chore(usb): cdc_queue now use uint32_t length by catbraincell · Pull Request #2550 · stm32duino/Arduino_Core_STM32

catbraincell · 2024-11-04T17:33:56Z

raise cdc buffer size 32KB limit due to uint16_t length variables.

This PR fixes bug: cdc queue limited
Now CDC_TRANSMIT_QUEUE_BUFFER_PACKET_NUMBER and/or CDC_RECEIVE_QUEUE_BUFFER_PACKET_NUMBER both can be greater than 512. which every buffer size is 64B, 64*512=32768

…E_BUFFER_PACKET_NUMBER>512 or CDC_RECEIVE_QUEUE_BUFFER_PACKET_NUMBER>512.

which allows CDC_TRANSMIT_QUEUE_BUFFER_PACKET_NUMBER>512 or CDC_RECEIVE_QUEUE_BUFFER_PACKET_NUMBER>512. Signed-off-by: warmonkey <luoshumymail@gmail.com>

fpistm · 2025-02-18T09:32:50Z

Hi @warmonkey
Sorry for the delay. 😢
I've dig into HAL to see why it was uint16_t and found the reason, even if changed to uint32_t it will be always truncated to uint16_t when accessing the PMA (R/W).

Arduino_Core_STM32/system/Drivers/STM32F1xx_HAL_Driver/Src/stm32f1xx_ll_usb.c

Line 2817 in 345138a

    
           void USB_WritePMA(USB_TypeDef const *USBx, uint8_t *pbUsrBuf, uint16_t wPMABufAddr, uint16_t wNBytes)

So I will not merge this PR, sorry.

catbraincell · 2025-05-20T10:04:15Z

Hi @warmonkey Sorry for the delay. 😢 I've dig into HAL to see why it was uint16_t and found the reason, even if changed to uint32_t it will be always truncated to uint16_t when accessing the PMA (R/W).

Arduino_Core_STM32/system/Drivers/STM32F1xx_HAL_Driver/Src/stm32f1xx_ll_usb.c

Line 2817 in 345138a

void USB_WritePMA(USB_TypeDef const *USBx, uint8_t *pbUsrBuf, uint16_t wPMABufAddr, uint16_t wNBytes)

So I will not merge this PR, sorry.

On STM32F4 USB_WritePMA will not be called. The actual function to send data over USB is USBD_LL_Transmit() -> ... -> USB_EPStartXfer()

There two paths can trigger the USBD_LL_Transmit(): USBSerial::write() -> CDC_continue_transmit(), USBD_CDC_TransmitCplt() -> CDC_continue_transmit() -> USBD_CDC_TransmitPacket() -> USBD_LL_Transmit() -> HAL_PCD_EP_Transmit() -> USB_EPStartXfer()

In CDC_continue_transmit() file: usbd_cdc_if.c line 345, it's calling buffer = CDC_TransmitQueue_ReadBlock(&TransmitQueue, &size); which returns the size to send.
In CDC_TransmitQueue_ReadBlock() file: cdc_queue.c line 88, the size value is limited to 65472 in my commit

as stated in stm32f4xx_ll_usb.c function USB_EPStartXfer line 782:

Program the transfer size and packet count as follows: 
xfersize = N * maxpacket + short_packet 
pktcnt = N + (short_packet_exist ? 1 : 0)
...
pktcnt = (uint16_t)((ep->xfer_len + ep->maxpacket - 1U) / ep->maxpacket);
USBx_INEP(epnum)->DIEPTSIZ |= (USB_OTG_DIEPTSIZ_PKTCNT & (pktcnt << 19));
...
USBx_INEP(epnum)->DIEPTSIZ |= (USB_OTG_DIEPTSIZ_XFRSIZ & ep->xfer_len);

from the code above we know that the size to send from CDC_TransmitQueue_ReadBlock() previously, is assigned to ep->xfer_len (at stm32f4xx_hal_pcd.c line 1940 for stm32f4), then write into OTG_FS_DIEPTSIZx or OTG_HS_DIEPTSIZx register as:

PKTCNT = (int)(xfer_len/MAX_PACKET_SIZE)     //64 for FS, 512 for HS
XFRSIZ = xfer_len

from RM0008 RM0090 RM0432 we know the PKTCNT has 10bits, max value 0x3ff (1023), XFRSIZ has 19bits, max value 0x7ffff (524287). 1023 * 512 = 523776 < 524287

but, it's not easy to know the USB is running in HS or FS mode.

In conclusion, we should limit the maximum size returned by CDC_TransmitQueue_ReadBlock() to OTG_MAX_PKTCNTUSB_FS_MAX_PACKET_SIZE which is 102364 (for now).

I will read pdev->ep_in[CDCInEpAdd & 0xFU].maxpacket instead in the next commit.

The CDC queue is a software buffer, it's not relevant to the PMA area. The PMA is located in 0x4000 6000(1) - 0x4000 63FF Shared USB/CAN SRAM 512 bytes (RM0008 Page 52) it has only 512 bytes shared with multiple endpoints. User should copy data from/to PMA manually.

Before I submit this PR I checked multiple times, compiled and deployed on actual products. I ran the test for at least 24hrs.

catbraincell · 2025-05-20T10:51:37Z

Im working on an improvement. Will use maxpacket in USBD_EndpointTypeDef instead of use a fixed value 64. Just finished but i need more real world test.

…ket() in CDC_continue_transmit()

catbraincell · 2026-07-22T02:07:45Z

@fpistm I think we should merge to make the feature available first. taking MAX_PACKET_SIZE = 64 is a conservative number so its correct but not optimal. iit will be slower on HS.

Copilot

Pull request overview

This PR updates the USB CDC queue and related call sites to remove the previous ~32KB limitation caused by uint16_t length/index variables, enabling larger CDC transmit/receive queues when packet counts exceed 512.

Changes:

Widened CDC queue indices/lengths from 16-bit to 32-bit and adjusted queue read/write logic accordingly.
Updated USBSerial read helpers to work with the widened receive-queue APIs.
Introduced a helper to compute the maximum TX transfer size and attempted to clamp TX chunking accordingly.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
libraries/USBDevice/src/USBSerial.cpp	Updates read helper locals to match widened receive-queue API types.
libraries/USBDevice/src/cdc/usbd_cdc.c	Adds `USBD_CDC_GetTxMaxSize()` helper to compute max TX transfer size.
libraries/USBDevice/src/cdc/usbd_cdc_if.c	Uses the new max TX size when selecting a transmit block.
libraries/USBDevice/src/cdc/cdc_queue.c	Widens queue counters and adds transmit block clamping logic.
libraries/USBDevice/inc/usbd_cdc.h	Declares the new `USBD_CDC_GetTxMaxSize()` API.
libraries/USBDevice/inc/cdc_queue.h	Widens queue struct fields and updates function signatures/macros.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

+#ifdef USE_USBD_COMPOSITE
+uint8_t USBD_CDC_GetTxMaxSize(USBD_HandleTypeDef *pdev, uint8_t ClassId)
+{
+  /* Get the Endpoints addresses allocated for this class instance */
+  CDCInEpAdd  = USBD_CoreGetEPAdd(pdev, USBD_EP_IN, USBD_EP_TYPE_BULK, ClassId);
+#else
+uint8_t USBD_CDC_GetTxMaxSize(USBD_HandleTypeDef *pdev)
+{
+#endif /* USE_USBD_COMPOSITE */
+
+  return pdev->ep_in[CDCInEpAdd & 0xFU].maxpacket * USB_MAX_PKTCNT;
+}


+#define CDC_TRANSMIT_MAX_BUFFER_SIZE    65472       //STM32 USB OTG DIEPTSIZ PKTCNT maximum 0x3ff
+#define CDC_RECEIVE_MAX_BUFFER_SIZE     CDC_QUEUE_MAX_PACKET_SIZE


 int CDC_TransmitQueue_ReadSize(CDC_TransmitQueue_TypeDef *queue);
 void CDC_TransmitQueue_Enqueue(CDC_TransmitQueue_TypeDef *queue, const uint8_t *buffer, uint32_t size);
-uint8_t *CDC_TransmitQueue_ReadBlock(CDC_TransmitQueue_TypeDef *queue, uint16_t *size);
+uint8_t *CDC_TransmitQueue_ReadBlock(CDC_TransmitQueue_TypeDef *queue, uint32_t *size);


 // Read flat block from queue biggest as possible, but max QUEUE_MAX_PACKET_SIZE
 uint8_t *CDC_TransmitQueue_ReadBlock(CDC_TransmitQueue_TypeDef *queue,
-                                     uint16_t *size)
+                                     uint32_t *size)
 {
  if (queue->write >= queue->read) {


catbraincell added 3 commits November 4, 2024 00:27

cdc_queue.cpp now use uint32_t length, which allows CDC_TRANSMIT_QUEU…

049a8c8

…E_BUFFER_PACKET_NUMBER>512 or CDC_RECEIVE_QUEUE_BUFFER_PACKET_NUMBER>512.

bugfix: print(0ULL) output nothing

9333e7f

chore(usb): cdc_queue now use uint32_t length

66c8c0a

which allows CDC_TRANSMIT_QUEUE_BUFFER_PACKET_NUMBER>512 or CDC_RECEIVE_QUEUE_BUFFER_PACKET_NUMBER>512. Signed-off-by: warmonkey <luoshumymail@gmail.com>

catbraincell closed this Nov 5, 2024

fpistm reopened this Nov 5, 2024

fpistm changed the title ~~cdc_queue.cpp change to uint32_t length~~ chore(usb): cdc_queue now use uint32_t length Nov 5, 2024

fpistm added the enhancement New feature or request label Nov 5, 2024

fpistm added this to the 2.9.0 milestone Nov 5, 2024

fpistm mentioned this pull request Nov 5, 2024

chore(usb): cdc_queue now use uint32_t length #2547

Closed

fpistm force-pushed the main branch from 750f997 to 66c8c0a Compare November 5, 2024 08:25

fpistm removed this from the 2.9.0 milestone Nov 22, 2024

fpistm closed this Feb 18, 2025

fpistm added the invalid This doesn't seem right label Feb 18, 2025

fpistm reopened this May 20, 2025

github-project-automation Bot moved this from Done to In progress in STM32 core based on ST HAL May 20, 2025

fpistm removed the invalid This doesn't seem right label May 20, 2025

catbraincell added 3 commits May 20, 2025 05:25

Read epin maxpacket to determine the max size of USBD_CDC_TransmitPac…

7b70b39

…ket() in CDC_continue_transmit()

Merge branch 'main' of https://github.com/warmonkey/Arduino_Core_STM32

934508d

Merge branch 'stm32duino:main' into main

125eafe

fpistm force-pushed the main branch from 352b9df to bc21685 Compare May 19, 2026 06:54

fpistm requested a review from Copilot July 22, 2026 07:03

Copilot started reviewing on behalf of fpistm July 22, 2026 07:04 View session

Copilot AI reviewed Jul 22, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

chore(usb): cdc_queue now use uint32_t length#2550

chore(usb): cdc_queue now use uint32_t length#2550
catbraincell wants to merge 6 commits into
stm32duino:mainfrom
catbraincell:main

catbraincell commented Nov 4, 2024 •

edited by fpistm

Loading

Uh oh!

fpistm commented Feb 18, 2025

Uh oh!

catbraincell commented May 20, 2025 •

edited

Loading

Uh oh!

catbraincell commented May 20, 2025 •

edited

Loading

Uh oh!

catbraincell commented Jul 22, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

		#define CDC_TRANSMIT_MAX_BUFFER_SIZE 65472 //STM32 USB OTG DIEPTSIZ PKTCNT maximum 0x3ff
		#define CDC_RECEIVE_MAX_BUFFER_SIZE CDC_QUEUE_MAX_PACKET_SIZE

Uh oh!

Conversation

catbraincell commented Nov 4, 2024 • edited by fpistm Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

fpistm commented Feb 18, 2025

Uh oh!

catbraincell commented May 20, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

catbraincell commented May 20, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

catbraincell commented Jul 22, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

catbraincell commented Nov 4, 2024 •

edited by fpistm

Loading

catbraincell commented May 20, 2025 •

edited

Loading

catbraincell commented May 20, 2025 •

edited

Loading