On 2 February 2017 at 12:17, Alex Bennée <alex.ben...@linaro.org> wrote: > Peter Maydell <peter.mayd...@linaro.org> writes: >> Your longjump is effectively skipping the normal "end of the TB" code, >> which is what usually does the set_condexec for you. At the end of a >> TB the expectation is that everything's been sync'd back to the CPU >> state structure. > > Hmm so as long as the tlb flush helpers don't set ARM_CP_SUPPRESS_TB_END > things should just work normally?
If they're register writes, yes -- we end the TB on writes but not on reads (on the assumption that only writes might modify state that requires a TB end). If the TLB instructions are reads then we can add a "force TB end" flag. > Is shouldn't matter if the TB with the > flush is chained to a new TB as the exit_request test should fire before > any more state changing operations happen? I think so, but again if you know that you're always going to exit the TB then there's no point setting it up so it might chain. thanks -- PMM
