Skip to contentSkip to conversation

Updates for ARC VDK platform

PATCH In review Alexey Brodkin 3 participants 15 messages Feb 1, 2017 – Feb 2, 2017

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover Updates for ARC VDK platform

This series improves ARC VDK support in upstream Linux kernel by:
1) Removal of UP configuration which is no longer supported by ARC VDK. Instead SMP platform with all but master cores halted is used to mimic UP system.

2) Adding off-the-tree patches that enable MMC card and UIO

Alexey Brodkin (3):
arc: vdk: Disable halt on reset
arc: vdk: Add support of MMC controller
arc: vdk: Add support of UIO

 arch/arc/boot/dts/vdk_axs10x_mb.dtsi    | 26 ++++++++++++++++++++++++++
 arch/arc/configs/vdk_hs38_smp_defconfig | 10 +++++++---
 2 files changed, 33 insertions(+), 3 deletions(-)
Signature · 1 line
2.7.4
#1 lore raw

PATCH 1/3 arc: vdk: Disable halt on reset

In recent VDKs ARC cores are configured as "run on reset" which made existing kernel configuration outdated to effect that slave cores never start execution of the code keeping only master online.

With that fix we're again in sync with VDK platform.

And while at it we regenerate defconfig via savedefconfig so default options are now excluded.

Signed-off-by Alexey Brodkin <abrodkin at synopsys.com>
---
 arch/arc/configs/vdk_hs38_smp_defconfig | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)
arch/arc/configs/vdk_hs38_smp_defconfig +1 -3
index 573028f19de7..f6361de41eec 100644
@@ -15,7 +15,7 @@ CONFIG_ARC_PLAT_AXS10X=y
 CONFIG_AXS103=y
 CONFIG_ISA_ARCV2=y
 CONFIG_SMP=y
-# CONFIG_ARC_TIMERS_64BIT is not set
+# CONFIG_ARC_SMP_HALT_ON_RESET is not set
 CONFIG_ARC_UBOOT_SUPPORT=y
 CONFIG_ARC_BUILTIN_DTB_NAME="vdk_hs38_smp"
 CONFIG_PREEMPT=y
@@ -56,7 +56,6 @@ CONFIG_NATIONAL_PHY=y
 CONFIG_MOUSE_PS2_TOUCHKIT=y
 CONFIG_SERIO_ARC_PS2=y
 # CONFIG_LEGACY_PTYS is not set
-# CONFIG_DEVKMEM is not set
 CONFIG_SERIAL_8250=y
 CONFIG_SERIAL_8250_CONSOLE=y
 CONFIG_SERIAL_8250_DW=y
@@ -80,7 +79,6 @@ CONFIG_USB_STORAGE=y
 CONFIG_USB_SERIAL=y
 # CONFIG_IOMMU_SUPPORT is not set
 CONFIG_EXT3_FS=y
-CONFIG_EXT4_FS=y
 CONFIG_MSDOS_FS=y
 CONFIG_VFAT_FS=y
 CONFIG_NTFS_FS=y
Signature · 1 line
2.7.4
#2 lore raw

Re: [PATCH 1/3] arc: vdk: Disable halt on reset

Quoted 20 lines · Alexey Brodkin

In recent VDKs ARC cores are configured as "run on reset" which made existing kernel configuration outdated to effect that slave cores never start execution of the code keeping only master online.

With that fix we're again in sync with VDK platform.

And while at it we regenerate defconfig via savedefconfig so default options are now excluded.

Signed-off-by Alexey Brodkin <abrodkin at synopsys.com>
---
 arch/arc/configs/vdk_hs38_smp_defconfig | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)
arch/arc/configs/vdk_hs38_smp_defconfig +0 -0
index 573028f19de7..f6361de41eec 100644

arch/arc/configs/vdk_hs38_smp_defconfig CONFIG_ARC_PLAT_AXS10X=y · line 18

Quoted 4 lines
arch/arc/configs/vdk_hs38_smp_defconfig +0 -1
@@ -15,7 +15,7 @@ CONFIG_ARC_PLAT_AXS10X=y
 CONFIG_AXS103=y
 CONFIG_ISA_ARCV2=y
 CONFIG_SMP=y
-# CONFIG_ARC_TIMERS_64BIT is not set

Are you sure abut this part. Ater the timers driver rework, this would enable GFRC for SMP builds and AFAIKR there were some issues with time with GFRC + nSIM etc..

Quoted 21 lines
arch/arc/configs/vdk_hs38_smp_defconfig +1 -2
+# CONFIG_ARC_SMP_HALT_ON_RESET is not set
 CONFIG_ARC_UBOOT_SUPPORT=y
 CONFIG_ARC_BUILTIN_DTB_NAME="vdk_hs38_smp"
 CONFIG_PREEMPT=y
@@ -56,7 +56,6 @@ CONFIG_NATIONAL_PHY=y
 CONFIG_MOUSE_PS2_TOUCHKIT=y
 CONFIG_SERIO_ARC_PS2=y
 # CONFIG_LEGACY_PTYS is not set
-# CONFIG_DEVKMEM is not set
 CONFIG_SERIAL_8250=y
 CONFIG_SERIAL_8250_CONSOLE=y
 CONFIG_SERIAL_8250_DW=y
@@ -80,7 +79,6 @@ CONFIG_USB_STORAGE=y
 CONFIG_USB_SERIAL=y
 # CONFIG_IOMMU_SUPPORT is not set
 CONFIG_EXT3_FS=y
-CONFIG_EXT4_FS=y
 CONFIG_MSDOS_FS=y
 CONFIG_VFAT_FS=y
 CONFIG_NTFS_FS=y
#3 lore raw

Re: [PATCH 1/3] arc: vdk: Disable halt on reset

Hi Vineet,

Quoted 17 lines · Vineet Gupta
Quoted 16 lines · Alexey Brodkin

In recent VDKs ARC cores are configured as "run on reset" which made existing kernel configuration outdated to effect that slave cores never start execution of the code keeping only master online.

With that fix we're again in sync with VDK platform.

And while at it we regenerate defconfig via savedefconfig so default options are now excluded.

Signed-off-by Alexey Brodkin <abrodkin at synopsys.com>

---
?arch/arc/configs/vdk_hs38_smp_defconfig | 4 +---
?1 file changed, 1 insertion(+), 3 deletions(-)

Quoted 12 lines
Quoted 9 lines
arch/arc/configs/vdk_hs38_smp_defconfig +0 -0
index 573028f19de7..f6361de41eec 100644
@@ -15,7 +15,7 @@ CONFIG_ARC_PLAT_AXS10X=y

?CONFIG_AXS103=y
?CONFIG_ISA_ARCV2=y
?CONFIG_SMP=y
-# CONFIG_ARC_TIMERS_64BIT is not set

Are you sure abut this part. Ater the timers driver rework, this would enable GFRC for SMP builds and AFAIKR there were some issues with time with GFRC + nSIM etc..

Not anymore :)

Probably I missed something in discussions.
As a matter of fact I did run-test resulting vmlinux and it worked very nice. More over?ARC_TIMERS_64BIT is selected automatically by?ISA_ARCV2, see
http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/arch/arc/Kconfig#n119

That said even if "# CONFIG_ARC_TIMERS_64BIT is not set" is left in place the option will be effectively enabled, no?

-Alexey

#4 lore raw

Re: [PATCH 1/3] arc: vdk: Disable halt on reset

Quoted 8 lines · Alexey Brodkin
Quoted 4 lines
Quoted 1 line

-# CONFIG_ARC_TIMERS_64BIT is not set

Are you sure abut this part. Ater the timers driver rework, this would enable GFRC for SMP builds and AFAIKR there were some issues with time with GFRC + nSIM etc..

Not anymore :)

Probably I missed something in discussions.

STAR 9000879565, 9000879563

Quoted 6 lines

As a matter of fact I did run-test resulting vmlinux and it worked very nice. More over ARC_TIMERS_64BIT is selected automatically by ISA_ARCV2, see
http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/arch/arc/Kconfig#n119

That said even if "# CONFIG_ARC_TIMERS_64BIT is not set" is left in place the option will be effectively enabled, no?

The whole point of adding this to defconfig is to override the default from Kconfig ?

#5 lore raw

Re: [PATCH 1/3] arc: vdk: Disable halt on reset

Hi Vineet,

Quoted 16 lines · Vineet Gupta
Quoted 11 lines · Alexey Brodkin
Quoted 7 lines
Quoted 4 lines

-# CONFIG_ARC_TIMERS_64BIT is not set

Are you sure abut this part. Ater the timers driver rework, this would enable GFRC for SMP builds and AFAIKR there were some issues with time with GFRC + nSIM etc..

Not anymore :)

Probably I missed something in discussions.

STAR 9000879565, 9000879563

Quoted 1 line
Quoted 8 lines
Quoted 6 lines

As a matter of fact I did run-test resulting vmlinux and it worked very nice. More over ARC_TIMERS_64BIT is selected automatically by ISA_ARCV2, see
http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/arch/arc/Kconfig#n119

That said even if "# CONFIG_ARC_TIMERS_64BIT is not set" is left in place the option will be effectively enabled, no?

The whole point of adding this to defconfig is to override the default from Kconfig ?

Not anymore :)

Since commit?c4c9a040ecb7 ("clocksource: import ARC timer driver"), see?http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=c4c9a040ecb7297e011e579f5a9cc280e42d725f we have this:
---------------------->8----------------------
config ISA_ARCV2
bool "ARC ISA v2"
select ARC_TIMERS_64BIT
---------------------->8----------------------

which really means if one selects ISA_ARCV2 then ARC_TIMERS_64BIT gets selected automatically and there's no way to override it from either menuconfig or defconfig.

Probably behavior that you meant was to keep a separate "config ARC_TIMERS_64BIT" and have it "default y if ISA_ARCV2".

-Alexey

#6 lore raw

Re: [PATCH 1/3] arc: vdk: Disable halt on reset

Quoted 12 lines · Alexey Brodkin
Quoted 1 line

The whole point of adding this to defconfig is to override the default from Kconfig ?

Not anymore :)

Since commit c4c9a040ecb7 ("clocksource: import ARC timer driver"),
see http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=c4c9a040ecb7297e011e579f5a9cc280e42d725f
we have this:
---------------------->8----------------------
config ISA_ARCV2
bool "ARC ISA v2"
select ARC_TIMERS_64BIT
---------------------->8----------------------

Quoted 2 lines

which really means if one selects ISA_ARCV2 then ARC_TIMERS_64BIT gets selected automatically and there's no way to override it from either menuconfig or defconfig.

Bummer - this means VDK based off 4.9+ kernel might be affected with nsim GFRC issue !

Quoted 2 lines

Probably behavior that you meant was to keep a separate "config ARC_TIMERS_64BIT" and have it "default y if ISA_ARCV2".

No, the whole point of moving it out of arch/arc was to increase test coverage etc so it was not tied to ISA_ARCV2 on purpose so that it would atleast build.

Lets see what Rudd has to say abt this. But GFRC can't be used this would need fixing after all by introducing an additional ARC_PLAT_CANT_USE_TIMERS_64BIT which is def_bool set to n, but selected y in VDK Kconfig.

-Vineet

#7 lore raw

RE: [PATCH 1/3] arc: vdk: Disable halt on reset

Hi Vineet,

Quoted 24 lines

-----Original Message-----
From: Vineet Gupta [mailto:vgupta at synopsys.com]
Sent: Wednesday, February 1, 2017 10:56 PM
To: Alexey Brodkin <Alexey.Brodkin at synopsys.com>
Cc: Ruud Derwig <rderwig at synopsys.com>; linux-kernel at vger.kernel.org; linux-snps-arc at lists.infradead.org
Subject: Re: [PATCH 1/3] arc: vdk: Disable halt on reset

Quoted 1 line · Alexey Brodkin
Quoted 1 line

The whole point of adding this to defconfig is to override the default from

Kconfig ?

Quoted 13 lines

Not anymore :)

Since commit c4c9a040ecb7 ("clocksource: import ARC timer driver"), see
http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/
?id=c4c9a040ecb7297e011e579f5a9cc280e42d725f
we have this:
---------------------->8----------------------
config ISA_ARCV2
bool "ARC ISA v2"
select ARC_TIMERS_64BIT
---------------------->8----------------------

Quoted 6 lines
Quoted 2 lines

which really means if one selects ISA_ARCV2 then ARC_TIMERS_64BIT gets selected automatically and there's no way to override it from either

menuconfig or defconfig.

Bummer - this means VDK based off 4.9+ kernel might be affected with nsim GFRC issue !

One note here - I think it only affects 4.10+ kernels.

-Aexey

#8 lore raw

RE: [PATCH 1/3] arc: vdk: Disable halt on reset

In reply to Vineet Gupta

Hi,

VDK config includes GRFC, not sure if it's working/used though.
The nSIM STAR is about that GRFC frequency is fixed in nSIM, but configurable for HW/customers. But if the frequency is configured correctly (same as cpu frequency), don't think there's an issue.

Ruud.
-----Original Message-----
From: Vineet Gupta
Sent: Wednesday, February 01, 2017 8:56 PM
To: Alexey Brodkin <Alexey.Brodkin at synopsys.com>
Cc: Ruud Derwig <rderwig at synopsys.com>; linux-kernel at vger.kernel.org; linux-snps-arc at lists.infradead.org
Subject: Re: [PATCH 1/3] arc: vdk: Disable halt on reset

Quoted 14 lines · Alexey Brodkin
Quoted 1 line

The whole point of adding this to defconfig is to override the default from Kconfig ?

Not anymore :)

Since commit c4c9a040ecb7 ("clocksource: import ARC timer driver"), see
http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/
?id=c4c9a040ecb7297e011e579f5a9cc280e42d725f
we have this:
---------------------->8----------------------
config ISA_ARCV2
bool "ARC ISA v2"
select ARC_TIMERS_64BIT
---------------------->8----------------------

Quoted 2 lines

which really means if one selects ISA_ARCV2 then ARC_TIMERS_64BIT gets selected automatically and there's no way to override it from either menuconfig or defconfig.

Bummer - this means VDK based off 4.9+ kernel might be affected with nsim GFRC issue !

Quoted 2 lines

Probably behavior that you meant was to keep a separate "config ARC_TIMERS_64BIT" and have it "default y if ISA_ARCV2".

No, the whole point of moving it out of arch/arc was to increase test coverage etc so it was not tied to ISA_ARCV2 on purpose so that it would atleast build.

Lets see what Rudd has to say abt this. But GFRC can't be used this would need fixing after all by introducing an additional ARC_PLAT_CANT_USE_TIMERS_64BIT which is def_bool set to n, but selected y in VDK Kconfig.

-Vineet

#9 lore raw

PATCH 2/3 arc: vdk: Add support of MMC controller

In reply to Alexey Brodkin

ARC VDK virtual platform emulates host MMC controller (DW Mobile Storage) and moreover rootfs is situated on that virtual card.

Signed-off-by Alexey Brodkin <abrodkin at synopsys.com>
---
 arch/arc/boot/dts/vdk_axs10x_mb.dtsi    | 18 ++++++++++++++++++
 arch/arc/configs/vdk_hs38_smp_defconfig |  4 ++++
 2 files changed, 22 insertions(+)
arch/arc/boot/dts/vdk_axs10x_mb.dtsi +18 -0
index 99498a4b4216..1953914b9f4f 100644
@@ -23,6 +23,12 @@
 				#clock-cells = <0>;
 			};
 
+			mmcclk: mmcclk {
+				compatible = "fixed-clock";
+				clock-frequency = <50000000>;
+				#clock-cells = <0>;
+			};
+
 			pguclk: pguclk {
 				#clock-cells = <0>;
 				compatible = "fixed-clock";
@@ -94,5 +100,17 @@
 			interrupts = <5>;
 			interrupt-names = "arc_ps2_irq";
 		};
+
+		mmc at 0x15000 {
+			compatible = "snps,dw-mshc";
+			reg = <0x15000 0x400>;
+			num-slots = <1>;
+			fifo-depth = <1024>;
+			card-detect-delay = <200>;
+			clocks = <&apbclk>, <&mmcclk>;
+			clock-names = "biu", "ciu";
+			interrupts = <7>;
+			bus-width = <4>;
+		};
 	};
 };
arch/arc/configs/vdk_hs38_smp_defconfig +4 -0
index f6361de41eec..e03413fbc784 100644
@@ -77,6 +77,10 @@ CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_OHCI_HCD_PLATFORM=y
 CONFIG_USB_STORAGE=y
 CONFIG_USB_SERIAL=y
+CONFIG_MMC=y
+CONFIG_MMC_SDHCI=y
+CONFIG_MMC_SDHCI_PLTFM=y
+CONFIG_MMC_DW=y
 # CONFIG_IOMMU_SUPPORT is not set
 CONFIG_EXT3_FS=y
 CONFIG_MSDOS_FS=y
Signature · 1 line
2.7.4
#10 lore raw

PATCH 3/3 arc: vdk: Add support of UIO

In reply to Alexey Brodkin

ARC VDK for EVSS uses UIO for communication with Embedded Vision Subsystem.

Signed-off-by Alexey Brodkin <abrodkin at synopsys.com>
---
 arch/arc/boot/dts/vdk_axs10x_mb.dtsi    | 8 ++++++++
 arch/arc/configs/vdk_hs38_smp_defconfig | 2 ++
 2 files changed, 10 insertions(+)
arch/arc/boot/dts/vdk_axs10x_mb.dtsi +8 -0
index 1953914b9f4f..f0df59b23e21 100644
@@ -112,5 +112,13 @@
 			interrupts = <7>;
 			bus-width = <4>;
 		};
+
+		/* Embedded Vision subsystem UIO mappings; only relevant for EV VDK */
+		uio_ev: uio at 0xD0000000 {
+			compatible = "generic-uio";
+			reg = <0xD0000000 0x2000 0xD1000000 0x2000 0x90000000 0x10000000 0xC0000000 0x10000000>;
+			reg-names = "ev_gsa", "ev_ctrl", "ev_shared_mem", "ev_code_mem";
+			interrupts = <23>;
+		};
 	};
 };
arch/arc/configs/vdk_hs38_smp_defconfig +2 -0
index e03413fbc784..1a9b19903c9a 100644
@@ -81,6 +81,8 @@ CONFIG_MMC=y
 CONFIG_MMC_SDHCI=y
 CONFIG_MMC_SDHCI_PLTFM=y
 CONFIG_MMC_DW=y
+CONFIG_UIO=y
+CONFIG_UIO_PDRV_GENIRQ=y
 # CONFIG_IOMMU_SUPPORT is not set
 CONFIG_EXT3_FS=y
 CONFIG_MSDOS_FS=y
Signature · 1 line
2.7.4
#11 lore raw
In reply to Alexey Brodkin
Quoted 4 lines · Alexey Brodkin

This series improves ARC VDK support in upstream Linux kernel by:
1) Removal of UP configuration which is no longer supported by ARC VDK. Instead SMP platform with all but master cores halted is used to mimic UP system.

Is this missing from the series ?

#12 lore raw

Hi Vineet,

Quoted 8 lines · Vineet Gupta
Quoted 5 lines · Alexey Brodkin

This series improves ARC VDK support in upstream Linux kernel by:
?1) Removal of UP configuration which is no longer supported by ARC VDK. ????Instead SMP platform with all but master cores halted is used to mimic ????UP system.

Is this missing from the series ?

Indeed, I was doing 2 things in parallel and so missed -1 patch :( I may either send v2 with missing patch included or save it for another series that ditches most of UP configs as well. What suits you better?

-Alexey

#13 lore raw
Quoted 4 lines · Alexey Brodkin

Hi Vineet,

Quoted 1 line · Vineet Gupta

On 02/01/2017 08:43 AM, Alexey Brodkin wrote:

Quoted 8 lines
Quoted 5 lines
Quoted 4 lines

This series improves ARC VDK support in upstream Linux kernel by:
1) Removal of UP configuration which is no longer supported by ARC VDK. Instead SMP platform with all but master cores halted is used to mimic UP system.

Is this missing from the series ?

Indeed, I was doing 2 things in parallel and so missed -1 patch :( I may either send v2 with missing patch included or save it for another series that ditches most of UP configs as well. What suits you better?

Latter would be better !

-Vineet

#14 lore raw

Hi Vineet,

Quoted 12 lines · Vineet Gupta
Quoted 11 lines · Alexey Brodkin

Hi Vineet,

Quoted 7 lines · Vineet Gupta
Quoted 5 lines · Alexey Brodkin

This series improves ARC VDK support in upstream Linux kernel by:
?1) Removal of UP configuration which is no longer supported by ARC VDK. ????Instead SMP platform with all but master cores halted is used to mimic ????UP system.

Quoted 6 lines
Quoted 4 lines
Quoted 1 line

Is this missing from the series ?

Indeed, I was doing 2 things in parallel and so missed -1 patch :( I may either send v2 with missing patch included or save it for another series that ditches most of UP configs as well. What suits you better?

Latter would be better !

So then you're going to apply existing series as it is sometime soon and in the meantime I'll prepare another one, right?

-Alexey