Supporting the Raspberry Pi 5

[rsta2]

There is some initial Circle support for the new Raspberry Pi 5 on the branch rpi5. Please see the file doc/rpi5.txt for more information.

[rsta2]

There are more commits for the Raspberry Pi 5 support on the branch rpi5.

[pottendo]

One thing 'solved': https://www.raspberrypi.com/documentation/computers/raspberry-pi-5.html#uart-connector
The console has changed to a different connector - which is unfortunately not accessible in my configuration, as I'm using a passive cooling solution, where this connector is not accessible... :-(

[pottendo]

...also tried with the newest release of compilers from here: https://developer.arm.com/downloads/-/arm-gnu-toolchain-downloads
gcc version 13.2.1 20231009 (Arm GNU Toolchain 13.2.rel1 (Build arm-13.7))
hosted on a 22.04 Ubuntu (x86) system.

[rsta2]

Thanks for appreciating Circle and for being interested in the Raspberry Pi 5 support.

There have been some recent modifications to the rpi5 branch, which I pushed yesterday. So please be sure to use the latest sources.

I don't know, which firmware you have in your EEPROM. There was an important firmware change recently. Please ensure, that you have the option kernel_address=0x80000 in your config.txt file. Otherwise the kernel image may be loaded to the wrong address and will not work.

Your Config.mk is OK. Please use the GCC-12.2.Rel1-based toolchain. That's the currently recommended one. The three files, you have listed, should be sufficient for the RPi 5 on the SD card. I did not test network boot. I cannot say, if it is working with Circle.

On the RPi 5 the serial console is by default the dedicated UART connector ("ttyS11"). If you want to use the UART at GPIO pins 14/15 ("ttyS1") you have to add the following option to Config.mk:

DEFINE += -DSERIAL_DEVICE_DEFAULT=0

The HDMI screen support in the RPi 5 firmware is not like the support for the earlier models. Many configuration options are missing. Raspberry Pi OS is generating the display output in the Linux kernel now, but Circle still relies on the firmware. Unfortunately some displays do not work any more with it. I don't think, the firmware honors your video= setting in cmdline.txt. So I don't know, how is the situation with your display.

On the RPi 4 and 5 you need the following option in Config.mk, if you want to run a sample program without an attached HDMI screen:

DEFINE += -DSCREEN_HEADLESS

Otherwise the 'CScreenDevice' instance will not initialize (returns FALSE) and the sample will not run.

[pottendo]

Thanks alot for the comprehensive feedback and hints! Much appreciated! I'll keep you informed about my progress!
bye, pottendo

[pottendo]

Hi,
Success!
I think kernel_address=0x80000 was the major hint!
Serial works also again - on pins 14/15.
I managed to merge this into circle-stdlib and even my app starts doing something:

logger: Circle 45.3.1 started on Raspberry Pi 5 (AArch64)
pottendo-kern: Interrupt done
00:00:00.13 timer: SpeedFactor is 7.69
00:00:00.13 pottendo-kern: mTimer done
00:00:00.66 emmc: Found a valid version 3.0x SD card
00:00:00.66 pottendo-kern: EMMC done
00:00:00.67 pottendo-kern: mounted drive: SD:
00:00:00.67 pottendo-kern: I2C done
00:00:00.67 pottendo-kern: pottendo-Pi1541 (0x0)
00:00:00.68 pottendo-log:: Maximum temp set to 85, dynamic adaption possible, curret freq = 1500MHz (max=2400MHz)
...

My screen seems to be not supported on that level... :-( All other PIs are happy, though? Seems Pi5 is a diva?!
Need to debug a bit as ethernet network and USB seems to be not working - I haven't adopted the source yet, and I found some #if RASPPI <= 4 in circle-stdlib; so a bit to do, but in general it seems to work well.

thanks again and happy hacking, pottendo

PS: btw, I'm booting over the network - easier workflow for me!

[rsta2]

You are welcome and thanks for the feedback. It's good, that it is working now so far. I guess, each HDMI display, which needed special settings in config.txt or cmdline.txt on earlier RPi models can have problems on the RPi 5 with Circle. But writing a kernel mode display driver to solve this would be a huge task. Unfortunately Ethernet is not supported yet. The NIC is in the RP1 southbridge now and needs a totally new driver. I had not the time to port it yet. USB should work. If it does not, please give it a try with SD card boot. I also had problems with USB boot here. Perhaps the device is not reset sufficiently to work at the moment.

[pottendo]

ok, will try - though I filed an issue already... sorry to be too quick.

[pottendo]

no, confirmed with SDCard boot:

00:00:00.23 xhci: HW reset failed
00:00:00.23 usb: Cannot initialize xHCI device #0
00:00:00.24 pottendo-kern: USBHCI failed

[aniplay62]

Thanks for your RPI(5) support. I use USPi under FreeRTOS/LwIP platform, updated since long time with your Circle updates (RP3B+, RPI4 and 5). I have also an USB problem on the RPI5 (devices detection problem during boot) who seems not exits when HDMI is not connected. Very strange, for example if RF keyboard works, I need to remove HDMI for get FTDI cable detected.
Else about network support , I added macb driver for LwIP from u-boot sources (polling for rx with a timer), it's easier than Linux sources.

[rsta2]

You are welcome. Yes, there seems to be still a strange problem with the new USB support for the RPi 5 in Circle. Honestly I did most of the USB testing without a HDMI display connected, because my display, which I'm normally using on the RPis does not work with the RPi 5. So your problem report is new to me. I have to test this with the display of my development PC or buy a new HDMI display. I will have a look on the u-Boot sources. Thanks for info.

[rsta2]

I think, I was a little quick with condemning the new Circle USB driver for the Raspberry Pi 5. I did a test with sample/39-umsdplugging and many USB devices now with a HDMI display attached and everything was working well.

@aniplay62 Did the problems, you described, happen with an original Circle sample with the latest source from the rpi5 branch? You wrote, you have ported the USB support to USPi, which is different.

[rsta2]

Unfortunately currently I do not have enough feedback to judge about the quality of the new Circle USB driver for the Raspberry Pi 5. More (also positive) feedback from other users would help.

To make it simple, I attached a binary of the sample/39-umsdplugging with all necessary files to be installed on a SD card or USB flash drive (if USB boot is enabled). This requires a HDMI display connected, which works with the RPi 5. After boot please attach USB flash drives (or other USB devices) you have and see, if they are detected. Thank you in advance!

umsdplugging-rpi5.zip

[aniplay62]

Your lastest change, delay after USB device reset, seems fix the problem. I tried with cold/warn (watchdog) reset. Tried with USPi ;-)

[rsta2]

OK. But in the end this doesn't tell, if the USB support for RPi 5 in native Circle is working well or not. I didn't fully realize in the first read, that you are using a port of the xHCI driver from Circle in your own project. This is out of scope for this discussion board.

[aniplay62]

This is my macb driver from u-boot sources, there are some changes for the RP1 like PHY reset via GPIO and leds settings. I reinit all after unplug/plug the cable. Else I think, it's changes for LwIP.
macb_rp1.txt

[rsta2]

@aniplay62 I think, that's what open source is all about. Thanks again.

[rsta2]

@pottendo I'll need a few days to sort this problem out and to complete this driver.

Yes, such a HDMI display driver would be great, but is probably an immense task.

[rsta2]

@pottendo The driver for the Ethernet NIC of the Raspberry Pi 5 is on the rpi5 branch now. Most of the network samples should work, very few and the docs still need an update (e.g. 06-ethernet, 38-bootloader).

@aniplay62 I used a coherent memory region to place the RX/TX rings and DMA buffers, because I had problems with the way the U-Boot driver was managing the data cache. I did not get it reliable under heavy load.

[pottendo]

Hi,
sorry for the delay - was a bit busy this week.
I can confirm that networking (Ethernet) works. Well done!
Just for curiosity: I assume WiFi is still to be done as it probably again needs a new(er) driver, right?
bye, pottendo

[rsta2]

Thank you for the positive feedback! I will try to support WLAN with the help of @aniplay62.

[aniplay62]

Hi, I have a wifi that works, if my base of work is Zerowi, I use also the Richard Miller's emmc file. I just changed the init routine, I don't know anything, but it works (it's the same chip as the RPI4).

int emmcinit(void)
{
uint32_t clk = (rpi_type < 5) ? rpi_vc_get_clock_rate(RPI_VC_CLOCK_ID_EMMC) : 54000000;
if(clk == 0)
{
clk = Extfreq;
board_printf("emmc: assuming external clock %lu Mhz\r\n", clk/1000000);
}
emmc.extclk = clk;
// board_printf("emmc control %x %x %x\r\n", RD(Control0), RD(Control1), RD(Control2));
WR(Control1, Srsthc); /* Reset the complete host /
delay(10);
while(RD(Control1) & Srsthc);
WR(Control1, Srstdata); / Reset the data */
delay(10);
WR(Control1, 0);
// added for RPI 5
if(rpi_type >= 5)
{
// Enable SD Bus Power VDD1 at 3.3V
uint32_t control0 = RD(Control0);
control0 |= 0x0F << 8;
WR(Control0, control0);
udelay(2000);
}
return(0);
}

[aniplay62]

No., but if FIQ is enabled with some method (<=> RPI4), where to use it?
You cannot redirect all PCIe interrupts on FIQ.
Put a wire on the pins reserved for external reset button, and program an interrupt on this GPIO 2712 pin, why not...

[rsta2]

I understand your problem. But design decisions have to be made, and perhaps they had other applications in mind than you. For example as a desktop PC replacement the RPi 5 is much more suitable than the RPi 4, because it has an immense CPU power. This is also good for synthesizers, which are often programmed using Circle. My goal is to support the RPi 5 as good as I can for the bare-metal applications, which are possible with it.

On the RPi 4 a specific ARM stub was used to have the chance to enter EL3 for FIQ setup. But on the RPi 5 the ARM stub is actually the Trusted Firmware, which is difficult to replace. There is a small chance, that the Trusted Firmware allows setting up the FIQ on the GIC, but I searched for related information and haven't found it so far.

It would be good for example to have the FIQ for the dedicated UART, which is connected to the UART socket. This UART is not in the RP1.

[aniplay62]

I try a also dedicated UART but it's a Zilog Z8523L (a Z85230 in3V3) used for synchronous frames on lot of GPIO pins ;-), GIC is a V2 but something is different with the RPI4, changes GICD_ITARGETSR produce no effects (re-reading is good, but it's always routed to core 0).

[rsta2]

Yes, some kind of the good old Z8530 SCC, I know it. ;)

According to my notes GICD_IGROUPR0 can only be accessed in secure mode, and GICD_CTLR + GICC_CTLR have a different layout in secure mode, which is needed to enable the FIQ. Secure mode is managed by the ARM Trusted Firmware. I guess, for the Raspberry Pi 5 a modified version is used, which is not available in source code for the public.

[aniplay62]

Ok, like me. The first time inside this machine, a
Falcon030, this is my old Falcon tower with a 68060.
Else I think there are no problem with GICD_ITARGETSR, it's my fault (32 bits address used). Sure for FIQ it's another story...

[rsta2]

The Raspberry Pi 5 support has been merged to the develop branch to be part of the next release. The branch rpi5 has been deleted.

[rsta2]

Circle 46 with initial Raspberry Pi 5 support has been released.

[rsta2]

There is initial I2S audio support for the Raspberry Pi 5 on the develop branch. It's currently interrupt driven. Some applications could have problems with the interrupt load, but MiniSynth Pi works and its repo has been updated. I'm still working on the DMA support. Also the currently used I2S clock is not very precise, because the Audio PLL is not supported yet.

[pottendo]

hi,
assuming that with DMA support, also things like:
m_PWMSoundDevice->Playback ((void *)Sample_bin, Sample_bin_size, 1, 8);
may work, I do have a simple testcase in my application.
Looking forward to try!
bye, pottendo

[rsta2]

That would already work with the interrupt-driven support, but the RPi 5 does not have a headphone jack for PWM output. That's why PWM sound output is not a priority. The new driver is for I2S sound, which requires an external interface. They are not expensive, but provide a better sound quality.

[HidenoriMatsubayashi]

Hi, I'm trying circle on RPI5 using raspberry pi probe (UART connector), but I cannot see any UART logs from /dev/ttyACM0. Do I need to add additional settings?

Welcome to minicom 2.8

OPTIONS: I18n 
Port /dev/ttyACM0, 09:55:44

Press CTRL-A Z for help on special keys


RPi: BOOTSYS release VERSION:30cc5f37 DATE: 2024/01/05 TIME: 15:57:40
BOOTMODE: 0x06 partition 0 build-ts BUILD_TIMESTAMP=1704470260 serial 74b7555e boardrev c04170 stc 818651
AON_RESET: 00000003 PM_RSTS 00001000
RP1_BOOT chip ID: 0x20001927
PM_RSTS: 0x00001000
part 00000000 reset_info 00000000
PMIC reset-event 00000000 rtc 00000000 alarm 00000000 enabled 0
uSD voltage 3.3V
Initialising SDRAM 'Samsung' 16Gb x2 total-size: 32 Gbit 4267
DDR 4267 1 0 32 152
RP1_BOOT chip ID: 0x20001927

RP1_BOOT chip ID: 0x20001927
RP1_BOOT: fw size 25968
PCI2 init
PCI2 reset
PCIe scan 00001de4:00000001
RP1_CHIP_INFO 20001927

RPi: BOOTLOADER release VERSION:30cc5f37 DATE: 2024/01/05 TIME: 15:57:40
BOOTMODE: 0x06 partition 0 build-ts BUILD_TIMESTAMP=1704470260 serial 74b7555e boardrev c04170 stc 3738784
AON_RESET: 00000003 PM_RSTS 00001000
M.2 PCIe HAT not detected.
usb_pd_init status 3
USB_PD CONFIG 0 41
Boot mode: SD (01) order f4
SD HOST: 200000000 CTL0: 0x00800000 BUS: 400000 Hz actual: 390625 HZ div: 512 (256) status: 0x1fff0000 delay: 276
SD HOST: 200000000 CTL0: 0x00800f00 BUS: 400000 Hz actual: 390625 HZ div: 512 (256) status: 0x1fff0000 delay: 276
OCR c0ff8000 [73]
CID: 00275048534438474230da82649200fb
CSD: 400e00325b59000039b37f800a400000
SD: bus-width: 4 spec: 2 SCR: 0x02358000 0x01000000
SD HOST: 200000000 CTL0: 0x00800f04 BUS: 50000000 Hz actual: 50000000 HZ div: 4 (2) status: 0x1fff0000 delay: 2
MBR: 0x00002000, 1048576 type: 0x0c
MBR: 0x00102000,14069760 type: 0x83
MBR: 0x00000000,       0 type: 0x00
MBR: 0x00000000,       0 type: 0x00
Trying partition: 0
type: 32 lba: 8192 'mkfs.fat' ' bootfs     ' clusters 261116 (4)
rsc 32 fat-sectors 2040 root dir cluster 2 sectors 0 entries 0
FAT32 clusters 261116
[sdcard] autoboot.txt not found
Select partition rsts 0 C(boot_partition) 0 EEPROM config 0 result 0
Trying partition: 0
type: 32 lba: 8192 'mkfs.fat' ' bootfs     ' clusters 261116 (4)
rsc 32 fat-sectors 2040 root dir cluster 2 sectors 0 entries 0
FAT32 clusters 261116
Read config.txt bytes      683 hnd 0x3f
[sdcard] pieeprom.upd not found
usb_max_current_enable default 0 max-current 3000
Read bcm2712-rpi-5-b.dtb bytes    75197 hnd 0x1048f
dt-match: compatible: raspberrypi,5-model-b match: brcm,bcm2712
dt-match: compatible: brcm,bcm2712 match: brcm,bcm2712
Selecting USB low current limit

[rsta2]

To see the logger output from the Circle sample programs via the UART connector of the Raspberry Pi 5, you have to add the option logdev=ttyS11 to the file cmdline.txt on the SD card. When there is no HDMI display attached, additionally the following option must be added to Config.mk before build:

DEFINE += -DSCREEN_HEADLESS

[HidenoriMatsubayashi]

Thank you for the response.

you have to add the option logdev=ttyS11 to the file cmdline.txt on the SD card.

I tried that, but it didn't work...

Before:

console=serial0,115200 console=tty1 root=PARTUUID=9e9dd766-02 rootfstype=ext4 fsck.repair=yes rootwait splash plymouth.ignore-serial-consoles cfg80211.ieee80211_regdom=GB

After:

console=serial0,115200 console=tty1 logdev=ttyS11 root=PARTUUID=9e9dd766-02 rootfstype=ext4 fsck.repair=yes rootwait splash plymouth.ignore-serial-consoles cfg80211.ieee80211_regdom=GB

[rsta2]

Can you please extract the contents of the attached archive to an empty SD card, connect your debugprobe (via the U socket) to the UART connector, run a terminal program and start the RPi 5 with this SD card. I tested it here using a debugprobe and it works. The RPi 5 blinks 5 times and then generates some log output from Circle sample/04-timer. This is the Config.mk, that I used:

PREFIX64 = aarch64-none-elf-
AARCH = 64
RASPPI = 5
DEFINE += -DSCREEN_HEADLESS
PUTTYSERIALPORT=/dev/ttyACM0
USERBAUD=115200

04-timer-rpi5.zip

[HidenoriMatsubayashi]

04-timer-rpi5.zip worked fine. Thanks a lot! I'll check the differences...

[rsta2]

The I2S DMA support for the Raspberry Pi 5 is on the develop branch. Unfortunately there is still an issue with it. With greater chunk sizes the audio quality is becoming bad for unknown reason. I did many tests, but unfortunately didn't found the reason so far. As a workaround the chunk size is currently limited to a maximum of 128. Smaller chunk sizes may give an even better quality. You can define USE_I2S_SOUND_IRQ to fall back to IRQ support, which does not have this problem.

[rsta2]

The I2S driver is now using the pll_audio as the clock source and the issue, mentioned above, is fixed. With this clock source the timing is much more accurate. The update is on the develop branch.

[rsta2]

PWM sound is implemented on the Raspberry Pi 5 now using the RP1 DMA controller. It's on the develop branch. You need an external adapter like the one, which is described here.

The left channel has to be connected to GPIO12 (SoC numbers), the right channel to GPIO13. Alternatively the left channel can be moved to GPIO18 by defining USE_GPIO18_FOR_LEFT_PWM and the the right channel to GPIO19 by defining USE_GPIO19_FOR_RIGHT_PWM.

[diyelectromusic]

Oh, and for what it's worth, I've an initial test build of MiniDexed for the Pi 5 now too: probonopd/MiniDexed#638

Massive thanks @rsta2 - circle is amazing :)

Kevin

[rsta2]

Thank you for testing this. I'm glad to see MiniDexed working on the RPi 5!

[pottendo]

hi,
great to see MiniDexed - by coincidence I've just plugged a minidexed-setup on my Pi4 - still using the pre-built binary.
However, my Roland FP90 (E-)Piano connected via USB to the Pi seems not to play the right samples - it seems to play a combination of samples, instead what I'm reading on the display; But I'm very new to MiniDexed, so I have to study what I'm doing wrong...

I'm still waiting to get a i2s audio interface for my PIs - will then also try to get it working on Pi5; will let you know.

bye, pottendo

[rsta2]

@pottendo If the problem with MiniDexed remains, you should ask a question in the MiniDexed discussion forum.

[pottendo]

No, there is no problem -I think.
Loaded some other Sound Bands and it sounds like it should.
Still i2s for better Sound quality to be checked.
Thx for the Response, pottendo