Hi Matt, On Sat, Sep 5, 2026 at 9:30 PM Matt Turner <[email protected]> wrote: > > Implement the perf register sampling interface for Alpha. This enables > perf to capture register state and user stack dumps with samples, > supporting --call-graph dwarf. > > The perf_regs enum exposes the registers available in pt_regs: r0-r8, > r16-r28, gp, pc, and ps. Registers r9-r15 are callee-saved and live > in switch_stack, not pt_regs, so they are not included. > > A switch statement maps perf register indices to pt_regs fields since > Alpha's pt_regs layout is non-contiguous (unlike architectures where > direct array indexing works). > > Assisted-by: Claude:claude-opus-4-6 > Signed-off-by: Matt Turner <[email protected]>
I think this patch still needs kernel-side work in addition to the matching tools/perf support. HAVE_PERF_USER_STACK_DUMP makes the generic perf code obtain the user stack pointer through user_stack_pointer(regs), which on Alpha returns regs->usp. However, the interrupt SAVE_ALL path does not initialize that field, and do_entInt() does not snapshot USP. A profiling interrupt taken from user mode may therefore attempt to copy the stack from a stale or uninitialized address. The proposed perf register ABI also does not include SP. A DWARF unwinder needs both the sampled stack contents and the corresponding stack pointer, so this cannot be completed solely by a later tools/perf patch. Would it make sense to drop this patch from the current series and submit the perf support separately, including USP capture on interrupt entry, SP in the kernel perf register ABI, and the matching tools/perf support? That would allow the rest of this series to proceed without waiting for the perf interface to be completed. Regards, Magnus
