Swatantra-66 opened a new pull request, #21884:
URL: https://github.com/apache/nuttx/pull/21884

   ## Summary
   
   POSIX IEEE Std 1003.1 and Linux specifications (`dup3(2)`) specify parameter 
validation, flag constraints, and descriptor handling for `dup3()`:
   
   1. **POSIX / Linux Flags Validation**:
      Linux `dup3(2)` explicitly mandates:
      > *"EINVAL (dup3()) flags contain an invalid value. Or, oldfd was equal 
to newfd."*
      The only valid flag for `dup3()` is `O_CLOEXEC` (or `0`). Previously, 
NuttX did not validate `flags`, silently accepting and ignoring invalid flag 
bits. Validates `(flags & ~O_CLOEXEC) != 0` returning `-EINVAL`.
   
   2. **Prevent Premature Heap Table Allocation on Invalid `oldfd`**:
      In `fdlist_dup3()`, `fdlist_extend()` was invoked to expand the file 
descriptor table on the heap *before* validating `fd1`. If `fd1` was invalid 
(e.g. `fd1 = -1` or an unopened descriptor), the task's file descriptor table 
was permanently expanded on the heap before `fdlist_get2()` returned `-EBADF`. 
Moving `fdlist_get2()` before `fdlist_extend()` validates `fd1` first, 
returning `-EBADF` with zero heap allocations and zero side-effects. If 
`fdlist_extend()` fails later, `file_put(filep1)` safely releases the acquired 
file reference.
   
   3. **Defensive NULL List Validation**:
      Defensively validates `list != NULL` returning `-EINVAL`, eliminating 
CWE-476 NULL pointer dereferences in `fdlist_limit(list)`.
   
   4. **Task-Specific Descriptor Resolution in `fdlist_dup2()`**:
      When `fd1 == fd2`, `fdlist_dup2()` previously called `file_get(fd1, 
&filep)` which queried `this_task()` rather than the provided `list`. Uses 
`fdlist_get2(list, fd1, &filep, NULL)` so the descriptor is looked up in the 
intended file list (e.g., during spawned task file actions in 
`sched/task/task_spawnparms.c`).
   
   ---
   
   ## Before vs. After Behavior
   
   | Scenario | Before | After |
   | :--- | :--- | :--- |
   | `dup3(fd1, fd2, 0x1234)` (invalid flags) | Silently ignored invalid bits, 
succeeded | Fails immediately with `ERROR`, `errno = EINVAL` |
   | `dup3(-1, 50, 0)` (invalid oldfd, large newfd) | Pre-allocated heap rows 
expanding task fd table before returning `-EBADF` | Returns `-EBADF` 
immediately; **zero heap memory allocated** |
   | `fdlist_dup3(NULL, ...)` | Dereferenced `NULL->fl_limit` (Kernel Panic / 
Crash) | Safely returns `-EINVAL` |
   | `posix_spawn` with `dup2(5, 5)` action | Checked descriptor `5` in 
caller's task instead of spawned task | Checks descriptor `5` in spawned task's 
descriptor table |
   
   ---
   
   ## Impact
   
   * **POSIX / Linux Conformance**: Rejects invalid flags with `-EINVAL`.
   * **Memory & Resource Efficiency**: Prevents failing `dup3()` invocations 
with invalid `oldfd` from permanently fragmenting and growing the process heap 
descriptor table.
   * **Security & Robustness**: Prevents NULL pointer dereferences and 
incorrect task descriptor queries during process spawning.
   
   ---
   
   ## Testing
   
   ### 1. Build Verification
   Cleanly built with `arm-none-eabi-gcc 10.3.1` on `stm32f103-minimum:nsh` 
(`tools/ci/testlist/arm-04.dat`):
   ```text
   $ ./tools/configure.sh stm32f103-minimum:nsh
   $ make -j8
   ...
   CC:  inode/fs_files.c
   ...
   LD: nuttx
   CP: nuttx.bin
   
   $ arm-none-eabi-size nuttx
      text         data     bss     dec     hex filename
     53723          468    3056   57247    df9f nuttx
   ```
   
   ### 2. Code Style Verification
   Checked with `nxstyle` inside Docker:
   ```text
   $ ./tools/nxstyle fs/inode/fs_files.c
   No errors, no warnings.
   ```
   
   ### 3. What Was Not Tested
   * Real hardware target (tested using cross-compilation for ARM Cortex-M3).
   * Protected and Kernel build modes.


-- 
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