PTRACE_SINGLEBLOCK support for BookE currently stops on the instruction after taken branches. This is different from the behavior on Server where it stops after all branches.
BookE was made to simulate Server by taking a single step after the branch taken exception. It is understood that the reason for making PTRACE_SINGLEBLOCK on BookE to simulate Server was to make the semantics exposed to user space identicial on both, but this is not really possible due to the fundamental difference that untaken branches do not trigger the branch taken exception in BookE. BookE ISA's branch taken exception triggers before a branch that will be taken executes. This allows software to examine the branch and the conditions under which it will be taken. It also means software can tell where basic blocks end (at least the ones which are terminated by taken branches). There are no architected registers that report the address of the branch instruction after it has executed. Server's branch trace exception triggers after a branch executes regardless of whether or not it was taken. The exception stops on the instruction after fall-through branches. Two mutually-exclusive patches are provided for RFC that expose BookE's branch taken debug exception behavior accessible through PTRACE_SINGLEBLOCK: - The first patch keeps the semantic behavior of the existing support by using the ptrace() addr parameter to select between the modes. This requires a new bit in the TIF as well as changes in kernel/ptrace.c. - The second patch makes PTRACE_SINGLEBLOCK reflect the BookE native behavior, which stops on the branch instruction. The changes are isolated to arch/powerpc/kernel/traps.c. IMHO, the only reason not to do the 2nd patch would be to maintain compatibility for any tools that actually rely on the inaccurate simulation of Server's behavior when run on a BookE system. Are there any tools that actually rely upon the behavior currently implemented for BookE in Linux -- SIGTRAP only after taken branches? Even if there are, it should be possible to modify such a tool to issue a PTRACE_SINGLESTEP after receiving the SIGTRAP on the branch to retain equivalent functionality. _______________________________________________ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
