Message ID | 1491957874-31600-24-git-send-email-andre.przywara@arm.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
Hi Andre, On 12/04/17 01:44, Andre Przywara wrote: > The INV command instructs the ITS to update the configuration data for > a given LPI by re-reading its entry from the property table. > We don't need to care so much about the priority value, but enabling > or disabling an LPI has some effect: We remove or push virtual LPIs > to their VCPUs, also check the virtual pending bit if an LPI gets enabled. > > Signed-off-by: Andre Przywara <andre.przywara@arm.com> > --- > xen/arch/arm/vgic-v3-its.c | 65 ++++++++++++++++++++++++++++++++++++++++++++++ > 1 file changed, 65 insertions(+) > > diff --git a/xen/arch/arm/vgic-v3-its.c b/xen/arch/arm/vgic-v3-its.c > index 09cb3af..f2789c5 100644 > --- a/xen/arch/arm/vgic-v3-its.c > +++ b/xen/arch/arm/vgic-v3-its.c > @@ -418,6 +418,68 @@ static int update_lpi_property(struct domain *d, uint32_t vlpi, > return 0; > } > > +/* > + * Checks whether an LPI that got enabled or disabled needs to change > + * something in the VGIC (added or removed from the LR or queues). > + * Must be called with the VCPU VGIC lock held. > + */ > +static void update_lpi_vgic_status(struct vcpu *v, struct pending_irq *p, > + uint32_t vlpi) p->irq should be equal to vlpi. No? > +{ > + ASSERT(spin_is_locked(&v->arch.vgic.lock)); The locking is likely to wrong here too (see patch #2). For instance with a MOVI then INV on interrupt enabled. > + > + if ( test_bit(GIC_IRQ_GUEST_ENABLED, &p->status) ) > + { > + if ( !list_empty(&p->inflight) && > + !test_bit(GIC_IRQ_GUEST_VISIBLE, &p->status) ) > + gic_raise_guest_irq(v, vlpi, p->lpi_priority); > + } > + else > + { > + clear_bit(GIC_IRQ_GUEST_ENABLED, &p->status); > + list_del_init(&p->lr_queue); > + } > +} > + > +static int its_handle_inv(struct virt_its *its, uint64_t *cmdptr) > +{ > + struct domain *d = its->d; > + uint32_t devid = its_cmd_get_deviceid(cmdptr); > + uint32_t eventid = its_cmd_get_id(cmdptr); > + struct pending_irq *p; > + unsigned long flags; > + struct vcpu *vcpu; > + uint32_t vlpi; > + int ret = -1; > + > + /* Translate the event into a vCPU/vLPI pair. */ > + if ( !read_itte(its, devid, eventid, &vcpu, &vlpi) ) > + return -1; > + > + if ( vlpi == INVALID_LPI ) > + return -1; > + > + spin_lock_irqsave(&vcpu->arch.vgic.lock, flags); > + > + p = d->arch.vgic.handler->lpi_to_pending(d, vlpi); > + if ( !p ) > + goto out_unlock; As said on v5, this could be simpler and use the pending_irqs in the device. That would be an improvement though. So a would be good. > + > + /* Read the property table and update our cached status. */ > + if ( update_lpi_property(d, vlpi, p) ) > + goto out_unlock; > + > + /* Check whether the LPI needs to go on a VCPU. */ > + update_lpi_vgic_status(vcpu, p, vlpi); > + > + ret = 0; > + > +out_unlock: > + spin_unlock_irqrestore(&vcpu->arch.vgic.lock, flags); > + > + return ret; > +} > + > static int its_handle_mapc(struct virt_its *its, uint64_t *cmdptr) > { > uint32_t collid = its_cmd_get_collection(cmdptr); > @@ -757,6 +819,9 @@ static int vgic_its_handle_cmds(struct domain *d, struct virt_its *its) > case GITS_CMD_INT: > ret = its_handle_int(its, command); > break; > + case GITS_CMD_INV: > + ret = its_handle_inv(its, command); > + break; > case GITS_CMD_MAPC: > ret = its_handle_mapc(its, command); > break; > Cheers,
Hi, On 12/04/17 18:20, Julien Grall wrote: > Hi Andre, > > On 12/04/17 01:44, Andre Przywara wrote: >> The INV command instructs the ITS to update the configuration data for >> a given LPI by re-reading its entry from the property table. >> We don't need to care so much about the priority value, but enabling >> or disabling an LPI has some effect: We remove or push virtual LPIs >> to their VCPUs, also check the virtual pending bit if an LPI gets >> enabled. >> >> Signed-off-by: Andre Przywara <andre.przywara@arm.com> >> --- >> xen/arch/arm/vgic-v3-its.c | 65 >> ++++++++++++++++++++++++++++++++++++++++++++++ >> 1 file changed, 65 insertions(+) >> >> diff --git a/xen/arch/arm/vgic-v3-its.c b/xen/arch/arm/vgic-v3-its.c >> index 09cb3af..f2789c5 100644 >> --- a/xen/arch/arm/vgic-v3-its.c >> +++ b/xen/arch/arm/vgic-v3-its.c >> @@ -418,6 +418,68 @@ static int update_lpi_property(struct domain *d, >> uint32_t vlpi, >> return 0; >> } >> >> +/* >> + * Checks whether an LPI that got enabled or disabled needs to change >> + * something in the VGIC (added or removed from the LR or queues). >> + * Must be called with the VCPU VGIC lock held. >> + */ >> +static void update_lpi_vgic_status(struct vcpu *v, struct pending_irq >> *p, >> + uint32_t vlpi) > > p->irq should be equal to vlpi. No? It is, by I liked the idea of having logically separate parameters expressed as such. But I removed vlpi now and am using p->irq instead. > >> +{ >> + ASSERT(spin_is_locked(&v->arch.vgic.lock)); > > The locking is likely to wrong here too (see patch #2). For instance > with a MOVI then INV on interrupt enabled. > >> + >> + if ( test_bit(GIC_IRQ_GUEST_ENABLED, &p->status) ) >> + { >> + if ( !list_empty(&p->inflight) && >> + !test_bit(GIC_IRQ_GUEST_VISIBLE, &p->status) ) >> + gic_raise_guest_irq(v, vlpi, p->lpi_priority); >> + } >> + else >> + { >> + clear_bit(GIC_IRQ_GUEST_ENABLED, &p->status); >> + list_del_init(&p->lr_queue); >> + } >> +} >> + >> +static int its_handle_inv(struct virt_its *its, uint64_t *cmdptr) >> +{ >> + struct domain *d = its->d; >> + uint32_t devid = its_cmd_get_deviceid(cmdptr); >> + uint32_t eventid = its_cmd_get_id(cmdptr); >> + struct pending_irq *p; >> + unsigned long flags; >> + struct vcpu *vcpu; >> + uint32_t vlpi; >> + int ret = -1; >> + >> + /* Translate the event into a vCPU/vLPI pair. */ >> + if ( !read_itte(its, devid, eventid, &vcpu, &vlpi) ) >> + return -1; >> + >> + if ( vlpi == INVALID_LPI ) >> + return -1; >> + >> + spin_lock_irqsave(&vcpu->arch.vgic.lock, flags); >> + >> + p = d->arch.vgic.handler->lpi_to_pending(d, vlpi); >> + if ( !p ) >> + goto out_unlock; > > As said on v5, this could be simpler and use the pending_irqs in the > device. That would be an improvement though. So a would be good. Originally I found it more straight-forward to use the one existing interface (the rbtree) we also use in the VGIC part, which would allow us to handle locking or ref-counting in one central place. But indeed the ITS command handling has all the data we need to find the pending_irq directly from the virtual device. So I replaced all lpi_to_pending() calls in those handlers with a new function gicv3_its_get_event_pending_irq(), which looks up the struct from an ITS/device/event triple. I take and keep the its->lock for the runtime of these functions, so those events and their memory will not vanish meanwhile. Does that make sense? Cheers, Andre. >> + >> + /* Read the property table and update our cached status. */ >> + if ( update_lpi_property(d, vlpi, p) ) >> + goto out_unlock; >> + >> + /* Check whether the LPI needs to go on a VCPU. */ >> + update_lpi_vgic_status(vcpu, p, vlpi); >> + >> + ret = 0; >> + >> +out_unlock: >> + spin_unlock_irqrestore(&vcpu->arch.vgic.lock, flags); >> + >> + return ret; >> +} >> + >> static int its_handle_mapc(struct virt_its *its, uint64_t *cmdptr) >> { >> uint32_t collid = its_cmd_get_collection(cmdptr); >> @@ -757,6 +819,9 @@ static int vgic_its_handle_cmds(struct domain *d, >> struct virt_its *its) >> case GITS_CMD_INT: >> ret = its_handle_int(its, command); >> break; >> + case GITS_CMD_INV: >> + ret = its_handle_inv(its, command); >> + break; >> case GITS_CMD_MAPC: >> ret = its_handle_mapc(its, command); >> break; >> > > Cheers, >
On 10/05/17 16:11, Andre Przywara wrote: > Hi, Hi Andre, > On 12/04/17 18:20, Julien Grall wrote: >> On 12/04/17 01:44, Andre Przywara wrote: >> >>> +{ >>> + ASSERT(spin_is_locked(&v->arch.vgic.lock)); >> >> The locking is likely to wrong here too (see patch #2). For instance >> with a MOVI then INV on interrupt enabled. >> >>> + >>> + if ( test_bit(GIC_IRQ_GUEST_ENABLED, &p->status) ) >>> + { >>> + if ( !list_empty(&p->inflight) && >>> + !test_bit(GIC_IRQ_GUEST_VISIBLE, &p->status) ) >>> + gic_raise_guest_irq(v, vlpi, p->lpi_priority); >>> + } >>> + else >>> + { >>> + clear_bit(GIC_IRQ_GUEST_ENABLED, &p->status); >>> + list_del_init(&p->lr_queue); >>> + } >>> +} >>> + >>> +static int its_handle_inv(struct virt_its *its, uint64_t *cmdptr) >>> +{ >>> + struct domain *d = its->d; >>> + uint32_t devid = its_cmd_get_deviceid(cmdptr); >>> + uint32_t eventid = its_cmd_get_id(cmdptr); >>> + struct pending_irq *p; >>> + unsigned long flags; >>> + struct vcpu *vcpu; >>> + uint32_t vlpi; >>> + int ret = -1; >>> + >>> + /* Translate the event into a vCPU/vLPI pair. */ >>> + if ( !read_itte(its, devid, eventid, &vcpu, &vlpi) ) >>> + return -1; >>> + >>> + if ( vlpi == INVALID_LPI ) >>> + return -1; >>> + >>> + spin_lock_irqsave(&vcpu->arch.vgic.lock, flags); >>> + >>> + p = d->arch.vgic.handler->lpi_to_pending(d, vlpi); >>> + if ( !p ) >>> + goto out_unlock; >> >> As said on v5, this could be simpler and use the pending_irqs in the >> device. That would be an improvement though. So a would be good. > > Originally I found it more straight-forward to use the one existing > interface (the rbtree) we also use in the VGIC part, which would allow > us to handle locking or ref-counting in one central place. > But indeed the ITS command handling has all the data we need to find the > pending_irq directly from the virtual device. > So I replaced all lpi_to_pending() calls in those handlers with a new > function gicv3_its_get_event_pending_irq(), which looks up the struct > from an ITS/device/event triple. > I take and keep the its->lock for the runtime of these functions, so > those events and their memory will not vanish meanwhile. > > Does that make sense? It makes sense to keep the ref-counting in one central place. But it is better to avoid reading guest memory and therefore avoid most of checking and overhead to translate the IPA to PA. That's why I suggested to use pending_irqs :). Cheers,
diff --git a/xen/arch/arm/vgic-v3-its.c b/xen/arch/arm/vgic-v3-its.c index 09cb3af..f2789c5 100644 --- a/xen/arch/arm/vgic-v3-its.c +++ b/xen/arch/arm/vgic-v3-its.c @@ -418,6 +418,68 @@ static int update_lpi_property(struct domain *d, uint32_t vlpi, return 0; } +/* + * Checks whether an LPI that got enabled or disabled needs to change + * something in the VGIC (added or removed from the LR or queues). + * Must be called with the VCPU VGIC lock held. + */ +static void update_lpi_vgic_status(struct vcpu *v, struct pending_irq *p, + uint32_t vlpi) +{ + ASSERT(spin_is_locked(&v->arch.vgic.lock)); + + if ( test_bit(GIC_IRQ_GUEST_ENABLED, &p->status) ) + { + if ( !list_empty(&p->inflight) && + !test_bit(GIC_IRQ_GUEST_VISIBLE, &p->status) ) + gic_raise_guest_irq(v, vlpi, p->lpi_priority); + } + else + { + clear_bit(GIC_IRQ_GUEST_ENABLED, &p->status); + list_del_init(&p->lr_queue); + } +} + +static int its_handle_inv(struct virt_its *its, uint64_t *cmdptr) +{ + struct domain *d = its->d; + uint32_t devid = its_cmd_get_deviceid(cmdptr); + uint32_t eventid = its_cmd_get_id(cmdptr); + struct pending_irq *p; + unsigned long flags; + struct vcpu *vcpu; + uint32_t vlpi; + int ret = -1; + + /* Translate the event into a vCPU/vLPI pair. */ + if ( !read_itte(its, devid, eventid, &vcpu, &vlpi) ) + return -1; + + if ( vlpi == INVALID_LPI ) + return -1; + + spin_lock_irqsave(&vcpu->arch.vgic.lock, flags); + + p = d->arch.vgic.handler->lpi_to_pending(d, vlpi); + if ( !p ) + goto out_unlock; + + /* Read the property table and update our cached status. */ + if ( update_lpi_property(d, vlpi, p) ) + goto out_unlock; + + /* Check whether the LPI needs to go on a VCPU. */ + update_lpi_vgic_status(vcpu, p, vlpi); + + ret = 0; + +out_unlock: + spin_unlock_irqrestore(&vcpu->arch.vgic.lock, flags); + + return ret; +} + static int its_handle_mapc(struct virt_its *its, uint64_t *cmdptr) { uint32_t collid = its_cmd_get_collection(cmdptr); @@ -757,6 +819,9 @@ static int vgic_its_handle_cmds(struct domain *d, struct virt_its *its) case GITS_CMD_INT: ret = its_handle_int(its, command); break; + case GITS_CMD_INV: + ret = its_handle_inv(its, command); + break; case GITS_CMD_MAPC: ret = its_handle_mapc(its, command); break;
The INV command instructs the ITS to update the configuration data for a given LPI by re-reading its entry from the property table. We don't need to care so much about the priority value, but enabling or disabling an LPI has some effect: We remove or push virtual LPIs to their VCPUs, also check the virtual pending bit if an LPI gets enabled. Signed-off-by: Andre Przywara <andre.przywara@arm.com> --- xen/arch/arm/vgic-v3-its.c | 65 ++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 65 insertions(+)