On Sun, Feb 21, 2021 at 5:08 PM Robert Middleton <osfan6...@gmail.com> wrote: > > > I don't have an actual Microchip Explorer16/32 which it sounds like > > you are working on, but if you were to put a draft PR up or point me > > at a branch I would be happy to try to run this against the simulator > > and see if we can get a little further. > > > > This is the simulator I was using https://github.com/sergev/qemu/wiki > > I did have to make a few changes to get it to build with a modern > > gcc/glibc > > > > This may be a typo on your side, but I'm testing with an MX, not an > MZ. I did just try it with that simulator, and I get an error about > writing to peripheral register 1f800600 being not supported.
That is TIMER1 I just wrote a patch for that. I'll try and get that in a little better place. > > Here's the branch that I'm working off of at the moment. It also has > a fix for defining CONFIG_SERIAL_TERMIOS with the PIC32, which doesn't > properly compile: > https://github.com/rm5248/incubator-nuttx/tree/pic-messing-around > > You should be able to get the proper configuration by doing > ./tools/configure.sh -E -l pic32mx-explorer16:nsh. Note that the > explorer16 has the output on UART2 - I don't remember if you can > switch which UART qemu is using to see that output. Switching it to > UART1 for the simulator should be fine though. Ok running this in the sim I was getting the init thread failing to be crated, so I enabled debug asserts and get this: ❯ ./mipsel-softmmu/qemu-system-mipsel -machine pic32mx7-explorer16 -nographic -monitor none -serial stdio -kernel ~/nuttx/wrk/nuttx/nuttx.hex -S -s Board: Microchip Explorer16 Processor: M4K RAM size: 128 kbytes Load file: '/home/bashton/nuttx/wrk/nuttx/nuttx.hex', 250820 bytes nx_start: Entry up_assert: Assertion failed at file:mm_heap/mm_initialize.c line: 82 up_dumpstate: sp: a0001b14 up_dumpstate: stack base: a0001c0c up_dumpstate: stack size: 00001000 up_stackdump: a0001b00: 9d03ad90 a0001b24 9d03a9bc a0001b00 a0001b00 a0001b1c 9d00cecc a0001b14 up_stackdump: a0001b20: a0001c0c 9d03adf0 00001000 a00003fc a0001b14 a0001c0c 00001000 a0001b44 up_stackdump: a0001b40: 9d00b990 00000000 9d03a9bc 9d03a9e8 9d03a104 00000052 00000000 a0001b64 up_stackdump: a0001b60: 9d006260 9d03a104 00000052 a0008061 9d0326cc a0001b7c 9d008d1c 9d03a104 up_stackdump: a0001b80: 00000052 bf8860c0 a0001b8c a00009a4 00000000 00000001 9d008700 a0001ba4 up_stackdump: a0001ba0: 9d008f78 a00009a4 a0001c0c 0001e3f4 bf8860c0 0000000c a0001bbc a0001bc4 up_stackdump: a0001bc0: 9d008bb4 a00009a4 a0001c0c 0001e3f4 00000003 a0001bdc 9d000380 a0001c0c up_stackdump: a0001be0: 0001e3f4 ffffffff 00000008 00000010 00000000 00000000 a000038c a0001c0c up_stackdump: a0001c00: 0001e3f4 00000000 9d000020 00000000 00000000 00000000 00000000 00000000 Dropping the breakpoint on _assert we can then see exactly what is going on here: Breakpoint 1, _assert (filename=0x9d03a104 "mm_heap/mm_initialize.c", linenum=82) at assert/lib_assert.c:36 36 up_assert(filename, linenum); (gdb) bt #0 _assert (filename=0x9d03a104 "mm_heap/mm_initialize.c", linenum=82) at assert/lib_assert.c:36 #1 0x9d008d1c in mm_addregion (heap=0xa00009a4 <g_mmheap>, heapstart=0xa0001c0c, heapsize=123892) at mm_heap/mm_initialize.c:82 #2 0x9d008f78 in mm_initialize (heap=0xa00009a4 <g_mmheap>, heapstart=0xa0001c0c, heapsize=123892) at mm_heap/mm_initialize.c:203 #3 0x9d008bb4 in umm_initialize (heap_start=0xa0001c0c, heap_size=123892) at umm_heap/umm_initialize.c:85 #4 0x9d000380 in nx_start () at init/nx_start.c:552 #5 0x9d000020 in __start_nuttx () at chip/pic32mx_head.S:582 Backtrace stopped: frame did not save the PC (gdb) up #1 0x9d008d1c in mm_addregion (heap=0xa00009a4 <g_mmheap>, heapstart=0xa0001c0c, heapsize=123892) at mm_heap/mm_initialize.c:82 82 DEBUGASSERT(heapsize <= MMSIZE_MAX + 1); (gdb) p heapsize $1 = 123892 (gdb) Then looking at the definition of MMSIZE_MAX we see "include/nuttx/mm/mm.h" 509L #ifdef CONFIG_MM_SMALL typedef uint16_t mmsize_t; # define MMSIZE_MAX UINT16_MAX #else typedef uint32_t mmsize_t; # define MMSIZE_MAX UINT32_MAX #endif Your configuration is using CONFIG_MM_SMALL so heap could not be allocated correctly and we would likely get all kinds of unexpected behavior. Resolving that then takes us to another assert failure: #5 0x9d00cea8 in up_irqinitialize () at chip/pic32mx_irq.c:116 116 up_prioritize_irq(PIC32MX_IRQSRC_CS0, (CHIP_SW0_PRIORITY << 2)); (gdb) down #4 0x9d00d4ec in up_prioritize_irq (irq=129, priority=24) at chip/pic32mx_irq.c:476 476 DEBUGASSERT((unsigned)irq < NR_IRQS && (unsigned)(priority >> 2) > 0); This now starts to explain why the context switch is not happening and up_swint0 is not triggered up_prioritize_irq should be called with the real interrupt number not the offset used when enabling and disabling interrupts: ===include/pic32mx/irq_5xx6xx7xx.h==== #define PIC32MX_IRQ_U5 51 /* Vector: 51, UART5 */ #define PIC32MX_IRQ_BAD 52 /* Not a real IRQ number */ #define NR_IRQS 52 /* Interrupt numbers. These should be used for enabling and disabling * interrupt sources. Note that there are more interrupt sources than * interrupt vectors and interrupt priorities. An offset of 128 is * used so that there is no overlap with the IRQ numbers and to avoid * errors due to misuse. */ #define PIC32MX_IRQSRC0_FIRST (128+0) #define PIC32MX_IRQSRC_CT (128+0) /* Vector: 0, Core Timer Interrupt */ So this change should resolve that: commit e4439d51d07fbc1357deb06dea6913ee0e3a92ab (HEAD -> pic-messing-around) Author: Brennan Ashton <bash...@brennanashton.com> Date: Sun Feb 21 19:54:58 2021 -0800 Fix context switch bug for pic32mx diff --git a/arch/mips/src/pic32mx/pic32mx_irq.c b/arch/mips/src/pic32mx/pic32mx_irq.c index a8df13ee79..6ac35e3aa5 100644 --- a/arch/mips/src/pic32mx/pic32mx_irq.c +++ b/arch/mips/src/pic32mx/pic32mx_irq.c @@ -113,7 +113,7 @@ void up_irqinitialize(void) /* Set the Software Interrupt0 to a special priority */ - up_prioritize_irq(PIC32MX_IRQSRC_CS0, (CHIP_SW0_PRIORITY << 2)); + up_prioritize_irq(PIC32MX_IRQ_CS0, (CHIP_SW0_PRIORITY << 2)); /* Set the BEV bit in the STATUS register */ We now see the shell: (note I did disable most of the debug features you had enabled like the ones for memory managment) ❯ ./mipsel-softmmu/qemu-system-mipsel -machine pic32mx7-explorer16 -nographic -monitor none -serial stdio -kernel ~/nuttx/wrk/nuttx/nuttx.hex Board: Microchip Explorer16 Processor: M4K RAM size: 128 kbytes Load file: '/home/bashton/nuttx/wrk/nuttx/nuttx.hex', 250312 bytes nx_start: Entry uart_register: Registering /dev/console uart_register: Registering /dev/ttyS0 uart_register: Registering /dev/ttyS1 nx_start_application: Starting init thread NuttShell (NSH) NutnxtX_s-1tart0: .0.1 nsh> CPU0: Beginning Idle Loop nsh> uname -a NuttX 10.0.1 198649273f-dirty Feb 21 2021 18:53:48 mips pic32mx-starterkit Hopefully this was helpful and gets you unblocked. --Brennan
