]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
19 years agoBump __FreeBSD_version for libalias and ng_nat.
glebius [Fri, 6 May 2005 15:04:18 +0000 (15:04 +0000)]
Bump __FreeBSD_version for libalias and ng_nat.

19 years agoAttach ng_nat and libalias to build.
glebius [Fri, 6 May 2005 14:47:54 +0000 (14:47 +0000)]
Attach ng_nat and libalias to build.

19 years agoDocument 'pentium-m' CPUTYPE.
pjd [Fri, 6 May 2005 13:14:48 +0000 (13:14 +0000)]
Document 'pentium-m' CPUTYPE.

19 years agoAdd a workaround for 64-bit archs: store unsigned long return value in
glebius [Fri, 6 May 2005 13:01:31 +0000 (13:01 +0000)]
Add a workaround for 64-bit archs: store unsigned long return value in
temporary variable, check it and then cast to in_addr_t.

19 years agoDocument FreeBSD-SA-05:06.iir, FreeBSD-SA-05:07.ldt, and
simon [Fri, 6 May 2005 11:46:31 +0000 (11:46 +0000)]
Document FreeBSD-SA-05:06.iir, FreeBSD-SA-05:07.ldt, and
FreeBSD-SA-05:08.kmem.

19 years agos/DEBUG/LIBALIAS_DEBUG/, since DEBUG is defined in LINT and
glebius [Fri, 6 May 2005 11:07:49 +0000 (11:07 +0000)]
s/DEBUG/LIBALIAS_DEBUG/, since DEBUG is defined in LINT and
not supported for kernel build.

19 years agoDon't try to enable event reporting for core file, it is not writable.
davidxu [Fri, 6 May 2005 11:01:15 +0000 (11:01 +0000)]
Don't try to enable event reporting for core file, it is not writable.

19 years agoFix race by using atomic operation, with this change, both libpthread
davidxu [Fri, 6 May 2005 07:37:01 +0000 (07:37 +0000)]
Fix race by using atomic operation, with this change, both libpthread
and libthr now can run profiling on SMP.

19 years agoIf we are going to
cperciva [Fri, 6 May 2005 02:50:00 +0000 (02:50 +0000)]
If we are going to
1. Copy a NULL-terminated string into a fixed-length buffer, and
2. copyout that buffer to userland,
we really ought to
0. Zero the entire buffer
first.

Security: FreeBSD-SA-05:08.kmem

19 years agoCorrectly validate inputs to the i386_get_ldt syscall.
cperciva [Fri, 6 May 2005 02:40:18 +0000 (02:40 +0000)]
Correctly validate inputs to the i386_get_ldt syscall.

Security: FreeBSD-SA-05:07.ldt

19 years agoCorrect improper permissions on /dev/iir. The earlier permissions
cperciva [Fri, 6 May 2005 02:33:46 +0000 (02:33 +0000)]
Correct improper permissions on /dev/iir.  The earlier permissions
of 0644 allowed for people to do Evil Things via ioctl(2).

Security: FreeBSD-SA-05:06.iir

19 years agoThe printf(9) `%p' conversion specifier puts an "0x" in
robert [Fri, 6 May 2005 00:15:57 +0000 (00:15 +0000)]
The printf(9) `%p' conversion specifier puts an "0x" in
front of the pointer value.  Therefore, remove the "0x"
from the format string.

19 years agoFix our NTFS readdir function.
robert [Fri, 6 May 2005 00:06:06 +0000 (00:06 +0000)]
Fix our NTFS readdir function.

To check a directory's in-use bitmap bit by bit, we use
a pointer to an 8 bit wide unsigned value.

The index used to dereference this pointer is calculated
by shifting the bit index right 3 bits.  Then we do a
logical AND with the bit# represented by the lower 3
bits of the bit index.

This is an idiomatic way of iterating through a bit map
with simple bitwise operations.

This commit fixes the bug that we only checked bits
3:0 of each 8 bit chunk, because we only used bits 1:0
of the bit index for the bit# in the current 8 bit value.
This resulted in files not being returned by getdirentries(2).

Change the type of the bit map pointer from `char *' to
`u_int8_t *'.

19 years agong_nat - a netgraph(4) node, which does NAT
glebius [Thu, 5 May 2005 23:41:21 +0000 (23:41 +0000)]
ng_nat - a netgraph(4) node, which does NAT

19 years agolibalias is now buildable as kernel module
glebius [Thu, 5 May 2005 22:43:04 +0000 (22:43 +0000)]
libalias is now buildable as kernel module

19 years agoCast 64 bit quantity to uintmax_t to print it with %jx. This is
wpaul [Thu, 5 May 2005 22:33:06 +0000 (22:33 +0000)]
Cast 64 bit quantity to uintmax_t to print it with %jx. This is
technically a no-op since uintmax_t is uint64_t on all currently
supported architectures, but we should use an explicit cast instead
of depending on this obscure coincidence.

19 years agoMore bits for kernel version:
glebius [Thu, 5 May 2005 22:00:32 +0000 (22:00 +0000)]
More bits for kernel version:
- copy inet_aton() from libc
- disable getservbyname() lookup and accept only numeric port

19 years agoAlways include alias.h before alias_local.h
glebius [Thu, 5 May 2005 21:55:17 +0000 (21:55 +0000)]
Always include alias.h before alias_local.h

19 years agoWhen used in kernel define NO_FW_PUNCH, NO_LOGGING, NO_USE_SOCKETS.
glebius [Thu, 5 May 2005 21:53:17 +0000 (21:53 +0000)]
When used in kernel define NO_FW_PUNCH, NO_LOGGING, NO_USE_SOCKETS.

19 years agoFix argument order for bcopy() in last commit.
glebius [Thu, 5 May 2005 21:40:49 +0000 (21:40 +0000)]
Fix argument order for bcopy() in last commit.

Noticed by: njl
Pointy hat to: glebius

19 years agoUse bcopy() instead of memmove().
glebius [Thu, 5 May 2005 21:10:51 +0000 (21:10 +0000)]
Use bcopy() instead of memmove().

19 years agoHide fflush(3) under ifdef DEBUG.
glebius [Thu, 5 May 2005 21:07:34 +0000 (21:07 +0000)]
Hide fflush(3) under ifdef DEBUG.

19 years agoThings required to build libalias as kernel module:
glebius [Thu, 5 May 2005 21:05:38 +0000 (21:05 +0000)]
Things required to build libalias as kernel module:
- kernel module declarations and handler.
- macros to map malloc(3) calls to malloc(9) ones.
- malloc(9) declarations.
- call finishoff() from module handler MOD_UNLOAD case
  instead of atexit(3).
- use panic(9) instead of abort(3)
- take time from time_second instead of gettimeofday(2)
- define INADDR_NONE

19 years agoAdd NO_USE_SOCKETS knob, which cuts off functionality socket binding.
glebius [Thu, 5 May 2005 20:25:12 +0000 (20:25 +0000)]
Add NO_USE_SOCKETS knob, which cuts off functionality socket binding.

19 years agoAdd NO_LOGGING knob, which cuts off functionality of debug logging to a file.
glebius [Thu, 5 May 2005 20:22:09 +0000 (20:22 +0000)]
Add NO_LOGGING knob, which cuts off functionality of debug logging to a file.

19 years agoPlay with includes so that libalias can be compiled both as userland
glebius [Thu, 5 May 2005 19:27:32 +0000 (19:27 +0000)]
Play with includes so that libalias can be compiled both as userland
library and kernel module.

19 years agoAdd quirk for TEAC USB floppy drives.
dwhite [Thu, 5 May 2005 18:48:41 +0000 (18:48 +0000)]
Add quirk for TEAC USB floppy drives.

19 years agoMove the pcb variable initialization earlier. This is cosmetic here, but
peter [Thu, 5 May 2005 18:26:45 +0000 (18:26 +0000)]
Move the pcb variable initialization earlier.  This is cosmetic here, but
in as-yet uncommitted code for 32 bit binary compatability on 64 bit
kernels, some of the 32 bit registers come from the pcb.  Moving the
initialization here means fill_regs32() etc are laid out the same.

19 years agoRemove unused (besides being initialized) variable.
peter [Thu, 5 May 2005 18:19:53 +0000 (18:19 +0000)]
Remove unused (besides being initialized) variable.

19 years agoUse smart quote entities instead of double ``ticks'' to create more
murray [Thu, 5 May 2005 17:10:37 +0000 (17:10 +0000)]
Use smart quote entities instead of double ``ticks'' to create more
aesthetically pleasing HTML output.

Tested with: mozilla, opera, lynx, links
Problematic with: w3m
Reviewed by: remko, keramida

19 years agoUse %jx instead of %qx to silence compiler warning on amd64.
wpaul [Thu, 5 May 2005 15:56:41 +0000 (15:56 +0000)]
Use %jx instead of %qx to silence compiler warning on amd64.

19 years agoCatching up with the latest crypto API changes in net80211.
avatar [Thu, 5 May 2005 13:16:30 +0000 (13:16 +0000)]
Catching up with the latest crypto API changes in net80211.

Reviewed by: sam

19 years agoICH5 is also supported by ichsmb
wilko [Thu, 5 May 2005 12:03:36 +0000 (12:03 +0000)]
ICH5 is also supported by ichsmb

19 years agoadd simple test script.
ume [Thu, 5 May 2005 11:26:47 +0000 (11:26 +0000)]
add simple test script.

19 years agoFix breakage on alpha.
takawata [Thu, 5 May 2005 07:02:51 +0000 (07:02 +0000)]
Fix breakage on alpha.

Pointed out by: hrs via IRC

19 years agoOnly check signal event, single threading event shouldn't be reported.
davidxu [Thu, 5 May 2005 06:42:02 +0000 (06:42 +0000)]
Only check signal event, single threading event shouldn't be reported.

19 years agoAvoid sleeping with mutex held in kern_ndis.c.
wpaul [Thu, 5 May 2005 06:14:59 +0000 (06:14 +0000)]
Avoid sleeping with mutex held in kern_ndis.c.

Remove unused fields from ndis_miniport_block.

Fix a bug in KeFlushQueuedDpcs() (we weren't calculating the kq pointer
correctly).

In if_ndis.c, clear the IFF_RUNNING flag before calling ndis_halt_nic().

Add some guards in kern_ndis.c to avoid letting anyone invoke ndis_get_info()
or ndis_set_info() if the NIC isn't fully initialized. Apparently, mdnsd
will sometimes try to invoke the ndis_ioctl() routine at exactly the
wrong moment (to futz with its multicast filters) when the interface
comes up, and can trigger a crash unless we guard against it.

19 years agoRemove extranaous free() of ASCII filename from NdisOpenFile().
wpaul [Thu, 5 May 2005 04:16:13 +0000 (04:16 +0000)]
Remove extranaous free() of ASCII filename from NdisOpenFile().

Oh, one additional change I forgot to mention in the last commit:
NdisOpenFile() was broken in the case for firmware files that were
pre-loaded as modules. When searching for the module in NdisOpenFile(),
we would match against a symbol name, which would contain the string
we were looking for, then save a pointer to the linker file handle.
Later, in NdisMapFile(), we would refer to the filename hung off
this handle when trying to find the starting address symbol. Only
problem is, this filename is different from the embedded symbol
name we're searching for, so the mapping would fail. I found this
problem while testing the AirGo driver, which requires a small
firmware file.

19 years agoThis commit makes a bunch of changes, some big, some not so big.
wpaul [Thu, 5 May 2005 03:56:09 +0000 (03:56 +0000)]
This commit makes a bunch of changes, some big, some not so big.

- Remove the old task threads from kern_ndis.c and reimplement them in
  subr_ntoskrnl.c, in order to more properly emulate the Windows DPC
  API. Each CPU gets its own DPC queue/thread, and each queue can
  have low, medium and high importance DPCs. New APIs implemented:
  KeSetTargetProcessorDpc(), KeSetImportanceDpc() and KeFlushQueuedDpcs().
  (This is the biggest change.)

- Fix a bug in NdisMInitializeTimer(): the k_dpc pointer in the
  nmt_timer embedded in the ndis_miniport_timer struct must be set
  to point to the DPC, also embedded in the struct. Failing to do
  this breaks dequeueing of DPCs submitted via timers, and in turn
  breaks cancelling timers.

- Fix a bug in KeCancelTimer(): if the timer is interted in the timer
  queue (i.e. the timeout callback is still pending), we have to both
  untimeout() the timer _and_ call KeRemoveQueueDpc() to nuke the DPC
  that might be pending. Failing to do this breaks cancellation of
  periodic timers, which always appear to be inserted in the timer queue.

- Make use of the nmt_nexttimer field in ndis_miniport_timer: keep a
  queue of pending timers and cancel them all in ndis_halt_nic(), prior
  to calling MiniportHalt(). Also call KeFlushQueuedDpcs() to make sure
  any DPCs queued by the timers have expired.

- Modify NdisMAllocateSharedMemory() and NdisMFreeSharedMemory() to keep
  track of both the virtual and physical addresses of the shared memory
  buffers that get handed out. The AirGo MIMO driver appears to have a bug
  in it: for one of the segments is allocates, it returns the wrong
  virtual address. This would confuse NdisMFreeSharedMemory() and cause
  a crash. Why it doesn't crash Windows too I have no idea (from reading
  the documentation for NdisMFreeSharedMemory(), it appears to be a violation
  of the API).

- Implement strstr(), strchr() and MmIsAddressValid().

- Implement IoAllocateWorkItem(), IoFreeWorkItem(), IoQueueWorkItem() and
  ExQueueWorkItem(). (This is the second biggest change.)

- Make NdisScheduleWorkItem() call ExQueueWorkItem(). (Note that the
  ExQueueWorkItem() API is deprecated by Microsoft, but NDIS still uses
  it, since NdisScheduleWorkItem() is incompatible with the IoXXXWorkItem()
  API.)

- Change if_ndis.c to use the NdisScheduleWorkItem() interface for scheduling
  tasks.

With all these changes and fixes, the AirGo MIMO driver for the Belkin
F5D8010 Pre-N card now works. Special thanks to Paul Robinson
(paul dawt robinson at pwermedia dawt net) for the loan of a card
for testing.

19 years agoBackout part of rev 1.71, which breaks the interfaces on IBM/Intel blade
dwhite [Thu, 5 May 2005 03:37:27 +0000 (03:37 +0000)]
Backout part of rev 1.71, which breaks the interfaces on IBM/Intel blade
servers.

PR: kern/68445
MFC after: 7 days

19 years agoTurn on PCB_FULLCTX in set_regs to fully restore context
davidxu [Wed, 4 May 2005 22:26:40 +0000 (22:26 +0000)]
Turn on PCB_FULLCTX in set_regs to fully restore context
set by debugger.

19 years ago - We need to inhert the OBJ_NEEDGIANT flag from the original object in
jeff [Wed, 4 May 2005 20:54:16 +0000 (20:54 +0000)]
 - We need to inhert the OBJ_NEEDGIANT flag from the original object in
   vm_object_split().

Spotted by: alc

19 years agoDefine HAVE_STRUCT_UTMP_UT_HOST in order for the REMOTEHOST environment
mp [Wed, 4 May 2005 20:21:57 +0000 (20:21 +0000)]
Define HAVE_STRUCT_UTMP_UT_HOST in order for the REMOTEHOST environment
variable to show up correctly.

Note: This is really a bug with the stock tcsh configure.in script not
including <sys/types.h> before <utmp.h>. But, since config.h is maintained
separate from the contrib sources, this is the correct fix.

Noticed by: ache

19 years agoif_mtu not ifp_mtu.
mlaier [Wed, 4 May 2005 19:51:06 +0000 (19:51 +0000)]
if_mtu not ifp_mtu.

19 years ago- add $FreeBSD$.
ume [Wed, 4 May 2005 19:40:30 +0000 (19:40 +0000)]
- add $FreeBSD$.
- enable -r option for regress target.

19 years agoChange m_uiotombuf so it will accept offset at which data should be copied
emax [Wed, 4 May 2005 18:55:03 +0000 (18:55 +0000)]
Change m_uiotombuf so it will accept offset at which data should be copied
to the mbuf. Offset cannot exceed MHLEN bytes. This is currently used to
fix Ethernet header alignment problem on alpha and sparc64. Also change all
users of m_uiotombuf to pass proper offset.

Reviewed by: jmg, sam
Tested by: Sten Spans "sten AT blinkenlights DOT nl"
MFC after: 1 week

19 years agoFix a bug in NFS/TCP where retransmissions would not reliably happen
ps [Wed, 4 May 2005 16:37:31 +0000 (16:37 +0000)]
Fix a bug in NFS/TCP where retransmissions would not reliably happen
if the server rebooted or tore down the connection for any reason.

Found by: Jonathan Noack.
Submitted by: Mohan Srinivasan.

19 years agoBring back fix from rev. 1.28 which was lost during the import.
mlaier [Wed, 4 May 2005 16:03:01 +0000 (16:03 +0000)]
Bring back fix from rev. 1.28 which was lost during the import.

19 years agoCombine rev. 1.29 and 1.30 to something that will make sense for future
mlaier [Wed, 4 May 2005 15:55:29 +0000 (15:55 +0000)]
Combine rev. 1.29 and 1.30 to something that will make sense for future
imports.

19 years agoMake LINT happy.
mlaier [Wed, 4 May 2005 15:29:28 +0000 (15:29 +0000)]
Make LINT happy.

19 years agoMake smbfs capable to use 16bit char set in filenames.
takawata [Wed, 4 May 2005 15:05:46 +0000 (15:05 +0000)]
Make smbfs capable to use 16bit char set in filenames.

PR:78110

19 years agoAdd more technical terms to reduce false positives when spellchecking
murray [Wed, 4 May 2005 14:19:36 +0000 (14:19 +0000)]
Add more technical terms to reduce false positives when spellchecking
FreeBSD documentation.

19 years agoIf we don't get a suggested MTU during path MTU discovery
andre [Wed, 4 May 2005 13:48:44 +0000 (13:48 +0000)]
If we don't get a suggested MTU during path MTU discovery
look up the packet size of the packet that generated the
response, step down the MTU by one step through ip_next_mtu()
and try again.

Suggested by: dwmalone

19 years agoCleanup IPFW2 ifdefs.
glebius [Wed, 4 May 2005 13:24:37 +0000 (13:24 +0000)]
Cleanup IPFW2 ifdefs.

19 years agoMakefile is not needed here.
glebius [Wed, 4 May 2005 13:24:12 +0000 (13:24 +0000)]
Makefile is not needed here.

19 years agoAdd another step of 1280 (gif(4) tunnels) to ip_next_mtu().
andre [Wed, 4 May 2005 13:23:54 +0000 (13:23 +0000)]
Add another step of 1280 (gif(4) tunnels) to ip_next_mtu().

19 years agoIPFW version 2 is the only option now in HEAD. Do not confuse
glebius [Wed, 4 May 2005 13:14:57 +0000 (13:14 +0000)]
IPFW version 2 is the only option now in HEAD. Do not confuse
users of future releases with instructions about building IPFW2
on RELENG_4.

19 years agoIPFW version 2 is the only option in HEAD and RELENG_5.
glebius [Wed, 4 May 2005 13:12:52 +0000 (13:12 +0000)]
IPFW version 2 is the only option in HEAD and RELENG_5.
Thus, cleanup unnecessary now ifdefs.

19 years agoPass icmp_error() the MTU argument directly instead of
andre [Wed, 4 May 2005 13:09:19 +0000 (13:09 +0000)]
Pass icmp_error() the MTU argument directly instead of
an interface pointer.  This simplifies a couple of uses
and removes some XXX workarounds.

19 years agoBuild userland libalias using src/sys/netinet/libalias.
glebius [Wed, 4 May 2005 12:49:44 +0000 (12:49 +0000)]
Build userland libalias using src/sys/netinet/libalias.

Reviewed by: ru
Repocopy by: peter

19 years agoAdd -m option (find module using modfind() and stat it).
fjoe [Wed, 4 May 2005 12:46:43 +0000 (12:46 +0000)]
Add -m option (find module using modfind() and stat it).

19 years ago- add $FreeBSD$.
ume [Wed, 4 May 2005 12:02:10 +0000 (12:02 +0000)]
- add $FreeBSD$.
- do service lookup.
- add options to use gethostbyname(3) or getipnodebyname(3) instead
  of getaddrinfo(3).
- add option to do reverse lookup.

19 years agoThis commit was generated by cvs2svn to compensate for changes in r145857,
ume [Wed, 4 May 2005 11:55:42 +0000 (11:55 +0000)]
This commit was generated by cvs2svn to compensate for changes in r145857,
which included commits to RCS files with non-trunk default branches.

19 years agoimport of parallel name resolution test.
ume [Wed, 4 May 2005 11:55:42 +0000 (11:55 +0000)]
import of parallel name resolution test.

Obtained from: NetBSD

19 years agoDocument the fact that accept(2) may return EINVAL when addrlen is
keramida [Wed, 4 May 2005 11:09:26 +0000 (11:09 +0000)]
Document the fact that accept(2) may return EINVAL when addrlen is
negative (in addition to returning EINVAL when called on a descriptor
that is not a socket).

Submitted by: Arne H Juul <arnej@europe.yahoo-inc.com>
PR: docs/80587

19 years agoIntroduce MAC Framework and MAC Policy entry points to label and control
rwatson [Wed, 4 May 2005 10:39:15 +0000 (10:39 +0000)]
Introduce MAC Framework and MAC Policy entry points to label and control
access to POSIX Semaphores:

mac_init_posix_sem()            Initialize label for POSIX semaphore
mac_create_posix_sem()          Create POSIX semaphore
mac_destroy_posix_sem()         Destroy POSIX semaphore
mac_check_posix_sem_destroy()   Check whether semaphore may be destroyed
mac_check_posix_sem_getvalue()  Check whether semaphore may be queried
mac_check_possix_sem_open()     Check whether semaphore may be opened
mac_check_posix_sem_post()      Check whether semaphore may be posted to
mac_check_posix_sem_unlink()    Check whether semaphore may be unlinked
mac_check_posix_sem_wait()      Check whether may wait on semaphore

Update Biba, MLS, Stub, and Test policies to implement these entry points.
For information flow policies, most semaphore operations are effectively
read/write.

Submitted by: Dandekar Hrishikesh <rishi_dandekar at sbcglobal dot net>
Sponsored by: DARPA, McAfee, SPARTA
Obtained from: TrustedBSD Project

19 years agoUse libutil in DPADD as well.
mlaier [Wed, 4 May 2005 08:57:12 +0000 (08:57 +0000)]
Use libutil in DPADD as well.

Submitted by: ru

19 years agoThe correct description for mode "w" is
des [Wed, 4 May 2005 08:12:44 +0000 (08:12 +0000)]
The correct description for mode "w" is
  (((truncate to zero length) or (create)) (text file)) (for writing)
and not
  ((truncate file to zero length) or (create text file)) (for writing)

MFC after: 1 week

19 years ago-introduce net.bpf sysctl instead of the less intuitive debug.*
csjp [Wed, 4 May 2005 03:09:28 +0000 (03:09 +0000)]
-introduce net.bpf sysctl instead of the less intuitive debug.*

    debug.bpf_bufsize is now net.bpf.bufsize
    debug.bpf_maxbufsize is now net.bpf.maxbufsize

-move function prototypes for bpf_drvinit and bpf_clone up to the
 top of the file with the others
-assert bpfd lock in catchpacket() and bpf_wakeup()

MFC after: 2 weeks

19 years agoA patch to support Palm Tungsten T via USB-Cradle.
julian [Wed, 4 May 2005 00:46:24 +0000 (00:46 +0000)]
A patch to support Palm Tungsten T via USB-Cradle.
not suer where it comes from but suspect kimoto at ohnolab.org

MFC after: 1 week

19 years agodo mutex lock for each yp calls. with this,
ume [Tue, 3 May 2005 20:30:31 +0000 (20:30 +0000)]
do mutex lock for each yp calls.  with this,
  http://cvsweb.netbsd.org/bsdweb.cgi/src/regress/lib/libpthread/resolv/
is working.

19 years agoMove definitions of 'struct kuser' and 'struct ksem' from uipc_sem.c
rwatson [Tue, 3 May 2005 20:21:24 +0000 (20:21 +0000)]
Move definitions of 'struct kuser' and 'struct ksem' from uipc_sem.c
to ksem.h so that they are accessible from the MAC Framework for the
purposes of labeling and enforcing additional protections.  #error
if these are included without _KERNEL, since they are not intended
(nor installed) for user application use.

Submitted by: Dandekar Hrishikesh <rishi_dandekar at sbcglobal dot net>
Sponsored by: DARPA, SPARTA
Obtained from: TrustedBSD Project

19 years agoBump __FreeBSD_version for pf 3.7 and inform about user visible changes.
mlaier [Tue, 3 May 2005 17:43:14 +0000 (17:43 +0000)]
Bump __FreeBSD_version for pf 3.7 and inform about user visible changes.

19 years agoUpdate ifconfig's pfsync module for 3.7 pfsync(4)
mlaier [Tue, 3 May 2005 16:59:14 +0000 (16:59 +0000)]
Update ifconfig's pfsync module for 3.7 pfsync(4)

19 years agoAdapt Makefiles for pfctl(8) and authpf(8) to 3.7 sources.
mlaier [Tue, 3 May 2005 16:57:38 +0000 (16:57 +0000)]
Adapt Makefiles for pfctl(8) and authpf(8) to 3.7 sources.

19 years agoResolve conflicts created during the import of pf 3.7 Some features are
mlaier [Tue, 3 May 2005 16:55:20 +0000 (16:55 +0000)]
Resolve conflicts created during the import of pf 3.7 Some features are
missing and will be implemented in a second step.  This is functional as is.

Tested by: freebsd-pf, pfsense.org
Obtained from: OpenBSD

19 years agoThis commit was generated by cvs2svn to compensate for changes in r145837,
mlaier [Tue, 3 May 2005 16:47:37 +0000 (16:47 +0000)]
This commit was generated by cvs2svn to compensate for changes in r145837,
which included commits to RCS files with non-trunk default branches.

19 years agoImport pf userland from OpenBSD 3.7 (OPENBSD_3_7 as of today)
mlaier [Tue, 3 May 2005 16:47:37 +0000 (16:47 +0000)]
Import pf userland from OpenBSD 3.7 (OPENBSD_3_7 as of today)

19 years agoResolve conflicts created during the import of pf 3.7 Some features are
mlaier [Tue, 3 May 2005 16:43:32 +0000 (16:43 +0000)]
Resolve conflicts created during the import of pf 3.7 Some features are
missing and will be implemented in a second step.  This is functional as is.

Tested by: freebsd-pf, pfsense.org
Obtained from: OpenBSD
X-MFC after: never (breaks API/ABI)

19 years ago - Initialize vfslocked correctly early enough for MAC to compile.
jeff [Tue, 3 May 2005 16:24:59 +0000 (16:24 +0000)]
 - Initialize vfslocked correctly early enough for MAC to compile.
 - Fix one place where we explicitly drop Giant!

Pointy hat to: me
Submitted by: Max Laier
Warned by: Tinderbox

19 years agoCleanup for getgrouplist(3):
delphij [Tue, 3 May 2005 16:20:03 +0000 (16:20 +0000)]
Cleanup for getgrouplist(3):
 - Use /*- instead of /* for copyright section
 - Include unistd.h for prototype of it
 - Sort and separate includes as described in style(9)
 - ANSIfy the function defination
 - Use const for the traversing iterator

19 years agofix signed/unsigned comparison warnings.
ume [Tue, 3 May 2005 16:13:31 +0000 (16:13 +0000)]
fix signed/unsigned comparison warnings.

Obtained from: NetBSD

19 years agoInclude paths.h for prototype of getbootfile(3).
delphij [Tue, 3 May 2005 15:28:07 +0000 (15:28 +0000)]
Include paths.h for prototype of getbootfile(3).

19 years ago- move to SCHED_4BSD per jeffr's comments on SCHED_ULE's state
grehan [Tue, 3 May 2005 11:56:05 +0000 (11:56 +0000)]
- move to SCHED_4BSD per jeffr's comments on SCHED_ULE's state
- enable MSDOSFS
- ehci is stable on the powerbook
- modules have been working for a long time.

19 years ago - Add a new object flag "OBJ_NEEDSGIANT". We set this flag if the
jeff [Tue, 3 May 2005 11:11:26 +0000 (11:11 +0000)]
 - Add a new object flag "OBJ_NEEDSGIANT".  We set this flag if the
   underlying vnode requires Giant.
 - In vm_fault only acquire Giant if the underlying object has NEEDSGIANT
   set.
 - In vm_object_shadow inherit the NEEDSGIANT flag from the backing object.

19 years ago - Set the v_object pointer after a successful VOP_OPEN(). This isn't a
jeff [Tue, 3 May 2005 11:05:33 +0000 (11:05 +0000)]
 - Set the v_object pointer after a successful VOP_OPEN().  This isn't a
   perfect solution as the lower vm object can change at unpredictable times
   if our lower vp happens to be on another unionfs, etc.

Submitted by: Oleg Sharoiko <os@rsu.ru>

19 years ago - Don't restrict the softdep stats to DEBUG kernels, they cost nothing to
jeff [Tue, 3 May 2005 11:03:29 +0000 (11:03 +0000)]
 - Don't restrict the softdep stats to DEBUG kernels, they cost nothing to
   export.  This was happening anyway since this file manually sets DEBUG.
 - Add a sysctl for the number of items on the worklist.
 - Use a more canonical loop restart in softdep_fsync_mountdev, it saves
   some code at the expense of a goto and makes me worry less about
   modifying a variable that should be private to the TAILQ_FOREACH_SAFE
   macro.

19 years ago - Remove two mtx_asserts that can incorrectly trigger if
jeff [Tue, 3 May 2005 10:58:05 +0000 (10:58 +0000)]
 - Remove two mtx_asserts that can incorrectly trigger if
   devstat_end_transaction is called from a fast interrupt.  Presently
   there is no way for mtx_assert to determine that we're not executing
   in a real thread context.

Submitted by: jhusted@isilon.com

19 years ago - A vnode may have made its way onto the free list while it was being
jeff [Tue, 3 May 2005 10:56:00 +0000 (10:56 +0000)]
 - A vnode may have made its way onto the free list while it was being
   vgone'd.  We must remove it from the freelist before returning in
   vtryrecycle() or we may get a duplicate free.

Reported by: kkenn

19 years ago - Use namei to acquire Giant for VFS if it is necessary. Drop the explicit
jeff [Tue, 3 May 2005 10:55:05 +0000 (10:55 +0000)]
 - Use namei to acquire Giant for VFS if it is necessary.  Drop the explicit
   Giant acquisition.
 - Remove GIANT_REQUIRED in the few remaining cases; the vm and vfs have
   both been locked.

19 years ago - Use NAMEI to pickup Giant if we need it in fpcheckstd().
jeff [Tue, 3 May 2005 10:52:22 +0000 (10:52 +0000)]
 - Use NAMEI to pickup Giant if we need it in fpcheckstd().

19 years ago - Neither of our image formats require Giant now that the vm and vfs have
jeff [Tue, 3 May 2005 10:51:38 +0000 (10:51 +0000)]
 - Neither of our image formats require Giant now that the vm and vfs have
   been locked.

19 years agoReshape the dma code to be a bit more flexible so it can cope with
sos [Tue, 3 May 2005 07:55:07 +0000 (07:55 +0000)]
Reshape the dma code to be a bit more flexible so it can cope with
new HW that has new and different demands.
Fix a few nits in former commit in this cleanup crusade.

Sponsored by: pair.com

19 years agoProperly mask off the status bits when checking to see if the ccb is still
scottl [Tue, 3 May 2005 07:11:19 +0000 (07:11 +0000)]
Properly mask off the status bits when checking to see if the ccb is still
valid to process.  This was causing deferred commands to be rejected due
to their extra status flag.

MFC After: 3 days

19 years agoThe driver looks like it can create valid 64-bit scatter-gather lists, so
scottl [Tue, 3 May 2005 05:44:42 +0000 (05:44 +0000)]
The driver looks like it can create valid 64-bit scatter-gather lists, so
don't restrict it to a 32-bit address space.  Also use the correct busdma
flags for the SRB memory area.

MFC After: 3 days

19 years agoThe kthread is disabled at this time, so don't try to wake it up on
scottl [Tue, 3 May 2005 05:42:03 +0000 (05:42 +0000)]
The kthread is disabled at this time, so don't try to wake it up on
shutdown.  This fixes a panic on reboot.

MFC After: 3 days

19 years agoPut gnome-lite and kde-lite on disc2 for ia64. The full gnome and kde
marcel [Tue, 3 May 2005 05:37:09 +0000 (05:37 +0000)]
Put gnome-lite and kde-lite on disc2 for ia64. The full gnome and kde
packages is causing the disc2 ISO to overflow in excess of 100MB.

MFC after: 2 days

19 years agoRemove only the packages sub-directory before recreating it. Do not
marcel [Tue, 3 May 2005 04:19:07 +0000 (04:19 +0000)]
Remove only the packages sub-directory before recreating it. Do not
remove the whole disc treei structure. This allows one to specify the
disc trees created by a prior release build (under R/cdrom) as the
destination without destroying the contents. This better integrates
with release building and makes further automation easier.

MFC after: 2 days

19 years agoAllow specifying which INDEX file to use by setting PKG_INDEX. By default
marcel [Tue, 3 May 2005 04:11:56 +0000 (04:11 +0000)]
Allow specifying which INDEX file to use by setting PKG_INDEX. By default
the INDEX file is taken from the package source tree as defined by the
PKG_TREE variable. This change allows using the (possibly incomplete)
packages on pointyhat.

MFC after: 2 days

19 years agoFix some busdma API violations in the dumpsys handler.
scottl [Mon, 2 May 2005 22:56:52 +0000 (22:56 +0000)]
Fix some busdma API violations in the dumpsys handler.

MFC After: 3 days