Manpages

NAME

nbd_set_strict_mode - control how strictly to follow NBD protocol

SYNOPSIS

#include <libnbd.h>
int nbd_set_strict_mode (
struct nbd_handle *h, uint32_t flags
);

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 "flags" argument is a bitmask, including zero or more of the following strictness flags:
"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 "LIBNBD_CMD_FLAG_FUA" when nbd_can_fua(3) returned false). If clear, this flag relies on the server to reject unexpected commands.

"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 "LIBNBD_CMD_FLAG_MASK"), or a flag not normally associated with a command (such as using "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 "LIBNBD_CMD_FLAG_DF" for nbd_pread_structured(3) gated by nbd_can_df(3)) are controlled by "LIBNBD_STRICT_COMMANDS" instead; and "LIBNBD_CMD_FLAG_PAYLOAD_LEN" is managed automatically by libnbd unless "LIBNBD_STRICT_AUTO_FLAG" is disabled.

Note that the NBD protocol only supports 16 bits of command flags, even though the libnbd API uses "uint32_t"; bits outside of the range permitted by the protocol are always a client-side error.

"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.

"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.

"LIBNBD_STRICT_ALIGN" = 0x10

If set, and the server provided minimum block sizes (see "LIBNBD_SIZE_MINIMUM" for nbd_get_block_size(3)), this flag rejects client requests that do not have length and offset aligned to the server’s minimum requirements. If clear, unaligned requests are sent to the server, where it is up to the server whether to honor or reject the request.

"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 "LIBNBD_SIZE_PAYLOAD" for 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.

"LIBNBD_STRICT_AUTO_FLAG" = 0x40

If set, commands that accept the "LIBNBD_CMD_FLAG_PAYLOAD_LEN" flag (such as nbd_pwrite(3) and 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.

For convenience, the constant "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 nbd_get_strict_mode(3) and modify that value, rather than blindly setting a constant value.

RETURN VALUE

If the call is successful the function returns 0.

ERRORS

On error -1 is returned.

Refer to "ERROR HANDLING" in libnbd(3) for how to get further details of the error.

The following parameters must not be NULL: "h". For more information see "Non-NULL parameters" in libnbd(3).

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

SEE ALSO

nbd_can_df(3), nbd_can_fua(3), nbd_create(3), nbd_get_block_size(3), nbd_get_strict_mode(3), nbd_pread_structured(3), nbd_pwrite(3), nbd_set_handshake_flags(3), nbd_stats_bytes_received(3), nbd_stats_bytes_sent(3), libnbd(3).

AUTHORS

Eric Blake

Richard W.M. Jones

COPYRIGHT

Copyright Red Hat

LICENSE

This library is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version.

This library is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details.

You should have received a copy of the GNU Lesser General Public License along with this library; if not, write to the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA