Re: [PATCH v2] riscv: perf: Drop defining `perf_instruction_pointer()` and `perf_misc_flags()`

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Tue, Nov 19, 2024 at 08:52:28AM +0100, Ingo Molnar wrote:
> 
> * Prabhakar <prabhakar.mahadev-lad.rj@xxxxxxxxxxxxxx> wrote:
> 
> > From: Lad Prabhakar <prabhakar.mahadev-lad.rj@xxxxxxxxxxxxxx>
> > 
> > With commit 2c47e7a74f44 ("perf/core: Correct perf sampling with guest
> > VMs"), the perf core now handles the functionality previously requiring
> > arch-specific definitions of `perf_instruction_pointer()` and
> > `perf_misc_flags()`. As these definitions are no longer necessary for
> > RISC-V, this patch removes their implementation and declarations.
> > 
> > This change also fixes the following build issue on RISC-V:
> > 
> >     ./include/linux/perf_event.h:1679:84: error: macro "perf_misc_flags" passed 2 arguments, but takes just 1
> >     ./include/linux/perf_event.h:1679:22: error: 'perf_misc_flags' redeclared as different kind of symbol
> >     ./include/linux/perf_event.h:1680:22: error: conflicting types for 'perf_instruction_pointer'; have 'long unsigned int(struct perf_event *, struct pt_regs *)'
> > 
> > The above errors arise from conflicts between the core definitions in
> > `linux/perf_event.h` and the RISC-V-specific definitions in
> > `arch/riscv/include/asm/perf_event.h`. Removing the RISC-V-specific
> > definitions resolves these issues and aligns the architecture with the
> > updated perf core.
> > 
> > Fixes: 2c47e7a74f44 ("perf/core: Correct perf sampling with guest VMs")
> 
> Yeah, so the Fixes tag is wrong - this is not a build bug
> with that commit, and your patch does not even apply to
> the perf events tree.
> 
> This is a semantic merge conflict that arises in linux-next - the
> riscv version of perf_instruction_pointer() function doesn't even
> exist in the perf tree...
> 
> AFAICS the problem is that the riscv tree applied this commit:
> 
>   5bb5ccb3e8d8 ("riscv: perf: add guest vs host distinction")
> 
> While the perf tree solved this in a more generic fashion:
> 
>   2c47e7a74f44 perf/core: Correct perf sampling with guest VMs
>   baff01f3d75f perf/x86: Refactor misc flag assignments
>   3e807cf07d96 perf/powerpc: Use perf_arch_instruction_pointer()
>   04782e63917d perf/core: Hoist perf_instruction_pointer() and perf_misc_flags()
>   e33ed362cf9e perf/arm: Drop unused functions
> 
> So I believe, assuming the perf version works fine on riscv
> (I haven't tested it), that the solution is to revert
> 5bb5ccb3e8d8 either in the riscv tree, or upon merging it.

Hi Ingo,

This patch isn't a complete revert of 5bb5ccb3e8d8. The early returns in
perf_callchain_user() and perf_callchain_kernel() should remain and the
CONFIG_PERF_EVENTS #ifdef in arch/riscv/include/asm/perf_event.h should
remain as well.

Thanks,
drew




[Index of Archives]     [Linux Samsung SOC]     [Linux Wireless]     [Linux Kernel]     [ATH6KL]     [Linux Bluetooth]     [Linux Netdev]     [Kernel Newbies]     [IDE]     [Security]     [Git]     [Netfilter]     [Bugtraq]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Linux ATA RAID]     [Samba]     [Device Mapper]

  Powered by Linux