]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
19 years agoCorrect two bugs that may result in incorrect CBCP response for
delphij [Mon, 6 Dec 2004 15:49:52 +0000 (15:49 +0000)]
Correct two bugs that may result in incorrect CBCP response for
administrator specified number.

The implementation in our version transmits extra zero byte that is a
violation of the Microsoft PPP Callback Control Protocol section 3.2[1].
This can lead to incorrect behavior on certain devices, as reported
in bin/50613.

Confirmed with: cbcp.c in ppp's cvs, rev. 1.4 and 1.15 [2]

Submitted by: Dmitry Pryanishnikov <dmitry at atlantis.dp.ua>
PR: bin/50613

[1] http://www.dsl.gr.jp/~manabe/PPxP/doc/Standards/draft-gidwani-ppp-callback-cp-00.txt
[2] http://cvs.samba.org/cgi-bin/cvsweb/ppp/pppd/cbcp.c
    http://www.samba.org/cvs.html

19 years agoFix -t option processing.
bland [Mon, 6 Dec 2004 15:38:24 +0000 (15:38 +0000)]
Fix -t option processing.
Partially initialized tm structure was passed into mktime(3).

Approved by: ru

19 years agoSome constification which doesn't require code rewrites.
harti [Mon, 6 Dec 2004 15:20:12 +0000 (15:20 +0000)]
Some constification which doesn't require code rewrites.

19 years agoThis commit was generated by cvs2svn to compensate for changes in r138451,
roberto [Mon, 6 Dec 2004 14:33:29 +0000 (14:33 +0000)]
This commit was generated by cvs2svn to compensate for changes in r138451,
which included commits to RCS files with non-trunk default branches.

19 years agoMerge from the main BK repository for ntp: put two midly annoying messages
roberto [Mon, 6 Dec 2004 14:33:29 +0000 (14:33 +0000)]
Merge from the main BK repository for ntp: put two midly annoying messages
under #ifdef DEBUG. Merge of revision 1.45 by H. Stenn.

Done on the vendor branch to minimise future imports.

Reminded by: obrien

19 years agoPrint a warning if running as !root for aio_md_test rather than failing
rwatson [Mon, 6 Dec 2004 13:15:23 +0000 (13:15 +0000)]
Print a warning if running as !root for aio_md_test rather than failing
the test.  Privilege is required in order to allocate an md device.

19 years agoAdd a few convenient functions in the mount_arg() family and collect the
phk [Mon, 6 Dec 2004 13:01:41 +0000 (13:01 +0000)]
Add a few convenient functions in the mount_arg() family and collect the
entire family at the end of the source file.

19 years agoAdd a basic aio functionality regression test, which simply writes and
rwatson [Mon, 6 Dec 2004 12:56:38 +0000 (12:56 +0000)]
Add a basic aio functionality regression test, which simply writes and
then reads from a fairly broad range of object types: regular file,
fifo, UNIX socketpair, pty, UNIX pipe, and an md device.  Not a deep
test of functionality, just a basic test that aio_write followed by
aio_read returns the correct data in a relatively timely manner.

Requested by: phk

19 years agoCollapse two almost identical license copies, preserving the rights of
phk [Mon, 6 Dec 2004 12:44:30 +0000 (12:44 +0000)]
Collapse two almost identical license copies, preserving the rights of
all listed authors, rightholders and contributors.

19 years agoRemove the kern.rootdev sysctl.
phk [Mon, 6 Dec 2004 12:40:45 +0000 (12:40 +0000)]
Remove the kern.rootdev sysctl.

Root filessytems (like NFS) don't have an associated disk device,
and even if they had, the exact semantics would be filesystem
dependent and should be implemented there.

19 years agoMake struct vfsopt{list} private to vfs_mount.c
phk [Mon, 6 Dec 2004 12:36:17 +0000 (12:36 +0000)]
Make struct vfsopt{list} private to vfs_mount.c

19 years agoFix warning
phk [Mon, 6 Dec 2004 12:34:28 +0000 (12:34 +0000)]
Fix warning

19 years agoUpdate prototypes for eventhandler_register() and eventhandler_find_list()
jkoshy [Mon, 6 Dec 2004 11:30:55 +0000 (11:30 +0000)]
Update prototypes for eventhandler_register() and eventhandler_find_list()
to match the code.

19 years agoStyle: fix indentation.
harti [Mon, 6 Dec 2004 11:30:36 +0000 (11:30 +0000)]
Style: fix indentation.

19 years agoWe don't have RAIDFrame anymore and it seems gvinum doesn't use SI_SUB_RAID,
pjd [Mon, 6 Dec 2004 11:28:02 +0000 (11:28 +0000)]
We don't have RAIDFrame anymore and it seems gvinum doesn't use SI_SUB_RAID,
so correct stale comment. The only SI_SUB_RAID consumer is gmirror right now.

19 years agoUse 'const char *' for a few prototypes.
jkoshy [Mon, 6 Dec 2004 10:53:40 +0000 (10:53 +0000)]
Use 'const char *' for a few prototypes.

Reviewed by: ru

19 years agoUnbreak sockstat(1) on systems without the divert protocol.
ru [Mon, 6 Dec 2004 09:28:05 +0000 (09:28 +0000)]
Unbreak sockstat(1) on systems without the divert protocol.

19 years agoRemove an unused macro.
harti [Mon, 6 Dec 2004 08:57:41 +0000 (08:57 +0000)]
Remove an unused macro.

19 years agoConstify arguments to Hash_FindEntry and Hash_CreateEntry.
harti [Mon, 6 Dec 2004 08:56:30 +0000 (08:56 +0000)]
Constify arguments to Hash_FindEntry and Hash_CreateEntry.

19 years agoRemove extra empty lines.
harti [Mon, 6 Dec 2004 08:52:02 +0000 (08:52 +0000)]
Remove extra empty lines.

19 years agoStyle: fix indentation, prototypes for functions even in comment.
harti [Mon, 6 Dec 2004 08:51:34 +0000 (08:51 +0000)]
Style: fix indentation, prototypes for functions even in comment.

19 years agoMake this work under debugging, e.g., "make -dl".
ru [Mon, 6 Dec 2004 08:51:30 +0000 (08:51 +0000)]
Make this work under debugging, e.g., "make -dl".

19 years agoUserland change corresponding to the change in kernel/userland communication
phk [Mon, 6 Dec 2004 08:32:09 +0000 (08:32 +0000)]
Userland change corresponding to the change in kernel/userland communication
for NFS locking.

19 years agoFor reasons unknown, the nfs locking code used a fifo to send requests to
phk [Mon, 6 Dec 2004 08:31:32 +0000 (08:31 +0000)]
For reasons unknown, the nfs locking code used a fifo to send requests to
userland and a dedicated system call to get replies.

The vnode-bypass of fifos broke this into a panic.

Ditch all the magic and create a device /dev/nfslock instead, and
use that for both directions apart from the shorter path, this is
also faster because the device driver runs Giant free using the
vnode bypass.

Noticed by: marcel

19 years agoAdd support for the memory-mapped PCI Express configuration mechanism. This
scottl [Mon, 6 Dec 2004 08:27:10 +0000 (08:27 +0000)]
Add support for the memory-mapped PCI Express configuration mechanism.  This
actually is a property of the northbridge and applies to all PCI/PCI-X/PCIe
devices in the system, though only PCIe devices will respond to registers
higher than 256.  This uses per-CPU pools of temporary mappings so that
the whole 256MB of configuration space doesn't have to be mapped all at
once.  While the sf_buf API was considered for this, the fact that it
requires sleep locks and can return failure made it unsuitable for this use.

For now only the Intel Grantsdale and Lindenhurst (925 and 752x) chipsets are
supported.  Since there doesn't appear to be a compatible way to determine
northbridge support, new chipsets will have to be explicitely added in the
future.

19 years agoThe 'start' command processes loader.conf variables, not loader.rc.
imp [Mon, 6 Dec 2004 05:30:31 +0000 (05:30 +0000)]
The 'start' command processes loader.conf variables, not loader.rc.
Fix comment to match.

19 years agoThis jumbo allocator has been removed from the kernel.
alc [Mon, 6 Dec 2004 04:24:17 +0000 (04:24 +0000)]
This jumbo allocator has been removed from the kernel.

19 years agoEnable amr(4) - scottl fixed when used with >4GB RAM.
obrien [Mon, 6 Dec 2004 02:50:31 +0000 (02:50 +0000)]
Enable amr(4) - scottl fixed when used with >4GB RAM.

19 years agoUpdate the Tigon 1 and 2 driver to use the sf_buf API for implementing
alc [Mon, 6 Dec 2004 00:43:40 +0000 (00:43 +0000)]
Update the Tigon 1 and 2 driver to use the sf_buf API for implementing
zero-copy receive of jumbo frames.  This eliminates the need for the
jumbo frame allocator implemented in kern/uipc_jumbo.c and sys/jumbo.h.
Remove it.

Note: Zero-copy receive of jumbo frames did not work without these changes;
I believe there was insufficient locking on the jumbo vm object.

Tested by: ken@
Discussed with: gallatin@

19 years agoFix a number of bugs and significantly alter the command execution path to
scottl [Sun, 5 Dec 2004 23:48:17 +0000 (23:48 +0000)]
Fix a number of bugs and significantly alter the command execution path to
properly support bounce buffers and resource shortages.  This allows the
driver to work properly and reliably with more than 4GB of RAM.  Of the
three data paths that exist in the driver, (block, CAM, ioctl), the ioctl
path has not been well tested with these changes due to difficulty with
finding an application that uses it that actually works.

Sponsored by: The FreeBSD Foundation and FreeBSD Systems, Inc.

19 years agoCode for automatic name completion in a tcsh(1) environment.
hmp [Sun, 5 Dec 2004 23:44:52 +0000 (23:44 +0000)]
Code for automatic name completion in a tcsh(1) environment.

PR: docs/44435
Submitted by: Slaven Rezic slaven dot rezic at berlin dot de

19 years agoUse ng_callout() instead of timeout(9).
glebius [Sun, 5 Dec 2004 22:58:13 +0000 (22:58 +0000)]
Use ng_callout() instead of timeout(9).

Approved by: julian (mentor)

19 years agoConvert GIANT_REQUIRED; in nfs_mountroot() to NET_ASSERT_GIANT(),
rwatson [Sun, 5 Dec 2004 22:53:17 +0000 (22:53 +0000)]
Convert GIANT_REQUIRED; in nfs_mountroot() to NET_ASSERT_GIANT(),
and annotate that nfs_mountroot assumes it is OK to step on the
values in the global NFSv3 diskless structure as the mountroot
function is called during a serialized part of the boot, before
any other NFS client activity occurs.

MFC after: 2 weeks

19 years agoConvert a GIANT_REQUIRED; into a NET_ASSERT_GIANT();, as sockets are
rwatson [Sun, 5 Dec 2004 22:50:09 +0000 (22:50 +0000)]
Convert a GIANT_REQUIRED; into a NET_ASSERT_GIANT();, as sockets are
now only conditionally protected by Giant based on debug.mpsafenet.

19 years agoMake sure to map the whole kernel into 1MB pages. Try to use the remaining
cognet [Sun, 5 Dec 2004 22:48:04 +0000 (22:48 +0000)]
Make sure to map the whole kernel into 1MB pages. Try to use the remaining
memory for things such as the kernel stack.

19 years agoAssert the tcptw inpcb lock in tcp_timer_2msl_reset(), as fields in
rwatson [Sun, 5 Dec 2004 22:47:29 +0000 (22:47 +0000)]
Assert the tcptw inpcb lock in tcp_timer_2msl_reset(), as fields in
the tcptw undergo non-atomic read-modify-writes.

MFC after: 2 weeks

19 years agoReactivate the use of the minidata cache.
cognet [Sun, 5 Dec 2004 22:47:25 +0000 (22:47 +0000)]
Reactivate the use of the minidata cache.

19 years agoDo not change the page directory and do not flush the TLB when switching to
cognet [Sun, 5 Dec 2004 22:46:59 +0000 (22:46 +0000)]
Do not change the page directory and do not flush the TLB when switching to
a kernel thread.

19 years agoRemove an unused field from the struct pv_entry.
cognet [Sun, 5 Dec 2004 22:46:30 +0000 (22:46 +0000)]
Remove an unused field from the struct pv_entry.
While I'm there, fix style.

19 years agoVFS_STATFS(mp, ...) is mostly called with &mp->mnt_stat, but a few cases
phk [Sun, 5 Dec 2004 22:41:02 +0000 (22:41 +0000)]
VFS_STATFS(mp, ...) is mostly called with &mp->mnt_stat, but a few cases
doesn't.  Most of the implementations have grown weeds for this so they
copy some fields from mnt_stat if the passed argument isn't that.

Fix this the cleaner way:  Always call the implementation on mnt_stat
and copy that in toto to the VFS_STATFS argument if different.

19 years agoFix null-pointer indirect function calls introduced in the previous
marcel [Sun, 5 Dec 2004 22:30:28 +0000 (22:30 +0000)]
Fix null-pointer indirect function calls introduced in the previous
commit. In the new world order, the transitive closure on the vector
operations is not precomputed. As such, it's unsafe to actually use
any of the function pointers in an indirect function call. They can
be null, and we need to use the default vector in that case.
This is mostly a quick fix for the four function pointers that are
ed explicitly. A more generic or scalable solution is likely to see
the light of day.

No pathos on: current@

19 years agoAssert inpcb lock in:
rwatson [Sun, 5 Dec 2004 22:27:53 +0000 (22:27 +0000)]
Assert inpcb lock in:

  tcpip_fillheaders()
  tcp_discardcb()
  tcp_close()
  tcp_notify()
  tcp_new_isn()
  tcp_xmit_bandwidth_limit()

Fix a locking comment in tcp_twstart(): the pcbinfo will be locked (and
is asserted).

MFC after: 2 weeks

19 years agoMinor grammer fix in comment.
rwatson [Sun, 5 Dec 2004 22:20:59 +0000 (22:20 +0000)]
Minor grammer fix in comment.

19 years agoPass the inpcb reference into ip_getmoptions() rather than just the
rwatson [Sun, 5 Dec 2004 22:08:37 +0000 (22:08 +0000)]
Pass the inpcb reference into ip_getmoptions() rather than just the
inp->inp_moptions pointer, so that ip_getmoptions() can perform
necessary locking when doing non-atomic reads.

Lock the inpcb by default to copy any data to local variables, then
unlock before performing sooptcopyout().

MFC after: 2 weeks

19 years agoDefine INP_UNLOCK_ASSERT() to assert that an inpcb is unlocked.
rwatson [Sun, 5 Dec 2004 22:07:14 +0000 (22:07 +0000)]
Define INP_UNLOCK_ASSERT() to assert that an inpcb is unlocked.

MFC after: 2 weeks

19 years agoCorrect a sanity check in vnode_pager_generic_putpages(). The cast used
alc [Sun, 5 Dec 2004 21:48:11 +0000 (21:48 +0000)]
Correct a sanity check in vnode_pager_generic_putpages().  The cast used
to implement the sanity check should have been changed when we converted
the implementation of vm_pindex_t from 32 to 64 bits.  (Thus, RELENG_4 is
not affected.)  The consequence of this error would be a legimate write to
an extremely large file being treated as an errant attempt to write meta-
data.

Discussed with: tegge@

19 years ago'all' argument for list_net_interfaces() is now unused, remove it.
pjd [Sun, 5 Dec 2004 21:45:36 +0000 (21:45 +0000)]
'all' argument for list_net_interfaces() is now unused, remove it.

19 years agoPush the inpcb argument into ip_setmoptions() when setting IP multicast
rwatson [Sun, 5 Dec 2004 21:38:33 +0000 (21:38 +0000)]
Push the inpcb argument into ip_setmoptions() when setting IP multicast
socket options, so that it is available for locking.

19 years agoMake sure the first argument to the user function is 16-byte aligned.
deischen [Sun, 5 Dec 2004 21:22:08 +0000 (21:22 +0000)]
Make sure the first argument to the user function is 16-byte aligned.

Submitted by: tegge

19 years agoWhen panicing in device_unbusy(), actually tell what device has the issue.
obrien [Sun, 5 Dec 2004 20:58:56 +0000 (20:58 +0000)]
When panicing in device_unbusy(), actually tell what device has the issue.

19 years agoForced commit to note that the previous revision, 1.233, should be:
rwatson [Sun, 5 Dec 2004 19:12:06 +0000 (19:12 +0000)]
Forced commit to note that the previous revision, 1.233, should be:

MFC after: 2 weeks

19 years agoStart working through inpcb locking for ip_ctloutput() by cleaning up
rwatson [Sun, 5 Dec 2004 19:11:09 +0000 (19:11 +0000)]
Start working through inpcb locking for ip_ctloutput() by cleaning up
modifications to the inpcb IP options mbuf:

- Lock the inpcb before passing it into ip_pcbopts() in order to prevent
  simulatenous reads and read-modify-writes that could result in races.
- Pass the inpcb reference into ip_pcbopts() instead of the option chain
  pointer in the inpcb.
- Assert the inpcb lock in ip_pcbots.
- Convert one or two uses of a pointer as a boolean or an integer
  comparison to a comparison with NULL for readability.

19 years ago- Use uint16_t to pass argument for NGM_NETFLOW_IFINFO, bump cookie.
glebius [Sun, 5 Dec 2004 14:30:38 +0000 (14:30 +0000)]
- Use uint16_t to pass argument for NGM_NETFLOW_IFINFO, bump cookie.
- Always check that index number passed from userland
  is <= NG_NETFLOW_MAXIFACES. [1]
- Increase NG_NETFLOW_MAXIFACES up to 512. [2]

Noticed by: Roman Palagin [1]
Requested by: Yuri Y. Bushmelev [2]
MFC after: 1 week

19 years agoShow divert(4) sockets as well.
ru [Sun, 5 Dec 2004 14:17:14 +0000 (14:17 +0000)]
Show divert(4) sockets as well.

19 years agoCorrect a misspelling in a comment.
rwatson [Sun, 5 Dec 2004 13:28:52 +0000 (13:28 +0000)]
Correct a misspelling in a comment.

19 years agoAcquire socket receive buffer mutex before appending and then waking up
rwatson [Sun, 5 Dec 2004 13:27:30 +0000 (13:27 +0000)]
Acquire socket receive buffer mutex before appending and then waking up
a receive socket in DDP.  This reduces the number of mutex operations
required to deliver to a socket by two, and is the model used in other
protocols.

19 years agoDDB_UNATTENDED is now called KDB_UNATTENDED.
simon [Sun, 5 Dec 2004 12:48:07 +0000 (12:48 +0000)]
DDB_UNATTENDED is now called KDB_UNATTENDED.

MFC after: 3 days

19 years agoIPv6 packets can contain headers (like options) before the TCP/UDP/ICMP6
dhartmei [Sun, 5 Dec 2004 12:15:43 +0000 (12:15 +0000)]
IPv6 packets can contain headers (like options) before the TCP/UDP/ICMP6
header. pf finds the first TCP/UDP/ICMP6 header to filter by traversing
the header chain. In the case where headers are skipped, the protocol
checksum verification used the wrong length (included the skipped headers),
leading to incorrectly mismatching checksums. Such IPv6 packets with
headers were silently dropped.

Discovered by: Bernhard Schmidt
MFC after: 1 week

19 years agoUse "ifconfig -l" instead of "list_network_interfaces all" in
rse [Sun, 5 Dec 2004 09:51:48 +0000 (09:51 +0000)]
Use "ifconfig -l" instead of "list_network_interfaces all" in
ifnet_rename() to support situations where rc.conf's $network_interfaces
variable is set to an explicit list of network interfaces (instead of
the default "auto").

Using "list_network_interfaces all" resulted in using
$network_interfaces for both interface _renaming_ and interface
_configuration_ which obviously cannot work either before (if the
new name is in $network_interfaces) or after (if the old name is in
$network_interfaces) renaming the interface.

19 years agofix typo: s/intefraces/interfaces/
rse [Sun, 5 Dec 2004 09:01:20 +0000 (09:01 +0000)]
fix typo: s/intefraces/interfaces/

19 years agoStart to add GIANT_REQUIRED; macros in places where giant is required
imp [Sun, 5 Dec 2004 07:55:30 +0000 (07:55 +0000)]
Start to add GIANT_REQUIRED; macros in places where giant is required
and that I've verified things seem to basically work.  I was able to
boot and hot plug usb devices.  Please let me know if this causes
problems for anybody.

The push down of giant has proceeded to the point that this will start
to matter more and more.

19 years agoSource changes to allow building a cross-debugger. Move solib.c and
marcel [Sun, 5 Dec 2004 06:59:01 +0000 (06:59 +0000)]
Source changes to allow building a cross-debugger. Move solib.c and
solib-svr4.c to the MD makefiles because they are native files for
alpha and sparc64, but target files for amd64, i386 and ia64.

Note that kgdb(1) does not yet build as a cross-debugger due to
libkvm.

19 years agoo Don't limit GPT as a rank 2 provider. Allow it to be connected
marcel [Sun, 5 Dec 2004 06:02:21 +0000 (06:02 +0000)]
o  Don't limit GPT as a rank 2 provider. Allow it to be connected
   anywhere in the DAG. This includes configurations that are not
   allowed by the EFI specification.
o  Reject a GPT partition table if it's not preceeded by a PMBR.
   There's no need to preserve the MBR partitioning anymore as GPT
   is mature and with the first bullet extending the applicability
   of GPT, it's better to be a bit more strict.

19 years agoNow as a full blown port.
obrien [Sun, 5 Dec 2004 04:16:36 +0000 (04:16 +0000)]
Now as a full blown port.

19 years agoFix previous commit. GDB_CROSS_DEBUGGER needs to be defined for all
marcel [Sun, 5 Dec 2004 04:08:22 +0000 (04:08 +0000)]
Fix previous commit. GDB_CROSS_DEBUGGER needs to be defined for all
subdirectories.

19 years agoGrab Giant around calls to DEVICE_SUSPEND/RESUME in acpi_SetSleepState().
njl [Sun, 5 Dec 2004 01:35:18 +0000 (01:35 +0000)]
Grab Giant around calls to DEVICE_SUSPEND/RESUME in acpi_SetSleepState().
If we are resuming non-MPSAFE drivers, they need Giant held for them.
This may fix some obscure suspend/resume problems.  It has fixed keyrate
setting problems that were triggered by cardbus (MPSAFE) changing the
ordering for syscons resume (non-MPSAFE).  Also, add some asserts that
Giant is held in our suspend/resume and shutdown methods.

Found by: iedowse
MFC after: 2 days

19 years agoAdd more frequently-used locale directories. This is in preparation
kris [Sat, 4 Dec 2004 23:30:36 +0000 (23:30 +0000)]
Add more frequently-used locale directories.  This is in preparation
for cleanup of pkg-plist files with respect to handling of the share/locale
subdirectories.

MFC after: 3 days

19 years agoDocument ACPI tunable hw.acpi.serialize_methods.
marks [Sat, 4 Dec 2004 23:26:46 +0000 (23:26 +0000)]
Document ACPI tunable hw.acpi.serialize_methods.
Also reorder tunables.

Approved by: njl
MFC after: 1 week

19 years agoReplace (inlined) pmap_pte() calls with smaller, faster code where
alc [Sat, 4 Dec 2004 22:02:31 +0000 (22:02 +0000)]
Replace (inlined) pmap_pte() calls with smaller, faster code where
possible, such as the inner loop of pmap_copy().

Remove two comments that apply to i386 but not amd64.

19 years agoWhen initializing device, set d_softc and d_no fields for all components,
pjd [Sat, 4 Dec 2004 21:20:59 +0000 (21:20 +0000)]
When initializing device, set d_softc and d_no fields for all components,
because we know it then and we need it when inserting a component which
wasn't destroyed while device was running.

Reported by: Michael Handler <handler@grendel.net>
MFC after: 1 week

19 years ago- Change license to standard 2 clause BSD. [1]
simon [Sat, 4 Dec 2004 19:01:20 +0000 (19:01 +0000)]
- Change license to standard 2 clause BSD. [1]
- Add a missing period.

Noticed by: ru
Approved by: Jorge Mario G. Mazo <jgutie11@eafit.edu.co> [1]

19 years agoRevert rev. 1.166 and remove sparc64/pci/ofw_pci_if.m from MFILES again,
marius [Sat, 4 Dec 2004 14:20:56 +0000 (14:20 +0000)]
Revert rev. 1.166 and remove sparc64/pci/ofw_pci_if.m from MFILES again,
sparc64/pci/ofw_pci.h is no longer required for compiling modules.

19 years agoCatch up with the new device name of sab(4). The entries for tty[a,b]
marius [Sat, 4 Dec 2004 14:03:45 +0000 (14:03 +0000)]
Catch up with the new device name of sab(4). The entries for tty[a,b]
can't be removed as ofw_console(4) and zs(4) use them so one has to
live with some complaints about non-existent devices at boot time and
remove the respective entries locally for now.

19 years agoRemove #if 0'ed rootfs mounting code.
phk [Sat, 4 Dec 2004 09:58:20 +0000 (09:58 +0000)]
Remove #if 0'ed rootfs mounting code.

19 years agoRemove embryonic rootfs mounting facility.
phk [Sat, 4 Dec 2004 09:57:38 +0000 (09:57 +0000)]
Remove embryonic rootfs mounting facility.

In the near future rootfs mounting will not require special handling
in the filesystems.

19 years agoReplace GNU tar with BSD tar.
obrien [Sat, 4 Dec 2004 08:24:36 +0000 (08:24 +0000)]
Replace GNU tar with BSD tar.

19 years agokmod.mk knows how to create empty opt_*.h files so let it deal with the
brooks [Fri, 3 Dec 2004 23:43:02 +0000 (23:43 +0000)]
kmod.mk knows how to create empty opt_*.h files so let it deal with the
NOINET6 case.

Reported by: ru

19 years agoImplement a function, mount_arg() for accumulating a list of mount parameters
phk [Fri, 3 Dec 2004 22:38:06 +0000 (22:38 +0000)]
Implement a function, mount_arg() for accumulating a list of mount parameters
to nmount.

Make kernel_mount() accept the output from mount_arg() and know how to
free the malloc'ed space.

Make kernel_vmount() use the new function.

19 years agoSort and wash #includes.
phk [Fri, 3 Dec 2004 21:29:25 +0000 (21:29 +0000)]
Sort and wash #includes.

19 years agoWhen omount() is called, check if the filesystem have a cmount method
phk [Fri, 3 Dec 2004 21:14:46 +0000 (21:14 +0000)]
When omount() is called, check if the filesystem have a cmount method
and if so call it.

The cmount method will gather and interpret omount() style arguments,
and issue a kern_[v]mount() call to execute the corresponding nmount
operation.

19 years agotypo in comment.
phk [Fri, 3 Dec 2004 20:36:55 +0000 (20:36 +0000)]
typo in comment.

19 years agoAdd vfs_cmount() method to vfs_ops, this is to convert old-style mount
phk [Fri, 3 Dec 2004 19:33:19 +0000 (19:33 +0000)]
Add vfs_cmount() method to vfs_ops, this is to convert old-style mount
args to nmount request.

19 years agoAdd early checks for MNT_ROOTFS since we need to allow it later on in
phk [Fri, 3 Dec 2004 19:25:44 +0000 (19:25 +0000)]
Add early checks for MNT_ROOTFS since we need to allow it later on in
the code path.

19 years agoRetire unused vfs_mount() function in the name of nmount migration.
phk [Fri, 3 Dec 2004 18:40:58 +0000 (18:40 +0000)]
Retire unused vfs_mount() function in the name of nmount migration.

19 years ago- Simplify pcn_probe() by moving vendor/device matching code to pcn_match().
mdodd [Fri, 3 Dec 2004 18:35:00 +0000 (18:35 +0000)]
- Simplify pcn_probe() by moving vendor/device matching code to pcn_match().
- Avoid LOR in pcn_probe() by removing useless mutex stuff.

19 years ago- Move chip ID code into separate function.
mdodd [Fri, 3 Dec 2004 18:21:30 +0000 (18:21 +0000)]
- Move chip ID code into separate function.
- Initialize sc->pcn_type during ATTACH as softc contents may not surivive
  from PROBE.
- Print out chip-id to assist with ongoing pcn(4) debugging efforts.

19 years agoDo not blindly pass linux filesystem specific mount data across.
phk [Fri, 3 Dec 2004 18:14:22 +0000 (18:14 +0000)]
Do not blindly pass linux filesystem specific mount data across.

19 years agoDon not call pp_down()/pp_up() form XX_tlf()/XX_tls() in non PPP mode
rik [Fri, 3 Dec 2004 16:57:15 +0000 (16:57 +0000)]
Don not call pp_down()/pp_up() form XX_tlf()/XX_tls() in non PPP mode
to privent running of PPP's state machine in non PPP mode.

MFC: after 3 days.

19 years agoAdditional register definitions.
mdodd [Fri, 3 Dec 2004 16:45:11 +0000 (16:45 +0000)]
Additional register definitions.

Obtained from:  NetBSD

19 years agoIntroduce vfs_byname_kld() which will try to load the filesystem
phk [Fri, 3 Dec 2004 16:11:01 +0000 (16:11 +0000)]
Introduce vfs_byname_kld() which will try to load the filesystem
as a module if possible.

Use it so we don't have linker magic in the middle of the already
complex mount code.

19 years agoFix description: the argument to setcontext(3) must have been previously
rse [Fri, 3 Dec 2004 14:10:04 +0000 (14:10 +0000)]
Fix description: the argument to setcontext(3) must have been previously
initialized "by a call to getcontext(3) or makecontext(3)" and not "by a
call to setcontext(3) or makecontext(3)".

MFC after: 3 days

19 years agoAdd a manual page for the kernel's EVENTHANDLER facility.
jkoshy [Fri, 3 Dec 2004 14:05:55 +0000 (14:05 +0000)]
Add a manual page for the kernel's EVENTHANDLER facility.

Reviewed by: ru

19 years agoRemove an extra space.
harti [Fri, 3 Dec 2004 13:03:16 +0000 (13:03 +0000)]
Remove an extra space.

Submitted by: ru

19 years agoFix breakage introduced on 64-bit platforms with my last commit. Need
harti [Fri, 3 Dec 2004 12:55:57 +0000 (12:55 +0000)]
Fix breakage introduced on 64-bit platforms with my last commit. Need
to change to size_t in a couple of other places too.

19 years agoMake NAMEI_DIAGNOSTIC compile again and add a stragic vprint()
phk [Fri, 3 Dec 2004 12:15:39 +0000 (12:15 +0000)]
Make NAMEI_DIAGNOSTIC compile again and add a stragic vprint()

19 years agoImprove vprint() a little bit: break long lines, reduce indent and tell
phk [Fri, 3 Dec 2004 12:09:34 +0000 (12:09 +0000)]
Improve vprint() a little bit:  break long lines, reduce indent and tell
if the VI_LOCK() is held.

19 years agoPlug a memory leak.
harti [Fri, 3 Dec 2004 12:07:03 +0000 (12:07 +0000)]
Plug a memory leak.

19 years agoPut macro arguments in paranthesis.
harti [Fri, 3 Dec 2004 12:02:14 +0000 (12:02 +0000)]
Put macro arguments in paranthesis.

Submitted by: johan

19 years agoMake sizes to be of type size_t and correct function arguments that
harti [Fri, 3 Dec 2004 11:59:30 +0000 (11:59 +0000)]
Make sizes to be of type size_t and correct function arguments that
should be Byte (as the numerous casts to Byte in the function calls show).

19 years agoAdd missing vop_bypass (returning EOPNOTSUPP).
phk [Fri, 3 Dec 2004 08:56:30 +0000 (08:56 +0000)]
Add missing vop_bypass (returning EOPNOTSUPP).

Tripped up: marks