Summary:
- I noticed that the results of the df command are incorrect for NFS.
- Finally, I found that obj_attributes is not needed to
communicate with the NFS server on Linux
- This commit fixes this issue
Impact:
- None
Testing:
- Tested with spresense:rndis_smp
Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com>
The use of uint64_t primitive type in NFS structures forces the compiler to align data on an 8-byte boundary.
As a result of this, unwanted gaps being created, which causes NFS to fail. (e.g., nfs_read/initialize the request)
Using nfsuint64 instead of uint64_t fixes this issue.
* fs/nfs: Remove all nfs_checkmount call. The check just waste cpu cycle since nobody will set nm_mounted to false, and remove the unused fields(nm_mounted and n_flags) and related flags too
* fs/nfs: Fix the definition not confirm to RFC 1813 and other minor issue(unused, typo, duplication, alignment...)
* fs/nfs: Always represent error with negative number and remove the unused arguments from function
* fs/nfs: Set socket receive timeout with nfs_args->timeo and fix warning:
nfs/nfs.h:59:28: warning: large integer implicitly truncated to unsigned type [-Woverflow]
#define NFS_TIMEO (1 * NFS_HZ) /* Default timeout = 1 second */
^
nfs/nfs_vfsops.c:1857:23: note: in expansion of macro 'NFS_TIMEO'
nprmt.timeo = NFS_TIMEO;
^
^~~~~~~~~
* fs/nfs: Update the file attribute correctly in nfs_filetruncate and simplify the attrbitue conversion between NFSv3 and NuttX
* fs/nfs: Remove the unfinished and buggy EXCLUSIVE creation mode
* fs/nfs: Call nfs_fsinfo in nfs_bind instead of nfs_statfs since we should update the buffer size before transfer happen, and handle the attribute existence variance correctly.
* fs/nfs: Shouldn't insert node into list again in nfs_dup and fix other typo issue
* fs/nfs: Should skip . and .. in nfs_readdir
* fs/nfs: Remove the unnecessary dynamic allocation and the duplicated root handle storage
Masayuki Ishikawa