* [PATCH 0/4] xics: Eliminate unnecessary class
@ 2019-09-24 4:59 David Gibson
2019-09-24 4:59 ` [PATCH 1/4] xics: Eliminate 'reject', 'resend' and 'eoi' class hooks David Gibson
` (5 more replies)
0 siblings, 6 replies; 30+ messages in thread
From: David Gibson @ 2019-09-24 4:59 UTC (permalink / raw)
To: qemu-ppc, clg; +Cc: gkurz, qemu-devel, David Gibson
The XICS interrupt controller device used to have separate subtypes
for the KVM and non-KVM variant of the device. That was a bad idea,
because it leaked information that should be entirely host-side
implementation specific to the kinda-sorta guest visible QOM class
names.
We eliminated the KVM specific class some time ago, but it's left
behind a distinction between the TYPE_ICS_BASE abstract class and
TYPE_ICS_SIMPLE subtype which no longer serves any purpose.
This series collapses the two types back into one.
David Gibson (4):
xics: Eliminate 'reject', 'resend' and 'eoi' class hooks
xics: Merge reset and realize hooks
xics: Rename misleading ics_simple_*() functions
xics: Merge TYPE_ICS_BASE and TYPE_ICS_SIMPLE classes
hw/intc/trace-events | 10 +-
hw/intc/xics.c | 211 ++++++++++++++----------------------------
hw/intc/xics_spapr.c | 12 +--
hw/ppc/pnv_psi.c | 6 +-
hw/ppc/spapr_irq.c | 6 +-
include/hw/ppc/xics.h | 30 +-----
6 files changed, 92 insertions(+), 183 deletions(-)
--
2.21.0
^ permalink raw reply [flat|nested] 30+ messages in thread* [PATCH 1/4] xics: Eliminate 'reject', 'resend' and 'eoi' class hooks 2019-09-24 4:59 [PATCH 0/4] xics: Eliminate unnecessary class David Gibson @ 2019-09-24 4:59 ` David Gibson 2019-09-24 5:23 ` Cédric Le Goater 2019-09-24 7:28 ` Greg Kurz 2019-09-24 4:59 ` [PATCH 2/4] xics: Merge reset and realize hooks David Gibson ` (4 subsequent siblings) 5 siblings, 2 replies; 30+ messages in thread From: David Gibson @ 2019-09-24 4:59 UTC (permalink / raw) To: qemu-ppc, clg; +Cc: gkurz, qemu-devel, David Gibson Currently ics_reject(), ics_resend() and ics_eoi() indirect through class methods. But there's only one implementation of each method, the one in TYPE_ICS_SIMPLE. TYPE_ICS_BASE has no implementation, but it's never instantiated, and has no other subtypes. So clean up by eliminating the method and just having ics_reject(), ics_resend() and ics_eoi() contain the logic directly. Signed-off-by: David Gibson <david@gibson.dropbear.id.au> --- hw/intc/trace-events | 4 ++-- hw/intc/xics.c | 54 +++++++++++-------------------------------- include/hw/ppc/xics.h | 4 ---- 3 files changed, 15 insertions(+), 47 deletions(-) diff --git a/hw/intc/trace-events b/hw/intc/trace-events index 719f46b516..fdc716c2cc 100644 --- a/hw/intc/trace-events +++ b/hw/intc/trace-events @@ -70,8 +70,8 @@ xics_ics_simple_set_irq_msi(int srcno, int nr) "set_irq_msi: srcno %d [irq 0x%x] xics_masked_pending(void) "set_irq_msi: masked pending" xics_ics_simple_set_irq_lsi(int srcno, int nr) "set_irq_lsi: srcno %d [irq 0x%x]" xics_ics_simple_write_xive(int nr, int srcno, int server, uint8_t priority) "ics_write_xive: irq 0x%x [src %d] server 0x%x prio 0x%x" -xics_ics_simple_reject(int nr, int srcno) "reject irq 0x%x [src %d]" -xics_ics_simple_eoi(int nr) "ics_eoi: irq 0x%x" +xics_ics_reject(int nr, int srcno) "reject irq 0x%x [src %d]" +xics_ics_eoi(int nr) "ics_eoi: irq 0x%x" # s390_flic_kvm.c flic_create_device(int err) "flic: create device failed %d" diff --git a/hw/intc/xics.c b/hw/intc/xics.c index b2fca2975c..93139b0189 100644 --- a/hw/intc/xics.c +++ b/hw/intc/xics.c @@ -98,32 +98,8 @@ void ics_pic_print_info(ICSState *ics, Monitor *mon) #define XISR(icp) (((icp)->xirr) & XISR_MASK) #define CPPR(icp) (((icp)->xirr) >> 24) -static void ics_reject(ICSState *ics, uint32_t nr) -{ - ICSStateClass *k = ICS_BASE_GET_CLASS(ics); - - if (k->reject) { - k->reject(ics, nr); - } -} - -void ics_resend(ICSState *ics) -{ - ICSStateClass *k = ICS_BASE_GET_CLASS(ics); - - if (k->resend) { - k->resend(ics); - } -} - -static void ics_eoi(ICSState *ics, int nr) -{ - ICSStateClass *k = ICS_BASE_GET_CLASS(ics); - - if (k->eoi) { - k->eoi(ics, nr); - } -} +static void ics_reject(ICSState *ics, uint32_t nr); +static void ics_eoi(ICSState *ics, uint32_t nr); static void icp_check_ipi(ICPState *icp) { @@ -427,7 +403,7 @@ Object *icp_create(Object *cpu, const char *type, XICSFabric *xi, Error **errp) /* * ICS: Source layer */ -static void ics_simple_resend_msi(ICSState *ics, int srcno) +static void ics_resend_msi(ICSState *ics, int srcno) { ICSIRQState *irq = ics->irqs + srcno; @@ -440,7 +416,7 @@ static void ics_simple_resend_msi(ICSState *ics, int srcno) } } -static void ics_simple_resend_lsi(ICSState *ics, int srcno) +static void ics_resend_lsi(ICSState *ics, int srcno) { ICSIRQState *irq = ics->irqs + srcno; @@ -478,7 +454,7 @@ static void ics_simple_set_irq_lsi(ICSState *ics, int srcno, int val) } else { irq->status &= ~XICS_STATUS_ASSERTED; } - ics_simple_resend_lsi(ics, srcno); + ics_resend_lsi(ics, srcno); } void ics_simple_set_irq(void *opaque, int srcno, int val) @@ -512,7 +488,7 @@ static void ics_simple_write_xive_msi(ICSState *ics, int srcno) static void ics_simple_write_xive_lsi(ICSState *ics, int srcno) { - ics_simple_resend_lsi(ics, srcno); + ics_resend_lsi(ics, srcno); } void ics_simple_write_xive(ICSState *ics, int srcno, int server, @@ -534,11 +510,11 @@ void ics_simple_write_xive(ICSState *ics, int srcno, int server, } } -static void ics_simple_reject(ICSState *ics, uint32_t nr) +static void ics_reject(ICSState *ics, uint32_t nr) { ICSIRQState *irq = ics->irqs + nr - ics->offset; - trace_xics_ics_simple_reject(nr, nr - ics->offset); + trace_xics_ics_reject(nr, nr - ics->offset); if (irq->flags & XICS_FLAGS_IRQ_MSI) { irq->status |= XICS_STATUS_REJECTED; } else if (irq->flags & XICS_FLAGS_IRQ_LSI) { @@ -546,26 +522,26 @@ static void ics_simple_reject(ICSState *ics, uint32_t nr) } } -static void ics_simple_resend(ICSState *ics) +void ics_resend(ICSState *ics) { int i; for (i = 0; i < ics->nr_irqs; i++) { /* FIXME: filter by server#? */ if (ics->irqs[i].flags & XICS_FLAGS_IRQ_LSI) { - ics_simple_resend_lsi(ics, i); + ics_resend_lsi(ics, i); } else { - ics_simple_resend_msi(ics, i); + ics_resend_msi(ics, i); } } } -static void ics_simple_eoi(ICSState *ics, uint32_t nr) +static void ics_eoi(ICSState *ics, uint32_t nr) { int srcno = nr - ics->offset; ICSIRQState *irq = ics->irqs + srcno; - trace_xics_ics_simple_eoi(nr); + trace_xics_ics_eoi(nr); if (ics->irqs[srcno].flags & XICS_FLAGS_IRQ_LSI) { irq->status &= ~XICS_STATUS_SENT; @@ -617,10 +593,6 @@ static void ics_simple_class_init(ObjectClass *klass, void *data) &isc->parent_realize); device_class_set_parent_reset(dc, ics_simple_reset, &isc->parent_reset); - - isc->reject = ics_simple_reject; - isc->resend = ics_simple_resend; - isc->eoi = ics_simple_eoi; } static const TypeInfo ics_simple_info = { diff --git a/include/hw/ppc/xics.h b/include/hw/ppc/xics.h index 64a2c8862a..34d7985b7c 100644 --- a/include/hw/ppc/xics.h +++ b/include/hw/ppc/xics.h @@ -106,10 +106,6 @@ struct ICSStateClass { DeviceRealize parent_realize; DeviceReset parent_reset; - - void (*reject)(ICSState *s, uint32_t irq); - void (*resend)(ICSState *s); - void (*eoi)(ICSState *s, uint32_t irq); }; struct ICSState { -- 2.21.0 ^ permalink raw reply related [flat|nested] 30+ messages in thread
* Re: [PATCH 1/4] xics: Eliminate 'reject', 'resend' and 'eoi' class hooks 2019-09-24 4:59 ` [PATCH 1/4] xics: Eliminate 'reject', 'resend' and 'eoi' class hooks David Gibson @ 2019-09-24 5:23 ` Cédric Le Goater 2019-09-24 7:28 ` Greg Kurz 1 sibling, 0 replies; 30+ messages in thread From: Cédric Le Goater @ 2019-09-24 5:23 UTC (permalink / raw) To: David Gibson, qemu-ppc; +Cc: gkurz, qemu-devel On 24/09/2019 06:59, David Gibson wrote: > Currently ics_reject(), ics_resend() and ics_eoi() indirect through > class methods. But there's only one implementation of each method, > the one in TYPE_ICS_SIMPLE. TYPE_ICS_BASE has no implementation, but > it's never instantiated, and has no other subtypes. > > So clean up by eliminating the method and just having ics_reject(), > ics_resend() and ics_eoi() contain the logic directly. > > Signed-off-by: David Gibson <david@gibson.dropbear.id.au> Reviewed-by: Cédric Le Goater <clg@kaod.org> C. > --- > hw/intc/trace-events | 4 ++-- > hw/intc/xics.c | 54 +++++++++++-------------------------------- > include/hw/ppc/xics.h | 4 ---- > 3 files changed, 15 insertions(+), 47 deletions(-) > > diff --git a/hw/intc/trace-events b/hw/intc/trace-events > index 719f46b516..fdc716c2cc 100644 > --- a/hw/intc/trace-events > +++ b/hw/intc/trace-events > @@ -70,8 +70,8 @@ xics_ics_simple_set_irq_msi(int srcno, int nr) "set_irq_msi: srcno %d [irq 0x%x] > xics_masked_pending(void) "set_irq_msi: masked pending" > xics_ics_simple_set_irq_lsi(int srcno, int nr) "set_irq_lsi: srcno %d [irq 0x%x]" > xics_ics_simple_write_xive(int nr, int srcno, int server, uint8_t priority) "ics_write_xive: irq 0x%x [src %d] server 0x%x prio 0x%x" > -xics_ics_simple_reject(int nr, int srcno) "reject irq 0x%x [src %d]" > -xics_ics_simple_eoi(int nr) "ics_eoi: irq 0x%x" > +xics_ics_reject(int nr, int srcno) "reject irq 0x%x [src %d]" > +xics_ics_eoi(int nr) "ics_eoi: irq 0x%x" > > # s390_flic_kvm.c > flic_create_device(int err) "flic: create device failed %d" > diff --git a/hw/intc/xics.c b/hw/intc/xics.c > index b2fca2975c..93139b0189 100644 > --- a/hw/intc/xics.c > +++ b/hw/intc/xics.c > @@ -98,32 +98,8 @@ void ics_pic_print_info(ICSState *ics, Monitor *mon) > #define XISR(icp) (((icp)->xirr) & XISR_MASK) > #define CPPR(icp) (((icp)->xirr) >> 24) > > -static void ics_reject(ICSState *ics, uint32_t nr) > -{ > - ICSStateClass *k = ICS_BASE_GET_CLASS(ics); > - > - if (k->reject) { > - k->reject(ics, nr); > - } > -} > - > -void ics_resend(ICSState *ics) > -{ > - ICSStateClass *k = ICS_BASE_GET_CLASS(ics); > - > - if (k->resend) { > - k->resend(ics); > - } > -} > - > -static void ics_eoi(ICSState *ics, int nr) > -{ > - ICSStateClass *k = ICS_BASE_GET_CLASS(ics); > - > - if (k->eoi) { > - k->eoi(ics, nr); > - } > -} > +static void ics_reject(ICSState *ics, uint32_t nr); > +static void ics_eoi(ICSState *ics, uint32_t nr); > > static void icp_check_ipi(ICPState *icp) > { > @@ -427,7 +403,7 @@ Object *icp_create(Object *cpu, const char *type, XICSFabric *xi, Error **errp) > /* > * ICS: Source layer > */ > -static void ics_simple_resend_msi(ICSState *ics, int srcno) > +static void ics_resend_msi(ICSState *ics, int srcno) > { > ICSIRQState *irq = ics->irqs + srcno; > > @@ -440,7 +416,7 @@ static void ics_simple_resend_msi(ICSState *ics, int srcno) > } > } > > -static void ics_simple_resend_lsi(ICSState *ics, int srcno) > +static void ics_resend_lsi(ICSState *ics, int srcno) > { > ICSIRQState *irq = ics->irqs + srcno; > > @@ -478,7 +454,7 @@ static void ics_simple_set_irq_lsi(ICSState *ics, int srcno, int val) > } else { > irq->status &= ~XICS_STATUS_ASSERTED; > } > - ics_simple_resend_lsi(ics, srcno); > + ics_resend_lsi(ics, srcno); > } > > void ics_simple_set_irq(void *opaque, int srcno, int val) > @@ -512,7 +488,7 @@ static void ics_simple_write_xive_msi(ICSState *ics, int srcno) > > static void ics_simple_write_xive_lsi(ICSState *ics, int srcno) > { > - ics_simple_resend_lsi(ics, srcno); > + ics_resend_lsi(ics, srcno); > } > > void ics_simple_write_xive(ICSState *ics, int srcno, int server, > @@ -534,11 +510,11 @@ void ics_simple_write_xive(ICSState *ics, int srcno, int server, > } > } > > -static void ics_simple_reject(ICSState *ics, uint32_t nr) > +static void ics_reject(ICSState *ics, uint32_t nr) > { > ICSIRQState *irq = ics->irqs + nr - ics->offset; > > - trace_xics_ics_simple_reject(nr, nr - ics->offset); > + trace_xics_ics_reject(nr, nr - ics->offset); > if (irq->flags & XICS_FLAGS_IRQ_MSI) { > irq->status |= XICS_STATUS_REJECTED; > } else if (irq->flags & XICS_FLAGS_IRQ_LSI) { > @@ -546,26 +522,26 @@ static void ics_simple_reject(ICSState *ics, uint32_t nr) > } > } > > -static void ics_simple_resend(ICSState *ics) > +void ics_resend(ICSState *ics) > { > int i; > > for (i = 0; i < ics->nr_irqs; i++) { > /* FIXME: filter by server#? */ > if (ics->irqs[i].flags & XICS_FLAGS_IRQ_LSI) { > - ics_simple_resend_lsi(ics, i); > + ics_resend_lsi(ics, i); > } else { > - ics_simple_resend_msi(ics, i); > + ics_resend_msi(ics, i); > } > } > } > > -static void ics_simple_eoi(ICSState *ics, uint32_t nr) > +static void ics_eoi(ICSState *ics, uint32_t nr) > { > int srcno = nr - ics->offset; > ICSIRQState *irq = ics->irqs + srcno; > > - trace_xics_ics_simple_eoi(nr); > + trace_xics_ics_eoi(nr); > > if (ics->irqs[srcno].flags & XICS_FLAGS_IRQ_LSI) { > irq->status &= ~XICS_STATUS_SENT; > @@ -617,10 +593,6 @@ static void ics_simple_class_init(ObjectClass *klass, void *data) > &isc->parent_realize); > device_class_set_parent_reset(dc, ics_simple_reset, > &isc->parent_reset); > - > - isc->reject = ics_simple_reject; > - isc->resend = ics_simple_resend; > - isc->eoi = ics_simple_eoi; > } > > static const TypeInfo ics_simple_info = { > diff --git a/include/hw/ppc/xics.h b/include/hw/ppc/xics.h > index 64a2c8862a..34d7985b7c 100644 > --- a/include/hw/ppc/xics.h > +++ b/include/hw/ppc/xics.h > @@ -106,10 +106,6 @@ struct ICSStateClass { > > DeviceRealize parent_realize; > DeviceReset parent_reset; > - > - void (*reject)(ICSState *s, uint32_t irq); > - void (*resend)(ICSState *s); > - void (*eoi)(ICSState *s, uint32_t irq); > }; > > struct ICSState { > ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 1/4] xics: Eliminate 'reject', 'resend' and 'eoi' class hooks 2019-09-24 4:59 ` [PATCH 1/4] xics: Eliminate 'reject', 'resend' and 'eoi' class hooks David Gibson 2019-09-24 5:23 ` Cédric Le Goater @ 2019-09-24 7:28 ` Greg Kurz 1 sibling, 0 replies; 30+ messages in thread From: Greg Kurz @ 2019-09-24 7:28 UTC (permalink / raw) To: David Gibson; +Cc: gkurz, qemu-ppc, clg, qemu-devel On Tue, 24 Sep 2019 14:59:49 +1000 David Gibson <david@gibson.dropbear.id.au> wrote: > Currently ics_reject(), ics_resend() and ics_eoi() indirect through > class methods. But there's only one implementation of each method, > the one in TYPE_ICS_SIMPLE. TYPE_ICS_BASE has no implementation, but > it's never instantiated, and has no other subtypes. > > So clean up by eliminating the method and just having ics_reject(), > ics_resend() and ics_eoi() contain the logic directly. > > Signed-off-by: David Gibson <david@gibson.dropbear.id.au> > --- Reviewed-by: Greg Kurz <groug@kaod.org> > hw/intc/trace-events | 4 ++-- > hw/intc/xics.c | 54 +++++++++++-------------------------------- > include/hw/ppc/xics.h | 4 ---- > 3 files changed, 15 insertions(+), 47 deletions(-) > > diff --git a/hw/intc/trace-events b/hw/intc/trace-events > index 719f46b516..fdc716c2cc 100644 > --- a/hw/intc/trace-events > +++ b/hw/intc/trace-events > @@ -70,8 +70,8 @@ xics_ics_simple_set_irq_msi(int srcno, int nr) "set_irq_msi: srcno %d [irq 0x%x] > xics_masked_pending(void) "set_irq_msi: masked pending" > xics_ics_simple_set_irq_lsi(int srcno, int nr) "set_irq_lsi: srcno %d [irq 0x%x]" > xics_ics_simple_write_xive(int nr, int srcno, int server, uint8_t priority) "ics_write_xive: irq 0x%x [src %d] server 0x%x prio 0x%x" > -xics_ics_simple_reject(int nr, int srcno) "reject irq 0x%x [src %d]" > -xics_ics_simple_eoi(int nr) "ics_eoi: irq 0x%x" > +xics_ics_reject(int nr, int srcno) "reject irq 0x%x [src %d]" > +xics_ics_eoi(int nr) "ics_eoi: irq 0x%x" > > # s390_flic_kvm.c > flic_create_device(int err) "flic: create device failed %d" > diff --git a/hw/intc/xics.c b/hw/intc/xics.c > index b2fca2975c..93139b0189 100644 > --- a/hw/intc/xics.c > +++ b/hw/intc/xics.c > @@ -98,32 +98,8 @@ void ics_pic_print_info(ICSState *ics, Monitor *mon) > #define XISR(icp) (((icp)->xirr) & XISR_MASK) > #define CPPR(icp) (((icp)->xirr) >> 24) > > -static void ics_reject(ICSState *ics, uint32_t nr) > -{ > - ICSStateClass *k = ICS_BASE_GET_CLASS(ics); > - > - if (k->reject) { > - k->reject(ics, nr); > - } > -} > - > -void ics_resend(ICSState *ics) > -{ > - ICSStateClass *k = ICS_BASE_GET_CLASS(ics); > - > - if (k->resend) { > - k->resend(ics); > - } > -} > - > -static void ics_eoi(ICSState *ics, int nr) > -{ > - ICSStateClass *k = ICS_BASE_GET_CLASS(ics); > - > - if (k->eoi) { > - k->eoi(ics, nr); > - } > -} > +static void ics_reject(ICSState *ics, uint32_t nr); > +static void ics_eoi(ICSState *ics, uint32_t nr); > > static void icp_check_ipi(ICPState *icp) > { > @@ -427,7 +403,7 @@ Object *icp_create(Object *cpu, const char *type, XICSFabric *xi, Error **errp) > /* > * ICS: Source layer > */ > -static void ics_simple_resend_msi(ICSState *ics, int srcno) > +static void ics_resend_msi(ICSState *ics, int srcno) > { > ICSIRQState *irq = ics->irqs + srcno; > > @@ -440,7 +416,7 @@ static void ics_simple_resend_msi(ICSState *ics, int srcno) > } > } > > -static void ics_simple_resend_lsi(ICSState *ics, int srcno) > +static void ics_resend_lsi(ICSState *ics, int srcno) > { > ICSIRQState *irq = ics->irqs + srcno; > > @@ -478,7 +454,7 @@ static void ics_simple_set_irq_lsi(ICSState *ics, int srcno, int val) > } else { > irq->status &= ~XICS_STATUS_ASSERTED; > } > - ics_simple_resend_lsi(ics, srcno); > + ics_resend_lsi(ics, srcno); > } > > void ics_simple_set_irq(void *opaque, int srcno, int val) > @@ -512,7 +488,7 @@ static void ics_simple_write_xive_msi(ICSState *ics, int srcno) > > static void ics_simple_write_xive_lsi(ICSState *ics, int srcno) > { > - ics_simple_resend_lsi(ics, srcno); > + ics_resend_lsi(ics, srcno); > } > > void ics_simple_write_xive(ICSState *ics, int srcno, int server, > @@ -534,11 +510,11 @@ void ics_simple_write_xive(ICSState *ics, int srcno, int server, > } > } > > -static void ics_simple_reject(ICSState *ics, uint32_t nr) > +static void ics_reject(ICSState *ics, uint32_t nr) > { > ICSIRQState *irq = ics->irqs + nr - ics->offset; > > - trace_xics_ics_simple_reject(nr, nr - ics->offset); > + trace_xics_ics_reject(nr, nr - ics->offset); > if (irq->flags & XICS_FLAGS_IRQ_MSI) { > irq->status |= XICS_STATUS_REJECTED; > } else if (irq->flags & XICS_FLAGS_IRQ_LSI) { > @@ -546,26 +522,26 @@ static void ics_simple_reject(ICSState *ics, uint32_t nr) > } > } > > -static void ics_simple_resend(ICSState *ics) > +void ics_resend(ICSState *ics) > { > int i; > > for (i = 0; i < ics->nr_irqs; i++) { > /* FIXME: filter by server#? */ > if (ics->irqs[i].flags & XICS_FLAGS_IRQ_LSI) { > - ics_simple_resend_lsi(ics, i); > + ics_resend_lsi(ics, i); > } else { > - ics_simple_resend_msi(ics, i); > + ics_resend_msi(ics, i); > } > } > } > > -static void ics_simple_eoi(ICSState *ics, uint32_t nr) > +static void ics_eoi(ICSState *ics, uint32_t nr) > { > int srcno = nr - ics->offset; > ICSIRQState *irq = ics->irqs + srcno; > > - trace_xics_ics_simple_eoi(nr); > + trace_xics_ics_eoi(nr); > > if (ics->irqs[srcno].flags & XICS_FLAGS_IRQ_LSI) { > irq->status &= ~XICS_STATUS_SENT; > @@ -617,10 +593,6 @@ static void ics_simple_class_init(ObjectClass *klass, void *data) > &isc->parent_realize); > device_class_set_parent_reset(dc, ics_simple_reset, > &isc->parent_reset); > - > - isc->reject = ics_simple_reject; > - isc->resend = ics_simple_resend; > - isc->eoi = ics_simple_eoi; > } > > static const TypeInfo ics_simple_info = { > diff --git a/include/hw/ppc/xics.h b/include/hw/ppc/xics.h > index 64a2c8862a..34d7985b7c 100644 > --- a/include/hw/ppc/xics.h > +++ b/include/hw/ppc/xics.h > @@ -106,10 +106,6 @@ struct ICSStateClass { > > DeviceRealize parent_realize; > DeviceReset parent_reset; > - > - void (*reject)(ICSState *s, uint32_t irq); > - void (*resend)(ICSState *s); > - void (*eoi)(ICSState *s, uint32_t irq); > }; > > struct ICSState { ^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH 2/4] xics: Merge reset and realize hooks 2019-09-24 4:59 [PATCH 0/4] xics: Eliminate unnecessary class David Gibson 2019-09-24 4:59 ` [PATCH 1/4] xics: Eliminate 'reject', 'resend' and 'eoi' class hooks David Gibson @ 2019-09-24 4:59 ` David Gibson 2019-09-24 5:26 ` Cédric Le Goater ` (2 more replies) 2019-09-24 4:59 ` [PATCH 3/4] xics: Rename misleading ics_simple_*() functions David Gibson ` (3 subsequent siblings) 5 siblings, 3 replies; 30+ messages in thread From: David Gibson @ 2019-09-24 4:59 UTC (permalink / raw) To: qemu-ppc, clg; +Cc: gkurz, qemu-devel, David Gibson Currently TYPE_XICS_BASE and TYPE_XICS_SIMPLE have their own reset and realize methods, using the standard technique for having the subtype call the supertype's methods before doing its own thing. But TYPE_XICS_SIMPLE is the only subtype of TYPE_XICS_BASE ever instantiated, so there's no point having the split here. Merge them together into just ics_reset() and ics_realize() functions. Signed-off-by: David Gibson <david@gibson.dropbear.id.au> --- hw/intc/xics.c | 97 ++++++++++++++++--------------------------- include/hw/ppc/xics.h | 3 -- 2 files changed, 35 insertions(+), 65 deletions(-) diff --git a/hw/intc/xics.c b/hw/intc/xics.c index 93139b0189..db0e532bd9 100644 --- a/hw/intc/xics.c +++ b/hw/intc/xics.c @@ -548,68 +548,13 @@ static void ics_eoi(ICSState *ics, uint32_t nr) } } -static void ics_simple_reset(DeviceState *dev) -{ - ICSStateClass *icsc = ICS_BASE_GET_CLASS(dev); - - icsc->parent_reset(dev); - - if (kvm_irqchip_in_kernel()) { - Error *local_err = NULL; - - ics_set_kvm_state(ICS_BASE(dev), &local_err); - if (local_err) { - error_report_err(local_err); - } - } -} - -static void ics_simple_reset_handler(void *dev) -{ - ics_simple_reset(dev); -} - -static void ics_simple_realize(DeviceState *dev, Error **errp) -{ - ICSState *ics = ICS_SIMPLE(dev); - ICSStateClass *icsc = ICS_BASE_GET_CLASS(ics); - Error *local_err = NULL; - - icsc->parent_realize(dev, &local_err); - if (local_err) { - error_propagate(errp, local_err); - return; - } - - qemu_register_reset(ics_simple_reset_handler, ics); -} - -static void ics_simple_class_init(ObjectClass *klass, void *data) -{ - DeviceClass *dc = DEVICE_CLASS(klass); - ICSStateClass *isc = ICS_BASE_CLASS(klass); - - device_class_set_parent_realize(dc, ics_simple_realize, - &isc->parent_realize); - device_class_set_parent_reset(dc, ics_simple_reset, - &isc->parent_reset); -} - -static const TypeInfo ics_simple_info = { - .name = TYPE_ICS_SIMPLE, - .parent = TYPE_ICS_BASE, - .instance_size = sizeof(ICSState), - .class_init = ics_simple_class_init, - .class_size = sizeof(ICSStateClass), -}; - static void ics_reset_irq(ICSIRQState *irq) { irq->priority = 0xff; irq->saved_priority = 0xff; } -static void ics_base_reset(DeviceState *dev) +static void ics_reset(DeviceState *dev) { ICSState *ics = ICS_BASE(dev); int i; @@ -625,17 +570,31 @@ static void ics_base_reset(DeviceState *dev) ics_reset_irq(ics->irqs + i); ics->irqs[i].flags = flags[i]; } + + if (kvm_irqchip_in_kernel()) { + Error *local_err = NULL; + + ics_set_kvm_state(ICS_BASE(dev), &local_err); + if (local_err) { + error_report_err(local_err); + } + } +} + +static void ics_reset_handler(void *dev) +{ + ics_reset(dev); } -static void ics_base_realize(DeviceState *dev, Error **errp) +static void ics_realize(DeviceState *dev, Error **errp) { ICSState *ics = ICS_BASE(dev); + Error *local_err = NULL; Object *obj; - Error *err = NULL; - obj = object_property_get_link(OBJECT(dev), ICS_PROP_XICS, &err); + obj = object_property_get_link(OBJECT(dev), ICS_PROP_XICS, &local_err); if (!obj) { - error_propagate_prepend(errp, err, + error_propagate_prepend(errp, local_err, "required link '" ICS_PROP_XICS "' not found: "); return; @@ -647,8 +606,22 @@ static void ics_base_realize(DeviceState *dev, Error **errp) return; } ics->irqs = g_malloc0(ics->nr_irqs * sizeof(ICSIRQState)); + + qemu_register_reset(ics_reset_handler, ics); +} + +static void ics_simple_class_init(ObjectClass *klass, void *data) +{ } +static const TypeInfo ics_simple_info = { + .name = TYPE_ICS_SIMPLE, + .parent = TYPE_ICS_BASE, + .instance_size = sizeof(ICSState), + .class_init = ics_simple_class_init, + .class_size = sizeof(ICSStateClass), +}; + static void ics_base_instance_init(Object *obj) { ICSState *ics = ICS_BASE(obj); @@ -725,9 +698,9 @@ static void ics_base_class_init(ObjectClass *klass, void *data) { DeviceClass *dc = DEVICE_CLASS(klass); - dc->realize = ics_base_realize; + dc->realize = ics_realize; dc->props = ics_base_properties; - dc->reset = ics_base_reset; + dc->reset = ics_reset; dc->vmsd = &vmstate_ics_base; } diff --git a/include/hw/ppc/xics.h b/include/hw/ppc/xics.h index 34d7985b7c..0eb39c2561 100644 --- a/include/hw/ppc/xics.h +++ b/include/hw/ppc/xics.h @@ -103,9 +103,6 @@ struct PnvICPState { struct ICSStateClass { DeviceClass parent_class; - - DeviceRealize parent_realize; - DeviceReset parent_reset; }; struct ICSState { -- 2.21.0 ^ permalink raw reply related [flat|nested] 30+ messages in thread
* Re: [PATCH 2/4] xics: Merge reset and realize hooks 2019-09-24 4:59 ` [PATCH 2/4] xics: Merge reset and realize hooks David Gibson @ 2019-09-24 5:26 ` Cédric Le Goater 2019-09-24 7:36 ` Greg Kurz 2019-09-24 9:44 ` Philippe Mathieu-Daudé 2 siblings, 0 replies; 30+ messages in thread From: Cédric Le Goater @ 2019-09-24 5:26 UTC (permalink / raw) To: David Gibson, qemu-ppc; +Cc: gkurz, qemu-devel On 24/09/2019 06:59, David Gibson wrote: > Currently TYPE_XICS_BASE and TYPE_XICS_SIMPLE have their own reset and > realize methods, using the standard technique for having the subtype > call the supertype's methods before doing its own thing. > > But TYPE_XICS_SIMPLE is the only subtype of TYPE_XICS_BASE ever > instantiated, so there's no point having the split here. Merge them > together into just ics_reset() and ics_realize() functions. > > Signed-off-by: David Gibson <david@gibson.dropbear.id.au> Reviewed-by: Cédric Le Goater <clg@kaod.org> C. > --- > hw/intc/xics.c | 97 ++++++++++++++++--------------------------- > include/hw/ppc/xics.h | 3 -- > 2 files changed, 35 insertions(+), 65 deletions(-) > > diff --git a/hw/intc/xics.c b/hw/intc/xics.c > index 93139b0189..db0e532bd9 100644 > --- a/hw/intc/xics.c > +++ b/hw/intc/xics.c > @@ -548,68 +548,13 @@ static void ics_eoi(ICSState *ics, uint32_t nr) > } > } > > -static void ics_simple_reset(DeviceState *dev) > -{ > - ICSStateClass *icsc = ICS_BASE_GET_CLASS(dev); > - > - icsc->parent_reset(dev); > - > - if (kvm_irqchip_in_kernel()) { > - Error *local_err = NULL; > - > - ics_set_kvm_state(ICS_BASE(dev), &local_err); > - if (local_err) { > - error_report_err(local_err); > - } > - } > -} > - > -static void ics_simple_reset_handler(void *dev) > -{ > - ics_simple_reset(dev); > -} > - > -static void ics_simple_realize(DeviceState *dev, Error **errp) > -{ > - ICSState *ics = ICS_SIMPLE(dev); > - ICSStateClass *icsc = ICS_BASE_GET_CLASS(ics); > - Error *local_err = NULL; > - > - icsc->parent_realize(dev, &local_err); > - if (local_err) { > - error_propagate(errp, local_err); > - return; > - } > - > - qemu_register_reset(ics_simple_reset_handler, ics); > -} > - > -static void ics_simple_class_init(ObjectClass *klass, void *data) > -{ > - DeviceClass *dc = DEVICE_CLASS(klass); > - ICSStateClass *isc = ICS_BASE_CLASS(klass); > - > - device_class_set_parent_realize(dc, ics_simple_realize, > - &isc->parent_realize); > - device_class_set_parent_reset(dc, ics_simple_reset, > - &isc->parent_reset); > -} > - > -static const TypeInfo ics_simple_info = { > - .name = TYPE_ICS_SIMPLE, > - .parent = TYPE_ICS_BASE, > - .instance_size = sizeof(ICSState), > - .class_init = ics_simple_class_init, > - .class_size = sizeof(ICSStateClass), > -}; > - > static void ics_reset_irq(ICSIRQState *irq) > { > irq->priority = 0xff; > irq->saved_priority = 0xff; > } > > -static void ics_base_reset(DeviceState *dev) > +static void ics_reset(DeviceState *dev) > { > ICSState *ics = ICS_BASE(dev); > int i; > @@ -625,17 +570,31 @@ static void ics_base_reset(DeviceState *dev) > ics_reset_irq(ics->irqs + i); > ics->irqs[i].flags = flags[i]; > } > + > + if (kvm_irqchip_in_kernel()) { > + Error *local_err = NULL; > + > + ics_set_kvm_state(ICS_BASE(dev), &local_err); > + if (local_err) { > + error_report_err(local_err); > + } > + } > +} > + > +static void ics_reset_handler(void *dev) > +{ > + ics_reset(dev); > } > > -static void ics_base_realize(DeviceState *dev, Error **errp) > +static void ics_realize(DeviceState *dev, Error **errp) > { > ICSState *ics = ICS_BASE(dev); > + Error *local_err = NULL; > Object *obj; > - Error *err = NULL; > > - obj = object_property_get_link(OBJECT(dev), ICS_PROP_XICS, &err); > + obj = object_property_get_link(OBJECT(dev), ICS_PROP_XICS, &local_err); > if (!obj) { > - error_propagate_prepend(errp, err, > + error_propagate_prepend(errp, local_err, > "required link '" ICS_PROP_XICS > "' not found: "); > return; > @@ -647,8 +606,22 @@ static void ics_base_realize(DeviceState *dev, Error **errp) > return; > } > ics->irqs = g_malloc0(ics->nr_irqs * sizeof(ICSIRQState)); > + > + qemu_register_reset(ics_reset_handler, ics); > +} > + > +static void ics_simple_class_init(ObjectClass *klass, void *data) > +{ > } > > +static const TypeInfo ics_simple_info = { > + .name = TYPE_ICS_SIMPLE, > + .parent = TYPE_ICS_BASE, > + .instance_size = sizeof(ICSState), > + .class_init = ics_simple_class_init, > + .class_size = sizeof(ICSStateClass), > +}; > + > static void ics_base_instance_init(Object *obj) > { > ICSState *ics = ICS_BASE(obj); > @@ -725,9 +698,9 @@ static void ics_base_class_init(ObjectClass *klass, void *data) > { > DeviceClass *dc = DEVICE_CLASS(klass); > > - dc->realize = ics_base_realize; > + dc->realize = ics_realize; > dc->props = ics_base_properties; > - dc->reset = ics_base_reset; > + dc->reset = ics_reset; > dc->vmsd = &vmstate_ics_base; > } > > diff --git a/include/hw/ppc/xics.h b/include/hw/ppc/xics.h > index 34d7985b7c..0eb39c2561 100644 > --- a/include/hw/ppc/xics.h > +++ b/include/hw/ppc/xics.h > @@ -103,9 +103,6 @@ struct PnvICPState { > > struct ICSStateClass { > DeviceClass parent_class; > - > - DeviceRealize parent_realize; > - DeviceReset parent_reset; > }; > > struct ICSState { > ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 2/4] xics: Merge reset and realize hooks 2019-09-24 4:59 ` [PATCH 2/4] xics: Merge reset and realize hooks David Gibson 2019-09-24 5:26 ` Cédric Le Goater @ 2019-09-24 7:36 ` Greg Kurz 2019-09-24 9:44 ` Philippe Mathieu-Daudé 2 siblings, 0 replies; 30+ messages in thread From: Greg Kurz @ 2019-09-24 7:36 UTC (permalink / raw) To: David Gibson; +Cc: gkurz, qemu-ppc, clg, qemu-devel On Tue, 24 Sep 2019 14:59:50 +1000 David Gibson <david@gibson.dropbear.id.au> wrote: > Currently TYPE_XICS_BASE and TYPE_XICS_SIMPLE have their own reset and > realize methods, using the standard technique for having the subtype > call the supertype's methods before doing its own thing. > > But TYPE_XICS_SIMPLE is the only subtype of TYPE_XICS_BASE ever > instantiated, so there's no point having the split here. Merge them > together into just ics_reset() and ics_realize() functions. > > Signed-off-by: David Gibson <david@gibson.dropbear.id.au> > --- Reviewed-by: Greg Kurz <groug@kaod.org> > hw/intc/xics.c | 97 ++++++++++++++++--------------------------- > include/hw/ppc/xics.h | 3 -- > 2 files changed, 35 insertions(+), 65 deletions(-) > > diff --git a/hw/intc/xics.c b/hw/intc/xics.c > index 93139b0189..db0e532bd9 100644 > --- a/hw/intc/xics.c > +++ b/hw/intc/xics.c > @@ -548,68 +548,13 @@ static void ics_eoi(ICSState *ics, uint32_t nr) > } > } > > -static void ics_simple_reset(DeviceState *dev) > -{ > - ICSStateClass *icsc = ICS_BASE_GET_CLASS(dev); > - > - icsc->parent_reset(dev); > - > - if (kvm_irqchip_in_kernel()) { > - Error *local_err = NULL; > - > - ics_set_kvm_state(ICS_BASE(dev), &local_err); > - if (local_err) { > - error_report_err(local_err); > - } > - } > -} > - > -static void ics_simple_reset_handler(void *dev) > -{ > - ics_simple_reset(dev); > -} > - > -static void ics_simple_realize(DeviceState *dev, Error **errp) > -{ > - ICSState *ics = ICS_SIMPLE(dev); > - ICSStateClass *icsc = ICS_BASE_GET_CLASS(ics); > - Error *local_err = NULL; > - > - icsc->parent_realize(dev, &local_err); > - if (local_err) { > - error_propagate(errp, local_err); > - return; > - } > - > - qemu_register_reset(ics_simple_reset_handler, ics); > -} > - > -static void ics_simple_class_init(ObjectClass *klass, void *data) > -{ > - DeviceClass *dc = DEVICE_CLASS(klass); > - ICSStateClass *isc = ICS_BASE_CLASS(klass); > - > - device_class_set_parent_realize(dc, ics_simple_realize, > - &isc->parent_realize); > - device_class_set_parent_reset(dc, ics_simple_reset, > - &isc->parent_reset); > -} > - > -static const TypeInfo ics_simple_info = { > - .name = TYPE_ICS_SIMPLE, > - .parent = TYPE_ICS_BASE, > - .instance_size = sizeof(ICSState), > - .class_init = ics_simple_class_init, > - .class_size = sizeof(ICSStateClass), > -}; > - > static void ics_reset_irq(ICSIRQState *irq) > { > irq->priority = 0xff; > irq->saved_priority = 0xff; > } > > -static void ics_base_reset(DeviceState *dev) > +static void ics_reset(DeviceState *dev) > { > ICSState *ics = ICS_BASE(dev); > int i; > @@ -625,17 +570,31 @@ static void ics_base_reset(DeviceState *dev) > ics_reset_irq(ics->irqs + i); > ics->irqs[i].flags = flags[i]; > } > + > + if (kvm_irqchip_in_kernel()) { > + Error *local_err = NULL; > + > + ics_set_kvm_state(ICS_BASE(dev), &local_err); > + if (local_err) { > + error_report_err(local_err); > + } > + } > +} > + > +static void ics_reset_handler(void *dev) > +{ > + ics_reset(dev); > } > > -static void ics_base_realize(DeviceState *dev, Error **errp) > +static void ics_realize(DeviceState *dev, Error **errp) > { > ICSState *ics = ICS_BASE(dev); > + Error *local_err = NULL; > Object *obj; > - Error *err = NULL; > > - obj = object_property_get_link(OBJECT(dev), ICS_PROP_XICS, &err); > + obj = object_property_get_link(OBJECT(dev), ICS_PROP_XICS, &local_err); > if (!obj) { > - error_propagate_prepend(errp, err, > + error_propagate_prepend(errp, local_err, > "required link '" ICS_PROP_XICS > "' not found: "); > return; > @@ -647,8 +606,22 @@ static void ics_base_realize(DeviceState *dev, Error **errp) > return; > } > ics->irqs = g_malloc0(ics->nr_irqs * sizeof(ICSIRQState)); > + > + qemu_register_reset(ics_reset_handler, ics); > +} > + > +static void ics_simple_class_init(ObjectClass *klass, void *data) > +{ > } > > +static const TypeInfo ics_simple_info = { > + .name = TYPE_ICS_SIMPLE, > + .parent = TYPE_ICS_BASE, > + .instance_size = sizeof(ICSState), > + .class_init = ics_simple_class_init, > + .class_size = sizeof(ICSStateClass), > +}; > + > static void ics_base_instance_init(Object *obj) > { > ICSState *ics = ICS_BASE(obj); > @@ -725,9 +698,9 @@ static void ics_base_class_init(ObjectClass *klass, void *data) > { > DeviceClass *dc = DEVICE_CLASS(klass); > > - dc->realize = ics_base_realize; > + dc->realize = ics_realize; > dc->props = ics_base_properties; > - dc->reset = ics_base_reset; > + dc->reset = ics_reset; > dc->vmsd = &vmstate_ics_base; > } > > diff --git a/include/hw/ppc/xics.h b/include/hw/ppc/xics.h > index 34d7985b7c..0eb39c2561 100644 > --- a/include/hw/ppc/xics.h > +++ b/include/hw/ppc/xics.h > @@ -103,9 +103,6 @@ struct PnvICPState { > > struct ICSStateClass { > DeviceClass parent_class; > - > - DeviceRealize parent_realize; > - DeviceReset parent_reset; > }; > > struct ICSState { ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 2/4] xics: Merge reset and realize hooks 2019-09-24 4:59 ` [PATCH 2/4] xics: Merge reset and realize hooks David Gibson 2019-09-24 5:26 ` Cédric Le Goater 2019-09-24 7:36 ` Greg Kurz @ 2019-09-24 9:44 ` Philippe Mathieu-Daudé 2019-09-24 11:40 ` David Gibson 2 siblings, 1 reply; 30+ messages in thread From: Philippe Mathieu-Daudé @ 2019-09-24 9:44 UTC (permalink / raw) To: David Gibson, qemu-ppc, clg; +Cc: gkurz, qemu-devel Hi David, On 9/24/19 6:59 AM, David Gibson wrote: > Currently TYPE_XICS_BASE and TYPE_XICS_SIMPLE have their own reset and > realize methods, using the standard technique for having the subtype > call the supertype's methods before doing its own thing. > > But TYPE_XICS_SIMPLE is the only subtype of TYPE_XICS_BASE ever > instantiated, so there's no point having the split here. Merge them > together into just ics_reset() and ics_realize() functions. > > Signed-off-by: David Gibson <david@gibson.dropbear.id.au> > --- > hw/intc/xics.c | 97 ++++++++++++++++--------------------------- > include/hw/ppc/xics.h | 3 -- > 2 files changed, 35 insertions(+), 65 deletions(-) > > diff --git a/hw/intc/xics.c b/hw/intc/xics.c > index 93139b0189..db0e532bd9 100644 > --- a/hw/intc/xics.c > +++ b/hw/intc/xics.c > @@ -548,68 +548,13 @@ static void ics_eoi(ICSState *ics, uint32_t nr) > } > } > > -static void ics_simple_reset(DeviceState *dev) > -{ > - ICSStateClass *icsc = ICS_BASE_GET_CLASS(dev); > - > - icsc->parent_reset(dev); > - > - if (kvm_irqchip_in_kernel()) { > - Error *local_err = NULL; > - > - ics_set_kvm_state(ICS_BASE(dev), &local_err); > - if (local_err) { > - error_report_err(local_err); > - } > - } > -} > - > -static void ics_simple_reset_handler(void *dev) > -{ > - ics_simple_reset(dev); > -} > - > -static void ics_simple_realize(DeviceState *dev, Error **errp) > -{ > - ICSState *ics = ICS_SIMPLE(dev); > - ICSStateClass *icsc = ICS_BASE_GET_CLASS(ics); > - Error *local_err = NULL; > - > - icsc->parent_realize(dev, &local_err); > - if (local_err) { > - error_propagate(errp, local_err); > - return; > - } > - > - qemu_register_reset(ics_simple_reset_handler, ics); > -} > - > -static void ics_simple_class_init(ObjectClass *klass, void *data) > -{ > - DeviceClass *dc = DEVICE_CLASS(klass); > - ICSStateClass *isc = ICS_BASE_CLASS(klass); > - > - device_class_set_parent_realize(dc, ics_simple_realize, > - &isc->parent_realize); > - device_class_set_parent_reset(dc, ics_simple_reset, > - &isc->parent_reset); > -} > - > -static const TypeInfo ics_simple_info = { > - .name = TYPE_ICS_SIMPLE, > - .parent = TYPE_ICS_BASE, > - .instance_size = sizeof(ICSState), > - .class_init = ics_simple_class_init, > - .class_size = sizeof(ICSStateClass), > -}; > - > static void ics_reset_irq(ICSIRQState *irq) > { > irq->priority = 0xff; > irq->saved_priority = 0xff; > } > > -static void ics_base_reset(DeviceState *dev) > +static void ics_reset(DeviceState *dev) > { > ICSState *ics = ICS_BASE(dev); > int i; > @@ -625,17 +570,31 @@ static void ics_base_reset(DeviceState *dev) > ics_reset_irq(ics->irqs + i); > ics->irqs[i].flags = flags[i]; > } > + > + if (kvm_irqchip_in_kernel()) { > + Error *local_err = NULL; > + > + ics_set_kvm_state(ICS_BASE(dev), &local_err); > + if (local_err) { > + error_report_err(local_err); > + } > + } > +} > + > +static void ics_reset_handler(void *dev) > +{ > + ics_reset(dev); > } > > -static void ics_base_realize(DeviceState *dev, Error **errp) > +static void ics_realize(DeviceState *dev, Error **errp) > { > ICSState *ics = ICS_BASE(dev); > + Error *local_err = NULL; Nit: This variable renaming is confusing, maybe another patch? > Object *obj; > - Error *err = NULL; > > - obj = object_property_get_link(OBJECT(dev), ICS_PROP_XICS, &err); > + obj = object_property_get_link(OBJECT(dev), ICS_PROP_XICS, &local_err); > if (!obj) { > - error_propagate_prepend(errp, err, > + error_propagate_prepend(errp, local_err, > "required link '" ICS_PROP_XICS > "' not found: "); > return; > @@ -647,8 +606,22 @@ static void ics_base_realize(DeviceState *dev, Error **errp) > return; > } > ics->irqs = g_malloc0(ics->nr_irqs * sizeof(ICSIRQState)); > + > + qemu_register_reset(ics_reset_handler, ics); Can you change this call by ... > +} > + > +static void ics_simple_class_init(ObjectClass *klass, void *data) > +{ ... this? DeviceClass *dc = DEVICE_CLASS(klass); dc->reset = ics_reset; > } > > +static const TypeInfo ics_simple_info = { > + .name = TYPE_ICS_SIMPLE, > + .parent = TYPE_ICS_BASE, But now reading here, why keep TYPE_ICS_BASE? It seems you can simplify further using directly: .parent = TYPE_DEVICE, > + .instance_size = sizeof(ICSState), > + .class_init = ics_simple_class_init, > + .class_size = sizeof(ICSStateClass), > +}; > + > static void ics_base_instance_init(Object *obj) > { > ICSState *ics = ICS_BASE(obj); > @@ -725,9 +698,9 @@ static void ics_base_class_init(ObjectClass *klass, void *data) > { > DeviceClass *dc = DEVICE_CLASS(klass); > > - dc->realize = ics_base_realize; > + dc->realize = ics_realize; > dc->props = ics_base_properties; > - dc->reset = ics_base_reset; > + dc->reset = ics_reset; > dc->vmsd = &vmstate_ics_base; > } > > diff --git a/include/hw/ppc/xics.h b/include/hw/ppc/xics.h > index 34d7985b7c..0eb39c2561 100644 > --- a/include/hw/ppc/xics.h > +++ b/include/hw/ppc/xics.h > @@ -103,9 +103,6 @@ struct PnvICPState { > > struct ICSStateClass { > DeviceClass parent_class; > - > - DeviceRealize parent_realize; > - DeviceReset parent_reset; > }; > > struct ICSState { > ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 2/4] xics: Merge reset and realize hooks 2019-09-24 9:44 ` Philippe Mathieu-Daudé @ 2019-09-24 11:40 ` David Gibson 0 siblings, 0 replies; 30+ messages in thread From: David Gibson @ 2019-09-24 11:40 UTC (permalink / raw) To: Philippe Mathieu-Daudé; +Cc: gkurz, qemu-ppc, clg, qemu-devel [-- Attachment #1: Type: text/plain, Size: 6605 bytes --] On Tue, Sep 24, 2019 at 11:44:01AM +0200, Philippe Mathieu-Daudé wrote: > Hi David, > > On 9/24/19 6:59 AM, David Gibson wrote: > > Currently TYPE_XICS_BASE and TYPE_XICS_SIMPLE have their own reset and > > realize methods, using the standard technique for having the subtype > > call the supertype's methods before doing its own thing. > > > > But TYPE_XICS_SIMPLE is the only subtype of TYPE_XICS_BASE ever > > instantiated, so there's no point having the split here. Merge them > > together into just ics_reset() and ics_realize() functions. > > > > Signed-off-by: David Gibson <david@gibson.dropbear.id.au> > > --- > > hw/intc/xics.c | 97 ++++++++++++++++--------------------------- > > include/hw/ppc/xics.h | 3 -- > > 2 files changed, 35 insertions(+), 65 deletions(-) > > > > diff --git a/hw/intc/xics.c b/hw/intc/xics.c > > index 93139b0189..db0e532bd9 100644 > > --- a/hw/intc/xics.c > > +++ b/hw/intc/xics.c > > @@ -548,68 +548,13 @@ static void ics_eoi(ICSState *ics, uint32_t nr) > > } > > } > > > > -static void ics_simple_reset(DeviceState *dev) > > -{ > > - ICSStateClass *icsc = ICS_BASE_GET_CLASS(dev); > > - > > - icsc->parent_reset(dev); > > - > > - if (kvm_irqchip_in_kernel()) { > > - Error *local_err = NULL; > > - > > - ics_set_kvm_state(ICS_BASE(dev), &local_err); > > - if (local_err) { > > - error_report_err(local_err); > > - } > > - } > > -} > > - > > -static void ics_simple_reset_handler(void *dev) > > -{ > > - ics_simple_reset(dev); > > -} > > - > > -static void ics_simple_realize(DeviceState *dev, Error **errp) > > -{ > > - ICSState *ics = ICS_SIMPLE(dev); > > - ICSStateClass *icsc = ICS_BASE_GET_CLASS(ics); > > - Error *local_err = NULL; > > - > > - icsc->parent_realize(dev, &local_err); > > - if (local_err) { > > - error_propagate(errp, local_err); > > - return; > > - } > > - > > - qemu_register_reset(ics_simple_reset_handler, ics); > > -} > > - > > -static void ics_simple_class_init(ObjectClass *klass, void *data) > > -{ > > - DeviceClass *dc = DEVICE_CLASS(klass); > > - ICSStateClass *isc = ICS_BASE_CLASS(klass); > > - > > - device_class_set_parent_realize(dc, ics_simple_realize, > > - &isc->parent_realize); > > - device_class_set_parent_reset(dc, ics_simple_reset, > > - &isc->parent_reset); > > -} > > - > > -static const TypeInfo ics_simple_info = { > > - .name = TYPE_ICS_SIMPLE, > > - .parent = TYPE_ICS_BASE, > > - .instance_size = sizeof(ICSState), > > - .class_init = ics_simple_class_init, > > - .class_size = sizeof(ICSStateClass), > > -}; > > - > > static void ics_reset_irq(ICSIRQState *irq) > > { > > irq->priority = 0xff; > > irq->saved_priority = 0xff; > > } > > > > -static void ics_base_reset(DeviceState *dev) > > +static void ics_reset(DeviceState *dev) > > { > > ICSState *ics = ICS_BASE(dev); > > int i; > > @@ -625,17 +570,31 @@ static void ics_base_reset(DeviceState *dev) > > ics_reset_irq(ics->irqs + i); > > ics->irqs[i].flags = flags[i]; > > } 2> > + > > + if (kvm_irqchip_in_kernel()) { > > + Error *local_err = NULL; > > + > > + ics_set_kvm_state(ICS_BASE(dev), &local_err); > > + if (local_err) { > > + error_report_err(local_err); > > + } > > + } > > +} > > + > > +static void ics_reset_handler(void *dev) > > +{ > > + ics_reset(dev); > > } > > > > -static void ics_base_realize(DeviceState *dev, Error **errp) > > +static void ics_realize(DeviceState *dev, Error **errp) > > { > > ICSState *ics = ICS_BASE(dev); > > + Error *local_err = NULL; > > Nit: This variable renaming is confusing, maybe another patch? Turns out this is basically gone in rework I've done since then anyway. > > Object *obj; > > - Error *err = NULL; > > > > - obj = object_property_get_link(OBJECT(dev), ICS_PROP_XICS, &err); > > + obj = object_property_get_link(OBJECT(dev), ICS_PROP_XICS, &local_err); > > if (!obj) { > > - error_propagate_prepend(errp, err, > > + error_propagate_prepend(errp, local_err, > > "required link '" ICS_PROP_XICS > > "' not found: "); > > return; > > @@ -647,8 +606,22 @@ static void ics_base_realize(DeviceState *dev, Error **errp) > > return; > > } > > ics->irqs = g_malloc0(ics->nr_irqs * sizeof(ICSIRQState)); > > + > > + qemu_register_reset(ics_reset_handler, ics); > > Can you change this call by ... > > > +} > > + > > +static void ics_simple_class_init(ObjectClass *klass, void *data) > > +{ > > ... this? > > DeviceClass *dc = DEVICE_CLASS(klass); > > dc->reset = ics_reset; > > > } > > > > +static const TypeInfo ics_simple_info = { > > + .name = TYPE_ICS_SIMPLE, > > + .parent = TYPE_ICS_BASE, > > But now reading here, why keep TYPE_ICS_BASE? > It seems you can simplify further using directly: As you noticed, this happens later in the series. > > .parent = TYPE_DEVICE, > > > + .instance_size = sizeof(ICSState), > > + .class_init = ics_simple_class_init, > > + .class_size = sizeof(ICSStateClass), > > +}; > > + > > static void ics_base_instance_init(Object *obj) > > { > > ICSState *ics = ICS_BASE(obj); > > @@ -725,9 +698,9 @@ static void ics_base_class_init(ObjectClass *klass, void *data) > > { > > DeviceClass *dc = DEVICE_CLASS(klass); > > > > - dc->realize = ics_base_realize; > > + dc->realize = ics_realize; > > dc->props = ics_base_properties; > > - dc->reset = ics_base_reset; > > + dc->reset = ics_reset; > > dc->vmsd = &vmstate_ics_base; > > } > > > > diff --git a/include/hw/ppc/xics.h b/include/hw/ppc/xics.h > > index 34d7985b7c..0eb39c2561 100644 > > --- a/include/hw/ppc/xics.h > > +++ b/include/hw/ppc/xics.h > > @@ -103,9 +103,6 @@ struct PnvICPState { > > > > struct ICSStateClass { > > DeviceClass parent_class; > > - > > - DeviceRealize parent_realize; > > - DeviceReset parent_reset; > > }; > > > > struct ICSState { > > > -- David Gibson | I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_ | _way_ _around_! http://www.ozlabs.org/~dgibson [-- Attachment #2: signature.asc --] [-- Type: application/pgp-signature, Size: 833 bytes --] ^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH 3/4] xics: Rename misleading ics_simple_*() functions 2019-09-24 4:59 [PATCH 0/4] xics: Eliminate unnecessary class David Gibson 2019-09-24 4:59 ` [PATCH 1/4] xics: Eliminate 'reject', 'resend' and 'eoi' class hooks David Gibson 2019-09-24 4:59 ` [PATCH 2/4] xics: Merge reset and realize hooks David Gibson @ 2019-09-24 4:59 ` David Gibson 2019-09-24 5:26 ` Cédric Le Goater 2019-09-24 7:38 ` Greg Kurz 2019-09-24 4:59 ` [PATCH 4/4] xics: Merge TYPE_ICS_BASE and TYPE_ICS_SIMPLE classes David Gibson ` (2 subsequent siblings) 5 siblings, 2 replies; 30+ messages in thread From: David Gibson @ 2019-09-24 4:59 UTC (permalink / raw) To: qemu-ppc, clg; +Cc: gkurz, qemu-devel, David Gibson There are a number of ics_simple_*() functions that aren't actually specific to TYPE_XICS_SIMPLE at all, and are equally valid on TYPE_XICS_BASE. Rename them to ics_*() accordingly. Signed-off-by: David Gibson <david@gibson.dropbear.id.au> --- hw/intc/trace-events | 6 +++--- hw/intc/xics.c | 29 ++++++++++++++--------------- hw/intc/xics_spapr.c | 12 ++++++------ hw/ppc/pnv_psi.c | 4 ++-- hw/ppc/spapr_irq.c | 2 +- include/hw/ppc/xics.h | 6 +++--- 6 files changed, 29 insertions(+), 30 deletions(-) diff --git a/hw/intc/trace-events b/hw/intc/trace-events index fdc716c2cc..527c3f76ca 100644 --- a/hw/intc/trace-events +++ b/hw/intc/trace-events @@ -66,10 +66,10 @@ xics_icp_accept(uint32_t old_xirr, uint32_t new_xirr) "icp_accept: XIRR 0x%"PRIx xics_icp_eoi(int server, uint32_t xirr, uint32_t new_xirr) "icp_eoi: server %d given XIRR 0x%"PRIx32" new XIRR 0x%"PRIx32 xics_icp_irq(int server, int nr, uint8_t priority) "cpu %d trying to deliver irq 0x%"PRIx32" priority 0x%x" xics_icp_raise(uint32_t xirr, uint8_t pending_priority) "raising IRQ new XIRR=0x%x new pending priority=0x%x" -xics_ics_simple_set_irq_msi(int srcno, int nr) "set_irq_msi: srcno %d [irq 0x%x]" +xics_ics_set_irq_msi(int srcno, int nr) "set_irq_msi: srcno %d [irq 0x%x]" xics_masked_pending(void) "set_irq_msi: masked pending" -xics_ics_simple_set_irq_lsi(int srcno, int nr) "set_irq_lsi: srcno %d [irq 0x%x]" -xics_ics_simple_write_xive(int nr, int srcno, int server, uint8_t priority) "ics_write_xive: irq 0x%x [src %d] server 0x%x prio 0x%x" +xics_ics_set_irq_lsi(int srcno, int nr) "set_irq_lsi: srcno %d [irq 0x%x]" +xics_ics_write_xive(int nr, int srcno, int server, uint8_t priority) "ics_write_xive: irq 0x%x [src %d] server 0x%x prio 0x%x" xics_ics_reject(int nr, int srcno) "reject irq 0x%x [src %d]" xics_ics_eoi(int nr) "ics_eoi: irq 0x%x" diff --git a/hw/intc/xics.c b/hw/intc/xics.c index db0e532bd9..9ae51bbc76 100644 --- a/hw/intc/xics.c +++ b/hw/intc/xics.c @@ -428,11 +428,11 @@ static void ics_resend_lsi(ICSState *ics, int srcno) } } -static void ics_simple_set_irq_msi(ICSState *ics, int srcno, int val) +static void ics_set_irq_msi(ICSState *ics, int srcno, int val) { ICSIRQState *irq = ics->irqs + srcno; - trace_xics_ics_simple_set_irq_msi(srcno, srcno + ics->offset); + trace_xics_ics_set_irq_msi(srcno, srcno + ics->offset); if (val) { if (irq->priority == 0xff) { @@ -444,11 +444,11 @@ static void ics_simple_set_irq_msi(ICSState *ics, int srcno, int val) } } -static void ics_simple_set_irq_lsi(ICSState *ics, int srcno, int val) +static void ics_set_irq_lsi(ICSState *ics, int srcno, int val) { ICSIRQState *irq = ics->irqs + srcno; - trace_xics_ics_simple_set_irq_lsi(srcno, srcno + ics->offset); + trace_xics_ics_set_irq_lsi(srcno, srcno + ics->offset); if (val) { irq->status |= XICS_STATUS_ASSERTED; } else { @@ -457,7 +457,7 @@ static void ics_simple_set_irq_lsi(ICSState *ics, int srcno, int val) ics_resend_lsi(ics, srcno); } -void ics_simple_set_irq(void *opaque, int srcno, int val) +void ics_set_irq(void *opaque, int srcno, int val) { ICSState *ics = (ICSState *)opaque; @@ -467,13 +467,13 @@ void ics_simple_set_irq(void *opaque, int srcno, int val) } if (ics->irqs[srcno].flags & XICS_FLAGS_IRQ_LSI) { - ics_simple_set_irq_lsi(ics, srcno, val); + ics_set_irq_lsi(ics, srcno, val); } else { - ics_simple_set_irq_msi(ics, srcno, val); + ics_set_irq_msi(ics, srcno, val); } } -static void ics_simple_write_xive_msi(ICSState *ics, int srcno) +static void ics_write_xive_msi(ICSState *ics, int srcno) { ICSIRQState *irq = ics->irqs + srcno; @@ -486,13 +486,13 @@ static void ics_simple_write_xive_msi(ICSState *ics, int srcno) icp_irq(ics, irq->server, srcno + ics->offset, irq->priority); } -static void ics_simple_write_xive_lsi(ICSState *ics, int srcno) +static void ics_write_xive_lsi(ICSState *ics, int srcno) { ics_resend_lsi(ics, srcno); } -void ics_simple_write_xive(ICSState *ics, int srcno, int server, - uint8_t priority, uint8_t saved_priority) +void ics_write_xive(ICSState *ics, int srcno, int server, + uint8_t priority, uint8_t saved_priority) { ICSIRQState *irq = ics->irqs + srcno; @@ -500,13 +500,12 @@ void ics_simple_write_xive(ICSState *ics, int srcno, int server, irq->priority = priority; irq->saved_priority = saved_priority; - trace_xics_ics_simple_write_xive(ics->offset + srcno, srcno, server, - priority); + trace_xics_ics_write_xive(ics->offset + srcno, srcno, server, priority); if (ics->irqs[srcno].flags & XICS_FLAGS_IRQ_LSI) { - ics_simple_write_xive_lsi(ics, srcno); + ics_write_xive_lsi(ics, srcno); } else { - ics_simple_write_xive_msi(ics, srcno); + ics_write_xive_msi(ics, srcno); } } diff --git a/hw/intc/xics_spapr.c b/hw/intc/xics_spapr.c index 6577be0d92..3e9444813a 100644 --- a/hw/intc/xics_spapr.c +++ b/hw/intc/xics_spapr.c @@ -179,7 +179,7 @@ static void rtas_set_xive(PowerPCCPU *cpu, SpaprMachineState *spapr, } srcno = nr - ics->offset; - ics_simple_write_xive(ics, srcno, server, priority, priority); + ics_write_xive(ics, srcno, server, priority, priority); rtas_st(rets, 0, RTAS_OUT_SUCCESS); } @@ -243,8 +243,8 @@ static void rtas_int_off(PowerPCCPU *cpu, SpaprMachineState *spapr, } srcno = nr - ics->offset; - ics_simple_write_xive(ics, srcno, ics->irqs[srcno].server, 0xff, - ics->irqs[srcno].priority); + ics_write_xive(ics, srcno, ics->irqs[srcno].server, 0xff, + ics->irqs[srcno].priority); rtas_st(rets, 0, RTAS_OUT_SUCCESS); } @@ -276,9 +276,9 @@ static void rtas_int_on(PowerPCCPU *cpu, SpaprMachineState *spapr, } srcno = nr - ics->offset; - ics_simple_write_xive(ics, srcno, ics->irqs[srcno].server, - ics->irqs[srcno].saved_priority, - ics->irqs[srcno].saved_priority); + ics_write_xive(ics, srcno, ics->irqs[srcno].server, + ics->irqs[srcno].saved_priority, + ics->irqs[srcno].saved_priority); rtas_st(rets, 0, RTAS_OUT_SUCCESS); } diff --git a/hw/ppc/pnv_psi.c b/hw/ppc/pnv_psi.c index 88ba8e7b9b..8ea81e9d8e 100644 --- a/hw/ppc/pnv_psi.c +++ b/hw/ppc/pnv_psi.c @@ -311,7 +311,7 @@ static void pnv_psi_set_xivr(PnvPsi *psi, uint32_t reg, uint64_t val) * do for now but a more accurate implementation would instead * use a fixed server/prio and a remapper of the generated irq. */ - ics_simple_write_xive(ics, src, server, prio, prio); + ics_write_xive(ics, src, server, prio, prio); } static uint64_t pnv_psi_reg_read(PnvPsi *psi, uint32_t offset, bool mmio) @@ -514,7 +514,7 @@ static void pnv_psi_power8_realize(DeviceState *dev, Error **errp) ics_set_irq_type(ics, i, true); } - psi->qirqs = qemu_allocate_irqs(ics_simple_set_irq, ics, ics->nr_irqs); + psi->qirqs = qemu_allocate_irqs(ics_set_irq, ics, ics->nr_irqs); /* XSCOM region for PSI registers */ pnv_xscom_region_init(&psi->xscom_regs, OBJECT(dev), &pnv_psi_xscom_ops, diff --git a/hw/ppc/spapr_irq.c b/hw/ppc/spapr_irq.c index d8f46b6797..ac189c5796 100644 --- a/hw/ppc/spapr_irq.c +++ b/hw/ppc/spapr_irq.c @@ -210,7 +210,7 @@ static void spapr_irq_set_irq_xics(void *opaque, int srcno, int val) { SpaprMachineState *spapr = opaque; - ics_simple_set_irq(spapr->ics, srcno, val); + ics_set_irq(spapr->ics, srcno, val); } static void spapr_irq_reset_xics(SpaprMachineState *spapr, Error **errp) diff --git a/include/hw/ppc/xics.h b/include/hw/ppc/xics.h index 0eb39c2561..92628e7cab 100644 --- a/include/hw/ppc/xics.h +++ b/include/hw/ppc/xics.h @@ -168,9 +168,9 @@ uint32_t icp_accept(ICPState *ss); uint32_t icp_ipoll(ICPState *ss, uint32_t *mfrr); void icp_eoi(ICPState *icp, uint32_t xirr); -void ics_simple_write_xive(ICSState *ics, int nr, int server, - uint8_t priority, uint8_t saved_priority); -void ics_simple_set_irq(void *opaque, int srcno, int val); +void ics_write_xive(ICSState *ics, int nr, int server, + uint8_t priority, uint8_t saved_priority); +void ics_set_irq(void *opaque, int srcno, int val); static inline bool ics_irq_free(ICSState *ics, uint32_t srcno) { -- 2.21.0 ^ permalink raw reply related [flat|nested] 30+ messages in thread
* Re: [PATCH 3/4] xics: Rename misleading ics_simple_*() functions 2019-09-24 4:59 ` [PATCH 3/4] xics: Rename misleading ics_simple_*() functions David Gibson @ 2019-09-24 5:26 ` Cédric Le Goater 2019-09-24 7:38 ` Greg Kurz 1 sibling, 0 replies; 30+ messages in thread From: Cédric Le Goater @ 2019-09-24 5:26 UTC (permalink / raw) To: David Gibson, qemu-ppc; +Cc: gkurz, qemu-devel On 24/09/2019 06:59, David Gibson wrote: > There are a number of ics_simple_*() functions that aren't actually > specific to TYPE_XICS_SIMPLE at all, and are equally valid on > TYPE_XICS_BASE. Rename them to ics_*() accordingly. > > Signed-off-by: David Gibson <david@gibson.dropbear.id.au> Reviewed-by: Cédric Le Goater <clg@kaod.org> C. > --- > hw/intc/trace-events | 6 +++--- > hw/intc/xics.c | 29 ++++++++++++++--------------- > hw/intc/xics_spapr.c | 12 ++++++------ > hw/ppc/pnv_psi.c | 4 ++-- > hw/ppc/spapr_irq.c | 2 +- > include/hw/ppc/xics.h | 6 +++--- > 6 files changed, 29 insertions(+), 30 deletions(-) > > diff --git a/hw/intc/trace-events b/hw/intc/trace-events > index fdc716c2cc..527c3f76ca 100644 > --- a/hw/intc/trace-events > +++ b/hw/intc/trace-events > @@ -66,10 +66,10 @@ xics_icp_accept(uint32_t old_xirr, uint32_t new_xirr) "icp_accept: XIRR 0x%"PRIx > xics_icp_eoi(int server, uint32_t xirr, uint32_t new_xirr) "icp_eoi: server %d given XIRR 0x%"PRIx32" new XIRR 0x%"PRIx32 > xics_icp_irq(int server, int nr, uint8_t priority) "cpu %d trying to deliver irq 0x%"PRIx32" priority 0x%x" > xics_icp_raise(uint32_t xirr, uint8_t pending_priority) "raising IRQ new XIRR=0x%x new pending priority=0x%x" > -xics_ics_simple_set_irq_msi(int srcno, int nr) "set_irq_msi: srcno %d [irq 0x%x]" > +xics_ics_set_irq_msi(int srcno, int nr) "set_irq_msi: srcno %d [irq 0x%x]" > xics_masked_pending(void) "set_irq_msi: masked pending" > -xics_ics_simple_set_irq_lsi(int srcno, int nr) "set_irq_lsi: srcno %d [irq 0x%x]" > -xics_ics_simple_write_xive(int nr, int srcno, int server, uint8_t priority) "ics_write_xive: irq 0x%x [src %d] server 0x%x prio 0x%x" > +xics_ics_set_irq_lsi(int srcno, int nr) "set_irq_lsi: srcno %d [irq 0x%x]" > +xics_ics_write_xive(int nr, int srcno, int server, uint8_t priority) "ics_write_xive: irq 0x%x [src %d] server 0x%x prio 0x%x" > xics_ics_reject(int nr, int srcno) "reject irq 0x%x [src %d]" > xics_ics_eoi(int nr) "ics_eoi: irq 0x%x" > > diff --git a/hw/intc/xics.c b/hw/intc/xics.c > index db0e532bd9..9ae51bbc76 100644 > --- a/hw/intc/xics.c > +++ b/hw/intc/xics.c > @@ -428,11 +428,11 @@ static void ics_resend_lsi(ICSState *ics, int srcno) > } > } > > -static void ics_simple_set_irq_msi(ICSState *ics, int srcno, int val) > +static void ics_set_irq_msi(ICSState *ics, int srcno, int val) > { > ICSIRQState *irq = ics->irqs + srcno; > > - trace_xics_ics_simple_set_irq_msi(srcno, srcno + ics->offset); > + trace_xics_ics_set_irq_msi(srcno, srcno + ics->offset); > > if (val) { > if (irq->priority == 0xff) { > @@ -444,11 +444,11 @@ static void ics_simple_set_irq_msi(ICSState *ics, int srcno, int val) > } > } > > -static void ics_simple_set_irq_lsi(ICSState *ics, int srcno, int val) > +static void ics_set_irq_lsi(ICSState *ics, int srcno, int val) > { > ICSIRQState *irq = ics->irqs + srcno; > > - trace_xics_ics_simple_set_irq_lsi(srcno, srcno + ics->offset); > + trace_xics_ics_set_irq_lsi(srcno, srcno + ics->offset); > if (val) { > irq->status |= XICS_STATUS_ASSERTED; > } else { > @@ -457,7 +457,7 @@ static void ics_simple_set_irq_lsi(ICSState *ics, int srcno, int val) > ics_resend_lsi(ics, srcno); > } > > -void ics_simple_set_irq(void *opaque, int srcno, int val) > +void ics_set_irq(void *opaque, int srcno, int val) > { > ICSState *ics = (ICSState *)opaque; > > @@ -467,13 +467,13 @@ void ics_simple_set_irq(void *opaque, int srcno, int val) > } > > if (ics->irqs[srcno].flags & XICS_FLAGS_IRQ_LSI) { > - ics_simple_set_irq_lsi(ics, srcno, val); > + ics_set_irq_lsi(ics, srcno, val); > } else { > - ics_simple_set_irq_msi(ics, srcno, val); > + ics_set_irq_msi(ics, srcno, val); > } > } > > -static void ics_simple_write_xive_msi(ICSState *ics, int srcno) > +static void ics_write_xive_msi(ICSState *ics, int srcno) > { > ICSIRQState *irq = ics->irqs + srcno; > > @@ -486,13 +486,13 @@ static void ics_simple_write_xive_msi(ICSState *ics, int srcno) > icp_irq(ics, irq->server, srcno + ics->offset, irq->priority); > } > > -static void ics_simple_write_xive_lsi(ICSState *ics, int srcno) > +static void ics_write_xive_lsi(ICSState *ics, int srcno) > { > ics_resend_lsi(ics, srcno); > } > > -void ics_simple_write_xive(ICSState *ics, int srcno, int server, > - uint8_t priority, uint8_t saved_priority) > +void ics_write_xive(ICSState *ics, int srcno, int server, > + uint8_t priority, uint8_t saved_priority) > { > ICSIRQState *irq = ics->irqs + srcno; > > @@ -500,13 +500,12 @@ void ics_simple_write_xive(ICSState *ics, int srcno, int server, > irq->priority = priority; > irq->saved_priority = saved_priority; > > - trace_xics_ics_simple_write_xive(ics->offset + srcno, srcno, server, > - priority); > + trace_xics_ics_write_xive(ics->offset + srcno, srcno, server, priority); > > if (ics->irqs[srcno].flags & XICS_FLAGS_IRQ_LSI) { > - ics_simple_write_xive_lsi(ics, srcno); > + ics_write_xive_lsi(ics, srcno); > } else { > - ics_simple_write_xive_msi(ics, srcno); > + ics_write_xive_msi(ics, srcno); > } > } > > diff --git a/hw/intc/xics_spapr.c b/hw/intc/xics_spapr.c > index 6577be0d92..3e9444813a 100644 > --- a/hw/intc/xics_spapr.c > +++ b/hw/intc/xics_spapr.c > @@ -179,7 +179,7 @@ static void rtas_set_xive(PowerPCCPU *cpu, SpaprMachineState *spapr, > } > > srcno = nr - ics->offset; > - ics_simple_write_xive(ics, srcno, server, priority, priority); > + ics_write_xive(ics, srcno, server, priority, priority); > > rtas_st(rets, 0, RTAS_OUT_SUCCESS); > } > @@ -243,8 +243,8 @@ static void rtas_int_off(PowerPCCPU *cpu, SpaprMachineState *spapr, > } > > srcno = nr - ics->offset; > - ics_simple_write_xive(ics, srcno, ics->irqs[srcno].server, 0xff, > - ics->irqs[srcno].priority); > + ics_write_xive(ics, srcno, ics->irqs[srcno].server, 0xff, > + ics->irqs[srcno].priority); > > rtas_st(rets, 0, RTAS_OUT_SUCCESS); > } > @@ -276,9 +276,9 @@ static void rtas_int_on(PowerPCCPU *cpu, SpaprMachineState *spapr, > } > > srcno = nr - ics->offset; > - ics_simple_write_xive(ics, srcno, ics->irqs[srcno].server, > - ics->irqs[srcno].saved_priority, > - ics->irqs[srcno].saved_priority); > + ics_write_xive(ics, srcno, ics->irqs[srcno].server, > + ics->irqs[srcno].saved_priority, > + ics->irqs[srcno].saved_priority); > > rtas_st(rets, 0, RTAS_OUT_SUCCESS); > } > diff --git a/hw/ppc/pnv_psi.c b/hw/ppc/pnv_psi.c > index 88ba8e7b9b..8ea81e9d8e 100644 > --- a/hw/ppc/pnv_psi.c > +++ b/hw/ppc/pnv_psi.c > @@ -311,7 +311,7 @@ static void pnv_psi_set_xivr(PnvPsi *psi, uint32_t reg, uint64_t val) > * do for now but a more accurate implementation would instead > * use a fixed server/prio and a remapper of the generated irq. > */ > - ics_simple_write_xive(ics, src, server, prio, prio); > + ics_write_xive(ics, src, server, prio, prio); > } > > static uint64_t pnv_psi_reg_read(PnvPsi *psi, uint32_t offset, bool mmio) > @@ -514,7 +514,7 @@ static void pnv_psi_power8_realize(DeviceState *dev, Error **errp) > ics_set_irq_type(ics, i, true); > } > > - psi->qirqs = qemu_allocate_irqs(ics_simple_set_irq, ics, ics->nr_irqs); > + psi->qirqs = qemu_allocate_irqs(ics_set_irq, ics, ics->nr_irqs); > > /* XSCOM region for PSI registers */ > pnv_xscom_region_init(&psi->xscom_regs, OBJECT(dev), &pnv_psi_xscom_ops, > diff --git a/hw/ppc/spapr_irq.c b/hw/ppc/spapr_irq.c > index d8f46b6797..ac189c5796 100644 > --- a/hw/ppc/spapr_irq.c > +++ b/hw/ppc/spapr_irq.c > @@ -210,7 +210,7 @@ static void spapr_irq_set_irq_xics(void *opaque, int srcno, int val) > { > SpaprMachineState *spapr = opaque; > > - ics_simple_set_irq(spapr->ics, srcno, val); > + ics_set_irq(spapr->ics, srcno, val); > } > > static void spapr_irq_reset_xics(SpaprMachineState *spapr, Error **errp) > diff --git a/include/hw/ppc/xics.h b/include/hw/ppc/xics.h > index 0eb39c2561..92628e7cab 100644 > --- a/include/hw/ppc/xics.h > +++ b/include/hw/ppc/xics.h > @@ -168,9 +168,9 @@ uint32_t icp_accept(ICPState *ss); > uint32_t icp_ipoll(ICPState *ss, uint32_t *mfrr); > void icp_eoi(ICPState *icp, uint32_t xirr); > > -void ics_simple_write_xive(ICSState *ics, int nr, int server, > - uint8_t priority, uint8_t saved_priority); > -void ics_simple_set_irq(void *opaque, int srcno, int val); > +void ics_write_xive(ICSState *ics, int nr, int server, > + uint8_t priority, uint8_t saved_priority); > +void ics_set_irq(void *opaque, int srcno, int val); > > static inline bool ics_irq_free(ICSState *ics, uint32_t srcno) > { > ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 3/4] xics: Rename misleading ics_simple_*() functions 2019-09-24 4:59 ` [PATCH 3/4] xics: Rename misleading ics_simple_*() functions David Gibson 2019-09-24 5:26 ` Cédric Le Goater @ 2019-09-24 7:38 ` Greg Kurz 1 sibling, 0 replies; 30+ messages in thread From: Greg Kurz @ 2019-09-24 7:38 UTC (permalink / raw) To: David Gibson; +Cc: gkurz, qemu-ppc, clg, qemu-devel On Tue, 24 Sep 2019 14:59:51 +1000 David Gibson <david@gibson.dropbear.id.au> wrote: > There are a number of ics_simple_*() functions that aren't actually > specific to TYPE_XICS_SIMPLE at all, and are equally valid on > TYPE_XICS_BASE. Rename them to ics_*() accordingly. > > Signed-off-by: David Gibson <david@gibson.dropbear.id.au> > --- Reviewed-by: Greg Kurz <groug@kaod.org> > hw/intc/trace-events | 6 +++--- > hw/intc/xics.c | 29 ++++++++++++++--------------- > hw/intc/xics_spapr.c | 12 ++++++------ > hw/ppc/pnv_psi.c | 4 ++-- > hw/ppc/spapr_irq.c | 2 +- > include/hw/ppc/xics.h | 6 +++--- > 6 files changed, 29 insertions(+), 30 deletions(-) > > diff --git a/hw/intc/trace-events b/hw/intc/trace-events > index fdc716c2cc..527c3f76ca 100644 > --- a/hw/intc/trace-events > +++ b/hw/intc/trace-events > @@ -66,10 +66,10 @@ xics_icp_accept(uint32_t old_xirr, uint32_t new_xirr) "icp_accept: XIRR 0x%"PRIx > xics_icp_eoi(int server, uint32_t xirr, uint32_t new_xirr) "icp_eoi: server %d given XIRR 0x%"PRIx32" new XIRR 0x%"PRIx32 > xics_icp_irq(int server, int nr, uint8_t priority) "cpu %d trying to deliver irq 0x%"PRIx32" priority 0x%x" > xics_icp_raise(uint32_t xirr, uint8_t pending_priority) "raising IRQ new XIRR=0x%x new pending priority=0x%x" > -xics_ics_simple_set_irq_msi(int srcno, int nr) "set_irq_msi: srcno %d [irq 0x%x]" > +xics_ics_set_irq_msi(int srcno, int nr) "set_irq_msi: srcno %d [irq 0x%x]" > xics_masked_pending(void) "set_irq_msi: masked pending" > -xics_ics_simple_set_irq_lsi(int srcno, int nr) "set_irq_lsi: srcno %d [irq 0x%x]" > -xics_ics_simple_write_xive(int nr, int srcno, int server, uint8_t priority) "ics_write_xive: irq 0x%x [src %d] server 0x%x prio 0x%x" > +xics_ics_set_irq_lsi(int srcno, int nr) "set_irq_lsi: srcno %d [irq 0x%x]" > +xics_ics_write_xive(int nr, int srcno, int server, uint8_t priority) "ics_write_xive: irq 0x%x [src %d] server 0x%x prio 0x%x" > xics_ics_reject(int nr, int srcno) "reject irq 0x%x [src %d]" > xics_ics_eoi(int nr) "ics_eoi: irq 0x%x" > > diff --git a/hw/intc/xics.c b/hw/intc/xics.c > index db0e532bd9..9ae51bbc76 100644 > --- a/hw/intc/xics.c > +++ b/hw/intc/xics.c > @@ -428,11 +428,11 @@ static void ics_resend_lsi(ICSState *ics, int srcno) > } > } > > -static void ics_simple_set_irq_msi(ICSState *ics, int srcno, int val) > +static void ics_set_irq_msi(ICSState *ics, int srcno, int val) > { > ICSIRQState *irq = ics->irqs + srcno; > > - trace_xics_ics_simple_set_irq_msi(srcno, srcno + ics->offset); > + trace_xics_ics_set_irq_msi(srcno, srcno + ics->offset); > > if (val) { > if (irq->priority == 0xff) { > @@ -444,11 +444,11 @@ static void ics_simple_set_irq_msi(ICSState *ics, int srcno, int val) > } > } > > -static void ics_simple_set_irq_lsi(ICSState *ics, int srcno, int val) > +static void ics_set_irq_lsi(ICSState *ics, int srcno, int val) > { > ICSIRQState *irq = ics->irqs + srcno; > > - trace_xics_ics_simple_set_irq_lsi(srcno, srcno + ics->offset); > + trace_xics_ics_set_irq_lsi(srcno, srcno + ics->offset); > if (val) { > irq->status |= XICS_STATUS_ASSERTED; > } else { > @@ -457,7 +457,7 @@ static void ics_simple_set_irq_lsi(ICSState *ics, int srcno, int val) > ics_resend_lsi(ics, srcno); > } > > -void ics_simple_set_irq(void *opaque, int srcno, int val) > +void ics_set_irq(void *opaque, int srcno, int val) > { > ICSState *ics = (ICSState *)opaque; > > @@ -467,13 +467,13 @@ void ics_simple_set_irq(void *opaque, int srcno, int val) > } > > if (ics->irqs[srcno].flags & XICS_FLAGS_IRQ_LSI) { > - ics_simple_set_irq_lsi(ics, srcno, val); > + ics_set_irq_lsi(ics, srcno, val); > } else { > - ics_simple_set_irq_msi(ics, srcno, val); > + ics_set_irq_msi(ics, srcno, val); > } > } > > -static void ics_simple_write_xive_msi(ICSState *ics, int srcno) > +static void ics_write_xive_msi(ICSState *ics, int srcno) > { > ICSIRQState *irq = ics->irqs + srcno; > > @@ -486,13 +486,13 @@ static void ics_simple_write_xive_msi(ICSState *ics, int srcno) > icp_irq(ics, irq->server, srcno + ics->offset, irq->priority); > } > > -static void ics_simple_write_xive_lsi(ICSState *ics, int srcno) > +static void ics_write_xive_lsi(ICSState *ics, int srcno) > { > ics_resend_lsi(ics, srcno); > } > > -void ics_simple_write_xive(ICSState *ics, int srcno, int server, > - uint8_t priority, uint8_t saved_priority) > +void ics_write_xive(ICSState *ics, int srcno, int server, > + uint8_t priority, uint8_t saved_priority) > { > ICSIRQState *irq = ics->irqs + srcno; > > @@ -500,13 +500,12 @@ void ics_simple_write_xive(ICSState *ics, int srcno, int server, > irq->priority = priority; > irq->saved_priority = saved_priority; > > - trace_xics_ics_simple_write_xive(ics->offset + srcno, srcno, server, > - priority); > + trace_xics_ics_write_xive(ics->offset + srcno, srcno, server, priority); > > if (ics->irqs[srcno].flags & XICS_FLAGS_IRQ_LSI) { > - ics_simple_write_xive_lsi(ics, srcno); > + ics_write_xive_lsi(ics, srcno); > } else { > - ics_simple_write_xive_msi(ics, srcno); > + ics_write_xive_msi(ics, srcno); > } > } > > diff --git a/hw/intc/xics_spapr.c b/hw/intc/xics_spapr.c > index 6577be0d92..3e9444813a 100644 > --- a/hw/intc/xics_spapr.c > +++ b/hw/intc/xics_spapr.c > @@ -179,7 +179,7 @@ static void rtas_set_xive(PowerPCCPU *cpu, SpaprMachineState *spapr, > } > > srcno = nr - ics->offset; > - ics_simple_write_xive(ics, srcno, server, priority, priority); > + ics_write_xive(ics, srcno, server, priority, priority); > > rtas_st(rets, 0, RTAS_OUT_SUCCESS); > } > @@ -243,8 +243,8 @@ static void rtas_int_off(PowerPCCPU *cpu, SpaprMachineState *spapr, > } > > srcno = nr - ics->offset; > - ics_simple_write_xive(ics, srcno, ics->irqs[srcno].server, 0xff, > - ics->irqs[srcno].priority); > + ics_write_xive(ics, srcno, ics->irqs[srcno].server, 0xff, > + ics->irqs[srcno].priority); > > rtas_st(rets, 0, RTAS_OUT_SUCCESS); > } > @@ -276,9 +276,9 @@ static void rtas_int_on(PowerPCCPU *cpu, SpaprMachineState *spapr, > } > > srcno = nr - ics->offset; > - ics_simple_write_xive(ics, srcno, ics->irqs[srcno].server, > - ics->irqs[srcno].saved_priority, > - ics->irqs[srcno].saved_priority); > + ics_write_xive(ics, srcno, ics->irqs[srcno].server, > + ics->irqs[srcno].saved_priority, > + ics->irqs[srcno].saved_priority); > > rtas_st(rets, 0, RTAS_OUT_SUCCESS); > } > diff --git a/hw/ppc/pnv_psi.c b/hw/ppc/pnv_psi.c > index 88ba8e7b9b..8ea81e9d8e 100644 > --- a/hw/ppc/pnv_psi.c > +++ b/hw/ppc/pnv_psi.c > @@ -311,7 +311,7 @@ static void pnv_psi_set_xivr(PnvPsi *psi, uint32_t reg, uint64_t val) > * do for now but a more accurate implementation would instead > * use a fixed server/prio and a remapper of the generated irq. > */ > - ics_simple_write_xive(ics, src, server, prio, prio); > + ics_write_xive(ics, src, server, prio, prio); > } > > static uint64_t pnv_psi_reg_read(PnvPsi *psi, uint32_t offset, bool mmio) > @@ -514,7 +514,7 @@ static void pnv_psi_power8_realize(DeviceState *dev, Error **errp) > ics_set_irq_type(ics, i, true); > } > > - psi->qirqs = qemu_allocate_irqs(ics_simple_set_irq, ics, ics->nr_irqs); > + psi->qirqs = qemu_allocate_irqs(ics_set_irq, ics, ics->nr_irqs); > > /* XSCOM region for PSI registers */ > pnv_xscom_region_init(&psi->xscom_regs, OBJECT(dev), &pnv_psi_xscom_ops, > diff --git a/hw/ppc/spapr_irq.c b/hw/ppc/spapr_irq.c > index d8f46b6797..ac189c5796 100644 > --- a/hw/ppc/spapr_irq.c > +++ b/hw/ppc/spapr_irq.c > @@ -210,7 +210,7 @@ static void spapr_irq_set_irq_xics(void *opaque, int srcno, int val) > { > SpaprMachineState *spapr = opaque; > > - ics_simple_set_irq(spapr->ics, srcno, val); > + ics_set_irq(spapr->ics, srcno, val); > } > > static void spapr_irq_reset_xics(SpaprMachineState *spapr, Error **errp) > diff --git a/include/hw/ppc/xics.h b/include/hw/ppc/xics.h > index 0eb39c2561..92628e7cab 100644 > --- a/include/hw/ppc/xics.h > +++ b/include/hw/ppc/xics.h > @@ -168,9 +168,9 @@ uint32_t icp_accept(ICPState *ss); > uint32_t icp_ipoll(ICPState *ss, uint32_t *mfrr); > void icp_eoi(ICPState *icp, uint32_t xirr); > > -void ics_simple_write_xive(ICSState *ics, int nr, int server, > - uint8_t priority, uint8_t saved_priority); > -void ics_simple_set_irq(void *opaque, int srcno, int val); > +void ics_write_xive(ICSState *ics, int nr, int server, > + uint8_t priority, uint8_t saved_priority); > +void ics_set_irq(void *opaque, int srcno, int val); > > static inline bool ics_irq_free(ICSState *ics, uint32_t srcno) > { ^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH 4/4] xics: Merge TYPE_ICS_BASE and TYPE_ICS_SIMPLE classes 2019-09-24 4:59 [PATCH 0/4] xics: Eliminate unnecessary class David Gibson ` (2 preceding siblings ...) 2019-09-24 4:59 ` [PATCH 3/4] xics: Rename misleading ics_simple_*() functions David Gibson @ 2019-09-24 4:59 ` David Gibson 2019-09-24 5:31 ` Cédric Le Goater ` (2 more replies) 2019-09-24 5:22 ` [PATCH 0/4] xics: Eliminate unnecessary class Cédric Le Goater 2019-09-24 9:47 ` Philippe Mathieu-Daudé 5 siblings, 3 replies; 30+ messages in thread From: David Gibson @ 2019-09-24 4:59 UTC (permalink / raw) To: qemu-ppc, clg; +Cc: gkurz, qemu-devel, David Gibson TYPE_ICS_SIMPLE is the only subtype of TYPE_ICS_BASE that's ever instantiated, and the only one we're ever likely to want. The existence of different classes is just a hang over from when we (misguidedly) had separate subtypes for the KVM and non-KVM version of the device. So, collapse the two classes together into just TYPE_ICS. Signed-off-by: David Gibson <david@gibson.dropbear.id.au> --- hw/intc/xics.c | 57 ++++++++++++++++--------------------------- hw/ppc/pnv_psi.c | 2 +- hw/ppc/spapr_irq.c | 4 +-- include/hw/ppc/xics.h | 17 ++----------- 4 files changed, 26 insertions(+), 54 deletions(-) diff --git a/hw/intc/xics.c b/hw/intc/xics.c index 9ae51bbc76..388dbba870 100644 --- a/hw/intc/xics.c +++ b/hw/intc/xics.c @@ -555,7 +555,7 @@ static void ics_reset_irq(ICSIRQState *irq) static void ics_reset(DeviceState *dev) { - ICSState *ics = ICS_BASE(dev); + ICSState *ics = ICS(dev); int i; uint8_t flags[ics->nr_irqs]; @@ -573,7 +573,7 @@ static void ics_reset(DeviceState *dev) if (kvm_irqchip_in_kernel()) { Error *local_err = NULL; - ics_set_kvm_state(ICS_BASE(dev), &local_err); + ics_set_kvm_state(ICS(dev), &local_err); if (local_err) { error_report_err(local_err); } @@ -587,7 +587,7 @@ static void ics_reset_handler(void *dev) static void ics_realize(DeviceState *dev, Error **errp) { - ICSState *ics = ICS_BASE(dev); + ICSState *ics = ICS(dev); Error *local_err = NULL; Object *obj; @@ -609,26 +609,14 @@ static void ics_realize(DeviceState *dev, Error **errp) qemu_register_reset(ics_reset_handler, ics); } -static void ics_simple_class_init(ObjectClass *klass, void *data) +static void ics_instance_init(Object *obj) { -} - -static const TypeInfo ics_simple_info = { - .name = TYPE_ICS_SIMPLE, - .parent = TYPE_ICS_BASE, - .instance_size = sizeof(ICSState), - .class_init = ics_simple_class_init, - .class_size = sizeof(ICSStateClass), -}; - -static void ics_base_instance_init(Object *obj) -{ - ICSState *ics = ICS_BASE(obj); + ICSState *ics = ICS(obj); ics->offset = XICS_IRQ_BASE; } -static int ics_base_pre_save(void *opaque) +static int ics_pre_save(void *opaque) { ICSState *ics = opaque; @@ -639,7 +627,7 @@ static int ics_base_pre_save(void *opaque) return 0; } -static int ics_base_post_load(void *opaque, int version_id) +static int ics_post_load(void *opaque, int version_id) { ICSState *ics = opaque; @@ -657,7 +645,7 @@ static int ics_base_post_load(void *opaque, int version_id) return 0; } -static const VMStateDescription vmstate_ics_base_irq = { +static const VMStateDescription vmstate_ics_irq = { .name = "ics/irq", .version_id = 2, .minimum_version_id = 1, @@ -671,46 +659,44 @@ static const VMStateDescription vmstate_ics_base_irq = { }, }; -static const VMStateDescription vmstate_ics_base = { +static const VMStateDescription vmstate_ics = { .name = "ics", .version_id = 1, .minimum_version_id = 1, - .pre_save = ics_base_pre_save, - .post_load = ics_base_post_load, + .pre_save = ics_pre_save, + .post_load = ics_post_load, .fields = (VMStateField[]) { /* Sanity check */ VMSTATE_UINT32_EQUAL(nr_irqs, ICSState, NULL), VMSTATE_STRUCT_VARRAY_POINTER_UINT32(irqs, ICSState, nr_irqs, - vmstate_ics_base_irq, + vmstate_ics_irq, ICSIRQState), VMSTATE_END_OF_LIST() }, }; -static Property ics_base_properties[] = { +static Property ics_properties[] = { DEFINE_PROP_UINT32("nr-irqs", ICSState, nr_irqs, 0), DEFINE_PROP_END_OF_LIST(), }; -static void ics_base_class_init(ObjectClass *klass, void *data) +static void ics_class_init(ObjectClass *klass, void *data) { DeviceClass *dc = DEVICE_CLASS(klass); dc->realize = ics_realize; - dc->props = ics_base_properties; + dc->props = ics_properties; dc->reset = ics_reset; - dc->vmsd = &vmstate_ics_base; + dc->vmsd = &vmstate_ics; } -static const TypeInfo ics_base_info = { - .name = TYPE_ICS_BASE, +static const TypeInfo ics_info = { + .name = TYPE_ICS, .parent = TYPE_DEVICE, - .abstract = true, .instance_size = sizeof(ICSState), - .instance_init = ics_base_instance_init, - .class_init = ics_base_class_init, - .class_size = sizeof(ICSStateClass), + .instance_init = ics_instance_init, + .class_init = ics_class_init, }; static const TypeInfo xics_fabric_info = { @@ -749,8 +735,7 @@ void ics_set_irq_type(ICSState *ics, int srcno, bool lsi) static void xics_register_types(void) { - type_register_static(&ics_simple_info); - type_register_static(&ics_base_info); + type_register_static(&ics_info); type_register_static(&icp_info); type_register_static(&xics_fabric_info); } diff --git a/hw/ppc/pnv_psi.c b/hw/ppc/pnv_psi.c index 8ea81e9d8e..a997f16bb4 100644 --- a/hw/ppc/pnv_psi.c +++ b/hw/ppc/pnv_psi.c @@ -469,7 +469,7 @@ static void pnv_psi_power8_instance_init(Object *obj) Pnv8Psi *psi8 = PNV8_PSI(obj); object_initialize_child(obj, "ics-psi", &psi8->ics, sizeof(psi8->ics), - TYPE_ICS_SIMPLE, &error_abort, NULL); + TYPE_ICS, &error_abort, NULL); } static const uint8_t irq_to_xivr[] = { diff --git a/hw/ppc/spapr_irq.c b/hw/ppc/spapr_irq.c index ac189c5796..6c45d2a3c0 100644 --- a/hw/ppc/spapr_irq.c +++ b/hw/ppc/spapr_irq.c @@ -98,7 +98,7 @@ static void spapr_irq_init_xics(SpaprMachineState *spapr, int nr_irqs, Object *obj; Error *local_err = NULL; - obj = object_new(TYPE_ICS_SIMPLE); + obj = object_new(TYPE_ICS); object_property_add_child(OBJECT(spapr), "ics", obj, &error_abort); object_property_add_const_link(obj, ICS_PROP_XICS, OBJECT(spapr), &error_fatal); @@ -109,7 +109,7 @@ static void spapr_irq_init_xics(SpaprMachineState *spapr, int nr_irqs, return; } - spapr->ics = ICS_BASE(obj); + spapr->ics = ICS(obj); xics_spapr_init(spapr); } diff --git a/include/hw/ppc/xics.h b/include/hw/ppc/xics.h index 92628e7cab..d8cf206a69 100644 --- a/include/hw/ppc/xics.h +++ b/include/hw/ppc/xics.h @@ -89,21 +89,8 @@ struct PnvICPState { uint32_t links[3]; }; -#define TYPE_ICS_BASE "ics-base" -#define ICS_BASE(obj) OBJECT_CHECK(ICSState, (obj), TYPE_ICS_BASE) - -/* Retain ics for sPAPR for migration from existing sPAPR guests */ -#define TYPE_ICS_SIMPLE "ics" -#define ICS_SIMPLE(obj) OBJECT_CHECK(ICSState, (obj), TYPE_ICS_SIMPLE) - -#define ICS_BASE_CLASS(klass) \ - OBJECT_CLASS_CHECK(ICSStateClass, (klass), TYPE_ICS_BASE) -#define ICS_BASE_GET_CLASS(obj) \ - OBJECT_GET_CLASS(ICSStateClass, (obj), TYPE_ICS_BASE) - -struct ICSStateClass { - DeviceClass parent_class; -}; +#define TYPE_ICS "ics" +#define ICS(obj) OBJECT_CHECK(ICSState, (obj), TYPE_ICS) struct ICSState { /*< private >*/ -- 2.21.0 ^ permalink raw reply related [flat|nested] 30+ messages in thread
* Re: [PATCH 4/4] xics: Merge TYPE_ICS_BASE and TYPE_ICS_SIMPLE classes 2019-09-24 4:59 ` [PATCH 4/4] xics: Merge TYPE_ICS_BASE and TYPE_ICS_SIMPLE classes David Gibson @ 2019-09-24 5:31 ` Cédric Le Goater 2019-09-24 11:41 ` David Gibson 2019-09-24 7:40 ` Greg Kurz 2019-09-24 9:46 ` Philippe Mathieu-Daudé 2 siblings, 1 reply; 30+ messages in thread From: Cédric Le Goater @ 2019-09-24 5:31 UTC (permalink / raw) To: David Gibson, qemu-ppc; +Cc: gkurz, qemu-devel On 24/09/2019 06:59, David Gibson wrote: > TYPE_ICS_SIMPLE is the only subtype of TYPE_ICS_BASE that's ever > instantiated, and the only one we're ever likely to want. The > existence of different classes is just a hang over from when we > (misguidedly) had separate subtypes for the KVM and non-KVM version of > the device. > > So, collapse the two classes together into just TYPE_ICS. Well, I have been maintaining another subclass for the PHB3 MSI but it has never been merged and it will require some rework. Anyhow the base ICS code is cleaner with that patch and it does not seem to break migration. > > Signed-off-by: David Gibson <david@gibson.dropbear.id.au> Reviewed-by: Cédric Le Goater <clg@kaod.org> C. > --- > hw/intc/xics.c | 57 ++++++++++++++++--------------------------- > hw/ppc/pnv_psi.c | 2 +- > hw/ppc/spapr_irq.c | 4 +-- > include/hw/ppc/xics.h | 17 ++----------- > 4 files changed, 26 insertions(+), 54 deletions(-) > > diff --git a/hw/intc/xics.c b/hw/intc/xics.c > index 9ae51bbc76..388dbba870 100644 > --- a/hw/intc/xics.c > +++ b/hw/intc/xics.c > @@ -555,7 +555,7 @@ static void ics_reset_irq(ICSIRQState *irq) > > static void ics_reset(DeviceState *dev) > { > - ICSState *ics = ICS_BASE(dev); > + ICSState *ics = ICS(dev); > int i; > uint8_t flags[ics->nr_irqs]; > > @@ -573,7 +573,7 @@ static void ics_reset(DeviceState *dev) > if (kvm_irqchip_in_kernel()) { > Error *local_err = NULL; > > - ics_set_kvm_state(ICS_BASE(dev), &local_err); > + ics_set_kvm_state(ICS(dev), &local_err); > if (local_err) { > error_report_err(local_err); > } > @@ -587,7 +587,7 @@ static void ics_reset_handler(void *dev) > > static void ics_realize(DeviceState *dev, Error **errp) > { > - ICSState *ics = ICS_BASE(dev); > + ICSState *ics = ICS(dev); > Error *local_err = NULL; > Object *obj; > > @@ -609,26 +609,14 @@ static void ics_realize(DeviceState *dev, Error **errp) > qemu_register_reset(ics_reset_handler, ics); > } > > -static void ics_simple_class_init(ObjectClass *klass, void *data) > +static void ics_instance_init(Object *obj) > { > -} > - > -static const TypeInfo ics_simple_info = { > - .name = TYPE_ICS_SIMPLE, > - .parent = TYPE_ICS_BASE, > - .instance_size = sizeof(ICSState), > - .class_init = ics_simple_class_init, > - .class_size = sizeof(ICSStateClass), > -}; > - > -static void ics_base_instance_init(Object *obj) > -{ > - ICSState *ics = ICS_BASE(obj); > + ICSState *ics = ICS(obj); > > ics->offset = XICS_IRQ_BASE; > } > > -static int ics_base_pre_save(void *opaque) > +static int ics_pre_save(void *opaque) > { > ICSState *ics = opaque; > > @@ -639,7 +627,7 @@ static int ics_base_pre_save(void *opaque) > return 0; > } > > -static int ics_base_post_load(void *opaque, int version_id) > +static int ics_post_load(void *opaque, int version_id) > { > ICSState *ics = opaque; > > @@ -657,7 +645,7 @@ static int ics_base_post_load(void *opaque, int version_id) > return 0; > } > > -static const VMStateDescription vmstate_ics_base_irq = { > +static const VMStateDescription vmstate_ics_irq = { > .name = "ics/irq", > .version_id = 2, > .minimum_version_id = 1, > @@ -671,46 +659,44 @@ static const VMStateDescription vmstate_ics_base_irq = { > }, > }; > > -static const VMStateDescription vmstate_ics_base = { > +static const VMStateDescription vmstate_ics = { > .name = "ics", > .version_id = 1, > .minimum_version_id = 1, > - .pre_save = ics_base_pre_save, > - .post_load = ics_base_post_load, > + .pre_save = ics_pre_save, > + .post_load = ics_post_load, > .fields = (VMStateField[]) { > /* Sanity check */ > VMSTATE_UINT32_EQUAL(nr_irqs, ICSState, NULL), > > VMSTATE_STRUCT_VARRAY_POINTER_UINT32(irqs, ICSState, nr_irqs, > - vmstate_ics_base_irq, > + vmstate_ics_irq, > ICSIRQState), > VMSTATE_END_OF_LIST() > }, > }; > > -static Property ics_base_properties[] = { > +static Property ics_properties[] = { > DEFINE_PROP_UINT32("nr-irqs", ICSState, nr_irqs, 0), > DEFINE_PROP_END_OF_LIST(), > }; > > -static void ics_base_class_init(ObjectClass *klass, void *data) > +static void ics_class_init(ObjectClass *klass, void *data) > { > DeviceClass *dc = DEVICE_CLASS(klass); > > dc->realize = ics_realize; > - dc->props = ics_base_properties; > + dc->props = ics_properties; > dc->reset = ics_reset; > - dc->vmsd = &vmstate_ics_base; > + dc->vmsd = &vmstate_ics; > } > > -static const TypeInfo ics_base_info = { > - .name = TYPE_ICS_BASE, > +static const TypeInfo ics_info = { > + .name = TYPE_ICS, > .parent = TYPE_DEVICE, > - .abstract = true, > .instance_size = sizeof(ICSState), > - .instance_init = ics_base_instance_init, > - .class_init = ics_base_class_init, > - .class_size = sizeof(ICSStateClass), > + .instance_init = ics_instance_init, > + .class_init = ics_class_init, > }; > > static const TypeInfo xics_fabric_info = { > @@ -749,8 +735,7 @@ void ics_set_irq_type(ICSState *ics, int srcno, bool lsi) > > static void xics_register_types(void) > { > - type_register_static(&ics_simple_info); > - type_register_static(&ics_base_info); > + type_register_static(&ics_info); > type_register_static(&icp_info); > type_register_static(&xics_fabric_info); > } > diff --git a/hw/ppc/pnv_psi.c b/hw/ppc/pnv_psi.c > index 8ea81e9d8e..a997f16bb4 100644 > --- a/hw/ppc/pnv_psi.c > +++ b/hw/ppc/pnv_psi.c > @@ -469,7 +469,7 @@ static void pnv_psi_power8_instance_init(Object *obj) > Pnv8Psi *psi8 = PNV8_PSI(obj); > > object_initialize_child(obj, "ics-psi", &psi8->ics, sizeof(psi8->ics), > - TYPE_ICS_SIMPLE, &error_abort, NULL); > + TYPE_ICS, &error_abort, NULL); > } > > static const uint8_t irq_to_xivr[] = { > diff --git a/hw/ppc/spapr_irq.c b/hw/ppc/spapr_irq.c > index ac189c5796..6c45d2a3c0 100644 > --- a/hw/ppc/spapr_irq.c > +++ b/hw/ppc/spapr_irq.c > @@ -98,7 +98,7 @@ static void spapr_irq_init_xics(SpaprMachineState *spapr, int nr_irqs, > Object *obj; > Error *local_err = NULL; > > - obj = object_new(TYPE_ICS_SIMPLE); > + obj = object_new(TYPE_ICS); > object_property_add_child(OBJECT(spapr), "ics", obj, &error_abort); > object_property_add_const_link(obj, ICS_PROP_XICS, OBJECT(spapr), > &error_fatal); > @@ -109,7 +109,7 @@ static void spapr_irq_init_xics(SpaprMachineState *spapr, int nr_irqs, > return; > } > > - spapr->ics = ICS_BASE(obj); > + spapr->ics = ICS(obj); > > xics_spapr_init(spapr); > } > diff --git a/include/hw/ppc/xics.h b/include/hw/ppc/xics.h > index 92628e7cab..d8cf206a69 100644 > --- a/include/hw/ppc/xics.h > +++ b/include/hw/ppc/xics.h > @@ -89,21 +89,8 @@ struct PnvICPState { > uint32_t links[3]; > }; > > -#define TYPE_ICS_BASE "ics-base" > -#define ICS_BASE(obj) OBJECT_CHECK(ICSState, (obj), TYPE_ICS_BASE) > - > -/* Retain ics for sPAPR for migration from existing sPAPR guests */ > -#define TYPE_ICS_SIMPLE "ics" > -#define ICS_SIMPLE(obj) OBJECT_CHECK(ICSState, (obj), TYPE_ICS_SIMPLE) > - > -#define ICS_BASE_CLASS(klass) \ > - OBJECT_CLASS_CHECK(ICSStateClass, (klass), TYPE_ICS_BASE) > -#define ICS_BASE_GET_CLASS(obj) \ > - OBJECT_GET_CLASS(ICSStateClass, (obj), TYPE_ICS_BASE) > - > -struct ICSStateClass { > - DeviceClass parent_class; > -}; > +#define TYPE_ICS "ics" > +#define ICS(obj) OBJECT_CHECK(ICSState, (obj), TYPE_ICS) > > struct ICSState { > /*< private >*/ > ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 4/4] xics: Merge TYPE_ICS_BASE and TYPE_ICS_SIMPLE classes 2019-09-24 5:31 ` Cédric Le Goater @ 2019-09-24 11:41 ` David Gibson 2019-09-24 14:06 ` Cédric Le Goater 0 siblings, 1 reply; 30+ messages in thread From: David Gibson @ 2019-09-24 11:41 UTC (permalink / raw) To: Cédric Le Goater; +Cc: gkurz, qemu-ppc, qemu-devel [-- Attachment #1: Type: text/plain, Size: 9084 bytes --] On Tue, Sep 24, 2019 at 07:31:44AM +0200, Cédric Le Goater wrote: > On 24/09/2019 06:59, David Gibson wrote: > > TYPE_ICS_SIMPLE is the only subtype of TYPE_ICS_BASE that's ever > > instantiated, and the only one we're ever likely to want. The > > existence of different classes is just a hang over from when we > > (misguidedly) had separate subtypes for the KVM and non-KVM version of > > the device. > > > > So, collapse the two classes together into just TYPE_ICS. > > > Well, I have been maintaining another subclass for the PHB3 MSI > but it has never been merged and it will require some rework. Well, if you did do this again, is there an actual need for it to be a subclass of ICS_BASE, and not ICS_SIMPLE? AFAICT the merged ICS class should be fine for pnv as well. > Anyhow the base ICS code is cleaner with that patch and it > does not seem to break migration. > > > > > Signed-off-by: David Gibson <david@gibson.dropbear.id.au> > > > Reviewed-by: Cédric Le Goater <clg@kaod.org> > > C. > > > > --- > > hw/intc/xics.c | 57 ++++++++++++++++--------------------------- > > hw/ppc/pnv_psi.c | 2 +- > > hw/ppc/spapr_irq.c | 4 +-- > > include/hw/ppc/xics.h | 17 ++----------- > > 4 files changed, 26 insertions(+), 54 deletions(-) > > > > diff --git a/hw/intc/xics.c b/hw/intc/xics.c > > index 9ae51bbc76..388dbba870 100644 > > --- a/hw/intc/xics.c > > +++ b/hw/intc/xics.c > > @@ -555,7 +555,7 @@ static void ics_reset_irq(ICSIRQState *irq) > > > > static void ics_reset(DeviceState *dev) > > { > > - ICSState *ics = ICS_BASE(dev); > > + ICSState *ics = ICS(dev); > > int i; > > uint8_t flags[ics->nr_irqs]; > > > > @@ -573,7 +573,7 @@ static void ics_reset(DeviceState *dev) > > if (kvm_irqchip_in_kernel()) { > > Error *local_err = NULL; > > > > - ics_set_kvm_state(ICS_BASE(dev), &local_err); > > + ics_set_kvm_state(ICS(dev), &local_err); > > if (local_err) { > > error_report_err(local_err); > > } > > @@ -587,7 +587,7 @@ static void ics_reset_handler(void *dev) > > > > static void ics_realize(DeviceState *dev, Error **errp) > > { > > - ICSState *ics = ICS_BASE(dev); > > + ICSState *ics = ICS(dev); > > Error *local_err = NULL; > > Object *obj; > > > > @@ -609,26 +609,14 @@ static void ics_realize(DeviceState *dev, Error **errp) > > qemu_register_reset(ics_reset_handler, ics); > > } > > > > -static void ics_simple_class_init(ObjectClass *klass, void *data) > > +static void ics_instance_init(Object *obj) > > { > > -} > > - > > -static const TypeInfo ics_simple_info = { > > - .name = TYPE_ICS_SIMPLE, > > - .parent = TYPE_ICS_BASE, > > - .instance_size = sizeof(ICSState), > > - .class_init = ics_simple_class_init, > > - .class_size = sizeof(ICSStateClass), > > -}; > > - > > -static void ics_base_instance_init(Object *obj) > > -{ > > - ICSState *ics = ICS_BASE(obj); > > + ICSState *ics = ICS(obj); > > > > ics->offset = XICS_IRQ_BASE; > > } > > > > -static int ics_base_pre_save(void *opaque) > > +static int ics_pre_save(void *opaque) > > { > > ICSState *ics = opaque; > > > > @@ -639,7 +627,7 @@ static int ics_base_pre_save(void *opaque) > > return 0; > > } > > > > -static int ics_base_post_load(void *opaque, int version_id) > > +static int ics_post_load(void *opaque, int version_id) > > { > > ICSState *ics = opaque; > > > > @@ -657,7 +645,7 @@ static int ics_base_post_load(void *opaque, int version_id) > > return 0; > > } > > > > -static const VMStateDescription vmstate_ics_base_irq = { > > +static const VMStateDescription vmstate_ics_irq = { > > .name = "ics/irq", > > .version_id = 2, > > .minimum_version_id = 1, > > @@ -671,46 +659,44 @@ static const VMStateDescription vmstate_ics_base_irq = { > > }, > > }; > > > > -static const VMStateDescription vmstate_ics_base = { > > +static const VMStateDescription vmstate_ics = { > > .name = "ics", > > .version_id = 1, > > .minimum_version_id = 1, > > - .pre_save = ics_base_pre_save, > > - .post_load = ics_base_post_load, > > + .pre_save = ics_pre_save, > > + .post_load = ics_post_load, > > .fields = (VMStateField[]) { > > /* Sanity check */ > > VMSTATE_UINT32_EQUAL(nr_irqs, ICSState, NULL), > > > > VMSTATE_STRUCT_VARRAY_POINTER_UINT32(irqs, ICSState, nr_irqs, > > - vmstate_ics_base_irq, > > + vmstate_ics_irq, > > ICSIRQState), > > VMSTATE_END_OF_LIST() > > }, > > }; > > > > -static Property ics_base_properties[] = { > > +static Property ics_properties[] = { > > DEFINE_PROP_UINT32("nr-irqs", ICSState, nr_irqs, 0), > > DEFINE_PROP_END_OF_LIST(), > > }; > > > > -static void ics_base_class_init(ObjectClass *klass, void *data) > > +static void ics_class_init(ObjectClass *klass, void *data) > > { > > DeviceClass *dc = DEVICE_CLASS(klass); > > > > dc->realize = ics_realize; > > - dc->props = ics_base_properties; > > + dc->props = ics_properties; > > dc->reset = ics_reset; > > - dc->vmsd = &vmstate_ics_base; > > + dc->vmsd = &vmstate_ics; > > } > > > > -static const TypeInfo ics_base_info = { > > - .name = TYPE_ICS_BASE, > > +static const TypeInfo ics_info = { > > + .name = TYPE_ICS, > > .parent = TYPE_DEVICE, > > - .abstract = true, > > .instance_size = sizeof(ICSState), > > - .instance_init = ics_base_instance_init, > > - .class_init = ics_base_class_init, > > - .class_size = sizeof(ICSStateClass), > > + .instance_init = ics_instance_init, > > + .class_init = ics_class_init, > > }; > > > > static const TypeInfo xics_fabric_info = { > > @@ -749,8 +735,7 @@ void ics_set_irq_type(ICSState *ics, int srcno, bool lsi) > > > > static void xics_register_types(void) > > { > > - type_register_static(&ics_simple_info); > > - type_register_static(&ics_base_info); > > + type_register_static(&ics_info); > > type_register_static(&icp_info); > > type_register_static(&xics_fabric_info); > > } > > diff --git a/hw/ppc/pnv_psi.c b/hw/ppc/pnv_psi.c > > index 8ea81e9d8e..a997f16bb4 100644 > > --- a/hw/ppc/pnv_psi.c > > +++ b/hw/ppc/pnv_psi.c > > @@ -469,7 +469,7 @@ static void pnv_psi_power8_instance_init(Object *obj) > > Pnv8Psi *psi8 = PNV8_PSI(obj); > > > > object_initialize_child(obj, "ics-psi", &psi8->ics, sizeof(psi8->ics), > > - TYPE_ICS_SIMPLE, &error_abort, NULL); > > + TYPE_ICS, &error_abort, NULL); > > } > > > > static const uint8_t irq_to_xivr[] = { > > diff --git a/hw/ppc/spapr_irq.c b/hw/ppc/spapr_irq.c > > index ac189c5796..6c45d2a3c0 100644 > > --- a/hw/ppc/spapr_irq.c > > +++ b/hw/ppc/spapr_irq.c > > @@ -98,7 +98,7 @@ static void spapr_irq_init_xics(SpaprMachineState *spapr, int nr_irqs, > > Object *obj; > > Error *local_err = NULL; > > > > - obj = object_new(TYPE_ICS_SIMPLE); > > + obj = object_new(TYPE_ICS); > > object_property_add_child(OBJECT(spapr), "ics", obj, &error_abort); > > object_property_add_const_link(obj, ICS_PROP_XICS, OBJECT(spapr), > > &error_fatal); > > @@ -109,7 +109,7 @@ static void spapr_irq_init_xics(SpaprMachineState *spapr, int nr_irqs, > > return; > > } > > > > - spapr->ics = ICS_BASE(obj); > > + spapr->ics = ICS(obj); > > > > xics_spapr_init(spapr); > > } > > diff --git a/include/hw/ppc/xics.h b/include/hw/ppc/xics.h > > index 92628e7cab..d8cf206a69 100644 > > --- a/include/hw/ppc/xics.h > > +++ b/include/hw/ppc/xics.h > > @@ -89,21 +89,8 @@ struct PnvICPState { > > uint32_t links[3]; > > }; > > > > -#define TYPE_ICS_BASE "ics-base" > > -#define ICS_BASE(obj) OBJECT_CHECK(ICSState, (obj), TYPE_ICS_BASE) > > - > > -/* Retain ics for sPAPR for migration from existing sPAPR guests */ > > -#define TYPE_ICS_SIMPLE "ics" > > -#define ICS_SIMPLE(obj) OBJECT_CHECK(ICSState, (obj), TYPE_ICS_SIMPLE) > > - > > -#define ICS_BASE_CLASS(klass) \ > > - OBJECT_CLASS_CHECK(ICSStateClass, (klass), TYPE_ICS_BASE) > > -#define ICS_BASE_GET_CLASS(obj) \ > > - OBJECT_GET_CLASS(ICSStateClass, (obj), TYPE_ICS_BASE) > > - > > -struct ICSStateClass { > > - DeviceClass parent_class; > > -}; > > +#define TYPE_ICS "ics" > > +#define ICS(obj) OBJECT_CHECK(ICSState, (obj), TYPE_ICS) > > > > struct ICSState { > > /*< private >*/ > > > -- David Gibson | I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_ | _way_ _around_! http://www.ozlabs.org/~dgibson [-- Attachment #2: signature.asc --] [-- Type: application/pgp-signature, Size: 833 bytes --] ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 4/4] xics: Merge TYPE_ICS_BASE and TYPE_ICS_SIMPLE classes 2019-09-24 11:41 ` David Gibson @ 2019-09-24 14:06 ` Cédric Le Goater 2019-09-25 1:46 ` David Gibson 0 siblings, 1 reply; 30+ messages in thread From: Cédric Le Goater @ 2019-09-24 14:06 UTC (permalink / raw) To: David Gibson; +Cc: Philippe Mathieu-Daudé, gkurz, qemu-ppc, qemu-devel On 24/09/2019 13:41, David Gibson wrote: > On Tue, Sep 24, 2019 at 07:31:44AM +0200, Cédric Le Goater wrote: >> On 24/09/2019 06:59, David Gibson wrote: >>> TYPE_ICS_SIMPLE is the only subtype of TYPE_ICS_BASE that's ever >>> instantiated, and the only one we're ever likely to want. The >>> existence of different classes is just a hang over from when we >>> (misguidedly) had separate subtypes for the KVM and non-KVM version of >>> the device. >>> >>> So, collapse the two classes together into just TYPE_ICS. >> >> >> Well, I have been maintaining another subclass for the PHB3 MSI >> but it has never been merged and it will require some rework. > > Well, if you did do this again, is there an actual need for it to be a > subclass of ICS_BASE, and not ICS_SIMPLE? AFAICT the merged ICS class > should be fine for pnv as well. the reject resend handlers might be an issue. Anyhow, let's merge this cleanup. PHB3 has been out of tree for too long. C. >> Anyhow the base ICS code is cleaner with that patch and it >> does not seem to break migration. >> >>> >>> Signed-off-by: David Gibson <david@gibson.dropbear.id.au> >> >> >> Reviewed-by: Cédric Le Goater <clg@kaod.org> >> >> C. >> >> >>> --- >>> hw/intc/xics.c | 57 ++++++++++++++++--------------------------- >>> hw/ppc/pnv_psi.c | 2 +- >>> hw/ppc/spapr_irq.c | 4 +-- >>> include/hw/ppc/xics.h | 17 ++----------- >>> 4 files changed, 26 insertions(+), 54 deletions(-) >>> >>> diff --git a/hw/intc/xics.c b/hw/intc/xics.c >>> index 9ae51bbc76..388dbba870 100644 >>> --- a/hw/intc/xics.c >>> +++ b/hw/intc/xics.c >>> @@ -555,7 +555,7 @@ static void ics_reset_irq(ICSIRQState *irq) >>> >>> static void ics_reset(DeviceState *dev) >>> { >>> - ICSState *ics = ICS_BASE(dev); >>> + ICSState *ics = ICS(dev); >>> int i; >>> uint8_t flags[ics->nr_irqs]; >>> >>> @@ -573,7 +573,7 @@ static void ics_reset(DeviceState *dev) >>> if (kvm_irqchip_in_kernel()) { >>> Error *local_err = NULL; >>> >>> - ics_set_kvm_state(ICS_BASE(dev), &local_err); >>> + ics_set_kvm_state(ICS(dev), &local_err); >>> if (local_err) { >>> error_report_err(local_err); >>> } >>> @@ -587,7 +587,7 @@ static void ics_reset_handler(void *dev) >>> >>> static void ics_realize(DeviceState *dev, Error **errp) >>> { >>> - ICSState *ics = ICS_BASE(dev); >>> + ICSState *ics = ICS(dev); >>> Error *local_err = NULL; >>> Object *obj; >>> >>> @@ -609,26 +609,14 @@ static void ics_realize(DeviceState *dev, Error **errp) >>> qemu_register_reset(ics_reset_handler, ics); >>> } >>> >>> -static void ics_simple_class_init(ObjectClass *klass, void *data) >>> +static void ics_instance_init(Object *obj) >>> { >>> -} >>> - >>> -static const TypeInfo ics_simple_info = { >>> - .name = TYPE_ICS_SIMPLE, >>> - .parent = TYPE_ICS_BASE, >>> - .instance_size = sizeof(ICSState), >>> - .class_init = ics_simple_class_init, >>> - .class_size = sizeof(ICSStateClass), >>> -}; >>> - >>> -static void ics_base_instance_init(Object *obj) >>> -{ >>> - ICSState *ics = ICS_BASE(obj); >>> + ICSState *ics = ICS(obj); >>> >>> ics->offset = XICS_IRQ_BASE; >>> } >>> >>> -static int ics_base_pre_save(void *opaque) >>> +static int ics_pre_save(void *opaque) >>> { >>> ICSState *ics = opaque; >>> >>> @@ -639,7 +627,7 @@ static int ics_base_pre_save(void *opaque) >>> return 0; >>> } >>> >>> -static int ics_base_post_load(void *opaque, int version_id) >>> +static int ics_post_load(void *opaque, int version_id) >>> { >>> ICSState *ics = opaque; >>> >>> @@ -657,7 +645,7 @@ static int ics_base_post_load(void *opaque, int version_id) >>> return 0; >>> } >>> >>> -static const VMStateDescription vmstate_ics_base_irq = { >>> +static const VMStateDescription vmstate_ics_irq = { >>> .name = "ics/irq", >>> .version_id = 2, >>> .minimum_version_id = 1, >>> @@ -671,46 +659,44 @@ static const VMStateDescription vmstate_ics_base_irq = { >>> }, >>> }; >>> >>> -static const VMStateDescription vmstate_ics_base = { >>> +static const VMStateDescription vmstate_ics = { >>> .name = "ics", >>> .version_id = 1, >>> .minimum_version_id = 1, >>> - .pre_save = ics_base_pre_save, >>> - .post_load = ics_base_post_load, >>> + .pre_save = ics_pre_save, >>> + .post_load = ics_post_load, >>> .fields = (VMStateField[]) { >>> /* Sanity check */ >>> VMSTATE_UINT32_EQUAL(nr_irqs, ICSState, NULL), >>> >>> VMSTATE_STRUCT_VARRAY_POINTER_UINT32(irqs, ICSState, nr_irqs, >>> - vmstate_ics_base_irq, >>> + vmstate_ics_irq, >>> ICSIRQState), >>> VMSTATE_END_OF_LIST() >>> }, >>> }; >>> >>> -static Property ics_base_properties[] = { >>> +static Property ics_properties[] = { >>> DEFINE_PROP_UINT32("nr-irqs", ICSState, nr_irqs, 0), >>> DEFINE_PROP_END_OF_LIST(), >>> }; >>> >>> -static void ics_base_class_init(ObjectClass *klass, void *data) >>> +static void ics_class_init(ObjectClass *klass, void *data) >>> { >>> DeviceClass *dc = DEVICE_CLASS(klass); >>> >>> dc->realize = ics_realize; >>> - dc->props = ics_base_properties; >>> + dc->props = ics_properties; >>> dc->reset = ics_reset; >>> - dc->vmsd = &vmstate_ics_base; >>> + dc->vmsd = &vmstate_ics; >>> } >>> >>> -static const TypeInfo ics_base_info = { >>> - .name = TYPE_ICS_BASE, >>> +static const TypeInfo ics_info = { >>> + .name = TYPE_ICS, >>> .parent = TYPE_DEVICE, >>> - .abstract = true, >>> .instance_size = sizeof(ICSState), >>> - .instance_init = ics_base_instance_init, >>> - .class_init = ics_base_class_init, >>> - .class_size = sizeof(ICSStateClass), >>> + .instance_init = ics_instance_init, >>> + .class_init = ics_class_init, >>> }; >>> >>> static const TypeInfo xics_fabric_info = { >>> @@ -749,8 +735,7 @@ void ics_set_irq_type(ICSState *ics, int srcno, bool lsi) >>> >>> static void xics_register_types(void) >>> { >>> - type_register_static(&ics_simple_info); >>> - type_register_static(&ics_base_info); >>> + type_register_static(&ics_info); >>> type_register_static(&icp_info); >>> type_register_static(&xics_fabric_info); >>> } >>> diff --git a/hw/ppc/pnv_psi.c b/hw/ppc/pnv_psi.c >>> index 8ea81e9d8e..a997f16bb4 100644 >>> --- a/hw/ppc/pnv_psi.c >>> +++ b/hw/ppc/pnv_psi.c >>> @@ -469,7 +469,7 @@ static void pnv_psi_power8_instance_init(Object *obj) >>> Pnv8Psi *psi8 = PNV8_PSI(obj); >>> >>> object_initialize_child(obj, "ics-psi", &psi8->ics, sizeof(psi8->ics), >>> - TYPE_ICS_SIMPLE, &error_abort, NULL); >>> + TYPE_ICS, &error_abort, NULL); >>> } >>> >>> static const uint8_t irq_to_xivr[] = { >>> diff --git a/hw/ppc/spapr_irq.c b/hw/ppc/spapr_irq.c >>> index ac189c5796..6c45d2a3c0 100644 >>> --- a/hw/ppc/spapr_irq.c >>> +++ b/hw/ppc/spapr_irq.c >>> @@ -98,7 +98,7 @@ static void spapr_irq_init_xics(SpaprMachineState *spapr, int nr_irqs, >>> Object *obj; >>> Error *local_err = NULL; >>> >>> - obj = object_new(TYPE_ICS_SIMPLE); >>> + obj = object_new(TYPE_ICS); >>> object_property_add_child(OBJECT(spapr), "ics", obj, &error_abort); >>> object_property_add_const_link(obj, ICS_PROP_XICS, OBJECT(spapr), >>> &error_fatal); >>> @@ -109,7 +109,7 @@ static void spapr_irq_init_xics(SpaprMachineState *spapr, int nr_irqs, >>> return; >>> } >>> >>> - spapr->ics = ICS_BASE(obj); >>> + spapr->ics = ICS(obj); >>> >>> xics_spapr_init(spapr); >>> } >>> diff --git a/include/hw/ppc/xics.h b/include/hw/ppc/xics.h >>> index 92628e7cab..d8cf206a69 100644 >>> --- a/include/hw/ppc/xics.h >>> +++ b/include/hw/ppc/xics.h >>> @@ -89,21 +89,8 @@ struct PnvICPState { >>> uint32_t links[3]; >>> }; >>> >>> -#define TYPE_ICS_BASE "ics-base" >>> -#define ICS_BASE(obj) OBJECT_CHECK(ICSState, (obj), TYPE_ICS_BASE) >>> - >>> -/* Retain ics for sPAPR for migration from existing sPAPR guests */ >>> -#define TYPE_ICS_SIMPLE "ics" >>> -#define ICS_SIMPLE(obj) OBJECT_CHECK(ICSState, (obj), TYPE_ICS_SIMPLE) >>> - >>> -#define ICS_BASE_CLASS(klass) \ >>> - OBJECT_CLASS_CHECK(ICSStateClass, (klass), TYPE_ICS_BASE) >>> -#define ICS_BASE_GET_CLASS(obj) \ >>> - OBJECT_GET_CLASS(ICSStateClass, (obj), TYPE_ICS_BASE) >>> - >>> -struct ICSStateClass { >>> - DeviceClass parent_class; >>> -}; >>> +#define TYPE_ICS "ics" >>> +#define ICS(obj) OBJECT_CHECK(ICSState, (obj), TYPE_ICS) >>> >>> struct ICSState { >>> /*< private >*/ >>> >> > ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 4/4] xics: Merge TYPE_ICS_BASE and TYPE_ICS_SIMPLE classes 2019-09-24 14:06 ` Cédric Le Goater @ 2019-09-25 1:46 ` David Gibson 2019-09-25 6:04 ` Cédric Le Goater 0 siblings, 1 reply; 30+ messages in thread From: David Gibson @ 2019-09-25 1:46 UTC (permalink / raw) To: Cédric Le Goater Cc: Philippe Mathieu-Daudé, gkurz, qemu-ppc, qemu-devel [-- Attachment #1: Type: text/plain, Size: 1501 bytes --] On Tue, Sep 24, 2019 at 04:06:02PM +0200, Cédric Le Goater wrote: > On 24/09/2019 13:41, David Gibson wrote: > > On Tue, Sep 24, 2019 at 07:31:44AM +0200, Cédric Le Goater wrote: > >> On 24/09/2019 06:59, David Gibson wrote: > >>> TYPE_ICS_SIMPLE is the only subtype of TYPE_ICS_BASE that's ever > >>> instantiated, and the only one we're ever likely to want. The > >>> existence of different classes is just a hang over from when we > >>> (misguidedly) had separate subtypes for the KVM and non-KVM version of > >>> the device. > >>> > >>> So, collapse the two classes together into just TYPE_ICS. > >> > >> > >> Well, I have been maintaining another subclass for the PHB3 MSI > >> but it has never been merged and it will require some rework. > > > > Well, if you did do this again, is there an actual need for it to be a > > subclass of ICS_BASE, and not ICS_SIMPLE? AFAICT the merged ICS class > > should be fine for pnv as well. > > the reject resend handlers might be an issue. Anyhow, let's merge this > cleanup. PHB3 has been out of tree for too long. Hrm, are you sure. I remember thinking the other day "whatever happened to that PHB3 patchset?". Is it actually broken, or has it just been a long time since it was posted, and therefore been forgotten by me. -- David Gibson | I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_ | _way_ _around_! http://www.ozlabs.org/~dgibson [-- Attachment #2: signature.asc --] [-- Type: application/pgp-signature, Size: 833 bytes --] ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 4/4] xics: Merge TYPE_ICS_BASE and TYPE_ICS_SIMPLE classes 2019-09-25 1:46 ` David Gibson @ 2019-09-25 6:04 ` Cédric Le Goater 2019-10-03 17:53 ` Cédric Le Goater 0 siblings, 1 reply; 30+ messages in thread From: Cédric Le Goater @ 2019-09-25 6:04 UTC (permalink / raw) To: David Gibson; +Cc: Philippe Mathieu-Daudé, gkurz, qemu-ppc, qemu-devel On 25/09/2019 03:46, David Gibson wrote: > On Tue, Sep 24, 2019 at 04:06:02PM +0200, Cédric Le Goater wrote: >> On 24/09/2019 13:41, David Gibson wrote: >>> On Tue, Sep 24, 2019 at 07:31:44AM +0200, Cédric Le Goater wrote: >>>> On 24/09/2019 06:59, David Gibson wrote: >>>>> TYPE_ICS_SIMPLE is the only subtype of TYPE_ICS_BASE that's ever >>>>> instantiated, and the only one we're ever likely to want. The >>>>> existence of different classes is just a hang over from when we >>>>> (misguidedly) had separate subtypes for the KVM and non-KVM version of >>>>> the device. >>>>> >>>>> So, collapse the two classes together into just TYPE_ICS. >>>> >>>> >>>> Well, I have been maintaining another subclass for the PHB3 MSI >>>> but it has never been merged and it will require some rework. >>> >>> Well, if you did do this again, is there an actual need for it to be a >>> subclass of ICS_BASE, and not ICS_SIMPLE? AFAICT the merged ICS class >>> should be fine for pnv as well. >> >> the reject resend handlers might be an issue. Anyhow, let's merge this >> cleanup. PHB3 has been out of tree for too long. > > Hrm, are you sure. I remember thinking the other day "whatever > happened to that PHB3 patchset?". Is it actually broken, It is not broken. PowerNV machines can boot rather complex PCI layouts on P8 (XICS) and P9 (XIVE). See the complex configuration examples here : https://github.com/legoater/qemu/wiki/PowerNV > or has it just been a long time since it was posted, and therefore > been forgotten by me. It hasn't been posted in a long time (+1 year). Here are the latest exchanges we had in November: http://patchwork.ozlabs.org/patch/951227/ Nothing really worrying : - some XICS infrastructure are needed (currently being removed) - check if we need a separate source model for the MSI/LSI - rework the PBCQ modeling to some extent. - misc cleanups. I have been quite busy this last year. I would appreciate if someone could take ownership of the PHB part. C. ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 4/4] xics: Merge TYPE_ICS_BASE and TYPE_ICS_SIMPLE classes 2019-09-25 6:04 ` Cédric Le Goater @ 2019-10-03 17:53 ` Cédric Le Goater 0 siblings, 0 replies; 30+ messages in thread From: Cédric Le Goater @ 2019-10-03 17:53 UTC (permalink / raw) To: David Gibson; +Cc: Philippe Mathieu-Daudé, gkurz, qemu-ppc, qemu-devel On 25/09/2019 08:04, Cédric Le Goater wrote: > On 25/09/2019 03:46, David Gibson wrote: >> On Tue, Sep 24, 2019 at 04:06:02PM +0200, Cédric Le Goater wrote: >>> On 24/09/2019 13:41, David Gibson wrote: >>>> On Tue, Sep 24, 2019 at 07:31:44AM +0200, Cédric Le Goater wrote: >>>>> On 24/09/2019 06:59, David Gibson wrote: >>>>>> TYPE_ICS_SIMPLE is the only subtype of TYPE_ICS_BASE that's ever >>>>>> instantiated, and the only one we're ever likely to want. The >>>>>> existence of different classes is just a hang over from when we >>>>>> (misguidedly) had separate subtypes for the KVM and non-KVM version of >>>>>> the device. >>>>>> >>>>>> So, collapse the two classes together into just TYPE_ICS. >>>>> >>>>> >>>>> Well, I have been maintaining another subclass for the PHB3 MSI >>>>> but it has never been merged and it will require some rework. >>>> >>>> Well, if you did do this again, is there an actual need for it to be a >>>> subclass of ICS_BASE, and not ICS_SIMPLE? AFAICT the merged ICS class >>>> should be fine for pnv as well. >>> >>> the reject resend handlers might be an issue. Anyhow, let's merge this >>> cleanup. PHB3 has been out of tree for too long. >> >> Hrm, are you sure. I remember thinking the other day "whatever >> happened to that PHB3 patchset?". Is it actually broken, > > It is not broken. > > PowerNV machines can boot rather complex PCI layouts on P8 (XICS) > and P9 (XIVE). See the complex configuration examples here : > > https://github.com/legoater/qemu/wiki/PowerNV > >> or has it just been a long time since it was posted, and therefore >> been forgotten by me. > > It hasn't been posted in a long time (+1 year). Here are the latest > exchanges we had in November: > > http://patchwork.ozlabs.org/patch/951227/ > > Nothing really worrying : > > - some XICS infrastructure are needed (currently being removed) > - check if we need a separate source model for the MSI/LSI > - rework the PBCQ modeling to some extent. > - misc cleanups. > > I have been quite busy this last year. I would appreciate if someone > could take ownership of the PHB part. Recent XICS changes have made support of PHB3 complex and I don't have time to keep it alive anymore. I am dropping the patch from my tree. In case someone wants to take over, it's here : https://github.com/legoater/qemu/commits/powernv-4.2-p8 Focus is on P9 and P10 now. Cheers, C. ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 4/4] xics: Merge TYPE_ICS_BASE and TYPE_ICS_SIMPLE classes 2019-09-24 4:59 ` [PATCH 4/4] xics: Merge TYPE_ICS_BASE and TYPE_ICS_SIMPLE classes David Gibson 2019-09-24 5:31 ` Cédric Le Goater @ 2019-09-24 7:40 ` Greg Kurz 2019-09-24 9:46 ` Philippe Mathieu-Daudé 2 siblings, 0 replies; 30+ messages in thread From: Greg Kurz @ 2019-09-24 7:40 UTC (permalink / raw) To: David Gibson; +Cc: gkurz, qemu-ppc, clg, qemu-devel On Tue, 24 Sep 2019 14:59:52 +1000 David Gibson <david@gibson.dropbear.id.au> wrote: > TYPE_ICS_SIMPLE is the only subtype of TYPE_ICS_BASE that's ever > instantiated, and the only one we're ever likely to want. The > existence of different classes is just a hang over from when we > (misguidedly) had separate subtypes for the KVM and non-KVM version of > the device. > > So, collapse the two classes together into just TYPE_ICS. > > Signed-off-by: David Gibson <david@gibson.dropbear.id.au> > --- Reviewed-by: Greg Kurz <groug@kaod.org> > hw/intc/xics.c | 57 ++++++++++++++++--------------------------- > hw/ppc/pnv_psi.c | 2 +- > hw/ppc/spapr_irq.c | 4 +-- > include/hw/ppc/xics.h | 17 ++----------- > 4 files changed, 26 insertions(+), 54 deletions(-) > > diff --git a/hw/intc/xics.c b/hw/intc/xics.c > index 9ae51bbc76..388dbba870 100644 > --- a/hw/intc/xics.c > +++ b/hw/intc/xics.c > @@ -555,7 +555,7 @@ static void ics_reset_irq(ICSIRQState *irq) > > static void ics_reset(DeviceState *dev) > { > - ICSState *ics = ICS_BASE(dev); > + ICSState *ics = ICS(dev); > int i; > uint8_t flags[ics->nr_irqs]; > > @@ -573,7 +573,7 @@ static void ics_reset(DeviceState *dev) > if (kvm_irqchip_in_kernel()) { > Error *local_err = NULL; > > - ics_set_kvm_state(ICS_BASE(dev), &local_err); > + ics_set_kvm_state(ICS(dev), &local_err); > if (local_err) { > error_report_err(local_err); > } > @@ -587,7 +587,7 @@ static void ics_reset_handler(void *dev) > > static void ics_realize(DeviceState *dev, Error **errp) > { > - ICSState *ics = ICS_BASE(dev); > + ICSState *ics = ICS(dev); > Error *local_err = NULL; > Object *obj; > > @@ -609,26 +609,14 @@ static void ics_realize(DeviceState *dev, Error **errp) > qemu_register_reset(ics_reset_handler, ics); > } > > -static void ics_simple_class_init(ObjectClass *klass, void *data) > +static void ics_instance_init(Object *obj) > { > -} > - > -static const TypeInfo ics_simple_info = { > - .name = TYPE_ICS_SIMPLE, > - .parent = TYPE_ICS_BASE, > - .instance_size = sizeof(ICSState), > - .class_init = ics_simple_class_init, > - .class_size = sizeof(ICSStateClass), > -}; > - > -static void ics_base_instance_init(Object *obj) > -{ > - ICSState *ics = ICS_BASE(obj); > + ICSState *ics = ICS(obj); > > ics->offset = XICS_IRQ_BASE; > } > > -static int ics_base_pre_save(void *opaque) > +static int ics_pre_save(void *opaque) > { > ICSState *ics = opaque; > > @@ -639,7 +627,7 @@ static int ics_base_pre_save(void *opaque) > return 0; > } > > -static int ics_base_post_load(void *opaque, int version_id) > +static int ics_post_load(void *opaque, int version_id) > { > ICSState *ics = opaque; > > @@ -657,7 +645,7 @@ static int ics_base_post_load(void *opaque, int version_id) > return 0; > } > > -static const VMStateDescription vmstate_ics_base_irq = { > +static const VMStateDescription vmstate_ics_irq = { > .name = "ics/irq", > .version_id = 2, > .minimum_version_id = 1, > @@ -671,46 +659,44 @@ static const VMStateDescription vmstate_ics_base_irq = { > }, > }; > > -static const VMStateDescription vmstate_ics_base = { > +static const VMStateDescription vmstate_ics = { > .name = "ics", > .version_id = 1, > .minimum_version_id = 1, > - .pre_save = ics_base_pre_save, > - .post_load = ics_base_post_load, > + .pre_save = ics_pre_save, > + .post_load = ics_post_load, > .fields = (VMStateField[]) { > /* Sanity check */ > VMSTATE_UINT32_EQUAL(nr_irqs, ICSState, NULL), > > VMSTATE_STRUCT_VARRAY_POINTER_UINT32(irqs, ICSState, nr_irqs, > - vmstate_ics_base_irq, > + vmstate_ics_irq, > ICSIRQState), > VMSTATE_END_OF_LIST() > }, > }; > > -static Property ics_base_properties[] = { > +static Property ics_properties[] = { > DEFINE_PROP_UINT32("nr-irqs", ICSState, nr_irqs, 0), > DEFINE_PROP_END_OF_LIST(), > }; > > -static void ics_base_class_init(ObjectClass *klass, void *data) > +static void ics_class_init(ObjectClass *klass, void *data) > { > DeviceClass *dc = DEVICE_CLASS(klass); > > dc->realize = ics_realize; > - dc->props = ics_base_properties; > + dc->props = ics_properties; > dc->reset = ics_reset; > - dc->vmsd = &vmstate_ics_base; > + dc->vmsd = &vmstate_ics; > } > > -static const TypeInfo ics_base_info = { > - .name = TYPE_ICS_BASE, > +static const TypeInfo ics_info = { > + .name = TYPE_ICS, > .parent = TYPE_DEVICE, > - .abstract = true, > .instance_size = sizeof(ICSState), > - .instance_init = ics_base_instance_init, > - .class_init = ics_base_class_init, > - .class_size = sizeof(ICSStateClass), > + .instance_init = ics_instance_init, > + .class_init = ics_class_init, > }; > > static const TypeInfo xics_fabric_info = { > @@ -749,8 +735,7 @@ void ics_set_irq_type(ICSState *ics, int srcno, bool lsi) > > static void xics_register_types(void) > { > - type_register_static(&ics_simple_info); > - type_register_static(&ics_base_info); > + type_register_static(&ics_info); > type_register_static(&icp_info); > type_register_static(&xics_fabric_info); > } > diff --git a/hw/ppc/pnv_psi.c b/hw/ppc/pnv_psi.c > index 8ea81e9d8e..a997f16bb4 100644 > --- a/hw/ppc/pnv_psi.c > +++ b/hw/ppc/pnv_psi.c > @@ -469,7 +469,7 @@ static void pnv_psi_power8_instance_init(Object *obj) > Pnv8Psi *psi8 = PNV8_PSI(obj); > > object_initialize_child(obj, "ics-psi", &psi8->ics, sizeof(psi8->ics), > - TYPE_ICS_SIMPLE, &error_abort, NULL); > + TYPE_ICS, &error_abort, NULL); > } > > static const uint8_t irq_to_xivr[] = { > diff --git a/hw/ppc/spapr_irq.c b/hw/ppc/spapr_irq.c > index ac189c5796..6c45d2a3c0 100644 > --- a/hw/ppc/spapr_irq.c > +++ b/hw/ppc/spapr_irq.c > @@ -98,7 +98,7 @@ static void spapr_irq_init_xics(SpaprMachineState *spapr, int nr_irqs, > Object *obj; > Error *local_err = NULL; > > - obj = object_new(TYPE_ICS_SIMPLE); > + obj = object_new(TYPE_ICS); > object_property_add_child(OBJECT(spapr), "ics", obj, &error_abort); > object_property_add_const_link(obj, ICS_PROP_XICS, OBJECT(spapr), > &error_fatal); > @@ -109,7 +109,7 @@ static void spapr_irq_init_xics(SpaprMachineState *spapr, int nr_irqs, > return; > } > > - spapr->ics = ICS_BASE(obj); > + spapr->ics = ICS(obj); > > xics_spapr_init(spapr); > } > diff --git a/include/hw/ppc/xics.h b/include/hw/ppc/xics.h > index 92628e7cab..d8cf206a69 100644 > --- a/include/hw/ppc/xics.h > +++ b/include/hw/ppc/xics.h > @@ -89,21 +89,8 @@ struct PnvICPState { > uint32_t links[3]; > }; > > -#define TYPE_ICS_BASE "ics-base" > -#define ICS_BASE(obj) OBJECT_CHECK(ICSState, (obj), TYPE_ICS_BASE) > - > -/* Retain ics for sPAPR for migration from existing sPAPR guests */ > -#define TYPE_ICS_SIMPLE "ics" > -#define ICS_SIMPLE(obj) OBJECT_CHECK(ICSState, (obj), TYPE_ICS_SIMPLE) > - > -#define ICS_BASE_CLASS(klass) \ > - OBJECT_CLASS_CHECK(ICSStateClass, (klass), TYPE_ICS_BASE) > -#define ICS_BASE_GET_CLASS(obj) \ > - OBJECT_GET_CLASS(ICSStateClass, (obj), TYPE_ICS_BASE) > - > -struct ICSStateClass { > - DeviceClass parent_class; > -}; > +#define TYPE_ICS "ics" > +#define ICS(obj) OBJECT_CHECK(ICSState, (obj), TYPE_ICS) > > struct ICSState { > /*< private >*/ ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 4/4] xics: Merge TYPE_ICS_BASE and TYPE_ICS_SIMPLE classes 2019-09-24 4:59 ` [PATCH 4/4] xics: Merge TYPE_ICS_BASE and TYPE_ICS_SIMPLE classes David Gibson 2019-09-24 5:31 ` Cédric Le Goater 2019-09-24 7:40 ` Greg Kurz @ 2019-09-24 9:46 ` Philippe Mathieu-Daudé 2 siblings, 0 replies; 30+ messages in thread From: Philippe Mathieu-Daudé @ 2019-09-24 9:46 UTC (permalink / raw) To: David Gibson, qemu-ppc, clg; +Cc: gkurz, qemu-devel On 9/24/19 6:59 AM, David Gibson wrote: > TYPE_ICS_SIMPLE is the only subtype of TYPE_ICS_BASE that's ever > instantiated, and the only one we're ever likely to want. The > existence of different classes is just a hang over from when we > (misguidedly) had separate subtypes for the KVM and non-KVM version of > the device. > > So, collapse the two classes together into just TYPE_ICS. > > Signed-off-by: David Gibson <david@gibson.dropbear.id.au> > --- > hw/intc/xics.c | 57 ++++++++++++++++--------------------------- > hw/ppc/pnv_psi.c | 2 +- > hw/ppc/spapr_irq.c | 4 +-- > include/hw/ppc/xics.h | 17 ++----------- > 4 files changed, 26 insertions(+), 54 deletions(-) > > diff --git a/hw/intc/xics.c b/hw/intc/xics.c > index 9ae51bbc76..388dbba870 100644 > --- a/hw/intc/xics.c > +++ b/hw/intc/xics.c > @@ -555,7 +555,7 @@ static void ics_reset_irq(ICSIRQState *irq) > > static void ics_reset(DeviceState *dev) > { > - ICSState *ics = ICS_BASE(dev); > + ICSState *ics = ICS(dev); > int i; > uint8_t flags[ics->nr_irqs]; > > @@ -573,7 +573,7 @@ static void ics_reset(DeviceState *dev) > if (kvm_irqchip_in_kernel()) { > Error *local_err = NULL; > > - ics_set_kvm_state(ICS_BASE(dev), &local_err); > + ics_set_kvm_state(ICS(dev), &local_err); > if (local_err) { > error_report_err(local_err); > } > @@ -587,7 +587,7 @@ static void ics_reset_handler(void *dev) > > static void ics_realize(DeviceState *dev, Error **errp) > { > - ICSState *ics = ICS_BASE(dev); > + ICSState *ics = ICS(dev); > Error *local_err = NULL; > Object *obj; > > @@ -609,26 +609,14 @@ static void ics_realize(DeviceState *dev, Error **errp) > qemu_register_reset(ics_reset_handler, ics); > } > > -static void ics_simple_class_init(ObjectClass *klass, void *data) > +static void ics_instance_init(Object *obj) > { > -} > - > -static const TypeInfo ics_simple_info = { > - .name = TYPE_ICS_SIMPLE, > - .parent = TYPE_ICS_BASE, > - .instance_size = sizeof(ICSState), > - .class_init = ics_simple_class_init, > - .class_size = sizeof(ICSStateClass), > -}; > - > -static void ics_base_instance_init(Object *obj) > -{ > - ICSState *ics = ICS_BASE(obj); > + ICSState *ics = ICS(obj); > > ics->offset = XICS_IRQ_BASE; > } > > -static int ics_base_pre_save(void *opaque) > +static int ics_pre_save(void *opaque) > { > ICSState *ics = opaque; > > @@ -639,7 +627,7 @@ static int ics_base_pre_save(void *opaque) > return 0; > } > > -static int ics_base_post_load(void *opaque, int version_id) > +static int ics_post_load(void *opaque, int version_id) > { > ICSState *ics = opaque; > > @@ -657,7 +645,7 @@ static int ics_base_post_load(void *opaque, int version_id) > return 0; > } > > -static const VMStateDescription vmstate_ics_base_irq = { > +static const VMStateDescription vmstate_ics_irq = { > .name = "ics/irq", > .version_id = 2, > .minimum_version_id = 1, > @@ -671,46 +659,44 @@ static const VMStateDescription vmstate_ics_base_irq = { > }, > }; > > -static const VMStateDescription vmstate_ics_base = { > +static const VMStateDescription vmstate_ics = { > .name = "ics", > .version_id = 1, > .minimum_version_id = 1, > - .pre_save = ics_base_pre_save, > - .post_load = ics_base_post_load, > + .pre_save = ics_pre_save, > + .post_load = ics_post_load, > .fields = (VMStateField[]) { > /* Sanity check */ > VMSTATE_UINT32_EQUAL(nr_irqs, ICSState, NULL), > > VMSTATE_STRUCT_VARRAY_POINTER_UINT32(irqs, ICSState, nr_irqs, > - vmstate_ics_base_irq, > + vmstate_ics_irq, > ICSIRQState), > VMSTATE_END_OF_LIST() > }, > }; > > -static Property ics_base_properties[] = { > +static Property ics_properties[] = { > DEFINE_PROP_UINT32("nr-irqs", ICSState, nr_irqs, 0), > DEFINE_PROP_END_OF_LIST(), > }; > > -static void ics_base_class_init(ObjectClass *klass, void *data) > +static void ics_class_init(ObjectClass *klass, void *data) > { > DeviceClass *dc = DEVICE_CLASS(klass); > > dc->realize = ics_realize; > - dc->props = ics_base_properties; > + dc->props = ics_properties; > dc->reset = ics_reset; > - dc->vmsd = &vmstate_ics_base; > + dc->vmsd = &vmstate_ics; > } > > -static const TypeInfo ics_base_info = { > - .name = TYPE_ICS_BASE, > +static const TypeInfo ics_info = { > + .name = TYPE_ICS, > .parent = TYPE_DEVICE, Ah now I see this, OK. > - .abstract = true, > .instance_size = sizeof(ICSState), > - .instance_init = ics_base_instance_init, > - .class_init = ics_base_class_init, > - .class_size = sizeof(ICSStateClass), > + .instance_init = ics_instance_init, > + .class_init = ics_class_init, > }; > > static const TypeInfo xics_fabric_info = { > @@ -749,8 +735,7 @@ void ics_set_irq_type(ICSState *ics, int srcno, bool lsi) > > static void xics_register_types(void) > { > - type_register_static(&ics_simple_info); > - type_register_static(&ics_base_info); > + type_register_static(&ics_info); > type_register_static(&icp_info); > type_register_static(&xics_fabric_info); > } > diff --git a/hw/ppc/pnv_psi.c b/hw/ppc/pnv_psi.c > index 8ea81e9d8e..a997f16bb4 100644 > --- a/hw/ppc/pnv_psi.c > +++ b/hw/ppc/pnv_psi.c > @@ -469,7 +469,7 @@ static void pnv_psi_power8_instance_init(Object *obj) > Pnv8Psi *psi8 = PNV8_PSI(obj); > > object_initialize_child(obj, "ics-psi", &psi8->ics, sizeof(psi8->ics), > - TYPE_ICS_SIMPLE, &error_abort, NULL); > + TYPE_ICS, &error_abort, NULL); > } > > static const uint8_t irq_to_xivr[] = { > diff --git a/hw/ppc/spapr_irq.c b/hw/ppc/spapr_irq.c > index ac189c5796..6c45d2a3c0 100644 > --- a/hw/ppc/spapr_irq.c > +++ b/hw/ppc/spapr_irq.c > @@ -98,7 +98,7 @@ static void spapr_irq_init_xics(SpaprMachineState *spapr, int nr_irqs, > Object *obj; > Error *local_err = NULL; > > - obj = object_new(TYPE_ICS_SIMPLE); > + obj = object_new(TYPE_ICS); > object_property_add_child(OBJECT(spapr), "ics", obj, &error_abort); > object_property_add_const_link(obj, ICS_PROP_XICS, OBJECT(spapr), > &error_fatal); > @@ -109,7 +109,7 @@ static void spapr_irq_init_xics(SpaprMachineState *spapr, int nr_irqs, > return; > } > > - spapr->ics = ICS_BASE(obj); > + spapr->ics = ICS(obj); > > xics_spapr_init(spapr); > } > diff --git a/include/hw/ppc/xics.h b/include/hw/ppc/xics.h > index 92628e7cab..d8cf206a69 100644 > --- a/include/hw/ppc/xics.h > +++ b/include/hw/ppc/xics.h > @@ -89,21 +89,8 @@ struct PnvICPState { > uint32_t links[3]; > }; > > -#define TYPE_ICS_BASE "ics-base" > -#define ICS_BASE(obj) OBJECT_CHECK(ICSState, (obj), TYPE_ICS_BASE) > - > -/* Retain ics for sPAPR for migration from existing sPAPR guests */ > -#define TYPE_ICS_SIMPLE "ics" > -#define ICS_SIMPLE(obj) OBJECT_CHECK(ICSState, (obj), TYPE_ICS_SIMPLE) > - > -#define ICS_BASE_CLASS(klass) \ > - OBJECT_CLASS_CHECK(ICSStateClass, (klass), TYPE_ICS_BASE) > -#define ICS_BASE_GET_CLASS(obj) \ > - OBJECT_GET_CLASS(ICSStateClass, (obj), TYPE_ICS_BASE) > - > -struct ICSStateClass { > - DeviceClass parent_class; > -}; > +#define TYPE_ICS "ics" > +#define ICS(obj) OBJECT_CHECK(ICSState, (obj), TYPE_ICS) > > struct ICSState { > /*< private >*/ > ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 0/4] xics: Eliminate unnecessary class 2019-09-24 4:59 [PATCH 0/4] xics: Eliminate unnecessary class David Gibson ` (3 preceding siblings ...) 2019-09-24 4:59 ` [PATCH 4/4] xics: Merge TYPE_ICS_BASE and TYPE_ICS_SIMPLE classes David Gibson @ 2019-09-24 5:22 ` Cédric Le Goater 2019-09-24 7:52 ` Greg Kurz 2019-09-24 9:47 ` Philippe Mathieu-Daudé 5 siblings, 1 reply; 30+ messages in thread From: Cédric Le Goater @ 2019-09-24 5:22 UTC (permalink / raw) To: David Gibson, qemu-ppc; +Cc: gkurz, qemu-devel On 24/09/2019 06:59, David Gibson wrote: > The XICS interrupt controller device used to have separate subtypes > for the KVM and non-KVM variant of the device. That was a bad idea, > because it leaked information that should be entirely host-side > implementation specific to the kinda-sorta guest visible QOM class > names. > > We eliminated the KVM specific class some time ago, but it's left > behind a distinction between the TYPE_ICS_BASE abstract class and > TYPE_ICS_SIMPLE subtype which no longer serves any purpose. > > This series collapses the two types back into one. OK. Is it migration compatible ? because this is why we have kept this subclass. I am glad we can remove it finally. Thanks, C. > > David Gibson (4): > xics: Eliminate 'reject', 'resend' and 'eoi' class hooks > xics: Merge reset and realize hooks > xics: Rename misleading ics_simple_*() functions > xics: Merge TYPE_ICS_BASE and TYPE_ICS_SIMPLE classes > > hw/intc/trace-events | 10 +- > hw/intc/xics.c | 211 ++++++++++++++---------------------------- > hw/intc/xics_spapr.c | 12 +-- > hw/ppc/pnv_psi.c | 6 +- > hw/ppc/spapr_irq.c | 6 +- > include/hw/ppc/xics.h | 30 +----- > 6 files changed, 92 insertions(+), 183 deletions(-) > ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 0/4] xics: Eliminate unnecessary class 2019-09-24 5:22 ` [PATCH 0/4] xics: Eliminate unnecessary class Cédric Le Goater @ 2019-09-24 7:52 ` Greg Kurz 2019-09-24 9:55 ` Cédric Le Goater 0 siblings, 1 reply; 30+ messages in thread From: Greg Kurz @ 2019-09-24 7:52 UTC (permalink / raw) To: Cédric Le Goater; +Cc: gkurz, qemu-ppc, qemu-devel, David Gibson On Tue, 24 Sep 2019 07:22:51 +0200 Cédric Le Goater <clg@kaod.org> wrote: > On 24/09/2019 06:59, David Gibson wrote: > > The XICS interrupt controller device used to have separate subtypes > > for the KVM and non-KVM variant of the device. That was a bad idea, > > because it leaked information that should be entirely host-side > > implementation specific to the kinda-sorta guest visible QOM class > > names. > > > > We eliminated the KVM specific class some time ago, but it's left > > behind a distinction between the TYPE_ICS_BASE abstract class and > > TYPE_ICS_SIMPLE subtype which no longer serves any purpose. > > > > This series collapses the two types back into one. > > OK. Is it migration compatible ? because this is why we have kept Yeah, the types themselves don't matter, only the format of the binary stream described in the VMStateDescription does. > this subclass. I am glad we can remove it finally. > I was thinking we were keeping that for pnv... > Thanks, > > C. > > > > > > David Gibson (4): > > xics: Eliminate 'reject', 'resend' and 'eoi' class hooks > > xics: Merge reset and realize hooks > > xics: Rename misleading ics_simple_*() functions > > xics: Merge TYPE_ICS_BASE and TYPE_ICS_SIMPLE classes > > > > hw/intc/trace-events | 10 +- > > hw/intc/xics.c | 211 ++++++++++++++---------------------------- > > hw/intc/xics_spapr.c | 12 +-- > > hw/ppc/pnv_psi.c | 6 +- > > hw/ppc/spapr_irq.c | 6 +- > > include/hw/ppc/xics.h | 30 +----- > > 6 files changed, 92 insertions(+), 183 deletions(-) > > > ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 0/4] xics: Eliminate unnecessary class 2019-09-24 7:52 ` Greg Kurz @ 2019-09-24 9:55 ` Cédric Le Goater 2019-09-24 10:04 ` Philippe Mathieu-Daudé 0 siblings, 1 reply; 30+ messages in thread From: Cédric Le Goater @ 2019-09-24 9:55 UTC (permalink / raw) To: Greg Kurz; +Cc: gkurz, qemu-ppc, qemu-devel, David Gibson On 24/09/2019 09:52, Greg Kurz wrote: > On Tue, 24 Sep 2019 07:22:51 +0200 > Cédric Le Goater <clg@kaod.org> wrote: > >> On 24/09/2019 06:59, David Gibson wrote: >>> The XICS interrupt controller device used to have separate subtypes >>> for the KVM and non-KVM variant of the device. That was a bad idea, >>> because it leaked information that should be entirely host-side >>> implementation specific to the kinda-sorta guest visible QOM class >>> names. >>> >>> We eliminated the KVM specific class some time ago, but it's left >>> behind a distinction between the TYPE_ICS_BASE abstract class and >>> TYPE_ICS_SIMPLE subtype which no longer serves any purpose. >>> >>> This series collapses the two types back into one. >> >> OK. Is it migration compatible ? because this is why we have kept > > Yeah, the types themselves don't matter, only the format of the > binary stream described in the VMStateDescription does. > >> this subclass. I am glad we can remove it finally. >> > > I was thinking we were keeping that for pnv... Yes, also. See the resend and reject handler in the code below. I have been maintaining this patch since QEMU 2.6. I think it is time for me to declare defeat on getting it merged and QEMU 4.2 will be the last rebase. C. /* * QEMU PowerPC PowerNV PHB3 model * * Copyright (c) 2014-2018, IBM Corporation. * * This code is licensed under the GPL version 2 or later. See the * COPYING file in the top-level directory. */ #include "qemu/osdep.h" #include "qemu/log.h" #include "qapi/error.h" #include "qemu-common.h" #include "hw/pci-host/pnv_phb3_regs.h" #include "hw/pci-host/pnv_phb3.h" #include "hw/ppc/pnv.h" #include "hw/pci/msi.h" #include "monitor/monitor.h" #include "hw/irq.h" #include "hw/qdev-properties.h" #include "sysemu/reset.h" static uint64_t phb3_msi_ive_addr(PnvPHB3 *phb, int srcno) { uint64_t ivtbar = phb->regs[PHB_IVT_BAR >> 3]; uint64_t phbctl = phb->regs[PHB_CONTROL >> 3]; if (!(ivtbar & PHB_IVT_BAR_ENABLE)) { qemu_log_mask(LOG_GUEST_ERROR, "Failed access to disable IVT BAR !"); return 0; } if (srcno >= (ivtbar & PHB_IVT_LENGTH_MASK)) { qemu_log_mask(LOG_GUEST_ERROR, "MSI out of bounds (%d vs 0x%"PRIx64")", srcno, (uint64_t) (ivtbar & PHB_IVT_LENGTH_MASK)); return 0; } ivtbar &= PHB_IVT_BASE_ADDRESS_MASK; if (phbctl & PHB_CTRL_IVE_128_BYTES) { return ivtbar + 128 * srcno; } else { return ivtbar + 16 * srcno; } } static bool phb3_msi_read_ive(PnvPHB3 *phb, int srcno, uint64_t *out_ive) { uint64_t ive_addr, ive; ive_addr = phb3_msi_ive_addr(phb, srcno); if (!ive_addr) { return false; } if (dma_memory_read(&address_space_memory, ive_addr, &ive, sizeof(ive))) { qemu_log_mask(LOG_GUEST_ERROR, "Failed to read IVE at 0x%" PRIx64, ive_addr); return false; } *out_ive = be64_to_cpu(ive); return true; } static void phb3_msi_set_p(Phb3MsiState *msi, int srcno, uint8_t gen) { uint64_t ive_addr; uint8_t p = 0x01 | (gen << 1); ive_addr = phb3_msi_ive_addr(msi->phb, srcno); if (!ive_addr) { return; } if (dma_memory_write(&address_space_memory, ive_addr + 4, &p, 1)) { qemu_log_mask(LOG_GUEST_ERROR, "Failed to write IVE (set P) at 0x%" PRIx64, ive_addr); } } static void phb3_msi_set_q(Phb3MsiState *msi, int srcno) { uint64_t ive_addr; uint8_t q = 0x01; ive_addr = phb3_msi_ive_addr(msi->phb, srcno); if (!ive_addr) { return; } if (dma_memory_write(&address_space_memory, ive_addr + 5, &q, 1)) { qemu_log_mask(LOG_GUEST_ERROR, "Failed to write IVE (set Q) at 0x%" PRIx64, ive_addr); } } static void phb3_msi_try_send(Phb3MsiState *msi, int srcno, bool force) { ICSState *ics = ICS_BASE(msi); uint64_t ive; uint64_t server, prio, pq, gen; if (!phb3_msi_read_ive(msi->phb, srcno, &ive)) { return; } server = GETFIELD(IODA2_IVT_SERVER, ive); prio = GETFIELD(IODA2_IVT_PRIORITY, ive); if (!force) { pq = GETFIELD(IODA2_IVT_Q, ive) | (GETFIELD(IODA2_IVT_P, ive) << 1); } else { pq = 0; } gen = GETFIELD(IODA2_IVT_GEN, ive); /* * The low order 2 bits are the link pointer (Type II interrupts). * Shift back to get a valid IRQ server. */ server >>= 2; switch (pq) { case 0: /* 00 */ if (prio == 0xff) { /* Masked, set Q */ phb3_msi_set_q(msi, srcno); } else { /* Enabled, set P and send */ phb3_msi_set_p(msi, srcno, gen); icp_irq(ics, server, srcno + ics->offset, prio); } break; case 2: /* 10 */ /* Already pending, set Q */ phb3_msi_set_q(msi, srcno); break; case 1: /* 01 */ case 3: /* 11 */ default: /* Just drop stuff if Q already set */ break; } } static void phb3_msi_set_irq(void *opaque, int srcno, int val) { Phb3MsiState *msi = PHB3_MSI(opaque); if (val) { phb3_msi_try_send(msi, srcno, false); } } void pnv_phb3_msi_send(Phb3MsiState *msi, uint64_t addr, uint16_t data, int32_t dev_pe) { ICSState *ics = ICS_BASE(msi); uint64_t ive; uint16_t pe; uint32_t src = ((addr >> 4) & 0xffff) | (data & 0x1f); if (src >= ics->nr_irqs) { qemu_log_mask(LOG_GUEST_ERROR, "MSI %d out of bounds", src); return; } if (dev_pe >= 0) { if (!phb3_msi_read_ive(msi->phb, src, &ive)) { return; } pe = GETFIELD(IODA2_IVT_PE, ive); if (pe != dev_pe) { qemu_log_mask(LOG_GUEST_ERROR, "MSI %d send by PE#%d but assigned to PE#%d", src, dev_pe, pe); return; } } qemu_irq_pulse(msi->qirqs[src]); } void pnv_phb3_msi_ffi(Phb3MsiState *msi, uint64_t val) { /* Emit interrupt */ pnv_phb3_msi_send(msi, val, 0, -1); /* Clear FFI lock */ msi->phb->regs[PHB_FFI_LOCK >> 3] = 0; } static void phb3_msi_reject(ICSState *ics, uint32_t nr) { Phb3MsiState *msi = PHB3_MSI(ics); unsigned int srcno = nr - ics->offset; unsigned int idx = srcno >> 6; unsigned int bit = 1ull << (srcno & 0x3f); assert(srcno < PHB3_MAX_MSI); msi->rba[idx] |= bit; msi->rba_sum |= (1u << idx); } static void phb3_msi_resend(ICSState *ics) { Phb3MsiState *msi = PHB3_MSI(ics); unsigned int i, j; if (msi->rba_sum == 0) { return; } for (i = 0; i < 32; i++) { if ((msi->rba_sum & (1u << i)) == 0) { continue; } msi->rba_sum &= ~(1u << i); for (j = 0; j < 64; j++) { if ((msi->rba[i] & (1ull << j)) == 0) { continue; } msi->rba[i] &= ~(1u << j); phb3_msi_try_send(msi, i * 64 + j, true); } } } static void phb3_msi_print_info(ICSState *ics, Monitor *mon) { Phb3MsiState *msi = PHB3_MSI(ics); int i; for (i = 0; i < ics->nr_irqs; i++) { uint64_t ive; if (!phb3_msi_read_ive(msi->phb, i, &ive)) { return; } if (GETFIELD(IODA2_IVT_PRIORITY, ive) == 0xff) { continue; } monitor_printf(mon, " %4x %c%c server=%04x prio=%02x gen=%d\n", ics->offset + i, GETFIELD(IODA2_IVT_P, ive) ? 'P' : '-', GETFIELD(IODA2_IVT_Q, ive) ? 'Q' : '-', (uint32_t) GETFIELD(IODA2_IVT_SERVER, ive) >> 2, (uint32_t) GETFIELD(IODA2_IVT_PRIORITY, ive), (uint32_t) GETFIELD(IODA2_IVT_GEN, ive)); } } static void phb3_msi_reset(DeviceState *dev) { Phb3MsiState *msi = PHB3_MSI(dev); ICSStateClass *icsc = ICS_BASE_GET_CLASS(dev); icsc->parent_reset(dev); memset(msi->rba, 0, sizeof(msi->rba)); msi->rba_sum = 0; } static void phb3_msi_reset_handler(void *dev) { phb3_msi_reset(dev); } void pnv_phb3_msi_update_config(Phb3MsiState *msi, uint32_t base, uint32_t count) { ICSState *ics = ICS_BASE(msi); if (count > PHB3_MAX_MSI) { count = PHB3_MAX_MSI; } ics->nr_irqs = count; ics->offset = base; } static void phb3_msi_realize(DeviceState *dev, Error **errp) { Phb3MsiState *msi = PHB3_MSI(dev); ICSState *ics = ICS_BASE(msi); ICSStateClass *icsc = ICS_BASE_GET_CLASS(ics); Object *obj; Error *local_err = NULL; icsc->parent_realize(dev, &local_err); if (local_err) { error_propagate(errp, local_err); return; } obj = object_property_get_link(OBJECT(dev), "phb", &local_err); if (!obj) { error_propagate(errp, local_err); error_prepend(errp, "required link 'phb' not found: "); return; } msi->phb = PNV_PHB3(obj); msi->qirqs = qemu_allocate_irqs(phb3_msi_set_irq, msi, ics->nr_irqs); qemu_register_reset(phb3_msi_reset_handler, dev); } static void phb3_msi_instance_init(Object *obj) { ICSState *ics = ICS_BASE(obj); /* Will be overriden later */ ics->offset = 0; } static void phb3_msi_class_init(ObjectClass *klass, void *data) { DeviceClass *dc = DEVICE_CLASS(klass); ICSStateClass *isc = ICS_BASE_CLASS(klass); device_class_set_parent_realize(dc, phb3_msi_realize, &isc->parent_realize); device_class_set_parent_reset(dc, phb3_msi_reset, &isc->parent_reset); isc->reject = phb3_msi_reject; isc->resend = phb3_msi_resend; isc->print_info = phb3_msi_print_info; } static const TypeInfo phb3_msi_info = { .name = TYPE_PHB3_MSI, .parent = TYPE_ICS_BASE, .instance_size = sizeof(Phb3MsiState), .class_init = phb3_msi_class_init, .class_size = sizeof(ICSStateClass), .instance_init = phb3_msi_instance_init, }; static void pnv_phb3_msi_register_types(void) { type_register_static(&phb3_msi_info); } type_init(pnv_phb3_msi_register_types) ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 0/4] xics: Eliminate unnecessary class 2019-09-24 9:55 ` Cédric Le Goater @ 2019-09-24 10:04 ` Philippe Mathieu-Daudé 2019-09-24 11:00 ` Cédric Le Goater 0 siblings, 1 reply; 30+ messages in thread From: Philippe Mathieu-Daudé @ 2019-09-24 10:04 UTC (permalink / raw) To: Cédric Le Goater, Greg Kurz Cc: gkurz, qemu-ppc, qemu-devel, David Gibson Hi Cédric, On 9/24/19 11:55 AM, Cédric Le Goater wrote: > On 24/09/2019 09:52, Greg Kurz wrote: >> On Tue, 24 Sep 2019 07:22:51 +0200 >> Cédric Le Goater <clg@kaod.org> wrote: >> >>> On 24/09/2019 06:59, David Gibson wrote: >>>> The XICS interrupt controller device used to have separate subtypes >>>> for the KVM and non-KVM variant of the device. That was a bad idea, >>>> because it leaked information that should be entirely host-side >>>> implementation specific to the kinda-sorta guest visible QOM class >>>> names. >>>> >>>> We eliminated the KVM specific class some time ago, but it's left >>>> behind a distinction between the TYPE_ICS_BASE abstract class and >>>> TYPE_ICS_SIMPLE subtype which no longer serves any purpose. >>>> >>>> This series collapses the two types back into one. >>> >>> OK. Is it migration compatible ? because this is why we have kept >> >> Yeah, the types themselves don't matter, only the format of the >> binary stream described in the VMStateDescription does. >> >>> this subclass. I am glad we can remove it finally. >>> >> >> I was thinking we were keeping that for pnv... > > > Yes, also. See the resend and reject handler in the code > below. > > > I have been maintaining this patch since QEMU 2.6. I think > it is time for me to declare defeat on getting it merged > and QEMU 4.2 will be the last rebase. Do you remember what is missing for being merged upstream? > /* > * QEMU PowerPC PowerNV PHB3 model > * > * Copyright (c) 2014-2018, IBM Corporation. > * > * This code is licensed under the GPL version 2 or later. See the > * COPYING file in the top-level directory. > */ > #include "qemu/osdep.h" > #include "qemu/log.h" > #include "qapi/error.h" > #include "qemu-common.h" > #include "hw/pci-host/pnv_phb3_regs.h" > #include "hw/pci-host/pnv_phb3.h" > #include "hw/ppc/pnv.h" > #include "hw/pci/msi.h" > #include "monitor/monitor.h" > #include "hw/irq.h" > #include "hw/qdev-properties.h" > #include "sysemu/reset.h" > > static uint64_t phb3_msi_ive_addr(PnvPHB3 *phb, int srcno) > { > uint64_t ivtbar = phb->regs[PHB_IVT_BAR >> 3]; > uint64_t phbctl = phb->regs[PHB_CONTROL >> 3]; > > if (!(ivtbar & PHB_IVT_BAR_ENABLE)) { > qemu_log_mask(LOG_GUEST_ERROR, "Failed access to disable IVT BAR !"); > return 0; > } > > if (srcno >= (ivtbar & PHB_IVT_LENGTH_MASK)) { > qemu_log_mask(LOG_GUEST_ERROR, "MSI out of bounds (%d vs 0x%"PRIx64")", > srcno, (uint64_t) (ivtbar & PHB_IVT_LENGTH_MASK)); > return 0; > } > > ivtbar &= PHB_IVT_BASE_ADDRESS_MASK; > > if (phbctl & PHB_CTRL_IVE_128_BYTES) { > return ivtbar + 128 * srcno; > } else { > return ivtbar + 16 * srcno; > } > } > > static bool phb3_msi_read_ive(PnvPHB3 *phb, int srcno, uint64_t *out_ive) > { > uint64_t ive_addr, ive; > > ive_addr = phb3_msi_ive_addr(phb, srcno); > if (!ive_addr) { > return false; > } > > if (dma_memory_read(&address_space_memory, ive_addr, &ive, sizeof(ive))) { > qemu_log_mask(LOG_GUEST_ERROR, "Failed to read IVE at 0x%" PRIx64, > ive_addr); > return false; > } > *out_ive = be64_to_cpu(ive); > > return true; > } > > static void phb3_msi_set_p(Phb3MsiState *msi, int srcno, uint8_t gen) > { > uint64_t ive_addr; > uint8_t p = 0x01 | (gen << 1); > > ive_addr = phb3_msi_ive_addr(msi->phb, srcno); > if (!ive_addr) { > return; > } > > if (dma_memory_write(&address_space_memory, ive_addr + 4, &p, 1)) { > qemu_log_mask(LOG_GUEST_ERROR, > "Failed to write IVE (set P) at 0x%" PRIx64, ive_addr); > } > } > > static void phb3_msi_set_q(Phb3MsiState *msi, int srcno) > { > uint64_t ive_addr; > uint8_t q = 0x01; > > ive_addr = phb3_msi_ive_addr(msi->phb, srcno); > if (!ive_addr) { > return; > } > > if (dma_memory_write(&address_space_memory, ive_addr + 5, &q, 1)) { > qemu_log_mask(LOG_GUEST_ERROR, > "Failed to write IVE (set Q) at 0x%" PRIx64, ive_addr); > } > } > > static void phb3_msi_try_send(Phb3MsiState *msi, int srcno, bool force) > { > ICSState *ics = ICS_BASE(msi); > uint64_t ive; > uint64_t server, prio, pq, gen; > > if (!phb3_msi_read_ive(msi->phb, srcno, &ive)) { > return; > } > > server = GETFIELD(IODA2_IVT_SERVER, ive); > prio = GETFIELD(IODA2_IVT_PRIORITY, ive); > if (!force) { > pq = GETFIELD(IODA2_IVT_Q, ive) | (GETFIELD(IODA2_IVT_P, ive) << 1); > } else { > pq = 0; > } > gen = GETFIELD(IODA2_IVT_GEN, ive); > > /* > * The low order 2 bits are the link pointer (Type II interrupts). > * Shift back to get a valid IRQ server. > */ > server >>= 2; > > switch (pq) { > case 0: /* 00 */ > if (prio == 0xff) { > /* Masked, set Q */ > phb3_msi_set_q(msi, srcno); > } else { > /* Enabled, set P and send */ > phb3_msi_set_p(msi, srcno, gen); > icp_irq(ics, server, srcno + ics->offset, prio); > } > break; > case 2: /* 10 */ > /* Already pending, set Q */ > phb3_msi_set_q(msi, srcno); > break; > case 1: /* 01 */ > case 3: /* 11 */ > default: > /* Just drop stuff if Q already set */ > break; > } > } > > static void phb3_msi_set_irq(void *opaque, int srcno, int val) > { > Phb3MsiState *msi = PHB3_MSI(opaque); > > if (val) { > phb3_msi_try_send(msi, srcno, false); > } > } > > > void pnv_phb3_msi_send(Phb3MsiState *msi, uint64_t addr, uint16_t data, > int32_t dev_pe) > { > ICSState *ics = ICS_BASE(msi); > uint64_t ive; > uint16_t pe; > uint32_t src = ((addr >> 4) & 0xffff) | (data & 0x1f); > > if (src >= ics->nr_irqs) { > qemu_log_mask(LOG_GUEST_ERROR, "MSI %d out of bounds", src); > return; > } > if (dev_pe >= 0) { > if (!phb3_msi_read_ive(msi->phb, src, &ive)) { > return; > } > pe = GETFIELD(IODA2_IVT_PE, ive); > if (pe != dev_pe) { > qemu_log_mask(LOG_GUEST_ERROR, > "MSI %d send by PE#%d but assigned to PE#%d", > src, dev_pe, pe); > return; > } > } > qemu_irq_pulse(msi->qirqs[src]); > } > > void pnv_phb3_msi_ffi(Phb3MsiState *msi, uint64_t val) > { > /* Emit interrupt */ > pnv_phb3_msi_send(msi, val, 0, -1); > > /* Clear FFI lock */ > msi->phb->regs[PHB_FFI_LOCK >> 3] = 0; > } > > static void phb3_msi_reject(ICSState *ics, uint32_t nr) > { > Phb3MsiState *msi = PHB3_MSI(ics); > unsigned int srcno = nr - ics->offset; > unsigned int idx = srcno >> 6; > unsigned int bit = 1ull << (srcno & 0x3f); > > assert(srcno < PHB3_MAX_MSI); > > msi->rba[idx] |= bit; > msi->rba_sum |= (1u << idx); > } > > static void phb3_msi_resend(ICSState *ics) > { > Phb3MsiState *msi = PHB3_MSI(ics); > unsigned int i, j; > > if (msi->rba_sum == 0) { > return; > } > > for (i = 0; i < 32; i++) { > if ((msi->rba_sum & (1u << i)) == 0) { > continue; > } > msi->rba_sum &= ~(1u << i); > for (j = 0; j < 64; j++) { > if ((msi->rba[i] & (1ull << j)) == 0) { > continue; > } > msi->rba[i] &= ~(1u << j); > phb3_msi_try_send(msi, i * 64 + j, true); > } > } > } > > static void phb3_msi_print_info(ICSState *ics, Monitor *mon) > { > Phb3MsiState *msi = PHB3_MSI(ics); > int i; > > for (i = 0; i < ics->nr_irqs; i++) { > uint64_t ive; > > if (!phb3_msi_read_ive(msi->phb, i, &ive)) { > return; > } > > if (GETFIELD(IODA2_IVT_PRIORITY, ive) == 0xff) { > continue; > } > > monitor_printf(mon, " %4x %c%c server=%04x prio=%02x gen=%d\n", > ics->offset + i, > GETFIELD(IODA2_IVT_P, ive) ? 'P' : '-', > GETFIELD(IODA2_IVT_Q, ive) ? 'Q' : '-', > (uint32_t) GETFIELD(IODA2_IVT_SERVER, ive) >> 2, > (uint32_t) GETFIELD(IODA2_IVT_PRIORITY, ive), > (uint32_t) GETFIELD(IODA2_IVT_GEN, ive)); > } > } > > static void phb3_msi_reset(DeviceState *dev) > { > Phb3MsiState *msi = PHB3_MSI(dev); > ICSStateClass *icsc = ICS_BASE_GET_CLASS(dev); > > icsc->parent_reset(dev); > > memset(msi->rba, 0, sizeof(msi->rba)); > msi->rba_sum = 0; > } > > static void phb3_msi_reset_handler(void *dev) > { > phb3_msi_reset(dev); > } > > void pnv_phb3_msi_update_config(Phb3MsiState *msi, uint32_t base, > uint32_t count) > { > ICSState *ics = ICS_BASE(msi); > > if (count > PHB3_MAX_MSI) { > count = PHB3_MAX_MSI; > } > ics->nr_irqs = count; > ics->offset = base; > } > > static void phb3_msi_realize(DeviceState *dev, Error **errp) > { > Phb3MsiState *msi = PHB3_MSI(dev); > ICSState *ics = ICS_BASE(msi); > ICSStateClass *icsc = ICS_BASE_GET_CLASS(ics); > Object *obj; > Error *local_err = NULL; > > icsc->parent_realize(dev, &local_err); > if (local_err) { > error_propagate(errp, local_err); > return; > } > > obj = object_property_get_link(OBJECT(dev), "phb", &local_err); > if (!obj) { > error_propagate(errp, local_err); > error_prepend(errp, "required link 'phb' not found: "); > return; > } > msi->phb = PNV_PHB3(obj); > > msi->qirqs = qemu_allocate_irqs(phb3_msi_set_irq, msi, ics->nr_irqs); > > qemu_register_reset(phb3_msi_reset_handler, dev); > } > > static void phb3_msi_instance_init(Object *obj) > { > ICSState *ics = ICS_BASE(obj); > > /* Will be overriden later */ > ics->offset = 0; > } > > static void phb3_msi_class_init(ObjectClass *klass, void *data) > { > DeviceClass *dc = DEVICE_CLASS(klass); > ICSStateClass *isc = ICS_BASE_CLASS(klass); > > device_class_set_parent_realize(dc, phb3_msi_realize, > &isc->parent_realize); > device_class_set_parent_reset(dc, phb3_msi_reset, > &isc->parent_reset); > > isc->reject = phb3_msi_reject; > isc->resend = phb3_msi_resend; > isc->print_info = phb3_msi_print_info; > } > > static const TypeInfo phb3_msi_info = { > .name = TYPE_PHB3_MSI, > .parent = TYPE_ICS_BASE, > .instance_size = sizeof(Phb3MsiState), > .class_init = phb3_msi_class_init, > .class_size = sizeof(ICSStateClass), > .instance_init = phb3_msi_instance_init, > }; > > static void pnv_phb3_msi_register_types(void) > { > type_register_static(&phb3_msi_info); > } > > type_init(pnv_phb3_msi_register_types) > ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 0/4] xics: Eliminate unnecessary class 2019-09-24 10:04 ` Philippe Mathieu-Daudé @ 2019-09-24 11:00 ` Cédric Le Goater 2019-09-26 1:28 ` David Gibson 0 siblings, 1 reply; 30+ messages in thread From: Cédric Le Goater @ 2019-09-24 11:00 UTC (permalink / raw) To: Philippe Mathieu-Daudé, Greg Kurz Cc: gkurz, qemu-ppc, qemu-devel, David Gibson On 24/09/2019 12:04, Philippe Mathieu-Daudé wrote: > Hi Cédric, > > On 9/24/19 11:55 AM, Cédric Le Goater wrote: >> On 24/09/2019 09:52, Greg Kurz wrote: >>> On Tue, 24 Sep 2019 07:22:51 +0200 >>> Cédric Le Goater <clg@kaod.org> wrote: >>> >>>> On 24/09/2019 06:59, David Gibson wrote: >>>>> The XICS interrupt controller device used to have separate subtypes >>>>> for the KVM and non-KVM variant of the device. That was a bad idea, >>>>> because it leaked information that should be entirely host-side >>>>> implementation specific to the kinda-sorta guest visible QOM class >>>>> names. >>>>> >>>>> We eliminated the KVM specific class some time ago, but it's left >>>>> behind a distinction between the TYPE_ICS_BASE abstract class and >>>>> TYPE_ICS_SIMPLE subtype which no longer serves any purpose. >>>>> >>>>> This series collapses the two types back into one. >>>> >>>> OK. Is it migration compatible ? because this is why we have kept >>> >>> Yeah, the types themselves don't matter, only the format of the >>> binary stream described in the VMStateDescription does. >>> >>>> this subclass. I am glad we can remove it finally. >>>> >>> >>> I was thinking we were keeping that for pnv... >> >> >> Yes, also. See the resend and reject handler in the code >> below. >> >> >> I have been maintaining this patch since QEMU 2.6. I think >> it is time for me to declare defeat on getting it merged >> and QEMU 4.2 will be the last rebase. > > Do you remember what is missing for being merged upstream? lack of interest I would say. Here is the patch. C. From b254a0279f5cf13a5db2db0bb209fe806af59ad0 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?C=C3=A9dric=20Le=20Goater?= <clg@kaod.org> Date: Wed, 18 Sep 2019 18:02:03 +0200 Subject: [PATCH] ppc/pnv: Add model for Power8 PHB3 PCIe Host bridge MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This is a model of the PCIe Host Bridge (PHB3) found on a Power8 processor. It includes the PowerBus logic interface (PBCQ), IOMMU support, a single PCIe Gen.3 Root Complex, and support for MSI and LSI interrupt sources as found on a Power8 system using the XICS interrupt controller. The Power8 processor comes in different flavors: Venice, Murano, Naple, each having a different number of PHBs. To make things simpler, the PHB3 model provides 3 per chip. Some platforms, like the Firestone, can also couple PHBs on the first chip to provide more bandwidth but this is too specific to model in QEMU. No default device layout is provided and PCI devices can be added on any of the available PCIe Root Port (pcie.0 .. 2 of a Power8 chip) with address 0x0 as the firwware (skiboot) only accepts a single device per root port. To run a simple system with a network and a storage adapters, use a command line options such as : -device e1000e,netdev=net0,mac=C0:FF:EE:00:00:02,bus=pcie.0,addr=0x0 -netdev bridge,id=net0,helper=/usr/libexec/qemu-bridge-helper,br=virbr0,id=hostnet0 -device megasas,id=scsi0,bus=pcie.1,addr=0x0 -drive file=$disk,if=none,id=drive-scsi0-0-0-0,format=qcow2,cache=none -device scsi-hd,bus=scsi0.0,channel=0,scsi-id=0,lun=0,drive=drive-scsi0-0-0-0,id=scsi0-0-0-0,bootindex=2 If more are needed, include a bridge. Multi chip is supported, each chip adding its set of PHB3 controllers and its PCI busses. The model doesn't emulate the EEH error handling (and may never do). Originally from Benjamin Herrenschmidt. I am taking ownership since I have been maintaining the model since QEMU 2.6. Signed-off-by: Cédric Le Goater <clg@kaod.org> --- include/hw/pci-host/pnv_phb3.h | 173 ++++ include/hw/pci-host/pnv_phb3_regs.h | 454 ++++++++++ include/hw/ppc/pnv.h | 7 + include/hw/ppc/pnv_xscom.h | 9 + include/hw/ppc/xics.h | 1 + hw/intc/xics.c | 2 +- hw/pci-host/pnv_phb3.c | 1191 +++++++++++++++++++++++++++ hw/pci-host/pnv_phb3_msi.c | 320 +++++++ hw/pci-host/pnv_phb3_pbcq.c | 347 ++++++++ hw/ppc/pnv.c | 75 +- hw/ppc/pnv_xscom.c | 11 +- hw/pci-host/Makefile.objs | 1 + hw/ppc/Kconfig | 2 + 13 files changed, 2588 insertions(+), 5 deletions(-) create mode 100644 include/hw/pci-host/pnv_phb3.h create mode 100644 include/hw/pci-host/pnv_phb3_regs.h create mode 100644 hw/pci-host/pnv_phb3.c create mode 100644 hw/pci-host/pnv_phb3_msi.c create mode 100644 hw/pci-host/pnv_phb3_pbcq.c diff --git a/include/hw/pci-host/pnv_phb3.h b/include/hw/pci-host/pnv_phb3.h new file mode 100644 index 000000000000..b282730a5b40 --- /dev/null +++ b/include/hw/pci-host/pnv_phb3.h @@ -0,0 +1,173 @@ +/* + * QEMU PowerPC PowerNV PHB3 model + * + * Copyright (c) 2014-2018, IBM Corporation. + * + * This code is licensed under the GPL version 2 or later. See the + * COPYING file in the top-level directory. + */ + +#ifndef PCI_HOST_PNV_PHB3_H +#define PCI_HOST_PNV_PHB3_H + +#include "hw/pci/pcie_host.h" +#include "hw/pci/pcie_port.h" +#include "hw/ppc/xics.h" + +typedef struct PnvPHB3 PnvPHB3; +typedef struct PnvChip PnvChip; + +/* + * PHB3 XICS Source for MSIs + */ +#define TYPE_PHB3_MSI "phb3-msi" +#define PHB3_MSI(obj) OBJECT_CHECK(Phb3MsiState, (obj), TYPE_PHB3_MSI) + +#define PHB3_MAX_MSI 2048 + +typedef struct Phb3MsiState { + ICSState ics; + qemu_irq *qirqs; + + PnvPHB3 *phb; + uint64_t rba[PHB3_MAX_MSI / 64]; + uint32_t rba_sum; +} Phb3MsiState; + +void pnv_phb3_msi_update_config(Phb3MsiState *msis, uint32_t base, + uint32_t count); +void pnv_phb3_msi_send(Phb3MsiState *msis, uint64_t addr, uint16_t data, + int32_t dev_pe); +void pnv_phb3_msi_ffi(Phb3MsiState *msis, uint64_t val); + + +/* We have one such address space wrapper per possible device + * under the PHB since they need to be assigned statically at + * qemu device creation time. The relationship to a PE is done + * later dynamically. This means we can potentially create a lot + * of these guys. Q35 stores them as some kind of radix tree but + * we never really need to do fast lookups so instead we simply + * keep a QLIST of them for now, we can add the radix if needed + * later on. + * + * We do cache the PE number to speed things up a bit though. + */ +typedef struct PnvPhb3DMASpace { + PCIBus *bus; + uint8_t devfn; + int pe_num; /* Cached PE number */ +#define PHB_INVALID_PE (-1) + PnvPHB3 *phb; + AddressSpace dma_as; + IOMMUMemoryRegion dma_mr; + MemoryRegion msi32_mr; + MemoryRegion msi64_mr; + bool msi32_mapped; + bool msi64_mapped; + QLIST_ENTRY(PnvPhb3DMASpace) list; +} PnvPhb3DMASpace; + +/* + * PHB3 Power Bus Common Queue + */ +#define TYPE_PNV_PBCQ "pnv-pbcq" +#define PNV_PBCQ(obj) OBJECT_CHECK(PnvPBCQState, (obj), TYPE_PNV_PBCQ) + +typedef struct PnvPBCQState { + DeviceState parent; + + uint32_t nest_xbase; + uint32_t spci_xbase; + uint32_t pci_xbase; +#define PBCQ_NEST_REGS_COUNT 0x46 +#define PBCQ_PCI_REGS_COUNT 0x15 +#define PBCQ_SPCI_REGS_COUNT 0x5 + + uint64_t nest_regs[PBCQ_NEST_REGS_COUNT]; + uint64_t spci_regs[PBCQ_SPCI_REGS_COUNT]; + uint64_t pci_regs[PBCQ_PCI_REGS_COUNT]; + MemoryRegion mmbar0; + MemoryRegion mmbar1; + MemoryRegion phbbar; + bool mmio0_mapped; + bool mmio1_mapped; + bool phb_mapped; + uint64_t mmio0_base; + uint64_t mmio0_size; + uint64_t mmio1_base; + uint64_t mmio1_size; + PnvPHB3 *phb; + + MemoryRegion xscom_nest_regs; + MemoryRegion xscom_pci_regs; + MemoryRegion xscom_spci_regs; +} PnvPBCQState; + +/* + * PHB3 PCIe Root port + */ +#define TYPE_PNV_PHB3_ROOT_BUS "pnv-phb3-root-bus" + +#define TYPE_PNV_PHB3_ROOT_PORT "pnv-phb3-root-port" + +typedef struct PnvPHB3RootPort { + PCIESlot parent_obj; +} PnvPHB3RootPort; + +/* + * PHB3 PCIe Host Bridge for PowerNV machines (POWER8) + */ +#define TYPE_PNV_PHB3 "pnv-phb3" +#define PNV_PHB3(obj) OBJECT_CHECK(PnvPHB3, (obj), TYPE_PNV_PHB3) + +#define PNV_PHB3_NUM_M64 16 +#define PNV_PHB3_NUM_REGS (0x1000 >> 3) +#define PNV_PHB3_NUM_LSI 8 +#define PNV_PHB3_NUM_PE 256 + +#define PCI_MMIO_TOTAL_SIZE (0x1ull << 60) + +struct PnvPHB3 { + PCIExpressHost parent_obj; + + PnvPHB3RootPort root; + + uint32_t chip_id; + uint32_t phb_id; + char bus_path[8]; + + uint64_t regs[PNV_PHB3_NUM_REGS]; + MemoryRegion mr_regs; + + MemoryRegion mr_m32; + MemoryRegion mr_m64[PNV_PHB3_NUM_M64]; + bool regs_mapped; + bool m32_mapped; + bool m64_mapped[PNV_PHB3_NUM_M64]; + MemoryRegion pci_mmio; + MemoryRegion pci_io; + + uint64_t ioda_LIST[8]; + uint64_t ioda_LXIVT[8]; + uint64_t ioda_TVT[512]; + uint64_t ioda_M64BT[16]; + uint64_t ioda_MDT[256]; + uint64_t ioda_PEEV[4]; + + uint32_t total_irq; + ICSState lsis; + qemu_irq *qirqs; + Phb3MsiState msis; + + PnvPBCQState pbcq; + + QLIST_HEAD(, PnvPhb3DMASpace) dma_spaces; +}; + +uint64_t pnv_phb3_reg_read(void *opaque, hwaddr off, unsigned size); +void pnv_phb3_reg_write(void *opaque, hwaddr off, uint64_t val, unsigned size); +void pnv_phb3_update_regions(PnvPHB3 *phb); +void pnv_phb3_remap_irqs(PnvPHB3 *phb); + + +#endif /* PCI_HOST_PNV_PHB3_H */ diff --git a/include/hw/pci-host/pnv_phb3_regs.h b/include/hw/pci-host/pnv_phb3_regs.h new file mode 100644 index 000000000000..8df83c321e86 --- /dev/null +++ b/include/hw/pci-host/pnv_phb3_regs.h @@ -0,0 +1,454 @@ +/* Copyright (c) 2013-2018, IBM Corporation. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or + * implied. + * + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#ifndef PCI_HOST_PNV_PHB3_REGS_H +#define PCI_HOST_PNV_PHB3_REGS_H + +#include "target/ppc/cpu.h" + +/* TODO: fix GETFIELD/SETFIELD macro definitions */ +#if HOST_LONG_BITS == 32 +# define MASK_TO_LSH(m) (__builtin_ffsll(m) - 1) +#elif HOST_LONG_BITS == 64 +# define MASK_TO_LSH(m) (__builtin_ffsl(m) - 1) +#else +# error Unknown sizeof long +#endif + +#define GETFIELD(m, v) (((v) & (m)) >> MASK_TO_LSH(m)) +#define SETFIELD(m, v, val) \ + (((v) & ~(m)) | ((((typeof(v))(val)) << MASK_TO_LSH(m)) & (m))) + +/* + * PBCQ XSCOM registers + */ + +#define PBCQ_NEST_IRSN_COMPARE 0x1a +#define PBCQ_NEST_IRSN_COMP PPC_BITMASK(0, 18) +#define PBCQ_NEST_IRSN_MASK 0x1b +#define PBCQ_NEST_LSI_SRC_ID 0x1f +#define PBCQ_NEST_LSI_SRC PPC_BITMASK(0, 7) +#define PBCQ_NEST_REGS_COUNT 0x46 +#define PBCQ_NEST_MMIO_BAR0 0x40 +#define PBCQ_NEST_MMIO_BAR1 0x41 +#define PBCQ_NEST_PHB_BAR 0x42 +#define PBCQ_NEST_MMIO_MASK0 0x43 +#define PBCQ_NEST_MMIO_MASK1 0x44 +#define PBCQ_NEST_BAR_EN 0x45 +#define PBCQ_NEST_BAR_EN_MMIO0 PPC_BIT(0) +#define PBCQ_NEST_BAR_EN_MMIO1 PPC_BIT(1) +#define PBCQ_NEST_BAR_EN_PHB PPC_BIT(2) +#define PBCQ_NEST_BAR_EN_IRSN_RX PPC_BIT(3) +#define PBCQ_NEST_BAR_EN_IRSN_TX PPC_BIT(4) + +#define PBCQ_PCI_REGS_COUNT 0x15 +#define PBCQ_PCI_BAR2 0x0b + +#define PBCQ_SPCI_REGS_COUNT 0x5 +#define PBCQ_SPCI_ASB_ADDR 0x0 +#define PBCQ_SPCI_ASB_STATUS 0x1 +#define PBCQ_SPCI_ASB_DATA 0x2 +#define PBCQ_SPCI_AIB_CAPP_EN 0x3 +#define PBCQ_SPCI_CAPP_SEC_TMR 0x4 + +/* + * PHB MMIO registers + */ + +/* PHB Fundamental register set A */ +#define PHB_LSI_SOURCE_ID 0x100 +#define PHB_LSI_SRC_ID PPC_BITMASK(5, 12) +#define PHB_DMA_CHAN_STATUS 0x110 +#define PHB_DMA_CHAN_ANY_ERR PPC_BIT(27) +#define PHB_DMA_CHAN_ANY_ERR1 PPC_BIT(28) +#define PHB_DMA_CHAN_ANY_FREEZE PPC_BIT(29) +#define PHB_CPU_LOADSTORE_STATUS 0x120 +#define PHB_CPU_LS_ANY_ERR PPC_BIT(27) +#define PHB_CPU_LS_ANY_ERR1 PPC_BIT(28) +#define PHB_CPU_LS_ANY_FREEZE PPC_BIT(29) +#define PHB_DMA_MSI_NODE_ID 0x128 +#define PHB_DMAMSI_NID_FIXED PPC_BIT(0) +#define PHB_DMAMSI_NID PPC_BITMASK(24, 31) +#define PHB_CONFIG_DATA 0x130 +#define PHB_LOCK0 0x138 +#define PHB_CONFIG_ADDRESS 0x140 +#define PHB_CA_ENABLE PPC_BIT(0) +#define PHB_CA_BUS PPC_BITMASK(4, 11) +#define PHB_CA_DEV PPC_BITMASK(12, 16) +#define PHB_CA_FUNC PPC_BITMASK(17, 19) +#define PHB_CA_REG PPC_BITMASK(20, 31) +#define PHB_CA_PE PPC_BITMASK(40, 47) +#define PHB_LOCK1 0x148 +#define PHB_IVT_BAR 0x150 +#define PHB_IVT_BAR_ENABLE PPC_BIT(0) +#define PHB_IVT_BASE_ADDRESS_MASK PPC_BITMASK(14, 48) +#define PHB_IVT_LENGTH_MASK PPC_BITMASK(52, 63) +#define PHB_RBA_BAR 0x158 +#define PHB_RBA_BAR_ENABLE PPC_BIT(0) +#define PHB_RBA_BASE_ADDRESS PPC_BITMASK(14, 55) +#define PHB_PHB3_CONFIG 0x160 +#define PHB_PHB3C_64B_TCE_EN PPC_BIT(2) +#define PHB_PHB3C_32BIT_MSI_EN PPC_BIT(8) +#define PHB_PHB3C_64BIT_MSI_EN PPC_BIT(14) +#define PHB_PHB3C_M32_EN PPC_BIT(16) +#define PHB_RTT_BAR 0x168 +#define PHB_RTT_BAR_ENABLE PPC_BIT(0) +#define PHB_RTT_BASE_ADDRESS_MASK PPC_BITMASK(14, 46) +#define PHB_PELTV_BAR 0x188 +#define PHB_PELTV_BAR_ENABLE PPC_BIT(0) +#define PHB_PELTV_BASE_ADDRESS PPC_BITMASK(14, 50) +#define PHB_M32_BASE_ADDR 0x190 +#define PHB_M32_BASE_MASK 0x198 +#define PHB_M32_START_ADDR 0x1a0 +#define PHB_PEST_BAR 0x1a8 +#define PHB_PEST_BAR_ENABLE PPC_BIT(0) +#define PHB_PEST_BASE_ADDRESS PPC_BITMASK(14, 51) +#define PHB_M64_UPPER_BITS 0x1f0 +#define PHB_INTREP_TIMER 0x1f8 +#define PHB_DMARD_SYNC 0x200 +#define PHB_DMARD_SYNC_START PPC_BIT(0) +#define PHB_DMARD_SYNC_COMPLETE PPC_BIT(1) +#define PHB_RTC_INVALIDATE 0x208 +#define PHB_RTC_INVALIDATE_ALL PPC_BIT(0) +#define PHB_RTC_INVALIDATE_RID PPC_BITMASK(16, 31) +#define PHB_TCE_KILL 0x210 +#define PHB_TCE_KILL_ALL PPC_BIT(0) +#define PHB_TCE_SPEC_CTL 0x218 +#define PHB_IODA_ADDR 0x220 +#define PHB_IODA_AD_AUTOINC PPC_BIT(0) +#define PHB_IODA_AD_TSEL PPC_BITMASK(11, 15) +#define PHB_IODA_AD_TADR PPC_BITMASK(55, 63) +#define PHB_IODA_DATA0 0x228 +#define PHB_FFI_REQUEST 0x238 +#define PHB_FFI_LOCK_CLEAR PPC_BIT(3) +#define PHB_FFI_REQUEST_ISN PPC_BITMASK(49, 59) +#define PHB_FFI_LOCK 0x240 +#define PHB_FFI_LOCK_STATE PPC_BIT(0) +#define PHB_XIVE_UPDATE 0x248 /* Broken in DD1 */ +#define PHB_PHB3_GEN_CAP 0x250 +#define PHB_PHB3_TCE_CAP 0x258 +#define PHB_PHB3_IRQ_CAP 0x260 +#define PHB_PHB3_EEH_CAP 0x268 +#define PHB_IVC_INVALIDATE 0x2a0 +#define PHB_IVC_INVALIDATE_ALL PPC_BIT(0) +#define PHB_IVC_INVALIDATE_SID PPC_BITMASK(16, 31) +#define PHB_IVC_UPDATE 0x2a8 +#define PHB_IVC_UPDATE_ENABLE_P PPC_BIT(0) +#define PHB_IVC_UPDATE_ENABLE_Q PPC_BIT(1) +#define PHB_IVC_UPDATE_ENABLE_SERVER PPC_BIT(2) +#define PHB_IVC_UPDATE_ENABLE_PRI PPC_BIT(3) +#define PHB_IVC_UPDATE_ENABLE_GEN PPC_BIT(4) +#define PHB_IVC_UPDATE_ENABLE_CON PPC_BIT(5) +#define PHB_IVC_UPDATE_GEN_MATCH PPC_BITMASK(6, 7) +#define PHB_IVC_UPDATE_SERVER PPC_BITMASK(8, 23) +#define PHB_IVC_UPDATE_PRI PPC_BITMASK(24, 31) +#define PHB_IVC_UPDATE_GEN PPC_BITMASK(32, 33) +#define PHB_IVC_UPDATE_P PPC_BITMASK(34, 34) +#define PHB_IVC_UPDATE_Q PPC_BITMASK(35, 35) +#define PHB_IVC_UPDATE_SID PPC_BITMASK(48, 63) +#define PHB_PAPR_ERR_INJ_CTL 0x2b0 +#define PHB_PAPR_ERR_INJ_CTL_INB PPC_BIT(0) +#define PHB_PAPR_ERR_INJ_CTL_OUTB PPC_BIT(1) +#define PHB_PAPR_ERR_INJ_CTL_STICKY PPC_BIT(2) +#define PHB_PAPR_ERR_INJ_CTL_CFG PPC_BIT(3) +#define PHB_PAPR_ERR_INJ_CTL_RD PPC_BIT(4) +#define PHB_PAPR_ERR_INJ_CTL_WR PPC_BIT(5) +#define PHB_PAPR_ERR_INJ_CTL_FREEZE PPC_BIT(6) +#define PHB_PAPR_ERR_INJ_ADDR 0x2b8 +#define PHB_PAPR_ERR_INJ_ADDR_MMIO PPC_BITMASK(16, 63) +#define PHB_PAPR_ERR_INJ_MASK 0x2c0 +#define PHB_PAPR_ERR_INJ_MASK_CFG PPC_BITMASK(4, 11) +#define PHB_PAPR_ERR_INJ_MASK_MMIO PPC_BITMASK(16, 63) +#define PHB_ETU_ERR_SUMMARY 0x2c8 + +/* UTL registers */ +#define UTL_SYS_BUS_CONTROL 0x400 +#define UTL_STATUS 0x408 +#define UTL_SYS_BUS_AGENT_STATUS 0x410 +#define UTL_SYS_BUS_AGENT_ERR_SEVERITY 0x418 +#define UTL_SYS_BUS_AGENT_IRQ_EN 0x420 +#define UTL_SYS_BUS_BURST_SZ_CONF 0x440 +#define UTL_REVISION_ID 0x448 +#define UTL_BCLK_DOMAIN_DBG1 0x460 +#define UTL_BCLK_DOMAIN_DBG2 0x468 +#define UTL_BCLK_DOMAIN_DBG3 0x470 +#define UTL_BCLK_DOMAIN_DBG4 0x478 +#define UTL_BCLK_DOMAIN_DBG5 0x480 +#define UTL_BCLK_DOMAIN_DBG6 0x488 +#define UTL_OUT_POST_HDR_BUF_ALLOC 0x4c0 +#define UTL_OUT_POST_DAT_BUF_ALLOC 0x4d0 +#define UTL_IN_POST_HDR_BUF_ALLOC 0x4e0 +#define UTL_IN_POST_DAT_BUF_ALLOC 0x4f0 +#define UTL_OUT_NP_BUF_ALLOC 0x500 +#define UTL_IN_NP_BUF_ALLOC 0x510 +#define UTL_PCIE_TAGS_ALLOC 0x520 +#define UTL_GBIF_READ_TAGS_ALLOC 0x530 +#define UTL_PCIE_PORT_CONTROL 0x540 +#define UTL_PCIE_PORT_STATUS 0x548 +#define UTL_PCIE_PORT_ERROR_SEV 0x550 +#define UTL_PCIE_PORT_IRQ_EN 0x558 +#define UTL_RC_STATUS 0x560 +#define UTL_RC_ERR_SEVERITY 0x568 +#define UTL_RC_IRQ_EN 0x570 +#define UTL_EP_STATUS 0x578 +#define UTL_EP_ERR_SEVERITY 0x580 +#define UTL_EP_ERR_IRQ_EN 0x588 +#define UTL_PCI_PM_CTRL1 0x590 +#define UTL_PCI_PM_CTRL2 0x598 +#define UTL_GP_CTL1 0x5a0 +#define UTL_GP_CTL2 0x5a8 +#define UTL_PCLK_DOMAIN_DBG1 0x5b0 +#define UTL_PCLK_DOMAIN_DBG2 0x5b8 +#define UTL_PCLK_DOMAIN_DBG3 0x5c0 +#define UTL_PCLK_DOMAIN_DBG4 0x5c8 + +/* PCI-E Stack registers */ +#define PHB_PCIE_SYSTEM_CONFIG 0x600 +#define PHB_PCIE_BUS_NUMBER 0x608 +#define PHB_PCIE_SYSTEM_TEST 0x618 +#define PHB_PCIE_LINK_MANAGEMENT 0x630 +#define PHB_PCIE_LM_LINK_ACTIVE PPC_BIT(8) +#define PHB_PCIE_DLP_TRAIN_CTL 0x640 +#define PHB_PCIE_DLP_TCTX_DISABLE PPC_BIT(1) +#define PHB_PCIE_DLP_TCRX_DISABLED PPC_BIT(16) +#define PHB_PCIE_DLP_INBAND_PRESENCE PPC_BIT(19) +#define PHB_PCIE_DLP_TC_DL_LINKUP PPC_BIT(21) +#define PHB_PCIE_DLP_TC_DL_PGRESET PPC_BIT(22) +#define PHB_PCIE_DLP_TC_DL_LINKACT PPC_BIT(23) +#define PHB_PCIE_SLOP_LOOPBACK_STATUS 0x648 +#define PHB_PCIE_SYS_LINK_INIT 0x668 +#define PHB_PCIE_UTL_CONFIG 0x670 +#define PHB_PCIE_DLP_CONTROL 0x678 +#define PHB_PCIE_UTL_ERRLOG1 0x680 +#define PHB_PCIE_UTL_ERRLOG2 0x688 +#define PHB_PCIE_UTL_ERRLOG3 0x690 +#define PHB_PCIE_UTL_ERRLOG4 0x698 +#define PHB_PCIE_DLP_ERRLOG1 0x6a0 +#define PHB_PCIE_DLP_ERRLOG2 0x6a8 +#define PHB_PCIE_DLP_ERR_STATUS 0x6b0 +#define PHB_PCIE_DLP_ERR_COUNTERS 0x6b8 +#define PHB_PCIE_UTL_ERR_INJECT 0x6c0 +#define PHB_PCIE_TLDLP_ERR_INJECT 0x6c8 +#define PHB_PCIE_LANE_EQ_CNTL0 0x6d0 +#define PHB_PCIE_LANE_EQ_CNTL1 0x6d8 +#define PHB_PCIE_LANE_EQ_CNTL2 0x6e0 +#define PHB_PCIE_LANE_EQ_CNTL3 0x6e8 +#define PHB_PCIE_STRAPPING 0x700 + +/* Fundamental register set B */ +#define PHB_VERSION 0x800 +#define PHB_RESET 0x808 +#define PHB_CONTROL 0x810 +#define PHB_CTRL_IVE_128_BYTES PPC_BIT(24) +#define PHB_AIB_RX_CRED_INIT_TIMER 0x818 +#define PHB_AIB_RX_CMD_CRED 0x820 +#define PHB_AIB_RX_DATA_CRED 0x828 +#define PHB_AIB_TX_CMD_CRED 0x830 +#define PHB_AIB_TX_DATA_CRED 0x838 +#define PHB_AIB_TX_CHAN_MAPPING 0x840 +#define PHB_AIB_TAG_ENABLE 0x858 +#define PHB_AIB_FENCE_CTRL 0x860 +#define PHB_TCE_TAG_ENABLE 0x868 +#define PHB_TCE_WATERMARK 0x870 +#define PHB_TIMEOUT_CTRL1 0x878 +#define PHB_TIMEOUT_CTRL2 0x880 +#define PHB_Q_DMA_R 0x888 +#define PHB_Q_DMA_R_QUIESCE_DMA PPC_BIT(0) +#define PHB_Q_DMA_R_AUTORESET PPC_BIT(1) +#define PHB_Q_DMA_R_DMA_RESP_STATUS PPC_BIT(4) +#define PHB_Q_DMA_R_MMIO_RESP_STATUS PPC_BIT(5) +#define PHB_Q_DMA_R_TCE_RESP_STATUS PPC_BIT(6) +#define PHB_AIB_TAG_STATUS 0x900 +#define PHB_TCE_TAG_STATUS 0x908 + +/* FIR & Error registers */ +#define PHB_LEM_FIR_ACCUM 0xc00 +#define PHB_LEM_FIR_AND_MASK 0xc08 +#define PHB_LEM_FIR_OR_MASK 0xc10 +#define PHB_LEM_ERROR_MASK 0xc18 +#define PHB_LEM_ERROR_AND_MASK 0xc20 +#define PHB_LEM_ERROR_OR_MASK 0xc28 +#define PHB_LEM_ACTION0 0xc30 +#define PHB_LEM_ACTION1 0xc38 +#define PHB_LEM_WOF 0xc40 +#define PHB_ERR_STATUS 0xc80 +#define PHB_ERR1_STATUS 0xc88 +#define PHB_ERR_INJECT 0xc90 +#define PHB_ERR_LEM_ENABLE 0xc98 +#define PHB_ERR_IRQ_ENABLE 0xca0 +#define PHB_ERR_FREEZE_ENABLE 0xca8 +#define PHB_ERR_AIB_FENCE_ENABLE 0xcb0 +#define PHB_ERR_LOG_0 0xcc0 +#define PHB_ERR_LOG_1 0xcc8 +#define PHB_ERR_STATUS_MASK 0xcd0 +#define PHB_ERR1_STATUS_MASK 0xcd8 + +#define PHB_OUT_ERR_STATUS 0xd00 +#define PHB_OUT_ERR1_STATUS 0xd08 +#define PHB_OUT_ERR_INJECT 0xd10 +#define PHB_OUT_ERR_LEM_ENABLE 0xd18 +#define PHB_OUT_ERR_IRQ_ENABLE 0xd20 +#define PHB_OUT_ERR_FREEZE_ENABLE 0xd28 +#define PHB_OUT_ERR_AIB_FENCE_ENABLE 0xd30 +#define PHB_OUT_ERR_LOG_0 0xd40 +#define PHB_OUT_ERR_LOG_1 0xd48 +#define PHB_OUT_ERR_STATUS_MASK 0xd50 +#define PHB_OUT_ERR1_STATUS_MASK 0xd58 + +#define PHB_INA_ERR_STATUS 0xd80 +#define PHB_INA_ERR1_STATUS 0xd88 +#define PHB_INA_ERR_INJECT 0xd90 +#define PHB_INA_ERR_LEM_ENABLE 0xd98 +#define PHB_INA_ERR_IRQ_ENABLE 0xda0 +#define PHB_INA_ERR_FREEZE_ENABLE 0xda8 +#define PHB_INA_ERR_AIB_FENCE_ENABLE 0xdb0 +#define PHB_INA_ERR_LOG_0 0xdc0 +#define PHB_INA_ERR_LOG_1 0xdc8 +#define PHB_INA_ERR_STATUS_MASK 0xdd0 +#define PHB_INA_ERR1_STATUS_MASK 0xdd8 + +#define PHB_INB_ERR_STATUS 0xe00 +#define PHB_INB_ERR1_STATUS 0xe08 +#define PHB_INB_ERR_INJECT 0xe10 +#define PHB_INB_ERR_LEM_ENABLE 0xe18 +#define PHB_INB_ERR_IRQ_ENABLE 0xe20 +#define PHB_INB_ERR_FREEZE_ENABLE 0xe28 +#define PHB_INB_ERR_AIB_FENCE_ENABLE 0xe30 +#define PHB_INB_ERR_LOG_0 0xe40 +#define PHB_INB_ERR_LOG_1 0xe48 +#define PHB_INB_ERR_STATUS_MASK 0xe50 +#define PHB_INB_ERR1_STATUS_MASK 0xe58 + +/* Performance monitor & Debug registers */ +#define PHB_TRACE_CONTROL 0xf80 +#define PHB_PERFMON_CONFIG 0xf88 +#define PHB_PERFMON_CTR0 0xf90 +#define PHB_PERFMON_CTR1 0xf98 +#define PHB_PERFMON_CTR2 0xfa0 +#define PHB_PERFMON_CTR3 0xfa8 +#define PHB_HOTPLUG_OVERRIDE 0xfb0 +#define PHB_HPOVR_FORCE_RESAMPLE PPC_BIT(9) +#define PHB_HPOVR_PRESENCE_A PPC_BIT(10) +#define PHB_HPOVR_PRESENCE_B PPC_BIT(11) +#define PHB_HPOVR_LINK_ACTIVE PPC_BIT(12) +#define PHB_HPOVR_LINK_BIFURCATED PPC_BIT(13) +#define PHB_HPOVR_LINK_LANE_SWAPPED PPC_BIT(14) + +/* + * IODA2 on-chip tables + */ + +#define IODA2_TBL_LIST 1 +#define IODA2_TBL_LXIVT 2 +#define IODA2_TBL_IVC_CAM 3 +#define IODA2_TBL_RBA 4 +#define IODA2_TBL_RCAM 5 +#define IODA2_TBL_MRT 6 +#define IODA2_TBL_PESTA 7 +#define IODA2_TBL_PESTB 8 +#define IODA2_TBL_TVT 9 +#define IODA2_TBL_TCAM 10 +#define IODA2_TBL_TDR 11 +#define IODA2_TBL_M64BT 16 +#define IODA2_TBL_M32DT 17 +#define IODA2_TBL_PEEV 20 + +/* LXIVT */ +#define IODA2_LXIVT_SERVER PPC_BITMASK(8, 23) +#define IODA2_LXIVT_PRIORITY PPC_BITMASK(24, 31) +#define IODA2_LXIVT_NODE_ID PPC_BITMASK(56, 63) + +/* IVT */ +#define IODA2_IVT_SERVER PPC_BITMASK(0, 23) +#define IODA2_IVT_PRIORITY PPC_BITMASK(24, 31) +#define IODA2_IVT_GEN PPC_BITMASK(37, 38) +#define IODA2_IVT_P PPC_BITMASK(39, 39) +#define IODA2_IVT_Q PPC_BITMASK(47, 47) +#define IODA2_IVT_PE PPC_BITMASK(48, 63) + +/* TVT */ +#define IODA2_TVT_TABLE_ADDR PPC_BITMASK(0, 47) +#define IODA2_TVT_NUM_LEVELS PPC_BITMASK(48, 50) +#define IODA2_TVE_1_LEVEL 0 +#define IODA2_TVE_2_LEVELS 1 +#define IODA2_TVE_3_LEVELS 2 +#define IODA2_TVE_4_LEVELS 3 +#define IODA2_TVE_5_LEVELS 4 +#define IODA2_TVT_TCE_TABLE_SIZE PPC_BITMASK(51, 55) +#define IODA2_TVT_IO_PSIZE PPC_BITMASK(59, 63) + +/* PESTA */ +#define IODA2_PESTA_MMIO_FROZEN PPC_BIT(0) + +/* PESTB */ +#define IODA2_PESTB_DMA_STOPPED PPC_BIT(0) + +/* M32DT */ +#define IODA2_M32DT_PE PPC_BITMASK(8, 15) + +/* M64BT */ +#define IODA2_M64BT_ENABLE PPC_BIT(0) +#define IODA2_M64BT_SINGLE_PE PPC_BIT(1) +#define IODA2_M64BT_BASE PPC_BITMASK(2, 31) +#define IODA2_M64BT_MASK PPC_BITMASK(34, 63) +#define IODA2_M64BT_SINGLE_BASE PPC_BITMASK(2, 26) +#define IODA2_M64BT_PE_HI PPC_BITMASK(27, 31) +#define IODA2_M64BT_SINGLE_MASK PPC_BITMASK(34, 58) +#define IODA2_M64BT_PE_LOW PPC_BITMASK(59, 63) + +/* + * IODA2 in-memory tables + */ + +/* PEST + * + * 2x8 bytes entries, PEST0 and PEST1 + */ + +#define IODA2_PEST0_MMIO_CAUSE PPC_BIT(2) +#define IODA2_PEST0_CFG_READ PPC_BIT(3) +#define IODA2_PEST0_CFG_WRITE PPC_BIT(4) +#define IODA2_PEST0_TTYPE PPC_BITMASK(5, 7) +#define PEST_TTYPE_DMA_WRITE 0 +#define PEST_TTYPE_MSI 1 +#define PEST_TTYPE_DMA_READ 2 +#define PEST_TTYPE_DMA_READ_RESP 3 +#define PEST_TTYPE_MMIO_LOAD 4 +#define PEST_TTYPE_MMIO_STORE 5 +#define PEST_TTYPE_OTHER 7 +#define IODA2_PEST0_CA_RETURN PPC_BIT(8) +#define IODA2_PEST0_UTL_RTOS_TIMEOUT PPC_BIT(8) /* Same bit as CA return */ +#define IODA2_PEST0_UR_RETURN PPC_BIT(9) +#define IODA2_PEST0_UTL_NONFATAL PPC_BIT(10) +#define IODA2_PEST0_UTL_FATAL PPC_BIT(11) +#define IODA2_PEST0_PARITY_UE PPC_BIT(13) +#define IODA2_PEST0_UTL_CORRECTABLE PPC_BIT(14) +#define IODA2_PEST0_UTL_INTERRUPT PPC_BIT(15) +#define IODA2_PEST0_MMIO_XLATE PPC_BIT(16) +#define IODA2_PEST0_IODA2_ERROR PPC_BIT(16) /* Same bit as MMIO xlate */ +#define IODA2_PEST0_TCE_PAGE_FAULT PPC_BIT(18) +#define IODA2_PEST0_TCE_ACCESS_FAULT PPC_BIT(19) +#define IODA2_PEST0_DMA_RESP_TIMEOUT PPC_BIT(20) +#define IODA2_PEST0_AIB_SIZE_INVALID PPC_BIT(21) +#define IODA2_PEST0_LEM_BIT PPC_BITMASK(26, 31) +#define IODA2_PEST0_RID PPC_BITMASK(32, 47) +#define IODA2_PEST0_MSI_DATA PPC_BITMASK(48, 63) + +#define IODA2_PEST1_FAIL_ADDR PPC_BITMASK(3, 63) + + +#endif /* PCI_HOST_PNV_PHB3_REGS_H */ diff --git a/include/hw/ppc/pnv.h b/include/hw/ppc/pnv.h index aa9d5dd16df1..8087937b132b 100644 --- a/include/hw/ppc/pnv.h +++ b/include/hw/ppc/pnv.h @@ -30,6 +30,7 @@ #include "hw/ppc/pnv_homer.h" #include "hw/ppc/pnv_xive.h" #include "hw/ppc/pnv_core.h" +#include "hw/pci-host/pnv_phb3.h" #define TYPE_PNV_CHIP "pnv-chip" #define PNV_CHIP(obj) OBJECT_CHECK(PnvChip, (obj), TYPE_PNV_CHIP) @@ -58,6 +59,8 @@ typedef struct PnvChip { uint64_t cores_mask; void *cores; + uint32_t num_phbs; + MemoryRegion xscom_mmio; MemoryRegion xscom; AddressSpace xscom_as; @@ -79,6 +82,9 @@ typedef struct Pnv8Chip { Pnv8Psi psi; PnvOCC occ; PnvHomer homer; + +#define PNV8_CHIP_PHB3_MAX 4 + PnvPHB3 phbs[PNV8_CHIP_PHB3_MAX]; } Pnv8Chip; #define TYPE_PNV9_CHIP "pnv9-chip" @@ -107,6 +113,7 @@ typedef struct PnvChipClass { PnvChipType chip_type; uint64_t chip_cfam_id; uint64_t cores_mask; + uint32_t num_phbs; DeviceRealize parent_realize; diff --git a/include/hw/ppc/pnv_xscom.h b/include/hw/ppc/pnv_xscom.h index 67641ed27800..6b7d0d908f8a 100644 --- a/include/hw/ppc/pnv_xscom.h +++ b/include/hw/ppc/pnv_xscom.h @@ -87,6 +87,15 @@ typedef struct PnvXScomInterfaceClass { #define PNV9_XSCOM_XIVE_BASE 0x5013000 #define PNV9_XSCOM_XIVE_SIZE 0x300 +#define PNV_XSCOM_PBCQ_NEST_BASE 0x2012000 +#define PNV_XSCOM_PBCQ_NEST_SIZE 0x46 + +#define PNV_XSCOM_PBCQ_PCI_BASE 0x9012000 +#define PNV_XSCOM_PBCQ_PCI_SIZE 0x15 + +#define PNV_XSCOM_PBCQ_SPCI_BASE 0x9013c00 +#define PNV_XSCOM_PBCQ_SPCI_SIZE 0x5 + extern void pnv_xscom_realize(PnvChip *chip, uint64_t size, Error **errp); extern int pnv_dt_xscom(PnvChip *chip, void *fdt, int offset); diff --git a/include/hw/ppc/xics.h b/include/hw/ppc/xics.h index 64a2c8862a72..ab0954ac58e5 100644 --- a/include/hw/ppc/xics.h +++ b/include/hw/ppc/xics.h @@ -174,6 +174,7 @@ void icp_set_mfrr(ICPState *icp, uint8_t mfrr); uint32_t icp_accept(ICPState *ss); uint32_t icp_ipoll(ICPState *ss, uint32_t *mfrr); void icp_eoi(ICPState *icp, uint32_t xirr); +void icp_irq(ICSState *ics, int server, int nr, uint8_t priority); void ics_simple_write_xive(ICSState *ics, int nr, int server, uint8_t priority, uint8_t saved_priority); diff --git a/hw/intc/xics.c b/hw/intc/xics.c index b2fca2975cc4..be1e0da058be 100644 --- a/hw/intc/xics.c +++ b/hw/intc/xics.c @@ -232,7 +232,7 @@ void icp_eoi(ICPState *icp, uint32_t xirr) } } -static void icp_irq(ICSState *ics, int server, int nr, uint8_t priority) +void icp_irq(ICSState *ics, int server, int nr, uint8_t priority) { ICPState *icp = xics_icp_get(ics->xics, server); diff --git a/hw/pci-host/pnv_phb3.c b/hw/pci-host/pnv_phb3.c new file mode 100644 index 000000000000..086ba01fa52f --- /dev/null +++ b/hw/pci-host/pnv_phb3.c @@ -0,0 +1,1191 @@ +/* + * QEMU PowerPC PowerNV PHB3 model + * + * Copyright (c) 2014-2018, IBM Corporation. + * + * This code is licensed under the GPL version 2 or later. See the + * COPYING file in the top-level directory. + */ +#include "qemu/osdep.h" +#include "qemu/log.h" +#include "qapi/visitor.h" +#include "qapi/error.h" +#include "qemu-common.h" +#include "hw/pci-host/pnv_phb3_regs.h" +#include "hw/pci-host/pnv_phb3.h" +#include "hw/pci/pcie_host.h" +#include "hw/pci/pcie_port.h" +#include "hw/ppc/pnv.h" +#include "hw/irq.h" +#include "hw/qdev-properties.h" + +#define phb3_error(fmt, ...) \ + qemu_log_mask(LOG_GUEST_ERROR, "phb3: " fmt "\n", ## __VA_ARGS__) + + +static PCIDevice *pnv_phb3_find_cfg_dev(PnvPHB3 *phb) +{ + PCIHostState *pci = PCI_HOST_BRIDGE(phb); + uint64_t addr = phb->regs[PHB_CONFIG_ADDRESS >> 3]; + uint8_t bus, devfn; + + if (!(addr >> 63)) { + return NULL; + } + bus = (addr >> 52) & 0xff; + devfn = (addr >> 44) & 0xff; + + return pci_find_device(pci->bus, bus, devfn); +} + +/* + * The CONFIG_DATA register expects little endian accesses, but as the + * region is big endian, we have to swap the value. + */ +static void pnv_phb3_config_write(PnvPHB3 *phb, unsigned off, + unsigned size, uint64_t val) +{ + uint32_t cfg_addr, limit; + PCIDevice *pdev; + + pdev = pnv_phb3_find_cfg_dev(phb); + if (!pdev) { + return; + } + cfg_addr = (phb->regs[PHB_CONFIG_ADDRESS >> 3] >> 32) & 0xffc; + cfg_addr |= off; + limit = pci_config_size(pdev); + if (limit <= cfg_addr) { + /* conventional pci device can be behind pcie-to-pci bridge. + 256 <= addr < 4K has no effects. */ + return; + } + switch (size) { + case 1: + break; + case 2: + val = bswap16(val); + break; + case 4: + val = bswap32(val); + break; + default: + g_assert_not_reached(); + } + pci_host_config_write_common(pdev, cfg_addr, limit, val, size); +} + +static uint64_t pnv_phb3_config_read(PnvPHB3 *phb, unsigned off, + unsigned size) +{ + uint32_t cfg_addr, limit; + PCIDevice *pdev; + uint64_t val; + + pdev = pnv_phb3_find_cfg_dev(phb); + if (!pdev) { + return ~0ull; + } + cfg_addr = (phb->regs[PHB_CONFIG_ADDRESS >> 3] >> 32) & 0xffc; + cfg_addr |= off; + limit = pci_config_size(pdev); + if (limit <= cfg_addr) { + /* conventional pci device can be behind pcie-to-pci bridge. + 256 <= addr < 4K has no effects. */ + return ~0ull; + } + val = pci_host_config_read_common(pdev, cfg_addr, limit, size); + switch (size) { + case 1: + return val; + case 2: + return bswap16(val); + case 4: + return bswap32(val); + default: + g_assert_not_reached(); + } +} + +static void pnv_phb3_check_m32(PnvPHB3 *phb) +{ + uint64_t base, start, size; + MemoryRegion *parent; + PnvPBCQState *pbcq = &phb->pbcq; + + if (phb->m32_mapped) { + memory_region_del_subregion(phb->mr_m32.container, &phb->mr_m32); + phb->m32_mapped = false; + } + + if (!(phb->regs[PHB_PHB3_CONFIG >> 3] & PHB_PHB3C_M32_EN)) { + return; + } + + /* Grab geometry from registers */ + base = phb->regs[PHB_M32_BASE_ADDR >> 3]; + start = phb->regs[PHB_M32_START_ADDR >> 3]; + size = ~(phb->regs[PHB_M32_BASE_MASK >> 3] | 0xfffc000000000000ull) + 1; + + /* Check if it matches an enabled MMIO region in the PBCQ */ + if (pbcq->mmio0_mapped && base >= pbcq->mmio0_base && + (base + size) <= (pbcq->mmio0_base + pbcq->mmio0_size)) { + parent = &pbcq->mmbar0; + base -= pbcq->mmio0_base; + } else if (pbcq->mmio1_mapped && base >= pbcq->mmio1_base && + (base + size) <= (pbcq->mmio1_base + pbcq->mmio1_size)) { + parent = &pbcq->mmbar1; + base -= pbcq->mmio1_base; + } else { + return; + } + + /* Create alias */ + memory_region_init_alias(&phb->mr_m32, OBJECT(phb), "phb3-m32", + &phb->pci_mmio, start, size); + memory_region_add_subregion(parent, base, &phb->mr_m32); + phb->m32_mapped = true; +} + +static void pnv_phb3_check_m64(PnvPHB3 *phb, uint32_t index) +{ + uint64_t base, start, size, m64; + MemoryRegion *parent; + PnvPBCQState *pbcq = &phb->pbcq; + + if (phb->m64_mapped[index]) { + /* Should we destroy it in RCU friendly way... ? */ + memory_region_del_subregion(phb->mr_m64[index].container, + &phb->mr_m64[index]); + phb->m64_mapped[index] = false; + } + + /* Get table entry */ + m64 = phb->ioda_M64BT[index]; + + if (!(m64 & IODA2_M64BT_ENABLE)) { + return; + } + + /* Grab geometry from registers */ + base = GETFIELD(IODA2_M64BT_BASE, m64) << 20; + if (m64 & IODA2_M64BT_SINGLE_PE) { + base &= ~0x1ffffffull; + } + size = GETFIELD(IODA2_M64BT_MASK, m64) << 20; + size |= 0xfffc000000000000ull; + size = ~size + 1; + start = base | (phb->regs[PHB_M64_UPPER_BITS >> 3]); + + /* Check if it matches an enabled MMIO region in the PBCQ */ + if (pbcq->mmio0_mapped && base >= pbcq->mmio0_base && + (base + size) <= (pbcq->mmio0_base + pbcq->mmio0_size)) { + parent = &pbcq->mmbar0; + base -= pbcq->mmio0_base; + } else if (pbcq->mmio1_mapped && base >= pbcq->mmio1_base && + (base + size) <= (pbcq->mmio1_base + pbcq->mmio1_size)) { + parent = &pbcq->mmbar1; + base -= pbcq->mmio1_base; + } else { + return; + } + + /* Create alias */ + memory_region_init_alias(&phb->mr_m64[index], OBJECT(phb), "phb3-m64", + &phb->pci_mmio, start, size); + memory_region_add_subregion(parent, base, &phb->mr_m64[index]); + phb->m64_mapped[index] = true; +} + +static void pnv_phb3_check_all_m64s(PnvPHB3 *phb) +{ + uint64_t i; + + for (i = 0; i < PNV_PHB3_NUM_M64; i++) { + pnv_phb3_check_m64(phb, i); + } +} + +static void pnv_phb3_lxivt_write(PnvPHB3 *phb, unsigned idx, uint64_t val) +{ + uint8_t server, prio; + + phb->ioda_LXIVT[idx] = val & (IODA2_LXIVT_SERVER | + IODA2_LXIVT_PRIORITY | + IODA2_LXIVT_NODE_ID); + server = GETFIELD(IODA2_LXIVT_SERVER, val); + prio = GETFIELD(IODA2_LXIVT_PRIORITY, val); + + /* + * The low order 2 bits are the link pointer (Type II interrupts). + * Shift back to get a valid IRQ server. + */ + server >>= 2; + + ics_simple_write_xive(&phb->lsis, idx, server, prio, prio); +} + +static uint64_t *pnv_phb3_ioda_access(PnvPHB3 *phb, + unsigned *out_table, unsigned *out_idx) +{ + uint64_t adreg = phb->regs[PHB_IODA_ADDR >> 3]; + unsigned int index = GETFIELD(PHB_IODA_AD_TADR, adreg); + unsigned int table = GETFIELD(PHB_IODA_AD_TSEL, adreg); + unsigned int mask; + uint64_t *tptr = NULL; + + switch (table) { + case IODA2_TBL_LIST: + tptr = phb->ioda_LIST; + mask = 7; + break; + case IODA2_TBL_LXIVT: + tptr = phb->ioda_LXIVT; + mask = 7; + break; + case IODA2_TBL_IVC_CAM: + case IODA2_TBL_RBA: + mask = 31; + break; + case IODA2_TBL_RCAM: + mask = 63; + break; + case IODA2_TBL_MRT: + mask = 7; + break; + case IODA2_TBL_PESTA: + case IODA2_TBL_PESTB: + mask = 255; + break; + case IODA2_TBL_TVT: + tptr = phb->ioda_TVT; + mask = 511; + break; + case IODA2_TBL_TCAM: + case IODA2_TBL_TDR: + mask = 63; + break; + case IODA2_TBL_M64BT: + tptr = phb->ioda_M64BT; + mask = 15; + break; + case IODA2_TBL_M32DT: + tptr = phb->ioda_MDT; + mask = 255; + break; + case IODA2_TBL_PEEV: + tptr = phb->ioda_PEEV; + mask = 3; + break; + default: + phb3_error("invalid IODA table %d", table); + return NULL; + } + index &= mask; + if (out_idx) { + *out_idx = index; + } + if (out_table) { + *out_table = table; + } + if (tptr) { + tptr += index; + } + if (adreg & PHB_IODA_AD_AUTOINC) { + index = (index + 1) & mask; + adreg = SETFIELD(PHB_IODA_AD_TADR, adreg, index); + } + phb->regs[PHB_IODA_ADDR >> 3] = adreg; + return tptr; +} + +static uint64_t pnv_phb3_ioda_read(PnvPHB3 *phb) +{ + unsigned table; + uint64_t *tptr; + + tptr = pnv_phb3_ioda_access(phb, &table, NULL); + if (!tptr) { + /* Return 0 on unsupported tables, not ff's */ + return 0; + } + return *tptr; +} + +static void pnv_phb3_ioda_write(PnvPHB3 *phb, uint64_t val) +{ + unsigned table, idx; + uint64_t *tptr; + + tptr = pnv_phb3_ioda_access(phb, &table, &idx); + if (!tptr) { + return; + } + + /* Handle side effects */ + switch (table) { + case IODA2_TBL_LXIVT: + pnv_phb3_lxivt_write(phb, idx, val); + break; + case IODA2_TBL_M64BT: + *tptr = val; + pnv_phb3_check_m64(phb, idx); + break; + default: + *tptr = val; + } +} + +/* This is called whenever the PHB LSI, MSI source ID register or + * the PBCQ irq filters are written. + */ +void pnv_phb3_remap_irqs(PnvPHB3 *phb) +{ + ICSState *ics = &phb->lsis; + uint32_t local, global, count, mask, comp; + uint64_t baren; + PnvPBCQState *pbcq = &phb->pbcq; + + /* First check if we are enabled. Unlike real HW we don't separate TX and RX + * so we enable if both are set + */ + baren = pbcq->nest_regs[PBCQ_NEST_BAR_EN]; + if (!(baren & PBCQ_NEST_BAR_EN_IRSN_RX) || + !(baren & PBCQ_NEST_BAR_EN_IRSN_TX)) { + ics->offset = 0; + return; + } + + /* Grab local LSI source ID */ + local = GETFIELD(PHB_LSI_SRC_ID, phb->regs[PHB_LSI_SOURCE_ID >> 3]) << 3; + + /* Grab global one and compare */ + global = GETFIELD(PBCQ_NEST_LSI_SRC, + pbcq->nest_regs[PBCQ_NEST_LSI_SRC_ID]) << 3; + if (global != local) { + /* This happens during initialization, let's come back when we + * are properly configured + */ + ics->offset = 0; + return; + } + + /* Get the base on the powerbus */ + comp = GETFIELD(PBCQ_NEST_IRSN_COMP, + pbcq->nest_regs[PBCQ_NEST_IRSN_COMPARE]); + mask = GETFIELD(PBCQ_NEST_IRSN_COMP, + pbcq->nest_regs[PBCQ_NEST_IRSN_MASK]); + count = ((~mask) + 1) & 0x7ffff; + phb->total_irq = count; + + /* Sanity checks */ + if ((global + PNV_PHB3_NUM_LSI) > count) { + phb3_error("LSIs out of reach: LSI base=%d total irq=%d", global, + count); + } + + if (count > 2048) { + phb3_error("More interrupts than supported: %d", count); + } + + if ((comp & mask) != comp) { + phb3_error("IRQ compare bits not in mask: comp=0x%x mask=0x%x", + comp, mask); + comp &= mask; + } + /* Setup LSI offset */ + ics->offset = comp + global; + + /* Setup MSI offset */ + pnv_phb3_msi_update_config(&phb->msis, comp, count - PNV_PHB3_NUM_LSI); +} + +static void pnv_phb3_lsi_src_id_write(PnvPHB3 *phb, uint64_t val) +{ + /* Sanitize content */ + val &= PHB_LSI_SRC_ID; + phb->regs[PHB_LSI_SOURCE_ID >> 3] = val; + pnv_phb3_remap_irqs(phb); +} + +static void pnv_phb3_rtc_invalidate(PnvPHB3 *phb, uint64_t val) +{ + PnvPhb3DMASpace *ds; + + /* Always invalidate all for now ... */ + QLIST_FOREACH(ds, &phb->dma_spaces, list) { + ds->pe_num = PHB_INVALID_PE; + } +} + + +static void pnv_phb3_update_msi_regions(PnvPhb3DMASpace *ds) +{ + uint64_t cfg = ds->phb->regs[PHB_PHB3_CONFIG >> 3]; + + if (cfg & PHB_PHB3C_32BIT_MSI_EN) { + if (!ds->msi32_mapped) { + memory_region_add_subregion(MEMORY_REGION(&ds->dma_mr), + 0xffff0000, &ds->msi32_mr); + ds->msi32_mapped = true; + } + } else { + if (ds->msi32_mapped) { + memory_region_del_subregion(MEMORY_REGION(&ds->dma_mr), + &ds->msi32_mr); + ds->msi32_mapped = false; + } + } + + if (cfg & PHB_PHB3C_64BIT_MSI_EN) { + if (!ds->msi64_mapped) { + memory_region_add_subregion(MEMORY_REGION(&ds->dma_mr), + (1ull << 60), &ds->msi64_mr); + ds->msi64_mapped = true; + } + } else { + if (ds->msi64_mapped) { + memory_region_del_subregion(MEMORY_REGION(&ds->dma_mr), + &ds->msi64_mr); + ds->msi64_mapped = false; + } + } +} + +static void pnv_phb3_update_all_msi_regions(PnvPHB3 *phb) +{ + PnvPhb3DMASpace *ds; + + QLIST_FOREACH(ds, &phb->dma_spaces, list) { + pnv_phb3_update_msi_regions(ds); + } +} + +void pnv_phb3_reg_write(void *opaque, hwaddr off, uint64_t val, unsigned size) +{ + PnvPHB3 *phb = opaque; + bool changed; + + /* Special case configuration data */ + if ((off & 0xfffc) == PHB_CONFIG_DATA) { + pnv_phb3_config_write(phb, off & 0x3, size, val); + return; + } + + /* Other registers are 64-bit only */ + if (size != 8 || off & 0x7) { + phb3_error("Invalid register access, offset: 0x%"PRIx64" size: %d", + off, size); + return; + } + + /* Handle masking & filtering */ + switch (off) { + case PHB_M64_UPPER_BITS: + val &= 0xfffc000000000000ull; + break; + case PHB_Q_DMA_R: + /* This is enough logic to make SW happy but we aren't actually + * quiescing the DMAs + */ + if (val & PHB_Q_DMA_R_AUTORESET) { + val = 0; + } else { + val &= PHB_Q_DMA_R_QUIESCE_DMA; + } + break; + /* LEM stuff */ + case PHB_LEM_FIR_AND_MASK: + phb->regs[PHB_LEM_FIR_ACCUM >> 3] &= val; + return; + case PHB_LEM_FIR_OR_MASK: + phb->regs[PHB_LEM_FIR_ACCUM >> 3] |= val; + return; + case PHB_LEM_ERROR_AND_MASK: + phb->regs[PHB_LEM_ERROR_MASK >> 3] &= val; + return; + case PHB_LEM_ERROR_OR_MASK: + phb->regs[PHB_LEM_ERROR_MASK >> 3] |= val; + return; + case PHB_LEM_WOF: + val = 0; + break; + } + + /* Record whether it changed */ + changed = phb->regs[off >> 3] != val; + + /* Store in register cache first */ + phb->regs[off >> 3] = val; + + /* Handle side effects */ + switch (off) { + case PHB_PHB3_CONFIG: + if (changed) { + pnv_phb3_update_all_msi_regions(phb); + } + /* fall through */ + case PHB_M32_BASE_ADDR: + case PHB_M32_BASE_MASK: + case PHB_M32_START_ADDR: + if (changed) { + pnv_phb3_check_m32(phb); + } + break; + case PHB_M64_UPPER_BITS: + if (changed) { + pnv_phb3_check_all_m64s(phb); + } + break; + case PHB_LSI_SOURCE_ID: + if (changed) { + pnv_phb3_lsi_src_id_write(phb, val); + } + break; + + /* IODA table accesses */ + case PHB_IODA_DATA0: + pnv_phb3_ioda_write(phb, val); + break; + + /* RTC invalidation */ + case PHB_RTC_INVALIDATE: + pnv_phb3_rtc_invalidate(phb, val); + break; + + /* FFI request */ + case PHB_FFI_REQUEST: + pnv_phb3_msi_ffi(&phb->msis, val); + break; + + /* Silent simple writes */ + case PHB_CONFIG_ADDRESS: + case PHB_IODA_ADDR: + case PHB_TCE_KILL: + case PHB_TCE_SPEC_CTL: + case PHB_PEST_BAR: + case PHB_PELTV_BAR: + case PHB_RTT_BAR: + case PHB_RBA_BAR: + case PHB_IVT_BAR: + case PHB_FFI_LOCK: + case PHB_LEM_FIR_ACCUM: + case PHB_LEM_ERROR_MASK: + case PHB_LEM_ACTION0: + case PHB_LEM_ACTION1: + break; + + /* Noise on anything else */ + default: + qemu_log_mask(LOG_UNIMP, "phb3: reg_write 0x%"PRIx64"=%"PRIx64"\n", + off, val); + } +} + +uint64_t pnv_phb3_reg_read(void *opaque, hwaddr off, unsigned size) +{ + PnvPHB3 *phb = opaque; + PCIHostState *pci = PCI_HOST_BRIDGE(phb); + uint64_t val; + + if ((off & 0xfffc) == PHB_CONFIG_DATA) { + return pnv_phb3_config_read(phb, off & 0x3, size); + } + + /* Other registers are 64-bit only */ + if (size != 8 || off & 0x7) { + phb3_error("Invalid register access, offset: 0x%"PRIx64" size: %d", + off, size); + return ~0ull; + } + + /* Default read from cache */ + val = phb->regs[off >> 3]; + + switch (off) { + /* Simulate venice DD2.0 */ + case PHB_VERSION: + return 0x000000a300000005ull; + case PHB_PCIE_SYSTEM_CONFIG: + return 0x441100fc30000000; + + /* IODA table accesses */ + case PHB_IODA_DATA0: + return pnv_phb3_ioda_read(phb); + + /* Link training always appears trained */ + case PHB_PCIE_DLP_TRAIN_CTL: + if (!pci_find_device(pci->bus, 1, 0)) + return 0; + return PHB_PCIE_DLP_INBAND_PRESENCE | PHB_PCIE_DLP_TC_DL_LINKACT; + + /* FFI Lock */ + case PHB_FFI_LOCK: + /* Set lock and return previous value */ + phb->regs[off >> 3] |= PHB_FFI_LOCK_STATE; + return val; + + /* DMA read sync: make it look like it's complete */ + case PHB_DMARD_SYNC: + return PHB_DMARD_SYNC_COMPLETE; + + /* Silent simple reads */ + case PHB_PHB3_CONFIG: + case PHB_M32_BASE_ADDR: + case PHB_M32_BASE_MASK: + case PHB_M32_START_ADDR: + case PHB_CONFIG_ADDRESS: + case PHB_IODA_ADDR: + case PHB_RTC_INVALIDATE: + case PHB_TCE_KILL: + case PHB_TCE_SPEC_CTL: + case PHB_PEST_BAR: + case PHB_PELTV_BAR: + case PHB_RTT_BAR: + case PHB_RBA_BAR: + case PHB_IVT_BAR: + case PHB_M64_UPPER_BITS: + case PHB_LEM_FIR_ACCUM: + case PHB_LEM_ERROR_MASK: + case PHB_LEM_ACTION0: + case PHB_LEM_ACTION1: + break; + + /* Noise on anything else */ + default: + qemu_log_mask(LOG_UNIMP, "phb3: reg_read 0x%"PRIx64"=%"PRIx64"\n", + off, val); + } + return val; +} + +static const MemoryRegionOps pnv_phb3_reg_ops = { + .read = pnv_phb3_reg_read, + .write = pnv_phb3_reg_write, + .valid.min_access_size = 1, + .valid.max_access_size = 8, + .impl.min_access_size = 1, + .impl.max_access_size = 8, + .endianness = DEVICE_BIG_ENDIAN, +}; + +static int pnv_phb3_map_irq(PCIDevice *pci_dev, int irq_num) +{ + /* Check that out properly ... */ + return irq_num & 3; +} + +static void pnv_phb3_set_irq(void *opaque, int irq_num, int level) +{ + PnvPHB3 *phb = opaque; + + /* LSI only ... */ + if (irq_num > 3) { + phb3_error("Unknown IRQ to set %d", irq_num); + } + qemu_set_irq(phb->qirqs[irq_num], level); +} + +static bool pnv_phb3_resolve_pe(PnvPhb3DMASpace *ds) +{ + uint64_t rtt, addr; + uint16_t rte; + int bus_num; + + /* Already resolved ? */ + if (ds->pe_num != PHB_INVALID_PE) { + return true; + } + + /* We need to lookup the RTT */ + rtt = ds->phb->regs[PHB_RTT_BAR >> 3]; + if (!(rtt & PHB_RTT_BAR_ENABLE)) { + phb3_error("DMA with RTT BAR disabled !"); + /* Set error bits ? fence ? ... */ + return false; + } + + /* Read RTE */ + bus_num = pci_bus_num(ds->bus); + addr = rtt & PHB_RTT_BASE_ADDRESS_MASK; + addr += 2 * ((bus_num << 8) | ds->devfn); + if (dma_memory_read(&address_space_memory, addr, &rte, sizeof(rte))) { + phb3_error("Failed to read RTT entry at 0x%"PRIx64, addr); + /* Set error bits ? fence ? ... */ + return false; + } + rte = be16_to_cpu(rte); + + /* Fail upon reading of invalid PE# */ + if (rte >= PNV_PHB3_NUM_PE) { + phb3_error("RTE for RID 0x%x invalid (%04x", ds->devfn, rte); + /* Set error bits ? fence ? ... */ + return false; + } + ds->pe_num = rte; + return true; +} + +static void pnv_phb3_translate_tve(PnvPhb3DMASpace *ds, hwaddr addr, + bool is_write, uint64_t tve, + IOMMUTLBEntry *tlb) +{ + uint64_t tta = GETFIELD(IODA2_TVT_TABLE_ADDR, tve); + int32_t lev = GETFIELD(IODA2_TVT_NUM_LEVELS, tve); + uint32_t tts = GETFIELD(IODA2_TVT_TCE_TABLE_SIZE, tve); + uint32_t tps = GETFIELD(IODA2_TVT_IO_PSIZE, tve); + + /* Invalid levels */ + if (lev > 4) { + phb3_error("Invalid #levels in TVE %d", lev); + return; + } + + /* IO Page Size of 0 means untranslated, else use TCEs */ + if (tps == 0) { + /* We only support non-translate in top window + * XXX FIX THAT, Venice/Murano support it on bottom window + * above 4G and Naples suports it on everything + */ + if (!(tve & PPC_BIT(51))) { + phb3_error("xlate for invalid non-translate TVE"); + return; + } + /* XXX Handle boundaries */ + + /* XXX Use 4k pages like q35 ... for now */ + tlb->iova = addr & 0xfffffffffffff000ull; + tlb->translated_addr = addr & 0x0003fffffffff000ull; + tlb->addr_mask = 0xfffull; + tlb->perm = IOMMU_RW; + } else { + uint32_t tce_shift, tbl_shift, sh; + uint64_t base, taddr, tce, tce_mask; + + /* TVE disabled ? */ + if (tts == 0) { + phb3_error("xlate for invalid translated TVE"); + return; + } + + /* Address bits per bottom level TCE entry */ + tce_shift = tps + 11; + + /* Address bits per table level */ + tbl_shift = tts + 8; + + /* Top level table base address */ + base = tta << 12; + + /* Total shift to first level */ + sh = tbl_shift * lev + tce_shift; + + /* XXX Multi-level untested */ + while ((lev--) >= 0) { + /* Grab the TCE address */ + taddr = base | (((addr >> sh) & ((1ul << tbl_shift) - 1)) << 3); + if (dma_memory_read(&address_space_memory, taddr, &tce, + sizeof(tce))) { + phb3_error("Failed to read TCE at 0x%"PRIx64, taddr); + return; + } + tce = be64_to_cpu(tce); + + /* Check permission for indirect TCE */ + if ((lev >= 0) && !(tce & 3)) { + phb3_error("Invalid indirect TCE at 0x%"PRIx64, taddr); + phb3_error(" xlate %"PRIx64":%c TVE=%"PRIx64, addr, + is_write ? 'W' : 'R', tve); + phb3_error(" tta=%"PRIx64" lev=%d tts=%d tps=%d", + tta, lev, tts, tps); + return; + } + sh -= tbl_shift; + base = tce & ~0xfffull; + } + + /* We exit the loop with TCE being the final TCE */ + tce_mask = ~((1ull << tce_shift) - 1); + tlb->iova = addr & tce_mask; + tlb->translated_addr = tce & tce_mask; + tlb->addr_mask = ~tce_mask; + tlb->perm = tce & 3; + if ((is_write & !(tce & 2)) || ((!is_write) && !(tce & 1))) { + phb3_error("TCE access fault at 0x%"PRIx64, taddr); + phb3_error(" xlate %"PRIx64":%c TVE=%"PRIx64, addr, + is_write ? 'W' : 'R', tve); + phb3_error(" tta=%"PRIx64" lev=%d tts=%d tps=%d", + tta, lev, tts, tps); + } + } +} + +static IOMMUTLBEntry pnv_phb3_translate_iommu(IOMMUMemoryRegion *iommu, + hwaddr addr, + IOMMUAccessFlags flag, + int iommu_idx) +{ + PnvPhb3DMASpace *ds = container_of(iommu, PnvPhb3DMASpace, dma_mr); + int tve_sel; + uint64_t tve, cfg; + IOMMUTLBEntry ret = { + .target_as = &address_space_memory, + .iova = addr, + .translated_addr = 0, + .addr_mask = ~(hwaddr)0, + .perm = IOMMU_NONE, + }; + + /* Resolve PE# */ + if (!pnv_phb3_resolve_pe(ds)) { + phb3_error("Failed to resolve PE# for bus @%p (%d) devfn 0x%x", + ds->bus, pci_bus_num(ds->bus), ds->devfn); + return ret; + } + + /* Check top bits */ + switch (addr >> 60) { + case 00: + /* DMA or 32-bit MSI ? */ + cfg = ds->phb->regs[PHB_PHB3_CONFIG >> 3]; + if ((cfg & PHB_PHB3C_32BIT_MSI_EN) && + ((addr & 0xffffffffffff0000ull) == 0xffff0000ull)) { + phb3_error("xlate on 32-bit MSI region"); + return ret; + } + /* Choose TVE XXX Use PHB3 Control Register */ + tve_sel = (addr >> 59) & 1; + tve = ds->phb->ioda_TVT[ds->pe_num * 2 + tve_sel]; + pnv_phb3_translate_tve(ds, addr, flag & IOMMU_WO, tve, &ret); + break; + case 01: + phb3_error("xlate on 64-bit MSI region"); + break; + default: + phb3_error("xlate on unsupported address 0x%"PRIx64, addr); + } + return ret; +} + +#define TYPE_PNV_PHB3_IOMMU_MEMORY_REGION "pnv-phb3-iommu-memory-region" +#define PNV_PHB3_IOMMU_MEMORY_REGION(obj) \ + OBJECT_CHECK(IOMMUMemoryRegion, (obj), TYPE_PNV_PHB3_IOMMU_MEMORY_REGION) + +static void pnv_phb3_iommu_memory_region_class_init(ObjectClass *klass, + void *data) +{ + IOMMUMemoryRegionClass *imrc = IOMMU_MEMORY_REGION_CLASS(klass); + + imrc->translate = pnv_phb3_translate_iommu; +} + +static const TypeInfo pnv_phb3_iommu_memory_region_info = { + .parent = TYPE_IOMMU_MEMORY_REGION, + .name = TYPE_PNV_PHB3_IOMMU_MEMORY_REGION, + .class_init = pnv_phb3_iommu_memory_region_class_init, +}; + +/* + * MSI/MSIX memory region implementation. + * The handler handles both MSI and MSIX. + */ +static void pnv_phb3_msi_write(void *opaque, hwaddr addr, + uint64_t data, unsigned size) +{ + PnvPhb3DMASpace *ds = opaque; + + /* Resolve PE# */ + if (!pnv_phb3_resolve_pe(ds)) { + phb3_error("Failed to resolve PE# for bus @%p (%d) devfn 0x%x", + ds->bus, pci_bus_num(ds->bus), ds->devfn); + return; + } + + pnv_phb3_msi_send(&ds->phb->msis, addr, data, ds->pe_num); +} + +/* There is no .read as the read result is undefined by PCI spec */ +static uint64_t pnv_phb3_msi_read(void *opaque, hwaddr addr, unsigned size) +{ + phb3_error("invalid read @ 0x%" HWADDR_PRIx, addr); + return -1; +} + +static const MemoryRegionOps pnv_phb3_msi_ops = { + .read = pnv_phb3_msi_read, + .write = pnv_phb3_msi_write, + .endianness = DEVICE_LITTLE_ENDIAN +}; + +static AddressSpace *pnv_phb3_dma_iommu(PCIBus *bus, void *opaque, int devfn) +{ + PnvPHB3 *phb = opaque; + PnvPhb3DMASpace *ds; + + QLIST_FOREACH(ds, &phb->dma_spaces, list) { + if (ds->bus == bus && ds->devfn == devfn) { + break; + } + } + + if (ds == NULL) { + ds = g_malloc0(sizeof(PnvPhb3DMASpace)); + ds->bus = bus; + ds->devfn = devfn; + ds->pe_num = PHB_INVALID_PE; + ds->phb = phb; + memory_region_init_iommu(&ds->dma_mr, sizeof(ds->dma_mr), + TYPE_PNV_PHB3_IOMMU_MEMORY_REGION, + OBJECT(phb), "phb3_iommu", UINT64_MAX); + address_space_init(&ds->dma_as, MEMORY_REGION(&ds->dma_mr), + "phb3_iommu"); + memory_region_init_io(&ds->msi32_mr, OBJECT(phb), &pnv_phb3_msi_ops, + ds, "msi32", 0x10000); + memory_region_init_io(&ds->msi64_mr, OBJECT(phb), &pnv_phb3_msi_ops, + ds, "msi64", 0x100000); + pnv_phb3_update_msi_regions(ds); + + QLIST_INSERT_HEAD(&phb->dma_spaces, ds, list); + } + return &ds->dma_as; +} + +static void pnv_phb3_instance_init(Object *obj) +{ + PnvPHB3 *phb = PNV_PHB3(obj); + + QLIST_INIT(&phb->dma_spaces); + + /* LSI sources */ + object_initialize(&phb->lsis, sizeof(phb->lsis), TYPE_ICS_SIMPLE); + object_property_add_child(obj, "lsi", OBJECT(&phb->lsis), NULL); + + /* Default init ... will be fixed by HW inits */ + phb->lsis.offset = 0; + + /* MSI sources */ + object_initialize(&phb->msis, sizeof(phb->msis), TYPE_PHB3_MSI); + object_property_add_const_link(OBJECT(&phb->msis), "phb", obj, + &error_abort); + object_property_add_child(obj, "msi", OBJECT(&phb->msis), NULL); + + /* Power Bus Common Queue */ + object_initialize(&phb->pbcq, sizeof(phb->pbcq), TYPE_PNV_PBCQ); + object_property_add_const_link(OBJECT(&phb->pbcq), "phb", obj, + &error_abort); + object_property_add_child(obj, "pbcq", OBJECT(&phb->pbcq), NULL); + + /* Root Port */ + object_initialize(&phb->root, sizeof(phb->root), TYPE_PNV_PHB3_ROOT_PORT); + object_property_add_child(obj, "root", OBJECT(&phb->root), NULL); + qdev_prop_set_int32(DEVICE(&phb->root), "addr", PCI_DEVFN(0, 0)); + qdev_prop_set_bit(DEVICE(&phb->root), "multifunction", false); +} + +static void pnv_phb3_realize(DeviceState *dev, Error **errp) +{ + PnvPHB3 *phb = PNV_PHB3(dev); + PCIHostState *pci = PCI_HOST_BRIDGE(dev); + PnvMachineState *pnv = PNV_MACHINE(qdev_get_machine()); + Error *local_err = NULL; + int i; + + if (!pnv_get_chip(pnv, phb->chip_id)) { + error_setg(errp, "invalid chip id: %d", phb->chip_id); + return; + } + + if (phb->phb_id >= PNV8_CHIP_PHB3_MAX) { + error_setg(errp, "invalid PHB index: %d", phb->phb_id); + return; + } + + /* LSI sources */ + object_property_add_const_link(OBJECT(&phb->lsis), "xics", OBJECT(pnv), + &error_abort); + object_property_set_int(OBJECT(&phb->lsis), PNV_PHB3_NUM_LSI, "nr-irqs", + &error_abort); + object_property_set_bool(OBJECT(&phb->lsis), true, "realized", &local_err); + if (local_err) { + error_propagate(errp, local_err); + return; + } + + for (i = 0; i < phb->lsis.nr_irqs; i++) { + ics_set_irq_type(&phb->lsis, i, true); + } + + phb->qirqs = qemu_allocate_irqs(ics_simple_set_irq, &phb->lsis, + phb->lsis.nr_irqs); + + /* MSI sources */ + object_property_add_const_link(OBJECT(&phb->msis), "xics", OBJECT(pnv), + &error_abort); + object_property_set_int(OBJECT(&phb->msis), PHB3_MAX_MSI, "nr-irqs", + &error_abort); + object_property_set_bool(OBJECT(&phb->msis), true, "realized", &local_err); + if (local_err) { + error_propagate(errp, local_err); + return; + } + + /* Power Bus Common Queue */ + object_property_set_bool(OBJECT(&phb->pbcq), true, "realized", &local_err); + if (local_err) { + error_propagate(errp, local_err); + return; + } + + /* Controller Registers */ + memory_region_init_io(&phb->mr_regs, OBJECT(phb), &pnv_phb3_reg_ops, phb, + "phb3-regs", 0x1000); + + /* PHB3 doesn't support IO space. However, qemu gets very upset if + * we don't have an IO region to anchor IO BARs onto so we just + * initialize one which we never hook up to anything + */ + memory_region_init(&phb->pci_io, OBJECT(phb), "pci-io", 0x10000); + memory_region_init(&phb->pci_mmio, OBJECT(phb), "pci-mmio", + PCI_MMIO_TOTAL_SIZE); + + pci->bus = pci_register_root_bus(dev, "root-bus", + pnv_phb3_set_irq, pnv_phb3_map_irq, phb, + &phb->pci_mmio, &phb->pci_io, + 0, 4, TYPE_PNV_PHB3_ROOT_BUS); + + pci_setup_iommu(pci->bus, pnv_phb3_dma_iommu, phb); + + /* Add a single Root port */ + qdev_prop_set_uint8(DEVICE(&phb->root), "chassis", phb->chip_id); + qdev_prop_set_uint16(DEVICE(&phb->root), "slot", phb->phb_id); + qdev_set_parent_bus(DEVICE(&phb->root), BUS(pci->bus)); + qdev_init_nofail(DEVICE(&phb->root)); +} + +void pnv_phb3_update_regions(PnvPHB3 *phb) +{ + PnvPBCQState *pbcq = &phb->pbcq; + + /* Unmap first always */ + if (phb->regs_mapped) { + memory_region_del_subregion(&pbcq->phbbar, &phb->mr_regs); + phb->regs_mapped = false; + } + + /* Map registers if enabled */ + if (pbcq->phb_mapped) { + /* XXX We should use the PHB BAR 2 register but we don't ... */ + memory_region_add_subregion(&pbcq->phbbar, 0, &phb->mr_regs); + phb->regs_mapped = true; + } + + /* Check/update m32 */ + if (phb->m32_mapped) { + pnv_phb3_check_m32(phb); + } + pnv_phb3_check_all_m64s(phb); +} + +static const char *pnv_phb3_root_bus_path(PCIHostState *host_bridge, + PCIBus *rootbus) +{ + PnvPHB3 *phb = PNV_PHB3(host_bridge); + + snprintf(phb->bus_path, sizeof(phb->bus_path), "00%02x:%02x", + phb->chip_id, phb->phb_id); + return phb->bus_path; +} + +static Property pnv_phb3_properties[] = { + DEFINE_PROP_UINT32("index", PnvPHB3, phb_id, 0), + DEFINE_PROP_UINT32("chip-id", PnvPHB3, chip_id, 0), + DEFINE_PROP_END_OF_LIST(), +}; + +static void pnv_phb3_class_init(ObjectClass *klass, void *data) +{ + PCIHostBridgeClass *hc = PCI_HOST_BRIDGE_CLASS(klass); + DeviceClass *dc = DEVICE_CLASS(klass); + + hc->root_bus_path = pnv_phb3_root_bus_path; + dc->realize = pnv_phb3_realize; + dc->props = pnv_phb3_properties; + set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories); +} + +static const TypeInfo pnv_phb3_type_info = { + .name = TYPE_PNV_PHB3, + .parent = TYPE_PCIE_HOST_BRIDGE, + .instance_size = sizeof(PnvPHB3), + .class_init = pnv_phb3_class_init, + .instance_init = pnv_phb3_instance_init, +}; + +static void pnv_phb3_root_bus_class_init(ObjectClass *klass, void *data) +{ + BusClass *k = BUS_CLASS(klass); + + /* + * PHB3 has only a single root complex. Enforce the limit on the + * parent bus + */ + k->max_dev = 1; +} + +static const TypeInfo pnv_phb3_root_bus_info = { + .name = TYPE_PNV_PHB3_ROOT_BUS, + .parent = TYPE_PCIE_BUS, + .class_init = pnv_phb3_root_bus_class_init, + .interfaces = (InterfaceInfo[]) { + { INTERFACE_PCIE_DEVICE }, + { } + }, +}; + +static void pnv_phb3_root_port_realize(DeviceState *dev, Error **errp) +{ + PCIERootPortClass *rpc = PCIE_ROOT_PORT_GET_CLASS(dev); + Error *local_err = NULL; + + rpc->parent_realize(dev, &local_err); + if (local_err) { + error_propagate(errp, local_err); + return; + } +} + +static void pnv_phb3_root_port_class_init(ObjectClass *klass, void *data) +{ + DeviceClass *dc = DEVICE_CLASS(klass); + PCIDeviceClass *k = PCI_DEVICE_CLASS(klass); + PCIERootPortClass *rpc = PCIE_ROOT_PORT_CLASS(klass); + + dc->desc = "IBM PHB3 PCIE Root Port"; + + device_class_set_parent_realize(dc, pnv_phb3_root_port_realize, + &rpc->parent_realize); + + k->vendor_id = PCI_VENDOR_ID_IBM; + k->device_id = 0x03dc; + k->revision = 0; + + rpc->exp_offset = 0x48; + rpc->aer_offset = 0x100; +} + +static const TypeInfo pnv_phb3_root_port_info = { + .name = TYPE_PNV_PHB3_ROOT_PORT, + .parent = TYPE_PCIE_ROOT_PORT, + .instance_size = sizeof(PnvPHB3RootPort), + .class_init = pnv_phb3_root_port_class_init, +}; + +static void pnv_phb3_register_types(void) +{ + type_register_static(&pnv_phb3_root_bus_info); + type_register_static(&pnv_phb3_root_port_info); + type_register_static(&pnv_phb3_type_info); + type_register_static(&pnv_phb3_iommu_memory_region_info); +} + +type_init(pnv_phb3_register_types) diff --git a/hw/pci-host/pnv_phb3_msi.c b/hw/pci-host/pnv_phb3_msi.c new file mode 100644 index 000000000000..e689c3c1e18c --- /dev/null +++ b/hw/pci-host/pnv_phb3_msi.c @@ -0,0 +1,320 @@ +/* + * QEMU PowerPC PowerNV PHB3 model + * + * Copyright (c) 2014-2018, IBM Corporation. + * + * This code is licensed under the GPL version 2 or later. See the + * COPYING file in the top-level directory. + */ +#include "qemu/osdep.h" +#include "qemu/log.h" +#include "qapi/error.h" +#include "qemu-common.h" +#include "hw/pci-host/pnv_phb3_regs.h" +#include "hw/pci-host/pnv_phb3.h" +#include "hw/ppc/pnv.h" +#include "hw/pci/msi.h" +#include "monitor/monitor.h" +#include "hw/irq.h" +#include "hw/qdev-properties.h" +#include "sysemu/reset.h" + +static uint64_t phb3_msi_ive_addr(PnvPHB3 *phb, int srcno) +{ + uint64_t ivtbar = phb->regs[PHB_IVT_BAR >> 3]; + uint64_t phbctl = phb->regs[PHB_CONTROL >> 3]; + + if (!(ivtbar & PHB_IVT_BAR_ENABLE)) { + qemu_log_mask(LOG_GUEST_ERROR, "Failed access to disable IVT BAR !"); + return 0; + } + + if (srcno >= (ivtbar & PHB_IVT_LENGTH_MASK)) { + qemu_log_mask(LOG_GUEST_ERROR, "MSI out of bounds (%d vs 0x%"PRIx64")", + srcno, (uint64_t) (ivtbar & PHB_IVT_LENGTH_MASK)); + return 0; + } + + ivtbar &= PHB_IVT_BASE_ADDRESS_MASK; + + if (phbctl & PHB_CTRL_IVE_128_BYTES) { + return ivtbar + 128 * srcno; + } else { + return ivtbar + 16 * srcno; + } +} + +static bool phb3_msi_read_ive(PnvPHB3 *phb, int srcno, uint64_t *out_ive) +{ + uint64_t ive_addr, ive; + + ive_addr = phb3_msi_ive_addr(phb, srcno); + if (!ive_addr) { + return false; + } + + if (dma_memory_read(&address_space_memory, ive_addr, &ive, sizeof(ive))) { + qemu_log_mask(LOG_GUEST_ERROR, "Failed to read IVE at 0x%" PRIx64, + ive_addr); + return false; + } + *out_ive = be64_to_cpu(ive); + + return true; +} + +static void phb3_msi_set_p(Phb3MsiState *msi, int srcno, uint8_t gen) +{ + uint64_t ive_addr; + uint8_t p = 0x01 | (gen << 1); + + ive_addr = phb3_msi_ive_addr(msi->phb, srcno); + if (!ive_addr) { + return; + } + + if (dma_memory_write(&address_space_memory, ive_addr + 4, &p, 1)) { + qemu_log_mask(LOG_GUEST_ERROR, + "Failed to write IVE (set P) at 0x%" PRIx64, ive_addr); + } +} + +static void phb3_msi_set_q(Phb3MsiState *msi, int srcno) +{ + uint64_t ive_addr; + uint8_t q = 0x01; + + ive_addr = phb3_msi_ive_addr(msi->phb, srcno); + if (!ive_addr) { + return; + } + + if (dma_memory_write(&address_space_memory, ive_addr + 5, &q, 1)) { + qemu_log_mask(LOG_GUEST_ERROR, + "Failed to write IVE (set Q) at 0x%" PRIx64, ive_addr); + } +} + +static void phb3_msi_try_send(Phb3MsiState *msi, int srcno, bool force) +{ + ICSState *ics = ICS_BASE(msi); + uint64_t ive; + uint64_t server, prio, pq, gen; + + if (!phb3_msi_read_ive(msi->phb, srcno, &ive)) { + return; + } + + server = GETFIELD(IODA2_IVT_SERVER, ive); + prio = GETFIELD(IODA2_IVT_PRIORITY, ive); + if (!force) { + pq = GETFIELD(IODA2_IVT_Q, ive) | (GETFIELD(IODA2_IVT_P, ive) << 1); + } else { + pq = 0; + } + gen = GETFIELD(IODA2_IVT_GEN, ive); + + /* + * The low order 2 bits are the link pointer (Type II interrupts). + * Shift back to get a valid IRQ server. + */ + server >>= 2; + + switch (pq) { + case 0: /* 00 */ + if (prio == 0xff) { + /* Masked, set Q */ + phb3_msi_set_q(msi, srcno); + } else { + /* Enabled, set P and send */ + phb3_msi_set_p(msi, srcno, gen); + icp_irq(ics, server, srcno + ics->offset, prio); + } + break; + case 2: /* 10 */ + /* Already pending, set Q */ + phb3_msi_set_q(msi, srcno); + break; + case 1: /* 01 */ + case 3: /* 11 */ + default: + /* Just drop stuff if Q already set */ + break; + } +} + +static void phb3_msi_set_irq(void *opaque, int srcno, int val) +{ + Phb3MsiState *msi = PHB3_MSI(opaque); + + if (val) { + phb3_msi_try_send(msi, srcno, false); + } +} + + +void pnv_phb3_msi_send(Phb3MsiState *msi, uint64_t addr, uint16_t data, + int32_t dev_pe) +{ + ICSState *ics = ICS_BASE(msi); + uint64_t ive; + uint16_t pe; + uint32_t src = ((addr >> 4) & 0xffff) | (data & 0x1f); + + if (src >= ics->nr_irqs) { + qemu_log_mask(LOG_GUEST_ERROR, "MSI %d out of bounds", src); + return; + } + if (dev_pe >= 0) { + if (!phb3_msi_read_ive(msi->phb, src, &ive)) { + return; + } + pe = GETFIELD(IODA2_IVT_PE, ive); + if (pe != dev_pe) { + qemu_log_mask(LOG_GUEST_ERROR, + "MSI %d send by PE#%d but assigned to PE#%d", + src, dev_pe, pe); + return; + } + } + qemu_irq_pulse(msi->qirqs[src]); +} + +void pnv_phb3_msi_ffi(Phb3MsiState *msi, uint64_t val) +{ + /* Emit interrupt */ + pnv_phb3_msi_send(msi, val, 0, -1); + + /* Clear FFI lock */ + msi->phb->regs[PHB_FFI_LOCK >> 3] = 0; +} + +static void phb3_msi_reject(ICSState *ics, uint32_t nr) +{ + Phb3MsiState *msi = PHB3_MSI(ics); + unsigned int srcno = nr - ics->offset; + unsigned int idx = srcno >> 6; + unsigned int bit = 1ull << (srcno & 0x3f); + + assert(srcno < PHB3_MAX_MSI); + + msi->rba[idx] |= bit; + msi->rba_sum |= (1u << idx); +} + +static void phb3_msi_resend(ICSState *ics) +{ + Phb3MsiState *msi = PHB3_MSI(ics); + unsigned int i, j; + + if (msi->rba_sum == 0) { + return; + } + + for (i = 0; i < 32; i++) { + if ((msi->rba_sum & (1u << i)) == 0) { + continue; + } + msi->rba_sum &= ~(1u << i); + for (j = 0; j < 64; j++) { + if ((msi->rba[i] & (1ull << j)) == 0) { + continue; + } + msi->rba[i] &= ~(1u << j); + phb3_msi_try_send(msi, i * 64 + j, true); + } + } +} + +static void phb3_msi_reset(DeviceState *dev) +{ + Phb3MsiState *msi = PHB3_MSI(dev); + ICSStateClass *icsc = ICS_BASE_GET_CLASS(dev); + + icsc->parent_reset(dev); + + memset(msi->rba, 0, sizeof(msi->rba)); + msi->rba_sum = 0; +} + +static void phb3_msi_reset_handler(void *dev) +{ + phb3_msi_reset(dev); +} + +void pnv_phb3_msi_update_config(Phb3MsiState *msi, uint32_t base, + uint32_t count) +{ + ICSState *ics = ICS_BASE(msi); + + if (count > PHB3_MAX_MSI) { + count = PHB3_MAX_MSI; + } + ics->nr_irqs = count; + ics->offset = base; +} + +static void phb3_msi_realize(DeviceState *dev, Error **errp) +{ + Phb3MsiState *msi = PHB3_MSI(dev); + ICSState *ics = ICS_BASE(msi); + ICSStateClass *icsc = ICS_BASE_GET_CLASS(ics); + Object *obj; + Error *local_err = NULL; + + icsc->parent_realize(dev, &local_err); + if (local_err) { + error_propagate(errp, local_err); + return; + } + + obj = object_property_get_link(OBJECT(dev), "phb", &local_err); + if (!obj) { + error_propagate(errp, local_err); + error_prepend(errp, "required link 'phb' not found: "); + return; + } + msi->phb = PNV_PHB3(obj); + + msi->qirqs = qemu_allocate_irqs(phb3_msi_set_irq, msi, ics->nr_irqs); + + qemu_register_reset(phb3_msi_reset_handler, dev); +} + +static void phb3_msi_instance_init(Object *obj) +{ + ICSState *ics = ICS_BASE(obj); + + /* Will be overriden later */ + ics->offset = 0; +} + +static void phb3_msi_class_init(ObjectClass *klass, void *data) +{ + DeviceClass *dc = DEVICE_CLASS(klass); + ICSStateClass *isc = ICS_BASE_CLASS(klass); + + device_class_set_parent_realize(dc, phb3_msi_realize, + &isc->parent_realize); + device_class_set_parent_reset(dc, phb3_msi_reset, + &isc->parent_reset); + + isc->reject = phb3_msi_reject; + isc->resend = phb3_msi_resend; +} + +static const TypeInfo phb3_msi_info = { + .name = TYPE_PHB3_MSI, + .parent = TYPE_ICS_BASE, + .instance_size = sizeof(Phb3MsiState), + .class_init = phb3_msi_class_init, + .class_size = sizeof(ICSStateClass), + .instance_init = phb3_msi_instance_init, +}; + +static void pnv_phb3_msi_register_types(void) +{ + type_register_static(&phb3_msi_info); +} + +type_init(pnv_phb3_msi_register_types) diff --git a/hw/pci-host/pnv_phb3_pbcq.c b/hw/pci-host/pnv_phb3_pbcq.c new file mode 100644 index 000000000000..34df3ac684e4 --- /dev/null +++ b/hw/pci-host/pnv_phb3_pbcq.c @@ -0,0 +1,347 @@ +/* + * QEMU PowerPC PowerNV PHB3 model + * + * Copyright (c) 2014-2018, IBM Corporation. + * + * This code is licensed under the GPL version 2 or later. See the + * COPYING file in the top-level directory. + */ +#include "qemu/osdep.h" +#include "qapi/error.h" +#include "qemu-common.h" +#include "qemu/log.h" +#include "target/ppc/cpu.h" +#include "hw/ppc/fdt.h" +#include "hw/pci-host/pnv_phb3_regs.h" +#include "hw/pci-host/pnv_phb3.h" +#include "hw/ppc/pnv_xscom.h" +#include "hw/pci/pci_bridge.h" +#include "hw/pci/pci_bus.h" + +#include <libfdt.h> + +static uint64_t pnv_pbcq_nest_xscom_read(void *opaque, hwaddr addr, + unsigned size) +{ + PnvPBCQState *pbcq = PNV_PBCQ(opaque); + uint32_t offset = addr >> 3; + + return pbcq->nest_regs[offset]; +} + +static uint64_t pnv_pbcq_pci_xscom_read(void *opaque, hwaddr addr, + unsigned size) +{ + PnvPBCQState *pbcq = PNV_PBCQ(opaque); + uint32_t offset = addr >> 3; + + return pbcq->pci_regs[offset]; +} + +static uint64_t pnv_pbcq_spci_xscom_read(void *opaque, hwaddr addr, + unsigned size) +{ + PnvPBCQState *pbcq = PNV_PBCQ(opaque); + uint32_t offset = addr >> 3; + + if (offset == PBCQ_SPCI_ASB_DATA) { + return pnv_phb3_reg_read(pbcq->phb, + pbcq->spci_regs[PBCQ_SPCI_ASB_ADDR], 8); + } + return pbcq->spci_regs[offset]; +} + +static void pnv_pbcq_update_map(PnvPBCQState *pbcq) +{ + uint64_t bar_en = pbcq->nest_regs[PBCQ_NEST_BAR_EN]; + uint64_t bar, mask, size; + + /* + * NOTE: This will really not work well if those are remapped + * after the PHB has created its sub regions. We could do better + * if we had a way to resize regions but we don't really care + * that much in practice as the stuff below really only happens + * once early during boot + */ + + /* Handle unmaps */ + if (pbcq->mmio0_mapped && !(bar_en & PBCQ_NEST_BAR_EN_MMIO0)) { + memory_region_del_subregion(get_system_memory(), &pbcq->mmbar0); + pbcq->mmio0_mapped = false; + } + if (pbcq->mmio1_mapped && !(bar_en & PBCQ_NEST_BAR_EN_MMIO1)) { + memory_region_del_subregion(get_system_memory(), &pbcq->mmbar1); + pbcq->mmio1_mapped = false; + } + if (pbcq->phb_mapped && !(bar_en & PBCQ_NEST_BAR_EN_PHB)) { + memory_region_del_subregion(get_system_memory(), &pbcq->phbbar); + pbcq->phb_mapped = false; + } + + /* Update PHB */ + pnv_phb3_update_regions(pbcq->phb); + + /* Handle maps */ + if (!pbcq->mmio0_mapped && (bar_en & PBCQ_NEST_BAR_EN_MMIO0)) { + bar = pbcq->nest_regs[PBCQ_NEST_MMIO_BAR0] >> 14; + mask = pbcq->nest_regs[PBCQ_NEST_MMIO_MASK0]; + size = ((~mask) >> 14) + 1; + memory_region_init(&pbcq->mmbar0, OBJECT(pbcq), "pbcq-mmio0", size); + memory_region_add_subregion(get_system_memory(), bar, &pbcq->mmbar0); + pbcq->mmio0_mapped = true; + pbcq->mmio0_base = bar; + pbcq->mmio0_size = size; + } + if (!pbcq->mmio1_mapped && (bar_en & PBCQ_NEST_BAR_EN_MMIO1)) { + bar = pbcq->nest_regs[PBCQ_NEST_MMIO_BAR1] >> 14; + mask = pbcq->nest_regs[PBCQ_NEST_MMIO_MASK1]; + size = ((~mask) >> 14) + 1; + memory_region_init(&pbcq->mmbar1, OBJECT(pbcq), "pbcq-mmio1", size); + memory_region_add_subregion(get_system_memory(), bar, &pbcq->mmbar1); + pbcq->mmio1_mapped = true; + pbcq->mmio1_base = bar; + pbcq->mmio1_size = size; + } + if (!pbcq->phb_mapped && (bar_en & PBCQ_NEST_BAR_EN_PHB)) { + bar = pbcq->nest_regs[PBCQ_NEST_PHB_BAR] >> 14; + size = 0x1000; + memory_region_init(&pbcq->phbbar, OBJECT(pbcq), "pbcq-phb", size); + memory_region_add_subregion(get_system_memory(), bar, &pbcq->phbbar); + pbcq->phb_mapped = true; + } + + /* Update PHB */ + pnv_phb3_update_regions(pbcq->phb); +} + +static void pnv_pbcq_nest_xscom_write(void *opaque, hwaddr addr, + uint64_t val, unsigned size) +{ + PnvPBCQState *pbcq = PNV_PBCQ(opaque); + uint32_t reg = addr >> 3; + + switch (reg) { + case PBCQ_NEST_MMIO_BAR0: + case PBCQ_NEST_MMIO_BAR1: + case PBCQ_NEST_MMIO_MASK0: + case PBCQ_NEST_MMIO_MASK1: + if (pbcq->nest_regs[PBCQ_NEST_BAR_EN] & + (PBCQ_NEST_BAR_EN_MMIO0 | + PBCQ_NEST_BAR_EN_MMIO1)) { + qemu_log_mask(LOG_GUEST_ERROR, + "PHB3: Changing enabled BAR unsupported\n"); + } + pbcq->nest_regs[reg] = val & 0xffffffffc0000000ull; + break; + case PBCQ_NEST_PHB_BAR: + if (pbcq->nest_regs[PBCQ_NEST_BAR_EN] & PBCQ_NEST_BAR_EN_PHB) { + qemu_log_mask(LOG_GUEST_ERROR, + "PHB3: Changing enabled BAR unsupported\n"); + } + pbcq->nest_regs[reg] = val & 0xfffffffffc000000ull; + break; + case PBCQ_NEST_BAR_EN: + pbcq->nest_regs[reg] = val & 0xf800000000000000ull; + pnv_pbcq_update_map(pbcq); + pnv_phb3_remap_irqs(pbcq->phb); + break; + case PBCQ_NEST_IRSN_COMPARE: + case PBCQ_NEST_IRSN_MASK: + pbcq->nest_regs[reg] = val & PBCQ_NEST_IRSN_COMP; + pnv_phb3_remap_irqs(pbcq->phb); + break; + case PBCQ_NEST_LSI_SRC_ID: + pbcq->nest_regs[reg] = val & PBCQ_NEST_LSI_SRC; + pnv_phb3_remap_irqs(pbcq->phb); + break; + } + + /* XXX Don't error out on other regs for now ... */ +} + +static void pnv_pbcq_pci_xscom_write(void *opaque, hwaddr addr, + uint64_t val, unsigned size) +{ + PnvPBCQState *pbcq = PNV_PBCQ(opaque); + uint32_t reg = addr >> 3; + + switch (reg) { + case PBCQ_PCI_BAR2: + pbcq->pci_regs[reg] = val & 0xfffffffffc000000ull; + pnv_pbcq_update_map(pbcq); + break; + } + + /* XXX Don't error out on other regs for now ... */ +} + +static void pnv_pbcq_spci_xscom_write(void *opaque, hwaddr addr, + uint64_t val, unsigned size) +{ + PnvPBCQState *pbcq = PNV_PBCQ(opaque); + uint32_t reg = addr >> 3; + + switch (reg) { + case PBCQ_SPCI_ASB_ADDR: + pbcq->spci_regs[reg] = val & 0xfff; + break; + case PBCQ_SPCI_ASB_STATUS: + pbcq->spci_regs[reg] &= ~val; + break; + case PBCQ_SPCI_ASB_DATA: + pnv_phb3_reg_write(pbcq->phb, pbcq->spci_regs[PBCQ_SPCI_ASB_ADDR], + val, 8); + break; + case PBCQ_SPCI_AIB_CAPP_EN: + case PBCQ_SPCI_CAPP_SEC_TMR: + break; + } + + /* XXX Don't error out on other regs for now ... */ +} + +static const MemoryRegionOps pnv_pbcq_nest_xscom_ops = { + .read = pnv_pbcq_nest_xscom_read, + .write = pnv_pbcq_nest_xscom_write, + .valid.min_access_size = 8, + .valid.max_access_size = 8, + .impl.min_access_size = 8, + .impl.max_access_size = 8, + .endianness = DEVICE_BIG_ENDIAN, +}; + +static const MemoryRegionOps pnv_pbcq_pci_xscom_ops = { + .read = pnv_pbcq_pci_xscom_read, + .write = pnv_pbcq_pci_xscom_write, + .valid.min_access_size = 8, + .valid.max_access_size = 8, + .impl.min_access_size = 8, + .impl.max_access_size = 8, + .endianness = DEVICE_BIG_ENDIAN, +}; + +static const MemoryRegionOps pnv_pbcq_spci_xscom_ops = { + .read = pnv_pbcq_spci_xscom_read, + .write = pnv_pbcq_spci_xscom_write, + .valid.min_access_size = 8, + .valid.max_access_size = 8, + .impl.min_access_size = 8, + .impl.max_access_size = 8, + .endianness = DEVICE_BIG_ENDIAN, +}; + +static void pnv_pbcq_default_bars(PnvPBCQState *pbcq) +{ + uint64_t mm0, mm1, reg; + PnvPHB3 *phb = pbcq->phb; + + mm0 = 0x3d00000000000ull + 0x4000000000ull * phb->chip_id + + 0x1000000000ull * phb->phb_id; + mm1 = 0x3ff8000000000ull + 0x0200000000ull * phb->chip_id + + 0x0080000000ull * phb->phb_id; + reg = 0x3fffe40000000ull + 0x0000400000ull * phb->chip_id + + 0x0000100000ull * phb->phb_id; + + pbcq->nest_regs[PBCQ_NEST_MMIO_BAR0] = mm0 << 14; + pbcq->nest_regs[PBCQ_NEST_MMIO_BAR1] = mm1 << 14; + pbcq->nest_regs[PBCQ_NEST_PHB_BAR] = reg << 14; + pbcq->nest_regs[PBCQ_NEST_MMIO_MASK0] = 0x3fff000000000ull << 14; + pbcq->nest_regs[PBCQ_NEST_MMIO_MASK1] = 0x3ffff80000000ull << 14; + pbcq->pci_regs[PBCQ_PCI_BAR2] = reg << 14; +} + +static void pnv_pbcq_realize(DeviceState *dev, Error **errp) +{ + PnvPBCQState *pbcq = PNV_PBCQ(dev); + PnvPHB3 *phb; + Object *obj; + Error *local_err = NULL; + char name[32]; + + obj = object_property_get_link(OBJECT(dev), "phb", &local_err); + if (!obj) { + error_propagate(errp, local_err); + error_prepend(errp, "required link 'phb' not found: "); + return; + } + phb = pbcq->phb = PNV_PHB3(obj); + + /* XXX Fix OPAL to do that: establish default BAR values */ + pnv_pbcq_default_bars(pbcq); + + /* Initialize the XSCOM region for the PBCQ registers */ + snprintf(name, sizeof(name), "xscom-pbcq-nest-%d.%d", + phb->chip_id, phb->phb_id); + pnv_xscom_region_init(&pbcq->xscom_nest_regs, OBJECT(dev), + &pnv_pbcq_nest_xscom_ops, pbcq, name, + PNV_XSCOM_PBCQ_NEST_SIZE); + snprintf(name, sizeof(name), "xscom-pbcq-pci-%d.%d", + phb->chip_id, phb->phb_id); + pnv_xscom_region_init(&pbcq->xscom_pci_regs, OBJECT(dev), + &pnv_pbcq_pci_xscom_ops, pbcq, name, + PNV_XSCOM_PBCQ_PCI_SIZE); + snprintf(name, sizeof(name), "xscom-pbcq-spci-%d.%d", + phb->chip_id, phb->phb_id); + pnv_xscom_region_init(&pbcq->xscom_spci_regs, OBJECT(dev), + &pnv_pbcq_spci_xscom_ops, pbcq, name, + PNV_XSCOM_PBCQ_SPCI_SIZE); +} + +static int pnv_pbcq_dt_xscom(PnvXScomInterface *dev, void *fdt, + int xscom_offset) +{ + const char compat[] = "ibm,power8-pbcq"; + PnvPHB3 *phb = PNV_PBCQ(dev)->phb; + char *name; + int offset; + uint32_t lpc_pcba = PNV_XSCOM_PBCQ_NEST_BASE + 0x400 * phb->phb_id; + uint32_t reg[] = { + cpu_to_be32(lpc_pcba), + cpu_to_be32(PNV_XSCOM_PBCQ_NEST_SIZE), + cpu_to_be32(PNV_XSCOM_PBCQ_PCI_BASE + 0x400 * phb->phb_id), + cpu_to_be32(PNV_XSCOM_PBCQ_PCI_SIZE), + cpu_to_be32(PNV_XSCOM_PBCQ_SPCI_BASE + 0x040 * phb->phb_id), + cpu_to_be32(PNV_XSCOM_PBCQ_SPCI_SIZE) + }; + + name = g_strdup_printf("pbcq@%x", lpc_pcba); + offset = fdt_add_subnode(fdt, xscom_offset, name); + _FDT(offset); + g_free(name); + + _FDT((fdt_setprop(fdt, offset, "reg", reg, sizeof(reg)))); + + _FDT((fdt_setprop_cell(fdt, offset, "ibm,phb-index", phb->phb_id))); + _FDT((fdt_setprop_cell(fdt, offset, "ibm,chip-id", phb->chip_id))); + _FDT((fdt_setprop(fdt, offset, "compatible", compat, + sizeof(compat)))); + return 0; +} + +static void pnv_pbcq_class_init(ObjectClass *klass, void *data) +{ + DeviceClass *dc = DEVICE_CLASS(klass); + PnvXScomInterfaceClass *xdc = PNV_XSCOM_INTERFACE_CLASS(klass); + + xdc->dt_xscom = pnv_pbcq_dt_xscom; + + dc->realize = pnv_pbcq_realize; +} + +static const TypeInfo pnv_pbcq_type_info = { + .name = TYPE_PNV_PBCQ, + .parent = TYPE_DEVICE, + .instance_size = sizeof(PnvPBCQState), + .class_init = pnv_pbcq_class_init, + .interfaces = (InterfaceInfo[]) { + { TYPE_PNV_XSCOM_INTERFACE }, + { } + } +}; + +static void pnv_pbcq_register_types(void) +{ + type_register_static(&pnv_pbcq_type_info); +} + +type_init(pnv_pbcq_register_types) diff --git a/hw/ppc/pnv.c b/hw/ppc/pnv.c index c8014d081a40..0dcc91ac101b 100644 --- a/hw/ppc/pnv.c +++ b/hw/ppc/pnv.c @@ -50,6 +50,7 @@ #include "hw/boards.h" #include "hw/char/serial.h" #include "hw/timer/mc146818rtc.h" +#include "hw/pci/msi.h" #include <libfdt.h> @@ -604,8 +605,13 @@ static ISABus *pnv_isa_create(PnvChip *chip, Error **errp) static void pnv_chip_power8_pic_print_info(PnvChip *chip, Monitor *mon) { Pnv8Chip *chip8 = PNV8_CHIP(chip); + int i; ics_pic_print_info(&chip8->psi.ics, mon); + for (i = 0; i < chip->num_phbs; i++) { + ics_pic_print_info(ICS_BASE(&chip8->phbs[i].msis), mon); + ics_pic_print_info(&chip8->phbs[i].lsis, mon); + } } static void pnv_chip_power9_pic_print_info(PnvChip *chip, Monitor *mon) @@ -639,6 +645,9 @@ static void pnv_init(MachineState *machine) DeviceState *dev; bool load_skiboot_from_pnor = false; + /* MSIs are supported on this platform */ + msi_nonbroken = true; + /* allocate RAM */ if (machine->ram_size < (1 * GiB)) { warn_report("skiboot may not work with < 1GB of RAM"); @@ -870,7 +879,10 @@ static void pnv_chip_power9_intc_create(PnvChip *chip, PowerPCCPU *cpu, static void pnv_chip_power8_instance_init(Object *obj) { + PnvChip *chip = PNV_CHIP(obj); Pnv8Chip *chip8 = PNV8_CHIP(obj); + PnvChipClass *pcc = PNV_CHIP_GET_CLASS(obj); + int i; object_initialize_child(obj, "psi", &chip8->psi, sizeof(chip8->psi), TYPE_PNV8_PSI, &error_abort, NULL); @@ -891,6 +903,17 @@ static void pnv_chip_power8_instance_init(Object *obj) TYPE_PNV8_HOMER, &error_abort, NULL); object_property_add_const_link(OBJECT(&chip8->homer), "chip", obj, &error_abort); + + for (i = 0; i < pcc->num_phbs; i++) { + object_initialize_child(obj, "phb[*]", &chip8->phbs[i], + sizeof(chip8->phbs[i]), TYPE_PNV_PHB3, + &error_abort, NULL); + } + + /* + * Number of PHBs is the chip default + */ + chip->num_phbs = pcc->num_phbs; } static void pnv_chip_icp_realize(Pnv8Chip *chip8, Error **errp) @@ -932,6 +955,7 @@ static void pnv_chip_power8_realize(DeviceState *dev, Error **errp) Pnv8Chip *chip8 = PNV8_CHIP(dev); Pnv8Psi *psi8 = &chip8->psi; Error *local_err = NULL; + int i; /* XSCOM bridge is first */ pnv_xscom_realize(chip, PNV_XSCOM_SIZE, &local_err); @@ -998,6 +1022,33 @@ static void pnv_chip_power8_realize(DeviceState *dev, Error **errp) } memory_region_add_subregion(get_system_memory(), PNV_HOMER_BASE(chip), &chip8->homer.regs); + + /* PHB3 controllers */ + for (i = 0; i < chip->num_phbs; i++) { + PnvPHB3 *phb = &chip8->phbs[i]; + PnvPBCQState *pbcq = &phb->pbcq; + + object_property_set_int(OBJECT(phb), i, "index", &error_fatal); + object_property_set_int(OBJECT(phb), chip->chip_id, "chip-id", + &error_fatal); + object_property_set_bool(OBJECT(phb), true, "realized", &local_err); + if (local_err) { + error_propagate(errp, local_err); + return; + } + qdev_set_parent_bus(DEVICE(phb), sysbus_get_default()); + + /* Populate the XSCOM address space. */ + pnv_xscom_add_subregion(chip, + PNV_XSCOM_PBCQ_NEST_BASE + 0x400 * phb->phb_id, + &pbcq->xscom_nest_regs); + pnv_xscom_add_subregion(chip, + PNV_XSCOM_PBCQ_PCI_BASE + 0x400 * phb->phb_id, + &pbcq->xscom_pci_regs); + pnv_xscom_add_subregion(chip, + PNV_XSCOM_PBCQ_SPCI_BASE + 0x040 * phb->phb_id, + &pbcq->xscom_spci_regs); + } } static void pnv_chip_power8e_class_init(ObjectClass *klass, void *data) @@ -1008,6 +1059,7 @@ static void pnv_chip_power8e_class_init(ObjectClass *klass, void *data) k->chip_type = PNV_CHIP_POWER8E; k->chip_cfam_id = 0x221ef04980000000ull; /* P8 Murano DD2.1 */ k->cores_mask = POWER8E_CORE_MASK; + k->num_phbs = 3; k->core_pir = pnv_chip_core_pir_p8; k->intc_create = pnv_chip_power8_intc_create; k->isa_create = pnv_chip_power8_isa_create; @@ -1027,6 +1079,7 @@ static void pnv_chip_power8_class_init(ObjectClass *klass, void *data) k->chip_type = PNV_CHIP_POWER8; k->chip_cfam_id = 0x220ea04980000000ull; /* P8 Venice DD2.0 */ k->cores_mask = POWER8_CORE_MASK; + k->num_phbs = 3; k->core_pir = pnv_chip_core_pir_p8; k->intc_create = pnv_chip_power8_intc_create; k->isa_create = pnv_chip_power8_isa_create; @@ -1046,6 +1099,7 @@ static void pnv_chip_power8nvl_class_init(ObjectClass *klass, void *data) k->chip_type = PNV_CHIP_POWER8NVL; k->chip_cfam_id = 0x120d304980000000ull; /* P8 Naples DD1.0 */ k->cores_mask = POWER8_CORE_MASK; + k->num_phbs = 3; k->core_pir = pnv_chip_core_pir_p8; k->intc_create = pnv_chip_power8_intc_create; k->isa_create = pnv_chip_power8nvl_isa_create; @@ -1215,6 +1269,7 @@ static void pnv_chip_power9_class_init(ObjectClass *klass, void *data) k->chip_type = PNV_CHIP_POWER9; k->chip_cfam_id = 0x220d104900008000ull; /* P9 Nimbus DD2.0 */ k->cores_mask = POWER9_CORE_MASK; + k->num_phbs = 0; k->core_pir = pnv_chip_core_pir_p9; k->intc_create = pnv_chip_power9_intc_create; k->isa_create = pnv_chip_power9_isa_create; @@ -1336,6 +1391,7 @@ static Property pnv_chip_properties[] = { DEFINE_PROP_UINT64("ram-size", PnvChip, ram_size, 0), DEFINE_PROP_UINT32("nr-cores", PnvChip, nr_cores, 1), DEFINE_PROP_UINT64("cores-mask", PnvChip, cores_mask, 0x0), + DEFINE_PROP_UINT32("num-phbs", PnvChip, num_phbs, 0), DEFINE_PROP_END_OF_LIST(), }; @@ -1352,14 +1408,23 @@ static void pnv_chip_class_init(ObjectClass *klass, void *data) static ICSState *pnv_ics_get(XICSFabric *xi, int irq) { PnvMachineState *pnv = PNV_MACHINE(xi); - int i; + int i, j; for (i = 0; i < pnv->num_chips; i++) { + PnvChip *chip = pnv->chips[i]; Pnv8Chip *chip8 = PNV8_CHIP(pnv->chips[i]); if (ics_valid_irq(&chip8->psi.ics, irq)) { return &chip8->psi.ics; } + for (j = 0; j < chip->num_phbs; j++) { + if (ics_valid_irq(&chip8->phbs[j].lsis, irq)) { + return &chip8->phbs[j].lsis; + } + if (ics_valid_irq(ICS_BASE(&chip8->phbs[j].msis), irq)) { + return ICS_BASE(&chip8->phbs[j].msis); + } + } } return NULL; } @@ -1367,11 +1432,17 @@ static ICSState *pnv_ics_get(XICSFabric *xi, int irq) static void pnv_ics_resend(XICSFabric *xi) { PnvMachineState *pnv = PNV_MACHINE(xi); - int i; + int i, j; for (i = 0; i < pnv->num_chips; i++) { + PnvChip *chip = pnv->chips[i]; Pnv8Chip *chip8 = PNV8_CHIP(pnv->chips[i]); + ics_resend(&chip8->psi.ics); + for (j = 0; j < chip->num_phbs; j++) { + ics_resend(&chip8->phbs[j].lsis); + ics_resend(ICS_BASE(&chip8->phbs[j].msis)); + } } } diff --git a/hw/ppc/pnv_xscom.c b/hw/ppc/pnv_xscom.c index f01d788a6545..4d769371c0b5 100644 --- a/hw/ppc/pnv_xscom.c +++ b/hw/ppc/pnv_xscom.c @@ -298,7 +298,10 @@ static int xscom_dt_child(Object *child, void *opaque) PnvXScomInterface *xd = PNV_XSCOM_INTERFACE(child); PnvXScomInterfaceClass *xc = PNV_XSCOM_INTERFACE_GET_CLASS(xd); - if (xc->dt_xscom) { + /* + * Only "realized" devices should be configured in the DT + */ + if (xc->dt_xscom && DEVICE(child)->realized) { _FDT((xc->dt_xscom(xd, args->fdt, args->xscom_offset))); } } @@ -345,7 +348,11 @@ int pnv_dt_xscom(PnvChip *chip, void *fdt, int root_offset) args.fdt = fdt; args.xscom_offset = xscom_offset; - object_child_foreach(OBJECT(chip), xscom_dt_child, &args); + /* Some PnvXScomInterface objects lie a bit deeper (PnvPBCQState) + * than the first layer, so we need to loop on the whole object + * hierarchy to catch them + */ + object_child_foreach_recursive(OBJECT(chip), xscom_dt_child, &args); return 0; } diff --git a/hw/pci-host/Makefile.objs b/hw/pci-host/Makefile.objs index a9cd3e022d53..4c6f825a8f06 100644 --- a/hw/pci-host/Makefile.objs +++ b/hw/pci-host/Makefile.objs @@ -19,3 +19,4 @@ common-obj-$(CONFIG_PCI_EXPRESS_GENERIC_BRIDGE) += gpex.o common-obj-$(CONFIG_PCI_EXPRESS_XILINX) += xilinx-pcie.o common-obj-$(CONFIG_PCI_EXPRESS_DESIGNWARE) += designware.o +obj-$(CONFIG_POWERNV) += pnv_phb3.o pnv_phb3_pbcq.o pnv_phb3_msi.o diff --git a/hw/ppc/Kconfig b/hw/ppc/Kconfig index f927ec9c7487..3c2aa5b02a45 100644 --- a/hw/ppc/Kconfig +++ b/hw/ppc/Kconfig @@ -26,6 +26,8 @@ config POWERNV select MC146818RTC select XICS select XIVE + select PCI + select PCIE_PORT config PPC405 bool -- 2.21.0 ^ permalink raw reply related [flat|nested] 30+ messages in thread
* Re: [PATCH 0/4] xics: Eliminate unnecessary class 2019-09-24 11:00 ` Cédric Le Goater @ 2019-09-26 1:28 ` David Gibson 0 siblings, 0 replies; 30+ messages in thread From: David Gibson @ 2019-09-26 1:28 UTC (permalink / raw) To: Cédric Le Goater Cc: qemu-ppc, gkurz, Philippe Mathieu-Daudé, Greg Kurz, qemu-devel [-- Attachment #1: Type: text/plain, Size: 108173 bytes --] On Tue, Sep 24, 2019 at 01:00:00PM +0200, Cédric Le Goater wrote: > On 24/09/2019 12:04, Philippe Mathieu-Daudé wrote: > > Hi Cédric, > > > > On 9/24/19 11:55 AM, Cédric Le Goater wrote: > >> On 24/09/2019 09:52, Greg Kurz wrote: > >>> On Tue, 24 Sep 2019 07:22:51 +0200 > >>> Cédric Le Goater <clg@kaod.org> wrote: > >>> > >>>> On 24/09/2019 06:59, David Gibson wrote: > >>>>> The XICS interrupt controller device used to have separate subtypes > >>>>> for the KVM and non-KVM variant of the device. That was a bad idea, > >>>>> because it leaked information that should be entirely host-side > >>>>> implementation specific to the kinda-sorta guest visible QOM class > >>>>> names. > >>>>> > >>>>> We eliminated the KVM specific class some time ago, but it's left > >>>>> behind a distinction between the TYPE_ICS_BASE abstract class and > >>>>> TYPE_ICS_SIMPLE subtype which no longer serves any purpose. > >>>>> > >>>>> This series collapses the two types back into one. > >>>> > >>>> OK. Is it migration compatible ? because this is why we have kept > >>> > >>> Yeah, the types themselves don't matter, only the format of the > >>> binary stream described in the VMStateDescription does. > >>> > >>>> this subclass. I am glad we can remove it finally. > >>>> > >>> > >>> I was thinking we were keeping that for pnv... > >> > >> > >> Yes, also. See the resend and reject handler in the code > >> below. > >> > >> > >> I have been maintaining this patch since QEMU 2.6. I think > >> it is time for me to declare defeat on getting it merged > >> and QEMU 4.2 will be the last rebase. > > > > Do you remember what is missing for being merged upstream? > > lack of interest I would say. Here is the patch. So, I had a look at what this is doing with ICS. AFAIT, what it really wants is something *more* abstract that ICS_BASE - the main reason for its own resend and reject hooks AFAICT is because it uses a different representation of the irq state - the RBA registers rather than the full set of state bits per irq that are in ICSState. I'm also not entirely convinced the current PHB3 draft patch properly keeps the stuff in ICSState properly synchronized with its own representation. I can see two ways we could handle that 1) just use ICS_SIMPLE and convert to the RBA representation at MMIO time, simple but slow (but maybe not enough that we'd care). 2) Have a different ics base class (let's call it ICS_ABSTRACT) that has the reject/resend/eoi hooks but whose instance has no state at all. In any case (2) is different enough from what we have now that I now tend to agree that the way forward is to collapse the types together as I've proposed, and if we resurrect the PHB3 patch we can reimplement it in the way it actually needs. > > C. > > >From b254a0279f5cf13a5db2db0bb209fe806af59ad0 Mon Sep 17 00:00:00 2001 > From: =?UTF-8?q?C=C3=A9dric=20Le=20Goater?= <clg@kaod.org> > Date: Wed, 18 Sep 2019 18:02:03 +0200 > Subject: [PATCH] ppc/pnv: Add model for Power8 PHB3 PCIe Host bridge > MIME-Version: 1.0 > Content-Type: text/plain; charset=UTF-8 > Content-Transfer-Encoding: 8bit > > This is a model of the PCIe Host Bridge (PHB3) found on a Power8 > processor. It includes the PowerBus logic interface (PBCQ), IOMMU > support, a single PCIe Gen.3 Root Complex, and support for MSI and LSI > interrupt sources as found on a Power8 system using the XICS interrupt > controller. > > The Power8 processor comes in different flavors: Venice, Murano, > Naple, each having a different number of PHBs. To make things simpler, > the PHB3 model provides 3 per chip. Some platforms, like the > Firestone, can also couple PHBs on the first chip to provide more > bandwidth but this is too specific to model in QEMU. > > No default device layout is provided and PCI devices can be added on > any of the available PCIe Root Port (pcie.0 .. 2 of a Power8 chip) > with address 0x0 as the firwware (skiboot) only accepts a single > device per root port. To run a simple system with a network and a > storage adapters, use a command line options such as : > > -device e1000e,netdev=net0,mac=C0:FF:EE:00:00:02,bus=pcie.0,addr=0x0 > -netdev bridge,id=net0,helper=/usr/libexec/qemu-bridge-helper,br=virbr0,id=hostnet0 > > -device megasas,id=scsi0,bus=pcie.1,addr=0x0 > -drive file=$disk,if=none,id=drive-scsi0-0-0-0,format=qcow2,cache=none > -device scsi-hd,bus=scsi0.0,channel=0,scsi-id=0,lun=0,drive=drive-scsi0-0-0-0,id=scsi0-0-0-0,bootindex=2 > > If more are needed, include a bridge. > > Multi chip is supported, each chip adding its set of PHB3 controllers > and its PCI busses. The model doesn't emulate the EEH error handling > (and may never do). > > Originally from Benjamin Herrenschmidt. I am taking ownership since I > have been maintaining the model since QEMU 2.6. > > Signed-off-by: Cédric Le Goater <clg@kaod.org> > --- > include/hw/pci-host/pnv_phb3.h | 173 ++++ > include/hw/pci-host/pnv_phb3_regs.h | 454 ++++++++++ > include/hw/ppc/pnv.h | 7 + > include/hw/ppc/pnv_xscom.h | 9 + > include/hw/ppc/xics.h | 1 + > hw/intc/xics.c | 2 +- > hw/pci-host/pnv_phb3.c | 1191 +++++++++++++++++++++++++++ > hw/pci-host/pnv_phb3_msi.c | 320 +++++++ > hw/pci-host/pnv_phb3_pbcq.c | 347 ++++++++ > hw/ppc/pnv.c | 75 +- > hw/ppc/pnv_xscom.c | 11 +- > hw/pci-host/Makefile.objs | 1 + > hw/ppc/Kconfig | 2 + > 13 files changed, 2588 insertions(+), 5 deletions(-) > create mode 100644 include/hw/pci-host/pnv_phb3.h > create mode 100644 include/hw/pci-host/pnv_phb3_regs.h > create mode 100644 hw/pci-host/pnv_phb3.c > create mode 100644 hw/pci-host/pnv_phb3_msi.c > create mode 100644 hw/pci-host/pnv_phb3_pbcq.c > > diff --git a/include/hw/pci-host/pnv_phb3.h b/include/hw/pci-host/pnv_phb3.h > new file mode 100644 > index 000000000000..b282730a5b40 > --- /dev/null > +++ b/include/hw/pci-host/pnv_phb3.h > @@ -0,0 +1,173 @@ > +/* > + * QEMU PowerPC PowerNV PHB3 model > + * > + * Copyright (c) 2014-2018, IBM Corporation. > + * > + * This code is licensed under the GPL version 2 or later. See the > + * COPYING file in the top-level directory. > + */ > + > +#ifndef PCI_HOST_PNV_PHB3_H > +#define PCI_HOST_PNV_PHB3_H > + > +#include "hw/pci/pcie_host.h" > +#include "hw/pci/pcie_port.h" > +#include "hw/ppc/xics.h" > + > +typedef struct PnvPHB3 PnvPHB3; > +typedef struct PnvChip PnvChip; > + > +/* > + * PHB3 XICS Source for MSIs > + */ > +#define TYPE_PHB3_MSI "phb3-msi" > +#define PHB3_MSI(obj) OBJECT_CHECK(Phb3MsiState, (obj), TYPE_PHB3_MSI) > + > +#define PHB3_MAX_MSI 2048 > + > +typedef struct Phb3MsiState { > + ICSState ics; > + qemu_irq *qirqs; > + > + PnvPHB3 *phb; > + uint64_t rba[PHB3_MAX_MSI / 64]; > + uint32_t rba_sum; > +} Phb3MsiState; > + > +void pnv_phb3_msi_update_config(Phb3MsiState *msis, uint32_t base, > + uint32_t count); > +void pnv_phb3_msi_send(Phb3MsiState *msis, uint64_t addr, uint16_t data, > + int32_t dev_pe); > +void pnv_phb3_msi_ffi(Phb3MsiState *msis, uint64_t val); > + > + > +/* We have one such address space wrapper per possible device > + * under the PHB since they need to be assigned statically at > + * qemu device creation time. The relationship to a PE is done > + * later dynamically. This means we can potentially create a lot > + * of these guys. Q35 stores them as some kind of radix tree but > + * we never really need to do fast lookups so instead we simply > + * keep a QLIST of them for now, we can add the radix if needed > + * later on. > + * > + * We do cache the PE number to speed things up a bit though. > + */ > +typedef struct PnvPhb3DMASpace { > + PCIBus *bus; > + uint8_t devfn; > + int pe_num; /* Cached PE number */ > +#define PHB_INVALID_PE (-1) > + PnvPHB3 *phb; > + AddressSpace dma_as; > + IOMMUMemoryRegion dma_mr; > + MemoryRegion msi32_mr; > + MemoryRegion msi64_mr; > + bool msi32_mapped; > + bool msi64_mapped; > + QLIST_ENTRY(PnvPhb3DMASpace) list; > +} PnvPhb3DMASpace; > + > +/* > + * PHB3 Power Bus Common Queue > + */ > +#define TYPE_PNV_PBCQ "pnv-pbcq" > +#define PNV_PBCQ(obj) OBJECT_CHECK(PnvPBCQState, (obj), TYPE_PNV_PBCQ) > + > +typedef struct PnvPBCQState { > + DeviceState parent; > + > + uint32_t nest_xbase; > + uint32_t spci_xbase; > + uint32_t pci_xbase; > +#define PBCQ_NEST_REGS_COUNT 0x46 > +#define PBCQ_PCI_REGS_COUNT 0x15 > +#define PBCQ_SPCI_REGS_COUNT 0x5 > + > + uint64_t nest_regs[PBCQ_NEST_REGS_COUNT]; > + uint64_t spci_regs[PBCQ_SPCI_REGS_COUNT]; > + uint64_t pci_regs[PBCQ_PCI_REGS_COUNT]; > + MemoryRegion mmbar0; > + MemoryRegion mmbar1; > + MemoryRegion phbbar; > + bool mmio0_mapped; > + bool mmio1_mapped; > + bool phb_mapped; > + uint64_t mmio0_base; > + uint64_t mmio0_size; > + uint64_t mmio1_base; > + uint64_t mmio1_size; > + PnvPHB3 *phb; > + > + MemoryRegion xscom_nest_regs; > + MemoryRegion xscom_pci_regs; > + MemoryRegion xscom_spci_regs; > +} PnvPBCQState; > + > +/* > + * PHB3 PCIe Root port > + */ > +#define TYPE_PNV_PHB3_ROOT_BUS "pnv-phb3-root-bus" > + > +#define TYPE_PNV_PHB3_ROOT_PORT "pnv-phb3-root-port" > + > +typedef struct PnvPHB3RootPort { > + PCIESlot parent_obj; > +} PnvPHB3RootPort; > + > +/* > + * PHB3 PCIe Host Bridge for PowerNV machines (POWER8) > + */ > +#define TYPE_PNV_PHB3 "pnv-phb3" > +#define PNV_PHB3(obj) OBJECT_CHECK(PnvPHB3, (obj), TYPE_PNV_PHB3) > + > +#define PNV_PHB3_NUM_M64 16 > +#define PNV_PHB3_NUM_REGS (0x1000 >> 3) > +#define PNV_PHB3_NUM_LSI 8 > +#define PNV_PHB3_NUM_PE 256 > + > +#define PCI_MMIO_TOTAL_SIZE (0x1ull << 60) > + > +struct PnvPHB3 { > + PCIExpressHost parent_obj; > + > + PnvPHB3RootPort root; > + > + uint32_t chip_id; > + uint32_t phb_id; > + char bus_path[8]; > + > + uint64_t regs[PNV_PHB3_NUM_REGS]; > + MemoryRegion mr_regs; > + > + MemoryRegion mr_m32; > + MemoryRegion mr_m64[PNV_PHB3_NUM_M64]; > + bool regs_mapped; > + bool m32_mapped; > + bool m64_mapped[PNV_PHB3_NUM_M64]; > + MemoryRegion pci_mmio; > + MemoryRegion pci_io; > + > + uint64_t ioda_LIST[8]; > + uint64_t ioda_LXIVT[8]; > + uint64_t ioda_TVT[512]; > + uint64_t ioda_M64BT[16]; > + uint64_t ioda_MDT[256]; > + uint64_t ioda_PEEV[4]; > + > + uint32_t total_irq; > + ICSState lsis; > + qemu_irq *qirqs; > + Phb3MsiState msis; > + > + PnvPBCQState pbcq; > + > + QLIST_HEAD(, PnvPhb3DMASpace) dma_spaces; > +}; > + > +uint64_t pnv_phb3_reg_read(void *opaque, hwaddr off, unsigned size); > +void pnv_phb3_reg_write(void *opaque, hwaddr off, uint64_t val, unsigned size); > +void pnv_phb3_update_regions(PnvPHB3 *phb); > +void pnv_phb3_remap_irqs(PnvPHB3 *phb); > + > + > +#endif /* PCI_HOST_PNV_PHB3_H */ > diff --git a/include/hw/pci-host/pnv_phb3_regs.h b/include/hw/pci-host/pnv_phb3_regs.h > new file mode 100644 > index 000000000000..8df83c321e86 > --- /dev/null > +++ b/include/hw/pci-host/pnv_phb3_regs.h > @@ -0,0 +1,454 @@ > +/* Copyright (c) 2013-2018, IBM Corporation. > + * > + * Licensed under the Apache License, Version 2.0 (the "License"); > + * you may not use this file except in compliance with the License. > + * You may obtain a copy of the License at > + * > + * http://www.apache.org/licenses/LICENSE-2.0 > + * > + * Unless required by applicable law or agreed to in writing, software > + * distributed under the License is distributed on an "AS IS" BASIS, > + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or > + * implied. > + * > + * See the License for the specific language governing permissions and > + * limitations under the License. > + */ > + > +#ifndef PCI_HOST_PNV_PHB3_REGS_H > +#define PCI_HOST_PNV_PHB3_REGS_H > + > +#include "target/ppc/cpu.h" > + > +/* TODO: fix GETFIELD/SETFIELD macro definitions */ > +#if HOST_LONG_BITS == 32 > +# define MASK_TO_LSH(m) (__builtin_ffsll(m) - 1) > +#elif HOST_LONG_BITS == 64 > +# define MASK_TO_LSH(m) (__builtin_ffsl(m) - 1) > +#else > +# error Unknown sizeof long > +#endif > + > +#define GETFIELD(m, v) (((v) & (m)) >> MASK_TO_LSH(m)) > +#define SETFIELD(m, v, val) \ > + (((v) & ~(m)) | ((((typeof(v))(val)) << MASK_TO_LSH(m)) & (m))) > + > +/* > + * PBCQ XSCOM registers > + */ > + > +#define PBCQ_NEST_IRSN_COMPARE 0x1a > +#define PBCQ_NEST_IRSN_COMP PPC_BITMASK(0, 18) > +#define PBCQ_NEST_IRSN_MASK 0x1b > +#define PBCQ_NEST_LSI_SRC_ID 0x1f > +#define PBCQ_NEST_LSI_SRC PPC_BITMASK(0, 7) > +#define PBCQ_NEST_REGS_COUNT 0x46 > +#define PBCQ_NEST_MMIO_BAR0 0x40 > +#define PBCQ_NEST_MMIO_BAR1 0x41 > +#define PBCQ_NEST_PHB_BAR 0x42 > +#define PBCQ_NEST_MMIO_MASK0 0x43 > +#define PBCQ_NEST_MMIO_MASK1 0x44 > +#define PBCQ_NEST_BAR_EN 0x45 > +#define PBCQ_NEST_BAR_EN_MMIO0 PPC_BIT(0) > +#define PBCQ_NEST_BAR_EN_MMIO1 PPC_BIT(1) > +#define PBCQ_NEST_BAR_EN_PHB PPC_BIT(2) > +#define PBCQ_NEST_BAR_EN_IRSN_RX PPC_BIT(3) > +#define PBCQ_NEST_BAR_EN_IRSN_TX PPC_BIT(4) > + > +#define PBCQ_PCI_REGS_COUNT 0x15 > +#define PBCQ_PCI_BAR2 0x0b > + > +#define PBCQ_SPCI_REGS_COUNT 0x5 > +#define PBCQ_SPCI_ASB_ADDR 0x0 > +#define PBCQ_SPCI_ASB_STATUS 0x1 > +#define PBCQ_SPCI_ASB_DATA 0x2 > +#define PBCQ_SPCI_AIB_CAPP_EN 0x3 > +#define PBCQ_SPCI_CAPP_SEC_TMR 0x4 > + > +/* > + * PHB MMIO registers > + */ > + > +/* PHB Fundamental register set A */ > +#define PHB_LSI_SOURCE_ID 0x100 > +#define PHB_LSI_SRC_ID PPC_BITMASK(5, 12) > +#define PHB_DMA_CHAN_STATUS 0x110 > +#define PHB_DMA_CHAN_ANY_ERR PPC_BIT(27) > +#define PHB_DMA_CHAN_ANY_ERR1 PPC_BIT(28) > +#define PHB_DMA_CHAN_ANY_FREEZE PPC_BIT(29) > +#define PHB_CPU_LOADSTORE_STATUS 0x120 > +#define PHB_CPU_LS_ANY_ERR PPC_BIT(27) > +#define PHB_CPU_LS_ANY_ERR1 PPC_BIT(28) > +#define PHB_CPU_LS_ANY_FREEZE PPC_BIT(29) > +#define PHB_DMA_MSI_NODE_ID 0x128 > +#define PHB_DMAMSI_NID_FIXED PPC_BIT(0) > +#define PHB_DMAMSI_NID PPC_BITMASK(24, 31) > +#define PHB_CONFIG_DATA 0x130 > +#define PHB_LOCK0 0x138 > +#define PHB_CONFIG_ADDRESS 0x140 > +#define PHB_CA_ENABLE PPC_BIT(0) > +#define PHB_CA_BUS PPC_BITMASK(4, 11) > +#define PHB_CA_DEV PPC_BITMASK(12, 16) > +#define PHB_CA_FUNC PPC_BITMASK(17, 19) > +#define PHB_CA_REG PPC_BITMASK(20, 31) > +#define PHB_CA_PE PPC_BITMASK(40, 47) > +#define PHB_LOCK1 0x148 > +#define PHB_IVT_BAR 0x150 > +#define PHB_IVT_BAR_ENABLE PPC_BIT(0) > +#define PHB_IVT_BASE_ADDRESS_MASK PPC_BITMASK(14, 48) > +#define PHB_IVT_LENGTH_MASK PPC_BITMASK(52, 63) > +#define PHB_RBA_BAR 0x158 > +#define PHB_RBA_BAR_ENABLE PPC_BIT(0) > +#define PHB_RBA_BASE_ADDRESS PPC_BITMASK(14, 55) > +#define PHB_PHB3_CONFIG 0x160 > +#define PHB_PHB3C_64B_TCE_EN PPC_BIT(2) > +#define PHB_PHB3C_32BIT_MSI_EN PPC_BIT(8) > +#define PHB_PHB3C_64BIT_MSI_EN PPC_BIT(14) > +#define PHB_PHB3C_M32_EN PPC_BIT(16) > +#define PHB_RTT_BAR 0x168 > +#define PHB_RTT_BAR_ENABLE PPC_BIT(0) > +#define PHB_RTT_BASE_ADDRESS_MASK PPC_BITMASK(14, 46) > +#define PHB_PELTV_BAR 0x188 > +#define PHB_PELTV_BAR_ENABLE PPC_BIT(0) > +#define PHB_PELTV_BASE_ADDRESS PPC_BITMASK(14, 50) > +#define PHB_M32_BASE_ADDR 0x190 > +#define PHB_M32_BASE_MASK 0x198 > +#define PHB_M32_START_ADDR 0x1a0 > +#define PHB_PEST_BAR 0x1a8 > +#define PHB_PEST_BAR_ENABLE PPC_BIT(0) > +#define PHB_PEST_BASE_ADDRESS PPC_BITMASK(14, 51) > +#define PHB_M64_UPPER_BITS 0x1f0 > +#define PHB_INTREP_TIMER 0x1f8 > +#define PHB_DMARD_SYNC 0x200 > +#define PHB_DMARD_SYNC_START PPC_BIT(0) > +#define PHB_DMARD_SYNC_COMPLETE PPC_BIT(1) > +#define PHB_RTC_INVALIDATE 0x208 > +#define PHB_RTC_INVALIDATE_ALL PPC_BIT(0) > +#define PHB_RTC_INVALIDATE_RID PPC_BITMASK(16, 31) > +#define PHB_TCE_KILL 0x210 > +#define PHB_TCE_KILL_ALL PPC_BIT(0) > +#define PHB_TCE_SPEC_CTL 0x218 > +#define PHB_IODA_ADDR 0x220 > +#define PHB_IODA_AD_AUTOINC PPC_BIT(0) > +#define PHB_IODA_AD_TSEL PPC_BITMASK(11, 15) > +#define PHB_IODA_AD_TADR PPC_BITMASK(55, 63) > +#define PHB_IODA_DATA0 0x228 > +#define PHB_FFI_REQUEST 0x238 > +#define PHB_FFI_LOCK_CLEAR PPC_BIT(3) > +#define PHB_FFI_REQUEST_ISN PPC_BITMASK(49, 59) > +#define PHB_FFI_LOCK 0x240 > +#define PHB_FFI_LOCK_STATE PPC_BIT(0) > +#define PHB_XIVE_UPDATE 0x248 /* Broken in DD1 */ > +#define PHB_PHB3_GEN_CAP 0x250 > +#define PHB_PHB3_TCE_CAP 0x258 > +#define PHB_PHB3_IRQ_CAP 0x260 > +#define PHB_PHB3_EEH_CAP 0x268 > +#define PHB_IVC_INVALIDATE 0x2a0 > +#define PHB_IVC_INVALIDATE_ALL PPC_BIT(0) > +#define PHB_IVC_INVALIDATE_SID PPC_BITMASK(16, 31) > +#define PHB_IVC_UPDATE 0x2a8 > +#define PHB_IVC_UPDATE_ENABLE_P PPC_BIT(0) > +#define PHB_IVC_UPDATE_ENABLE_Q PPC_BIT(1) > +#define PHB_IVC_UPDATE_ENABLE_SERVER PPC_BIT(2) > +#define PHB_IVC_UPDATE_ENABLE_PRI PPC_BIT(3) > +#define PHB_IVC_UPDATE_ENABLE_GEN PPC_BIT(4) > +#define PHB_IVC_UPDATE_ENABLE_CON PPC_BIT(5) > +#define PHB_IVC_UPDATE_GEN_MATCH PPC_BITMASK(6, 7) > +#define PHB_IVC_UPDATE_SERVER PPC_BITMASK(8, 23) > +#define PHB_IVC_UPDATE_PRI PPC_BITMASK(24, 31) > +#define PHB_IVC_UPDATE_GEN PPC_BITMASK(32, 33) > +#define PHB_IVC_UPDATE_P PPC_BITMASK(34, 34) > +#define PHB_IVC_UPDATE_Q PPC_BITMASK(35, 35) > +#define PHB_IVC_UPDATE_SID PPC_BITMASK(48, 63) > +#define PHB_PAPR_ERR_INJ_CTL 0x2b0 > +#define PHB_PAPR_ERR_INJ_CTL_INB PPC_BIT(0) > +#define PHB_PAPR_ERR_INJ_CTL_OUTB PPC_BIT(1) > +#define PHB_PAPR_ERR_INJ_CTL_STICKY PPC_BIT(2) > +#define PHB_PAPR_ERR_INJ_CTL_CFG PPC_BIT(3) > +#define PHB_PAPR_ERR_INJ_CTL_RD PPC_BIT(4) > +#define PHB_PAPR_ERR_INJ_CTL_WR PPC_BIT(5) > +#define PHB_PAPR_ERR_INJ_CTL_FREEZE PPC_BIT(6) > +#define PHB_PAPR_ERR_INJ_ADDR 0x2b8 > +#define PHB_PAPR_ERR_INJ_ADDR_MMIO PPC_BITMASK(16, 63) > +#define PHB_PAPR_ERR_INJ_MASK 0x2c0 > +#define PHB_PAPR_ERR_INJ_MASK_CFG PPC_BITMASK(4, 11) > +#define PHB_PAPR_ERR_INJ_MASK_MMIO PPC_BITMASK(16, 63) > +#define PHB_ETU_ERR_SUMMARY 0x2c8 > + > +/* UTL registers */ > +#define UTL_SYS_BUS_CONTROL 0x400 > +#define UTL_STATUS 0x408 > +#define UTL_SYS_BUS_AGENT_STATUS 0x410 > +#define UTL_SYS_BUS_AGENT_ERR_SEVERITY 0x418 > +#define UTL_SYS_BUS_AGENT_IRQ_EN 0x420 > +#define UTL_SYS_BUS_BURST_SZ_CONF 0x440 > +#define UTL_REVISION_ID 0x448 > +#define UTL_BCLK_DOMAIN_DBG1 0x460 > +#define UTL_BCLK_DOMAIN_DBG2 0x468 > +#define UTL_BCLK_DOMAIN_DBG3 0x470 > +#define UTL_BCLK_DOMAIN_DBG4 0x478 > +#define UTL_BCLK_DOMAIN_DBG5 0x480 > +#define UTL_BCLK_DOMAIN_DBG6 0x488 > +#define UTL_OUT_POST_HDR_BUF_ALLOC 0x4c0 > +#define UTL_OUT_POST_DAT_BUF_ALLOC 0x4d0 > +#define UTL_IN_POST_HDR_BUF_ALLOC 0x4e0 > +#define UTL_IN_POST_DAT_BUF_ALLOC 0x4f0 > +#define UTL_OUT_NP_BUF_ALLOC 0x500 > +#define UTL_IN_NP_BUF_ALLOC 0x510 > +#define UTL_PCIE_TAGS_ALLOC 0x520 > +#define UTL_GBIF_READ_TAGS_ALLOC 0x530 > +#define UTL_PCIE_PORT_CONTROL 0x540 > +#define UTL_PCIE_PORT_STATUS 0x548 > +#define UTL_PCIE_PORT_ERROR_SEV 0x550 > +#define UTL_PCIE_PORT_IRQ_EN 0x558 > +#define UTL_RC_STATUS 0x560 > +#define UTL_RC_ERR_SEVERITY 0x568 > +#define UTL_RC_IRQ_EN 0x570 > +#define UTL_EP_STATUS 0x578 > +#define UTL_EP_ERR_SEVERITY 0x580 > +#define UTL_EP_ERR_IRQ_EN 0x588 > +#define UTL_PCI_PM_CTRL1 0x590 > +#define UTL_PCI_PM_CTRL2 0x598 > +#define UTL_GP_CTL1 0x5a0 > +#define UTL_GP_CTL2 0x5a8 > +#define UTL_PCLK_DOMAIN_DBG1 0x5b0 > +#define UTL_PCLK_DOMAIN_DBG2 0x5b8 > +#define UTL_PCLK_DOMAIN_DBG3 0x5c0 > +#define UTL_PCLK_DOMAIN_DBG4 0x5c8 > + > +/* PCI-E Stack registers */ > +#define PHB_PCIE_SYSTEM_CONFIG 0x600 > +#define PHB_PCIE_BUS_NUMBER 0x608 > +#define PHB_PCIE_SYSTEM_TEST 0x618 > +#define PHB_PCIE_LINK_MANAGEMENT 0x630 > +#define PHB_PCIE_LM_LINK_ACTIVE PPC_BIT(8) > +#define PHB_PCIE_DLP_TRAIN_CTL 0x640 > +#define PHB_PCIE_DLP_TCTX_DISABLE PPC_BIT(1) > +#define PHB_PCIE_DLP_TCRX_DISABLED PPC_BIT(16) > +#define PHB_PCIE_DLP_INBAND_PRESENCE PPC_BIT(19) > +#define PHB_PCIE_DLP_TC_DL_LINKUP PPC_BIT(21) > +#define PHB_PCIE_DLP_TC_DL_PGRESET PPC_BIT(22) > +#define PHB_PCIE_DLP_TC_DL_LINKACT PPC_BIT(23) > +#define PHB_PCIE_SLOP_LOOPBACK_STATUS 0x648 > +#define PHB_PCIE_SYS_LINK_INIT 0x668 > +#define PHB_PCIE_UTL_CONFIG 0x670 > +#define PHB_PCIE_DLP_CONTROL 0x678 > +#define PHB_PCIE_UTL_ERRLOG1 0x680 > +#define PHB_PCIE_UTL_ERRLOG2 0x688 > +#define PHB_PCIE_UTL_ERRLOG3 0x690 > +#define PHB_PCIE_UTL_ERRLOG4 0x698 > +#define PHB_PCIE_DLP_ERRLOG1 0x6a0 > +#define PHB_PCIE_DLP_ERRLOG2 0x6a8 > +#define PHB_PCIE_DLP_ERR_STATUS 0x6b0 > +#define PHB_PCIE_DLP_ERR_COUNTERS 0x6b8 > +#define PHB_PCIE_UTL_ERR_INJECT 0x6c0 > +#define PHB_PCIE_TLDLP_ERR_INJECT 0x6c8 > +#define PHB_PCIE_LANE_EQ_CNTL0 0x6d0 > +#define PHB_PCIE_LANE_EQ_CNTL1 0x6d8 > +#define PHB_PCIE_LANE_EQ_CNTL2 0x6e0 > +#define PHB_PCIE_LANE_EQ_CNTL3 0x6e8 > +#define PHB_PCIE_STRAPPING 0x700 > + > +/* Fundamental register set B */ > +#define PHB_VERSION 0x800 > +#define PHB_RESET 0x808 > +#define PHB_CONTROL 0x810 > +#define PHB_CTRL_IVE_128_BYTES PPC_BIT(24) > +#define PHB_AIB_RX_CRED_INIT_TIMER 0x818 > +#define PHB_AIB_RX_CMD_CRED 0x820 > +#define PHB_AIB_RX_DATA_CRED 0x828 > +#define PHB_AIB_TX_CMD_CRED 0x830 > +#define PHB_AIB_TX_DATA_CRED 0x838 > +#define PHB_AIB_TX_CHAN_MAPPING 0x840 > +#define PHB_AIB_TAG_ENABLE 0x858 > +#define PHB_AIB_FENCE_CTRL 0x860 > +#define PHB_TCE_TAG_ENABLE 0x868 > +#define PHB_TCE_WATERMARK 0x870 > +#define PHB_TIMEOUT_CTRL1 0x878 > +#define PHB_TIMEOUT_CTRL2 0x880 > +#define PHB_Q_DMA_R 0x888 > +#define PHB_Q_DMA_R_QUIESCE_DMA PPC_BIT(0) > +#define PHB_Q_DMA_R_AUTORESET PPC_BIT(1) > +#define PHB_Q_DMA_R_DMA_RESP_STATUS PPC_BIT(4) > +#define PHB_Q_DMA_R_MMIO_RESP_STATUS PPC_BIT(5) > +#define PHB_Q_DMA_R_TCE_RESP_STATUS PPC_BIT(6) > +#define PHB_AIB_TAG_STATUS 0x900 > +#define PHB_TCE_TAG_STATUS 0x908 > + > +/* FIR & Error registers */ > +#define PHB_LEM_FIR_ACCUM 0xc00 > +#define PHB_LEM_FIR_AND_MASK 0xc08 > +#define PHB_LEM_FIR_OR_MASK 0xc10 > +#define PHB_LEM_ERROR_MASK 0xc18 > +#define PHB_LEM_ERROR_AND_MASK 0xc20 > +#define PHB_LEM_ERROR_OR_MASK 0xc28 > +#define PHB_LEM_ACTION0 0xc30 > +#define PHB_LEM_ACTION1 0xc38 > +#define PHB_LEM_WOF 0xc40 > +#define PHB_ERR_STATUS 0xc80 > +#define PHB_ERR1_STATUS 0xc88 > +#define PHB_ERR_INJECT 0xc90 > +#define PHB_ERR_LEM_ENABLE 0xc98 > +#define PHB_ERR_IRQ_ENABLE 0xca0 > +#define PHB_ERR_FREEZE_ENABLE 0xca8 > +#define PHB_ERR_AIB_FENCE_ENABLE 0xcb0 > +#define PHB_ERR_LOG_0 0xcc0 > +#define PHB_ERR_LOG_1 0xcc8 > +#define PHB_ERR_STATUS_MASK 0xcd0 > +#define PHB_ERR1_STATUS_MASK 0xcd8 > + > +#define PHB_OUT_ERR_STATUS 0xd00 > +#define PHB_OUT_ERR1_STATUS 0xd08 > +#define PHB_OUT_ERR_INJECT 0xd10 > +#define PHB_OUT_ERR_LEM_ENABLE 0xd18 > +#define PHB_OUT_ERR_IRQ_ENABLE 0xd20 > +#define PHB_OUT_ERR_FREEZE_ENABLE 0xd28 > +#define PHB_OUT_ERR_AIB_FENCE_ENABLE 0xd30 > +#define PHB_OUT_ERR_LOG_0 0xd40 > +#define PHB_OUT_ERR_LOG_1 0xd48 > +#define PHB_OUT_ERR_STATUS_MASK 0xd50 > +#define PHB_OUT_ERR1_STATUS_MASK 0xd58 > + > +#define PHB_INA_ERR_STATUS 0xd80 > +#define PHB_INA_ERR1_STATUS 0xd88 > +#define PHB_INA_ERR_INJECT 0xd90 > +#define PHB_INA_ERR_LEM_ENABLE 0xd98 > +#define PHB_INA_ERR_IRQ_ENABLE 0xda0 > +#define PHB_INA_ERR_FREEZE_ENABLE 0xda8 > +#define PHB_INA_ERR_AIB_FENCE_ENABLE 0xdb0 > +#define PHB_INA_ERR_LOG_0 0xdc0 > +#define PHB_INA_ERR_LOG_1 0xdc8 > +#define PHB_INA_ERR_STATUS_MASK 0xdd0 > +#define PHB_INA_ERR1_STATUS_MASK 0xdd8 > + > +#define PHB_INB_ERR_STATUS 0xe00 > +#define PHB_INB_ERR1_STATUS 0xe08 > +#define PHB_INB_ERR_INJECT 0xe10 > +#define PHB_INB_ERR_LEM_ENABLE 0xe18 > +#define PHB_INB_ERR_IRQ_ENABLE 0xe20 > +#define PHB_INB_ERR_FREEZE_ENABLE 0xe28 > +#define PHB_INB_ERR_AIB_FENCE_ENABLE 0xe30 > +#define PHB_INB_ERR_LOG_0 0xe40 > +#define PHB_INB_ERR_LOG_1 0xe48 > +#define PHB_INB_ERR_STATUS_MASK 0xe50 > +#define PHB_INB_ERR1_STATUS_MASK 0xe58 > + > +/* Performance monitor & Debug registers */ > +#define PHB_TRACE_CONTROL 0xf80 > +#define PHB_PERFMON_CONFIG 0xf88 > +#define PHB_PERFMON_CTR0 0xf90 > +#define PHB_PERFMON_CTR1 0xf98 > +#define PHB_PERFMON_CTR2 0xfa0 > +#define PHB_PERFMON_CTR3 0xfa8 > +#define PHB_HOTPLUG_OVERRIDE 0xfb0 > +#define PHB_HPOVR_FORCE_RESAMPLE PPC_BIT(9) > +#define PHB_HPOVR_PRESENCE_A PPC_BIT(10) > +#define PHB_HPOVR_PRESENCE_B PPC_BIT(11) > +#define PHB_HPOVR_LINK_ACTIVE PPC_BIT(12) > +#define PHB_HPOVR_LINK_BIFURCATED PPC_BIT(13) > +#define PHB_HPOVR_LINK_LANE_SWAPPED PPC_BIT(14) > + > +/* > + * IODA2 on-chip tables > + */ > + > +#define IODA2_TBL_LIST 1 > +#define IODA2_TBL_LXIVT 2 > +#define IODA2_TBL_IVC_CAM 3 > +#define IODA2_TBL_RBA 4 > +#define IODA2_TBL_RCAM 5 > +#define IODA2_TBL_MRT 6 > +#define IODA2_TBL_PESTA 7 > +#define IODA2_TBL_PESTB 8 > +#define IODA2_TBL_TVT 9 > +#define IODA2_TBL_TCAM 10 > +#define IODA2_TBL_TDR 11 > +#define IODA2_TBL_M64BT 16 > +#define IODA2_TBL_M32DT 17 > +#define IODA2_TBL_PEEV 20 > + > +/* LXIVT */ > +#define IODA2_LXIVT_SERVER PPC_BITMASK(8, 23) > +#define IODA2_LXIVT_PRIORITY PPC_BITMASK(24, 31) > +#define IODA2_LXIVT_NODE_ID PPC_BITMASK(56, 63) > + > +/* IVT */ > +#define IODA2_IVT_SERVER PPC_BITMASK(0, 23) > +#define IODA2_IVT_PRIORITY PPC_BITMASK(24, 31) > +#define IODA2_IVT_GEN PPC_BITMASK(37, 38) > +#define IODA2_IVT_P PPC_BITMASK(39, 39) > +#define IODA2_IVT_Q PPC_BITMASK(47, 47) > +#define IODA2_IVT_PE PPC_BITMASK(48, 63) > + > +/* TVT */ > +#define IODA2_TVT_TABLE_ADDR PPC_BITMASK(0, 47) > +#define IODA2_TVT_NUM_LEVELS PPC_BITMASK(48, 50) > +#define IODA2_TVE_1_LEVEL 0 > +#define IODA2_TVE_2_LEVELS 1 > +#define IODA2_TVE_3_LEVELS 2 > +#define IODA2_TVE_4_LEVELS 3 > +#define IODA2_TVE_5_LEVELS 4 > +#define IODA2_TVT_TCE_TABLE_SIZE PPC_BITMASK(51, 55) > +#define IODA2_TVT_IO_PSIZE PPC_BITMASK(59, 63) > + > +/* PESTA */ > +#define IODA2_PESTA_MMIO_FROZEN PPC_BIT(0) > + > +/* PESTB */ > +#define IODA2_PESTB_DMA_STOPPED PPC_BIT(0) > + > +/* M32DT */ > +#define IODA2_M32DT_PE PPC_BITMASK(8, 15) > + > +/* M64BT */ > +#define IODA2_M64BT_ENABLE PPC_BIT(0) > +#define IODA2_M64BT_SINGLE_PE PPC_BIT(1) > +#define IODA2_M64BT_BASE PPC_BITMASK(2, 31) > +#define IODA2_M64BT_MASK PPC_BITMASK(34, 63) > +#define IODA2_M64BT_SINGLE_BASE PPC_BITMASK(2, 26) > +#define IODA2_M64BT_PE_HI PPC_BITMASK(27, 31) > +#define IODA2_M64BT_SINGLE_MASK PPC_BITMASK(34, 58) > +#define IODA2_M64BT_PE_LOW PPC_BITMASK(59, 63) > + > +/* > + * IODA2 in-memory tables > + */ > + > +/* PEST > + * > + * 2x8 bytes entries, PEST0 and PEST1 > + */ > + > +#define IODA2_PEST0_MMIO_CAUSE PPC_BIT(2) > +#define IODA2_PEST0_CFG_READ PPC_BIT(3) > +#define IODA2_PEST0_CFG_WRITE PPC_BIT(4) > +#define IODA2_PEST0_TTYPE PPC_BITMASK(5, 7) > +#define PEST_TTYPE_DMA_WRITE 0 > +#define PEST_TTYPE_MSI 1 > +#define PEST_TTYPE_DMA_READ 2 > +#define PEST_TTYPE_DMA_READ_RESP 3 > +#define PEST_TTYPE_MMIO_LOAD 4 > +#define PEST_TTYPE_MMIO_STORE 5 > +#define PEST_TTYPE_OTHER 7 > +#define IODA2_PEST0_CA_RETURN PPC_BIT(8) > +#define IODA2_PEST0_UTL_RTOS_TIMEOUT PPC_BIT(8) /* Same bit as CA return */ > +#define IODA2_PEST0_UR_RETURN PPC_BIT(9) > +#define IODA2_PEST0_UTL_NONFATAL PPC_BIT(10) > +#define IODA2_PEST0_UTL_FATAL PPC_BIT(11) > +#define IODA2_PEST0_PARITY_UE PPC_BIT(13) > +#define IODA2_PEST0_UTL_CORRECTABLE PPC_BIT(14) > +#define IODA2_PEST0_UTL_INTERRUPT PPC_BIT(15) > +#define IODA2_PEST0_MMIO_XLATE PPC_BIT(16) > +#define IODA2_PEST0_IODA2_ERROR PPC_BIT(16) /* Same bit as MMIO xlate */ > +#define IODA2_PEST0_TCE_PAGE_FAULT PPC_BIT(18) > +#define IODA2_PEST0_TCE_ACCESS_FAULT PPC_BIT(19) > +#define IODA2_PEST0_DMA_RESP_TIMEOUT PPC_BIT(20) > +#define IODA2_PEST0_AIB_SIZE_INVALID PPC_BIT(21) > +#define IODA2_PEST0_LEM_BIT PPC_BITMASK(26, 31) > +#define IODA2_PEST0_RID PPC_BITMASK(32, 47) > +#define IODA2_PEST0_MSI_DATA PPC_BITMASK(48, 63) > + > +#define IODA2_PEST1_FAIL_ADDR PPC_BITMASK(3, 63) > + > + > +#endif /* PCI_HOST_PNV_PHB3_REGS_H */ > diff --git a/include/hw/ppc/pnv.h b/include/hw/ppc/pnv.h > index aa9d5dd16df1..8087937b132b 100644 > --- a/include/hw/ppc/pnv.h > +++ b/include/hw/ppc/pnv.h > @@ -30,6 +30,7 @@ > #include "hw/ppc/pnv_homer.h" > #include "hw/ppc/pnv_xive.h" > #include "hw/ppc/pnv_core.h" > +#include "hw/pci-host/pnv_phb3.h" > > #define TYPE_PNV_CHIP "pnv-chip" > #define PNV_CHIP(obj) OBJECT_CHECK(PnvChip, (obj), TYPE_PNV_CHIP) > @@ -58,6 +59,8 @@ typedef struct PnvChip { > uint64_t cores_mask; > void *cores; > > + uint32_t num_phbs; > + > MemoryRegion xscom_mmio; > MemoryRegion xscom; > AddressSpace xscom_as; > @@ -79,6 +82,9 @@ typedef struct Pnv8Chip { > Pnv8Psi psi; > PnvOCC occ; > PnvHomer homer; > + > +#define PNV8_CHIP_PHB3_MAX 4 > + PnvPHB3 phbs[PNV8_CHIP_PHB3_MAX]; > } Pnv8Chip; > > #define TYPE_PNV9_CHIP "pnv9-chip" > @@ -107,6 +113,7 @@ typedef struct PnvChipClass { > PnvChipType chip_type; > uint64_t chip_cfam_id; > uint64_t cores_mask; > + uint32_t num_phbs; > > DeviceRealize parent_realize; > > diff --git a/include/hw/ppc/pnv_xscom.h b/include/hw/ppc/pnv_xscom.h > index 67641ed27800..6b7d0d908f8a 100644 > --- a/include/hw/ppc/pnv_xscom.h > +++ b/include/hw/ppc/pnv_xscom.h > @@ -87,6 +87,15 @@ typedef struct PnvXScomInterfaceClass { > #define PNV9_XSCOM_XIVE_BASE 0x5013000 > #define PNV9_XSCOM_XIVE_SIZE 0x300 > > +#define PNV_XSCOM_PBCQ_NEST_BASE 0x2012000 > +#define PNV_XSCOM_PBCQ_NEST_SIZE 0x46 > + > +#define PNV_XSCOM_PBCQ_PCI_BASE 0x9012000 > +#define PNV_XSCOM_PBCQ_PCI_SIZE 0x15 > + > +#define PNV_XSCOM_PBCQ_SPCI_BASE 0x9013c00 > +#define PNV_XSCOM_PBCQ_SPCI_SIZE 0x5 > + > extern void pnv_xscom_realize(PnvChip *chip, uint64_t size, Error **errp); > extern int pnv_dt_xscom(PnvChip *chip, void *fdt, int offset); > > diff --git a/include/hw/ppc/xics.h b/include/hw/ppc/xics.h > index 64a2c8862a72..ab0954ac58e5 100644 > --- a/include/hw/ppc/xics.h > +++ b/include/hw/ppc/xics.h > @@ -174,6 +174,7 @@ void icp_set_mfrr(ICPState *icp, uint8_t mfrr); > uint32_t icp_accept(ICPState *ss); > uint32_t icp_ipoll(ICPState *ss, uint32_t *mfrr); > void icp_eoi(ICPState *icp, uint32_t xirr); > +void icp_irq(ICSState *ics, int server, int nr, uint8_t priority); > > void ics_simple_write_xive(ICSState *ics, int nr, int server, > uint8_t priority, uint8_t saved_priority); > diff --git a/hw/intc/xics.c b/hw/intc/xics.c > index b2fca2975cc4..be1e0da058be 100644 > --- a/hw/intc/xics.c > +++ b/hw/intc/xics.c > @@ -232,7 +232,7 @@ void icp_eoi(ICPState *icp, uint32_t xirr) > } > } > > -static void icp_irq(ICSState *ics, int server, int nr, uint8_t priority) > +void icp_irq(ICSState *ics, int server, int nr, uint8_t priority) > { > ICPState *icp = xics_icp_get(ics->xics, server); > > diff --git a/hw/pci-host/pnv_phb3.c b/hw/pci-host/pnv_phb3.c > new file mode 100644 > index 000000000000..086ba01fa52f > --- /dev/null > +++ b/hw/pci-host/pnv_phb3.c > @@ -0,0 +1,1191 @@ > +/* > + * QEMU PowerPC PowerNV PHB3 model > + * > + * Copyright (c) 2014-2018, IBM Corporation. > + * > + * This code is licensed under the GPL version 2 or later. See the > + * COPYING file in the top-level directory. > + */ > +#include "qemu/osdep.h" > +#include "qemu/log.h" > +#include "qapi/visitor.h" > +#include "qapi/error.h" > +#include "qemu-common.h" > +#include "hw/pci-host/pnv_phb3_regs.h" > +#include "hw/pci-host/pnv_phb3.h" > +#include "hw/pci/pcie_host.h" > +#include "hw/pci/pcie_port.h" > +#include "hw/ppc/pnv.h" > +#include "hw/irq.h" > +#include "hw/qdev-properties.h" > + > +#define phb3_error(fmt, ...) \ > + qemu_log_mask(LOG_GUEST_ERROR, "phb3: " fmt "\n", ## __VA_ARGS__) > + > + > +static PCIDevice *pnv_phb3_find_cfg_dev(PnvPHB3 *phb) > +{ > + PCIHostState *pci = PCI_HOST_BRIDGE(phb); > + uint64_t addr = phb->regs[PHB_CONFIG_ADDRESS >> 3]; > + uint8_t bus, devfn; > + > + if (!(addr >> 63)) { > + return NULL; > + } > + bus = (addr >> 52) & 0xff; > + devfn = (addr >> 44) & 0xff; > + > + return pci_find_device(pci->bus, bus, devfn); > +} > + > +/* > + * The CONFIG_DATA register expects little endian accesses, but as the > + * region is big endian, we have to swap the value. > + */ > +static void pnv_phb3_config_write(PnvPHB3 *phb, unsigned off, > + unsigned size, uint64_t val) > +{ > + uint32_t cfg_addr, limit; > + PCIDevice *pdev; > + > + pdev = pnv_phb3_find_cfg_dev(phb); > + if (!pdev) { > + return; > + } > + cfg_addr = (phb->regs[PHB_CONFIG_ADDRESS >> 3] >> 32) & 0xffc; > + cfg_addr |= off; > + limit = pci_config_size(pdev); > + if (limit <= cfg_addr) { > + /* conventional pci device can be behind pcie-to-pci bridge. > + 256 <= addr < 4K has no effects. */ > + return; > + } > + switch (size) { > + case 1: > + break; > + case 2: > + val = bswap16(val); > + break; > + case 4: > + val = bswap32(val); > + break; > + default: > + g_assert_not_reached(); > + } > + pci_host_config_write_common(pdev, cfg_addr, limit, val, size); > +} > + > +static uint64_t pnv_phb3_config_read(PnvPHB3 *phb, unsigned off, > + unsigned size) > +{ > + uint32_t cfg_addr, limit; > + PCIDevice *pdev; > + uint64_t val; > + > + pdev = pnv_phb3_find_cfg_dev(phb); > + if (!pdev) { > + return ~0ull; > + } > + cfg_addr = (phb->regs[PHB_CONFIG_ADDRESS >> 3] >> 32) & 0xffc; > + cfg_addr |= off; > + limit = pci_config_size(pdev); > + if (limit <= cfg_addr) { > + /* conventional pci device can be behind pcie-to-pci bridge. > + 256 <= addr < 4K has no effects. */ > + return ~0ull; > + } > + val = pci_host_config_read_common(pdev, cfg_addr, limit, size); > + switch (size) { > + case 1: > + return val; > + case 2: > + return bswap16(val); > + case 4: > + return bswap32(val); > + default: > + g_assert_not_reached(); > + } > +} > + > +static void pnv_phb3_check_m32(PnvPHB3 *phb) > +{ > + uint64_t base, start, size; > + MemoryRegion *parent; > + PnvPBCQState *pbcq = &phb->pbcq; > + > + if (phb->m32_mapped) { > + memory_region_del_subregion(phb->mr_m32.container, &phb->mr_m32); > + phb->m32_mapped = false; > + } > + > + if (!(phb->regs[PHB_PHB3_CONFIG >> 3] & PHB_PHB3C_M32_EN)) { > + return; > + } > + > + /* Grab geometry from registers */ > + base = phb->regs[PHB_M32_BASE_ADDR >> 3]; > + start = phb->regs[PHB_M32_START_ADDR >> 3]; > + size = ~(phb->regs[PHB_M32_BASE_MASK >> 3] | 0xfffc000000000000ull) + 1; > + > + /* Check if it matches an enabled MMIO region in the PBCQ */ > + if (pbcq->mmio0_mapped && base >= pbcq->mmio0_base && > + (base + size) <= (pbcq->mmio0_base + pbcq->mmio0_size)) { > + parent = &pbcq->mmbar0; > + base -= pbcq->mmio0_base; > + } else if (pbcq->mmio1_mapped && base >= pbcq->mmio1_base && > + (base + size) <= (pbcq->mmio1_base + pbcq->mmio1_size)) { > + parent = &pbcq->mmbar1; > + base -= pbcq->mmio1_base; > + } else { > + return; > + } > + > + /* Create alias */ > + memory_region_init_alias(&phb->mr_m32, OBJECT(phb), "phb3-m32", > + &phb->pci_mmio, start, size); > + memory_region_add_subregion(parent, base, &phb->mr_m32); > + phb->m32_mapped = true; > +} > + > +static void pnv_phb3_check_m64(PnvPHB3 *phb, uint32_t index) > +{ > + uint64_t base, start, size, m64; > + MemoryRegion *parent; > + PnvPBCQState *pbcq = &phb->pbcq; > + > + if (phb->m64_mapped[index]) { > + /* Should we destroy it in RCU friendly way... ? */ > + memory_region_del_subregion(phb->mr_m64[index].container, > + &phb->mr_m64[index]); > + phb->m64_mapped[index] = false; > + } > + > + /* Get table entry */ > + m64 = phb->ioda_M64BT[index]; > + > + if (!(m64 & IODA2_M64BT_ENABLE)) { > + return; > + } > + > + /* Grab geometry from registers */ > + base = GETFIELD(IODA2_M64BT_BASE, m64) << 20; > + if (m64 & IODA2_M64BT_SINGLE_PE) { > + base &= ~0x1ffffffull; > + } > + size = GETFIELD(IODA2_M64BT_MASK, m64) << 20; > + size |= 0xfffc000000000000ull; > + size = ~size + 1; > + start = base | (phb->regs[PHB_M64_UPPER_BITS >> 3]); > + > + /* Check if it matches an enabled MMIO region in the PBCQ */ > + if (pbcq->mmio0_mapped && base >= pbcq->mmio0_base && > + (base + size) <= (pbcq->mmio0_base + pbcq->mmio0_size)) { > + parent = &pbcq->mmbar0; > + base -= pbcq->mmio0_base; > + } else if (pbcq->mmio1_mapped && base >= pbcq->mmio1_base && > + (base + size) <= (pbcq->mmio1_base + pbcq->mmio1_size)) { > + parent = &pbcq->mmbar1; > + base -= pbcq->mmio1_base; > + } else { > + return; > + } > + > + /* Create alias */ > + memory_region_init_alias(&phb->mr_m64[index], OBJECT(phb), "phb3-m64", > + &phb->pci_mmio, start, size); > + memory_region_add_subregion(parent, base, &phb->mr_m64[index]); > + phb->m64_mapped[index] = true; > +} > + > +static void pnv_phb3_check_all_m64s(PnvPHB3 *phb) > +{ > + uint64_t i; > + > + for (i = 0; i < PNV_PHB3_NUM_M64; i++) { > + pnv_phb3_check_m64(phb, i); > + } > +} > + > +static void pnv_phb3_lxivt_write(PnvPHB3 *phb, unsigned idx, uint64_t val) > +{ > + uint8_t server, prio; > + > + phb->ioda_LXIVT[idx] = val & (IODA2_LXIVT_SERVER | > + IODA2_LXIVT_PRIORITY | > + IODA2_LXIVT_NODE_ID); > + server = GETFIELD(IODA2_LXIVT_SERVER, val); > + prio = GETFIELD(IODA2_LXIVT_PRIORITY, val); > + > + /* > + * The low order 2 bits are the link pointer (Type II interrupts). > + * Shift back to get a valid IRQ server. > + */ > + server >>= 2; > + > + ics_simple_write_xive(&phb->lsis, idx, server, prio, prio); > +} > + > +static uint64_t *pnv_phb3_ioda_access(PnvPHB3 *phb, > + unsigned *out_table, unsigned *out_idx) > +{ > + uint64_t adreg = phb->regs[PHB_IODA_ADDR >> 3]; > + unsigned int index = GETFIELD(PHB_IODA_AD_TADR, adreg); > + unsigned int table = GETFIELD(PHB_IODA_AD_TSEL, adreg); > + unsigned int mask; > + uint64_t *tptr = NULL; > + > + switch (table) { > + case IODA2_TBL_LIST: > + tptr = phb->ioda_LIST; > + mask = 7; > + break; > + case IODA2_TBL_LXIVT: > + tptr = phb->ioda_LXIVT; > + mask = 7; > + break; > + case IODA2_TBL_IVC_CAM: > + case IODA2_TBL_RBA: > + mask = 31; > + break; > + case IODA2_TBL_RCAM: > + mask = 63; > + break; > + case IODA2_TBL_MRT: > + mask = 7; > + break; > + case IODA2_TBL_PESTA: > + case IODA2_TBL_PESTB: > + mask = 255; > + break; > + case IODA2_TBL_TVT: > + tptr = phb->ioda_TVT; > + mask = 511; > + break; > + case IODA2_TBL_TCAM: > + case IODA2_TBL_TDR: > + mask = 63; > + break; > + case IODA2_TBL_M64BT: > + tptr = phb->ioda_M64BT; > + mask = 15; > + break; > + case IODA2_TBL_M32DT: > + tptr = phb->ioda_MDT; > + mask = 255; > + break; > + case IODA2_TBL_PEEV: > + tptr = phb->ioda_PEEV; > + mask = 3; > + break; > + default: > + phb3_error("invalid IODA table %d", table); > + return NULL; > + } > + index &= mask; > + if (out_idx) { > + *out_idx = index; > + } > + if (out_table) { > + *out_table = table; > + } > + if (tptr) { > + tptr += index; > + } > + if (adreg & PHB_IODA_AD_AUTOINC) { > + index = (index + 1) & mask; > + adreg = SETFIELD(PHB_IODA_AD_TADR, adreg, index); > + } > + phb->regs[PHB_IODA_ADDR >> 3] = adreg; > + return tptr; > +} > + > +static uint64_t pnv_phb3_ioda_read(PnvPHB3 *phb) > +{ > + unsigned table; > + uint64_t *tptr; > + > + tptr = pnv_phb3_ioda_access(phb, &table, NULL); > + if (!tptr) { > + /* Return 0 on unsupported tables, not ff's */ > + return 0; > + } > + return *tptr; > +} > + > +static void pnv_phb3_ioda_write(PnvPHB3 *phb, uint64_t val) > +{ > + unsigned table, idx; > + uint64_t *tptr; > + > + tptr = pnv_phb3_ioda_access(phb, &table, &idx); > + if (!tptr) { > + return; > + } > + > + /* Handle side effects */ > + switch (table) { > + case IODA2_TBL_LXIVT: > + pnv_phb3_lxivt_write(phb, idx, val); > + break; > + case IODA2_TBL_M64BT: > + *tptr = val; > + pnv_phb3_check_m64(phb, idx); > + break; > + default: > + *tptr = val; > + } > +} > + > +/* This is called whenever the PHB LSI, MSI source ID register or > + * the PBCQ irq filters are written. > + */ > +void pnv_phb3_remap_irqs(PnvPHB3 *phb) > +{ > + ICSState *ics = &phb->lsis; > + uint32_t local, global, count, mask, comp; > + uint64_t baren; > + PnvPBCQState *pbcq = &phb->pbcq; > + > + /* First check if we are enabled. Unlike real HW we don't separate TX and RX > + * so we enable if both are set > + */ > + baren = pbcq->nest_regs[PBCQ_NEST_BAR_EN]; > + if (!(baren & PBCQ_NEST_BAR_EN_IRSN_RX) || > + !(baren & PBCQ_NEST_BAR_EN_IRSN_TX)) { > + ics->offset = 0; > + return; > + } > + > + /* Grab local LSI source ID */ > + local = GETFIELD(PHB_LSI_SRC_ID, phb->regs[PHB_LSI_SOURCE_ID >> 3]) << 3; > + > + /* Grab global one and compare */ > + global = GETFIELD(PBCQ_NEST_LSI_SRC, > + pbcq->nest_regs[PBCQ_NEST_LSI_SRC_ID]) << 3; > + if (global != local) { > + /* This happens during initialization, let's come back when we > + * are properly configured > + */ > + ics->offset = 0; > + return; > + } > + > + /* Get the base on the powerbus */ > + comp = GETFIELD(PBCQ_NEST_IRSN_COMP, > + pbcq->nest_regs[PBCQ_NEST_IRSN_COMPARE]); > + mask = GETFIELD(PBCQ_NEST_IRSN_COMP, > + pbcq->nest_regs[PBCQ_NEST_IRSN_MASK]); > + count = ((~mask) + 1) & 0x7ffff; > + phb->total_irq = count; > + > + /* Sanity checks */ > + if ((global + PNV_PHB3_NUM_LSI) > count) { > + phb3_error("LSIs out of reach: LSI base=%d total irq=%d", global, > + count); > + } > + > + if (count > 2048) { > + phb3_error("More interrupts than supported: %d", count); > + } > + > + if ((comp & mask) != comp) { > + phb3_error("IRQ compare bits not in mask: comp=0x%x mask=0x%x", > + comp, mask); > + comp &= mask; > + } > + /* Setup LSI offset */ > + ics->offset = comp + global; > + > + /* Setup MSI offset */ > + pnv_phb3_msi_update_config(&phb->msis, comp, count - PNV_PHB3_NUM_LSI); > +} > + > +static void pnv_phb3_lsi_src_id_write(PnvPHB3 *phb, uint64_t val) > +{ > + /* Sanitize content */ > + val &= PHB_LSI_SRC_ID; > + phb->regs[PHB_LSI_SOURCE_ID >> 3] = val; > + pnv_phb3_remap_irqs(phb); > +} > + > +static void pnv_phb3_rtc_invalidate(PnvPHB3 *phb, uint64_t val) > +{ > + PnvPhb3DMASpace *ds; > + > + /* Always invalidate all for now ... */ > + QLIST_FOREACH(ds, &phb->dma_spaces, list) { > + ds->pe_num = PHB_INVALID_PE; > + } > +} > + > + > +static void pnv_phb3_update_msi_regions(PnvPhb3DMASpace *ds) > +{ > + uint64_t cfg = ds->phb->regs[PHB_PHB3_CONFIG >> 3]; > + > + if (cfg & PHB_PHB3C_32BIT_MSI_EN) { > + if (!ds->msi32_mapped) { > + memory_region_add_subregion(MEMORY_REGION(&ds->dma_mr), > + 0xffff0000, &ds->msi32_mr); > + ds->msi32_mapped = true; > + } > + } else { > + if (ds->msi32_mapped) { > + memory_region_del_subregion(MEMORY_REGION(&ds->dma_mr), > + &ds->msi32_mr); > + ds->msi32_mapped = false; > + } > + } > + > + if (cfg & PHB_PHB3C_64BIT_MSI_EN) { > + if (!ds->msi64_mapped) { > + memory_region_add_subregion(MEMORY_REGION(&ds->dma_mr), > + (1ull << 60), &ds->msi64_mr); > + ds->msi64_mapped = true; > + } > + } else { > + if (ds->msi64_mapped) { > + memory_region_del_subregion(MEMORY_REGION(&ds->dma_mr), > + &ds->msi64_mr); > + ds->msi64_mapped = false; > + } > + } > +} > + > +static void pnv_phb3_update_all_msi_regions(PnvPHB3 *phb) > +{ > + PnvPhb3DMASpace *ds; > + > + QLIST_FOREACH(ds, &phb->dma_spaces, list) { > + pnv_phb3_update_msi_regions(ds); > + } > +} > + > +void pnv_phb3_reg_write(void *opaque, hwaddr off, uint64_t val, unsigned size) > +{ > + PnvPHB3 *phb = opaque; > + bool changed; > + > + /* Special case configuration data */ > + if ((off & 0xfffc) == PHB_CONFIG_DATA) { > + pnv_phb3_config_write(phb, off & 0x3, size, val); > + return; > + } > + > + /* Other registers are 64-bit only */ > + if (size != 8 || off & 0x7) { > + phb3_error("Invalid register access, offset: 0x%"PRIx64" size: %d", > + off, size); > + return; > + } > + > + /* Handle masking & filtering */ > + switch (off) { > + case PHB_M64_UPPER_BITS: > + val &= 0xfffc000000000000ull; > + break; > + case PHB_Q_DMA_R: > + /* This is enough logic to make SW happy but we aren't actually > + * quiescing the DMAs > + */ > + if (val & PHB_Q_DMA_R_AUTORESET) { > + val = 0; > + } else { > + val &= PHB_Q_DMA_R_QUIESCE_DMA; > + } > + break; > + /* LEM stuff */ > + case PHB_LEM_FIR_AND_MASK: > + phb->regs[PHB_LEM_FIR_ACCUM >> 3] &= val; > + return; > + case PHB_LEM_FIR_OR_MASK: > + phb->regs[PHB_LEM_FIR_ACCUM >> 3] |= val; > + return; > + case PHB_LEM_ERROR_AND_MASK: > + phb->regs[PHB_LEM_ERROR_MASK >> 3] &= val; > + return; > + case PHB_LEM_ERROR_OR_MASK: > + phb->regs[PHB_LEM_ERROR_MASK >> 3] |= val; > + return; > + case PHB_LEM_WOF: > + val = 0; > + break; > + } > + > + /* Record whether it changed */ > + changed = phb->regs[off >> 3] != val; > + > + /* Store in register cache first */ > + phb->regs[off >> 3] = val; > + > + /* Handle side effects */ > + switch (off) { > + case PHB_PHB3_CONFIG: > + if (changed) { > + pnv_phb3_update_all_msi_regions(phb); > + } > + /* fall through */ > + case PHB_M32_BASE_ADDR: > + case PHB_M32_BASE_MASK: > + case PHB_M32_START_ADDR: > + if (changed) { > + pnv_phb3_check_m32(phb); > + } > + break; > + case PHB_M64_UPPER_BITS: > + if (changed) { > + pnv_phb3_check_all_m64s(phb); > + } > + break; > + case PHB_LSI_SOURCE_ID: > + if (changed) { > + pnv_phb3_lsi_src_id_write(phb, val); > + } > + break; > + > + /* IODA table accesses */ > + case PHB_IODA_DATA0: > + pnv_phb3_ioda_write(phb, val); > + break; > + > + /* RTC invalidation */ > + case PHB_RTC_INVALIDATE: > + pnv_phb3_rtc_invalidate(phb, val); > + break; > + > + /* FFI request */ > + case PHB_FFI_REQUEST: > + pnv_phb3_msi_ffi(&phb->msis, val); > + break; > + > + /* Silent simple writes */ > + case PHB_CONFIG_ADDRESS: > + case PHB_IODA_ADDR: > + case PHB_TCE_KILL: > + case PHB_TCE_SPEC_CTL: > + case PHB_PEST_BAR: > + case PHB_PELTV_BAR: > + case PHB_RTT_BAR: > + case PHB_RBA_BAR: > + case PHB_IVT_BAR: > + case PHB_FFI_LOCK: > + case PHB_LEM_FIR_ACCUM: > + case PHB_LEM_ERROR_MASK: > + case PHB_LEM_ACTION0: > + case PHB_LEM_ACTION1: > + break; > + > + /* Noise on anything else */ > + default: > + qemu_log_mask(LOG_UNIMP, "phb3: reg_write 0x%"PRIx64"=%"PRIx64"\n", > + off, val); > + } > +} > + > +uint64_t pnv_phb3_reg_read(void *opaque, hwaddr off, unsigned size) > +{ > + PnvPHB3 *phb = opaque; > + PCIHostState *pci = PCI_HOST_BRIDGE(phb); > + uint64_t val; > + > + if ((off & 0xfffc) == PHB_CONFIG_DATA) { > + return pnv_phb3_config_read(phb, off & 0x3, size); > + } > + > + /* Other registers are 64-bit only */ > + if (size != 8 || off & 0x7) { > + phb3_error("Invalid register access, offset: 0x%"PRIx64" size: %d", > + off, size); > + return ~0ull; > + } > + > + /* Default read from cache */ > + val = phb->regs[off >> 3]; > + > + switch (off) { > + /* Simulate venice DD2.0 */ > + case PHB_VERSION: > + return 0x000000a300000005ull; > + case PHB_PCIE_SYSTEM_CONFIG: > + return 0x441100fc30000000; > + > + /* IODA table accesses */ > + case PHB_IODA_DATA0: > + return pnv_phb3_ioda_read(phb); > + > + /* Link training always appears trained */ > + case PHB_PCIE_DLP_TRAIN_CTL: > + if (!pci_find_device(pci->bus, 1, 0)) > + return 0; > + return PHB_PCIE_DLP_INBAND_PRESENCE | PHB_PCIE_DLP_TC_DL_LINKACT; > + > + /* FFI Lock */ > + case PHB_FFI_LOCK: > + /* Set lock and return previous value */ > + phb->regs[off >> 3] |= PHB_FFI_LOCK_STATE; > + return val; > + > + /* DMA read sync: make it look like it's complete */ > + case PHB_DMARD_SYNC: > + return PHB_DMARD_SYNC_COMPLETE; > + > + /* Silent simple reads */ > + case PHB_PHB3_CONFIG: > + case PHB_M32_BASE_ADDR: > + case PHB_M32_BASE_MASK: > + case PHB_M32_START_ADDR: > + case PHB_CONFIG_ADDRESS: > + case PHB_IODA_ADDR: > + case PHB_RTC_INVALIDATE: > + case PHB_TCE_KILL: > + case PHB_TCE_SPEC_CTL: > + case PHB_PEST_BAR: > + case PHB_PELTV_BAR: > + case PHB_RTT_BAR: > + case PHB_RBA_BAR: > + case PHB_IVT_BAR: > + case PHB_M64_UPPER_BITS: > + case PHB_LEM_FIR_ACCUM: > + case PHB_LEM_ERROR_MASK: > + case PHB_LEM_ACTION0: > + case PHB_LEM_ACTION1: > + break; > + > + /* Noise on anything else */ > + default: > + qemu_log_mask(LOG_UNIMP, "phb3: reg_read 0x%"PRIx64"=%"PRIx64"\n", > + off, val); > + } > + return val; > +} > + > +static const MemoryRegionOps pnv_phb3_reg_ops = { > + .read = pnv_phb3_reg_read, > + .write = pnv_phb3_reg_write, > + .valid.min_access_size = 1, > + .valid.max_access_size = 8, > + .impl.min_access_size = 1, > + .impl.max_access_size = 8, > + .endianness = DEVICE_BIG_ENDIAN, > +}; > + > +static int pnv_phb3_map_irq(PCIDevice *pci_dev, int irq_num) > +{ > + /* Check that out properly ... */ > + return irq_num & 3; > +} > + > +static void pnv_phb3_set_irq(void *opaque, int irq_num, int level) > +{ > + PnvPHB3 *phb = opaque; > + > + /* LSI only ... */ > + if (irq_num > 3) { > + phb3_error("Unknown IRQ to set %d", irq_num); > + } > + qemu_set_irq(phb->qirqs[irq_num], level); > +} > + > +static bool pnv_phb3_resolve_pe(PnvPhb3DMASpace *ds) > +{ > + uint64_t rtt, addr; > + uint16_t rte; > + int bus_num; > + > + /* Already resolved ? */ > + if (ds->pe_num != PHB_INVALID_PE) { > + return true; > + } > + > + /* We need to lookup the RTT */ > + rtt = ds->phb->regs[PHB_RTT_BAR >> 3]; > + if (!(rtt & PHB_RTT_BAR_ENABLE)) { > + phb3_error("DMA with RTT BAR disabled !"); > + /* Set error bits ? fence ? ... */ > + return false; > + } > + > + /* Read RTE */ > + bus_num = pci_bus_num(ds->bus); > + addr = rtt & PHB_RTT_BASE_ADDRESS_MASK; > + addr += 2 * ((bus_num << 8) | ds->devfn); > + if (dma_memory_read(&address_space_memory, addr, &rte, sizeof(rte))) { > + phb3_error("Failed to read RTT entry at 0x%"PRIx64, addr); > + /* Set error bits ? fence ? ... */ > + return false; > + } > + rte = be16_to_cpu(rte); > + > + /* Fail upon reading of invalid PE# */ > + if (rte >= PNV_PHB3_NUM_PE) { > + phb3_error("RTE for RID 0x%x invalid (%04x", ds->devfn, rte); > + /* Set error bits ? fence ? ... */ > + return false; > + } > + ds->pe_num = rte; > + return true; > +} > + > +static void pnv_phb3_translate_tve(PnvPhb3DMASpace *ds, hwaddr addr, > + bool is_write, uint64_t tve, > + IOMMUTLBEntry *tlb) > +{ > + uint64_t tta = GETFIELD(IODA2_TVT_TABLE_ADDR, tve); > + int32_t lev = GETFIELD(IODA2_TVT_NUM_LEVELS, tve); > + uint32_t tts = GETFIELD(IODA2_TVT_TCE_TABLE_SIZE, tve); > + uint32_t tps = GETFIELD(IODA2_TVT_IO_PSIZE, tve); > + > + /* Invalid levels */ > + if (lev > 4) { > + phb3_error("Invalid #levels in TVE %d", lev); > + return; > + } > + > + /* IO Page Size of 0 means untranslated, else use TCEs */ > + if (tps == 0) { > + /* We only support non-translate in top window > + * XXX FIX THAT, Venice/Murano support it on bottom window > + * above 4G and Naples suports it on everything > + */ > + if (!(tve & PPC_BIT(51))) { > + phb3_error("xlate for invalid non-translate TVE"); > + return; > + } > + /* XXX Handle boundaries */ > + > + /* XXX Use 4k pages like q35 ... for now */ > + tlb->iova = addr & 0xfffffffffffff000ull; > + tlb->translated_addr = addr & 0x0003fffffffff000ull; > + tlb->addr_mask = 0xfffull; > + tlb->perm = IOMMU_RW; > + } else { > + uint32_t tce_shift, tbl_shift, sh; > + uint64_t base, taddr, tce, tce_mask; > + > + /* TVE disabled ? */ > + if (tts == 0) { > + phb3_error("xlate for invalid translated TVE"); > + return; > + } > + > + /* Address bits per bottom level TCE entry */ > + tce_shift = tps + 11; > + > + /* Address bits per table level */ > + tbl_shift = tts + 8; > + > + /* Top level table base address */ > + base = tta << 12; > + > + /* Total shift to first level */ > + sh = tbl_shift * lev + tce_shift; > + > + /* XXX Multi-level untested */ > + while ((lev--) >= 0) { > + /* Grab the TCE address */ > + taddr = base | (((addr >> sh) & ((1ul << tbl_shift) - 1)) << 3); > + if (dma_memory_read(&address_space_memory, taddr, &tce, > + sizeof(tce))) { > + phb3_error("Failed to read TCE at 0x%"PRIx64, taddr); > + return; > + } > + tce = be64_to_cpu(tce); > + > + /* Check permission for indirect TCE */ > + if ((lev >= 0) && !(tce & 3)) { > + phb3_error("Invalid indirect TCE at 0x%"PRIx64, taddr); > + phb3_error(" xlate %"PRIx64":%c TVE=%"PRIx64, addr, > + is_write ? 'W' : 'R', tve); > + phb3_error(" tta=%"PRIx64" lev=%d tts=%d tps=%d", > + tta, lev, tts, tps); > + return; > + } > + sh -= tbl_shift; > + base = tce & ~0xfffull; > + } > + > + /* We exit the loop with TCE being the final TCE */ > + tce_mask = ~((1ull << tce_shift) - 1); > + tlb->iova = addr & tce_mask; > + tlb->translated_addr = tce & tce_mask; > + tlb->addr_mask = ~tce_mask; > + tlb->perm = tce & 3; > + if ((is_write & !(tce & 2)) || ((!is_write) && !(tce & 1))) { > + phb3_error("TCE access fault at 0x%"PRIx64, taddr); > + phb3_error(" xlate %"PRIx64":%c TVE=%"PRIx64, addr, > + is_write ? 'W' : 'R', tve); > + phb3_error(" tta=%"PRIx64" lev=%d tts=%d tps=%d", > + tta, lev, tts, tps); > + } > + } > +} > + > +static IOMMUTLBEntry pnv_phb3_translate_iommu(IOMMUMemoryRegion *iommu, > + hwaddr addr, > + IOMMUAccessFlags flag, > + int iommu_idx) > +{ > + PnvPhb3DMASpace *ds = container_of(iommu, PnvPhb3DMASpace, dma_mr); > + int tve_sel; > + uint64_t tve, cfg; > + IOMMUTLBEntry ret = { > + .target_as = &address_space_memory, > + .iova = addr, > + .translated_addr = 0, > + .addr_mask = ~(hwaddr)0, > + .perm = IOMMU_NONE, > + }; > + > + /* Resolve PE# */ > + if (!pnv_phb3_resolve_pe(ds)) { > + phb3_error("Failed to resolve PE# for bus @%p (%d) devfn 0x%x", > + ds->bus, pci_bus_num(ds->bus), ds->devfn); > + return ret; > + } > + > + /* Check top bits */ > + switch (addr >> 60) { > + case 00: > + /* DMA or 32-bit MSI ? */ > + cfg = ds->phb->regs[PHB_PHB3_CONFIG >> 3]; > + if ((cfg & PHB_PHB3C_32BIT_MSI_EN) && > + ((addr & 0xffffffffffff0000ull) == 0xffff0000ull)) { > + phb3_error("xlate on 32-bit MSI region"); > + return ret; > + } > + /* Choose TVE XXX Use PHB3 Control Register */ > + tve_sel = (addr >> 59) & 1; > + tve = ds->phb->ioda_TVT[ds->pe_num * 2 + tve_sel]; > + pnv_phb3_translate_tve(ds, addr, flag & IOMMU_WO, tve, &ret); > + break; > + case 01: > + phb3_error("xlate on 64-bit MSI region"); > + break; > + default: > + phb3_error("xlate on unsupported address 0x%"PRIx64, addr); > + } > + return ret; > +} > + > +#define TYPE_PNV_PHB3_IOMMU_MEMORY_REGION "pnv-phb3-iommu-memory-region" > +#define PNV_PHB3_IOMMU_MEMORY_REGION(obj) \ > + OBJECT_CHECK(IOMMUMemoryRegion, (obj), TYPE_PNV_PHB3_IOMMU_MEMORY_REGION) > + > +static void pnv_phb3_iommu_memory_region_class_init(ObjectClass *klass, > + void *data) > +{ > + IOMMUMemoryRegionClass *imrc = IOMMU_MEMORY_REGION_CLASS(klass); > + > + imrc->translate = pnv_phb3_translate_iommu; > +} > + > +static const TypeInfo pnv_phb3_iommu_memory_region_info = { > + .parent = TYPE_IOMMU_MEMORY_REGION, > + .name = TYPE_PNV_PHB3_IOMMU_MEMORY_REGION, > + .class_init = pnv_phb3_iommu_memory_region_class_init, > +}; > + > +/* > + * MSI/MSIX memory region implementation. > + * The handler handles both MSI and MSIX. > + */ > +static void pnv_phb3_msi_write(void *opaque, hwaddr addr, > + uint64_t data, unsigned size) > +{ > + PnvPhb3DMASpace *ds = opaque; > + > + /* Resolve PE# */ > + if (!pnv_phb3_resolve_pe(ds)) { > + phb3_error("Failed to resolve PE# for bus @%p (%d) devfn 0x%x", > + ds->bus, pci_bus_num(ds->bus), ds->devfn); > + return; > + } > + > + pnv_phb3_msi_send(&ds->phb->msis, addr, data, ds->pe_num); > +} > + > +/* There is no .read as the read result is undefined by PCI spec */ > +static uint64_t pnv_phb3_msi_read(void *opaque, hwaddr addr, unsigned size) > +{ > + phb3_error("invalid read @ 0x%" HWADDR_PRIx, addr); > + return -1; > +} > + > +static const MemoryRegionOps pnv_phb3_msi_ops = { > + .read = pnv_phb3_msi_read, > + .write = pnv_phb3_msi_write, > + .endianness = DEVICE_LITTLE_ENDIAN > +}; > + > +static AddressSpace *pnv_phb3_dma_iommu(PCIBus *bus, void *opaque, int devfn) > +{ > + PnvPHB3 *phb = opaque; > + PnvPhb3DMASpace *ds; > + > + QLIST_FOREACH(ds, &phb->dma_spaces, list) { > + if (ds->bus == bus && ds->devfn == devfn) { > + break; > + } > + } > + > + if (ds == NULL) { > + ds = g_malloc0(sizeof(PnvPhb3DMASpace)); > + ds->bus = bus; > + ds->devfn = devfn; > + ds->pe_num = PHB_INVALID_PE; > + ds->phb = phb; > + memory_region_init_iommu(&ds->dma_mr, sizeof(ds->dma_mr), > + TYPE_PNV_PHB3_IOMMU_MEMORY_REGION, > + OBJECT(phb), "phb3_iommu", UINT64_MAX); > + address_space_init(&ds->dma_as, MEMORY_REGION(&ds->dma_mr), > + "phb3_iommu"); > + memory_region_init_io(&ds->msi32_mr, OBJECT(phb), &pnv_phb3_msi_ops, > + ds, "msi32", 0x10000); > + memory_region_init_io(&ds->msi64_mr, OBJECT(phb), &pnv_phb3_msi_ops, > + ds, "msi64", 0x100000); > + pnv_phb3_update_msi_regions(ds); > + > + QLIST_INSERT_HEAD(&phb->dma_spaces, ds, list); > + } > + return &ds->dma_as; > +} > + > +static void pnv_phb3_instance_init(Object *obj) > +{ > + PnvPHB3 *phb = PNV_PHB3(obj); > + > + QLIST_INIT(&phb->dma_spaces); > + > + /* LSI sources */ > + object_initialize(&phb->lsis, sizeof(phb->lsis), TYPE_ICS_SIMPLE); > + object_property_add_child(obj, "lsi", OBJECT(&phb->lsis), NULL); > + > + /* Default init ... will be fixed by HW inits */ > + phb->lsis.offset = 0; > + > + /* MSI sources */ > + object_initialize(&phb->msis, sizeof(phb->msis), TYPE_PHB3_MSI); > + object_property_add_const_link(OBJECT(&phb->msis), "phb", obj, > + &error_abort); > + object_property_add_child(obj, "msi", OBJECT(&phb->msis), NULL); > + > + /* Power Bus Common Queue */ > + object_initialize(&phb->pbcq, sizeof(phb->pbcq), TYPE_PNV_PBCQ); > + object_property_add_const_link(OBJECT(&phb->pbcq), "phb", obj, > + &error_abort); > + object_property_add_child(obj, "pbcq", OBJECT(&phb->pbcq), NULL); > + > + /* Root Port */ > + object_initialize(&phb->root, sizeof(phb->root), TYPE_PNV_PHB3_ROOT_PORT); > + object_property_add_child(obj, "root", OBJECT(&phb->root), NULL); > + qdev_prop_set_int32(DEVICE(&phb->root), "addr", PCI_DEVFN(0, 0)); > + qdev_prop_set_bit(DEVICE(&phb->root), "multifunction", false); > +} > + > +static void pnv_phb3_realize(DeviceState *dev, Error **errp) > +{ > + PnvPHB3 *phb = PNV_PHB3(dev); > + PCIHostState *pci = PCI_HOST_BRIDGE(dev); > + PnvMachineState *pnv = PNV_MACHINE(qdev_get_machine()); > + Error *local_err = NULL; > + int i; > + > + if (!pnv_get_chip(pnv, phb->chip_id)) { > + error_setg(errp, "invalid chip id: %d", phb->chip_id); > + return; > + } > + > + if (phb->phb_id >= PNV8_CHIP_PHB3_MAX) { > + error_setg(errp, "invalid PHB index: %d", phb->phb_id); > + return; > + } > + > + /* LSI sources */ > + object_property_add_const_link(OBJECT(&phb->lsis), "xics", OBJECT(pnv), > + &error_abort); > + object_property_set_int(OBJECT(&phb->lsis), PNV_PHB3_NUM_LSI, "nr-irqs", > + &error_abort); > + object_property_set_bool(OBJECT(&phb->lsis), true, "realized", &local_err); > + if (local_err) { > + error_propagate(errp, local_err); > + return; > + } > + > + for (i = 0; i < phb->lsis.nr_irqs; i++) { > + ics_set_irq_type(&phb->lsis, i, true); > + } > + > + phb->qirqs = qemu_allocate_irqs(ics_simple_set_irq, &phb->lsis, > + phb->lsis.nr_irqs); > + > + /* MSI sources */ > + object_property_add_const_link(OBJECT(&phb->msis), "xics", OBJECT(pnv), > + &error_abort); > + object_property_set_int(OBJECT(&phb->msis), PHB3_MAX_MSI, "nr-irqs", > + &error_abort); > + object_property_set_bool(OBJECT(&phb->msis), true, "realized", &local_err); > + if (local_err) { > + error_propagate(errp, local_err); > + return; > + } > + > + /* Power Bus Common Queue */ > + object_property_set_bool(OBJECT(&phb->pbcq), true, "realized", &local_err); > + if (local_err) { > + error_propagate(errp, local_err); > + return; > + } > + > + /* Controller Registers */ > + memory_region_init_io(&phb->mr_regs, OBJECT(phb), &pnv_phb3_reg_ops, phb, > + "phb3-regs", 0x1000); > + > + /* PHB3 doesn't support IO space. However, qemu gets very upset if > + * we don't have an IO region to anchor IO BARs onto so we just > + * initialize one which we never hook up to anything > + */ > + memory_region_init(&phb->pci_io, OBJECT(phb), "pci-io", 0x10000); > + memory_region_init(&phb->pci_mmio, OBJECT(phb), "pci-mmio", > + PCI_MMIO_TOTAL_SIZE); > + > + pci->bus = pci_register_root_bus(dev, "root-bus", > + pnv_phb3_set_irq, pnv_phb3_map_irq, phb, > + &phb->pci_mmio, &phb->pci_io, > + 0, 4, TYPE_PNV_PHB3_ROOT_BUS); > + > + pci_setup_iommu(pci->bus, pnv_phb3_dma_iommu, phb); > + > + /* Add a single Root port */ > + qdev_prop_set_uint8(DEVICE(&phb->root), "chassis", phb->chip_id); > + qdev_prop_set_uint16(DEVICE(&phb->root), "slot", phb->phb_id); > + qdev_set_parent_bus(DEVICE(&phb->root), BUS(pci->bus)); > + qdev_init_nofail(DEVICE(&phb->root)); > +} > + > +void pnv_phb3_update_regions(PnvPHB3 *phb) > +{ > + PnvPBCQState *pbcq = &phb->pbcq; > + > + /* Unmap first always */ > + if (phb->regs_mapped) { > + memory_region_del_subregion(&pbcq->phbbar, &phb->mr_regs); > + phb->regs_mapped = false; > + } > + > + /* Map registers if enabled */ > + if (pbcq->phb_mapped) { > + /* XXX We should use the PHB BAR 2 register but we don't ... */ > + memory_region_add_subregion(&pbcq->phbbar, 0, &phb->mr_regs); > + phb->regs_mapped = true; > + } > + > + /* Check/update m32 */ > + if (phb->m32_mapped) { > + pnv_phb3_check_m32(phb); > + } > + pnv_phb3_check_all_m64s(phb); > +} > + > +static const char *pnv_phb3_root_bus_path(PCIHostState *host_bridge, > + PCIBus *rootbus) > +{ > + PnvPHB3 *phb = PNV_PHB3(host_bridge); > + > + snprintf(phb->bus_path, sizeof(phb->bus_path), "00%02x:%02x", > + phb->chip_id, phb->phb_id); > + return phb->bus_path; > +} > + > +static Property pnv_phb3_properties[] = { > + DEFINE_PROP_UINT32("index", PnvPHB3, phb_id, 0), > + DEFINE_PROP_UINT32("chip-id", PnvPHB3, chip_id, 0), > + DEFINE_PROP_END_OF_LIST(), > +}; > + > +static void pnv_phb3_class_init(ObjectClass *klass, void *data) > +{ > + PCIHostBridgeClass *hc = PCI_HOST_BRIDGE_CLASS(klass); > + DeviceClass *dc = DEVICE_CLASS(klass); > + > + hc->root_bus_path = pnv_phb3_root_bus_path; > + dc->realize = pnv_phb3_realize; > + dc->props = pnv_phb3_properties; > + set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories); > +} > + > +static const TypeInfo pnv_phb3_type_info = { > + .name = TYPE_PNV_PHB3, > + .parent = TYPE_PCIE_HOST_BRIDGE, > + .instance_size = sizeof(PnvPHB3), > + .class_init = pnv_phb3_class_init, > + .instance_init = pnv_phb3_instance_init, > +}; > + > +static void pnv_phb3_root_bus_class_init(ObjectClass *klass, void *data) > +{ > + BusClass *k = BUS_CLASS(klass); > + > + /* > + * PHB3 has only a single root complex. Enforce the limit on the > + * parent bus > + */ > + k->max_dev = 1; > +} > + > +static const TypeInfo pnv_phb3_root_bus_info = { > + .name = TYPE_PNV_PHB3_ROOT_BUS, > + .parent = TYPE_PCIE_BUS, > + .class_init = pnv_phb3_root_bus_class_init, > + .interfaces = (InterfaceInfo[]) { > + { INTERFACE_PCIE_DEVICE }, > + { } > + }, > +}; > + > +static void pnv_phb3_root_port_realize(DeviceState *dev, Error **errp) > +{ > + PCIERootPortClass *rpc = PCIE_ROOT_PORT_GET_CLASS(dev); > + Error *local_err = NULL; > + > + rpc->parent_realize(dev, &local_err); > + if (local_err) { > + error_propagate(errp, local_err); > + return; > + } > +} > + > +static void pnv_phb3_root_port_class_init(ObjectClass *klass, void *data) > +{ > + DeviceClass *dc = DEVICE_CLASS(klass); > + PCIDeviceClass *k = PCI_DEVICE_CLASS(klass); > + PCIERootPortClass *rpc = PCIE_ROOT_PORT_CLASS(klass); > + > + dc->desc = "IBM PHB3 PCIE Root Port"; > + > + device_class_set_parent_realize(dc, pnv_phb3_root_port_realize, > + &rpc->parent_realize); > + > + k->vendor_id = PCI_VENDOR_ID_IBM; > + k->device_id = 0x03dc; > + k->revision = 0; > + > + rpc->exp_offset = 0x48; > + rpc->aer_offset = 0x100; > +} > + > +static const TypeInfo pnv_phb3_root_port_info = { > + .name = TYPE_PNV_PHB3_ROOT_PORT, > + .parent = TYPE_PCIE_ROOT_PORT, > + .instance_size = sizeof(PnvPHB3RootPort), > + .class_init = pnv_phb3_root_port_class_init, > +}; > + > +static void pnv_phb3_register_types(void) > +{ > + type_register_static(&pnv_phb3_root_bus_info); > + type_register_static(&pnv_phb3_root_port_info); > + type_register_static(&pnv_phb3_type_info); > + type_register_static(&pnv_phb3_iommu_memory_region_info); > +} > + > +type_init(pnv_phb3_register_types) > diff --git a/hw/pci-host/pnv_phb3_msi.c b/hw/pci-host/pnv_phb3_msi.c > new file mode 100644 > index 000000000000..e689c3c1e18c > --- /dev/null > +++ b/hw/pci-host/pnv_phb3_msi.c > @@ -0,0 +1,320 @@ > +/* > + * QEMU PowerPC PowerNV PHB3 model > + * > + * Copyright (c) 2014-2018, IBM Corporation. > + * > + * This code is licensed under the GPL version 2 or later. See the > + * COPYING file in the top-level directory. > + */ > +#include "qemu/osdep.h" > +#include "qemu/log.h" > +#include "qapi/error.h" > +#include "qemu-common.h" > +#include "hw/pci-host/pnv_phb3_regs.h" > +#include "hw/pci-host/pnv_phb3.h" > +#include "hw/ppc/pnv.h" > +#include "hw/pci/msi.h" > +#include "monitor/monitor.h" > +#include "hw/irq.h" > +#include "hw/qdev-properties.h" > +#include "sysemu/reset.h" > + > +static uint64_t phb3_msi_ive_addr(PnvPHB3 *phb, int srcno) > +{ > + uint64_t ivtbar = phb->regs[PHB_IVT_BAR >> 3]; > + uint64_t phbctl = phb->regs[PHB_CONTROL >> 3]; > + > + if (!(ivtbar & PHB_IVT_BAR_ENABLE)) { > + qemu_log_mask(LOG_GUEST_ERROR, "Failed access to disable IVT BAR !"); > + return 0; > + } > + > + if (srcno >= (ivtbar & PHB_IVT_LENGTH_MASK)) { > + qemu_log_mask(LOG_GUEST_ERROR, "MSI out of bounds (%d vs 0x%"PRIx64")", > + srcno, (uint64_t) (ivtbar & PHB_IVT_LENGTH_MASK)); > + return 0; > + } > + > + ivtbar &= PHB_IVT_BASE_ADDRESS_MASK; > + > + if (phbctl & PHB_CTRL_IVE_128_BYTES) { > + return ivtbar + 128 * srcno; > + } else { > + return ivtbar + 16 * srcno; > + } > +} > + > +static bool phb3_msi_read_ive(PnvPHB3 *phb, int srcno, uint64_t *out_ive) > +{ > + uint64_t ive_addr, ive; > + > + ive_addr = phb3_msi_ive_addr(phb, srcno); > + if (!ive_addr) { > + return false; > + } > + > + if (dma_memory_read(&address_space_memory, ive_addr, &ive, sizeof(ive))) { > + qemu_log_mask(LOG_GUEST_ERROR, "Failed to read IVE at 0x%" PRIx64, > + ive_addr); > + return false; > + } > + *out_ive = be64_to_cpu(ive); > + > + return true; > +} > + > +static void phb3_msi_set_p(Phb3MsiState *msi, int srcno, uint8_t gen) > +{ > + uint64_t ive_addr; > + uint8_t p = 0x01 | (gen << 1); > + > + ive_addr = phb3_msi_ive_addr(msi->phb, srcno); > + if (!ive_addr) { > + return; > + } > + > + if (dma_memory_write(&address_space_memory, ive_addr + 4, &p, 1)) { > + qemu_log_mask(LOG_GUEST_ERROR, > + "Failed to write IVE (set P) at 0x%" PRIx64, ive_addr); > + } > +} > + > +static void phb3_msi_set_q(Phb3MsiState *msi, int srcno) > +{ > + uint64_t ive_addr; > + uint8_t q = 0x01; > + > + ive_addr = phb3_msi_ive_addr(msi->phb, srcno); > + if (!ive_addr) { > + return; > + } > + > + if (dma_memory_write(&address_space_memory, ive_addr + 5, &q, 1)) { > + qemu_log_mask(LOG_GUEST_ERROR, > + "Failed to write IVE (set Q) at 0x%" PRIx64, ive_addr); > + } > +} > + > +static void phb3_msi_try_send(Phb3MsiState *msi, int srcno, bool force) > +{ > + ICSState *ics = ICS_BASE(msi); > + uint64_t ive; > + uint64_t server, prio, pq, gen; > + > + if (!phb3_msi_read_ive(msi->phb, srcno, &ive)) { > + return; > + } > + > + server = GETFIELD(IODA2_IVT_SERVER, ive); > + prio = GETFIELD(IODA2_IVT_PRIORITY, ive); > + if (!force) { > + pq = GETFIELD(IODA2_IVT_Q, ive) | (GETFIELD(IODA2_IVT_P, ive) << 1); > + } else { > + pq = 0; > + } > + gen = GETFIELD(IODA2_IVT_GEN, ive); > + > + /* > + * The low order 2 bits are the link pointer (Type II interrupts). > + * Shift back to get a valid IRQ server. > + */ > + server >>= 2; > + > + switch (pq) { > + case 0: /* 00 */ > + if (prio == 0xff) { > + /* Masked, set Q */ > + phb3_msi_set_q(msi, srcno); > + } else { > + /* Enabled, set P and send */ > + phb3_msi_set_p(msi, srcno, gen); > + icp_irq(ics, server, srcno + ics->offset, prio); > + } > + break; > + case 2: /* 10 */ > + /* Already pending, set Q */ > + phb3_msi_set_q(msi, srcno); > + break; > + case 1: /* 01 */ > + case 3: /* 11 */ > + default: > + /* Just drop stuff if Q already set */ > + break; > + } > +} > + > +static void phb3_msi_set_irq(void *opaque, int srcno, int val) > +{ > + Phb3MsiState *msi = PHB3_MSI(opaque); > + > + if (val) { > + phb3_msi_try_send(msi, srcno, false); > + } > +} > + > + > +void pnv_phb3_msi_send(Phb3MsiState *msi, uint64_t addr, uint16_t data, > + int32_t dev_pe) > +{ > + ICSState *ics = ICS_BASE(msi); > + uint64_t ive; > + uint16_t pe; > + uint32_t src = ((addr >> 4) & 0xffff) | (data & 0x1f); > + > + if (src >= ics->nr_irqs) { > + qemu_log_mask(LOG_GUEST_ERROR, "MSI %d out of bounds", src); > + return; > + } > + if (dev_pe >= 0) { > + if (!phb3_msi_read_ive(msi->phb, src, &ive)) { > + return; > + } > + pe = GETFIELD(IODA2_IVT_PE, ive); > + if (pe != dev_pe) { > + qemu_log_mask(LOG_GUEST_ERROR, > + "MSI %d send by PE#%d but assigned to PE#%d", > + src, dev_pe, pe); > + return; > + } > + } > + qemu_irq_pulse(msi->qirqs[src]); > +} > + > +void pnv_phb3_msi_ffi(Phb3MsiState *msi, uint64_t val) > +{ > + /* Emit interrupt */ > + pnv_phb3_msi_send(msi, val, 0, -1); > + > + /* Clear FFI lock */ > + msi->phb->regs[PHB_FFI_LOCK >> 3] = 0; > +} > + > +static void phb3_msi_reject(ICSState *ics, uint32_t nr) > +{ > + Phb3MsiState *msi = PHB3_MSI(ics); > + unsigned int srcno = nr - ics->offset; > + unsigned int idx = srcno >> 6; > + unsigned int bit = 1ull << (srcno & 0x3f); > + > + assert(srcno < PHB3_MAX_MSI); > + > + msi->rba[idx] |= bit; > + msi->rba_sum |= (1u << idx); > +} > + > +static void phb3_msi_resend(ICSState *ics) > +{ > + Phb3MsiState *msi = PHB3_MSI(ics); > + unsigned int i, j; > + > + if (msi->rba_sum == 0) { > + return; > + } > + > + for (i = 0; i < 32; i++) { > + if ((msi->rba_sum & (1u << i)) == 0) { > + continue; > + } > + msi->rba_sum &= ~(1u << i); > + for (j = 0; j < 64; j++) { > + if ((msi->rba[i] & (1ull << j)) == 0) { > + continue; > + } > + msi->rba[i] &= ~(1u << j); > + phb3_msi_try_send(msi, i * 64 + j, true); > + } > + } > +} > + > +static void phb3_msi_reset(DeviceState *dev) > +{ > + Phb3MsiState *msi = PHB3_MSI(dev); > + ICSStateClass *icsc = ICS_BASE_GET_CLASS(dev); > + > + icsc->parent_reset(dev); > + > + memset(msi->rba, 0, sizeof(msi->rba)); > + msi->rba_sum = 0; > +} > + > +static void phb3_msi_reset_handler(void *dev) > +{ > + phb3_msi_reset(dev); > +} > + > +void pnv_phb3_msi_update_config(Phb3MsiState *msi, uint32_t base, > + uint32_t count) > +{ > + ICSState *ics = ICS_BASE(msi); > + > + if (count > PHB3_MAX_MSI) { > + count = PHB3_MAX_MSI; > + } > + ics->nr_irqs = count; > + ics->offset = base; > +} > + > +static void phb3_msi_realize(DeviceState *dev, Error **errp) > +{ > + Phb3MsiState *msi = PHB3_MSI(dev); > + ICSState *ics = ICS_BASE(msi); > + ICSStateClass *icsc = ICS_BASE_GET_CLASS(ics); > + Object *obj; > + Error *local_err = NULL; > + > + icsc->parent_realize(dev, &local_err); > + if (local_err) { > + error_propagate(errp, local_err); > + return; > + } > + > + obj = object_property_get_link(OBJECT(dev), "phb", &local_err); > + if (!obj) { > + error_propagate(errp, local_err); > + error_prepend(errp, "required link 'phb' not found: "); > + return; > + } > + msi->phb = PNV_PHB3(obj); > + > + msi->qirqs = qemu_allocate_irqs(phb3_msi_set_irq, msi, ics->nr_irqs); > + > + qemu_register_reset(phb3_msi_reset_handler, dev); > +} > + > +static void phb3_msi_instance_init(Object *obj) > +{ > + ICSState *ics = ICS_BASE(obj); > + > + /* Will be overriden later */ > + ics->offset = 0; > +} > + > +static void phb3_msi_class_init(ObjectClass *klass, void *data) > +{ > + DeviceClass *dc = DEVICE_CLASS(klass); > + ICSStateClass *isc = ICS_BASE_CLASS(klass); > + > + device_class_set_parent_realize(dc, phb3_msi_realize, > + &isc->parent_realize); > + device_class_set_parent_reset(dc, phb3_msi_reset, > + &isc->parent_reset); > + > + isc->reject = phb3_msi_reject; > + isc->resend = phb3_msi_resend; > +} > + > +static const TypeInfo phb3_msi_info = { > + .name = TYPE_PHB3_MSI, > + .parent = TYPE_ICS_BASE, > + .instance_size = sizeof(Phb3MsiState), > + .class_init = phb3_msi_class_init, > + .class_size = sizeof(ICSStateClass), > + .instance_init = phb3_msi_instance_init, > +}; > + > +static void pnv_phb3_msi_register_types(void) > +{ > + type_register_static(&phb3_msi_info); > +} > + > +type_init(pnv_phb3_msi_register_types) > diff --git a/hw/pci-host/pnv_phb3_pbcq.c b/hw/pci-host/pnv_phb3_pbcq.c > new file mode 100644 > index 000000000000..34df3ac684e4 > --- /dev/null > +++ b/hw/pci-host/pnv_phb3_pbcq.c > @@ -0,0 +1,347 @@ > +/* > + * QEMU PowerPC PowerNV PHB3 model > + * > + * Copyright (c) 2014-2018, IBM Corporation. > + * > + * This code is licensed under the GPL version 2 or later. See the > + * COPYING file in the top-level directory. > + */ > +#include "qemu/osdep.h" > +#include "qapi/error.h" > +#include "qemu-common.h" > +#include "qemu/log.h" > +#include "target/ppc/cpu.h" > +#include "hw/ppc/fdt.h" > +#include "hw/pci-host/pnv_phb3_regs.h" > +#include "hw/pci-host/pnv_phb3.h" > +#include "hw/ppc/pnv_xscom.h" > +#include "hw/pci/pci_bridge.h" > +#include "hw/pci/pci_bus.h" > + > +#include <libfdt.h> > + > +static uint64_t pnv_pbcq_nest_xscom_read(void *opaque, hwaddr addr, > + unsigned size) > +{ > + PnvPBCQState *pbcq = PNV_PBCQ(opaque); > + uint32_t offset = addr >> 3; > + > + return pbcq->nest_regs[offset]; > +} > + > +static uint64_t pnv_pbcq_pci_xscom_read(void *opaque, hwaddr addr, > + unsigned size) > +{ > + PnvPBCQState *pbcq = PNV_PBCQ(opaque); > + uint32_t offset = addr >> 3; > + > + return pbcq->pci_regs[offset]; > +} > + > +static uint64_t pnv_pbcq_spci_xscom_read(void *opaque, hwaddr addr, > + unsigned size) > +{ > + PnvPBCQState *pbcq = PNV_PBCQ(opaque); > + uint32_t offset = addr >> 3; > + > + if (offset == PBCQ_SPCI_ASB_DATA) { > + return pnv_phb3_reg_read(pbcq->phb, > + pbcq->spci_regs[PBCQ_SPCI_ASB_ADDR], 8); > + } > + return pbcq->spci_regs[offset]; > +} > + > +static void pnv_pbcq_update_map(PnvPBCQState *pbcq) > +{ > + uint64_t bar_en = pbcq->nest_regs[PBCQ_NEST_BAR_EN]; > + uint64_t bar, mask, size; > + > + /* > + * NOTE: This will really not work well if those are remapped > + * after the PHB has created its sub regions. We could do better > + * if we had a way to resize regions but we don't really care > + * that much in practice as the stuff below really only happens > + * once early during boot > + */ > + > + /* Handle unmaps */ > + if (pbcq->mmio0_mapped && !(bar_en & PBCQ_NEST_BAR_EN_MMIO0)) { > + memory_region_del_subregion(get_system_memory(), &pbcq->mmbar0); > + pbcq->mmio0_mapped = false; > + } > + if (pbcq->mmio1_mapped && !(bar_en & PBCQ_NEST_BAR_EN_MMIO1)) { > + memory_region_del_subregion(get_system_memory(), &pbcq->mmbar1); > + pbcq->mmio1_mapped = false; > + } > + if (pbcq->phb_mapped && !(bar_en & PBCQ_NEST_BAR_EN_PHB)) { > + memory_region_del_subregion(get_system_memory(), &pbcq->phbbar); > + pbcq->phb_mapped = false; > + } > + > + /* Update PHB */ > + pnv_phb3_update_regions(pbcq->phb); > + > + /* Handle maps */ > + if (!pbcq->mmio0_mapped && (bar_en & PBCQ_NEST_BAR_EN_MMIO0)) { > + bar = pbcq->nest_regs[PBCQ_NEST_MMIO_BAR0] >> 14; > + mask = pbcq->nest_regs[PBCQ_NEST_MMIO_MASK0]; > + size = ((~mask) >> 14) + 1; > + memory_region_init(&pbcq->mmbar0, OBJECT(pbcq), "pbcq-mmio0", size); > + memory_region_add_subregion(get_system_memory(), bar, &pbcq->mmbar0); > + pbcq->mmio0_mapped = true; > + pbcq->mmio0_base = bar; > + pbcq->mmio0_size = size; > + } > + if (!pbcq->mmio1_mapped && (bar_en & PBCQ_NEST_BAR_EN_MMIO1)) { > + bar = pbcq->nest_regs[PBCQ_NEST_MMIO_BAR1] >> 14; > + mask = pbcq->nest_regs[PBCQ_NEST_MMIO_MASK1]; > + size = ((~mask) >> 14) + 1; > + memory_region_init(&pbcq->mmbar1, OBJECT(pbcq), "pbcq-mmio1", size); > + memory_region_add_subregion(get_system_memory(), bar, &pbcq->mmbar1); > + pbcq->mmio1_mapped = true; > + pbcq->mmio1_base = bar; > + pbcq->mmio1_size = size; > + } > + if (!pbcq->phb_mapped && (bar_en & PBCQ_NEST_BAR_EN_PHB)) { > + bar = pbcq->nest_regs[PBCQ_NEST_PHB_BAR] >> 14; > + size = 0x1000; > + memory_region_init(&pbcq->phbbar, OBJECT(pbcq), "pbcq-phb", size); > + memory_region_add_subregion(get_system_memory(), bar, &pbcq->phbbar); > + pbcq->phb_mapped = true; > + } > + > + /* Update PHB */ > + pnv_phb3_update_regions(pbcq->phb); > +} > + > +static void pnv_pbcq_nest_xscom_write(void *opaque, hwaddr addr, > + uint64_t val, unsigned size) > +{ > + PnvPBCQState *pbcq = PNV_PBCQ(opaque); > + uint32_t reg = addr >> 3; > + > + switch (reg) { > + case PBCQ_NEST_MMIO_BAR0: > + case PBCQ_NEST_MMIO_BAR1: > + case PBCQ_NEST_MMIO_MASK0: > + case PBCQ_NEST_MMIO_MASK1: > + if (pbcq->nest_regs[PBCQ_NEST_BAR_EN] & > + (PBCQ_NEST_BAR_EN_MMIO0 | > + PBCQ_NEST_BAR_EN_MMIO1)) { > + qemu_log_mask(LOG_GUEST_ERROR, > + "PHB3: Changing enabled BAR unsupported\n"); > + } > + pbcq->nest_regs[reg] = val & 0xffffffffc0000000ull; > + break; > + case PBCQ_NEST_PHB_BAR: > + if (pbcq->nest_regs[PBCQ_NEST_BAR_EN] & PBCQ_NEST_BAR_EN_PHB) { > + qemu_log_mask(LOG_GUEST_ERROR, > + "PHB3: Changing enabled BAR unsupported\n"); > + } > + pbcq->nest_regs[reg] = val & 0xfffffffffc000000ull; > + break; > + case PBCQ_NEST_BAR_EN: > + pbcq->nest_regs[reg] = val & 0xf800000000000000ull; > + pnv_pbcq_update_map(pbcq); > + pnv_phb3_remap_irqs(pbcq->phb); > + break; > + case PBCQ_NEST_IRSN_COMPARE: > + case PBCQ_NEST_IRSN_MASK: > + pbcq->nest_regs[reg] = val & PBCQ_NEST_IRSN_COMP; > + pnv_phb3_remap_irqs(pbcq->phb); > + break; > + case PBCQ_NEST_LSI_SRC_ID: > + pbcq->nest_regs[reg] = val & PBCQ_NEST_LSI_SRC; > + pnv_phb3_remap_irqs(pbcq->phb); > + break; > + } > + > + /* XXX Don't error out on other regs for now ... */ > +} > + > +static void pnv_pbcq_pci_xscom_write(void *opaque, hwaddr addr, > + uint64_t val, unsigned size) > +{ > + PnvPBCQState *pbcq = PNV_PBCQ(opaque); > + uint32_t reg = addr >> 3; > + > + switch (reg) { > + case PBCQ_PCI_BAR2: > + pbcq->pci_regs[reg] = val & 0xfffffffffc000000ull; > + pnv_pbcq_update_map(pbcq); > + break; > + } > + > + /* XXX Don't error out on other regs for now ... */ > +} > + > +static void pnv_pbcq_spci_xscom_write(void *opaque, hwaddr addr, > + uint64_t val, unsigned size) > +{ > + PnvPBCQState *pbcq = PNV_PBCQ(opaque); > + uint32_t reg = addr >> 3; > + > + switch (reg) { > + case PBCQ_SPCI_ASB_ADDR: > + pbcq->spci_regs[reg] = val & 0xfff; > + break; > + case PBCQ_SPCI_ASB_STATUS: > + pbcq->spci_regs[reg] &= ~val; > + break; > + case PBCQ_SPCI_ASB_DATA: > + pnv_phb3_reg_write(pbcq->phb, pbcq->spci_regs[PBCQ_SPCI_ASB_ADDR], > + val, 8); > + break; > + case PBCQ_SPCI_AIB_CAPP_EN: > + case PBCQ_SPCI_CAPP_SEC_TMR: > + break; > + } > + > + /* XXX Don't error out on other regs for now ... */ > +} > + > +static const MemoryRegionOps pnv_pbcq_nest_xscom_ops = { > + .read = pnv_pbcq_nest_xscom_read, > + .write = pnv_pbcq_nest_xscom_write, > + .valid.min_access_size = 8, > + .valid.max_access_size = 8, > + .impl.min_access_size = 8, > + .impl.max_access_size = 8, > + .endianness = DEVICE_BIG_ENDIAN, > +}; > + > +static const MemoryRegionOps pnv_pbcq_pci_xscom_ops = { > + .read = pnv_pbcq_pci_xscom_read, > + .write = pnv_pbcq_pci_xscom_write, > + .valid.min_access_size = 8, > + .valid.max_access_size = 8, > + .impl.min_access_size = 8, > + .impl.max_access_size = 8, > + .endianness = DEVICE_BIG_ENDIAN, > +}; > + > +static const MemoryRegionOps pnv_pbcq_spci_xscom_ops = { > + .read = pnv_pbcq_spci_xscom_read, > + .write = pnv_pbcq_spci_xscom_write, > + .valid.min_access_size = 8, > + .valid.max_access_size = 8, > + .impl.min_access_size = 8, > + .impl.max_access_size = 8, > + .endianness = DEVICE_BIG_ENDIAN, > +}; > + > +static void pnv_pbcq_default_bars(PnvPBCQState *pbcq) > +{ > + uint64_t mm0, mm1, reg; > + PnvPHB3 *phb = pbcq->phb; > + > + mm0 = 0x3d00000000000ull + 0x4000000000ull * phb->chip_id + > + 0x1000000000ull * phb->phb_id; > + mm1 = 0x3ff8000000000ull + 0x0200000000ull * phb->chip_id + > + 0x0080000000ull * phb->phb_id; > + reg = 0x3fffe40000000ull + 0x0000400000ull * phb->chip_id + > + 0x0000100000ull * phb->phb_id; > + > + pbcq->nest_regs[PBCQ_NEST_MMIO_BAR0] = mm0 << 14; > + pbcq->nest_regs[PBCQ_NEST_MMIO_BAR1] = mm1 << 14; > + pbcq->nest_regs[PBCQ_NEST_PHB_BAR] = reg << 14; > + pbcq->nest_regs[PBCQ_NEST_MMIO_MASK0] = 0x3fff000000000ull << 14; > + pbcq->nest_regs[PBCQ_NEST_MMIO_MASK1] = 0x3ffff80000000ull << 14; > + pbcq->pci_regs[PBCQ_PCI_BAR2] = reg << 14; > +} > + > +static void pnv_pbcq_realize(DeviceState *dev, Error **errp) > +{ > + PnvPBCQState *pbcq = PNV_PBCQ(dev); > + PnvPHB3 *phb; > + Object *obj; > + Error *local_err = NULL; > + char name[32]; > + > + obj = object_property_get_link(OBJECT(dev), "phb", &local_err); > + if (!obj) { > + error_propagate(errp, local_err); > + error_prepend(errp, "required link 'phb' not found: "); > + return; > + } > + phb = pbcq->phb = PNV_PHB3(obj); > + > + /* XXX Fix OPAL to do that: establish default BAR values */ > + pnv_pbcq_default_bars(pbcq); > + > + /* Initialize the XSCOM region for the PBCQ registers */ > + snprintf(name, sizeof(name), "xscom-pbcq-nest-%d.%d", > + phb->chip_id, phb->phb_id); > + pnv_xscom_region_init(&pbcq->xscom_nest_regs, OBJECT(dev), > + &pnv_pbcq_nest_xscom_ops, pbcq, name, > + PNV_XSCOM_PBCQ_NEST_SIZE); > + snprintf(name, sizeof(name), "xscom-pbcq-pci-%d.%d", > + phb->chip_id, phb->phb_id); > + pnv_xscom_region_init(&pbcq->xscom_pci_regs, OBJECT(dev), > + &pnv_pbcq_pci_xscom_ops, pbcq, name, > + PNV_XSCOM_PBCQ_PCI_SIZE); > + snprintf(name, sizeof(name), "xscom-pbcq-spci-%d.%d", > + phb->chip_id, phb->phb_id); > + pnv_xscom_region_init(&pbcq->xscom_spci_regs, OBJECT(dev), > + &pnv_pbcq_spci_xscom_ops, pbcq, name, > + PNV_XSCOM_PBCQ_SPCI_SIZE); > +} > + > +static int pnv_pbcq_dt_xscom(PnvXScomInterface *dev, void *fdt, > + int xscom_offset) > +{ > + const char compat[] = "ibm,power8-pbcq"; > + PnvPHB3 *phb = PNV_PBCQ(dev)->phb; > + char *name; > + int offset; > + uint32_t lpc_pcba = PNV_XSCOM_PBCQ_NEST_BASE + 0x400 * phb->phb_id; > + uint32_t reg[] = { > + cpu_to_be32(lpc_pcba), > + cpu_to_be32(PNV_XSCOM_PBCQ_NEST_SIZE), > + cpu_to_be32(PNV_XSCOM_PBCQ_PCI_BASE + 0x400 * phb->phb_id), > + cpu_to_be32(PNV_XSCOM_PBCQ_PCI_SIZE), > + cpu_to_be32(PNV_XSCOM_PBCQ_SPCI_BASE + 0x040 * phb->phb_id), > + cpu_to_be32(PNV_XSCOM_PBCQ_SPCI_SIZE) > + }; > + > + name = g_strdup_printf("pbcq@%x", lpc_pcba); > + offset = fdt_add_subnode(fdt, xscom_offset, name); > + _FDT(offset); > + g_free(name); > + > + _FDT((fdt_setprop(fdt, offset, "reg", reg, sizeof(reg)))); > + > + _FDT((fdt_setprop_cell(fdt, offset, "ibm,phb-index", phb->phb_id))); > + _FDT((fdt_setprop_cell(fdt, offset, "ibm,chip-id", phb->chip_id))); > + _FDT((fdt_setprop(fdt, offset, "compatible", compat, > + sizeof(compat)))); > + return 0; > +} > + > +static void pnv_pbcq_class_init(ObjectClass *klass, void *data) > +{ > + DeviceClass *dc = DEVICE_CLASS(klass); > + PnvXScomInterfaceClass *xdc = PNV_XSCOM_INTERFACE_CLASS(klass); > + > + xdc->dt_xscom = pnv_pbcq_dt_xscom; > + > + dc->realize = pnv_pbcq_realize; > +} > + > +static const TypeInfo pnv_pbcq_type_info = { > + .name = TYPE_PNV_PBCQ, > + .parent = TYPE_DEVICE, > + .instance_size = sizeof(PnvPBCQState), > + .class_init = pnv_pbcq_class_init, > + .interfaces = (InterfaceInfo[]) { > + { TYPE_PNV_XSCOM_INTERFACE }, > + { } > + } > +}; > + > +static void pnv_pbcq_register_types(void) > +{ > + type_register_static(&pnv_pbcq_type_info); > +} > + > +type_init(pnv_pbcq_register_types) > diff --git a/hw/ppc/pnv.c b/hw/ppc/pnv.c > index c8014d081a40..0dcc91ac101b 100644 > --- a/hw/ppc/pnv.c > +++ b/hw/ppc/pnv.c > @@ -50,6 +50,7 @@ > #include "hw/boards.h" > #include "hw/char/serial.h" > #include "hw/timer/mc146818rtc.h" > +#include "hw/pci/msi.h" > > #include <libfdt.h> > > @@ -604,8 +605,13 @@ static ISABus *pnv_isa_create(PnvChip *chip, Error **errp) > static void pnv_chip_power8_pic_print_info(PnvChip *chip, Monitor *mon) > { > Pnv8Chip *chip8 = PNV8_CHIP(chip); > + int i; > > ics_pic_print_info(&chip8->psi.ics, mon); > + for (i = 0; i < chip->num_phbs; i++) { > + ics_pic_print_info(ICS_BASE(&chip8->phbs[i].msis), mon); > + ics_pic_print_info(&chip8->phbs[i].lsis, mon); > + } > } > > static void pnv_chip_power9_pic_print_info(PnvChip *chip, Monitor *mon) > @@ -639,6 +645,9 @@ static void pnv_init(MachineState *machine) > DeviceState *dev; > bool load_skiboot_from_pnor = false; > > + /* MSIs are supported on this platform */ > + msi_nonbroken = true; > + > /* allocate RAM */ > if (machine->ram_size < (1 * GiB)) { > warn_report("skiboot may not work with < 1GB of RAM"); > @@ -870,7 +879,10 @@ static void pnv_chip_power9_intc_create(PnvChip *chip, PowerPCCPU *cpu, > > static void pnv_chip_power8_instance_init(Object *obj) > { > + PnvChip *chip = PNV_CHIP(obj); > Pnv8Chip *chip8 = PNV8_CHIP(obj); > + PnvChipClass *pcc = PNV_CHIP_GET_CLASS(obj); > + int i; > > object_initialize_child(obj, "psi", &chip8->psi, sizeof(chip8->psi), > TYPE_PNV8_PSI, &error_abort, NULL); > @@ -891,6 +903,17 @@ static void pnv_chip_power8_instance_init(Object *obj) > TYPE_PNV8_HOMER, &error_abort, NULL); > object_property_add_const_link(OBJECT(&chip8->homer), "chip", obj, > &error_abort); > + > + for (i = 0; i < pcc->num_phbs; i++) { > + object_initialize_child(obj, "phb[*]", &chip8->phbs[i], > + sizeof(chip8->phbs[i]), TYPE_PNV_PHB3, > + &error_abort, NULL); > + } > + > + /* > + * Number of PHBs is the chip default > + */ > + chip->num_phbs = pcc->num_phbs; > } > > static void pnv_chip_icp_realize(Pnv8Chip *chip8, Error **errp) > @@ -932,6 +955,7 @@ static void pnv_chip_power8_realize(DeviceState *dev, Error **errp) > Pnv8Chip *chip8 = PNV8_CHIP(dev); > Pnv8Psi *psi8 = &chip8->psi; > Error *local_err = NULL; > + int i; > > /* XSCOM bridge is first */ > pnv_xscom_realize(chip, PNV_XSCOM_SIZE, &local_err); > @@ -998,6 +1022,33 @@ static void pnv_chip_power8_realize(DeviceState *dev, Error **errp) > } > memory_region_add_subregion(get_system_memory(), PNV_HOMER_BASE(chip), > &chip8->homer.regs); > + > + /* PHB3 controllers */ > + for (i = 0; i < chip->num_phbs; i++) { > + PnvPHB3 *phb = &chip8->phbs[i]; > + PnvPBCQState *pbcq = &phb->pbcq; > + > + object_property_set_int(OBJECT(phb), i, "index", &error_fatal); > + object_property_set_int(OBJECT(phb), chip->chip_id, "chip-id", > + &error_fatal); > + object_property_set_bool(OBJECT(phb), true, "realized", &local_err); > + if (local_err) { > + error_propagate(errp, local_err); > + return; > + } > + qdev_set_parent_bus(DEVICE(phb), sysbus_get_default()); > + > + /* Populate the XSCOM address space. */ > + pnv_xscom_add_subregion(chip, > + PNV_XSCOM_PBCQ_NEST_BASE + 0x400 * phb->phb_id, > + &pbcq->xscom_nest_regs); > + pnv_xscom_add_subregion(chip, > + PNV_XSCOM_PBCQ_PCI_BASE + 0x400 * phb->phb_id, > + &pbcq->xscom_pci_regs); > + pnv_xscom_add_subregion(chip, > + PNV_XSCOM_PBCQ_SPCI_BASE + 0x040 * phb->phb_id, > + &pbcq->xscom_spci_regs); > + } > } > > static void pnv_chip_power8e_class_init(ObjectClass *klass, void *data) > @@ -1008,6 +1059,7 @@ static void pnv_chip_power8e_class_init(ObjectClass *klass, void *data) > k->chip_type = PNV_CHIP_POWER8E; > k->chip_cfam_id = 0x221ef04980000000ull; /* P8 Murano DD2.1 */ > k->cores_mask = POWER8E_CORE_MASK; > + k->num_phbs = 3; > k->core_pir = pnv_chip_core_pir_p8; > k->intc_create = pnv_chip_power8_intc_create; > k->isa_create = pnv_chip_power8_isa_create; > @@ -1027,6 +1079,7 @@ static void pnv_chip_power8_class_init(ObjectClass *klass, void *data) > k->chip_type = PNV_CHIP_POWER8; > k->chip_cfam_id = 0x220ea04980000000ull; /* P8 Venice DD2.0 */ > k->cores_mask = POWER8_CORE_MASK; > + k->num_phbs = 3; > k->core_pir = pnv_chip_core_pir_p8; > k->intc_create = pnv_chip_power8_intc_create; > k->isa_create = pnv_chip_power8_isa_create; > @@ -1046,6 +1099,7 @@ static void pnv_chip_power8nvl_class_init(ObjectClass *klass, void *data) > k->chip_type = PNV_CHIP_POWER8NVL; > k->chip_cfam_id = 0x120d304980000000ull; /* P8 Naples DD1.0 */ > k->cores_mask = POWER8_CORE_MASK; > + k->num_phbs = 3; > k->core_pir = pnv_chip_core_pir_p8; > k->intc_create = pnv_chip_power8_intc_create; > k->isa_create = pnv_chip_power8nvl_isa_create; > @@ -1215,6 +1269,7 @@ static void pnv_chip_power9_class_init(ObjectClass *klass, void *data) > k->chip_type = PNV_CHIP_POWER9; > k->chip_cfam_id = 0x220d104900008000ull; /* P9 Nimbus DD2.0 */ > k->cores_mask = POWER9_CORE_MASK; > + k->num_phbs = 0; > k->core_pir = pnv_chip_core_pir_p9; > k->intc_create = pnv_chip_power9_intc_create; > k->isa_create = pnv_chip_power9_isa_create; > @@ -1336,6 +1391,7 @@ static Property pnv_chip_properties[] = { > DEFINE_PROP_UINT64("ram-size", PnvChip, ram_size, 0), > DEFINE_PROP_UINT32("nr-cores", PnvChip, nr_cores, 1), > DEFINE_PROP_UINT64("cores-mask", PnvChip, cores_mask, 0x0), > + DEFINE_PROP_UINT32("num-phbs", PnvChip, num_phbs, 0), > DEFINE_PROP_END_OF_LIST(), > }; > > @@ -1352,14 +1408,23 @@ static void pnv_chip_class_init(ObjectClass *klass, void *data) > static ICSState *pnv_ics_get(XICSFabric *xi, int irq) > { > PnvMachineState *pnv = PNV_MACHINE(xi); > - int i; > + int i, j; > > for (i = 0; i < pnv->num_chips; i++) { > + PnvChip *chip = pnv->chips[i]; > Pnv8Chip *chip8 = PNV8_CHIP(pnv->chips[i]); > > if (ics_valid_irq(&chip8->psi.ics, irq)) { > return &chip8->psi.ics; > } > + for (j = 0; j < chip->num_phbs; j++) { > + if (ics_valid_irq(&chip8->phbs[j].lsis, irq)) { > + return &chip8->phbs[j].lsis; > + } > + if (ics_valid_irq(ICS_BASE(&chip8->phbs[j].msis), irq)) { > + return ICS_BASE(&chip8->phbs[j].msis); > + } > + } > } > return NULL; > } > @@ -1367,11 +1432,17 @@ static ICSState *pnv_ics_get(XICSFabric *xi, int irq) > static void pnv_ics_resend(XICSFabric *xi) > { > PnvMachineState *pnv = PNV_MACHINE(xi); > - int i; > + int i, j; > > for (i = 0; i < pnv->num_chips; i++) { > + PnvChip *chip = pnv->chips[i]; > Pnv8Chip *chip8 = PNV8_CHIP(pnv->chips[i]); > + > ics_resend(&chip8->psi.ics); > + for (j = 0; j < chip->num_phbs; j++) { > + ics_resend(&chip8->phbs[j].lsis); > + ics_resend(ICS_BASE(&chip8->phbs[j].msis)); > + } > } > } > > diff --git a/hw/ppc/pnv_xscom.c b/hw/ppc/pnv_xscom.c > index f01d788a6545..4d769371c0b5 100644 > --- a/hw/ppc/pnv_xscom.c > +++ b/hw/ppc/pnv_xscom.c > @@ -298,7 +298,10 @@ static int xscom_dt_child(Object *child, void *opaque) > PnvXScomInterface *xd = PNV_XSCOM_INTERFACE(child); > PnvXScomInterfaceClass *xc = PNV_XSCOM_INTERFACE_GET_CLASS(xd); > > - if (xc->dt_xscom) { > + /* > + * Only "realized" devices should be configured in the DT > + */ > + if (xc->dt_xscom && DEVICE(child)->realized) { > _FDT((xc->dt_xscom(xd, args->fdt, args->xscom_offset))); > } > } > @@ -345,7 +348,11 @@ int pnv_dt_xscom(PnvChip *chip, void *fdt, int root_offset) > args.fdt = fdt; > args.xscom_offset = xscom_offset; > > - object_child_foreach(OBJECT(chip), xscom_dt_child, &args); > + /* Some PnvXScomInterface objects lie a bit deeper (PnvPBCQState) > + * than the first layer, so we need to loop on the whole object > + * hierarchy to catch them > + */ > + object_child_foreach_recursive(OBJECT(chip), xscom_dt_child, &args); > return 0; > } > > diff --git a/hw/pci-host/Makefile.objs b/hw/pci-host/Makefile.objs > index a9cd3e022d53..4c6f825a8f06 100644 > --- a/hw/pci-host/Makefile.objs > +++ b/hw/pci-host/Makefile.objs > @@ -19,3 +19,4 @@ common-obj-$(CONFIG_PCI_EXPRESS_GENERIC_BRIDGE) += gpex.o > common-obj-$(CONFIG_PCI_EXPRESS_XILINX) += xilinx-pcie.o > > common-obj-$(CONFIG_PCI_EXPRESS_DESIGNWARE) += designware.o > +obj-$(CONFIG_POWERNV) += pnv_phb3.o pnv_phb3_pbcq.o pnv_phb3_msi.o > diff --git a/hw/ppc/Kconfig b/hw/ppc/Kconfig > index f927ec9c7487..3c2aa5b02a45 100644 > --- a/hw/ppc/Kconfig > +++ b/hw/ppc/Kconfig > @@ -26,6 +26,8 @@ config POWERNV > select MC146818RTC > select XICS > select XIVE > + select PCI > + select PCIE_PORT > > config PPC405 > bool -- David Gibson | I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_ | _way_ _around_! http://www.ozlabs.org/~dgibson [-- Attachment #2: signature.asc --] [-- Type: application/pgp-signature, Size: 833 bytes --] ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 0/4] xics: Eliminate unnecessary class 2019-09-24 4:59 [PATCH 0/4] xics: Eliminate unnecessary class David Gibson ` (4 preceding siblings ...) 2019-09-24 5:22 ` [PATCH 0/4] xics: Eliminate unnecessary class Cédric Le Goater @ 2019-09-24 9:47 ` Philippe Mathieu-Daudé 2019-09-24 10:06 ` Greg Kurz 5 siblings, 1 reply; 30+ messages in thread From: Philippe Mathieu-Daudé @ 2019-09-24 9:47 UTC (permalink / raw) To: David Gibson, qemu-ppc, clg; +Cc: gkurz, qemu-devel On 9/24/19 6:59 AM, David Gibson wrote: > The XICS interrupt controller device used to have separate subtypes > for the KVM and non-KVM variant of the device. That was a bad idea, > because it leaked information that should be entirely host-side > implementation specific to the kinda-sorta guest visible QOM class > names. > > We eliminated the KVM specific class some time ago, but it's left > behind a distinction between the TYPE_ICS_BASE abstract class and > TYPE_ICS_SIMPLE subtype which no longer serves any purpose. > > This series collapses the two types back into one. > > David Gibson (4): > xics: Eliminate 'reject', 'resend' and 'eoi' class hooks > xics: Merge reset and realize hooks > xics: Rename misleading ics_simple_*() functions > xics: Merge TYPE_ICS_BASE and TYPE_ICS_SIMPLE classes Please remove the qemu_register_reset() call in hw/intc/xics.c, then for the series: Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com> ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 0/4] xics: Eliminate unnecessary class 2019-09-24 9:47 ` Philippe Mathieu-Daudé @ 2019-09-24 10:06 ` Greg Kurz 2019-09-24 10:22 ` Philippe Mathieu-Daudé 0 siblings, 1 reply; 30+ messages in thread From: Greg Kurz @ 2019-09-24 10:06 UTC (permalink / raw) To: Philippe Mathieu-Daudé Cc: qemu-devel, gkurz, qemu-ppc, clg, David Gibson On Tue, 24 Sep 2019 11:47:51 +0200 Philippe Mathieu-Daudé <philmd@redhat.com> wrote: > On 9/24/19 6:59 AM, David Gibson wrote: > > The XICS interrupt controller device used to have separate subtypes > > for the KVM and non-KVM variant of the device. That was a bad idea, > > because it leaked information that should be entirely host-side > > implementation specific to the kinda-sorta guest visible QOM class > > names. > > > > We eliminated the KVM specific class some time ago, but it's left > > behind a distinction between the TYPE_ICS_BASE abstract class and > > TYPE_ICS_SIMPLE subtype which no longer serves any purpose. > > > > This series collapses the two types back into one. > > > > David Gibson (4): > > xics: Eliminate 'reject', 'resend' and 'eoi' class hooks > > xics: Merge reset and realize hooks > > xics: Rename misleading ics_simple_*() functions > > xics: Merge TYPE_ICS_BASE and TYPE_ICS_SIMPLE classes > > Please remove the qemu_register_reset() call in hw/intc/xics.c, No. This is needed because the XICS devices don't sit in a bus and dc->reset doesn't get called by anyone. > then for the series: > Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com> > ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 0/4] xics: Eliminate unnecessary class 2019-09-24 10:06 ` Greg Kurz @ 2019-09-24 10:22 ` Philippe Mathieu-Daudé 0 siblings, 0 replies; 30+ messages in thread From: Philippe Mathieu-Daudé @ 2019-09-24 10:22 UTC (permalink / raw) To: Greg Kurz; +Cc: qemu-devel, gkurz, qemu-ppc, clg, David Gibson On 9/24/19 12:06 PM, Greg Kurz wrote: > On Tue, 24 Sep 2019 11:47:51 +0200 > Philippe Mathieu-Daudé <philmd@redhat.com> wrote: > >> On 9/24/19 6:59 AM, David Gibson wrote: >>> The XICS interrupt controller device used to have separate subtypes >>> for the KVM and non-KVM variant of the device. That was a bad idea, >>> because it leaked information that should be entirely host-side >>> implementation specific to the kinda-sorta guest visible QOM class >>> names. >>> >>> We eliminated the KVM specific class some time ago, but it's left >>> behind a distinction between the TYPE_ICS_BASE abstract class and >>> TYPE_ICS_SIMPLE subtype which no longer serves any purpose. >>> >>> This series collapses the two types back into one. >>> >>> David Gibson (4): >>> xics: Eliminate 'reject', 'resend' and 'eoi' class hooks >>> xics: Merge reset and realize hooks >>> xics: Rename misleading ics_simple_*() functions >>> xics: Merge TYPE_ICS_BASE and TYPE_ICS_SIMPLE classes >> >> Please remove the qemu_register_reset() call in hw/intc/xics.c, > > No. This is needed because the XICS devices don't sit in a bus and > dc->reset doesn't get called by anyone. Oh right, now I remember a previous discussion. Please add a comment about this before the qemu_register_reset() call! >> then for the series: >> Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com> Reviewed-by stands :) ^ permalink raw reply [flat|nested] 30+ messages in thread
end of thread, other threads:[~2019-10-03 17:54 UTC | newest] Thread overview: 30+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2019-09-24 4:59 [PATCH 0/4] xics: Eliminate unnecessary class David Gibson 2019-09-24 4:59 ` [PATCH 1/4] xics: Eliminate 'reject', 'resend' and 'eoi' class hooks David Gibson 2019-09-24 5:23 ` Cédric Le Goater 2019-09-24 7:28 ` Greg Kurz 2019-09-24 4:59 ` [PATCH 2/4] xics: Merge reset and realize hooks David Gibson 2019-09-24 5:26 ` Cédric Le Goater 2019-09-24 7:36 ` Greg Kurz 2019-09-24 9:44 ` Philippe Mathieu-Daudé 2019-09-24 11:40 ` David Gibson 2019-09-24 4:59 ` [PATCH 3/4] xics: Rename misleading ics_simple_*() functions David Gibson 2019-09-24 5:26 ` Cédric Le Goater 2019-09-24 7:38 ` Greg Kurz 2019-09-24 4:59 ` [PATCH 4/4] xics: Merge TYPE_ICS_BASE and TYPE_ICS_SIMPLE classes David Gibson 2019-09-24 5:31 ` Cédric Le Goater 2019-09-24 11:41 ` David Gibson 2019-09-24 14:06 ` Cédric Le Goater 2019-09-25 1:46 ` David Gibson 2019-09-25 6:04 ` Cédric Le Goater 2019-10-03 17:53 ` Cédric Le Goater 2019-09-24 7:40 ` Greg Kurz 2019-09-24 9:46 ` Philippe Mathieu-Daudé 2019-09-24 5:22 ` [PATCH 0/4] xics: Eliminate unnecessary class Cédric Le Goater 2019-09-24 7:52 ` Greg Kurz 2019-09-24 9:55 ` Cédric Le Goater 2019-09-24 10:04 ` Philippe Mathieu-Daudé 2019-09-24 11:00 ` Cédric Le Goater 2019-09-26 1:28 ` David Gibson 2019-09-24 9:47 ` Philippe Mathieu-Daudé 2019-09-24 10:06 ` Greg Kurz 2019-09-24 10:22 ` Philippe Mathieu-Daudé
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.