On Fri, Oct 30, 2026 at 2:56 PM Richard Henderson <[email protected]> wrote: > Both already run either on the CPU's own thread or with it stopped, or reach > another CPU exactly as cpu_single_step() does, which is where the previous > patch put the same kind of update.
"The previous patch" doesn't apply anymore after the split. That was my wording, sorry. > In system mode gdb inserts breakpoints with the vCPUs stopped, so there is no > window at all. True for gdb, but not for BP_CPU breakpoints. On ppc64, spr_write_ciabr() inserts one from the running vCPU and doesn't end the TB, so old blocks keep probing each other until the next return to the main loop. That's the same window as goto_tb, so I don't think it needs code, but the message should say so. > + /* The first breakpoint takes the CPU off the inline dispatch path. */ The CF_DERIVED comment and the one above tcg_update_cflags() in include/system/tcg.h still only list single-step, one-insn-per-tb and nochain. Both need breakpoints added.
