Hi Stefan, I copied some macro definitions from freescale SDK, and I have validated that it can kick off the remaining 3 cores. Below is the patched repos/base-hw/src/core/include/spec/imx6/board.h. (Also attached).

/* * \brief Board driver * \author Stefan Kalkowski * \author Martin Stein * \date 2014-02-25 */

/* * Copyright (C) 2014-2015 Genode Labs GmbH * * This file is part of the Genode OS framework, which is distributed * under the terms of the GNU General Public License version 2. */

#ifndef _BOARD_H_ #define _BOARD_H_

#include <spec/imx/board_support.h> #include <spec/cortex_a9/board_support.h>

#define REGS_SRC_BASE (0x020d8000) //!< Base address for SRC. #define HW_SRC_SCR_ADDR (REGS_SRC_BASE + 0x0) #define HW_SRC_GPR3_ADDR (REGS_SRC_BASE + 0x28) #define HW_SRC_GPR5_ADDR (REGS_SRC_BASE + 0x30) #define HW_SRC_GPR7_ADDR (REGS_SRC_BASE + 0x38) #define HW_SRC_SCR (*(volatile hw_src_scr_t *) HW_SRC_SCR_ADDR) #define HW_SRC_GPR3 (*(volatile hw_src_gpr3_t *) HW_SRC_GPR3_ADDR) #define HW_SRC_GPR5 (*(volatile hw_src_gpr5_t *) HW_SRC_GPR5_ADDR) #define HW_SRC_GPR7 (*(volatile hw_src_gpr7_t *) HW_SRC_GPR7_ADDR)

typedef unsigned int reg32_t;

typedef union _hw_src_gpr3 { reg32_t U; struct _hw_src_gpr3_bitfields { unsigned RESERVED0 : 32; //!< [31:0] Read/write bits, for general purpose } B; } hw_src_gpr3_t;

typedef union _hw_src_gpr7 { reg32_t U; struct _hw_src_gpr7_bitfields { unsigned RESERVED0 : 32; //!< [31:0] Read/write bits, for general purpose } B; } hw_src_gpr7_t;

typedef union _hw_src_gpr5 { reg32_t U; struct _hw_src_gpr5_bitfields { unsigned RESERVED0 : 32; //!< [31:0] Read/write bits, for general purpose } B; } hw_src_gpr5_t;

typedef union _hw_src_scr { reg32_t U; struct _hw_src_scr_bitfields { unsigned WARM_RESET_ENABLE : 1; //!< [0] WARM reset enable bit. unsigned SW_GPU_RST : 1; //!< [1] Software reset for gpu unsigned SW_VPU_RST : 1; //!< [2] Software reset for vpu unsigned SW_IPU1_RST : 1; //!< [3] Software reset for ipu1 unsigned SW_OPEN_VG_RST : 1; //!< [4] Software reset for open_vg unsigned WARM_RST_BYPASS_COUNT : 2; //!< [6:5] Defines the ckil cycles to count before bypassing the MMDC acknowledge for WARM reset. unsigned MASK_WDOG_RST : 4; //!< [10:7] Mask wdog_rst_b source. unsigned EIM_RST : 1; //!< [11] EIM reset is needed in order to reconfigure the eim chip select. unsigned SW_IPU2_RST : 1; //!< [12] Software reset for ipu2 unsigned CORE0_RST : 1; //!< [13] Software reset for core0 only. unsigned CORE1_RST : 1; //!< [14] Software reset for core1 only. unsigned CORE2_RST : 1; //!< [15] Software reset for core2 only. unsigned CORE3_RST : 1; //!< [16] Software reset for core3 only. unsigned CORE0_DBG_RST : 1; //!< [17] Software reset for core0 debug only. unsigned CORE1_DBG_RST : 1; //!< [18] Software reset for core1 debug only. unsigned CORE2_DBG_RST : 1; //!< [19] Software reset for core2 debug only. unsigned CORE3_DBG_RST : 1; //!< [20] Software reset for core3 debug only. unsigned CORES_DBG_RST : 1; //!< [21] Software reset for debug of arm platform only. unsigned CORE1_ENABLE : 1; //!< [22] CPU core1 enable. unsigned CORE2_ENABLE : 1; //!< [23] CPU core2 enable. unsigned CORE3_ENABLE : 1; //!< [24] CPU core3 enable. unsigned DBG_RST_MSK_PG : 1; //!< [25] Do not assert debug resets after power gating event of cpu unsigned RESERVED0 : 6; //!< [31:26] Reserved } B; } hw_src_scr_t;

#define HW_SRC_GPR3_WR(v) (HW_SRC_GPR3.U = (v)) #define HW_SRC_GPR5_WR(v) (HW_SRC_GPR5.U = (v)) #define HW_SRC_GPR7_WR(v) (HW_SRC_GPR7.U = (v))

namespace Genode { /** * Board driver */ class Board : public Imx::Board, public Cortex_a9::Board { public: static void secondary_cpus_ip(void * const ip) { HW_SRC_GPR3_WR((uint32_t) ip); HW_SRC_SCR.B.CORE1_ENABLE = 1; HW_SRC_GPR5_WR((uint32_t) ip); HW_SRC_SCR.B.CORE2_ENABLE = 1; HW_SRC_GPR7_WR((uint32_t) ip); HW_SRC_SCR.B.CORE3_ENABLE = 1;

return; }

static bool is_smp(){ return true; } }; }

#endif /* _BOARD_H_ */

Thanks for your amazing work. Best regards, Le Guan

On Fri, Oct 2, 2015 at 8:34 AM, Stefan Kalkowski < stefan.kalkowski@...1...> wrote:

Hi Le Guan,

can you provide the changes you did, so we are able to reproduce it?

I know there are still issues with SMP on Cortex A9 in our current implementation. They are targeted by the work of Martin Stein on issue #1312 in our issue tracker:

https://github.com/genodelabs/genode/issues/1312

However, the provided solution only uses the Pandaboard, and is not yet ready for our master branch. But given your knowledge regarding the scratch register used to provide the entrypoints to the additional cpu cores, we could test it with the Wandboard.

Regards Stefan

On 09/30/2015 12:07 AM, Le Guan wrote:

...

Hello everyone, I wonder if the current release supports SMP in base-hw. I read from the genode book that base-hw has well supported SMP long time ago. But the default configure in wandBoard sets NR_OF_CPUS to 1. When I change that value to 4 and implement several auxiliary SMP functions and run the program, there seems to be some error happen. Below is the output:

core 0 enabled kernel initialized core 1 enabled core 2 enabled core 3 enabled Genode 15.08-25-g5cf8c49 <local changes> int main(): --- create local services --- int main(): --- start init ---

The booting is stuck here. I have implemented the corresponding Cpu::start_secondary_cpus, Cpu::primary_id() and Cpu::executing_id().

What

...

else should I implement to boot SMP?

Thanks in advance! Best regards, Le Guan

---

...

---

genode-main mailing list genode-main@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/genode-main

-- Stefan Kalkowski Genode Labs

http://www.genode-labs.com/ · http://genode.org/

---

---

genode-main mailing list genode-main@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/genode-main

Hi Le Guan,

thank you for your kind words, and also for sharing your insights with us!

Triggered by your mail, I've also played around with the "system reset controller" using the i.MX6 Wandboard on Friday. However, I did not incorporated the results yet. Although the additional cores are booted correctly, our mp_server.run and affinity.run scripts, which test out the SMP functionality, triggered faults on this board. I assume Genode's current master branch is still missing some cache/shareability fixups regarding Cortex A9 CPUs, which is addressed in the issue #1312 that I've mentioned in my last mail.

We definitely will target those issues soon. In the meantime, if you need to use multiple cores on i.MX6, I suggest to look at Martin Stein's issue branch for all changes regarding cache maintainance and shareability attributes.

Kind Regards Stefan

On 10/02/2015 07:52 PM, Le Guan wrote:

Hi Stefan, The last post doesn't render correctly. I copied some macro definitions from freescale SDK, and I have validated that it can kick off the remaining 3 cores. Please find the patched repos/base-hw/src/core/include/spec/imx6/board.h in the attachment.

Thanks for your amazing work. Best regards, Le Guan

On Fri, Oct 2, 2015 at 8:34 AM, Stefan Kalkowski < stefan.kalkowski@...1...> wrote:

...

Hi Le Guan,

can you provide the changes you did, so we are able to reproduce it?

I know there are still issues with SMP on Cortex A9 in our current implementation. They are targeted by the work of Martin Stein on issue #1312 in our issue tracker:

https://github.com/genodelabs/genode/issues/1312

However, the provided solution only uses the Pandaboard, and is not yet ready for our master branch. But given your knowledge regarding the scratch register used to provide the entrypoints to the additional cpu cores, we could test it with the Wandboard.

Regards Stefan

On 09/30/2015 12:07 AM, Le Guan wrote:

...

Hello everyone, I wonder if the current release supports SMP in base-hw. I read from the genode book that base-hw has well supported SMP long time ago. But the default configure in wandBoard sets NR_OF_CPUS to 1. When I change that value to 4 and implement several auxiliary SMP functions and run the program, there seems to be some error happen. Below is the output:

core 0 enabled kernel initialized core 1 enabled core 2 enabled core 3 enabled Genode 15.08-25-g5cf8c49 <local changes> int main(): --- create local services --- int main(): --- start init ---

The booting is stuck here. I have implemented the corresponding Cpu::start_secondary_cpus, Cpu::primary_id() and Cpu::executing_id().

What

...

else should I implement to boot SMP?

Thanks in advance! Best regards, Le Guan

---

...

---

genode-main mailing list genode-main@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/genode-main

-- Stefan Kalkowski Genode Labs

http://www.genode-labs.com/ · http://genode.org/

---

---

genode-main mailing list genode-main@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/genode-main

---

---

genode-main mailing list genode-main@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/genode-main