| Message ID | 20140507100818.C714A196A9@mono.eik.bme.hu |
|---|---|
| State | New |
| Headers | show |
27.02.2014 05:05, BALATON Zoltan wrote: Shoudl we actually make it machine-specific, to keep even prog-if value of these things the same as before for older machine types? I dunno. mst says we should, I think this is not a very important property to keep. Thanks, /mjt > v2: resubmission after pc-2.1 is added with the multiport case > > hw/char/serial-pci.c | 2 ++ > 1 file changed, 2 insertions(+) > > diff --git a/hw/char/serial-pci.c b/hw/char/serial-pci.c > index 991c99f..f3f5e07 100644 > --- a/hw/char/serial-pci.c > +++ b/hw/char/serial-pci.c > @@ -60,6 +60,7 @@ static int serial_pci_init(PCIDevice *dev) > return -1; > } > > + pci->dev.config[PCI_CLASS_PROG] = 0x02; /* 16550 compatible */ > pci->dev.config[PCI_INTERRUPT_PIN] = 0x01; > s->irq = pci_allocate_irq(&pci->dev); > > @@ -101,6 +102,7 @@ static int multi_serial_pci_init(PCIDevice *dev) > assert(pci->ports > 0); > assert(pci->ports <= PCI_SERIAL_MAX_PORTS); > > + pci->dev.config[PCI_CLASS_PROG] = 0x02; /* 16550 compatible */ > pci->dev.config[PCI_INTERRUPT_PIN] = 0x01; > memory_region_init(&pci->iobar, OBJECT(pci), "multiserial", 8 * pci->ports); > pci_register_bar(&pci->dev, 0, PCI_BASE_ADDRESS_SPACE_IO, &pci->iobar); >
On Thu, 8 May 2014, Michael Tokarev wrote: > 27.02.2014 05:05, BALATON Zoltan wrote: > > Shoudl we actually make it machine-specific, to keep even prog-if value > of these things the same as before for older machine types? I dunno. > mst says we should, I think this is not a very important property to keep. So what's the decision? Can it be taken as it is now or do you want any more changes? (If you want it version specific then please tell me how to do that or show me an example because I don't know how to do that.) Regards, BALATON Zoltan > > Thanks, > > /mjt > >> v2: resubmission after pc-2.1 is added with the multiport case >> >> hw/char/serial-pci.c | 2 ++ >> 1 file changed, 2 insertions(+) >> >> diff --git a/hw/char/serial-pci.c b/hw/char/serial-pci.c >> index 991c99f..f3f5e07 100644 >> --- a/hw/char/serial-pci.c >> +++ b/hw/char/serial-pci.c >> @@ -60,6 +60,7 @@ static int serial_pci_init(PCIDevice *dev) >> return -1; >> } >> >> + pci->dev.config[PCI_CLASS_PROG] = 0x02; /* 16550 compatible */ >> pci->dev.config[PCI_INTERRUPT_PIN] = 0x01; >> s->irq = pci_allocate_irq(&pci->dev); >> >> @@ -101,6 +102,7 @@ static int multi_serial_pci_init(PCIDevice *dev) >> assert(pci->ports > 0); >> assert(pci->ports <= PCI_SERIAL_MAX_PORTS); >> >> + pci->dev.config[PCI_CLASS_PROG] = 0x02; /* 16550 compatible */ >> pci->dev.config[PCI_INTERRUPT_PIN] = 0x01; >> memory_region_init(&pci->iobar, OBJECT(pci), "multiserial", 8 * pci->ports); >> pci_register_bar(&pci->dev, 0, PCI_BASE_ADDRESS_SPACE_IO, &pci->iobar); >> > >
On Wed, May 14, 2014 at 01:39:03AM +0200, BALATON Zoltan wrote: > On Thu, 8 May 2014, Michael Tokarev wrote: > >27.02.2014 05:05, BALATON Zoltan wrote: > > > >Shoudl we actually make it machine-specific, to keep even prog-if value > >of these things the same as before for older machine types? I dunno. > >mst says we should, I think this is not a very important property to keep. > > So what's the decision? Can it be taken as it is now or do you want > any more changes? (If you want it version specific then please tell > me how to do that or show me an example because I don't know how to > do that.) > > Regards, > BALATON Zoltan If you change it migration to old qemu breaks. See commit aa93200b88fb1071eaf21bf766711762ed4630e2 Author: Gabriel L. Somlo <gsomlo@gmail.com> Date: Mon May 5 10:52:51 2014 -0400 apic: use emulated lapic version 0x14 on pc machines >= 2.1 as an example on how to do it. > > > >Thanks, > > > >/mjt > > > >> v2: resubmission after pc-2.1 is added with the multiport case > >> > >> hw/char/serial-pci.c | 2 ++ > >> 1 file changed, 2 insertions(+) > >> > >>diff --git a/hw/char/serial-pci.c b/hw/char/serial-pci.c > >>index 991c99f..f3f5e07 100644 > >>--- a/hw/char/serial-pci.c > >>+++ b/hw/char/serial-pci.c > >>@@ -60,6 +60,7 @@ static int serial_pci_init(PCIDevice *dev) > >> return -1; > >> } > >> > >>+ pci->dev.config[PCI_CLASS_PROG] = 0x02; /* 16550 compatible */ > >> pci->dev.config[PCI_INTERRUPT_PIN] = 0x01; > >> s->irq = pci_allocate_irq(&pci->dev); > >> > >>@@ -101,6 +102,7 @@ static int multi_serial_pci_init(PCIDevice *dev) > >> assert(pci->ports > 0); > >> assert(pci->ports <= PCI_SERIAL_MAX_PORTS); > >> > >>+ pci->dev.config[PCI_CLASS_PROG] = 0x02; /* 16550 compatible */ > >> pci->dev.config[PCI_INTERRUPT_PIN] = 0x01; > >> memory_region_init(&pci->iobar, OBJECT(pci), "multiserial", 8 * pci->ports); > >> pci_register_bar(&pci->dev, 0, PCI_BASE_ADDRESS_SPACE_IO, &pci->iobar); > >> > > > >
On Wed, 14 May 2014, Michael S. Tsirkin wrote: > On Wed, May 14, 2014 at 01:39:03AM +0200, BALATON Zoltan wrote: >> On Thu, 8 May 2014, Michael Tokarev wrote: >>> 27.02.2014 05:05, BALATON Zoltan wrote: >>> >>> Shoudl we actually make it machine-specific, to keep even prog-if value >>> of these things the same as before for older machine types? I dunno. >>> mst says we should, I think this is not a very important property to keep. >> >> So what's the decision? Can it be taken as it is now or do you want >> any more changes? (If you want it version specific then please tell >> me how to do that or show me an example because I don't know how to >> do that.) >> > > If you change it migration to old qemu breaks. This is unlikely as this pci serial card is not used by default in any machine version AFAIK. Is it still desired to keep backward compatibility for a device that can only be added by users explicitely? Regards, BALATON Zoltan > See > commit aa93200b88fb1071eaf21bf766711762ed4630e2 > Author: Gabriel L. Somlo <gsomlo@gmail.com> > Date: Mon May 5 10:52:51 2014 -0400 > > apic: use emulated lapic version 0x14 on pc machines >= 2.1 > as an example on how to do it. >>> >>> Thanks, >>> >>> /mjt >>> >>>> v2: resubmission after pc-2.1 is added with the multiport case >>>> >>>> hw/char/serial-pci.c | 2 ++ >>>> 1 file changed, 2 insertions(+) >>>> >>>> diff --git a/hw/char/serial-pci.c b/hw/char/serial-pci.c >>>> index 991c99f..f3f5e07 100644 >>>> --- a/hw/char/serial-pci.c >>>> +++ b/hw/char/serial-pci.c >>>> @@ -60,6 +60,7 @@ static int serial_pci_init(PCIDevice *dev) >>>> return -1; >>>> } >>>> >>>> + pci->dev.config[PCI_CLASS_PROG] = 0x02; /* 16550 compatible */ >>>> pci->dev.config[PCI_INTERRUPT_PIN] = 0x01; >>>> s->irq = pci_allocate_irq(&pci->dev); >>>> >>>> @@ -101,6 +102,7 @@ static int multi_serial_pci_init(PCIDevice *dev) >>>> assert(pci->ports > 0); >>>> assert(pci->ports <= PCI_SERIAL_MAX_PORTS); >>>> >>>> + pci->dev.config[PCI_CLASS_PROG] = 0x02; /* 16550 compatible */ >>>> pci->dev.config[PCI_INTERRUPT_PIN] = 0x01; >>>> memory_region_init(&pci->iobar, OBJECT(pci), "multiserial", 8 * pci->ports); >>>> pci_register_bar(&pci->dev, 0, PCI_BASE_ADDRESS_SPACE_IO, &pci->iobar); >>>> >>> >>> > >
Il 14/05/2014 12:59, BALATON Zoltan ha scritto: > On Wed, 14 May 2014, Michael S. Tsirkin wrote: >> On Wed, May 14, 2014 at 01:39:03AM +0200, BALATON Zoltan wrote: >>> On Thu, 8 May 2014, Michael Tokarev wrote: >>>> 27.02.2014 05:05, BALATON Zoltan wrote: >>>> >>>> Shoudl we actually make it machine-specific, to keep even prog-if value >>>> of these things the same as before for older machine types? I dunno. >>>> mst says we should, I think this is not a very important property to >>>> keep. >>> >>> So what's the decision? Can it be taken as it is now or do you want >>> any more changes? (If you want it version specific then please tell >>> me how to do that or show me an example because I don't know how to >>> do that.) >>> >> >> If you change it migration to old qemu breaks. > > This is unlikely as this pci serial card is not used by default in any > machine version AFAIK. Is it still desired to keep backward > compatibility for a device that can only be added by users explicitely? Yes, after all not even virtio-blk is used by default. Paolo
On Wed, May 14, 2014 at 12:59:15PM +0200, BALATON Zoltan wrote: > On Wed, 14 May 2014, Michael S. Tsirkin wrote: > >On Wed, May 14, 2014 at 01:39:03AM +0200, BALATON Zoltan wrote: > >>On Thu, 8 May 2014, Michael Tokarev wrote: > >>>27.02.2014 05:05, BALATON Zoltan wrote: > >>> > >>>Shoudl we actually make it machine-specific, to keep even prog-if value > >>>of these things the same as before for older machine types? I dunno. > >>>mst says we should, I think this is not a very important property to keep. > >> > >>So what's the decision? Can it be taken as it is now or do you want > >>any more changes? (If you want it version specific then please tell > >>me how to do that or show me an example because I don't know how to > >>do that.) > >> > > > >If you change it migration to old qemu breaks. > > This is unlikely as this pci serial card is not used by default in > any machine version AFAIK. Is it still desired to keep backward > compatibility for a device that can only be added by users > explicitely? > > Regards, > BALATON Zoltan Yes. > >See > >commit aa93200b88fb1071eaf21bf766711762ed4630e2 > >Author: Gabriel L. Somlo <gsomlo@gmail.com> > >Date: Mon May 5 10:52:51 2014 -0400 > > > > apic: use emulated lapic version 0x14 on pc machines >= 2.1 > >as an example on how to do it. > >>> > >>>Thanks, > >>> > >>>/mjt > >>> > >>>>v2: resubmission after pc-2.1 is added with the multiport case > >>>> > >>>>hw/char/serial-pci.c | 2 ++ > >>>>1 file changed, 2 insertions(+) > >>>> > >>>>diff --git a/hw/char/serial-pci.c b/hw/char/serial-pci.c > >>>>index 991c99f..f3f5e07 100644 > >>>>--- a/hw/char/serial-pci.c > >>>>+++ b/hw/char/serial-pci.c > >>>>@@ -60,6 +60,7 @@ static int serial_pci_init(PCIDevice *dev) > >>>> return -1; > >>>> } > >>>> > >>>>+ pci->dev.config[PCI_CLASS_PROG] = 0x02; /* 16550 compatible */ > >>>> pci->dev.config[PCI_INTERRUPT_PIN] = 0x01; > >>>> s->irq = pci_allocate_irq(&pci->dev); > >>>> > >>>>@@ -101,6 +102,7 @@ static int multi_serial_pci_init(PCIDevice *dev) > >>>> assert(pci->ports > 0); > >>>> assert(pci->ports <= PCI_SERIAL_MAX_PORTS); > >>>> > >>>>+ pci->dev.config[PCI_CLASS_PROG] = 0x02; /* 16550 compatible */ > >>>> pci->dev.config[PCI_INTERRUPT_PIN] = 0x01; > >>>> memory_region_init(&pci->iobar, OBJECT(pci), "multiserial", 8 * pci->ports); > >>>> pci_register_bar(&pci->dev, 0, PCI_BASE_ADDRESS_SPACE_IO, &pci->iobar); > >>>> > >>> > >>> > > > >
diff --git a/hw/char/serial-pci.c b/hw/char/serial-pci.c index 991c99f..f3f5e07 100644 --- a/hw/char/serial-pci.c +++ b/hw/char/serial-pci.c @@ -60,6 +60,7 @@ static int serial_pci_init(PCIDevice *dev) return -1; } + pci->dev.config[PCI_CLASS_PROG] = 0x02; /* 16550 compatible */ pci->dev.config[PCI_INTERRUPT_PIN] = 0x01; s->irq = pci_allocate_irq(&pci->dev); @@ -101,6 +102,7 @@ static int multi_serial_pci_init(PCIDevice *dev) assert(pci->ports > 0); assert(pci->ports <= PCI_SERIAL_MAX_PORTS); + pci->dev.config[PCI_CLASS_PROG] = 0x02; /* 16550 compatible */ pci->dev.config[PCI_INTERRUPT_PIN] = 0x01; memory_region_init(&pci->iobar, OBJECT(pci), "multiserial", 8 * pci->ports); pci_register_bar(&pci->dev, 0, PCI_BASE_ADDRESS_SPACE_IO, &pci->iobar);
Signed-off-by: BALATON Zoltan <balaton@eik.bme.hu> --- v2: resubmission after pc-2.1 is added with the multiport case hw/char/serial-pci.c | 2 ++ 1 file changed, 2 insertions(+)