| Message ID | 20140819213304.19537.2834.stgit@joelaarch64.amd.com |
|---|---|
| State | New |
| Headers | show |
On 19 August 2014 22:33, Joel Schopp <joel.schopp@amd.com> wrote: > I'm running on a system with 8 cpus and it would be nice to have qemu > support all of them. The attached patch does that and has been tested. > > That said, I'm not sure if 8 is enough or if we want to bump this even higher > now before systems with many more cpus come along. 255 anyone? 8 is the limit for a GICv2 system. But yes, we could have 8 CPUs here; I think this was just copied over from the vexpress boards (which have 4 because that hardware has 4.) thanks -- PMM
On 08/19/2014 05:43 PM, Peter Maydell wrote: > On 19 August 2014 22:33, Joel Schopp <joel.schopp@amd.com> wrote: >> I'm running on a system with 8 cpus and it would be nice to have qemu >> support all of them. The attached patch does that and has been tested. >> >> That said, I'm not sure if 8 is enough or if we want to bump this even higher >> now before systems with many more cpus come along. 255 anyone? > 8 is the limit for a GICv2 system. But yes, we could have 8 CPUs > here; I think this was just copied over from the vexpress boards > (which have 4 because that hardware has 4.) So the question is do we want to wait for the GICv3 changes to be merged to bump this to 255? If we do want to wait then applying my patch that bumps it to 8 is the right thing to do now. -Joel
On 20 August 2014 16:08, Joel Schopp <joel.schopp@amd.com> wrote: > > On 08/19/2014 05:43 PM, Peter Maydell wrote: >> On 19 August 2014 22:33, Joel Schopp <joel.schopp@amd.com> wrote: >>> I'm running on a system with 8 cpus and it would be nice to have qemu >>> support all of them. The attached patch does that and has been tested. >>> >>> That said, I'm not sure if 8 is enough or if we want to bump this even higher >>> now before systems with many more cpus come along. 255 anyone? >> 8 is the limit for a GICv2 system. But yes, we could have 8 CPUs >> here; I think this was just copied over from the vexpress boards >> (which have 4 because that hardware has 4.) > So the question is do we want to wait for the GICv3 changes to be merged > to bump this to 255? If we do want to wait then applying my patch that > bumps it to 8 is the right thing to do now. I'm not currently aware of any GICv3 support patches for QEMU (I know there is some kernel work); it's the view the guest sees that matters, and for QEMU based guests that has to be v2 currently. So yes, we should bump to 8 for now. -- PMM
On 19 August 2014 22:33, Joel Schopp <joel.schopp@amd.com> wrote: > I'm running on a system with 8 cpus and it would be nice to have qemu > support all of them. The attached patch does that and has been tested. > > That said, I'm not sure if 8 is enough or if we want to bump this even higher > now before systems with many more cpus come along. 255 anyone? > > Cc: Peter Maydell <peter.maydell@linaro.org> > Signed-off-by: Joel Schopp <joel.schopp@amd.com> > --- > hw/arm/virt.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/hw/arm/virt.c b/hw/arm/virt.c > index 06f4fad..7a09260 100644 > --- a/hw/arm/virt.c > +++ b/hw/arm/virt.c > @@ -605,7 +605,7 @@ static QEMUMachine machvirt_a15_machine = { > .name = "virt", > .desc = "ARM Virtual Machine", > .init = machvirt_init, > - .max_cpus = 4, > + .max_cpus = 8, > .has_dynamic_sysbus = true, > }; Applied to target-arm.next, thanks. (I had to fix up a conflict because you seem to have based this patch on something other than upstream master -- .has_dynamic_sysbus doesn't exist there.) -- PMM
diff --git a/hw/arm/virt.c b/hw/arm/virt.c index 06f4fad..7a09260 100644 --- a/hw/arm/virt.c +++ b/hw/arm/virt.c @@ -605,7 +605,7 @@ static QEMUMachine machvirt_a15_machine = { .name = "virt", .desc = "ARM Virtual Machine", .init = machvirt_init, - .max_cpus = 4, + .max_cpus = 8, .has_dynamic_sysbus = true, };
I'm running on a system with 8 cpus and it would be nice to have qemu support all of them. The attached patch does that and has been tested. That said, I'm not sure if 8 is enough or if we want to bump this even higher now before systems with many more cpus come along. 255 anyone? Cc: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Joel Schopp <joel.schopp@amd.com> --- hw/arm/virt.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)