DYNAMIC segment

Han Gao (1):
       riscv: vector: Fix context save/restore with xtheadvector

Xi Ruoyao (1):
       RISC-V: vDSO: Correct inline assembly constraints in the 
getrandom syscall wrapper

  arch/riscv/include/asm/runtime-const.h  |  2 +-
  arch/riscv/include/asm/vdso/getrandom.h |  2 +-
  arch/riscv/include/asm/vector.h         | 12 ++++++------
  arch/riscv/kernel/vdso/vdso.lds.S       |  2 +-
  4 files changed, 9 insertions(+), 9 deletions(-)


_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv

Re: [GIT PULL] RISC-V Fixes for 6.16-rc1

[Palmer Dabbelt]

On Tue, 10 Jun 2025 13:13:20 PDT (-0700), Alexandre Ghiti wrote:
> The following changes since commit 19272b37aa4f83ca52bdf9c16d5d81bdd1354494:
>
>    Linux 6.16-rc1 (2025-06-08 13:44:43 -0700)
>
> are available in the Git repository at:
>
>    git://git.kernel.org/pub/scm/linux/kernel/git/alexghiti/linux 
> tags/riscv-fixes-6.16-rc1
>
> for you to fetch changes up to 0e7815e464847acbe5d9adfb351c68f822bbaa7e:
>
>    RISC-V: vDSO: Correct inline assembly constraints in the getrandom 
> syscall wrapper (2025-06-10 08:43:46 +0000)
>
> ----------------------------------------------------------------
> riscv fixes for 6.16-rc1
>
> - A fix for the newly introduced getrandom vdso where clang optimizes 
> away a register variable which is both an input and an output parameter
> - Another fix for vdso to prevent an error from llvm-readelf
> - A fix for the runtime constant support which did not handle the zero 
> case correctly
> - A fix for theadvector where we did not save all the vector registers, 
> only a few of them

This one is exposing some of the workflow issues again: there's 
duplicate patches in here, and I let it hang out overnight to give 
linux-next a chance to send the "there's duplicate patches" email and 
didn't get anything.  I've had issues with gmail eating linux-next 
messages before, so sorry if I'm just missing them.  So not quite sure 
what's going on, but I think that's part of the problem.

There's also no "b4 ty"-type messages, which makes it hard to sort out 
the duplicates manually (or I'm missing them, one of your other messages 
didn't make patchwork).

> ----------------------------------------------------------------
>
> Note that there are still important pending fixes that need reviews and 
> tests:
>
> - raid6 
> (https://lore.kernel.org/linux-riscv/20250610101234.1100660-1-zhangchunyan@iscas.ac.cn/T/#t)

I'll just grab the that t0 fix -- I understand it enough to take it 
without a reproducer, I was going to yesterday but sort of just got 
tired.  I'll go post on the thread...

> - initialization of vdso static data 
> (https://lore.kernel.org/linux-riscv/20250522073327.246668-1-wangjingwei@iscas.ac.cn/)

There's a v3 with some commentary over here: 
https://lore.kernel.org/linux-riscv/mhng-FC7E1D2C-D4E1-490E-9363-508518B62FE5@palmerdabbelt-mac/

>
> So I may come with another PR this week.
>
> ----------------------------------------------------------------
>
> Alexandre Ghiti (1):
>        Merge patch "fix riscv runtime constant support"
>
> Charles Mirabile (1):
>        riscv: fix runtime constant support for nommu kernels

This also seems to be based on an arbitrary patch, which is kind of 
clunky.  It seems like you've got some mix of merging and rebasing going 
on, which is probably also part of why your PR from the last merge 
window had the rebase issue?

> Fangrui Song (1):
>        riscv: vdso: Exclude .rodata from the PT_DYNAMIC segment
>
> Han Gao (1):
>        riscv: vector: Fix context save/restore with xtheadvector
>
> Xi Ruoyao (1):
>        RISC-V: vDSO: Correct inline assembly constraints in the 
> getrandom syscall wrapper
>
>   arch/riscv/include/asm/runtime-const.h  |  2 +-
>   arch/riscv/include/asm/vdso/getrandom.h |  2 +-
>   arch/riscv/include/asm/vector.h         | 12 ++++++------
>   arch/riscv/kernel/vdso/vdso.lds.S       |  2 +-
>   4 files changed, 9 insertions(+), 9 deletions(-)
>
>
> _______________________________________________
> linux-riscv mailing list
> linux-riscv@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-riscv

So I just rebased this to drop the duplicates.  It's on the tester, 
should show up soon...

_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv

Re: [GIT PULL] RISC-V Fixes for 6.16-rc1

[Alexandre Ghiti]

On 6/12/25 21:19, Palmer Dabbelt wrote:
> On Tue, 10 Jun 2025 13:13:20 PDT (-0700), Alexandre Ghiti wrote:
>> The following changes since commit 
>> 19272b37aa4f83ca52bdf9c16d5d81bdd1354494:
>>
>>    Linux 6.16-rc1 (2025-06-08 13:44:43 -0700)
>>
>> are available in the Git repository at:
>>
>>    git://git.kernel.org/pub/scm/linux/kernel/git/alexghiti/linux 
>> tags/riscv-fixes-6.16-rc1
>>
>> for you to fetch changes up to 0e7815e464847acbe5d9adfb351c68f822bbaa7e:
>>
>>    RISC-V: vDSO: Correct inline assembly constraints in the getrandom 
>> syscall wrapper (2025-06-10 08:43:46 +0000)
>>
>> ----------------------------------------------------------------
>> riscv fixes for 6.16-rc1
>>
>> - A fix for the newly introduced getrandom vdso where clang optimizes 
>> away a register variable which is both an input and an output parameter
>> - Another fix for vdso to prevent an error from llvm-readelf
>> - A fix for the runtime constant support which did not handle the 
>> zero case correctly
>> - A fix for theadvector where we did not save all the vector 
>> registers, only a few of them
>
> This one is exposing some of the workflow issues again: there's 
> duplicate patches in here, and I let it hang out overnight to give 
> linux-next a chance to send the "there's duplicate patches" email and 
> didn't get anything.  I've had issues with gmail eating linux-next 
> messages before, so sorry if I'm just missing them.  So not quite sure 
> what's going on, but I think that's part of the problem.


I don't see duplicate patches though, how do you see them? What causes them?


>
> There's also no "b4 ty"-type messages, which makes it hard to sort out 
> the duplicates manually (or I'm missing them, one of your other 
> messages didn't make patchwork).


I did not setup b4 ty yet, I will.


>
>> ----------------------------------------------------------------
>>
>> Note that there are still important pending fixes that need reviews 
>> and tests:
>>
>> - raid6 
>> (https://lore.kernel.org/linux-riscv/20250610101234.1100660-1-zhangchunyan@iscas.ac.cn/T/#t)
>
> I'll just grab the that t0 fix -- I understand it enough to take it 
> without a reproducer, I was going to yesterday but sort of just got 
> tired.  I'll go post on the thread...
>
>> - initialization of vdso static data 
>> (https://lore.kernel.org/linux-riscv/20250522073327.246668-1-wangjingwei@iscas.ac.cn/)
>
> There's a v3 with some commentary over here: 
> https://lore.kernel.org/linux-riscv/mhng-FC7E1D2C-D4E1-490E-9363-508518B62FE5@palmerdabbelt-mac/
>
>>
>> So I may come with another PR this week.
>>
>> ----------------------------------------------------------------
>>
>> Alexandre Ghiti (1):
>>        Merge patch "fix riscv runtime constant support"
>>
>> Charles Mirabile (1):
>>        riscv: fix runtime constant support for nommu kernels
>
> This also seems to be based on an arbitrary patch, which is kind of 
> clunky.  It seems like you've got some mix of merging and rebasing 
> going on, which is probably also part of why your PR from the last 
> merge window had the rebase issue?


Ok, I need to address how I merge patches, I guess I should have changed 
the merge base: how do you proceed?

Regarding the previous PR, the problem was that the second PR depended 
on patches that were merged in -rc6 and the first PR was sent on -rc3, 
so what should be done here?

1. As I proposed on irc, rebase for-next on -rc6 and apply both PRs (the 
first was not merged yet) and this is why I rebased the second PR on 
-rc6. I also remember we talked about rebasing for-next on each -rcX, we 
already had the problem where for-next was broken because it was based 
on -rcM and we needed a fix from another tree from -rcN.
2. A back-merge of -rc6 into the second PR based on -rc3? Sounds wrong 
but I'll give it a try just to see how it looks.
3. Other solution?


>
>> Fangrui Song (1):
>>        riscv: vdso: Exclude .rodata from the PT_DYNAMIC segment
>>
>> Han Gao (1):
>>        riscv: vector: Fix context save/restore with xtheadvector
>>
>> Xi Ruoyao (1):
>>        RISC-V: vDSO: Correct inline assembly constraints in the 
>> getrandom syscall wrapper
>>
>>   arch/riscv/include/asm/runtime-const.h  |  2 +-
>>   arch/riscv/include/asm/vdso/getrandom.h |  2 +-
>>   arch/riscv/include/asm/vector.h         | 12 ++++++------
>>   arch/riscv/kernel/vdso/vdso.lds.S       |  2 +-
>>   4 files changed, 9 insertions(+), 9 deletions(-)
>>
>>
>> _______________________________________________
>> linux-riscv mailing list
>> linux-riscv@lists.infradead.org
>> http://lists.infradead.org/mailman/listinfo/linux-riscv
>
> So I just rebased this to drop the duplicates.  It's on the tester, 
> should show up soon...
>
> _______________________________________________
> linux-riscv mailing list
> linux-riscv@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-riscv

_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv