eLinux.org does not support HTTP any more and uses HTTPS instead, which is not supported by Circle. Because this sample tries to fetch a page using HTTP from this server, a HTTP error 301 (Moved permanently) will be reported. Unfortunately there is currently no solution on this.

Hello, I'm having a problem while testing with raspberry pi 3.
The blink example works and i can see the led blinking, but as soon i insert a serial code, the rasp hangs, also if i try to run other examples i receive nothing at the serial. Could it be a different bootcode problem? If so, could you post the boot binaries tou used to perform test with raspberry pi 3? Thanks

//
// kernel.cpp
//
// Circle - A C++ bare metal environment for Raspberry Pi
// Copyright (C) 2014-2016  R. Stange <[email protected]>
// 
// 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 3 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.
//
// You should have received a copy of the GNU General Public License
// along with this program.  If not, see <http://www.gnu.org/licenses/>.
//
#include "kernel.h"
#include <circle/gpiopin.h>
#include <circle/timer.h>

CKernel::CKernel (void)
{
}

CKernel::~CKernel (void)
{
}

boolean CKernel::Initialize (void)
{
	return TRUE;
}

TShutdownMode CKernel::Run (void)
{
	CGPIOPin AudioLeft (41, GPIOModeOutput);
	CGPIOPin AudioRight (40, GPIOModeOutput);

	// flash the Act LED 10 times and click on audio (3.5mm headphone jack)
	for (unsigned i = 1; i <= 10; i++)
	{
		m_ActLED.On ();
		AudioLeft.Invert ();
		AudioRight.Invert ();
		CTimer::SimpleMsDelay (200);

		m_ActLED.Off ();
		CTimer::SimpleMsDelay (500);
	}

	return ShutdownReboot;
}

Hi. I'm trying to run file access example. It keeps saying that "partition not found". I wonder why there is no ListDevices method in CDeviceNameService. It could help me to find correct device name.

Andrzej

Hi. As you probably know the default speed settings for raspberry zero are slightly higher than settings for raspberry 1. Since you use fixed values for both raspberry 1 and zero i had to make some modifications in circle core to change speed settings:

1. make CCPUThrottle::SetClockRate public accessible

	CCPUThrottle::SetClockRate(CLOCK_ID_CORE, 340 * 1000 * 1000, true);
	CCPUThrottle::SetClockRate(CLOCK_ID_ARM, 1000 * 1000 * 1000, true);

#define CLOCK_ID_SDRAM		8

	CCPUThrottle::SetClockRate(CLOCK_ID_SDRAM, 450 * 1000 * 1000, true);

2. add voltage related tag to property tags:

#define PROPTAG_GET_VOLTAGE	    	0x00030003
#define PROPTAG_SET_VOLTAGE	    	0x00038003

struct TPropertyTagVoltage
{
	TPropertyTag	Tag;
	u32		nVoltageId;
	#define VOLTAGE_ID_CORE		1
	#define VOLTAGE_ID_SDRAM_C	2
	#define VOLTAGE_ID_SDRAM_P	3
	#define VOLTAGE_ID_SDRAM_I	4
	u32		value;
};

struct TPropertyTagSetVoltage
{
	TPropertyTag	Tag;
	u32		nVoltageId;
	u32		value;
};

TagVoltage.nVoltageId = VOLTAGE_ID_CORE;
TagVoltage.value = 1350000; //1,35V
if (Tags.GetTag(PROPTAG_SET_VOLTAGE, &TagVoltage, sizeof TagVoltage, 4))
{
	Message.Format ("CORE Voltage %d %d\r\n", TagVoltage.nVoltageId, TagVoltage.value);
	m_Screen.Write ((const char *) Message, Message.GetLength ());		//return 0;
}

Are you planning to implement such routines in nearest future?

I'm trying to attach an interrupt handler to the rising edge of a GPIO pin. My current test setup is a push button with a pulldown resistor attached to GPIO pin 26 of a Raspberry Pi Zero. Pushing the button toggles the pin from LOW to HIGH, which I can verify with a multimeter and a logic analyzer.

The problem ist that the interrupt handler is being called all the time. Even if I connect GPIO 26 directly to GND, this behavior doesn't change. The symptom is a continuous stream of "Interrupt rising edge" messages on the console.

The code is derived from the code in softserial.cpp. The following is the relevant portion of a minimal test example (full project available if necessary). Is there a problem with the code, or do I have a hardware problem?

CKernel::CKernel (void)
:	m_Screen (m_Options.GetWidth (), m_Options.GetHeight ()),
	m_Timer (&m_Interrupt),
	m_Logger (m_Options.GetLogLevel (), &m_Timer),
	m_GPIOManager (&m_Interrupt)
{
}

CKernel::~CKernel (void)
{
}

bool CKernel::Initialize (void)
{
	bool bOK = true;

	if (bOK)
	{
		bOK = m_Screen.Initialize ();
	}

	if (bOK)
	{
		CDevice *pTarget = m_DeviceNameService.GetDevice (m_Options.GetLogDevice (), FALSE);
		if (pTarget == 0)
		{
			pTarget = &m_Screen;
		}

		bOK = m_Logger.Initialize (pTarget);
	}

	if (bOK)
	{
		bOK = m_Interrupt.Initialize ();
	}

	if (bOK)
	{
		bOK = m_Timer.Initialize ();
	}

	if (bOK)
	{
		bOK = m_GPIOManager.Initialize ();
	}

	return bOK;
}

void CKernel::InterruptHandler (void)
{
	m_Logger.Write("CKernel::InterruptHandler", LogNotice, "Interrupt rising edge");
}

void CKernel::InterruptStub (void *pParam)
{
	CKernel *pThis = (CKernel *) pParam;
	assert (pThis != 0);

	pThis->InterruptHandler ();
}

TShutdownMode CKernel::Run (void)
{
	m_Logger.Write (FromKernel, LogNotice, "Compile time: " __DATE__ " " __TIME__);

	CGPIOPin clockPin(26, GPIOModeInput, &m_GPIOManager);

	clockPin.ConnectInterrupt (InterruptStub, this);
	clockPin.EnableInterrupt (GPIOInterruptOnRisingEdge);

	while (true);

	return ShutdownReboot;
}

Keyboard is recognized as standard int3-1-1 device, but the key presses aren't taken correctly. Sometimes, when a modifier key is pressed, the event pair press/release are shown seconds later from the real key press.

Reported by José Luis Sánchez

Can Circle be used with QEMU?

Hi, I studied your implementation of scheduler.
Compare to the usual fork-and-join approach, your scheduler is kind of an polling based approach that the scheduler is activated on yield and run the next possible task.

Does my statement correctly describe your scheduler's behavior?

One limitation for this event based approach does not use multicores to explore true parallelism?

I have submitted pull request #19 to fix some minor code issues reported by clang. An additional notable warning reported by clang which probably should be investigated is the following:

bcmframebuffer.cpp:93:36: warning: comparison of constant 256 with expression of type 'u8' (aka 'unsigned char') is always true [-Wtautological-constant-out-of-range-compare]
    if (m_pInfo->Depth <= 8 && nIndex < 256)
                               ~~~~~~ ^ ~~~

As nIndex is declared as u8 in the parameter list of CBcmFrameBuffer::SetPalette, the comparison nIndex < 256 is always true.

When a sample that uses timers is loaded using u-boot, the execution fails. The reason is that u-boot changes the CP15 register VBAR (section 3.2.43, p.253 from ARM1176JZF-S TRM). I'm using a PI B+, but the same register is at ARMv7 & ARMv8.
Setting this CP15 register to zero, resolves the problem.

Btw, u-boot modifies too the bit-1 "Strict Alignment Fault Checking" in CP15 register c1,c0, 0 (section 3.2.7, p.176 from ARM1176JZF-S TRM) setting to "enabled". I can't say now how impact this in Circle.

Best regards ;)

Is there any plan to support lastest Raspbeery Pi 2? It will be worthy if it is possible.

I would like to use the PWMSounddevice with the RPI Zero but this requires the standard pins (40,etc) to be changed because the Zero doesn't have those audio pins available. I did see you had a quick solution to this in the Oct 1, 2017 commit, but this does not seem to work (anymore).

Any pointers on how this can be done currently?

Windows 8.1 Mingw environment,

Rules.mk CFLAGS may cause an compile error like the following.

arm-none-eabi-g++ -march=armv5 -mtune=arm1176jzf-s -Wall -Wno-psabi -fsigned-char -fno-builtin -nostdinc -nostdlib -undef -D__circle__ -I ../include -I ../addon -O -fno-exceptions -fno-rtti -std=c++0x -c -o synchronize.o synchronize.cpp
C:\Users\Miso\AppData\Local\Temp\ccBIvGL4.s: Assembler messages:
C:\Users\Miso\AppData\Local\Temp\ccBIvGL4.s:27: Error: selected processor does not support ARM mode cpsid i' C:\Users\Miso\AppData\Local\Temp\ccBIvGL4.s:79: Error: selected processor does not support ARM modecpsie i'
make: *** [synchronize.o] Error 1

I could fix the compile error with change of CFLAGS as the following.

CFLAGS += -march=armv6j -mfloat-abi=hard -mtune=arm1176jzf-s -Wall -Wno-psabi -fsigned-char -fno-builtin -nostdinc -nostdlib
-undef -D__circle__ -I $(CIRCLEHOME)/include -I $(CIRCLEHOME)/addon -O #-DNDEBUG

There will be no more compile error on Windows.

In CGPIOPin::Write you use PeripheralEntry and PeripheralExit to protect the memory-write that toggles the GPIO-Pin.
Is that really necessary? Because on "Real-Linux" one can change the GPIO state with a simple memory write without needing any special barriers? Or are the barriers provided by the Linux kernel in that case and hidden away from the C code running on Linux? (I would be quite surprised about that, though.)

Hi,

Is there any interest in porting (some of) these bits to the even-more-bare-metal VC4 side? Some things might be pretty useful there, I think, and could probably be made to work (i.e. "blobless") maybe almost just by tweaking base addresses and recompiling.

Hello and thank you for your project. I am using your library to characterize the power/energy behavior of the raspPI3b. I am trying to disable the prefetching unit of the CPU. According to the arm manual I need to set a value to CPUACTLR when processor is idle. I tried to add code in startup.S which reads CPUACTLR and then writes back the same value to the register. However the booting freezes. The code is added after the VBAR is reseted in startup.S and looks like :

mrrc    p15, 0, r2, r3, c15 
mcrr    p15, 0, r2, r3, c15  

Could u please provide me some guidance on how to do it?

Hello and congratulations for your awesome project. I am testing the 26-cpustress sample on a raspberry Pi 3b. I am setting all the compilation options for this platform and I compile successfully the 26th sample. The single threaded/CPU version is executing correctly. Afterwards I follow the instructions and i define ARM_ALLOW_MULTI_CORE in include/circle/sysconfig.h. I re-complie all the libraries and re-execute the sample. I am not sure, but i think that everything is executed on top of CPU 0. Am I missing something? Is there any implementation in which all 4 cores of my PI simultaneously execute a part of the computation (Simillar to the POSIX pthreads)? Thank you.

Hi,

The CString::Format method don't add the needed '0' with a format as "%04d". With "%04u" the zero filler is formatted as expected.

Both Circle and Circle64 have the code for CursorLeft missing in lib/screen.cpp

lib/screen.cpp

void CScreenDevice::Write (char chChar)
{
...
        case ScreenStateBracket:
...
                case 'C':
                        CursorRight ();
                        m_nState = ScreenStateStart;
                        break;

                case 'D': // Added by AMR, missing in original
                        CursorLeft ();
                        m_nState = ScreenStateStart;
                        break;

                case 'H':
                        CursorHome ();
                        m_nState = ScreenStateStart;
                        break;
...
}

This is more a question than a bug report: I wonder if you tried to use the board as USB-device (but not as host like in your examples) in order to e.g. let it appear as serial interface on a connected host PC (USB CDC device implementation).

Anyway: well done, this is a lot of useful code!

Hello
First Of all, thanks for this awesome codes, I would Like to Display Farsi Text On the Screen.
I found a Font.h file, I converted a farsi Font Like "B nazanin" to the bitmap hex file, and replaced it, but it did not work, Do you have any Idea or solution for this?

Thank You so much
Niloofar

When a bare image is loaded using u-boot the TLB cache maintains contents that generate problems when MMU is reenabled by Circle. This can be tested using sample 03-screentext, that shows strange lines over the text at random places, and last lines completely shattered when the screen resolution is low, as 640x480. When the MMU is disabled (calling memory(FALSE) in kernel.cpp constructor) all screen is showed correctly.

As specified in ARM Cortex-A Series Programmer's Guide (DEN0013D), section 8.8, after disable MMU is needed to invalidate L1-Cache, TLB and branch predictor (point 2 from sample code). For this issue the relevant action is invalidate TLB, that resolves the problem.

I've modified the file startup.S as:

mov r0, #0
mcr p15, 0, r0, c12, c0, 0 /* reset VBAR (if changed by u-boot) /
mcr p15, 0, r0, c7, c5, 0 / Invalidate Instruction Cache /
mcr p15, 0, r0, c7, c5, 6 / Invalidate branch prediction array /
mcr p15, 0, r0, c8, c7, 0 / Invalidate entire Unified Main TLB /
mcr p15, 0, r0, c7, c5, 4 / need to be changed for ISB instruction on PI-2/3 */

Hi,

While it's freshly new, this project seems really active. I think he could be a game changer in the Raspberry Pi's bare-metal world because it is more concrete than other projects I found.

dwelch67 opened the way with all these great "how-to" examples but circle is more about making all theses peripherals work together rather than one by one, right ?

I think encouraging the open source community to contribute is a must, 'cause I would to, but I don't want to mess up your plans. That's the reason why I think doing a list of priority features in your README (or through issues?) is important. I don't want to fork to make my own version, but rather contribute to make this one it more useful.

BTW, keep it up, this repository is full of interesting things for everyone (And I strongly consider myself part of "everyone" 😄)

http://www.airspayce.com/mikem/bcm2835/

BCM2835 C library has good SPI fuction.
But currently Circle does not provide SPI function at this moment.
Is there any plan to support SPI?

I'm just curious: Would it be possible to make the circle socket api look like normal sockets, so one can use a default C library that uses sockets and it would work out of the box?

Probably a lot of work, though.

When doing my research on FIQ interrupt handlers i found some information, that one can use an __attribute__ ((interrupt ("FIQ"))) or __attribute__ ((interrupt ("IRQ"))) to tell the compiler that a function should be used as an interrupt handler.
I don't see any information or usage of this in circle. Is there a special reason for this?

This is more a question than an issue.

I'm trying to process signals that have a 1 MHz clock. My ultimate goal is to simulate a device that is attached to the bus of a 6502 processor.

At the moment I'm trying to mirror the 1 MHz clock signal from one input GPIO pin to an output GPIO pin with the following code:

kernel.h:

#ifndef _kernel_h
#define _kernel_h

#include <circle/memory.h>
#include <circle/actled.h>
#include <circle/koptions.h>
#include <circle/devicenameservice.h>
#include <circle/screen.h>
#include <circle/serial.h>
#include <circle/exceptionhandler.h>
#include <circle/interrupt.h>
#include <circle/timer.h>
#include <circle/logger.h>
#include <circle/types.h>
#include <circle/gpiomanager.h>

enum TShutdownMode
{
	ShutdownNone,
	ShutdownHalt,
	ShutdownReboot
};

class CKernel
{
public:
	CKernel (void);
	~CKernel (void);

	bool Initialize (void);

	TShutdownMode Run (void);

private:
	// do not change this order
	CMemorySystem		m_Memory;
	CKernelOptions		m_Options;
	CDeviceNameService	m_DeviceNameService;
	CScreenDevice		m_Screen;
	CExceptionHandler	m_ExceptionHandler;
	CInterruptSystem	m_Interrupt;
	CTimer				m_Timer;
	CLogger				m_Logger;
	CGPIOManager		m_GPIOManager;

	void InterruptHandler (void);
	static void InterruptStub (void *pParam);

	CGPIOPin clockPin;

	bool volatile state;
};

#endif

kernel.cpp:

#include "kernel.h"

#include <assert.h>

static const char FromKernel[] = "kernel";

CKernel::CKernel (void)
:	m_Screen (m_Options.GetWidth (), m_Options.GetHeight ()),
	m_Timer (&m_Interrupt),
	m_Logger (m_Options.GetLogLevel (), &m_Timer),
	m_GPIOManager (&m_Interrupt),
	clockPin(26, GPIOModeInput, &m_GPIOManager),
	state(false)
{
}

CKernel::~CKernel (void)
{
}

bool CKernel::Initialize (void)
{
	bool bOK = true;

	if (bOK)
	{
		bOK = m_Screen.Initialize ();
	}

	if (bOK)
	{
		CDevice *pTarget = m_DeviceNameService.GetDevice (m_Options.GetLogDevice (), FALSE);
		if (pTarget == 0)
		{
			pTarget = &m_Screen;
		}

		bOK = m_Logger.Initialize (pTarget);
	}

	if (bOK)
	{
		bOK = m_Interrupt.Initialize ();
	}

	if (bOK)
	{
		bOK = m_Timer.Initialize ();
	}

	if (bOK)
	{
		bOK = m_GPIOManager.Initialize ();
	}

	return bOK;
}

void CKernel::InterruptHandler (void)
{
	state = clockPin.Read();
}

void CKernel::InterruptStub (void *pParam)
{
	CKernel *pThis = (CKernel *) pParam;
	assert (pThis != 0);

	pThis->InterruptHandler ();
}

TShutdownMode CKernel::Run (void)
{
	m_Logger.Write (FromKernel, LogNotice, "Compile time: " __DATE__ " " __TIME__);
	m_Logger.Write (FromKernel, LogNotice, "Connecting interrupts to clock pin");

	clockPin.ConnectInterrupt (InterruptStub, this);
	clockPin.EnableInterrupt (GPIOInterruptOnRisingEdge);
	clockPin.EnableInterrupt2 (GPIOInterruptOnFallingEdge);

	unsigned const clockMirrorPin = 10;
	m_Logger.Write (FromKernel, LogNotice, "Mirroring clock input to GPIO pin %u", clockMirrorPin);
	CGPIOPin clockOutPin(clockMirrorPin, GPIOModeOutput);

	m_Logger.Write (FromKernel, LogNotice, "Looping...");

	bool oldState = state;
	while (true) {
		bool const newState = state;

		if (newState != oldState) {
			if (newState) {
				clockOutPin.Write(HIGH);
			}
			else {
				clockOutPin.Write(LOW);
			}
			oldState = newState;
		}
	}

	return ShutdownReboot;
}

I use a logic analyzer to compare the input and the output signal, and the output signal does not keep up with the input signal, e.g. it misses arbitrary transitions of the input clock signal.

I also have version of the code that switches the output pin directly from the interrupt handling routine, and that bypasses the search for the right interrupt callback in CGPIOManager::InterruptHandler(), because I know that only one specific pin triggers an interrupt, but that is not much better, and the code is too large to post here.

So my question is, is it realistic trying to implement this with circle at all, or should I bury this idea?

Thank you

Circle should get multi-core support on the Raspberry Pi 2. This will take some time to complete.

Hi,

I forked your repository and have been compiling with the arm-none-eabi- toolchain to put the circle bare metal OS on my Raspberry Pi. I have been trying to experiment with writing to a 1 GB USB (FAT 32) drive plugged directly into the USB port on my Pi. After using make in the ../sample/15_files/ directory, the code successfully boots on the Pi, but hangs after the "Compile time" line that is printed to the monitor. I added an "else" to the "if" statement in kernel.cpp that checks to see if the partition is found and it is successfully being found. It seems that the partition is not being mounted properly. Do you have any idea why the code does not get past the mounting "if" statement? Should I be modifying the defined name of the partition at the beginning of kernel.cpp?

I've attached a screen shot of the modified code (2 else statements added). The code never reaches the second if statement. Any suggestions would be greatly appreciated!

According to this there is a new revision 1.2 of the Raspberry Pi 2 out which is based on the BCM2837 SoC (not on BCM2836 any more). The BCM2837 was previously used on the Raspberry Pi 3 only.

The Raspberry Pi 2 v1.2 is currently not supported by Circle. Because I do not own such device I cannot test it with Circle and cannot get it working.

Do I understand it correctly that currently it is not possible to connect a GPIO pin via class CGPIOPin to both a falling-edge and rising-edge interrupt? I looked at the "BCM2835 ARM Peripherals" document, and I think it should be possible technically to do that.

To me it looks like the buffer size is pretty much ignored when calling CSocket::Receive. The only thing it does is overwrite the number of received bytes in case less than nLength bytes were received?

So i cannot really trust the number of received bytes and also cannot trust receive to not write beyond the provided pBuffer-Pointer?

Is this intentional? Am I missing something here?

Pwmsound works fine. But USB module and Pwmsound makes confilct. It is very easy to dectect. Adding pwnsound function into USB sample code. Sound makes tick noise recursively to the end of play.

Hello,

I am dong some baremetal development for the Raspberry Pi and could find as good as no documentation in the public domain about the USB core except for the base address. I see that you have implemented USB support so I have a few questions:

1. Is there a standard register structure for USB devices or is it specific to implementations?
2. Where can I find the usb register descriptions for the BCM2835?
3. I am trying to develop a very basic usb driver which does not have to be generic at all. Let us just say that I want to use a keyboard as a fancy keypad connected over USB. Are there enough Infos available (I mean without having to reverse engineer code myself) in the public domain for me to even attempt this or should I just drop the idea?

Many thanks,
Szoshi

I want to use math.h in Circle.
So I tried to add the library but "No such file or directory" error is occured.
How I add more libraries in the sources?

Hi, awesome work! Tested Mouse and Keyboard and it works great.
There is no support for gamepads ?
Thanks for sharing!

hi
i try:

C:\Users\test>tftp -i 169.254.0.111 GET kernel.img kernel.img

I see on circle serial console:
00:12:25.68 tftpd: Transfer timed out
00:12:52.97 tftpd: Incoming read request from 169.254.229.229
00:13:37.98 tftpd: Transfer timed out
00:13:37.98 tftpd: Incoming write request from 169.254.229.229
00:13:43.04 tftpd: Transfer timed out
00:13:43.21 tftpd: Incoming write request from 169.254.229.229
00:13:48.22 tftpd: Transfer timed out

Hello and thank you for your project. I am using your library to characterize the power/energy behavior of the raspPI3b. I am trying to disable the prefetching unit of the CPU. According to the arm manual I need to set a value to CPUACTLR when processor is idle. I tried to add code in startup.S which reads CPUACTLR and then writes back another value to the register. However when i read and print the register the resutls are the same.

mrrc    p15, 0, r2, r3, c15 
bic r2, r2, #28672
orr r2, r2, #253755392
mcrr    p15, 0, r2, r3, c15  

Could u please provide me some guidance on how to do it?

Hi Rene,

From last firmware update, the secondary CPUs don't start correctly. The reason is this change:

raspberrypi/tools#76

I've applied this small modification to multicore.cpp:

write32 (ARM_LOCAL_MAILBOX3_SET0 + 0x10 * nCore, (u32) &_start_secondary);
asm volatile ("sev");

So, the CPUs are starting again.

Best regards

Hi!

I've come across some memory leak problem in web server implementation.
I can reproduce problem on example 21-webserver, on raspberry pi2.
I first tried with ARM_ALLOW_MULTI_CORE + REALTIME settings in sysconfig.h,
then with ARM_ALLOW_MULTI_CORE only setting, then with default settings, problem remain the same.
What I do is flooding webserver with following command from my pc:
curl -s "http://192.168.1.118?[1-10000]"
What happens is that always after 457 successful replies raspberry hangs.

I did some rough debugging, found that:

• with mem_info() I see that things get allocated but never freed, it looks like any connection is taking about 1 block of 262144, 2 blocks of 4096, 1 block of 1024;

• with TCP_DEBUG I've seen that connections stay stuck in FIN-WAIT-1 state, as if TCPTimerTimeWait never get called, so state never goes further to CLOSING and tcpconnection never get deleted;

• what is strange is that this happens only if I continuously do "fast" writing on screen (without scrolling): I continuously call "\b[H" sequence to go home with screen cursor, then call mem_info(); this results in seeing very fast update of memory consumption; in this conditions, example 21 hangs with even 1 connection, and connection last seen state is FIN-WAIT-1;

• if I remove the mem_info() thing, but keep the TCP_DEBUG thing, I see from the tcp log that 1 connection correctly terminates (goes to CLOSED state, and then I see "Deleted TCB" message); a connection flood anyway terminates with some not-well defined hang;

That's what I've seen so far, hope it is useful :)
I may try some further debugging next days...

I've found the following code (at https://github.com/rsta2/circle/blob/master/lib/bcmpropertytags.cpp#L94):

    u32 *pEndTag = (u32 *) (pBuffer->Tags + nTagSize);
    *pEndTag = PROPTAG_END;

#if RASPPI != 3
    CleanDataCache ();
    DataSyncBarrier ();
#endif

    u32 nBufferAddress = GPU_MEM_BASE + (u32) pBuffer;
    if (m_MailBox.WriteRead (nBufferAddress) != nBufferAddress)
    {
        return FALSE;
    }

#if RASPPI != 3
    InvalidateDataCache ();
    DataSyncBarrier ();
#else
    DataMemBarrier ();
#endif

I understand why you clear the data-cache and what the barriers are for, what I don't understand is why you don't clear the datacache and do the dsb when it's a rpi3 and after the writeread there's also a difference between the models. I think it's because of the different arm-cores, but what is the important point to know about the difference between the models regarding dmb, dsb and cachebehavior?

Hello René,

I'm testing the new code in develop branch. I've found as issue with the LED handling with a keyb Logitech K120 keyboard that not exist when I use a HP keyb (USB 2.0).

I've found another problem with new gamepad code while using a Sony DualShock Sixaxis for PS3 (original, not clone). Both errors are shown at boot:

00:00:00.71 usbdev0-1: Device ven424-9514, dev9-0-2 found
00:00:00.72 usbdev0-1: Interface int9-0-1 found
00:00:00.72 usbdev0-1: Function is not supported
00:00:00.73 usbdev0-1: Interface int9-0-2 found
00:00:00.73 usbdev0-1: Using device/interface int9-0-2
00:00:02.94 usbdev0-1: Device ven424-ec00 found
00:00:02.94 usbdev0-1: Using device/interface ven424-ec00
00:00:03.11 usbdev1-2: Device ven46d-c31c found
00:00:03.12 usbdev1-2: Interface int3-1-1 found
00:00:03.12 usbdev1-2: Using device/interface int3-1-1
00:00:03.13 usbdev1-2: Interface int3-0-0 found
00:00:03.13 usbdev1-2: Using device/interface int3-0-0
00:00:03.30 usbdev1-3: Device ven46d-c03e found
00:00:03.30 usbdev1-3: Interface int3-1-2 found
00:00:03.31 usbdev1-3: Using device/interface int3-1-2
00:00:03.49 usbdev1-5: Device ven54c-268 found
00:00:03.49 usbdev1-5: Interface int3-0-0 found
00:00:03.50 usbdev1-5: Using device/interface int3-0-0
00:00:03.55 smsc951x: MAC address is B8:27:EB:F6:EA:B3
00:00:03.56 usbhub: Port 1: Device configured
00:00:03.62 dwhci: Transaction failed (status 0xA)
00:00:03.63 usbpad: Cannot get HID report descriptor
00:00:03.63 usbhub: Port 2: Device configured
00:00:03.69 usbhub: Port 3: Device configured
00:00:03.75 dwhci: Transaction failed (status 0xA)
00:00:03.75 usbpad: Cannot get HID report descriptor
00:00:03.76 usbhub: Port 5: Cannot configure device
00:00:03.76 dwroot: Device configured

I have connected Logitech keyb, a mouse and PS3 DualShock.

Hi. Sorry for my bad english.

I would like to use CSPIMaster together with CSPIMasterDMA. I want to send few bytes synchronously (CSPIMaster) than much more bytes asynchronously (CSPIMasterDMA). I manager to do this by changing

	write32 (ARM_SPI0_CS,   (read32 (ARM_SPI0_CS) & ~CS_CS)
			      | (nChipSelect << CS_CS__SHIFT)
			      | CS_CLEAR_RX | CS_CLEAR_TX
			      | CS_TA);

to

	write32 (ARM_SPI0_CS,   (read32 (ARM_SPI0_CS) & ~CS_CS **& ~CS_DMAEN & ~CS_ADCS**)
			      | (nChipSelect << CS_CS__SHIFT)
			      | CS_CLEAR_RX | CS_CLEAR_TX
			      | CS_TA);

Do you think it's correct way to do this? Using DMA and callbacks for sending 1 or 2 bytes is very inconvienent. Should i expect any issues caused by switching between DMA and synchronous mode ?

Thank you.

Hi,

I hope this is the correct way to ask a question like this.

I am making games using this library and I would like to be able to have the game wait for the vertical sync, to avoid tearing. I've read up a bit on the subject and found two potential ways to do this.

1. Use the Set virtual offset tag, as discussed here. < https://www.raspberrypi.org/forums/viewtopic.php?t=19073 > So I've tried calling CBcmFrameBuffer::SetVirtualOffset and it returns with no error (TRUE) but the game is still not waiting for VSync.

So then I found this.

2. Enable fake_vsync_irq in config.txt and listen to IRQ 48, as discussed here < raspberrypi/firmware#67 > So I tried connecting an IRQ handler like so:
   m_Interrupt.ConnectIRQ( ARM_IRQ_SMI, IRQParty, (void*)0 );
   But for some reason that function never returns. If I add logging before and after that function call I see that the log message before the call is printed but not the message after.

My IRQ handler looks like this:

volatile uint32 testValue;

void IRQParty( void* _pData )
{
	testValue++;
	write32( 0x20600000, 0 );
}

Soooooo I guess my question is, how can I wait for vsync using the Circle library?

Thank you very much for an awesome library. Super easy to use and very lightweight! My game is running within a second after the power switch is flipped, rather than first having to wait for ages for Raspbian to boot up. That is fantastic!

Hi.

I am executing a cpu heavy application which uses circle. The application iteratively performs some logistics ( reading inputs, allocating memory) and then it turns into a computative intensive benchmark. I want to boot raspberry PI with a low frequency ( 600Mhz ) perform all the logistics in the low frequency ( 600Mhz) and then turn frequency on the Max ( 1200 Mhz).
On a pseudo code this looks like :

Boot....
Application:
while ( True ){
setFreqToLow(600)
PerformLogistics()
setFreqToHight(1200)
PerformComputations()
}

When I boot PI using the fast option it boots with the high frequency ( 1200 ) and then it toggles between 600Mhz-1200 Mhz normally. When I dont use the "fast" option the PI boots on the low (600Mhz ) frequency but it cannot set the high frequency afterwards. The CCPUThrottle::SetSpeed(CPUSpeedMaximum,True) returns "0".

Any idea why this happens? Thank you very much in advance.

I am currently using the socket api by Circle.

On the Raspi side, I opened a socket and created a TCP connection with a server that is running the "SimpleHTTPServer" module in Python.
I was able to make a HTTP GET request to ask for a file from this server, and receive it completely.

This works for files smaller than ~2MB.

When I try to receive a file that is 4MB, it will pause after receiving 1408 blocks of the file (each block is 1460 bytes). I have a loop that keeps calling receive() and the ending condition is whether or not its return value is 0.

When I check on the server side, it says that the connection is closed by the remote host. Therefore the connection was closed by the Raspi side.

Do you have any clue why this is happening?

Thanks in advance.

When the sample code starts, you can ping the pi. Between five to twenty seconds after the start, all freezes, no keyboard response, no mouse response and no net response. Sometimes, freezes immediately, sometimes freezes after one minute. I've found that increment the timer interrupt from 100 to 200 or 250 ints/sec, mitigates the problem.

Reported by @jsanchezv

Hi Rene,

I'm working again with Circle and have some questions that I want comment with you.

I'm using the CPWMSoundDevice class and, examining his code, you say that supports 12 bit samples at 44.1 sampling frequency. But in code I can see:

assert (nBitsPerSample == 8 || nBitsPerSample == 16);

so, the accepted sizes are 8 & 16. I would prefer to send 12-bit samples and the driver don't need resample on-the-fly. If I'm not wrong, the sampling freq. is configured with:

write32 (ARM_PWM_RNG1, PWM_RANGE);
write32 (ARM_PWM_RNG2, PWM_RANGE);

at the Run method. PWM_RANGE is defined as:

#define PWM_RANGE (CLOCK_FREQ / CLOCK_DIVIDER / SAMPLE_RATE)

so, the PWM_RANGE for 44.1 kHz is programmed with value 5668 because the result is truncated to lower integer. With this number as divisor, the resulting freq is 44107.26888 kHz. With 5669 could be 44099.48845, near from the intended sampling freq than 44107.

Anyway, would it be possible to change the sampling freq without recompiling the library?

My problem is that I need to generate the samples on-the-fly, and a frequency as 44099.48 is a problem to generate the samples. Trying different numbers, I've found two divisors interesting to me, that generates freqs. of 25 kHz and 31.25 kHz. I don't know when is possible to change the #define and expect the generated freq. would be the desired. I don't understand very well the MASH filters principle, and I depend of a correct freq to synchronize another things.

At another hand, could exist a method to know the PWM state without polling the driver? (using a callback from DMACompletionRoutine when nTransferSize = 0?, i.e.). I'm calling repeatedly to Playback method but that's very sensible to USB transfers (I guess) and it's easy to hear clicks!, pops!, and all sort of dirty sounds. The worst part is with a PI B+, that can't copes with all the work to be done.

It's very important to my app that can be a reliable method to generate the sound samples and, probably, use the PWM as a fine clock.

Best regards

First of all thanks for providing this huge set of examples.

I'm facing problem while doing floating point arithmetic operations on global variables. The error log is given below,
kernel.o: In function std::sin(float)': /usr/arm-none-eabi/include/c++/6.3.1/cmath:449: undefined reference to sinf'
kernel.o: In function std::tan(float)': /usr/arm-none-eabi/include/c++/6.3.1/cmath:506: undefined reference to tanf'
kernel.o: In function std::sqrt(float)': /usr/arm-none-eabi/include/c++/6.3.1/cmath:487: undefined reference to sqrtf'
../../Rules.mk:74: recipe for target 'kernel7.img' failed
make: *** [kernel7.img] Error 1

If I declare the variables in local scope or as static in global scope then the bare metal program compiles properly.
Do you have any idea why this linker issue are persisting?

Hi, I'm trying to use circle with my raspberry 3.
I could compile the 01 sample (arm-none-eabi- as PREFIX, RASPPI=3 in Rules.mk).
I have copied the boot files in the sd, and the kernel.img generated by gcc.
Unfortunately I can't see anything happening (if I understand correctly, the ACT led is the one near the power led, green).
Am I doing something wrong? How can I know?