casaroli opened a new pull request, #21795:
URL: https://github.com/apache/nuttx/pull/21795

   depends-on: apache/nuttx/pull/21778
   
   ## Summary
   
   `TIOCGWINSZ` and `TIOCSWINSZ` worked only on the telnet driver. The serial 
driver and the pseudo-terminal driver returned `ENOTTY`, so a program could not 
get the terminal size, `tcgetwinsize()` failed on a serial console, and 
`openpty()` could not set the size.
   
   This PR:
   
   - keeps a `struct winsize` in each serial device (`struct uart_dev_s`);
   - keeps one size for the two ends of a pseudo-terminal, because the size 
belongs to the terminal, not to one end;
   - sends `SIGWINCH` to the foreground PID when the size changes (for a 
pseudo-terminal: the foreground PID of the slave), if `CONFIG_TTY_SIGINT` or 
`CONFIG_TTY_SIGTSTP` keeps that PID. The same size again sends no signal. The 
pseudo-terminal sends the signal after it releases its lock: in a flat build 
the signal handler of the caller runs inside `nxsig_kill()`, and a handler 
usually calls `TIOCGWINSZ` again.
   
   The size is 0 rows and 0 columns until a program sets it. The second commit 
describes the ioctls in `Documentation/os/drivers/character/serial.rst`.
   
   ## Impact
   
   New ioctls on serial and pseudo-terminal devices. `struct uart_dev_s` and 
the pseudo-terminal device pair get a `struct winsize` (8 bytes each). No new 
option. No change for programs that do not use the ioctls; `SIGWINCH` has no 
default action, so a program without a handler ignores it. All build modes: the 
drivers read and write the `struct winsize` of the caller as they do for 
`TCGETS` and `TCSETS`.
   
   ## Testing
   
   Test program `winsize.c` (the `SIGWINCH` handler reads the size again with 
`TIOCGWINSZ` on the same terminal), with `CONFIG_PSEUDOTERM`, 
`CONFIG_PSEUDOTERM_SUSV1`, `CONFIG_SERIAL_TERMIOS`, `CONFIG_TTY_SIGINT`, 
`CONFIG_TTY_SIGTSTP` and `CONFIG_SIG_DEFAULT`, on qemu-armv8a:knsh (kernel 
build, the program from `/system/bin`) and qemu-armv8a:nsh (flat build, the 
program as a builtin), QEMU with HVF:
   
   | Case | master (knsh and nsh) | this PR (knsh and nsh) |
   |---|---|---|
   | serial console: set + get, 1 `SIGWINCH`, the handler sees the new size | 
`ENOTTY` | PASS |
   | serial console: the same size again, no `SIGWINCH` | not run (`ENOTTY` 
above) | PASS |
   | `openpty()` size, read on the slave | `ENOTTY` | 24x80, PASS |
   | the same size on the master | `ENOTTY` | 24x80, PASS |
   | `TIOCSWINSZ` on the master: the slave sees it, 1 `SIGWINCH`, the handler 
sees the new size | `ENOTTY` | PASS |
   | the same size again: no `SIGWINCH` | `ENOTTY` | PASS |
   | `TIOCSWINSZ` on the slave: the master sees it, 1 `SIGWINCH`, the handler 
sees the new size | `ENOTTY` | PASS |
   | `SIGWINCH` with the default action: the program goes on | `ENOTTY` | PASS |
   | (knsh) a `fork()` child is the foreground process of the slave: 
`TIOCSWINSZ` on the master in the parent sends it `SIGWINCH` | FAIL | PASS |
   
   ostest exits with status 0 on qemu-armv8a:knsh and qemu-armv8a:nsh with 
these options, on master and with this PR.
   
   Builds with `-Werror`, each OK: qemu-armv8a:nsh (flat, arm64) without the 
options, with all of them, with `CONFIG_PSEUDOTERM` but without 
`CONFIG_TTY_SIGINT`, `CONFIG_TTY_SIGTSTP` and `CONFIG_SERIAL_TERMIOS`, and with 
`CONFIG_TTY_SIGTSTP` only; qemu-armv8a:knsh (kernel, arm64) without the options 
and with them but without `CONFIG_SERIAL_TERMIOS`; with all the options: 
mps2-an500:knsh (protected, Cortex-M7), rv-virt:pnsh64 (protected, rv64), 
rv-virt:knsh64 (kernel, rv64), sim:nsh and esp32s3-devkit:nsh (Xtensa). 
lm3s6965-ek:qemu-protected builds as it is; with the options its kernel flash 
overflows on master too (by 5984 bytes; 6200 with this PR).
   
   `tools/checkpatch.sh -c -u -m -g` passes.
   


-- 
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