Commit 57c07e27 authored by Jeremy Allison's avatar Jeremy Allison Committed by Gerald (Jerry) Carter
Browse files

r1304: Turns out non-io opens that cause oplock breaks are a *different* set 

of desired access flags from those that cause sharing violations.
Oplock breaks are caused by : ~(SYNCHRONIZE_ACCESS|FILE_READ_ATTRIBUTES|FILE_WRITE_ATTRIBUTES)
Sharing violations are caused by : ~(SYNCHRONIZE_ACCESS|READ_CONTROL_ACCESS|FILE_READ_ATTRIBUTES|FILE_WRITE_ATTRIBUTES)
We now pass the torture rename test again.
I still need to work out if subsequent opens will cause sharing violations with an existing open
mode of SYNCHRONIZE_ACCESS|READ_CONTROL_ACCESS|FILE_READ_ATTRIBUTES|FILE_WRITE_ATTRIBUTES;
Jeremy.
(This used to be commit 38002bfc1c0981a2dba3f2319624c1ef055a3648)
parent b04c7822

source3/smbd/open.c

+2 −2
Original line number Diff line number Diff line
@@ -381,13 +381,13 @@ static BOOL check_share_mode(connection_struct *conn, share_mode_entry *share, i 
	if(!lp_share_modes(SNUM(conn)))

		return True;



	if (desired_access & ~(SYNCHRONIZE_ACCESS|FILE_READ_ATTRIBUTES|FILE_WRITE_ATTRIBUTES)) {

	if (desired_access & ~(SYNCHRONIZE_ACCESS|READ_CONTROL_ACCESS|FILE_READ_ATTRIBUTES|FILE_WRITE_ATTRIBUTES)) {

		non_io_open_request = False;

	} else {

		non_io_open_request = True;

	}



	if (share->desired_access & ~(SYNCHRONIZE_ACCESS|FILE_READ_ATTRIBUTES|FILE_WRITE_ATTRIBUTES)) {

	if (share->desired_access & ~(SYNCHRONIZE_ACCESS|READ_CONTROL_ACCESS|FILE_READ_ATTRIBUTES|FILE_WRITE_ATTRIBUTES)) {

		non_io_open_existing = False;

	} else {

		non_io_open_existing = True;