sys/quota.h; they're meant to be used by applications and were
stuffed into the wrong place by accident last week and got
committed that way yesterday.
If various checks are omitted, the CMSG_NXTHDR macro expands to
(struct cmsghdr *)((char *)(cmsg) + \
_ALIGN(((struct cmsghdr *)(cmsg))->cmsg_len))
Although there is no alignment problem (assuming cmsg is properly aligned
and _ALIGN is correct), this violates -Wcast-align on strict-alignment
architectures. Therefore an intermediate cast to void * is appropriate here.
There is no workaround other than not using -Wcast-align.
Taken from FreeBSD commit r220742 by jilles
used by drivers: a short name for the quiet/naive case and a string
to override the "pcidevs" based name by one provided by the driver,
ride on yesterday's kernel minor version bump
All porting work was done by TOYOKURA Atsushi, with
several minor adjustments for integration by me and nonaka@.
His SL-C700 first running NetBSD/zaurus was demonstrated on
NetBSD booth at Open Source Canference 2011 Tokyo/Fall:
http://www.NetBSD.org/gallery/events.html#opensourceconf2011-tokyofall
Note GENERIC kernel works on SL-C750/760/860 (and C1000/3x00) but for now
SL-C700 requires its own kernel with special KERNEL_BASE_PHYS address
due to zbsdmod.o loader restriction (probably caused its small 32MB RAM).
The remaining references in the kernel are in vfs_quotactl.c, the
compat_50 code for the old quotactl (to be fixed up), and the
code compiled from src/common/lib/libquota.
ufsclass2qtype -> quota_idtype_to_ufs
qtype2ufsclass -> quota_idtype_from_ufs
The reason for the direction of "ufs" changing is that the old names
were among the symbols using "ufs" to mean "fs-independent". So the
old names were for translating "ufsclass" (fs-independent quota id
type) to "qtype" (ufs-specific quota id type) and vice versa.
These functions are used in only two places, both of which are
inappropriate, so at some point they should probably be removed.
They're also identity transformations so not particularly helpful,
unless one were to make a careful and concerted effort to distinguish
the ufs quota code numbers from the fs-independent ones. This has not
been done and is probably impossible without support from a program
verifier, and maybe not even then.
They are static inline, so no compat concerns arise.
Also adjust the symbols they use to avoid <quota/quotaprop.h>.
logical order (as opposed to the previous order, which accumulated
arbitrarily), remove the separate codes for argument encoding as
there's now a 1-1 mapping between ops and argument substructures,
and assert in VFS_QUOTACTL() itself that the op in the args structure
matches the op passed directly.
This change requires a kernel version bump.
he