On 18.01.2018 16:58, Michael Mueller wrote: > > > On 18.01.18 15:33, David Hildenbrand wrote: >> On 18.01.2018 15:29, Michael Mueller wrote: >>> >>> >>> On 17.01.18 15:35, David Hildenbrand wrote: >>>> On 16.01.2018 21:02, Christian Borntraeger wrote: >>>>> From: Michael Mueller <mimu@xxxxxxxxxxxxxxxxxx> >>>>> >>>>> The patch implements routines to access the GISA to test and modify >>>>> its Interruption Pending Mask (IPM) from the host side. >>>>> >>>>> Signed-off-by: Michael Mueller <mimu@xxxxxxxxxxxxxxxxxx> >>>>> Reviewed-by: Pierre Morel <pmorel@xxxxxxxxxxxxxxxxxx> >>>>> Reviewed-by: Halil Pasic <pasic@xxxxxxxxxxxxxxxxxx> >>>>> Reviewed-by: Christian Borntraeger <borntraeger@xxxxxxxxxx> >>>>> Signed-off-by: Christian Borntraeger <borntraeger@xxxxxxxxxx> >>>>> --- >>>>> arch/s390/kvm/interrupt.c | 23 +++++++++++++++++++++++ >>>>> arch/s390/kvm/kvm-s390.h | 4 ++++ >>>>> 2 files changed, 27 insertions(+) >>>>> >>>>> diff --git a/arch/s390/kvm/interrupt.c b/arch/s390/kvm/interrupt.c >>>>> index b94173560dcf..dfdecff302d2 100644 >>>>> --- a/arch/s390/kvm/interrupt.c >>>>> +++ b/arch/s390/kvm/interrupt.c >>>>> @@ -2682,3 +2682,26 @@ int kvm_s390_get_irq_state(struct kvm_vcpu *vcpu, __u8 __user *buf, int len) >>>>> >>>>> return n; >>>>> } >>>>> + >>>>> +#define IPM_BIT_OFFSET (offsetof(struct kvm_s390_gisa, ipm) * 8) >>>> >>>> 8 -> BITS_PER_BYTE, but ... >>>> >>>> Am I wrong or can we only modify the 8 ipm bits this way? But we >>>> want/have to do it in an atomic fashion? >>>> >>>> Using an unsigned long seems wrong, because we "rewrite" more than we >>>> should. Esp. everything beyond ipm. oi / ni and friends are not >>>> available on older machines. >>>> >>>> What about something as simple as the following >>>> >>>> >>>> +void kvm_s390_gisa_set_ipm_gisc(struct kvm_s390_gisa *gisa, u8 gisc) >>>> +{ >>>> + u8 value = (0x80 >> gisc); >>>> + >>>> + __sync_fetch_and_or(&gisa->ipm, value); >>>> +} >>>> + >>>> >>> >>> Nobody is using this compiler build-in in the kernel and a quick compile >>> shows that it produces an ORK and CS instead of a LAOG beside a lot of >>> supporting instructions. Unfortunately it's not saving what you promise >>> ... ;) Will not change. >>> >> >> Using unsigned long * bitmap operations to modify an u8 type atomically >> just seems very very wrong > > > I have to reconsider this... > >> > Actually, the problem is there are no atomic byte-based operations on s390x without Interlocked-Access-Facility 2. Even __sync_fetch_and_or(&gisa->ipm, value) falls back to a Compare-And-Swap loop. And Compare-And-Swap also operates at least on 32bit. So I assume there isn't too much we can do about it. As storage locations following the u8 are also written - but in an atomic matter, it should in general not matter. But can we avoid starting the bitmap at the beginning of the gisa? What about something like this: +void kvm_s390_gisa_set_ipm_gisc(struct kvm_s390_gisa *gisa, u8 gisc) +{ + set_bit_inv(gisc, (unsigned long *) &gisa->ipm); +} -- Thanks, David / dhildenb -- To unsubscribe from this list: send the line "unsubscribe linux-s390" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html