thr3ads.net - Xen devel - [PATCH 0/4] xen/arm: assemble support for Allwinner A31 [May 2013]

If this information is useful, please help other people find it:
Share via:

Bamvor Jian Zhang

2013-May-31 12:53 UTC

[PATCH 0/4] xen/arm: assemble support for Allwinner A31

These series patch enable Allwinner A31(code name sun6i) support in
assemble. with these patches, the cpu 0 of sun6i SOC could successful
boot into the c environment.

Bamvor Jian Zhang (4):
  xen/arm: introduce Cortex-A7 support
  xen/arm: introduce Allwinner sun6i SOC basic support
  xen/arm: enable early printk for sun6i
  xen/arm: enable switch to hyper mode for sun6i

 xen/arch/arm/Rules.mk                 |  5 ++++
 xen/arch/arm/arm32/Makefile           |  1 +
 xen/arch/arm/arm32/debug-sun6i.inc    | 50 +++++++++++++++++++++++++++++++++++
 xen/arch/arm/arm32/head.S             | 11 ++++++++
 xen/arch/arm/arm32/mode_switch.S      | 13 +++++++--
 xen/arch/arm/arm32/proc-ca7.S         | 36 +++++++++++++++++++++++++
 xen/include/asm-arm/platforms/sun6i.h | 28 ++++++++++++++++++++
 xen/include/asm-arm/processor-ca7.h   | 25 ++++++++++++++++++
 8 files changed, 167 insertions(+), 2 deletions(-)
 create mode 100644 xen/arch/arm/arm32/debug-sun6i.inc
 create mode 100644 xen/arch/arm/arm32/proc-ca7.S
 create mode 100644 xen/include/asm-arm/platforms/sun6i.h
 create mode 100644 xen/include/asm-arm/processor-ca7.h

-- 
1.8.1.4

Bamvor Jian Zhang

2013-May-31 12:53 UTC

head link

[PATCH 1/4] xen/arm: introduce Cortex-A7 support

There are still some different registers compare with Cortex-A7
and Cortex-A15, such as ACTLR. So, the dedicated Cortex-A7 code
is introduced, including the Cortex-A7 initialization function
in assemble.

Signed-off-by: Bamvor Jian Zhang <bjzhang@suse.com>
---
 xen/arch/arm/arm32/Makefile         |  1 +
 xen/arch/arm/arm32/head.S           | 11 +++++++++++
 xen/arch/arm/arm32/proc-ca7.S       | 36 ++++++++++++++++++++++++++++++++++++
 xen/include/asm-arm/processor-ca7.h | 25 +++++++++++++++++++++++++
 4 files changed, 73 insertions(+)
 create mode 100644 xen/arch/arm/arm32/proc-ca7.S
 create mode 100644 xen/include/asm-arm/processor-ca7.h

diff --git a/xen/arch/arm/arm32/Makefile b/xen/arch/arm/arm32/Makefile
index aaf277a..09b5256 100644
--- a/xen/arch/arm/arm32/Makefile
+++ b/xen/arch/arm/arm32/Makefile
@@ -3,6 +3,7 @@ subdir-y += lib
 obj-y += entry.o
 obj-y += mode_switch.o
 obj-y += proc-ca15.o
+obj-y += proc-ca7.o
 
 obj-y += traps.o
 obj-y += domain.o
diff --git a/xen/arch/arm/arm32/head.S b/xen/arch/arm/arm32/head.S
index 0588d54..d3b849b 100644
--- a/xen/arch/arm/arm32/head.S
+++ b/xen/arch/arm/arm32/head.S
@@ -20,6 +20,7 @@
 #include <asm/config.h>
 #include <asm/page.h>
 #include <asm/processor-ca15.h>
+#include <asm/processor-ca7.h>
 #include <asm/asm_defns.h>
 
 #define ZIMAGE_MAGIC_NUMBER 0x016f2818
@@ -195,8 +196,18 @@ skip_bss:
         /* Is this a Cortex A15? */
         ldr   r1, =(CORTEX_A15_ID)
         teq   r0, r1
+        bne   test_ca7
         bleq  cortex_a15_init
+        PRINT("- cortex-a15 init done -\r\n")
+        b     cpu_init_done
+        /* Is this a Cortex A7? */
+test_ca7:
+        ldr   r1, =(CORTEX_A7_ID)
+        teq   r0, r1
+        bleq  cortex_a7_init
+        PRINT("- cortex-a7 init done -\r\n")
 
+cpu_init_done:
         /* Set up memory attribute type tables */
         ldr   r0, =MAIR0VAL
         ldr   r1, =MAIR1VAL
diff --git a/xen/arch/arm/arm32/proc-ca7.S b/xen/arch/arm/arm32/proc-ca7.S
new file mode 100644
index 0000000..e0c1bc2
--- /dev/null
+++ b/xen/arch/arm/arm32/proc-ca7.S
@@ -0,0 +1,36 @@
+/*
+ * xen/arch/arm/proc-ca7.S
+ *
+ * Cortex A7 specific initializations
+ *
+ * Bamvor Jian Zhang <bjzhang@suse.com>
+ * Copyright (c) 2013 SUSE
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+#include <asm/asm_defns.h>
+#include <asm/processor-ca7.h>
+
+.globl cortex_a7_init
+cortex_a7_init:
+        /* Set up the SMP bit in ACTLR */
+        mrc   CP32(r0, ACTLR)
+        orr   r0, r0, #(ACTLR_CA7_SMP) /* enable SMP bit */
+        mcr   CP32(r0, ACTLR)
+        mov   pc, lr
+
+/*
+ * Local variables:
+ * mode: ASM
+ * indent-tabs-mode: nil
+ * End:
+ */
diff --git a/xen/include/asm-arm/processor-ca7.h
b/xen/include/asm-arm/processor-ca7.h
new file mode 100644
index 0000000..b773c34
--- /dev/null
+++ b/xen/include/asm-arm/processor-ca7.h
@@ -0,0 +1,25 @@
+#ifndef __ASM_ARM_PROCESSOR_CA7_H
+#define __ASM_ARM_PROCESSOR_CA7_H
+
+
+#define CORTEX_A7_ID     (0x410FC070)
+
+/* ACTLR Auxiliary Control Register, Cortex A7 */
+#define ACTLR_CA7_DDI                (1<<28)
+#define ACTLR_CA7_DDVM               (1<<15)
+/* 2 bits */
+#define ACTLR_CA7_L1PCTL             (1<<13)
+#define ACTLR_CA7_L1RADIS            (1<<12)
+#define ACTLR_CA7_L2RADIS            (1<<11)
+#define ACTLR_CA7_DODMBS             (1<<10)
+#define ACTLR_CA7_SMP                (1<<6)
+
+#endif /* __ASM_ARM_PROCESSOR_CA7_H */
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
-- 
1.8.1.4

Bamvor Jian Zhang

2013-May-31 12:53 UTC

head link

[PATCH 2/4] xen/arm: introduce Allwinner sun6i SOC basic support

Allwinner A31(code name sun6i) is a quad-core Cortex-A7 SOC. Allwinner has
released the source code in http://git.rhombus-tech.net.

This patch define registers and macros for early printk and mode switch.
kick cpu is not implemneted yet. because the secondary cpu boot up
sequence is a little bit different from exynos5.

Signed-off-by: Bamvor Jian Zhang <bjzhang@suse.com>
---
 xen/arch/arm/arm32/debug-sun6i.inc    | 50 +++++++++++++++++++++++++++++++++++
 xen/include/asm-arm/platforms/sun6i.h | 28 ++++++++++++++++++++
 2 files changed, 78 insertions(+)
 create mode 100644 xen/arch/arm/arm32/debug-sun6i.inc
 create mode 100644 xen/include/asm-arm/platforms/sun6i.h

diff --git a/xen/arch/arm/arm32/debug-sun6i.inc
b/xen/arch/arm/arm32/debug-sun6i.inc
new file mode 100644
index 0000000..eaa1197
--- /dev/null
+++ b/xen/arch/arm/arm32/debug-sun6i.inc
@@ -0,0 +1,50 @@
+/*
+ * xen/arch/arm/arm32/debug-sun6i.inc
+ *
+ * Allwinner sun6i specific debug code
+ *
+ * Bamvor Jian Zhang <bjzhang@suse.com>
+ * Copyright (c) 2013 SUSE
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+/* sun6i UART initialization
+ * rb: register which contains the UART base address
+ * rc: scratch register 1 (unused here)
+ * rd: scratch register 2 (unused here) */
+.macro early_uart_init rb, rc, rd
+/* assume the uart already init by bootloader */
+.endm
+
+/* sun6i UART wait UART to be ready to transmit
+ * rb: register which contains the UART base address
+ * rc: scratch register */
+.macro early_uart_ready rb, rc
+1:
+        ldr   \rc, [\rb, #0x7c]     /* <- UART_USR (Status Register) */
+        tst   \rc, #0x2             /* Check BUSY bit */
+        beq   1b                    /* Wait for the UART to be ready */
+.endm
+
+/* sun6i UART transmit character
+ * rb: register which contains the UART base address
+ * rt: register which contains the character to transmit */
+.macro early_uart_transmit rb, rt
+        str   \rt, [\rb]            /* -> UART_THR (Transmit Holding
Register) */
+.endm
+
+/*
+ * Local variables:
+ * mode: ASM
+ * indent-tabs-mode: nil
+ * End:
+ */
diff --git a/xen/include/asm-arm/platforms/sun6i.h
b/xen/include/asm-arm/platforms/sun6i.h
new file mode 100644
index 0000000..f910ad6
--- /dev/null
+++ b/xen/include/asm-arm/platforms/sun6i.h
@@ -0,0 +1,28 @@
+#ifndef __ASM_ARM_PLATFORMS_SUN6I_H
+#define __ASM_ASM_PLATFORMS_SUN6I_H
+
+/* Constants below is only used in assembly because the DTS is not yet parsed
*/
+#ifdef __ASSEMBLY__
+
+/* GIC Base Address */
+#define SUN6I_GIC_BASE_ADDRESS       0x01c82000
+
+/* Timer''s frequency */
+/* \TODO I do not know yet, assume it 24MHz. check it later */
+#define SUN6I_TIMER_FREQUENCY       (24 * 1000 * 1000) /* 24 MHz */
+
+/* Arndale machine ID */
+/* XXX 3894 is downstream number for sun6i. 4137 is upstream number for sun6i
*/
+#define MACH_TYPE_SUN6I             4137
+
+#endif /* __ASSEMBLY__ */
+
+#endif /* __ASM_ARM_PLATFORMS_SUN6I_H */
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
-- 
1.8.1.4

Bamvor Jian Zhang

2013-May-31 12:53 UTC

head link

[PATCH 3/4] xen/arm: enable early printk for sun6i

enable early printk for sun6i base on the macro introduce by previous
patch.

Signed-off-by: Bamvor Jian Zhang <bjzhang@suse.com>
---
 xen/arch/arm/Rules.mk | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/xen/arch/arm/Rules.mk b/xen/arch/arm/Rules.mk
index 422ed04..59d29c4 100644
--- a/xen/arch/arm/Rules.mk
+++ b/xen/arch/arm/Rules.mk
@@ -64,6 +64,11 @@ EARLY_PRINTK_INC := pl011
 EARLY_PRINTK_BAUD := 115200
 EARLY_UART_BASE_ADDRESS := 0xfff36000
 endif
+ifeq ($(CONFIG_EARLY_PRINTK), sun6i)
+EARLY_PRINTK_INC := sun6i
+EARLY_PRINTK_BAUD := 115200
+EARLY_UART_BASE_ADDRESS := 0x01c28000
+endif
 
 ifneq ($(EARLY_PRINTK_INC),)
 EARLY_PRINTK := y
-- 
1.8.1.4

Bamvor Jian Zhang

2013-May-31 12:53 UTC

head link

[PATCH 4/4] xen/arm: enable switch to hyper mode for sun6i

Signed-off-by: Bamvor Jian Zhang <bjzhang@suse.com>
---
 xen/arch/arm/arm32/mode_switch.S | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)

diff --git a/xen/arch/arm/arm32/mode_switch.S b/xen/arch/arm/arm32/mode_switch.S
index c92a1cf..0fc26e0 100644
--- a/xen/arch/arm/arm32/mode_switch.S
+++ b/xen/arch/arm/arm32/mode_switch.S
@@ -21,6 +21,7 @@
 #include <asm/page.h>
 #include <asm/platforms/vexpress.h>
 #include <asm/platforms/exynos5.h>
+#include <asm/platforms/sun6i.h>
 #include <asm/asm_defns.h>
 #include <asm/gic.h>
 
@@ -90,15 +91,23 @@ enter_hyp_mode:
         bic   r0, r0, #0xe           /* Clear EA, FIQ and IRQ */
         mcr   CP32(r0, SCR)
 
-        ldr   r2, =MACH_TYPE_SMDK5250   /* r4 := Arndale machine ID */
         /* By default load Arndale defaults values */
+        ldr   r2, =MACH_TYPE_SMDK5250   /* r4 := Arndale machine ID */
         ldr   r0, =EXYNOS5_TIMER_FREQUENCY  /* r0 := timer''s frequency
*/
         ldr   r1, =EXYNOS5_GIC_BASE_ADDRESS /* r1 := GIC base address */
-        /* If it''s not the Arndale machine ID, load VE values */
         teq   r5, r2
+        beq   board_set_done
+        /* If it''s not the Arndale machine ID, try sun6i */
+        ldr   r2, =MACH_TYPE_SUN6I          /* r2 := sun6i machine ID */
+        ldr   r0, =SUN6I_TIMER_FREQUENCY    /* r0 := timer''s frequency
*/
+        ldr   r1, =SUN6I_GIC_BASE_ADDRESS   /* r1 := GIC base address */
+        teq   r5, r2
+        beq   board_set_done
+        /* If it''s not the sun6i machine ID, load VE values */
         ldrne r0, =V2M_TIMER_FREQUENCY
         ldrne r1, =V2M_GIC_BASE_ADDRESS
 
+board_set_done:
         /* Ugly: the system timer''s frequency register is only
          * programmable in Secure state.  Since we don''t know where
its
          * memory-mapped control registers live, we can''t find out the
-- 
1.8.1.4

Ian Campbell

2013-May-31 15:14 UTC

head link

Re: [PATCH 0/4] xen/arm: assemble support for Allwinner A31

On Fri, 2013-05-31 at 20:53 +0800, Bamvor Jian Zhang
wrote:> These series patch enable Allwinner A31(code name sun6i) support in
> assemble. with these patches, the cpu 0 of sun6i SOC could successful
> boot into the c environment.
I likely won''t have a chance to look at the code properly until next
week but I just wanted to say "cool" ;-) Nice to see support for a
second core AND another SoC!

You are using the MELE A1000 platform, right?
> Bamvor Jian Zhang (4):
>   xen/arm: introduce Cortex-A7 support
>   xen/arm: introduce Allwinner sun6i SOC basic support
>   xen/arm: enable early printk for sun6i
>   xen/arm: enable switch to hyper mode for sun6i
> 
>  xen/arch/arm/Rules.mk                 |  5 ++++
>  xen/arch/arm/arm32/Makefile           |  1 +
>  xen/arch/arm/arm32/debug-sun6i.inc    | 50
+++++++++++++++++++++++++++++++++++
>  xen/arch/arm/arm32/head.S             | 11 ++++++++
>  xen/arch/arm/arm32/mode_switch.S      | 13 +++++++--
>  xen/arch/arm/arm32/proc-ca7.S         | 36 +++++++++++++++++++++++++
>  xen/include/asm-arm/platforms/sun6i.h | 28 ++++++++++++++++++++
>  xen/include/asm-arm/processor-ca7.h   | 25 ++++++++++++++++++
>  8 files changed, 167 insertions(+), 2 deletions(-)
>  create mode 100644 xen/arch/arm/arm32/debug-sun6i.inc
>  create mode 100644 xen/arch/arm/arm32/proc-ca7.S
>  create mode 100644 xen/include/asm-arm/platforms/sun6i.h
>  create mode 100644 xen/include/asm-arm/processor-ca7.h
>

Bamvor Jian Zhang

2013-May-31 15:49 UTC

head link

Re: [PATCH 0/4] xen/arm: assemble support for Allwinner A31

>>> Ian Campbell  2013-5-31 at 23:14 >>>
> On Fri, 2013-05-31 at 20:53 +0800, Bamvor Jian Zhang wrote:
>> These series patch enable Allwinner A31(code name sun6i) support in
>> assemble. with these patches, the cpu 0 of sun6i SOC could successful
>> boot into the c environment.
> I likely won''t have a chance to look at the code properly until
next
> week but I just wanted to say "cool" ;-)
thanks in advance.> Nice to see support for a second core AND another SoC!
> You are using the MELE A1000 platform, right?yeah, mele A1000g quad core. In China, it is mele M9 powered by A31 and
mele M5 powered by A20.
i am not sure whether the a1000g and m9 is the exact same
one.>> Bamvor Jian Zhang (4):
>>   xen/arm: introduce Cortex-A7 support
>>   xen/arm: introduce Allwinner sun6i SOC basic support
>>   xen/arm: enable early printk for sun6i
>>   xen/arm: enable switch to hyper mode for sun6i
>>
>>  xen/arch/arm/Rules.mk                 |  5 ++++
>>  xen/arch/arm/arm32/Makefile           |  1 +
>>  xen/arch/arm/arm32/debug-sun6i.inc    | 50
+++++++++++++++++++++++++++++++++++
>>  xen/arch/arm/arm32/head.S             | 11 ++++++++
>>  xen/arch/arm/arm32/mode_switch.S      | 13 +++++++--
>>  xen/arch/arm/arm32/proc-ca7.S         | 36 +++++++++++++++++++++++++
>>  xen/include/asm-arm/platforms/sun6i.h | 28 ++++++++++++++++++++
>>  xen/include/asm-arm/processor-ca7.h   | 25 ++++++++++++++++++
>>  8 files changed, 167 insertions(+), 2 deletions(-)
>>  create mode 100644 xen/arch/arm/arm32/debug-sun6i.inc
>>  create mode 100644 xen/arch/arm/arm32/proc-ca7.S
>>  create mode 100644 xen/include/asm-arm/platforms/sun6i.h
>>  create mode 100644 xen/include/asm-arm/processor-ca7.h

Ian Campbell

2013-Jun-12 09:57 UTC

head link

Re: [PATCH 1/4] xen/arm: introduce Cortex-A7 support

On Fri, 2013-05-31 at 20:53 +0800, Bamvor Jian Zhang
wrote:> diff --git a/xen/arch/arm/arm32/head.S b/xen/arch/arm/arm32/head.S
> index 0588d54..d3b849b 100644
> --- a/xen/arch/arm/arm32/head.S
> +++ b/xen/arch/arm/arm32/head.S
> @@ -20,6 +20,7 @@
>  #include <asm/config.h>
>  #include <asm/page.h>
>  #include <asm/processor-ca15.h>
> +#include <asm/processor-ca7.h>
>  #include <asm/asm_defns.h>
>  
>  #define ZIMAGE_MAGIC_NUMBER 0x016f2818
> @@ -195,8 +196,18 @@ skip_bss:
>          /* Is this a Cortex A15? */
>          ldr   r1, =(CORTEX_A15_ID)
>          teq   r0, r1
> +        bne   test_ca7
>          bleq  cortex_a15_init
> +        PRINT("- cortex-a15 init done -\r\n")
> +        b     cpu_init_done
> +        /* Is this a Cortex A7? */
> +test_ca7:
> +        ldr   r1, =(CORTEX_A7_ID)
> +        teq   r0, r1
> +        bleq  cortex_a7_init
> +        PRINT("- cortex-a7 init done -\r\n")
I think now that we have a second processor here it is time to go for a
more scalable approach. A linker table driven thing, similar to what
Julien did for the platform (PLATFORM_START et al) or the Linux proc
table thing seems like the right answer. At the moment it would just
contain the MIDR mask and match values plus an optional init function
pointer, in time I suspect it will grow other functionality.
> +cpu_init_done:
>          /* Set up memory attribute type tables */
>          ldr   r0, =MAIR0VAL
>          ldr   r1, =MAIR1VAL
> diff --git a/xen/arch/arm/arm32/proc-ca7.S b/xen/arch/arm/arm32/proc-ca7.S
> new file mode 100644
> index 0000000..e0c1bc2
> --- /dev/null
> +++ b/xen/arch/arm/arm32/proc-ca7.S
> @@ -0,0 +1,36 @@
> +/*
> + * xen/arch/arm/proc-ca7.S
> + *
> + * Cortex A7 specific initializations
> + *
> + * Bamvor Jian Zhang <bjzhang@suse.com>
> + * Copyright (c) 2013 SUSE
> + *
> + * This program is free software; you can redistribute it and/or modify
> + * it under the terms of the GNU General Public License as published by
> + * the Free Software Foundation; either version 2 of the License, or
> + * (at your option) any later version.
> + *
> + * This program is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> + * GNU General Public License for more details.
> + */
> +
> +#include <asm/asm_defns.h>
> +#include <asm/processor-ca7.h>
> +
> +.globl cortex_a7_init
> +cortex_a7_init:
> +        /* Set up the SMP bit in ACTLR */
> +        mrc   CP32(r0, ACTLR)
> +        orr   r0, r0, #(ACTLR_CA7_SMP) /* enable SMP bit */
> +        mcr   CP32(r0, ACTLR)
> +        mov   pc, lr
> +
> +/*
> + * Local variables:
> + * mode: ASM
> + * indent-tabs-mode: nil
> + * End:
> + */
> diff --git a/xen/include/asm-arm/processor-ca7.h
b/xen/include/asm-arm/processor-ca7.h
> new file mode 100644
> index 0000000..b773c34
> --- /dev/null
> +++ b/xen/include/asm-arm/processor-ca7.h
> @@ -0,0 +1,25 @@
> +#ifndef __ASM_ARM_PROCESSOR_CA7_H
> +#define __ASM_ARM_PROCESSOR_CA7_H
> +
> +
> +#define CORTEX_A7_ID     (0x410FC070)
> +
> +/* ACTLR Auxiliary Control Register, Cortex A7 */
> +#define ACTLR_CA7_DDI                (1<<28)
> +#define ACTLR_CA7_DDVM               (1<<15)
> +/* 2 bits */
> +#define ACTLR_CA7_L1PCTL             (1<<13)
> +#define ACTLR_CA7_L1RADIS            (1<<12)
> +#define ACTLR_CA7_L2RADIS            (1<<11)
> +#define ACTLR_CA7_DODMBS             (1<<10)
> +#define ACTLR_CA7_SMP                (1<<6)
Is the content of this register really different between A7 and A15?

At least CA7_SMP is the same as CA15_SMP, which suggests we could share
the init function (to set the SMP bit) and have both a7 and a15 in the
same proc-v7.S file.
> +
> +#endif /* __ASM_ARM_PROCESSOR_CA7_H */
> +/*
> + * Local variables:
> + * mode: C
> + * c-file-style: "BSD"
> + * c-basic-offset: 4
> + * indent-tabs-mode: nil
> + * End:
> + */

Ian Campbell

2013-Jun-12 10:17 UTC

head link

Re: [PATCH 2/4] xen/arm: introduce Allwinner sun6i SOC basic support

On Fri, 2013-05-31 at 20:53 +0800, Bamvor Jian Zhang
wrote:> Allwinner A31(code name sun6i) is a quad-core Cortex-A7 SOC. Allwinner has
> released the source code in http://git.rhombus-tech.net.
> 
> This patch define registers and macros for early printk and mode switch.
> kick cpu is not implemneted yet. because the secondary cpu boot up
> sequence is a little bit different from exynos5.
> 
> Signed-off-by: Bamvor Jian Zhang <bjzhang@suse.com>
> ---
>  xen/arch/arm/arm32/debug-sun6i.inc    | 50
+++++++++++++++++++++++++++++++++++
>  xen/include/asm-arm/platforms/sun6i.h | 28 ++++++++++++++++++++
>  2 files changed, 78 insertions(+)
>  create mode 100644 xen/arch/arm/arm32/debug-sun6i.inc
>  create mode 100644 xen/include/asm-arm/platforms/sun6i.h
> 
> diff --git a/xen/arch/arm/arm32/debug-sun6i.inc
b/xen/arch/arm/arm32/debug-sun6i.inc
> new file mode 100644
> index 0000000..eaa1197
> --- /dev/null
> +++ b/xen/arch/arm/arm32/debug-sun6i.inc
> @@ -0,0 +1,50 @@
> +/*
> + * xen/arch/arm/arm32/debug-sun6i.inc
> + *
> + * Allwinner sun6i specific debug code
> + *
> + * Bamvor Jian Zhang <bjzhang@suse.com>
> + * Copyright (c) 2013 SUSE
> + *
> + * This program is free software; you can redistribute it and/or modify
> + * it under the terms of the GNU General Public License as published by
> + * the Free Software Foundation; either version 2 of the License, or
> + * (at your option) any later version.
> + *
> + * This program is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> + * GNU General Public License for more details.
> + */
> +
> +/* sun6i UART initialization
This does appear to use a distinct set of registers from either the
existing pl011 or exynos4xxx UART driver so it should be a separate
driver. Do you know if this particular logic block is specific to sn6i
or is it used elsewhere (e.g. in all sun*i)? Do you know if there is a
more generic name for the part?

Not a big deal -- we can always rename to be more generic as we see
other examples of the same UART.
> + * rb: register which contains the UART base address
> + * rc: scratch register 1 (unused here)
> + * rd: scratch register 2 (unused here) */
> +.macro early_uart_init rb, rc, rd
> +/* assume the uart already init by bootloader */
> +.endm
I think Andre''s refactoring of the UART init stuff in and around commit
130645eee340608c12b632535d3f983508348ba1 lets you avoid this by just not
setting EARLY_PRINTK_INIT_UART.
> +
> +/* sun6i UART wait UART to be ready to transmit
> + * rb: register which contains the UART base address
> + * rc: scratch register */
> +.macro early_uart_ready rb, rc
> +1:
> +        ldr   \rc, [\rb, #0x7c]     /* <- UART_USR (Status Register) */
> +        tst   \rc, #0x2             /* Check BUSY bit */
> +        beq   1b                    /* Wait for the UART to be ready */
> +.endm
> +
> +/* sun6i UART transmit character
> + * rb: register which contains the UART base address
> + * rt: register which contains the character to transmit */
> +.macro early_uart_transmit rb, rt
> +        str   \rt, [\rb]            /* -> UART_THR (Transmit Holding
Register) */
> +.endm
> +
> +/*
> + * Local variables:
> + * mode: ASM
> + * indent-tabs-mode: nil
> + * End:
> + */
> diff --git a/xen/include/asm-arm/platforms/sun6i.h
b/xen/include/asm-arm/platforms/sun6i.h
> new file mode 100644
> index 0000000..f910ad6
> --- /dev/null
> +++ b/xen/include/asm-arm/platforms/sun6i.h
> @@ -0,0 +1,28 @@
> +#ifndef __ASM_ARM_PLATFORMS_SUN6I_H
> +#define __ASM_ASM_PLATFORMS_SUN6I_H
> +
> +/* Constants below is only used in assembly because the DTS is not yet
parsed */
> +#ifdef __ASSEMBLY__
> +
> +/* GIC Base Address */
> +#define SUN6I_GIC_BASE_ADDRESS       0x01c82000
> +
> +/* Timer''s frequency */
> +/* \TODO I do not know yet, assume it 24MHz. check it later */
> +#define SUN6I_TIMER_FREQUENCY       (24 * 1000 * 1000) /* 24 MHz */
> +
> +/* Arndale machine ID */
Comment is out of date, but I hope we can avoid needing to know machine
numbers entirely some where down the line. (I expect this will be
clearer after my reply to #4)
> +/* XXX 3894 is downstream number for sun6i. 4137 is upstream number for
sun6i */
> +#define MACH_TYPE_SUN6I             4137
> +
> +#endif /* __ASSEMBLY__ */
> +
> +#endif /* __ASM_ARM_PLATFORMS_SUN6I_H */
> +/*
> + * Local variables:
> + * mode: C
> + * c-file-style: "BSD"
> + * c-basic-offset: 4
> + * indent-tabs-mode: nil
> + * End:
> + */

Ian Campbell

2013-Jun-12 10:17 UTC

head link

Re: [PATCH 3/4] xen/arm: enable early printk for sun6i

On Fri, 2013-05-31 at 20:53 +0800, Bamvor Jian Zhang
wrote:> enable early printk for sun6i base on the macro introduce by previous
> patch.
FWIW I think you could squash them together.
> Signed-off-by: Bamvor Jian Zhang <bjzhang@suse.com>
> ---
>  xen/arch/arm/Rules.mk | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/xen/arch/arm/Rules.mk b/xen/arch/arm/Rules.mk
> index 422ed04..59d29c4 100644
> --- a/xen/arch/arm/Rules.mk
> +++ b/xen/arch/arm/Rules.mk
> @@ -64,6 +64,11 @@ EARLY_PRINTK_INC := pl011
>  EARLY_PRINTK_BAUD := 115200
>  EARLY_UART_BASE_ADDRESS := 0xfff36000
>  endif
> +ifeq ($(CONFIG_EARLY_PRINTK), sun6i)
> +EARLY_PRINTK_INC := sun6i
> +EARLY_PRINTK_BAUD := 115200
Since you assume the bootloader did the init I don''t think you use
this?
> +EARLY_UART_BASE_ADDRESS := 0x01c28000
> +endif
>  
>  ifneq ($(EARLY_PRINTK_INC),)
>  EARLY_PRINTK := y

Ian Campbell

2013-Jun-12 10:17 UTC

head link

Re: [PATCH 4/4] xen/arm: enable switch to hyper mode for sun6i

On Fri, 2013-05-31 at 20:53 +0800, Bamvor Jian Zhang
wrote:> Signed-off-by: Bamvor Jian Zhang <bjzhang@suse.com>
> ---
>  xen/arch/arm/arm32/mode_switch.S | 13 +++++++++++--
>  1 file changed, 11 insertions(+), 2 deletions(-)
> 
> diff --git a/xen/arch/arm/arm32/mode_switch.S
b/xen/arch/arm/arm32/mode_switch.S
> index c92a1cf..0fc26e0 100644
> --- a/xen/arch/arm/arm32/mode_switch.S
> +++ b/xen/arch/arm/arm32/mode_switch.S
> @@ -21,6 +21,7 @@
>  #include <asm/page.h>
>  #include <asm/platforms/vexpress.h>
>  #include <asm/platforms/exynos5.h>
> +#include <asm/platforms/sun6i.h>
>  #include <asm/asm_defns.h>
>  #include <asm/gic.h>
>  
> @@ -90,15 +91,23 @@ enter_hyp_mode:
I''d really like to avoid adding more boards here. Accepting the Arndale
stuff was a pragmatic decision for the Xen 4.3 release but for 4.4 we
need to transition to something better and delete this file completely.

The way we want to fix this is to fix the bootloader, to correctly
launch kernels in NS HYP mode (Andre has patches to u-boot which I think
are upstream already) and to setup the GIC, timers etc for the specific
platform.

Updating the bootloader would be the ideal/preferred way to fix things
but in practice there will be cases where the bootloader cannot be fixed
e.g. because upstream u-boot does not support the platform well, or
because the methods for unbricking the board when things go wrong are
unavailable/unreliable. In those cases we want to remove this code from
Xen by using a boot-wrapper style shim like [0] or [1] which would run
between the bootloader and Xen and fix up all this sort of thing such
that it appears to Xen like it was loaded by a correct bootloader.

In this way we can maintain a clean separation between the main Xen code
base and the hacks required for initialisation on a specific platform.

Of course the bootwrapper can hardcode all sorts of system specific
stuff since it will be expected to be compiled for the specific
platform.

I don''t know which one applies to the sun6i platform -- it looks to me
like upstream u-boot support is well under way and the presence of FEL
mode makes unbricking the system pretty easy (I''ve had to use it on my
own A31 based tablet and it seems OK).

Ian.

[0] http://xenbits.xen.org/gitweb/?p=people/ianc/boot-wrapper.git;a=summary
[1] https://git.linaro.org/gitweb?p=arm/models/boot-wrapper.git;a=summary

>          bic   r0, r0, #0xe           /* Clear EA, FIQ and IRQ */
>          mcr   CP32(r0, SCR)
>  
> -        ldr   r2, =MACH_TYPE_SMDK5250   /* r4 := Arndale machine ID */
>          /* By default load Arndale defaults values */
> +        ldr   r2, =MACH_TYPE_SMDK5250   /* r4 := Arndale machine ID */
>          ldr   r0, =EXYNOS5_TIMER_FREQUENCY  /* r0 := timer''s
frequency */
>          ldr   r1, =EXYNOS5_GIC_BASE_ADDRESS /* r1 := GIC base address */
> -        /* If it''s not the Arndale machine ID, load VE values */
>          teq   r5, r2
> +        beq   board_set_done
> +        /* If it''s not the Arndale machine ID, try sun6i */
> +        ldr   r2, =MACH_TYPE_SUN6I          /* r2 := sun6i machine ID */
> +        ldr   r0, =SUN6I_TIMER_FREQUENCY    /* r0 := timer''s
frequency */
> +        ldr   r1, =SUN6I_GIC_BASE_ADDRESS   /* r1 := GIC base address */
> +        teq   r5, r2
> +        beq   board_set_done
> +        /* If it''s not the sun6i machine ID, load VE values */
>          ldrne r0, =V2M_TIMER_FREQUENCY
>          ldrne r1, =V2M_GIC_BASE_ADDRESS
>  
> +board_set_done:
>          /* Ugly: the system timer''s frequency register is only
>           * programmable in Secure state.  Since we don''t know
where its
>           * memory-mapped control registers live, we can''t find
out the

Bamvor Jian Zhang

2013-Jun-18 06:34 UTC

head link

Re: [PATCH 1/4] xen/arm: introduce Cortex-A7 support

Ian Campbell wrote:> On Fri, 2013-05-31 at 20:53 +0800, Bamvor Jian Zhang wrote:
>> diff --git a/xen/arch/arm/arm32/head.S b/xen/arch/arm/arm32/head.S
>> index 0588d54..d3b849b 100644
>> --- a/xen/arch/arm/arm32/head.S
>> +++ b/xen/arch/arm/arm32/head.S
>> @@ -20,6 +20,7 @@
>>  #include <asm/config.h>
>>  #include <asm/page.h>
>>  #include <asm/processor-ca15.h>
>> +#include <asm/processor-ca7.h>
>>  #include <asm/asm_defns.h>
>>  
>>  #define ZIMAGE_MAGIC_NUMBER 0x016f2818
>> @@ -195,8 +196,18 @@ skip_bss:
>>          /* Is this a Cortex A15? */
>>          ldr   r1, =(CORTEX_A15_ID)
>>          teq   r0, r1
>> +        bne   test_ca7
>>          bleq  cortex_a15_init
>> +        PRINT("- cortex-a15 init done -\r\n")
>> +        b     cpu_init_done
>> +        /* Is this a Cortex A7? */
>> +test_ca7:
>> +        ldr   r1, =(CORTEX_A7_ID)
>> +        teq   r0, r1
>> +        bleq  cortex_a7_init
>> +        PRINT("- cortex-a7 init done -\r\n")
>
> I think now that we have a second processor here it is time to go for a
> more scalable approach. A linker table driven thing, similar to what
> Julien did for the platform (PLATFORM_START et al) or the Linux proc
> table thing seems like the right answer. At the moment it would just
> contain the MIDR mask and match values plus an optional init function
> pointer, in time I suspect it will grow other functionality.got it. how about add something like __lookup_processor_type and proc-xxx.S
like linux kernel.>
>> +cpu_init_done:
>>          /* Set up memory attribute type tables */
>>          ldr   r0, =MAIR0VAL
>>          ldr   r1, =MAIR1VAL
>> diff --git a/xen/arch/arm/arm32/proc-ca7.S
b/xen/arch/arm/arm32/proc-ca7.S
>> new file mode 100644
>> index 0000000..e0c1bc2
>> --- /dev/null
>> +++ b/xen/arch/arm/arm32/proc-ca7.S
>> @@ -0,0 +1,36 @@
>> +/*
>> + * xen/arch/arm/proc-ca7.S
>> + *
>> + * Cortex A7 specific initializations
>> + *
>> + * Bamvor Jian Zhang <bjzhang@suse.com>
>> + * Copyright (c) 2013 SUSE
>> + *
>> + * This program is free software; you can redistribute it and/or
modify
>> + * it under the terms of the GNU General Public License as published
by
>> + * the Free Software Foundation; either version 2 of the License, or
>> + * (at your option) any later version.
>> + *
>> + * This program is distributed in the hope that it will be useful,
>> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
>> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
>> + * GNU General Public License for more details.
>> + */
>> +
>> +#include <asm/asm_defns.h>
>> +#include <asm/processor-ca7.h>
>> +
>> +.globl cortex_a7_init
>> +cortex_a7_init:
>> +        /* Set up the SMP bit in ACTLR */
>> +        mrc   CP32(r0, ACTLR)
>> +        orr   r0, r0, #(ACTLR_CA7_SMP) /* enable SMP bit */
>> +        mcr   CP32(r0, ACTLR)
>> +        mov   pc, lr
>> +
>> +/*
>> + * Local variables:
>> + * mode: ASM
>> + * indent-tabs-mode: nil
>> + * End:
>> + */
>> diff --git a/xen/include/asm-arm/processor-ca7.h
b/xen/include/asm-arm/processor-ca7.h
>> new file mode 100644
>> index 0000000..b773c34
>> --- /dev/null
>> +++ b/xen/include/asm-arm/processor-ca7.h
>> @@ -0,0 +1,25 @@
>> +#ifndef __ASM_ARM_PROCESSOR_CA7_H
>> +#define __ASM_ARM_PROCESSOR_CA7_H
>> +
>> +
>> +#define CORTEX_A7_ID     (0x410FC070)
>> +
>> +/* ACTLR Auxiliary Control Register, Cortex A7 */
>> +#define ACTLR_CA7_DDI                (1<<28)
>> +#define ACTLR_CA7_DDVM               (1<<15)
>> +/* 2 bits */
>> +#define ACTLR_CA7_L1PCTL             (1<<13)
>> +#define ACTLR_CA7_L1RADIS            (1<<12)
>> +#define ACTLR_CA7_L2RADIS            (1<<11)
>> +#define ACTLR_CA7_DODMBS             (1<<10)
>> +#define ACTLR_CA7_SMP                (1<<6)
>
> Is the content of this register really different between A7 and A15?only SMP bit is same. ACTLR is a implementation defined
register.> At least CA7_SMP is the same as CA15_SMP, which suggests we could share
> the init function (to set the SMP bit) and have both a7 and a15 in the
> same proc-v7.S file.Ok. i will add proc-v7.S. although i do not know whether a12 is same or not.
we could deal with that later.>> +
>> +#endif /* __ASM_ARM_PROCESSOR_CA7_H */
>> +/*
>> + * Local variables:
>> + * mode: C
>> + * c-file-style: "BSD"
>> + * c-basic-offset: 4
>> + * indent-tabs-mode: nil
>> + * End:
>> + */

Bamvor Jian Zhang

2013-Jun-18 06:35 UTC

head link

Re: [PATCH 2/4] xen/arm: introduce Allwinner sun6i SOC basic support

Ian Campbell wrote:> On Fri, 2013-05-31 at 20:53 +0800, Bamvor Jian Zhang wrote:
>> Allwinner A31(code name sun6i) is a quad-core Cortex-A7 SOC. Allwinner
has
>> released the source code in http://git.rhombus-tech.net.
>> 
>> This patch define registers and macros for early printk and mode
switch.
>> kick cpu is not implemneted yet. because the secondary cpu boot up
>> sequence is a little bit different from exynos5.
>> 
>> Signed-off-by: Bamvor Jian Zhang <bjzhang@suse.com>
>> ---
>>  xen/arch/arm/arm32/debug-sun6i.inc    | 50
+++++++++++++++++++++++++++++++++++
>>  xen/include/asm-arm/platforms/sun6i.h | 28 ++++++++++++++++++++
>>  2 files changed, 78 insertions(+)
>>  create mode 100644 xen/arch/arm/arm32/debug-sun6i.inc
>>  create mode 100644 xen/include/asm-arm/platforms/sun6i.h
>> 
>> diff --git a/xen/arch/arm/arm32/debug-sun6i.inc
b/xen/arch/arm/arm32/debug-sun6i.inc
>> new file mode 100644
>> index 0000000..eaa1197
>> --- /dev/null
>> +++ b/xen/arch/arm/arm32/debug-sun6i.inc
>> @@ -0,0 +1,50 @@
>> +/*
>> + * xen/arch/arm/arm32/debug-sun6i.inc
>> + *
>> + * Allwinner sun6i specific debug code
>> + *
>> + * Bamvor Jian Zhang <bjzhang@suse.com>
>> + * Copyright (c) 2013 SUSE
>> + *
>> + * This program is free software; you can redistribute it and/or
modify
>> + * it under the terms of the GNU General Public License as published
by
>> + * the Free Software Foundation; either version 2 of the License, or
>> + * (at your option) any later version.
>> + *
>> + * This program is distributed in the hope that it will be useful,
>> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
>> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
>> + * GNU General Public License for more details.
>> + */
>> +
>> +/* sun6i UART initialization
>
> This does appear to use a distinct set of registers from either the
> existing pl011 or exynos4xxx UART driver so it should be a separate
> driver. Do you know if this particular logic block is specific to sn6i
> or is it used elsewhere (e.g. in all sun*i)? Do you know if there is a
> more generic name for the part?sun4i, sun5i use the Synopsys DesignWare 8250 IP. and sun6i is also the
8250 compatiblity one. how about i create a debug-8250.inc and update
it when other 8250 device add.>
> Not a big deal -- we can always rename to be more generic as we see
> other examples of the same UART.
>
>> + * rb: register which contains the UART base address
>> + * rc: scratch register 1 (unused here)
>> + * rd: scratch register 2 (unused here) */
>> +.macro early_uart_init rb, rc, rd
>> +/* assume the uart already init by bootloader */
>> +.endm
>
> I think Andre''s refactoring of the UART init stuff in and around
commit
> 130645eee340608c12b632535d3f983508348ba1 lets you avoid this by just not
> setting EARLY_PRINTK_INIT_UART.yes. I saw Andre'' patch. when i write this piece of code
Andre''s patch is
not commit. i will modify this in next version.>
>> +
>> +/* sun6i UART wait UART to be ready to transmit
>> + * rb: register which contains the UART base address
>> + * rc: scratch register */
>> +.macro early_uart_ready rb, rc
>> +1:
>> +        ldr   \rc, [\rb, #0x7c]     /* <- UART_USR (Status
Register) */
>> +        tst   \rc, #0x2             /* Check BUSY bit */
>> +        beq   1b                    /* Wait for the UART to be ready
*/
>> +.endm
>> +
>> +/* sun6i UART transmit character
>> + * rb: register which contains the UART base address
>> + * rt: register which contains the character to transmit */
>> +.macro early_uart_transmit rb, rt
>> +        str   \rt, [\rb]            /* -> UART_THR (Transmit
Holding Register) */
>> +.endm
>> +
>> +/*
>> + * Local variables:
>> + * mode: ASM
>> + * indent-tabs-mode: nil
>> + * End:
>> + */
>> diff --git a/xen/include/asm-arm/platforms/sun6i.h
b/xen/include/asm-arm/platforms/sun6i.h
>> new file mode 100644
>> index 0000000..f910ad6
>> --- /dev/null
>> +++ b/xen/include/asm-arm/platforms/sun6i.h
>> @@ -0,0 +1,28 @@
>> +#ifndef __ASM_ARM_PLATFORMS_SUN6I_H
>> +#define __ASM_ASM_PLATFORMS_SUN6I_H
>> +
>> +/* Constants below is only used in assembly because the DTS is not yet
parsed */
>> +#ifdef __ASSEMBLY__
>> +
>> +/* GIC Base Address */
>> +#define SUN6I_GIC_BASE_ADDRESS       0x01c82000
>> +
>> +/* Timer''s frequency */
>> +/* \TODO I do not know yet, assume it 24MHz. check it later */
>> +#define SUN6I_TIMER_FREQUENCY       (24 * 1000 * 1000) /* 24 MHz */
>> +
>> +/* Arndale machine ID */
>
> Comment is out of date, but I hope we can avoid needing to know machine
> numbers entirely some where down the line. (I expect this will be
> clearer after my reply to #4)
>
>> +/* XXX 3894 is downstream number for sun6i. 4137 is upstream number
for sun6i */
>> +#define MACH_TYPE_SUN6I             4137
>> +
>> +#endif /* __ASSEMBLY__ */
>> +
>> +#endif /* __ASM_ARM_PLATFORMS_SUN6I_H */
>> +/*
>> + * Local variables:
>> + * mode: C
>> + * c-file-style: "BSD"
>> + * c-basic-offset: 4
>> + * indent-tabs-mode: nil
>> + * End:
>> + */

Bamvor Jian Zhang

2013-Jun-18 06:35 UTC

head link

ReL Re: [PATCH 3/4] xen/arm: enable early printk for sun6i

Ian Campbell wrote:> On Fri, 2013-05-31 at 20:53 +0800, Bamvor Jian Zhang wrote:
>> enable early printk for sun6i base on the macro introduce by previous
>> patch.
>
> FWIW I think you could squash them together.
Ok.>
>> Signed-off-by: Bamvor Jian Zhang <bjzhang@suse.com>
>> ---
>>  xen/arch/arm/Rules.mk | 5 +++++
>>  1 file changed, 5 insertions(+)
>> 
>> diff --git a/xen/arch/arm/Rules.mk b/xen/arch/arm/Rules.mk
>> index 422ed04..59d29c4 100644
>> --- a/xen/arch/arm/Rules.mk
>> +++ b/xen/arch/arm/Rules.mk
>> @@ -64,6 +64,11 @@ EARLY_PRINTK_INC := pl011
>>  EARLY_PRINTK_BAUD := 115200
>>  EARLY_UART_BASE_ADDRESS := 0xfff36000
>>  endif
>> +ifeq ($(CONFIG_EARLY_PRINTK), sun6i)
>> +EARLY_PRINTK_INC := sun6i
>> +EARLY_PRINTK_BAUD := 115200
>
> Since you assume the bootloader did the init I don''t think you use
this?it just a remider for me which the baudrate bootloader used. i will remove
it in next version.>
>> +EARLY_UART_BASE_ADDRESS := 0x01c28000
>> +endif
>>  
>>  ifneq ($(EARLY_PRINTK_INC),)
>>  EARLY_PRINTK := y

Bamvor Jian Zhang

2013-Jun-18 06:38 UTC

head link

Re: [PATCH 4/4] xen/arm: enable switch to hyper mode for sun6i

Ian Campbell wrote:> On Fri, 2013-05-31 at 20:53 +0800, Bamvor Jian Zhang wrote:
>> Signed-off-by: Bamvor Jian Zhang <bjzhang@suse.com>
>> ---
>>  xen/arch/arm/arm32/mode_switch.S | 13 +++++++++++--
>>  1 file changed, 11 insertions(+), 2 deletions(-)
>> 
>> diff --git a/xen/arch/arm/arm32/mode_switch.S
b/xen/arch/arm/arm32/mode_switch.S
>> index c92a1cf..0fc26e0 100644
>> --- a/xen/arch/arm/arm32/mode_switch.S
>> +++ b/xen/arch/arm/arm32/mode_switch.S
>> @@ -21,6 +21,7 @@
>>  #include <asm/page.h>
>>  #include <asm/platforms/vexpress.h>
>>  #include <asm/platforms/exynos5.h>
>> +#include <asm/platforms/sun6i.h>
>>  #include <asm/asm_defns.h>
>>  #include <asm/gic.h>
>>  
>> @@ -90,15 +91,23 @@ enter_hyp_mode:
>
> I''d really like to avoid adding more boards here. Accepting the
Arndale
> stuff was a pragmatic decision for the Xen 4.3 release but for 4.4 we
> need to transition to something better and delete this file completely.
>
> The way we want to fix this is to fix the bootloader, to correctly
> launch kernels in NS HYP mode (Andre has patches to u-boot which I think
> are upstream already) and to setup the GIC, timers etc for the specific
> platform.
>
> Updating the bootloader would be the ideal/preferred way to fix things
> but in practice there will be cases where the bootloader cannot be fixed
> e.g. because upstream u-boot does not support the platform well, or
> because the methods for unbricking the board when things go wrong are
> unavailable/unreliable. In those cases we want to remove this code from
> Xen by using a boot-wrapper style shim like [0] or [1] which would run
> between the bootloader and Xen and fix up all this sort of thing such
> that it appears to Xen like it was loaded by a correct bootloader.
>
> In this way we can maintain a clean separation between the main Xen code
> base and the hacks required for initialisation on a specific platform.
>
> Of course the bootwrapper can hardcode all sorts of system specific
> stuff since it will be expected to be compiled for the specific
> platform.
>
> I don''t know which one applies to the sun6i platform -- it looks
to me
> like upstream u-boot support is well under way and the presence of FEL
> mode makes unbricking the system pretty easy (I''ve had to use it
on my
> own A31 based tablet and it seems OK).i will investigate it. do i have change to got other part of my patch ack
without this bootwrapper or u-boot support? > Ian.
>
> [0] http://xenbits.xen.org/gitweb/?p=people/ianc/boot-wrapper.git;a=summary
> [1] https://git.linaro.org/gitweb?p=arm/models/boot-wrapper.git;a=summary
>
>
>>          bic   r0, r0, #0xe           /* Clear EA, FIQ and IRQ */
>>          mcr   CP32(r0, SCR)
>>  
>> -        ldr   r2, =MACH_TYPE_SMDK5250   /* r4 := Arndale machine ID */
>>          /* By default load Arndale defaults values */
>> +        ldr   r2, =MACH_TYPE_SMDK5250   /* r4 := Arndale machine ID */
>>          ldr   r0, =EXYNOS5_TIMER_FREQUENCY  /* r0 := timer''s
frequency */
>>          ldr   r1, =EXYNOS5_GIC_BASE_ADDRESS /* r1 := GIC base address
*/
>> -        /* If it''s not the Arndale machine ID, load VE values
*/
>>          teq   r5, r2
>> +        beq   board_set_done
>> +        /* If it''s not the Arndale machine ID, try sun6i */
>> +        ldr   r2, =MACH_TYPE_SUN6I          /* r2 := sun6i machine ID
*/
>> +        ldr   r0, =SUN6I_TIMER_FREQUENCY    /* r0 := timer''s
frequency */
>> +        ldr   r1, =SUN6I_GIC_BASE_ADDRESS   /* r1 := GIC base address
*/
>> +        teq   r5, r2
>> +        beq   board_set_done
>> +        /* If it''s not the sun6i machine ID, load VE values
*/
>>          ldrne r0, =V2M_TIMER_FREQUENCY
>>          ldrne r1, =V2M_GIC_BASE_ADDRESS
>>  
>> +board_set_done:
>>          /* Ugly: the system timer''s frequency register is
only
>>           * programmable in Secure state.  Since we don''t know
where its
>>           * memory-mapped control registers live, we can''t
find out the

Bamvor Jian Zhang

2013-Jun-18 06:49 UTC

head link

Re: [PATCH 3/4] xen/arm: enable early printk for sun6i

resend it because of the typo in previous reply

Ian Campbell wrote:> On Fri, 2013-05-31 at 20:53 +0800, Bamvor Jian Zhang wrote:
>> enable early printk for sun6i base on the macro introduce by previous
>> patch.
>
> FWIW I think you could squash them together.
Ok.>
>> Signed-off-by: Bamvor Jian Zhang <bjzhang@suse.com>
>> ---
>>  xen/arch/arm/Rules.mk | 5 +++++
>>  1 file changed, 5 insertions(+)
>> 
>> diff --git a/xen/arch/arm/Rules.mk b/xen/arch/arm/Rules.mk
>> index 422ed04..59d29c4 100644
>> --- a/xen/arch/arm/Rules.mk
>> +++ b/xen/arch/arm/Rules.mk
>> @@ -64,6 +64,11 @@ EARLY_PRINTK_INC := pl011
>>  EARLY_PRINTK_BAUD := 115200
>>  EARLY_UART_BASE_ADDRESS := 0xfff36000
>>  endif
>> +ifeq ($(CONFIG_EARLY_PRINTK), sun6i)
>> +EARLY_PRINTK_INC := sun6i
>> +EARLY_PRINTK_BAUD := 115200
>
> Since you assume the bootloader did the init I don''t think you use
this?it just a remider for me which the baudrate bootloader used. i will remove
it in next version.>
>> +EARLY_UART_BASE_ADDRESS := 0x01c28000
>> +endif
>>  
>>  ifneq ($(EARLY_PRINTK_INC),)
>>  EARLY_PRINTK := y

Ian Campbell

2013-Jun-18 09:24 UTC

head link

Re: [PATCH 1/4] xen/arm: introduce Cortex-A7 support

On Tue, 2013-06-18 at 00:34 -0600, Bamvor Jian Zhang
wrote:> Ian Campbell wrote:
> > On Fri, 2013-05-31 at 20:53 +0800, Bamvor Jian Zhang wrote:
> >> diff --git a/xen/arch/arm/arm32/head.S b/xen/arch/arm/arm32/head.S
> >> index 0588d54..d3b849b 100644
> >> --- a/xen/arch/arm/arm32/head.S
> >> +++ b/xen/arch/arm/arm32/head.S
> >> @@ -20,6 +20,7 @@
> >>  #include <asm/config.h>
> >>  #include <asm/page.h>
> >>  #include <asm/processor-ca15.h>
> >> +#include <asm/processor-ca7.h>
> >>  #include <asm/asm_defns.h>
> >>  
> >>  #define ZIMAGE_MAGIC_NUMBER 0x016f2818
> >> @@ -195,8 +196,18 @@ skip_bss:
> >>          /* Is this a Cortex A15? */
> >>          ldr   r1, =(CORTEX_A15_ID)
> >>          teq   r0, r1
> >> +        bne   test_ca7
> >>          bleq  cortex_a15_init
> >> +        PRINT("- cortex-a15 init done -\r\n")
> >> +        b     cpu_init_done
> >> +        /* Is this a Cortex A7? */
> >> +test_ca7:
> >> +        ldr   r1, =(CORTEX_A7_ID)
> >> +        teq   r0, r1
> >> +        bleq  cortex_a7_init
> >> +        PRINT("- cortex-a7 init done -\r\n")
> >
> > I think now that we have a second processor here it is time to go for
a
> > more scalable approach. A linker table driven thing, similar to what
> > Julien did for the platform (PLATFORM_START et al) or the Linux proc
> > table thing seems like the right answer. At the moment it would just
> > contain the MIDR mask and match values plus an optional init function
> > pointer, in time I suspect it will grow other functionality.
> got it. how about add something like __lookup_processor_type and proc-xxx.S
> like linux kernel.
This sounds like a good plan.
> >> +/* ACTLR Auxiliary Control Register, Cortex A7 */
> >> +#define ACTLR_CA7_DDI                (1<<28)
> >> +#define ACTLR_CA7_DDVM               (1<<15)
> >> +/* 2 bits */
> >> +#define ACTLR_CA7_L1PCTL             (1<<13)
> >> +#define ACTLR_CA7_L1RADIS            (1<<12)
> >> +#define ACTLR_CA7_L2RADIS            (1<<11)
> >> +#define ACTLR_CA7_DODMBS             (1<<10)
> >> +#define ACTLR_CA7_SMP                (1<<6)
> >
> > Is the content of this register really different between A7 and A15?
> only SMP bit is same. ACTLR is a implementation defined register.
I knew that but given the close architectural relationship between the
A7 and the A15 (i.e. they are big.LITTLE compatible with each other) I
wondered if the implementation defined stuff might be similar.
> > At least CA7_SMP is the same as CA15_SMP, which suggests we could
share
> > the init function (to set the SMP bit) and have both a7 and a15 in the
> > same proc-v7.S file.
> Ok. i will add proc-v7.S. although i do not know whether a12 is same or
not.
> we could deal with that later.
Right.

Ian.

Ian Campbell

2013-Jun-18 09:53 UTC

head link

Re: [PATCH 2/4] xen/arm: introduce Allwinner sun6i SOC basic support

On Tue, 2013-06-18 at 00:35 -0600, Bamvor Jian Zhang
wrote:> > This does appear to use a distinct set of registers from either the
> > existing pl011 or exynos4xxx UART driver so it should be a separate
> > driver. Do you know if this particular logic block is specific to sn6i
> > or is it used elsewhere (e.g. in all sun*i)? Do you know if there is a
> > more generic name for the part?
> sun4i, sun5i use the Synopsys DesignWare 8250 IP. and sun6i is also the
> 8250 compatiblity one. how about i create a debug-8250.inc and update
> it when other 8250 device add.
Yes if this is an 8250 compatible implementation the debug-8250.inc is
the right answer.

OOI Did you find the UART easily available on your platform? IOW was
soldering required...
> > Not a big deal -- we can always rename to be more generic as we see
> > other examples of the same UART.
> >
> >> + * rb: register which contains the UART base address
> >> + * rc: scratch register 1 (unused here)
> >> + * rd: scratch register 2 (unused here) */
> >> +.macro early_uart_init rb, rc, rd
> >> +/* assume the uart already init by bootloader */
> >> +.endm
> >
> > I think Andre''s refactoring of the UART init stuff in and
around commit
> > 130645eee340608c12b632535d3f983508348ba1 lets you avoid this by just
not
> > setting EARLY_PRINTK_INIT_UART.
> yes. I saw Andre'' patch. when i write this piece of code
Andre''s patch is
> not commit. i will modify this in next version.
Thanks.
> >
> >> +
> >> +/* sun6i UART wait UART to be ready to transmit
> >> + * rb: register which contains the UART base address
> >> + * rc: scratch register */
> >> +.macro early_uart_ready rb, rc
> >> +1:
> >> +        ldr   \rc, [\rb, #0x7c]     /* <- UART_USR (Status
Register) */
I think this register is a designware extension to the 8250. The Linux
sunxi debug driver is instead checking the UART_LSR for the TEMT or THRE
bits.
> >> +        tst   \rc, #0x2             /* Check BUSY bit */
> >> +        beq   1b                    /* Wait for the UART to be
ready */
> >> +.endm
Ian.

Ian Campbell

2013-Jun-18 09:54 UTC

head link

Re: ReL Re: [PATCH 3/4] xen/arm: enable early printk for sun6i

On Tue, 2013-06-18 at 00:35 -0600, Bamvor Jian Zhang
wrote:> >> @@ -64,6 +64,11 @@ EARLY_PRINTK_INC := pl011
> >>  EARLY_PRINTK_BAUD := 115200
> >>  EARLY_UART_BASE_ADDRESS := 0xfff36000
> >>  endif
> >> +ifeq ($(CONFIG_EARLY_PRINTK), sun6i)
> >> +EARLY_PRINTK_INC := sun6i
> >> +EARLY_PRINTK_BAUD := 115200
> >
> > Since you assume the bootloader did the init I don''t think
you use this?
> it just a remider for me which the baudrate bootloader used.
That''s not unreasonable. You could stick it in a comment:
# 8250 uart configured at 115200 by bootloader

Ian.

Ian Campbell

2013-Jun-18 10:00 UTC

head link

Re: [PATCH 4/4] xen/arm: enable switch to hyper mode for sun6i

On Tue, 2013-06-18 at 00:38 -0600, Bamvor Jian Zhang
wrote:> > I don''t know which one applies to the sun6i platform -- it
looks to me
> > like upstream u-boot support is well under way and the presence of FEL
> > mode makes unbricking the system pretty easy (I''ve had to use
it on my
> > own A31 based tablet and it seems OK).
> i will investigate it. do i have change to got other part of my patch ack
> without this bootwrapper or u-boot support? 
Not sure what you are asking but if it is "can patches #1-#3 be accepted
without this change" then the answer is yes IMHO.

In the meantime while we work on bootwrapper/u-boot support generally we
should also patch this code to not fallback to the vexpress for
everything, if the platform isn''t explicitly either Arndale or vexpress
it should do nothing.

Ian.

Bamvor Jian Zhang

2013-Jun-18 10:23 UTC

head link

Re: [PATCH 2/4] xen/arm: introduce Allwinner sun6i SOC basic support

"Ian Campbell <Ian.Campbell@citrix.com>"> On Tue, 2013-06-18
at 00:35 -0600, Bamvor Jian Zhang wrote:>> > This does appear to use a distinct set of registers from either
the
>> > existing pl011 or exynos4xxx UART driver so it should be a
separate
>> > driver. Do you know if this particular logic block is specific to
sn6i
>> > or is it used elsewhere (e.g. in all sun*i)? Do you know if there
is a
>> > more generic name for the part? 
>> sun4i, sun5i use the Synopsys DesignWare 8250 IP. and sun6i is also the
>> 8250 compatiblity one. how about i create a debug-8250.inc and update 
>> it when other 8250 device add.
>  
> Yes if this is an 8250 compatible implementation the debug-8250.inc is 
> the right answer.
>  
> OOI Did you find the UART easily available on your platform? IOW was 
> soldering required...i use the mele M9 STB which include the same connector of uart as mele
A1000.>> > Not a big deal -- we can always rename to be more generic as we
see
>> > other examples of the same UART. 
>> > 
>> >> + * rb: register which contains the UART base address 
>> >> + * rc: scratch register 1 (unused here) 
>> >> + * rd: scratch register 2 (unused here) */ 
>> >> +.macro early_uart_init rb, rc, rd 
>> >> +/* assume the uart already init by bootloader */ 
>> >> +.endm 
>> > 
>> > I think Andre''s refactoring of the UART init stuff in and
around commit
>> > 130645eee340608c12b632535d3f983508348ba1 lets you avoid this by
just not
>> > setting EARLY_PRINTK_INIT_UART.
>  
>> yes. I saw Andre'' patch. when i write this piece of code
Andre''s patch is
>> not commit. i will modify this in next version.
>  
> Thanks.
>  
>> > 
>> >> + 
>> >> +/* sun6i UART wait UART to be ready to transmit 
>> >> + * rb: register which contains the UART base address 
>> >> + * rc: scratch register */ 
>> >> +.macro early_uart_ready rb, rc 
>> >> +1: 
>> >> +        ldr   \rc, [\rb, #0x7c]     /* <- UART_USR (Status
Register) */
>  
> I think this register is a designware extension to the 8250. The Linux 
> sunxi debug driver is instead checking the UART_LSR for the TEMT or THRE 
> bits.
>  
>> >> +        tst   \rc, #0x2             /* Check BUSY bit */ 
>> >> +        beq   1b                    /* Wait for the UART to
be ready */
>> >> +.endm
>  
> Ian.

Bamvor Jian Zhang

2013-Jun-18 10:39 UTC

head link

Re: [PATCH 4/4] xen/arm: enable switch to hyper mode for sun6i

Ian Campbell wrote:> On Tue, 2013-06-18 at 00:38 -0600, Bamvor Jian Zhang wrote:
> > > I don''t know which one applies to the sun6i platform --
it looks to me
> > > like upstream u-boot support is well under way and the presence
of FEL
> > > mode makes unbricking the system pretty easy (I''ve had
to use it on my
> > > own A31 based tablet and it seems OK).
> 
> > i will investigate it. do i have change to got other part of my patch
ack
> > without this bootwrapper or u-boot support? 
> 
> Not sure what you are asking but if it is "can patches #1-#3 be
accepted
> without this change" then the answer is yes IMHO.
yes, this is what i mean. sorry for the typo.> 
> In the meantime while we work on bootwrapper/u-boot support generally we
> should also patch this code to not fallback to the vexpress for
> everything, if the platform isn''t explicitly either Arndale or
vexpress
> it should do nothing.So, i could check the vexpress machine id in enter_hyp_mode when it is not
Arndale. then raise error and hang if both check fails.> 
> Ian.
> Bamvor

Ian Campbell

2013-Jun-18 10:45 UTC

head link

Re: [PATCH 4/4] xen/arm: enable switch to hyper mode for sun6i

On Tue, 2013-06-18 at 04:39 -0600, Bamvor Jian Zhang
wrote:> Ian Campbell wrote:
> > On Tue, 2013-06-18 at 00:38 -0600, Bamvor Jian Zhang wrote:
> > > > I don''t know which one applies to the sun6i
platform -- it looks to me
> > > > like upstream u-boot support is well under way and the
presence of FEL
> > > > mode makes unbricking the system pretty easy (I''ve
had to use it on my
> > > > own A31 based tablet and it seems OK).
> > 
> > > i will investigate it. do i have change to got other part of my
patch ack
> > > without this bootwrapper or u-boot support? 
> > 
> > Not sure what you are asking but if it is "can patches #1-#3 be
accepted
> > without this change" then the answer is yes IMHO.
> yes, this is what i mean. sorry for the typo.
> > 
> > In the meantime while we work on bootwrapper/u-boot support generally
we
> > should also patch this code to not fallback to the vexpress for
> > everything, if the platform isn''t explicitly either Arndale
or vexpress
> > it should do nothing.
> So, i could check the vexpress machine id in enter_hyp_mode when it is not
> Arndale. then raise error and hang if both check fails.
It should return without doing anything rather than raising an error,
but otherwise yes.

Ian.

Reasonably Related Threads

Search for more apparently analagous threads

Xen devel - May 2013 - [PATCH 0/4] xen/arm: assemble support for Allwinner A31

[PATCH 0/4] xen/arm: assemble support for Allwinner A31

[PATCH 1/4] xen/arm: introduce Cortex-A7 support

[PATCH 2/4] xen/arm: introduce Allwinner sun6i SOC basic support

[PATCH 3/4] xen/arm: enable early printk for sun6i

[PATCH 4/4] xen/arm: enable switch to hyper mode for sun6i

Re: [PATCH 0/4] xen/arm: assemble support for Allwinner A31

Re: [PATCH 0/4] xen/arm: assemble support for Allwinner A31

Re: [PATCH 1/4] xen/arm: introduce Cortex-A7 support

Re: [PATCH 2/4] xen/arm: introduce Allwinner sun6i SOC basic support

Re: [PATCH 3/4] xen/arm: enable early printk for sun6i

Re: [PATCH 4/4] xen/arm: enable switch to hyper mode for sun6i

Re: [PATCH 1/4] xen/arm: introduce Cortex-A7 support

Re: [PATCH 2/4] xen/arm: introduce Allwinner sun6i SOC basic support

ReL Re: [PATCH 3/4] xen/arm: enable early printk for sun6i

Re: [PATCH 4/4] xen/arm: enable switch to hyper mode for sun6i

Re: [PATCH 3/4] xen/arm: enable early printk for sun6i

Re: [PATCH 1/4] xen/arm: introduce Cortex-A7 support

Re: [PATCH 2/4] xen/arm: introduce Allwinner sun6i SOC basic support

Re: ReL Re: [PATCH 3/4] xen/arm: enable early printk for sun6i

Re: [PATCH 4/4] xen/arm: enable switch to hyper mode for sun6i

Re: [PATCH 2/4] xen/arm: introduce Allwinner sun6i SOC basic support

Re: [PATCH 4/4] xen/arm: enable switch to hyper mode for sun6i

Re: [PATCH 4/4] xen/arm: enable switch to hyper mode for sun6i

Reasonably Related Threads