Segher Boessenkool wrote:
>> Split the Cell BE support into generic and platform
>> dependant parts.
>>
>> Creates a new config variable CONFIG_PPC_IBM_CELL_BLADE.
>> The existing CONFIG_PPC_CELL is now used to denote the
>> generic Cell processor support. Also renames spu_priv1.c
>> to spu_priv1_mmio.c.
>
>> config CELL_IIC
>> - depends on PPC_CELL
>> + depends on PPC_IBM_CELL_BLADE
>> bool
>> default y
>
> Erm no... Make this
>
> depends on PPC_CELL && !SOME_HYPERVISOR_THING
>
> to reflect the _real_ dependency please? Similar in a few other
> places perhaps.
Seems CELL_IIC is never used. Must be some stale variable,
so I removed it. Arnd, could you ack this.
Segher, a problem with your suggestion is that our
makefiles don't have as rich a set of logical ops as the
config files. Its easy to express 'build A if B', but not
so easy to do 'build A if not C'. To make this work
cleanly I made PPC_CELL denote !SOME_HYPERVISOR_THING,
so I can have constructions like this in the makefile:
obj-$(CONFIG_PPC_CELL) += ...
I also got rid of SPUFS_PRIV1_MMIO, since SPUFS_PRIV1_MMIO
just meant spufs with !SOME_HYPERVISOR_THING.
Updated patch follows.
-Geoff
Split the Cell BPA support into generic and platform
dependant parts.
Creates new config variable CONFIG_PPC_IBM_CELL_BLADE.
Also renames spu_priv1.c to spu_priv1_mmio.c.
Signed-off-by: Geoff Levand <[email protected]>
Index: cell--alp--2/arch/powerpc/Kconfig
===================================================================
--- cell--alp--2.orig/arch/powerpc/Kconfig 2006-05-01 15:13:22.000000000 -0700
+++ cell--alp--2/arch/powerpc/Kconfig 2006-05-01 15:16:38.000000000 -0700
@@ -391,15 +391,20 @@
For more informations, refer to <http://www.970eval.com>
config PPC_CELL
- bool " Cell Broadband Processor Architecture"
+ bool
+ default n
+
+config PPC_IBM_CELL_BLADE
+ bool " IBM Cell Blade"
depends on PPC_MULTIPLATFORM && PPC64
+ select PPC_CELL
select PPC_RTAS
select MMIO_NVRAM
select PPC_UDBG_16550
config PPC_SYSTEMSIM
bool " IBM Full System Simulator (systemsim) support"
- depends on PPC_CELL || PPC_PSERIES || PPC_MAPLE
+ depends on PPC_IBM_CELL_BLADE || PPC_PSERIES || PPC_MAPLE
help
Support booting resulting image under IBMs Full System Simulator.
If you enable this option, you are able to select device
@@ -450,11 +455,6 @@
depends on PPC_MAPLE
default y
-config CELL_IIC
- depends on PPC_CELL
- bool
- default y
-
config IBMVIO
depends on PPC_PSERIES || PPC_ISERIES
bool
Index: cell--alp--2/arch/powerpc/configs/cell_defconfig
===================================================================
--- cell--alp--2.orig/arch/powerpc/configs/cell_defconfig 2006-05-01 15:13:22.000000000 -0700
+++ cell--alp--2/arch/powerpc/configs/cell_defconfig 2006-05-01 15:13:23.000000000 -0700
@@ -116,6 +116,7 @@
# CONFIG_PPC_PMAC is not set
# CONFIG_PPC_MAPLE is not set
CONFIG_PPC_CELL=y
+CONFIG_PPC_IBM_CELL_BLADE=y
CONFIG_PPC_SYSTEMSIM=y
# CONFIG_U3_DART is not set
CONFIG_PPC_RTAS=y
@@ -123,7 +124,6 @@
CONFIG_RTAS_PROC=y
CONFIG_RTAS_FLASH=y
CONFIG_MMIO_NVRAM=y
-CONFIG_CELL_IIC=y
# CONFIG_PPC_MPC106 is not set
# CONFIG_CPU_FREQ is not set
# CONFIG_WANT_EARLY_SERIAL is not set
@@ -133,6 +133,7 @@
#
CONFIG_BE_DD2=y
CONFIG_SPU_FS=m
+CONFIG_SPU_BASE=y
CONFIG_SPUFS_MMAP=y
#
Index: cell--alp--2/arch/powerpc/configs/systemsim_defconfig
===================================================================
--- cell--alp--2.orig/arch/powerpc/configs/systemsim_defconfig 2006-05-01 15:13:22.000000000 -0700
+++ cell--alp--2/arch/powerpc/configs/systemsim_defconfig 2006-05-01 15:13:23.000000000 -0700
@@ -117,6 +117,7 @@
# CONFIG_PPC_PMAC is not set
CONFIG_PPC_MAPLE=y
CONFIG_PPC_CELL=y
+CONFIG_PPC_IBM_CELL_BLADE=y
CONFIG_PPC_SYSTEMSIM=y
CONFIG_SYSTEMSIM_IDLE=y
CONFIG_XICS=y
@@ -128,7 +129,6 @@
# CONFIG_RTAS_FLASH is not set
CONFIG_MMIO_NVRAM=y
CONFIG_MPIC_BROKEN_U3=y
-CONFIG_CELL_IIC=y
CONFIG_IBMVIO=y
# CONFIG_IBMEBUS is not set
# CONFIG_PPC_MPC106 is not set
@@ -139,6 +139,7 @@
# Cell Broadband Engine options
#
CONFIG_SPU_FS=m
+CONFIG_SPU_BASE=y
#
# Kernel options
Index: cell--alp--2/arch/powerpc/platforms/cell/Kconfig
===================================================================
--- cell--alp--2.orig/arch/powerpc/platforms/cell/Kconfig 2006-05-01 15:13:22.000000000 -0700
+++ cell--alp--2/arch/powerpc/platforms/cell/Kconfig 2006-05-01 15:13:23.000000000 -0700
@@ -11,10 +11,15 @@
or may crash other CPUs.
Say 'n' here unless you expect to run on DD2.0 only.
+config SPU_BASE
+ bool
+ default n
+
config SPU_FS
tristate "SPU file system"
default m
depends on PPC_CELL
+ select SPU_BASE
help
The SPU file system is used to access Synergistic Processing
Units on machines implementing the Broadband Processor
Index: cell--alp--2/arch/powerpc/platforms/cell/Makefile
===================================================================
--- cell--alp--2.orig/arch/powerpc/platforms/cell/Makefile 2006-05-01 15:13:22.000000000 -0700
+++ cell--alp--2/arch/powerpc/platforms/cell/Makefile 2006-05-01 15:17:58.000000000 -0700
@@ -1,14 +1,14 @@
-obj-y += interrupt.o iommu.o setup.o spider-pic.o
-obj-y += pervasive.o pci.o
-
-obj-$(CONFIG_SMP) += smp.o
+obj-$(CONFIG_PPC_CELL) += interrupt.o iommu.o setup.o \
+ spider-pic.o pervasive.o pci.o
+ifeq ($(CONFIG_SMP),y)
+obj-$(CONFIG_PPC_CELL) += smp.o
+endif
# needed only when building loadable spufs.ko
-spufs-modular-$(CONFIG_SPU_FS) += spu_syscalls.o
-obj-y += $(spufs-modular-m)
-
-# always needed in kernel
-spufs-builtin-$(CONFIG_SPU_FS) += spu_callbacks.o spu_base.o spu_priv1.o
-obj-y += $(spufs-builtin-y) $(spufs-builtin-m)
-
-obj-$(CONFIG_SPU_FS) += spufs/
+spufs-modular-$(CONFIG_SPU_FS) += spu_syscalls.o
+obj-$(CONFIG_SPU_BASE) += spu_callbacks.o spu_base.o \
+ $(spufs-modular-m)
+ifdef CONFIG_SPU_FS
+obj-$(CONFIG_PPC_CELL) += spu_priv1_mmio.o
+endif
+obj-$(CONFIG_SPU_FS) += spufs/
Index: cell--alp--2/arch/powerpc/platforms/cell/spu_priv1.c
===================================================================
--- cell--alp--2.orig/arch/powerpc/platforms/cell/spu_priv1.c 2006-05-01 15:13:22.000000000 -0700
+++ cell--alp--2/arch/powerpc/platforms/cell/spu_priv1.c 2006-04-21 13:04:41.284693750 -0700
@@ -1,142 +0,0 @@
-/*
- * access to SPU privileged registers
- */
-#include <linux/module.h>
-
-#include <asm/io.h>
-#include <asm/spu.h>
-
-void spu_int_mask_and(struct spu *spu, int class, u64 mask)
-{
- u64 old_mask;
-
- old_mask = in_be64(&spu->priv1->int_mask_RW[class]);
- out_be64(&spu->priv1->int_mask_RW[class], old_mask & mask);
-}
-EXPORT_SYMBOL_GPL(spu_int_mask_and);
-
-void spu_int_mask_or(struct spu *spu, int class, u64 mask)
-{
- u64 old_mask;
-
- old_mask = in_be64(&spu->priv1->int_mask_RW[class]);
- out_be64(&spu->priv1->int_mask_RW[class], old_mask | mask);
-}
-EXPORT_SYMBOL_GPL(spu_int_mask_or);
-
-void spu_int_mask_set(struct spu *spu, int class, u64 mask)
-{
- out_be64(&spu->priv1->int_mask_RW[class], mask);
-}
-EXPORT_SYMBOL_GPL(spu_int_mask_set);
-
-u64 spu_int_mask_get(struct spu *spu, int class)
-{
- return in_be64(&spu->priv1->int_mask_RW[class]);
-}
-EXPORT_SYMBOL_GPL(spu_int_mask_get);
-
-void spu_int_stat_clear(struct spu *spu, int class, u64 stat)
-{
- out_be64(&spu->priv1->int_stat_RW[class], stat);
-}
-EXPORT_SYMBOL_GPL(spu_int_stat_clear);
-
-u64 spu_int_stat_get(struct spu *spu, int class)
-{
- return in_be64(&spu->priv1->int_stat_RW[class]);
-}
-EXPORT_SYMBOL_GPL(spu_int_stat_get);
-
-void spu_int_route_set(struct spu *spu, u64 route)
-{
- out_be64(&spu->priv1->int_route_RW, route);
-}
-EXPORT_SYMBOL_GPL(spu_int_route_set);
-
-u64 spu_mfc_dar_get(struct spu *spu)
-{
- return in_be64(&spu->priv1->mfc_dar_RW);
-}
-EXPORT_SYMBOL_GPL(spu_mfc_dar_get);
-
-u64 spu_mfc_dsisr_get(struct spu *spu)
-{
- return in_be64(&spu->priv1->mfc_dsisr_RW);
-}
-EXPORT_SYMBOL_GPL(spu_mfc_dsisr_get);
-
-void spu_mfc_dsisr_set(struct spu *spu, u64 dsisr)
-{
- out_be64(&spu->priv1->mfc_dsisr_RW, dsisr);
-}
-EXPORT_SYMBOL_GPL(spu_mfc_dsisr_set);
-
-void spu_mfc_sdr_set(struct spu *spu, u64 sdr)
-{
- out_be64(&spu->priv1->mfc_sdr_RW, sdr);
-}
-EXPORT_SYMBOL_GPL(spu_mfc_sdr_set);
-
-void spu_mfc_sr1_set(struct spu *spu, u64 sr1)
-{
- out_be64(&spu->priv1->mfc_sr1_RW, sr1);
-}
-EXPORT_SYMBOL_GPL(spu_mfc_sr1_set);
-
-u64 spu_mfc_sr1_get(struct spu *spu)
-{
- return in_be64(&spu->priv1->mfc_sr1_RW);
-}
-EXPORT_SYMBOL_GPL(spu_mfc_sr1_get);
-
-void spu_mfc_tclass_id_set(struct spu *spu, u64 tclass_id)
-{
- out_be64(&spu->priv1->mfc_tclass_id_RW, tclass_id);
-}
-EXPORT_SYMBOL_GPL(spu_mfc_tclass_id_set);
-
-u64 spu_mfc_tclass_id_get(struct spu *spu)
-{
- return in_be64(&spu->priv1->mfc_tclass_id_RW);
-}
-EXPORT_SYMBOL_GPL(spu_mfc_tclass_id_get);
-
-void spu_smm_pgsz_set(struct spu *spu, u64 pgsz)
-{
- u64 smm_hid;
- smm_hid = in_be64(&spu->priv1->smm_hid);
- smm_hid &= ~(0xfull << 60);
- smm_hid |= pgsz << 60;
- out_be64(&spu->priv1->smm_hid, smm_hid);
-}
-
-void spu_tlb_invalidate(struct spu *spu)
-{
- out_be64(&spu->priv1->tlb_invalidate_entry_W, 0ul);
-}
-EXPORT_SYMBOL_GPL(spu_tlb_invalidate);
-
-void spu_resource_allocation_groupID_set(struct spu *spu, u64 id)
-{
- out_be64(&spu->priv1->resource_allocation_groupID_RW, id);
-}
-EXPORT_SYMBOL_GPL(spu_resource_allocation_groupID_set);
-
-u64 spu_resource_allocation_groupID_get(struct spu *spu)
-{
- return in_be64(&spu->priv1->resource_allocation_groupID_RW);
-}
-EXPORT_SYMBOL_GPL(spu_resource_allocation_groupID_get);
-
-void spu_resource_allocation_enable_set(struct spu *spu, u64 enable)
-{
- out_be64(&spu->priv1->resource_allocation_enable_RW, enable);
-}
-EXPORT_SYMBOL_GPL(spu_resource_allocation_enable_set);
-
-u64 spu_resource_allocation_enable_get(struct spu *spu)
-{
- return in_be64(&spu->priv1->resource_allocation_enable_RW);
-}
-EXPORT_SYMBOL_GPL(spu_resource_allocation_enable_get);
Index: cell--alp--2/arch/powerpc/platforms/cell/spu_priv1_mmio.c
===================================================================
--- cell--alp--2.orig/arch/powerpc/platforms/cell/spu_priv1_mmio.c 2006-04-21 13:04:41.284693750 -0700
+++ cell--alp--2/arch/powerpc/platforms/cell/spu_priv1_mmio.c 2006-05-01 15:16:49.000000000 -0700
@@ -0,0 +1,142 @@
+/*
+ * access to SPU privileged registers
+ */
+#include <linux/module.h>
+
+#include <asm/io.h>
+#include <asm/spu.h>
+
+void spu_int_mask_and(struct spu *spu, int class, u64 mask)
+{
+ u64 old_mask;
+
+ old_mask = in_be64(&spu->priv1->int_mask_RW[class]);
+ out_be64(&spu->priv1->int_mask_RW[class], old_mask & mask);
+}
+EXPORT_SYMBOL_GPL(spu_int_mask_and);
+
+void spu_int_mask_or(struct spu *spu, int class, u64 mask)
+{
+ u64 old_mask;
+
+ old_mask = in_be64(&spu->priv1->int_mask_RW[class]);
+ out_be64(&spu->priv1->int_mask_RW[class], old_mask | mask);
+}
+EXPORT_SYMBOL_GPL(spu_int_mask_or);
+
+void spu_int_mask_set(struct spu *spu, int class, u64 mask)
+{
+ out_be64(&spu->priv1->int_mask_RW[class], mask);
+}
+EXPORT_SYMBOL_GPL(spu_int_mask_set);
+
+u64 spu_int_mask_get(struct spu *spu, int class)
+{
+ return in_be64(&spu->priv1->int_mask_RW[class]);
+}
+EXPORT_SYMBOL_GPL(spu_int_mask_get);
+
+void spu_int_stat_clear(struct spu *spu, int class, u64 stat)
+{
+ out_be64(&spu->priv1->int_stat_RW[class], stat);
+}
+EXPORT_SYMBOL_GPL(spu_int_stat_clear);
+
+u64 spu_int_stat_get(struct spu *spu, int class)
+{
+ return in_be64(&spu->priv1->int_stat_RW[class]);
+}
+EXPORT_SYMBOL_GPL(spu_int_stat_get);
+
+void spu_int_route_set(struct spu *spu, u64 route)
+{
+ out_be64(&spu->priv1->int_route_RW, route);
+}
+EXPORT_SYMBOL_GPL(spu_int_route_set);
+
+u64 spu_mfc_dar_get(struct spu *spu)
+{
+ return in_be64(&spu->priv1->mfc_dar_RW);
+}
+EXPORT_SYMBOL_GPL(spu_mfc_dar_get);
+
+u64 spu_mfc_dsisr_get(struct spu *spu)
+{
+ return in_be64(&spu->priv1->mfc_dsisr_RW);
+}
+EXPORT_SYMBOL_GPL(spu_mfc_dsisr_get);
+
+void spu_mfc_dsisr_set(struct spu *spu, u64 dsisr)
+{
+ out_be64(&spu->priv1->mfc_dsisr_RW, dsisr);
+}
+EXPORT_SYMBOL_GPL(spu_mfc_dsisr_set);
+
+void spu_mfc_sdr_set(struct spu *spu, u64 sdr)
+{
+ out_be64(&spu->priv1->mfc_sdr_RW, sdr);
+}
+EXPORT_SYMBOL_GPL(spu_mfc_sdr_set);
+
+void spu_mfc_sr1_set(struct spu *spu, u64 sr1)
+{
+ out_be64(&spu->priv1->mfc_sr1_RW, sr1);
+}
+EXPORT_SYMBOL_GPL(spu_mfc_sr1_set);
+
+u64 spu_mfc_sr1_get(struct spu *spu)
+{
+ return in_be64(&spu->priv1->mfc_sr1_RW);
+}
+EXPORT_SYMBOL_GPL(spu_mfc_sr1_get);
+
+void spu_mfc_tclass_id_set(struct spu *spu, u64 tclass_id)
+{
+ out_be64(&spu->priv1->mfc_tclass_id_RW, tclass_id);
+}
+EXPORT_SYMBOL_GPL(spu_mfc_tclass_id_set);
+
+u64 spu_mfc_tclass_id_get(struct spu *spu)
+{
+ return in_be64(&spu->priv1->mfc_tclass_id_RW);
+}
+EXPORT_SYMBOL_GPL(spu_mfc_tclass_id_get);
+
+void spu_smm_pgsz_set(struct spu *spu, u64 pgsz)
+{
+ u64 smm_hid;
+ smm_hid = in_be64(&spu->priv1->smm_hid);
+ smm_hid &= ~(0xfull << 60);
+ smm_hid |= pgsz << 60;
+ out_be64(&spu->priv1->smm_hid, smm_hid);
+}
+
+void spu_tlb_invalidate(struct spu *spu)
+{
+ out_be64(&spu->priv1->tlb_invalidate_entry_W, 0ul);
+}
+EXPORT_SYMBOL_GPL(spu_tlb_invalidate);
+
+void spu_resource_allocation_groupID_set(struct spu *spu, u64 id)
+{
+ out_be64(&spu->priv1->resource_allocation_groupID_RW, id);
+}
+EXPORT_SYMBOL_GPL(spu_resource_allocation_groupID_set);
+
+u64 spu_resource_allocation_groupID_get(struct spu *spu)
+{
+ return in_be64(&spu->priv1->resource_allocation_groupID_RW);
+}
+EXPORT_SYMBOL_GPL(spu_resource_allocation_groupID_get);
+
+void spu_resource_allocation_enable_set(struct spu *spu, u64 enable)
+{
+ out_be64(&spu->priv1->resource_allocation_enable_RW, enable);
+}
+EXPORT_SYMBOL_GPL(spu_resource_allocation_enable_set);
+
+u64 spu_resource_allocation_enable_get(struct spu *spu)
+{
+ return in_be64(&spu->priv1->resource_allocation_enable_RW);
+}
+EXPORT_SYMBOL_GPL(spu_resource_allocation_enable_get);
Index: cell--alp--2/drivers/net/Kconfig
===================================================================
--- cell--alp--2.orig/drivers/net/Kconfig 2006-05-01 15:13:22.000000000 -0700
+++ cell--alp--2/drivers/net/Kconfig 2006-05-01 15:13:23.000000000 -0700
@@ -2179,7 +2179,7 @@
config SPIDER_NET
tristate "Spider Gigabit Ethernet driver"
- depends on PCI && PPC_CELL
+ depends on PCI && PPC_IBM_CELL_BLADE
select FW_LOADER
help
This driver supports the Gigabit Ethernet chips present on the
> Segher, a problem with your suggestion is that our
> makefiles don't have as rich a set of logical ops as the
> config files. Its easy to express 'build A if B', but not
> so easy to do 'build A if not C'. To make this work
> cleanly I made PPC_CELL denote !SOME_HYPERVISOR_THING,
> so I can have constructions like this in the makefile:
Not just that, but we can have a kernel image supporting both
the "raw" hardware _and_ stuff with a hypervisor underneath.
All CONFIG_<whatever> should always be used as a positive,
never a negative. My bad :-)
So it really should be
depends on PPC_CELL_NATIVE
or similar. Having PPC_CELL mean "native" / "raw" is not the
way to go, there will be many many hypervisors in the future,
it would be nice to have PPC_CELL mean just that, "support for
the Cell architecture" in general, kernels running on various
hypervisors will see the hardware virtualised to varying degrees.
Segher
On Tuesday 02 May 2006 01:09, Segher Boessenkool wrote:
> So it really should be
>
> ????????depends on PPC_CELL_NATIVE
>
> or similar. ?Having PPC_CELL mean "native" / "raw" is not the
> way to go, there will be many many hypervisors in the future,
> it would be nice to have PPC_CELL mean just that, "support for
> the Cell architecture" in general, kernels running on various
> hypervisors will see the hardware virtualised to varying degrees.
>
Yes, good point.
Arnd <><
On Tuesday 02 May 2006 00:51, Geoff Levand wrote:
> Seems CELL_IIC is never used. Must be some stale variable,
> so I removed it. Arnd, could you ack this.
Yes, I used it before when there was more code shared between
Cell and pSeries, but that is no longer the case.
> Segher, a problem with your suggestion is that our
> makefiles don't have as rich a set of logical ops as the
> config files. Its easy to express 'build A if B', but not
> so easy to do 'build A if not C'. To make this work
> cleanly I made PPC_CELL denote !SOME_HYPERVISOR_THING,
> so I can have constructions like this in the makefile:
>
> obj-$(CONFIG_PPC_CELL) += ...
>
> I also got rid of SPUFS_PRIV1_MMIO, since SPUFS_PRIV1_MMIO
> just meant spufs with !SOME_HYPERVISOR_THING.
>
I guess that one should really be (SPU_FS && CELL_NATIVE),
using the option Segher suggested now.
> ===================================================================
> --- cell--alp--2.orig/arch/powerpc/Kconfig 2006-05-01 15:13:22.000000000 -0700
> +++ cell--alp--2/arch/powerpc/Kconfig 2006-05-01 15:16:38.000000000 -0700
> @@ -391,15 +391,20 @@
> For more informations, refer to <http://www.970eval.com>
>
> config PPC_CELL
> - bool " Cell Broadband Processor Architecture"
> + bool
> + default n
> +
> +config PPC_IBM_CELL_BLADE
> + bool " IBM Cell Blade"
> depends on PPC_MULTIPLATFORM && PPC64
> + select PPC_CELL
> select PPC_RTAS
> select MMIO_NVRAM
> select PPC_UDBG_16550
>
> config PPC_SYSTEMSIM
> bool " IBM Full System Simulator (systemsim) support"
> - depends on PPC_CELL || PPC_PSERIES || PPC_MAPLE
> + depends on PPC_IBM_CELL_BLADE || PPC_PSERIES || PPC_MAPLE
> help
> Support booting resulting image under IBMs Full System Simulator.
> If you enable this option, you are able to select device
whoops, this one should not be there at all. Note that I updated
your previous patch as well to fit into the series for submission,
and that did not include systemsim.
> Index: cell--alp--2/arch/powerpc/platforms/cell/Kconfig
> ===================================================================
> --- cell--alp--2.orig/arch/powerpc/platforms/cell/Kconfig 2006-05-01 15:13:22.000000000 -0700
> +++ cell--alp--2/arch/powerpc/platforms/cell/Kconfig 2006-05-01 15:13:23.000000000 -0700
> @@ -11,10 +11,15 @@
> or may crash other CPUs.
> Say 'n' here unless you expect to run on DD2.0 only.
>
> +config SPU_BASE
> + bool
> + default n
> +
> config SPU_FS
> tristate "SPU file system"
> default m
> depends on PPC_CELL
> + select SPU_BASE
> help
> The SPU file system is used to access Synergistic Processing
> Units on machines implementing the Broadband Processor
> Index: cell--alp--2/arch/powerpc/platforms/cell/Makefile
> ===================================================================
> --- cell--alp--2.orig/arch/powerpc/platforms/cell/Makefile 2006-05-01 15:13:22.000000000 -0700
> +++ cell--alp--2/arch/powerpc/platforms/cell/Makefile 2006-05-01 15:17:58.000000000 -0700
> @@ -1,14 +1,14 @@
> -obj-y += interrupt.o iommu.o setup.o spider-pic.o
> -obj-y += pervasive.o pci.o
> -
> -obj-$(CONFIG_SMP) += smp.o
> +obj-$(CONFIG_PPC_CELL) += interrupt.o iommu.o setup.o \
> + spider-pic.o pervasive.o pci.o
> +ifeq ($(CONFIG_SMP),y)
> +obj-$(CONFIG_PPC_CELL) += smp.o
> +endif
>
> # needed only when building loadable spufs.ko
> -spufs-modular-$(CONFIG_SPU_FS) += spu_syscalls.o
> -obj-y += $(spufs-modular-m)
> -
> -# always needed in kernel
> -spufs-builtin-$(CONFIG_SPU_FS) += spu_callbacks.o spu_base.o spu_priv1.o
> -obj-y += $(spufs-builtin-y) $(spufs-builtin-m)
> -
> -obj-$(CONFIG_SPU_FS) += spufs/
> +spufs-modular-$(CONFIG_SPU_FS) += spu_syscalls.o
> +obj-$(CONFIG_SPU_BASE) += spu_callbacks.o spu_base.o \
> + $(spufs-modular-m)
> +ifdef CONFIG_SPU_FS
> +obj-$(CONFIG_PPC_CELL) += spu_priv1_mmio.o
> +endif
I guess this could then become something like
spu-priv1-$(CONFIG_PPC_CELL_NATIVE) += spu_priv1_mmio.o
spufs-modular-$(CONFIG_SPU_FS) += spu_syscalls.o
obj-$(CONFIG_SPU_BASE) += spu_callbacks.o spu_base.o \
$(spufs-modular-m) \
$(spu-priv1-y)
> Index: cell--alp--2/drivers/net/Kconfig
> ===================================================================
> --- cell--alp--2.orig/drivers/net/Kconfig 2006-05-01 15:13:22.000000000 -0700
> +++ cell--alp--2/drivers/net/Kconfig 2006-05-01 15:13:23.000000000 -0700
> @@ -2179,7 +2179,7 @@
>
> config SPIDER_NET
> tristate "Spider Gigabit Ethernet driver"
> - depends on PCI && PPC_CELL
> + depends on PCI && PPC_IBM_CELL_BLADE
> select FW_LOADER
> help
> This driver supports the Gigabit Ethernet chips present on the
Hmm, I'm also no longer sure if this is right. In theory, spidernet
could be used in all sorts of products, wether they are using the
same bridge chip or just the gigabit ethernet macro from it.
For now, I guess you can just leave this one alone if you respin
the patch another time. It's disabled by default, so the dependency
can be updated the next time we get a user in _addition_ to PPC_CELL.
Arnd <><
>> config SPIDER_NET
>> tristate "Spider Gigabit Ethernet driver"
>> - depends on PCI && PPC_CELL
>> + depends on PCI && PPC_IBM_CELL_BLADE
>> select FW_LOADER
>> help
>> This driver supports the Gigabit Ethernet chips present on the
>
> Hmm, I'm also no longer sure if this is right. In theory, spidernet
> could be used in all sorts of products, wether they are using the
> same bridge chip or just the gigabit ethernet macro from it.
>
> For now, I guess you can just leave this one alone if you respin
> the patch another time. It's disabled by default, so the dependency
> can be updated the next time we get a user in _addition_ to PPC_CELL.
Is there any reason the driver wouldn't build and/or run on other
platforms? If so, fix it. If not, just make it
depends on PCI
?
Segher
On Mon, 2006-05-01 at 15:51 -0700, Geoff Levand wrote:
> Segher, a problem with your suggestion is that our
> makefiles don't have as rich a set of logical ops as the
> config files. Its easy to express 'build A if B', but not
> so easy to do 'build A if not C'. To make this work
> cleanly I made PPC_CELL denote !SOME_HYPERVISOR_THING,
> so I can have constructions like this in the makefile:
>
> obj-$(CONFIG_PPC_CELL) += ...
Hi Geoff,
I've been ignoring this discussion, but now that I read it I think this
is all kinda backwards.
PPC_CELL should not denote !SOME_HYPERVISOR, it should just mean "basic
cell support", ie. PPC_CELL gets you platforms/cell built in.
Then we can have SOME_HYPERVISOR which _adds_ support for that
hypervisor. And PPC_CELL_BLADE which selects things which are actually
specific to that hardware, like SPIDERNET etc.
But SOME_HYPERVISOR should not remove support for running on bare metal,
it should just give you the option of running on the hypervisor. Yes
that may require testing things at runtime, that's what
firmware_has_feature() is for.
The goal should be that we have one kernel which can boot on all Cell
implementations. In fact the ultimate goal is to have one kernel that
can boot any platform under powerpc, that's a way off still, but we
don't want to start going backwards.
cheers
--
Michael Ellerman
IBM OzLabs
wwweb: http://michael.ellerman.id.au
phone: +61 2 6212 1183 (tie line 70 21183)
We do not inherit the earth from our ancestors,
we borrow it from our children. - S.M.A.R.T Person
On Tuesday 02 May 2006 02:06, Segher Boessenkool wrote:
> Is there any reason the driver wouldn't build and/or run on other
> platforms? ?If so, fix it. ?If not, just make it
>
> ????????depends on PCI
Well, it could run on other platforms, except:
- it requires a few properties in the device tree (local-mac-address,
firmware), so it should also depend on PPC
- It's not actually PCI at all, but on an internal bus that has
something close enough to a PCI config space.
Arnd <><
On Mon, May 01, 2006 at 03:51:35PM -0700, Geoff Levand wrote:
> makefiles don't have as rich a set of logical ops as the
> config files. Its easy to express 'build A if B', but not
> so easy to do 'build A if not C'. To make this work
> cleanly I made PPC_CELL denote !SOME_HYPERVISOR_THING,
> so I can have constructions like this in the makefile:
>
> obj-$(CONFIG_PPC_CELL) += ...
>
> I also got rid of SPUFS_PRIV1_MMIO, since SPUFS_PRIV1_MMIO
> just meant spufs with !SOME_HYPERVISOR_THING.
The Kconfig files has lots of nice operators.
Anyway, until we actually have the cell common hypervisor interface
out in the public adding any support for it in Kconfig or the
Makefile is completely pointless, and a patch like this even when
done correctly shouldn't go in.
> Split the Cell BPA support into generic and platform
> dependant parts.
>
> Creates new config variable CONFIG_PPC_IBM_CELL_BLADE.
That's wrong. Most of these files will be needed to support
the PS3 when running on bare hardware. The right option will
be CELL_SPIDER_HARDWARE, which is right now the only cell
hardware generation supported. It'll become meaningfull when
adding axon support.
Arnd Bergmann wrote:
>> I also got rid of SPUFS_PRIV1_MMIO, since SPUFS_PRIV1_MMIO
>> just meant spufs with !SOME_HYPERVISOR_THING.
>>
>
> I guess that one should really be (SPU_FS && CELL_NATIVE),
> using the option Segher suggested now.
OK. I set it up that way. Updated patch below.
>> config PPC_SYSTEMSIM
>> bool " IBM Full System Simulator (systemsim) support"
>> - depends on PPC_CELL || PPC_PSERIES || PPC_MAPLE
>> + depends on PPC_IBM_CELL_BLADE || PPC_PSERIES || PPC_MAPLE
>> help
>> Support booting resulting image under IBMs Full System Simulator.
>> If you enable this option, you are able to select device
>
> whoops, this one should not be there at all. Note that I updated
> your previous patch as well to fit into the series for submission,
> and that did not include systemsim.
Sorry, didn't notice you changed it. Fixed now.
>> # needed only when building loadable spufs.ko
>> -spufs-modular-$(CONFIG_SPU_FS) += spu_syscalls.o
>> -obj-y += $(spufs-modular-m)
>> -
>> -# always needed in kernel
>> -spufs-builtin-$(CONFIG_SPU_FS) += spu_callbacks.o spu_base.o spu_priv1.o
>> -obj-y += $(spufs-builtin-y) $(spufs-builtin-m)
>> -
>> -obj-$(CONFIG_SPU_FS) += spufs/
>> +spufs-modular-$(CONFIG_SPU_FS) += spu_syscalls.o
>> +obj-$(CONFIG_SPU_BASE) += spu_callbacks.o spu_base.o \
>> + $(spufs-modular-m)
>> +ifdef CONFIG_SPU_FS
>> +obj-$(CONFIG_PPC_CELL) += spu_priv1_mmio.o
>> +endif
>
> I guess this could then become something like
>
> spu-priv1-$(CONFIG_PPC_CELL_NATIVE) += spu_priv1_mmio.o
> spufs-modular-$(CONFIG_SPU_FS) += spu_syscalls.o
> obj-$(CONFIG_SPU_BASE) += spu_callbacks.o spu_base.o \
> $(spufs-modular-m) \
> $(spu-priv1-y)
Yes, a nice way.
>> --- cell--alp--2.orig/drivers/net/Kconfig 2006-05-01 15:13:22.000000000 -0700
>> +++ cell--alp--2/drivers/net/Kconfig 2006-05-01 15:13:23.000000000 -0700
>> @@ -2179,7 +2179,7 @@
>>
>> config SPIDER_NET
>> tristate "Spider Gigabit Ethernet driver"
>> - depends on PCI && PPC_CELL
>> + depends on PCI && PPC_IBM_CELL_BLADE
>> select FW_LOADER
>> help
>> This driver supports the Gigabit Ethernet chips present on the
>
> Hmm, I'm also no longer sure if this is right. In theory, spidernet
> could be used in all sorts of products, wether they are using the
> same bridge chip or just the gigabit ethernet macro from it.
>
> For now, I guess you can just leave this one alone if you respin
> the patch another time. It's disabled by default, so the dependency
> can be updated the next time we get a user in _addition_ to PPC_CELL.
OK, based on your other mail I just left it as PCI && PPC_IBM_CELL_BLADE.
We can change it when another system uses it.
-Geoff
Split the Cell BE support into generic and platform dependent parts.
Creates new config variables PPC_CELL_NATIVE and PPC_IBM_CELL_BLADE.
The existing CONFIG_PPC_CELL is now used to denote the generic
Cell processor support.
Also renames spu_priv1.c to spu_priv1_mmio.c.
Signed-off-by: Geoff Levand <[email protected]>
Index: cell--alp--3/arch/powerpc/Kconfig
===================================================================
--- cell--alp--3.orig/arch/powerpc/Kconfig 2006-05-02 10:10:52.000000000 -0700
+++ cell--alp--3/arch/powerpc/Kconfig 2006-05-02 10:11:42.000000000 -0700
@@ -391,8 +391,18 @@
For more informations, refer to <http://www.970eval.com>
config PPC_CELL
- bool " Cell Broadband Processor Architecture"
+ bool
+ default n
+
+config PPC_CELL_NATIVE
+ bool
+ select PPC_CELL
+ default n
+
+config PPC_IBM_CELL_BLADE
+ bool " IBM Cell Blade"
depends on PPC_MULTIPLATFORM && PPC64
+ select PPC_CELL_NATIVE
select PPC_RTAS
select MMIO_NVRAM
select PPC_UDBG_16550
@@ -439,11 +449,6 @@
depends on PPC_MAPLE
default y
-config CELL_IIC
- depends on PPC_CELL
- bool
- default y
-
config IBMVIO
depends on PPC_PSERIES || PPC_ISERIES
bool
Index: cell--alp--3/arch/powerpc/configs/cell_defconfig
===================================================================
--- cell--alp--3.orig/arch/powerpc/configs/cell_defconfig 2006-05-02 10:10:52.000000000 -0700
+++ cell--alp--3/arch/powerpc/configs/cell_defconfig 2006-05-02 10:12:22.000000000 -0700
@@ -118,13 +118,14 @@
# CONFIG_PPC_PMAC is not set
# CONFIG_PPC_MAPLE is not set
CONFIG_PPC_CELL=y
+CONFIG_PPC_CELL_NATIVE=y
+CONFIG_PPC_IBM_CELL_BLADE=y
# CONFIG_U3_DART is not set
CONFIG_PPC_RTAS=y
# CONFIG_RTAS_ERROR_LOGGING is not set
CONFIG_RTAS_PROC=y
CONFIG_RTAS_FLASH=y
CONFIG_MMIO_NVRAM=y
-CONFIG_CELL_IIC=y
# CONFIG_PPC_MPC106 is not set
# CONFIG_CPU_FREQ is not set
# CONFIG_WANT_EARLY_SERIAL is not set
@@ -133,6 +134,7 @@
# Cell Broadband Engine options
#
CONFIG_SPU_FS=m
+CONFIG_SPU_BASE=y
CONFIG_SPUFS_MMAP=y
#
Index: cell--alp--3/arch/powerpc/platforms/cell/Kconfig
===================================================================
--- cell--alp--3.orig/arch/powerpc/platforms/cell/Kconfig 2006-05-02 10:10:52.000000000 -0700
+++ cell--alp--3/arch/powerpc/platforms/cell/Kconfig 2006-05-02 10:14:05.000000000 -0700
@@ -5,11 +5,16 @@
tristate "SPU file system"
default m
depends on PPC_CELL
+ select SPU_BASE
help
The SPU file system is used to access Synergistic Processing
Units on machines implementing the Broadband Processor
Architecture.
+config SPU_BASE
+ bool
+ default n
+
config SPUFS_MMAP
bool
depends on SPU_FS && SPARSEMEM && !PPC_64K_PAGES
Index: cell--alp--3/arch/powerpc/platforms/cell/Makefile
===================================================================
--- cell--alp--3.orig/arch/powerpc/platforms/cell/Makefile 2006-05-02 10:10:52.000000000 -0700
+++ cell--alp--3/arch/powerpc/platforms/cell/Makefile 2006-05-02 10:39:27.000000000 -0700
@@ -1,14 +1,14 @@
-obj-y += interrupt.o iommu.o setup.o spider-pic.o
-obj-y += pervasive.o
+obj-$(CONFIG_PPC_CELL_NATIVE) += interrupt.o iommu.o setup.o \
+ spider-pic.o pervasive.o
-obj-$(CONFIG_SMP) += smp.o
+ifeq ($(CONFIG_SMP),y)
+obj-$(CONFIG_PPC_CELL_NATIVE) += smp.o
+endif
# needed only when building loadable spufs.ko
-spufs-modular-$(CONFIG_SPU_FS) += spu_syscalls.o
-obj-y += $(spufs-modular-m)
+spufs-modular-$(CONFIG_SPU_FS) += spu_syscalls.o
+spu-priv1-$(CONFIG_PPC_CELL_NATIVE) += spu_priv1_mmio.o
-# always needed in kernel
-spufs-builtin-$(CONFIG_SPU_FS) += spu_callbacks.o spu_base.o spu_priv1.o
-obj-y += $(spufs-builtin-y) $(spufs-builtin-m)
-
-obj-$(CONFIG_SPU_FS) += spufs/
+obj-$(CONFIG_SPU_BASE) += spu_callbacks.o spu_base.o \
+ $(spufs-modular-m) $(spu-priv1-y)
+obj-$(CONFIG_SPU_FS) += spufs/
Index: cell--alp--3/arch/powerpc/platforms/cell/spu_priv1.c
===================================================================
--- cell--alp--3.orig/arch/powerpc/platforms/cell/spu_priv1.c 2006-05-02 10:10:52.000000000 -0700
+++ cell--alp--3/arch/powerpc/platforms/cell/spu_priv1.c 2006-05-01 17:06:59.032678000 -0700
@@ -1,133 +0,0 @@
-/*
- * access to SPU privileged registers
- */
-#include <linux/module.h>
-
-#include <asm/io.h>
-#include <asm/spu.h>
-
-void spu_int_mask_and(struct spu *spu, int class, u64 mask)
-{
- u64 old_mask;
-
- old_mask = in_be64(&spu->priv1->int_mask_RW[class]);
- out_be64(&spu->priv1->int_mask_RW[class], old_mask & mask);
-}
-EXPORT_SYMBOL_GPL(spu_int_mask_and);
-
-void spu_int_mask_or(struct spu *spu, int class, u64 mask)
-{
- u64 old_mask;
-
- old_mask = in_be64(&spu->priv1->int_mask_RW[class]);
- out_be64(&spu->priv1->int_mask_RW[class], old_mask | mask);
-}
-EXPORT_SYMBOL_GPL(spu_int_mask_or);
-
-void spu_int_mask_set(struct spu *spu, int class, u64 mask)
-{
- out_be64(&spu->priv1->int_mask_RW[class], mask);
-}
-EXPORT_SYMBOL_GPL(spu_int_mask_set);
-
-u64 spu_int_mask_get(struct spu *spu, int class)
-{
- return in_be64(&spu->priv1->int_mask_RW[class]);
-}
-EXPORT_SYMBOL_GPL(spu_int_mask_get);
-
-void spu_int_stat_clear(struct spu *spu, int class, u64 stat)
-{
- out_be64(&spu->priv1->int_stat_RW[class], stat);
-}
-EXPORT_SYMBOL_GPL(spu_int_stat_clear);
-
-u64 spu_int_stat_get(struct spu *spu, int class)
-{
- return in_be64(&spu->priv1->int_stat_RW[class]);
-}
-EXPORT_SYMBOL_GPL(spu_int_stat_get);
-
-void spu_int_route_set(struct spu *spu, u64 route)
-{
- out_be64(&spu->priv1->int_route_RW, route);
-}
-EXPORT_SYMBOL_GPL(spu_int_route_set);
-
-u64 spu_mfc_dar_get(struct spu *spu)
-{
- return in_be64(&spu->priv1->mfc_dar_RW);
-}
-EXPORT_SYMBOL_GPL(spu_mfc_dar_get);
-
-u64 spu_mfc_dsisr_get(struct spu *spu)
-{
- return in_be64(&spu->priv1->mfc_dsisr_RW);
-}
-EXPORT_SYMBOL_GPL(spu_mfc_dsisr_get);
-
-void spu_mfc_dsisr_set(struct spu *spu, u64 dsisr)
-{
- out_be64(&spu->priv1->mfc_dsisr_RW, dsisr);
-}
-EXPORT_SYMBOL_GPL(spu_mfc_dsisr_set);
-
-void spu_mfc_sdr_set(struct spu *spu, u64 sdr)
-{
- out_be64(&spu->priv1->mfc_sdr_RW, sdr);
-}
-EXPORT_SYMBOL_GPL(spu_mfc_sdr_set);
-
-void spu_mfc_sr1_set(struct spu *spu, u64 sr1)
-{
- out_be64(&spu->priv1->mfc_sr1_RW, sr1);
-}
-EXPORT_SYMBOL_GPL(spu_mfc_sr1_set);
-
-u64 spu_mfc_sr1_get(struct spu *spu)
-{
- return in_be64(&spu->priv1->mfc_sr1_RW);
-}
-EXPORT_SYMBOL_GPL(spu_mfc_sr1_get);
-
-void spu_mfc_tclass_id_set(struct spu *spu, u64 tclass_id)
-{
- out_be64(&spu->priv1->mfc_tclass_id_RW, tclass_id);
-}
-EXPORT_SYMBOL_GPL(spu_mfc_tclass_id_set);
-
-u64 spu_mfc_tclass_id_get(struct spu *spu)
-{
- return in_be64(&spu->priv1->mfc_tclass_id_RW);
-}
-EXPORT_SYMBOL_GPL(spu_mfc_tclass_id_get);
-
-void spu_tlb_invalidate(struct spu *spu)
-{
- out_be64(&spu->priv1->tlb_invalidate_entry_W, 0ul);
-}
-EXPORT_SYMBOL_GPL(spu_tlb_invalidate);
-
-void spu_resource_allocation_groupID_set(struct spu *spu, u64 id)
-{
- out_be64(&spu->priv1->resource_allocation_groupID_RW, id);
-}
-EXPORT_SYMBOL_GPL(spu_resource_allocation_groupID_set);
-
-u64 spu_resource_allocation_groupID_get(struct spu *spu)
-{
- return in_be64(&spu->priv1->resource_allocation_groupID_RW);
-}
-EXPORT_SYMBOL_GPL(spu_resource_allocation_groupID_get);
-
-void spu_resource_allocation_enable_set(struct spu *spu, u64 enable)
-{
- out_be64(&spu->priv1->resource_allocation_enable_RW, enable);
-}
-EXPORT_SYMBOL_GPL(spu_resource_allocation_enable_set);
-
-u64 spu_resource_allocation_enable_get(struct spu *spu)
-{
- return in_be64(&spu->priv1->resource_allocation_enable_RW);
-}
-EXPORT_SYMBOL_GPL(spu_resource_allocation_enable_get);
Index: cell--alp--3/arch/powerpc/platforms/cell/spu_priv1_mmio.c
===================================================================
--- cell--alp--3.orig/arch/powerpc/platforms/cell/spu_priv1_mmio.c 2006-05-01 17:06:59.032678000 -0700
+++ cell--alp--3/arch/powerpc/platforms/cell/spu_priv1_mmio.c 2006-05-02 10:11:42.000000000 -0700
@@ -0,0 +1,133 @@
+/*
+ * access to SPU privileged registers
+ */
+#include <linux/module.h>
+
+#include <asm/io.h>
+#include <asm/spu.h>
+
+void spu_int_mask_and(struct spu *spu, int class, u64 mask)
+{
+ u64 old_mask;
+
+ old_mask = in_be64(&spu->priv1->int_mask_RW[class]);
+ out_be64(&spu->priv1->int_mask_RW[class], old_mask & mask);
+}
+EXPORT_SYMBOL_GPL(spu_int_mask_and);
+
+void spu_int_mask_or(struct spu *spu, int class, u64 mask)
+{
+ u64 old_mask;
+
+ old_mask = in_be64(&spu->priv1->int_mask_RW[class]);
+ out_be64(&spu->priv1->int_mask_RW[class], old_mask | mask);
+}
+EXPORT_SYMBOL_GPL(spu_int_mask_or);
+
+void spu_int_mask_set(struct spu *spu, int class, u64 mask)
+{
+ out_be64(&spu->priv1->int_mask_RW[class], mask);
+}
+EXPORT_SYMBOL_GPL(spu_int_mask_set);
+
+u64 spu_int_mask_get(struct spu *spu, int class)
+{
+ return in_be64(&spu->priv1->int_mask_RW[class]);
+}
+EXPORT_SYMBOL_GPL(spu_int_mask_get);
+
+void spu_int_stat_clear(struct spu *spu, int class, u64 stat)
+{
+ out_be64(&spu->priv1->int_stat_RW[class], stat);
+}
+EXPORT_SYMBOL_GPL(spu_int_stat_clear);
+
+u64 spu_int_stat_get(struct spu *spu, int class)
+{
+ return in_be64(&spu->priv1->int_stat_RW[class]);
+}
+EXPORT_SYMBOL_GPL(spu_int_stat_get);
+
+void spu_int_route_set(struct spu *spu, u64 route)
+{
+ out_be64(&spu->priv1->int_route_RW, route);
+}
+EXPORT_SYMBOL_GPL(spu_int_route_set);
+
+u64 spu_mfc_dar_get(struct spu *spu)
+{
+ return in_be64(&spu->priv1->mfc_dar_RW);
+}
+EXPORT_SYMBOL_GPL(spu_mfc_dar_get);
+
+u64 spu_mfc_dsisr_get(struct spu *spu)
+{
+ return in_be64(&spu->priv1->mfc_dsisr_RW);
+}
+EXPORT_SYMBOL_GPL(spu_mfc_dsisr_get);
+
+void spu_mfc_dsisr_set(struct spu *spu, u64 dsisr)
+{
+ out_be64(&spu->priv1->mfc_dsisr_RW, dsisr);
+}
+EXPORT_SYMBOL_GPL(spu_mfc_dsisr_set);
+
+void spu_mfc_sdr_set(struct spu *spu, u64 sdr)
+{
+ out_be64(&spu->priv1->mfc_sdr_RW, sdr);
+}
+EXPORT_SYMBOL_GPL(spu_mfc_sdr_set);
+
+void spu_mfc_sr1_set(struct spu *spu, u64 sr1)
+{
+ out_be64(&spu->priv1->mfc_sr1_RW, sr1);
+}
+EXPORT_SYMBOL_GPL(spu_mfc_sr1_set);
+
+u64 spu_mfc_sr1_get(struct spu *spu)
+{
+ return in_be64(&spu->priv1->mfc_sr1_RW);
+}
+EXPORT_SYMBOL_GPL(spu_mfc_sr1_get);
+
+void spu_mfc_tclass_id_set(struct spu *spu, u64 tclass_id)
+{
+ out_be64(&spu->priv1->mfc_tclass_id_RW, tclass_id);
+}
+EXPORT_SYMBOL_GPL(spu_mfc_tclass_id_set);
+
+u64 spu_mfc_tclass_id_get(struct spu *spu)
+{
+ return in_be64(&spu->priv1->mfc_tclass_id_RW);
+}
+EXPORT_SYMBOL_GPL(spu_mfc_tclass_id_get);
+
+void spu_tlb_invalidate(struct spu *spu)
+{
+ out_be64(&spu->priv1->tlb_invalidate_entry_W, 0ul);
+}
+EXPORT_SYMBOL_GPL(spu_tlb_invalidate);
+
+void spu_resource_allocation_groupID_set(struct spu *spu, u64 id)
+{
+ out_be64(&spu->priv1->resource_allocation_groupID_RW, id);
+}
+EXPORT_SYMBOL_GPL(spu_resource_allocation_groupID_set);
+
+u64 spu_resource_allocation_groupID_get(struct spu *spu)
+{
+ return in_be64(&spu->priv1->resource_allocation_groupID_RW);
+}
+EXPORT_SYMBOL_GPL(spu_resource_allocation_groupID_get);
+
+void spu_resource_allocation_enable_set(struct spu *spu, u64 enable)
+{
+ out_be64(&spu->priv1->resource_allocation_enable_RW, enable);
+}
+EXPORT_SYMBOL_GPL(spu_resource_allocation_enable_set);
+
+u64 spu_resource_allocation_enable_get(struct spu *spu)
+{
+ return in_be64(&spu->priv1->resource_allocation_enable_RW);
+}
+EXPORT_SYMBOL_GPL(spu_resource_allocation_enable_get);
Index: cell--alp--3/drivers/net/Kconfig
===================================================================
--- cell--alp--3.orig/drivers/net/Kconfig 2006-05-02 10:10:52.000000000 -0700
+++ cell--alp--3/drivers/net/Kconfig 2006-05-02 10:11:42.000000000 -0700
@@ -2171,7 +2171,7 @@
config SPIDER_NET
tristate "Spider Gigabit Ethernet driver"
- depends on PCI && PPC_CELL
+ depends on PCI && PPC_IBM_CELL_BLADE
select FW_LOADER
help
This driver supports the Gigabit Ethernet chips present on the
Michael Ellerman wrote:
> On Mon, 2006-05-01 at 15:51 -0700, Geoff Levand wrote:
>> Segher, a problem with your suggestion is that our
>> makefiles don't have as rich a set of logical ops as the
>> config files. Its easy to express 'build A if B', but not
>> so easy to do 'build A if not C'. To make this work
>> cleanly I made PPC_CELL denote !SOME_HYPERVISOR_THING,
>> so I can have constructions like this in the makefile:
>>
>> obj-$(CONFIG_PPC_CELL) += ...
>
> Hi Geoff,
>
> I've been ignoring this discussion, but now that I read it I think this
> is all kinda backwards.
>
> PPC_CELL should not denote !SOME_HYPERVISOR, it should just mean "basic
> cell support", ie. PPC_CELL gets you platforms/cell built in.
Yes, that's the way I originally made it, and I switched it back
to that in the latest patch.
> Then we can have SOME_HYPERVISOR which _adds_ support for that
> hypervisor. And PPC_CELL_BLADE which selects things which are actually
> specific to that hardware, like SPIDERNET etc.
> But SOME_HYPERVISOR should not remove support for running on bare metal,
> it should just give you the option of running on the hypervisor. Yes
> that may require testing things at runtime, that's what
> firmware_has_feature() is for.
I feel you're missing one thing though, we need PPC_CELL_NATIVE. We
don't want to build that in if we don't need it. Here's what I setup:
PPC_CELL = make descends into platforms/cell
PPC_CELL_NATIVE = add bare metal support
PPC_IBM_CELL_BLADE = add blade device drivers, etc.
> The goal should be that we have one kernel which can boot on all Cell
> implementations. In fact the ultimate goal is to have one kernel that
> can boot any platform under powerpc, that's a way off still, but we
> don't want to start going backwards.
Yes, that's the whole idea of this patch. It comes from my patch set
'cell: support multi-platform image'... But we also need to be able
to build in only the support needed. This is an important requirement
for consumer products, to reduce the image size.
In a certain class of products the kernel image and read-only parts
of the file system are stored on flash. A smaller kernel means more
space for applications. Also, a smaller kernel image loads faster.
Device startup time is very important for many consumer products.
-Geoff
Am Tuesday 02 May 2006 20:20 schrieb Geoff Levand:
> Split the Cell BE support into generic and platform dependent parts.
>
> Creates new config variables PPC_CELL_NATIVE and PPC_IBM_CELL_BLADE.
> The existing CONFIG_PPC_CELL is now used to denote the generic
> Cell processor support.
>
> Also renames spu_priv1.c to spu_priv1_mmio.c.
Ok, looks good now.
> Signed-off-by: Geoff Levand <[email protected]>
Acked-by: Arnd Bergmann <[email protected]>
>> Is there any reason the driver wouldn't build and/or run on other
>> platforms? If so, fix it. If not, just make it
>>
>> depends on PCI
>
> Well, it could run on other platforms, except:
>
> - it requires a few properties in the device tree (local-mac-address,
> firmware), so it should also depend on PPC
The portions of code that require OF should have appropriate #ifdef
guards.
> - It's not actually PCI at all, but on an internal bus that has
> something close enough to a PCI config space.
Our emulation should be good enough; if not, holler (off-list).
Segher
Am Wednesday 03 May 2006 01:38 schrieb Segher Boessenkool:
> > - it requires a few properties in the device tree (local-mac-address,
> > ? firmware), so it should also depend on PPC
>
> The portions of code that require OF should have appropriate #ifdef ?
> guards.
Why should we? Getting the mac address and the firmware into the
chip is rather essential to make it work. When there is an #ifdef
around that code, it will only produce a non-working driver that can
be compiled everywhere instead of a driver that can only be compiled
on platforms where it has a chance of working.
If someone has the need to make it work somewhere else, it's easy
enough to change to code to whatever other method is used to set up
the chip.
Arnd <><
Segher Boessenkool writes:
> > Well, it could run on other platforms, except:
> >
> > - it requires a few properties in the device tree (local-mac-address,
> > firmware), so it should also depend on PPC
>
> The portions of code that require OF should have appropriate #ifdef
> guards.
So you're suggesting that we change the Makefile so we can *add*
ifdefs? Usually we do it the other way around. :)
Paul.
>>> Well, it could run on other platforms, except:
>>>
>>> - it requires a few properties in the device tree (local-mac-
>>> address,
>>> firmware), so it should also depend on PPC
>>
>> The portions of code that require OF should have appropriate #ifdef
>> guards.
>
> So you're suggesting that we change the Makefile so we can *add*
> ifdefs? Usually we do it the other way around. :)
Yeah, what was I thinking. So use some platform hook instead.
But Arnd of course is right; if the driver (currently) only works
on a certain platform, just mark it as such in the Makefile (erm,
Kconfig file).
Hey, we should probably do that with 90% of all drivers. But that
is a discussion for some other day :-)
Segher
> Yeah, what was I thinking. So use some platform hook instead.
You must be smoking good stuff :)
> But Arnd of course is right; if the driver (currently) only works
> on a certain platform, just mark it as such in the Makefile (erm,
> Kconfig file).
Exactly :) I don't see any point in adding hooks or ifdef's or anything
fancy like that to guard from something that doesn't exist in real life:
that is building that driver on non-cell :) Thus Kconfig is the way to
go.