yangsong8-a1 opened a new pull request, #22158:
URL: https://github.com/apache/nuttx/pull/22158

   ## Summary
   
   The serial upper half calls `setup()` and `shutdown()` with 
`uart_spinlock()` held (since #21746).  `usbhost_cdcacm` and `usbhost_ft232r` 
took `priv->lock` (a mutex) there, and `shutdown()` called `usbhost_destroy()` 
when the device had been unplugged while open.
   
   From reading the code, unplug-then-close was broken on its own too: 
`usbhost_destroy()` frees the class instance, which contains the `uart_dev_s` 
that `uart_close()` keeps using, and its `unregister_driver()` reaches 
`uart_unlink()`, which takes the `closelock` that `uart_close()` holds.
   
   Use the upper half's `release()` method instead: `setup()`/`shutdown()` only 
update `crefs` under the existing spinlock / critical section; 
`usbhost_disconnected()` unregisters the device at once (unless 
`usbhost_connect()` is running); `usbhost_release()` frees everything after the 
last close.
   
   ## Impact
   
   USB host CDC/ACM and FT232R serial devices.  After an unplug the 
`/dev/ttyACMn` / `/dev/ttyUSBn` node is removed at once instead of at the last 
close.
   
   ## Testing
   
   Ubuntu 22.04, gcc 13.1.0, QEMU 6.2.0.
   
   - Build: `sim:usbhost` (CDC/ACM), `qemu-intel64:nsh` + xHCI PCI + 
`CONFIG_USBHOST_FT232R` (SMP=2 and SMP=n), no warnings.
   - Runtime, FT232R with QEMU `-device qemu-xhci -device usb-serial`: device 
enumerated as `/dev/ttyUSB0`; 40 `echo ... > /dev/ttyUSB0` (open, write, close) 
all reached the host chardev, SMP=2 and SMP=n, same as master.
   
   Not tested: the unplug path (QEMU `device_del` is not reported by the xHCI 
driver, on master either, so `usbhost_disconnected()` never runs); CDC/ACM at 
runtime; real hardware.
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to