[Lldb-commits] [PATCH] D147816: Clarify how watchpoint description in stop packets work, fix AArch64 unintended behavior

[Muhammad Omair Javaid via Phabricator via lldb-commits]

omjavaid added a comment.

I agree that silent continue was wrong and should be fixed. I tried for to remember what I was trying to do when I wrote that patch ... still dont remember much but digged out some information that may be useful for this patch review.

We had a bunch of funny behaving hardware mostly Nexus phones with different types of watchpoint behavior being implement by every vendor.

>From our local record i found that stp issue was never fixed. Some vendor machines reported correct hit_address while some didnt. In LLVM we do have bug report for another of these issues in one of the cases where STP instruction can trigger multiple watchpoints located side by side. https://bugs.llvm.org/show_bug.cgi?id=30758

On Linux ptrace is responsible for reporting a watchpoint hit address and also responsible for setting/unsetting watchpoints. In case of Arm64 ptrace while reporting watchpoints performs some heuristic based calculations to exactly cater for the case you have mentioned where access reports a address out of range. See watchpoint_handler code here:https://elixir.bootlin.com/linux/latest/source/arch/arm64/kernel/hw_breakpoint.c#L754

And this comment copied from same file :
/*

Arm64 hardware does not always report a watchpoint hit address that matches
one of the watchpoints set. It can also report an address "near" the
watchpoint if a single instruction access both watched and unwatched
addresses. There is no straight-forward way, short of disassembling the
offending instruction, to map that address back to the watchpoint. This
function computes the distance of the memory access from the watchpoint as a
heuristic for the likelihood that a given access triggered the watchpoint. *
See Section D2 <https://reviews.llvm.org/D2>.10.5 "Determining the memory location that caused a Watchpoint
exception" of ARMv8 Architecture Reference Manual for details. *
The function returns the distance of the address from the bytes watched by
the watchpoint. In case of an exact match, it returns 0. */


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D147816/new/

https://reviews.llvm.org/D147816