On 11/09/2017 18:40, Peter Maydell wrote: > On 11 September 2017 at 15:45, Thomas Huth <[email protected]> wrote: >> On 11.09.2017 16:39, Peter Maydell wrote: >>> On 11 September 2017 at 15:20, Laurent Vivier <[email protected]> wrote: >>>> Commit fd5d23babf (hmp: fix "dump-quest-memory" segfault) >>>> fixes the problem for i386, do the same for arm. >>>> >>>> Running QEMU with >>>> qemu-system-aarch64 -M none -nographic -m 256 >>>> and executing >>>> dump-guest-memory /dev/null 0 8192 >>>> results in segfault >>>> >>>> Fix by checking if we have CPU. >>>> >>>> Signed-off-by: Laurent Vivier <[email protected]> >>> >>> It seems a little arbitrary to assume that if there's no >>> CPU what you wanted was a 32-bit little-endian dump. >>> >>> Why do we have a machine without a CPU anyway ? >> >> The "none" machine is always started without a default CPU. > > If it has no CPU then how can we create a core dump for it? > We don't (in theory) even know whether it's x86 or ARM. > (One day we may support multiple CPU architectures in > one QEMU binary...) > > If the theory is hotplug-later then we're a bit stuck > because we need to know information now that we can't > know until the CPU is actually hotplugged.
As we have memory we should be able to dump memory, even without CPU. But I can also do as proposed by Thomas and return -1 to cancel the dump if there is no CPU. Thanks, Laurent
