| Message ID | 1574436472-6017-1-git-send-email-bmeng.cn@gmail.com |
|---|---|
| State | New |
| Headers | show |
| Series | riscv: virt: Allow PCI address 0 | expand |
On Fri, 22 Nov 2019 07:27:52 PST (-0800), bmeng.cn@gmail.com wrote: > When testing e1000 with the virt machine, e1000's I/O space cannot > be accessed. Debugging shows that the I/O BAR (BAR1) is correctly > written with address 0 plus I/O enable bit, but QEMU's "info pci" > shows that: > > Bus 0, device 1, function 0: > Ethernet controller: PCI device 8086:100e > ... > BAR1: I/O at 0xffffffffffffffff [0x003e]. > ... > > It turns out we should set pci_allow_0_address to true to allow 0 > PCI address, otherwise pci_bar_address() treats such address as > PCI_BAR_UNMAPPED. > > Signed-off-by: Bin Meng <bmeng.cn@gmail.com> > --- > > hw/riscv/virt.c | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/hw/riscv/virt.c b/hw/riscv/virt.c > index 23f340d..411bef5 100644 > --- a/hw/riscv/virt.c > +++ b/hw/riscv/virt.c > @@ -603,6 +603,7 @@ static void riscv_virt_machine_class_init(ObjectClass *oc, void *data) > mc->init = riscv_virt_board_init; > mc->max_cpus = 8; > mc->default_cpu_type = VIRT_CPU; > + mc->pci_allow_0_address = true; > } > > static const TypeInfo riscv_virt_machine_typeinfo = { Reviewed-by: Palmer Dabbelt <palmerdabbelt@google.com> I've put this on for-next, as I don't think this is 4.2 material. Thanks!
Hi Palmer, On Sat, Nov 23, 2019 at 6:41 AM Palmer Dabbelt <palmerdabbelt@google.com> wrote: > > On Fri, 22 Nov 2019 07:27:52 PST (-0800), bmeng.cn@gmail.com wrote: > > When testing e1000 with the virt machine, e1000's I/O space cannot > > be accessed. Debugging shows that the I/O BAR (BAR1) is correctly > > written with address 0 plus I/O enable bit, but QEMU's "info pci" > > shows that: > > > > Bus 0, device 1, function 0: > > Ethernet controller: PCI device 8086:100e > > ... > > BAR1: I/O at 0xffffffffffffffff [0x003e]. > > ... > > > > It turns out we should set pci_allow_0_address to true to allow 0 > > PCI address, otherwise pci_bar_address() treats such address as > > PCI_BAR_UNMAPPED. > > > > Signed-off-by: Bin Meng <bmeng.cn@gmail.com> > > --- > > > > hw/riscv/virt.c | 1 + > > 1 file changed, 1 insertion(+) > > > > diff --git a/hw/riscv/virt.c b/hw/riscv/virt.c > > index 23f340d..411bef5 100644 > > --- a/hw/riscv/virt.c > > +++ b/hw/riscv/virt.c > > @@ -603,6 +603,7 @@ static void riscv_virt_machine_class_init(ObjectClass *oc, void *data) > > mc->init = riscv_virt_board_init; > > mc->max_cpus = 8; > > mc->default_cpu_type = VIRT_CPU; > > + mc->pci_allow_0_address = true; > > } > > > > static const TypeInfo riscv_virt_machine_typeinfo = { > > Reviewed-by: Palmer Dabbelt <palmerdabbelt@google.com> > > I've put this on for-next, as I don't think this is 4.2 material. Looks you missed picking up this one :( Regards, Bin
On Sun, 02 Feb 2020 07:41:25 PST (-0800), bmeng.cn@gmail.com wrote: > Hi Palmer, > > On Sat, Nov 23, 2019 at 6:41 AM Palmer Dabbelt <palmerdabbelt@google.com> wrote: >> >> On Fri, 22 Nov 2019 07:27:52 PST (-0800), bmeng.cn@gmail.com wrote: >> > When testing e1000 with the virt machine, e1000's I/O space cannot >> > be accessed. Debugging shows that the I/O BAR (BAR1) is correctly >> > written with address 0 plus I/O enable bit, but QEMU's "info pci" >> > shows that: >> > >> > Bus 0, device 1, function 0: >> > Ethernet controller: PCI device 8086:100e >> > ... >> > BAR1: I/O at 0xffffffffffffffff [0x003e]. >> > ... >> > >> > It turns out we should set pci_allow_0_address to true to allow 0 >> > PCI address, otherwise pci_bar_address() treats such address as >> > PCI_BAR_UNMAPPED. >> > >> > Signed-off-by: Bin Meng <bmeng.cn@gmail.com> >> > --- >> > >> > hw/riscv/virt.c | 1 + >> > 1 file changed, 1 insertion(+) >> > >> > diff --git a/hw/riscv/virt.c b/hw/riscv/virt.c >> > index 23f340d..411bef5 100644 >> > --- a/hw/riscv/virt.c >> > +++ b/hw/riscv/virt.c >> > @@ -603,6 +603,7 @@ static void riscv_virt_machine_class_init(ObjectClass *oc, void *data) >> > mc->init = riscv_virt_board_init; >> > mc->max_cpus = 8; >> > mc->default_cpu_type = VIRT_CPU; >> > + mc->pci_allow_0_address = true; >> > } >> > >> > static const TypeInfo riscv_virt_machine_typeinfo = { >> >> Reviewed-by: Palmer Dabbelt <palmerdabbelt@google.com> >> >> I've put this on for-next, as I don't think this is 4.2 material. > > Looks you missed picking up this one :( Sorry about that, it's back on for-master.
diff --git a/hw/riscv/virt.c b/hw/riscv/virt.c index 23f340d..411bef5 100644 --- a/hw/riscv/virt.c +++ b/hw/riscv/virt.c @@ -603,6 +603,7 @@ static void riscv_virt_machine_class_init(ObjectClass *oc, void *data) mc->init = riscv_virt_board_init; mc->max_cpus = 8; mc->default_cpu_type = VIRT_CPU; + mc->pci_allow_0_address = true; } static const TypeInfo riscv_virt_machine_typeinfo = {
When testing e1000 with the virt machine, e1000's I/O space cannot be accessed. Debugging shows that the I/O BAR (BAR1) is correctly written with address 0 plus I/O enable bit, but QEMU's "info pci" shows that: Bus 0, device 1, function 0: Ethernet controller: PCI device 8086:100e ... BAR1: I/O at 0xffffffffffffffff [0x003e]. ... It turns out we should set pci_allow_0_address to true to allow 0 PCI address, otherwise pci_bar_address() treats such address as PCI_BAR_UNMAPPED. Signed-off-by: Bin Meng <bmeng.cn@gmail.com> --- hw/riscv/virt.c | 1 + 1 file changed, 1 insertion(+)