I can not run rtems 5.1 smp correctly on bsp xilinx-zynqmp

[smallphd at aliyun.com]

I think the key problem is the bsp_start_hook_0 function which will be executed in all cores.
However, this function does not distinguish each core. It means all four cores run the same code.
The right procedure is core 0 run the init code while other cores wait for it to complete.


BSP_START_TEXT_SECTION void bsp_start_hook_0(void)
{
  uint32_t sctlr_val;

  sctlr_val = arm_cp15_get_control();
  sctlr_val |= ARM_CP15_CTRL_CP15BEN;
  arm_cp15_set_control( sctlr_val );

  /*
   * Current U-boot loader seems to start kernel image
   * with I and D caches on and MMU enabled.
   * If RTEMS application image finds that cache is on
   * during startup then disable caches.
   */
  if ( sctlr_val & (ARM_CP15_CTRL_I | ARM_CP15_CTRL_C | ARM_CP15_CTRL_M ) ) {
    if ( sctlr_val & (ARM_CP15_CTRL_C | ARM_CP15_CTRL_M ) ) {
      /*
       * If the data cache is on then ensure that it is clean
       * before switching off to be extra carefull.
       */
      arm_cp15_data_cache_clean_all_levels();
    }
    arm_cp15_flush_prefetch_buffer();
    sctlr_val &= ~ ( ARM_CP15_CTRL_I | ARM_CP15_CTRL_C | ARM_CP15_CTRL_M | ARM_CP15_CTRL_A );
    arm_cp15_set_control( sctlr_val );
  }
  arm_cp15_instruction_cache_invalidate();
  /*
   * The care should be taken there that no shared levels
   * are invalidated by secondary CPUs in SMP case.
   * It is not problem on Zynq because level of coherency
   * is L1 only and higher level is not maintained and seen
   * by CP15. So no special care to limit levels on the secondary
   * are required there.
   */
  arm_cp15_data_cache_invalidate_all_levels();
  arm_cp15_branch_predictor_invalidate_all();
  arm_cp15_tlb_invalidate();
  arm_cp15_flush_prefetch_buffer();
}



smallphd at aliyun.com
 
From: Kinsey Moore
Date: 2020-10-06 21:56
To: joel at rtems.org; smallphd at aliyun.com
CC: Gedare Bloom; devel
Subject: RE: Re: I can not run rtems 5.1 smp correctly on bsp xilinx-zynqmp
Is it possible that the application was only started on the first core instead of all cores? Is it possible to check the execution state of the other cores?
 
Kinsey
 
From: Joel Sherrill <joel at rtems.org> 
Sent: Tuesday, October 6, 2020 08:11
To: smallphd at aliyun.com
Cc: Gedare Bloom <gedare at rtems.org>; devel <devel at rtems.org>; Kinsey Moore <kinsey.moore at oarcorp.com>
Subject: Re: Re: I can not run rtems 5.1 smp correctly on bsp xilinx-zynqmp
 
Maybe Kinsey has an idea. 
 
On Tue, Oct 6, 2020 at 1:29 AM smallphd at aliyun.com <smallphd at aliyun.com> wrote:
The board is Ultra96 board with JTAG boot.
And yes, there is a bspsmp.c in rtems-5.1\bsps\arm\xilinx-zynqmp\start\
 


smallphd at aliyun.com
 
From: Gedare Bloom
Date: 2020-10-06 00:16
To: smallphd at aliyun.com
CC: devel
Subject: Re: I can not run rtems 5.1 smp correctly on bsp xilinx-zynqmp
It should, I believe. The bsp has bspsmp.c file.
 
How did you configure (../rtems/configure)?
 
What board/target do you run it on?
 
On Mon, Oct 5, 2020 at 2:24 AM smallphd at aliyun.com <smallphd at aliyun.com> wrote:
> 
> Hi, all
> I compile rtems 5.1 with bsp xilinx-zynqmp. The single core mode is ok.
> But I can not use 2 or 3 or 4 cores of this bsp.
> After analysing the source code, I found there is no code to address smp condition. Does rtems 5.1 indeed not support smp mode for xilinx-zynqmp bsp?
> 
> ________________________________
> smallphd at aliyun.com
> _______________________________________________
> devel mailing list
> devel at rtems.org
> http://lists.rtems.org/mailman/listinfo/devel
_______________________________________________
devel mailing list
devel at rtems.org
http://lists.rtems.org/mailman/listinfo/devel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rtems.org/pipermail/devel/attachments/20201009/0d3e5893/attachment-0001.html>