Fs over hs #50

AndrewCapon · 2023-12-18T06:31:37Z

Yet another go at getting USB Host on Giga reliable.

We now have three defines in USBUserConf.h for testing.

ARC_USB_FULL_SIZE and ARC_FS_OVER_HS are mutually exclusive so be careful not to enable both!

ARC_USB_FULL_SIZE=1 will run the existing code.
ARC_FS_OVER_HS=1 will run the new code, where output transfers are split into packets and input transfers are done full size.

ARC_NO_RESUBMITREQUEST=1 will use the HCCHAR register, which I thought would fix the NAK issue, it doesn't seem to do this in all cases. Specifically setup control data.

ARC_NO_RESUBMITREQUEST=0 will resubmit any out packets, not fully tested with all devices.

So use:

run existing code for testing:

#define ARC_USB_FULL_SIZE (1)
#define ARC_FS_OVER_HS (0)
#define ARC_NO_RESUBMITREQUEST (0)

run new code with HCCHAR changes for testing:

#define ARC_USB_FULL_SIZE (0)
#define ARC_FS_OVER_HS (1)
#define ARC_NO_RESUBMITREQUEST (1)

run new code with retransmit for testing:

#define ARC_USB_FULL_SIZE (0)
#define ARC_FS_OVER_HS (1)
#define ARC_NO_RESUBMITREQUEST (0)

The last case is the one setup by default.

KurtE · 2023-12-18T12:53:31Z

@AndrewCapon @facchinm @mjs513 - I will give this version a try with the different devices we implemented and hopefully everything will work 🤞

Quick note: I will try to see if the issue with the keyboard with a built in hub issue... It may be something similar to the issue, I originally had with it on the Teensy, about the sequences of messages sent tot he hub. More details in the thread:
https://forum.pjrc.com/index.php?threads/usb-host-mouse-driver.45740/page-5#post-155745

AndrewCapon · 2023-12-18T13:09:13Z

I just tried a keyboard with a hub with my multi driver stuff.

Basically the underlying mbed stuff registers a hub, but nothing connected to it. I guess the keyboard is running off of the hub. I tried plugging a memory stick into the keyboard and the underlying mbed stuff did not even register a connection.

I'm guessing Arduino_USBHostMbed5/src/USBHostHub/USBHostHub.cpp is the culprit

KurtE · 2023-12-18T13:19:45Z

Thanks, will take a look, once I have had some more coffee... It is only about 5:15am.

Note: this is an independent issue and should not hold up integration of this PR.

github-actions · 2023-12-18T14:41:38Z

Memory usage change @ ac8c565

Board	flash	%	RAM for global variables	%
`arduino:mbed_giga:giga`	🔺 0 - +64	0.0 - 0.0	0 - 0	0.0 - 0.0
`arduino:mbed_opta:opta`	🔺 0 - +64	0.0 - 0.0	0 - 0	0.0 - 0.0
`arduino:mbed_portenta:envie_m7`	N/A	N/A	N/A	N/A

Click for full report table

Board	`examples/DirList` flash	%	`examples/DirList` RAM for global variables	%	`examples/FileRead` flash	%	`examples/FileRead` RAM for global variables	%	`examples/FileWrite` flash	%	`examples/FileWrite` RAM for global variables	%	`examples/OptaDirList` flash	%	`examples/OptaDirList` RAM for global variables	%	`examples/PortentaOTA` flash	%	`examples/PortentaOTA` RAM for global variables	%
`arduino:mbed_giga:giga`	0	0.0	0	0.0	64	0.0	0	0.0	0	0.0	0	0.0	0	0.0	0	0.0	0	0.0	0	0.0
`arduino:mbed_opta:opta`	0	0.0	0	0.0	64	0.0	0	0.0	0	0.0	0	0.0	64	0.0	0	0.0	64	0.0	0	0.0
`arduino:mbed_portenta:envie_m7`	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A

Click for full report CSV

Board,examples/DirList<br>flash,%,examples/DirList<br>RAM for global variables,%,examples/FileRead<br>flash,%,examples/FileRead<br>RAM for global variables,%,examples/FileWrite<br>flash,%,examples/FileWrite<br>RAM for global variables,%,examples/OptaDirList<br>flash,%,examples/OptaDirList<br>RAM for global variables,%,examples/PortentaOTA<br>flash,%,examples/PortentaOTA<br>RAM for global variables,%
arduino:mbed_giga:giga,0,0.0,0,0.0,64,0.0,0,0.0,0,0.0,0,0.0,0,0.0,0,0.0,0,0.0,0,0.0
arduino:mbed_opta:opta,0,0.0,0,0.0,64,0.0,0,0.0,0,0.0,0,0.0,64,0.0,0,0.0,64,0.0,0,0.0
arduino:mbed_portenta:envie_m7,N/A,N/A,N/A,N/A,N/A,N/A,N/A,N/A,N/A,N/A,N/A,N/A,N/A,N/A,N/A,N/A,N/A,N/A,N/A,N/A

per1234

Fix the typos in comments detected by the repository's automated spell checker as well as some additional typos I identified through a manual review.

per1234 · 2024-01-25T07:29:51Z

src/targets/TARGET_STM/USBEndpoint_STM.cpp

+// This effects the GIGA as it has no external PHY which is required in order to run at HS.
+//
+// 1. Transmitting length larger then packetsize can cause future issues reading, the NAK nightmare.
+// 2. Receiving multiple seperate packets can lead to lost data and the need to retry.
+//
+// So for transmitting we split the data into sizes of <= packet size, we initially set of a single transfer here 


Suggested change

// This effects the GIGA as it has no external PHY which is required in order to run at HS.

//

// 1. Transmitting length larger then packetsize can cause future issues reading, the NAK nightmare.

// 2. Receiving multiple seperate packets can lead to lost data and the need to retry.

//

// So for transmitting we split the data into sizes of <= packet size, we initially set of a single transfer here

// This affects the GIGA as it has no external PHY which is required in order to run at HS.

//

// 1. Transmitting length larger than packetsize can cause future issues reading, the NAK nightmare.

// 2. Receiving multiple separate packets can lead to lost data and the need to retry.

//

// So for transmitting we split the data into sizes of <= packet size, we initially set off a single transfer here

Fix typos in comment

per1234 · 2024-01-25T07:31:34Z

src/targets/TARGET_STM/USBHALHost_STM.cpp

+// This effects the GIGA as it has no external PHY which is required in order to run at HS.
+//
+// 1. Transmitting length larger then packetsize can cause future issues reading, the NAK nightmare.
+// 2. Receiving multiple seperate packets can lead to lost data and the need to retry.
+//
+// So for transmitting we split the data into sizes of <= packet size and individually transfer them, only calling pPriv->transferCompleted() when all data is sent.
+// For receiving we receive all the data and call pPriv->transferCompleted().
+//
+// If we get NAKS on the control OUT EPs we make sure that USB_OTG_HCCHAR_CHDIS ans USB_OTG_HCCHAR_CHENA are reset to the correct values.


Suggested change

// This effects the GIGA as it has no external PHY which is required in order to run at HS.

//

// 1. Transmitting length larger then packetsize can cause future issues reading, the NAK nightmare.

// 2. Receiving multiple seperate packets can lead to lost data and the need to retry.

//

// So for transmitting we split the data into sizes of <= packet size and individually transfer them, only calling pPriv->transferCompleted() when all data is sent.

// For receiving we receive all the data and call pPriv->transferCompleted().

//

// If we get NAKS on the control OUT EPs we make sure that USB_OTG_HCCHAR_CHDIS ans USB_OTG_HCCHAR_CHENA are reset to the correct values.

// This affects the GIGA as it has no external PHY which is required in order to run at HS.

//

// 1. Transmitting length larger than packetsize can cause future issues reading, the NAK nightmare.

// 2. Receiving multiple separate packets can lead to lost data and the need to retry.

//

// So for transmitting we split the data into sizes of <= packet size and individually transfer them, only calling pPriv->transferCompleted() when all data is sent.

// For receiving we receive all the data and call pPriv->transferCompleted().

//

// If we get NAKS on the control OUT EPs we make sure that USB_OTG_HCCHAR_CHDIS and USB_OTG_HCCHAR_CHENA are reset to the correct values.

per1234 · 2024-01-25T07:31:59Z

src/targets/TARGET_STM/USBHALHost_STM.cpp

+    if ((endpointType == EP_TYPE_INTR)) 
+    {
+      // Disable the channel interrupt and retransfer below
+      // TOD: really this retransfer should be queued up and transfered on the SOF interupt based on interval from descriptor.


Suggested change

// TOD: really this retransfer should be queued up and transfered on the SOF interupt based on interval from descriptor.

// TOD: really this retransfer should be queued up and transferred on the SOF interrupt based on interval from descriptor.

per1234 · 2024-01-25T07:34:16Z

src/targets/TARGET_STM/USBHALHost_STM.cpp

+						uint32_t tmpreg;
+						tmpreg = USBx_HC(uChannel)->HCCHAR;
+						tmpreg &= ~USB_OTG_HCCHAR_CHDIS;
+						tmpreg |= USB_OTG_HCCHAR_CHENA;
+						USBx_HC(uChannel)->HCCHAR = tmpreg;


Suggested change

uint32_t tmpreg;

tmpreg = USBx_HC(uChannel)->HCCHAR;

tmpreg &= ~USB_OTG_HCCHAR_CHDIS;

tmpreg |= USB_OTG_HCCHAR_CHENA;

USBx_HC(uChannel)->HCCHAR = tmpreg;

uint32_t tmpreg;

tmpreg = USBx_HC(uChannel)->HCCHAR;

tmpreg &= ~USB_OTG_HCCHAR_CHDIS;

tmpreg |= USB_OTG_HCCHAR_CHENA;

USBx_HC(uChannel)->HCCHAR = tmpreg;

Use spaces instead of tabs for indentation, as is done throughout the rest of the file.

per1234 · 2024-01-25T07:34:30Z

src/targets/TARGET_STM/USBHALHost_STM.cpp

+
+        case URB_DONE:
+        {
+          // first calculate how much we transferred, HAL metrics aonly work for IN EPs


Suggested change

// first calculate how much we transferred, HAL metrics aonly work for IN EPs

// first calculate how much we transferred, HAL metrics only work for IN EPs

AndrewCapon added 3 commits December 15, 2023 11:58

Initial version

3672be0

Fix interupt EP tranfer len issue

f6d051a

Add ARC_NO_RESUBMITREQUEST, if 0 will resbmit transfers on NAK

ac8c565

AndrewCapon mentioned this pull request Dec 18, 2023

Fix USB Host issues #42

Merged

per1234 suggested changes Jan 25, 2024

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fs over hs #50

Fs over hs #50

AndrewCapon commented Dec 18, 2023 •

edited

Loading

KurtE commented Dec 18, 2023 •

edited

Loading

AndrewCapon commented Dec 18, 2023

KurtE commented Dec 18, 2023

github-actions bot commented Dec 18, 2023

per1234 left a comment

per1234 Jan 25, 2024

per1234 Jan 25, 2024

per1234 Jan 25, 2024

per1234 Jan 25, 2024

per1234 Jan 25, 2024

	// TOD: really this retransfer should be queued up and transfered on the SOF interupt based on interval from descriptor.
	// TOD: really this retransfer should be queued up and transferred on the SOF interrupt based on interval from descriptor.

	// first calculate how much we transferred, HAL metrics aonly work for IN EPs
	// first calculate how much we transferred, HAL metrics only work for IN EPs

Fs over hs #50

Are you sure you want to change the base?

Fs over hs #50

Conversation

AndrewCapon commented Dec 18, 2023 • edited Loading

KurtE commented Dec 18, 2023 • edited Loading

AndrewCapon commented Dec 18, 2023

KurtE commented Dec 18, 2023

github-actions bot commented Dec 18, 2023

per1234 left a comment

Choose a reason for hiding this comment

per1234 Jan 25, 2024

Choose a reason for hiding this comment

per1234 Jan 25, 2024

Choose a reason for hiding this comment

per1234 Jan 25, 2024

Choose a reason for hiding this comment

per1234 Jan 25, 2024

Choose a reason for hiding this comment

per1234 Jan 25, 2024

Choose a reason for hiding this comment

AndrewCapon commented Dec 18, 2023 •

edited

Loading

KurtE commented Dec 18, 2023 •

edited

Loading