On 21.06.2018 17:49, Christian Borntraeger wrote: > > > On 04/24/2018 12:18 PM, David Hildenbrand wrote: >> Calling pause_all_vcpus()/resume_all_vcpus() from a VCPU thread might >> not be the best idea. As pause_all_vcpus() temporarily drops the qemu >> mutex, two parallel calls to pause_all_vcpus() can be active at a time, >> resulting in a deadlock. (either by two VCPUs or by the main thread and a >> VCPU) >> >> Let's handle it via the main loop instead, as suggested by Paolo. If we >> would have two parallel reset requests by two different VCPUs at the >> same time, the last one would win. >> >> We use the existing ipl device to handle it. The nice side effect is >> that we can get rid of reipl_requested. >> >> This change implies that all reset handling now goes via the common >> path, so "no-reboot" handling is now active for all kinds of reboots. > > Ok, this breaks the s390 IPL process when -no-reboot is specified. > The bios does a diagnose 308 subcode 1 to jump to the final image while > at the same time resetting all devices. This is now blocked with -no-reboot > (although it is actually the boot)
Indeed, one corner case I missed, thanks for looking into it. I wonder if we could turn that into a allow "one S390_RESET_MODIFIED_CLEAR ignoring -no-reboot" if booting from bios using a simple flag. Then s/SHUTDOWN_CAUSE_GUEST_RESET/SHUTDOWN_CAUSE_BIOS_RESET/ > > > I have noticed that with virt-install on iso images since virt-install > specifies -no-reboot. > > Something like this seems to help but it is not a nice solution. > > diff --git a/hw/s390x/ipl.c b/hw/s390x/ipl.c > index 0d67349004..7b32698eaa 100644 > --- a/hw/s390x/ipl.c > +++ b/hw/s390x/ipl.c > @@ -534,8 +534,14 @@ void s390_ipl_reset_request(CPUState *cs, enum > s390_reset reset_type) > */ > ipl->iplb_valid = s390_gen_initial_iplb(ipl); > } > + qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET); > + } else if (reset_type == S390_RESET_MODIFIED_CLEAR || > + reset_type == S390_RESET_LOAD_NORMAL) { > + /* ignore -no-reboot */ > + qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET_FORCE); > + } else { > + qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET); > } > - qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET); > /* as this is triggered by a CPU, make sure to exit the loop */ > if (tcg_enabled()) { > cpu_loop_exit(cs); > diff --git a/include/sysemu/sysemu.h b/include/sysemu/sysemu.h > index e893f72f3b..e9b11fd6cb 100644 > --- a/include/sysemu/sysemu.h > +++ b/include/sysemu/sysemu.h > @@ -44,6 +44,9 @@ typedef enum ShutdownCause { > turns that into a shutdown */ > SHUTDOWN_CAUSE_GUEST_PANIC, /* Guest panicked, and command line turns > that into a shutdown */ > + SHUTDOWN_CAUSE_GUEST_RESET_FORCE,/* Guest reset that should ignore > --no-reboot > + useful for the subsystem resets on s390 > done > + for kdump and kexec */ I don't think it is that ugly. > SHUTDOWN_CAUSE__MAX, > } ShutdownCause; > > diff --git a/vl.c b/vl.c > index b3426e03d0..18f379e833 100644 > --- a/vl.c > +++ b/vl.c > @@ -1674,7 +1674,9 @@ void qemu_system_guest_panicked(GuestPanicInformation > *info) > > void qemu_system_reset_request(ShutdownCause reason) > { > - if (no_reboot) { > + if (reason == SHUTDOWN_CAUSE_GUEST_RESET_FORCE) { > + reset_requested = SHUTDOWN_CAUSE_GUEST_RESET; > + } else if (no_reboot) { > shutdown_requested = reason; > } else { > reset_requested = reason; > -- Thanks, David / dhildenb