On 30.07.19 18:46, Peter Maydell wrote: > On Tue, 30 Jul 2019 at 17:05, Andrey Shinkevich > <andrey.shinkevich@xxxxxxxxxxxxx> wrote: >> >> Not the whole structure is initialized before passing it to the KVM. >> Reduce the number of Valgrind reports. >> >> Signed-off-by: Andrey Shinkevich <andrey.shinkevich@xxxxxxxxxxxxx> > > Does it even make sense to try to valgrind a KVM-enabled run > of QEMU? As soon as we run the guest it will make modifications > to memory which Valgrind can't track; and I don't think > Valgrind supports the KVM_RUN ioctl anyway... As long as we do not care about the guest memory, it does make sense and it does find bugs. See also https://www.linux-kvm.org/page/KVM_Forum_2014 https://www.linux-kvm.org/images/d/d2/03x07-Valgrind.pdf Unfortunately I wasnt able to follow up on those.
