Sorry for my poor english.

On Linux 2.4.x system(I test 2.4.17 and 2.4.20), smbd which is compiled with 
HAVE_KERNEL_OPLOCKS_LINUX sometime can't catch oplock break signal from linux 
kernel.

I cut-and-try various test and make sure reproduce condition.
To reprodule problem, do below procedure.
 1) Windows Client open server's file via smbd ,and oplock granted.
 2) Keep 1) handle open pass the time until call timeout_processing().
    Probabry the time is  60seconds. 
   (local.h -  #define SMBD_SELECT_TIMEOUT (60) )
 3) When the time pass, do "cat <opened-file>".
    "cat" can't open the file instantly,and waiting. 
@@And smbd don't catch oplock break signal from kernel, so smbd don't send 
oplock break LockingX request to Windows client.
 
 4) After 45seconds(/proc/sys/fs/lease-break-time), "cat" open the file.
    But ofcource smbd and Windows client oplock break. 
    It's bad file sharing status.

I search what is ill when timeout_processing(). 
And I found calling change_to_root_user() in timeout_processing() cause smbd to 
lose ability to catch oplock break signal from kernel.

Then, I search workaround by various cut-and-try,and I found.
In smbd/oplock_linux.c/linux_setlease(), do become_root() and unbecome_root() 
as below. I don't know this way is correct or incorrect.But do this, problem is 
not reproduce.
---------------------------------------------------------
static int linux_setlease(int fd, int leasetype)
{
	int ret;

         become_root() /******** ADD ************/
	if (fcntl(fd, F_SETSIG, RT_SIGNAL_LEASE) == -1) {
		DEBUG(3,("Failed to set signal handler for kernel lease\n"));
         @@@@@unbecome_root() /************ ADD ************/
		return -1;
	}

	ret = fcntl(fd, F_SETLEASE, leasetype);
	if (ret == -1 && errno == EACCES) {
		set_capability(CAP_LEASE);
		ret = fcntl(fd, F_SETLEASE, leasetype);
	}
@@@@@unbecome_root() /************** ADD **************/
	return ret;
}
---------------------------------------------------------
 
I don't understand linux kernel. But I guess below checking code impact to this 
problem.

--linux/fs/fcntl.c----------------------------------------------------
static void send_sigio_to_task(struct task_struct *p,
			       struct fown_struct *fown, 
			       int fd,
			       int reason)
{
	if ((fown->euid != 0) &&
	    (fown->euid ^ p->suid) && (fown->euid ^ p->uid) &&
	    (fown->uid ^ p->suid) && (fown->uid ^ p->uid))
		return;
<snip>
---------------------------------------------------------

And I think that call become_root() when F_SETLEASE avoid above checking code 
(fown->euid != 0),because below code.

---------------------------------------------------------
int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
{
 <snip>
	list_add(&fl->fl_link, &file_lock_list);
	filp->f_owner.pid = current->pid;
	filp->f_owner.uid = current->uid;
	filp->f_owner.euid = current->euid;   /******** THIS *********/
out_unlock:
	unlock_kernel();
	return error;
}
---------------------------------------------------------