On 11 Aug 2010, at 16:14, Hans-Werner Paulsen wrote:
> time 961.188544, pid 5152: Analyze RPC op 13 conn 0xffffffffc53572c0 code 0x0
> user 0x41629d76
This is a SetLock
> time 961.188660, pid 5152: Analyze RPC op 15 conn 0xffffffffc53572c0 code 0x0
> user 0x41629d76
This is a ReleaseLock
That sequence of actions means that it's afs_posix_lock_file() that is failing
- that's just a compatibility wrapper around Linux's posix_lock_file function,
which we use to enrol our looks in the kernel lock management scheme. However,
when posix_lock_file fails it returns a kernel form (negative) error code - not
a positive AFS error code, and so afs_convert_code mangles it.
The fix for the first part is (apologies for any patch fuzz):
diff --git a/src/afs/LINUX/osi_vnodeops.c b/src/afs/LINUX/osi_vnodeops.c
index 4beed54..a280d0e 100644
--- a/src/afs/LINUX/osi_vnodeops.c
+++ b/src/afs/LINUX/osi_vnodeops.c
@@ -480,7 +480,7 @@ afs_linux_lock(struct file *fp, int cmd, struct file_lock *f
#endif /* F_GETLK64 && F_GETLK != F_GETLK64 */
AFS_GLOCK();
- code = afs_lockctl(vcp, &flock, cmd, credp);
+ code = afs_convert_code(afs_lockctl(vcp, &flock, cmd, credp));
AFS_GUNLOCK();
if ((code == 0 || flp->fl_type == F_UNLCK) &&
@@ -517,7 +517,7 @@ afs_linux_lock(struct file *fp, int cmd, struct file_lock *f
flp->fl_end = flock.l_start + flock.l_len - 1;
crfree(credp);
- return afs_convert_code(code);
+ return code;
}
#ifdef STRUCT_FILE_OPERATIONS_HAS_FLOCK
Once you've applied this, I would be interested to know what error your test
now returns ...
Cheers,
Simon.
_______________________________________________
OpenAFS-devel mailing list
[email protected]
https://lists.openafs.org/mailman/listinfo/openafs-devel
