yangsong8-a1 opened a new pull request, #21782:
URL: https://github.com/apache/nuttx/pull/21782

   ## Summary
   
   When pty_register2() fails to register the slave device, it unregisters the
   master with unregister_driver(). That calls pty_unlink(), which calls
   pty_destroy() and frees devpair because pp_nopen is 0. The error path then
   used devpair again (nxmutex_destroy, nxsem_destroy, kmm_free).
   
   - Return right after unregister_driver() when pseudo-fs operations are
     enabled, because pty_unlink() has already freed devpair.
   - In pty_destroy(), unregister the slave before releasing the minor
     number, so the minor cannot be reused while /dev/pts/N still exists.
   - Follow-up: on that path pty_destroy() also freed the minor, and
     ptmx_open() freed it a second time (DEBUGASSERT in ptmx_minor_free()).
     It also removed the /dev/pts/N node, which may belong to another driver.
     Take a reference before unregistering the master, so pty_unlink() keeps
     the pair, and free it on the normal error path. The caller frees the
     minor once.
   
   ## Impact
   
   Only the pty_register2() error path and the order of operations in
   pty_destroy(). Normal open and close of ptys does not change.
   
   ## Testing
   
   Ubuntu 22.04, gcc 13.1. sim:nsh with CONFIG_PSEUDOTERM (SUSv1) and
   CONFIG_MM_KASAN (CONFIG_DEBUG_ASSERTIONS enabled), no new warnings. A test
   app opens and closes /dev/ptmx 20 times, with data both ways, then
   registers a dummy driver at /dev/pts/0 so that the next /dev/ptmx open
   fails in pty_register2():
   
   Before:
   ```
   normal: 20 open/close, fails=0, /dev/pts entries=0
   inject: register dummy /dev/pts/0 -> 0
   kasan_report: kasan detected a read access error, address at 
0x7bec5b52411c,size is 1
   ```
   After:
   ```
   normal: 20 open/close, fails=0, /dev/pts entries=0
   inject: register dummy /dev/pts/0 -> 0
   inject: open /dev/ptmx -> -1 errno=17 (expect -1 EEXIST=17)
   after: 5 open/close, fails=0
   ```
   


-- 
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