AlmAck opened a new pull request, #20001:
URL: https://github.com/apache/nuttx/pull/20001
## Summary
`filemtd_initialize()` opens its backing file with
```c
mode = O_RDONLY | O_WRONLY | O_CLOEXEC;
```
Commit 6161c73639 ("include/fcntl.h: remove O_RDOK/O_WROK aliases")
introduced this when it replaced the non-standard `O_RDOK | O_WROK`
pair, describing the change as a pure text substitution. That held while
the access mode was a genuine bitmask: `O_RDONLY` was `(1 << 0)`,
`O_WRONLY` was `(1 << 1)`, `O_RDWR` was both bits, and `O_ACCMODE` was
defined as an alias for `O_RDWR`. OR-ing the two was meaningful and
produced `O_RDWR`.
Commit 9e141acab3 ("include/fcntl.h: align open flags with Linux
values") then made the low two bits an *enumeration* — `O_RDONLY` 0,
`O_WRONLY` 1, `O_RDWR` 2 — and `O_ACCMODE` stopped being an alias for
`O_RDWR`, becoming an independent mask of 3. OR-ing two members of that
enumeration is no longer meaningful:
```
O_RDONLY | O_WRONLY == 0 | 1 == 1
1 & O_ACCMODE == O_WRONLY
```
The file is opened write-only, and `fs/vfs/fs_read.c:202` rejects every
read on it with `-EACCES`.
### Is this a pattern?
It looks like an isolated miss rather than a systematic one. Grepping
the tree:
* Two access-mode constants OR-ed together — this call site only.
* Bit-testing the mode instead of masking — one site,
`fs/xipfs/xipfs_vfs.c:606`, which happens to be correct under the new
values (and would have been wrong under the old ones, so it was
clearly written after the change).
* Unmasked equality against an access mode — none.
* Host/guest translation — the `NUTTX_O_*` mirror in
`include/nuttx/fs/hostfs.h` was updated in lockstep and
`host_oflags_convert()` switches on `flags & NUTTX_O_ACCMODE`.
## Impact
Affects every user of `filemtd_initialize()` — the simulator's
file-backed MTD, `testing/fs`, and any board that layers an MTD over a
file. Reads through the device fail, so any filesystem mounted on it
fails to come up.
On the simulator it surfaces as a LittleFS mount of a filemtd-backed
partition returning `-ENOSPC`, after which nothing that lives on that
volume works: the resource pack cannot be read, fonts load with zero
metrics, and a FlashDB partition on the same device logs out-of-bound
writes. None of those point at the open mode.
## Testing
**Evidence available (captured on this tree):**
Host: Linux x86_64, GCC 15. Board: `sim`, configuration with a 4 MiB
file-backed MTD (`filemtd_initialize`) partitioned into an mtdconfig
partition, a LittleFS volume and a third raw partition.
Before, mounting the LittleFS volume on the filemtd partition:
```
sim_storage: mount /dev/rblflash -> /mnt/fs: errno 28
[E/FAL] (fal_partition_write:455) Partition write error! Partition address
out of bound.
resource alloc of -1 bytes failed
GFont 0x40166fcc has line_height=0; falling back to default
```
After, same binary and same backing file, only this patch applied:
```
sim_storage: /dev/config + /dev/rblflash @/mnt/fs + /dev/rblflash_ts ready
(fs 959 + tsdb 64 erase blks, 4096 B/erase blk)
seeded /mnt/fs/system.pbpack (147398 bytes)
pbpack table cached: 22 entries
```
The volume mounts, autoformat works, files read back, and the third
partition writes cleanly.
--
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]
