On Fri, 15 Jan 2021 at 21:43, Alistair Francis <alistai...@gmail.com> wrote: > > On Fri, Jan 15, 2021 at 6:09 AM Bin Meng <bmeng...@gmail.com> wrote: > > > > On Fri, Jan 15, 2021 at 9:55 PM Peter Maydell <peter.mayd...@linaro.org> > > wrote > > > printf is definitely the wrong thing... you need to either report > > > the error back to the guest if the interface the guest is using > > > has a facility for reporting read/write failures, or log or report > > > it to the user using one of our APIs for that. > > > > It seems the hardware does not have a mechanism to report to the > > software when hardware cannot fulfill the task requested by software. > > > > I checked all existence of block_pwrite() callers. It looks like this > > is not handled consistently. Some indeed call printf(), some call > > error_setg_errno(), some call fprintf(stderr), some call qemu_log() > > ... > > Logging a guest error seems like the best bet, I'm not really sure > what else we would do.
Looking at the other options, I think error_report() of some kind is probably the best bet here. thanks -- PMM