1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
|
=head1 NAME
nbd_set_strict_mode - control how strictly to follow NBD protocol
=head1 SYNOPSIS
#include <libnbd.h>
int nbd_set_strict_mode (
struct nbd_handle *h, uint32_t flags
);
=head1 DESCRIPTION
By default, libnbd tries to detect requests that would trigger
undefined behavior in the NBD protocol, and rejects them client
side without causing any network traffic, rather than risking
undefined server behavior. However, for integration testing, it
can be handy to relax the strictness of libnbd, to coerce it into
sending such requests over the network for testing the robustness
of the server in dealing with such traffic.
The C<flags> argument is a bitmask, including zero or more of the
following strictness flags:
=over 4
=item C<LIBNBD_STRICT_COMMANDS> = 0x1
If set, this flag rejects client requests that do not comply with the
set of advertised server flags (for example, attempting a write on
a read-only server, or attempting to use C<LIBNBD_CMD_FLAG_FUA> when
L<nbd_can_fua(3)> returned false). If clear, this flag relies on the
server to reject unexpected commands.
=item C<LIBNBD_STRICT_FLAGS> = 0x2
If set, this flag rejects client requests that attempt to set a
command flag not recognized by libnbd (those outside of
C<LIBNBD_CMD_FLAG_MASK>), or a flag not normally associated with
a command (such as using C<LIBNBD_CMD_FLAG_FUA> on a read command).
If clear, all flags are sent on to the server, even if sending such
a flag may cause the server to change its reply in a manner that
confuses libnbd, perhaps causing deadlock or ending the connection.
Flags that are known by libnbd as associated with a given command
(such as C<LIBNBD_CMD_FLAG_DF> for L<nbd_pread_structured(3)> gated
by L<nbd_can_df(3)>) are controlled by C<LIBNBD_STRICT_COMMANDS>
instead; and C<LIBNBD_CMD_FLAG_PAYLOAD_LEN> is managed automatically
by libnbd unless C<LIBNBD_STRICT_AUTO_FLAG> is disabled.
Note that the NBD protocol only supports 16 bits of command flags,
even though the libnbd API uses C<uint32_t>; bits outside of the
range permitted by the protocol are always a client-side error.
=item C<LIBNBD_STRICT_BOUNDS> = 0x4
If set, this flag rejects client requests that would exceed the export
bounds without sending any traffic to the server. If clear, this flag
relies on the server to detect out-of-bounds requests.
=item C<LIBNBD_STRICT_ZERO_SIZE> = 0x8
If set, this flag rejects client requests with length 0. If clear,
this permits zero-length requests to the server, which may produce
undefined results.
=item C<LIBNBD_STRICT_ALIGN> = 0x10
If set, and the server provided minimum block sizes (see
C<LIBNBD_SIZE_MINIMUM> for L<nbd_get_block_size(3)>), this
flag rejects client requests that do not have length and offset
aligned to the server's minimum requirements, except for an
unaligned length that extends to the end of an image with
an overall unaligned size. If clear, all unaligned requests
are sent to the server, where it is up to the server whether
to honor or reject the request.
=item C<LIBNBD_STRICT_PAYLOAD> = 0x20
If set, the client refuses to send a command to the server
with more than libnbd's outgoing payload maximum (see
C<LIBNBD_SIZE_PAYLOAD> for L<nbd_get_block_size(3)>), whether
or not the server advertised a block size maximum. If clear,
oversize requests up to 64MiB may be attempted, although
requests larger than 32MiB are liable to cause some servers to
disconnect.
=item C<LIBNBD_STRICT_AUTO_FLAG> = 0x40
If set, commands that accept the C<LIBNBD_CMD_FLAG_PAYLOAD_LEN>
flag (such as L<nbd_pwrite(3)> and C<nbd_block_status_filter(3)>)
ignore the presence or absence of that flag from the caller,
instead sending the value over the wire that matches the
server's expectations based on whether extended headers were
negotiated when the connection was made. If clear, the caller
takes on the responsibility for whether the payload length
flag is set or clear during the affected command, which can
be useful during integration testing but is more likely to
lead to undefined behavior.
=back
For convenience, the constant C<LIBNBD_STRICT_MASK> is available to
describe all strictness flags supported by this build of libnbd.
Future versions of libnbd may add further flags, which are likely
to be enabled by default for additional client-side filtering. As
such, when attempting to relax only one specific bit while keeping
remaining checks at the client side, it is wiser to first call
L<nbd_get_strict_mode(3)> and modify that value, rather than
blindly setting a constant value.
=head1 RETURN VALUE
If the call is successful the function returns C<0>.
=head1 ERRORS
On error C<-1> is returned.
Refer to L<libnbd(3)/ERROR HANDLING>
for how to get further details of the error.
The following parameters must not be NULL: C<h>.
For more information see L<libnbd(3)/Non-NULL parameters>.
=head1 VERSION
This function first appeared in libnbd 1.6.
If you need to test if this function is available at compile time
check if the following macro is defined:
#define LIBNBD_HAVE_NBD_SET_STRICT_MODE 1
=head1 SEE ALSO
L<nbd_can_df(3)>,
L<nbd_can_fua(3)>,
L<nbd_create(3)>,
L<nbd_get_block_size(3)>,
L<nbd_get_strict_mode(3)>,
L<nbd_pread_structured(3)>,
L<nbd_pwrite(3)>,
L<nbd_set_handshake_flags(3)>,
L<nbd_stats_bytes_received(3)>,
L<nbd_stats_bytes_sent(3)>,
L<libnbd(3)>.
=head1 AUTHORS
Eric Blake
Richard W.M. Jones
=head1 COPYRIGHT
Copyright Red Hat
|
