This collection of routines provides a disklabel management interface to kernel device drivers. These routines are classified as machine- or architecture-dependent because of restrictions imposed by the machine architecture and boot-strapping code on the location of the label, or because cooperation with other operating systems requires specialized conversion code.
readdisklabel() attempts to read a disklabel from the device identified by
dev, using the device strategy routine passed in
strat. Note that a buffer structure is required to pass to the strategy routine; it needs to be acquired and parameterized for the intended I/O operation, and disposed of when the operation has completed. Some fields in the disklabel passed in
lp may be pre-initialized by the caller in order to meet device driver requirements for the I/O operation initiated to get to the disklabel data on the medium. In particular, the field “d_secsize”, if non-zero, is used by
readdisklabel() to get an appropriately sized buffer to pass to the device strategy routine. Unspecified fields in
lp should be set to zero. If the medium does not contain a native disklabel that can be read in directly,
readdisklabel() may resort to constructing a label from other machine-dependent information using the provided buffer passed in the
clp argument. If a disk label can not be found or constructed, a string containing an approximated description of the failure mode is returned. Otherwise the
NULL string is returned.
writedisklabel() stores disk label information contained in the disk label structure given by
lp on the device identified by
dev. Like
readdisklabel(), it acquires and sets up an I/O buffer to pass to the strategy routine
strat.
writedisklabel() may elect to do a machine-dependent conversion of the native disk label structure (using the buffer pointed at by
clp), to store the disk label onto the medium in a format complying with architectural constraints.
writedisklabel() returns 0 on success and
EINVAL if the disk label specifies invalid or inconvertible values. Otherwise, any error condition reported by the device strategy routine in the buffer's “
b_error” field is returned.
setdisklabel() checks a proposed new disk label passed in
nlp for some amount of basic sanity. This includes a check on attempts to change the location, or reduce the size, of an existing disk partition that is currently in use by the system. The current disposition of the disk partitions is made available through
olp and
openmask, which provide, respectively, the existing disk label and a bit mask identifying the partitions that are currently in use. Failure to pass on “basic sanity”, results in a
EINVAL return value, while a vetoed update of the partition layout is signaled by a
EBUSY return value. Otherwise, 0 is returned.
bounds_check_with_label() is used to check whether a device transfer described by
bp to the device identified by
dev, is properly contained within a disk partition of the disk with label
lp. If this check fails,
bounds_check_with_label() sets the buffer's “
b_error” field to
EINVAL, sets the
B_ERROR flag in “
b_flags”, and returns -1. If the argument
wlabel is zero, and the transfer is a write operation, a check is done if the transfer would overwrite (a portion of) the disklabel area on the medium. If that is the case,
EROFS is set in “
b_error”, the
B_ERROR flag is set in “
b_flags”, and -1 is returned. Note that
wlabel should be set to a non-zero value if the intended operation is expected to install or update the disk label. Programs that intend to do so using the raw device interface should notify the driver by using a
DIOCWLABEL ioctl function.