]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
19 years agoNote addition of est(4)
njl [Sun, 20 Feb 2005 20:30:47 +0000 (20:30 +0000)]
Note addition of est(4)

19 years agoHook EST up to the build.
njl [Sun, 20 Feb 2005 20:29:04 +0000 (20:29 +0000)]
Hook EST up to the build.

19 years agoAdd the Enhanced SpeedStep driver (EST). Currently, this driver only works
njl [Sun, 20 Feb 2005 20:27:59 +0000 (20:27 +0000)]
Add the Enhanced SpeedStep driver (EST).  Currently, this driver only works
on the previous generation of Pentium-M processors (Banias).  Support for
Dothan and later processors involves working with acpi_perf(4) to extract
information about supported states.  This driver should work on MP systems
including HTT.  It is experimental and may have a few bugs but has been
tested to not crash at least.

Thanks to Colin Percival for his initial work on this driver.

19 years agoIn unp_attach(), allow uma_zalloc to zero the new unpcb rather than
rwatson [Sun, 20 Feb 2005 20:05:11 +0000 (20:05 +0000)]
In unp_attach(), allow uma_zalloc to zero the new unpcb rather than
explicitly using bzero().

Update copyright.

MFC after: 3 days

19 years agoFix merging botch I made in last version.
imp [Sun, 20 Feb 2005 19:59:29 +0000 (19:59 +0000)]
Fix merging botch I made in last version.

19 years agoPerliminary support for Fujitsu SCSI LAN combo card. At least for the
imp [Sun, 20 Feb 2005 19:40:29 +0000 (19:40 +0000)]
Perliminary support for Fujitsu SCSI LAN combo card.  At least for the
LAN side of things.  It appears that all that's necessary is to relax
the check on the DLCR2 register in the probe.

19 years agoBack out part of 1.23. This was a work in progress to support the
imp [Sun, 20 Feb 2005 19:38:38 +0000 (19:38 +0000)]
Back out part of 1.23.  This was a work in progress to support the
Fujitsu LAN SCSI card's ethernet side.

19 years agoForced commit after moving mb86960.h from i386/isa/ic to dev/fe. It
imp [Sun, 20 Feb 2005 19:33:13 +0000 (19:33 +0000)]
Forced commit after moving mb86960.h from i386/isa/ic to dev/fe.  It
is only used by the fe driver, and isn't MD.

19 years agoMinor optimizations to compile out bits that aren't necessary when
imp [Sun, 20 Feb 2005 18:44:45 +0000 (18:44 +0000)]
Minor optimizations to compile out bits that aren't necessary when
certain options aren't enabled.

Submitted by: Chiharu Shibata-san (chi at bd mbn or jp)

19 years agoMinor style(9) tweaks.
imp [Sun, 20 Feb 2005 18:39:54 +0000 (18:39 +0000)]
Minor style(9) tweaks.

19 years agoCorrect typo in comment.
imp [Sun, 20 Feb 2005 18:24:09 +0000 (18:24 +0000)]
Correct typo in comment.

Submitted by: Chiharu Shibata-san (chi at bd mbn or jp)

19 years agoThis commit was generated by cvs2svn to compensate for changes in r142129,
mikeh [Sun, 20 Feb 2005 17:33:34 +0000 (17:33 +0000)]
This commit was generated by cvs2svn to compensate for changes in r142129,
which included commits to RCS files with non-trunk default branches.

19 years agoImport the latest CVS version of lukemftp.
mikeh [Sun, 20 Feb 2005 17:33:34 +0000 (17:33 +0000)]
Import the latest CVS version of lukemftp.

Short list of changes:

 * SIGINT termination from auto-fetch.
 * Less trusting of remote filenames during auto mgets.
 * Improved RFC2616 compliancy.
 * Fix globs when using ftp reget (from mat@).
 * Limit send buffer size.

19 years agoPrefer NULL to returning 0 cast to a pointer type.
rwatson [Sun, 20 Feb 2005 15:56:13 +0000 (15:56 +0000)]
Prefer NULL to returning 0 cast to a pointer type.

MFC after: 3 days

19 years agoIn soreceive(), when considering delivery to a socket in SS_ISCONFIRMING,
rwatson [Sun, 20 Feb 2005 15:54:44 +0000 (15:54 +0000)]
In soreceive(), when considering delivery to a socket in SS_ISCONFIRMING,
only call the protocol's pru_rcvd() if the protocol has the flag
PR_WANTRCVD set.  This brings that instance of pru_rcvd() into line with
the rest, which do check the flag.

MFC after: 3 days

19 years agoWhen creating a new FFS file system, the block size will indirectly
delphij [Sun, 20 Feb 2005 11:32:49 +0000 (11:32 +0000)]
When creating a new FFS file system, the block size will indirectly
affect the largest file size that is allowed by the file system.
On the other hand, when creating a snapshot, the snapshot file will
appear as it is as big as the file system itself.  Hence we will not
be able to create a snapshot on large file systems with small block
sizes.

Add a warning about this, and gives some hints to correct the issue.

Reviewed by:    mckusick
MFC After:      1 week

19 years agoThe recomputation of file system summary at mount time can be a
delphij [Sun, 20 Feb 2005 08:02:15 +0000 (08:02 +0000)]
The recomputation of file system summary at mount time can be a
very slow process, especially for large file systems that is just
recovered from a crash.

Since the summary is already re-sync'ed every 30 second, we will
not lag behind too much after a crash.  With this consideration
in mind, it is more reasonable to transfer the responsibility to
background fsck, to reduce the delay after a crash.

Add a new sysctl variable, vfs.ffs.compute_summary_at_mount, to
control this behavior.  When set to nonzero, we will get the
"old" behavior, that the summary is computed immediately at mount
time.

Add five new sysctl variables to adjust ndir, nbfree, nifree,
nffree and numclusters respectively.  Teach fsck_ffs about these
API, however, intentionally not to check the existence, since
kernels without these sysctls must have recomputed the summary
and hence no adjustments are necessary.

This change has eliminated the usual tens of minutes of delay of
mounting large dirty volumes.

Reviewed by: mckusick
MFC After: 1 week

19 years agoNew release notes: ULE fixes, mpsafevfs, NDISulator for x86-64
bmah [Sun, 20 Feb 2005 07:27:16 +0000 (07:27 +0000)]
New release notes:  ULE fixes, mpsafevfs, NDISulator for x86-64
binaries, sk(4) ALTQ and jumbo frames on Yukons, cd9660 less chatty,
burncd(8) fixate eject works, mknod(8) deprecated, NOFOO -> NO_FOO,
ppp(8) echo, tcpdrop(8), whois(1) -k.

Modified release note:  In netcat note, provide cross-reference to
nc(1) manpage.

Just another random walk through the commit messages...

19 years agoWhen creating a new FFS file system, the block size will indirectly
delphij [Sun, 20 Feb 2005 06:33:18 +0000 (06:33 +0000)]
When creating a new FFS file system, the block size will indirectly
affect the largest file size that is allowed by the file system.
On the other hand, when creating a snapshot, the snapshot file will
appear as it is as big as the file system itself.  Hence we will not
be able to create a file system on large file systems with small
block sizes.

Add a warning about this, and gives some hints to correct the issue.

Reviewed by: mckusick
MFC After: 1 week

19 years agoMove assignment of UNIX domain socket pcb during unp_attach() outside
rwatson [Sun, 20 Feb 2005 04:18:22 +0000 (04:18 +0000)]
Move assignment of UNIX domain socket pcb during unp_attach() outside
of the global UNIX domain socket mutex: no protection is needed that
early in the setup of the UNIX domain socket and socket structures.

MFC after: 3 days

19 years agoAdd a note that all four HME chips on a SUNW,qfe card share the same PROM,
marius [Sun, 20 Feb 2005 01:52:43 +0000 (01:52 +0000)]
Add a note that all four HME chips on a SUNW,qfe card share the same PROM,
making it a bit more clear why we can read four NAs from the VPD.

MFC after: 1 day

19 years agoRemoving the #endif as well sounds like a good idea.
cognet [Sun, 20 Feb 2005 01:26:01 +0000 (01:26 +0000)]
Removing the #endif as well sounds like a good idea.

19 years agoIn cpufunc_control, uncomment the code responsible for returning the old
cognet [Sun, 20 Feb 2005 01:24:46 +0000 (01:24 +0000)]
In cpufunc_control, uncomment the code responsible for returning the old
state of the control register.

19 years agoAdd the "freq_settings" sysctl to each device that registers with cpufreq
njl [Sun, 20 Feb 2005 00:59:15 +0000 (00:59 +0000)]
Add the "freq_settings" sysctl to each device that registers with cpufreq
so their individual settings can be seen separately for debugging.

19 years agoRemove the note about em(4) not working with bridge(4) when hardware
brueffer [Sun, 20 Feb 2005 00:09:38 +0000 (00:09 +0000)]
Remove the note about em(4) not working with bridge(4) when hardware
assisted VLAN processing is enabled from the BUGS section. This is
expected to work now.

19 years agoFixate the hash bucket size to 4K. We were using the default,
ru [Sat, 19 Feb 2005 23:59:49 +0000 (23:59 +0000)]
Fixate the hash bucket size to 4K.  We were using the default,
which is stat.st_blksize (i.e., PAGE_SIZE).

This change causes the .db files that were cross-compiled on
another platform to be identical to the natively built ones.

Tested on: alpha->amd64 build

19 years agoFix the same problem that was fixed in rev. 1.6 and got reintroduced
ru [Sat, 19 Feb 2005 21:32:21 +0000 (21:32 +0000)]
Fix the same problem that was fixed in rev. 1.6 and got reintroduced
in rev. 1.8 -- make btxld(8) a cross-tool for i386 and now amd64.

Tested on: alpha (which has a different idea of __LDPGSZ)

19 years agoUse a common multi-inclusion protection, and add such a
ru [Sat, 19 Feb 2005 21:16:48 +0000 (21:16 +0000)]
Use a common multi-inclusion protection, and add such a
protection to alpha/include/exec.h.

19 years agoNew release notes: OLDCARD removed, ath(4) transmit rate control
bmah [Sat, 19 Feb 2005 20:10:38 +0000 (20:10 +0000)]
New release notes: OLDCARD removed, ath(4) transmit rate control
modules, wlan(4) overhaul with particular mention of the wlan_wep
module, amr(4) PAE-safe (+MFC), matcd(4) removed, wd(4) removed for
pc98, ifconfig(8) reorg and update.

Culled from random skimming of HEAD commit messages.

Somewhat belatedly bump copyright date while here.

19 years agosendmail 8.13.3 has been merged
gshapiro [Sat, 19 Feb 2005 17:19:01 +0000 (17:19 +0000)]
sendmail 8.13.3 has been merged

19 years agosimply ignore unknown address family.
ume [Sat, 19 Feb 2005 16:54:26 +0000 (16:54 +0000)]
simply ignore unknown address family.

MFC after: 1 week

19 years agoteach IPv6 to `systat -netstat'.
ume [Sat, 19 Feb 2005 15:10:19 +0000 (15:10 +0000)]
teach IPv6 to `systat -netstat'.

Tested by: kuriyama
MFC after: 1 week

19 years agoReimplement recursion protection, checking whether current thread holds
glebius [Sat, 19 Feb 2005 14:41:49 +0000 (14:41 +0000)]
Reimplement recursion protection, checking whether current thread holds
sockbuf mutex.

Reviewed by: rwatson

19 years agoBump __FreeBSD_version for the Intel C/C++ compiler GCC-compatibility
marius [Sat, 19 Feb 2005 13:48:53 +0000 (13:48 +0000)]
Bump __FreeBSD_version for the Intel C/C++ compiler GCC-compatibility
fixes in src/include/stdbool.h rev. 1.7 and src/sys/i386/include/_types.h
rev. 1.10, required by ports/lang/icc.

19 years agoThe Intel C/C++ compiler doesn't require a typedef for _Bool when compiling
marius [Sat, 19 Feb 2005 13:47:33 +0000 (13:47 +0000)]
The Intel C/C++ compiler doesn't require a typedef for _Bool when compiling
C source so don't declare one when using the GCC-compatibility (defaulting
to GCC 2.95.0) of ICC.

Submitted by: netchild
MFC after: 1 week

19 years agoTogether with the changes to compile kernels with the Intel C/C++ compiler
marius [Sat, 19 Feb 2005 13:46:40 +0000 (13:46 +0000)]
Together with the changes to compile kernels with the Intel C/C++ compiler
preliminary support for using the GCC-compatibility of ICC was committed
but couldn't be tested at that time due to problems with ICC itself. Since
ICC 8.1 it's possible to use its GCC-compatibility under FreeBSD and it
turned out that a typedef for __gnuc_va_list is required in that case.
Revert the part of rev. 1.8 which #ifdef'ed out __gnuc_va_list for ICC.

MFC after: 1 week

19 years agoTry to unbreak the vnode locking around vop_reclaim() (based mostly on
phk [Sat, 19 Feb 2005 11:44:57 +0000 (11:44 +0000)]
Try to unbreak the vnode locking around vop_reclaim() (based mostly on
patch from kan@).

Pull bufobj_invalbuf() out of vinvalbuf() and make g_vfs call it on
close.  This is not yet a generally safe function, but for this very
specific use it is safe.  This solves the problem with buffers not
being flushed by unmount or after failed mount attempts.

19 years agoAdd lorder(1) to the list of bootstrap-tools.
ru [Sat, 19 Feb 2005 10:25:42 +0000 (10:25 +0000)]
Add lorder(1) to the list of bootstrap-tools.

19 years agoAllow to cross-build amd64 on non-i386.
ru [Sat, 19 Feb 2005 10:23:34 +0000 (10:23 +0000)]
Allow to cross-build amd64 on non-i386.

19 years agoFix a small bug in firmcvt: outfile must be strdup()ed.
wpaul [Sat, 19 Feb 2005 07:37:01 +0000 (07:37 +0000)]
Fix a small bug in firmcvt: outfile must be strdup()ed.
Also, add conditional code to allow different invokations for objcopy
depending on whether we're compiled on an i386 arch or amd64 arch, so
that we can produce x86-64 object files on amd64.

19 years agoWhen clearing a fragment, it's possible that the length is zero.
delphij [Sat, 19 Feb 2005 07:31:33 +0000 (07:31 +0000)]
When clearing a fragment, it's possible that the length is zero.

Reviewed by: mckusick
MFC After: 1 week

19 years agoAdd a new field to struct cf_setting for special values. These are driver-
njl [Sat, 19 Feb 2005 06:13:26 +0000 (06:13 +0000)]
Add a new field to struct cf_setting for special values.  These are driver-
specific values that other components may want to use.  Add support to
acpi_perf(4) to export the control and status values via this field.

19 years agoDon't restart a timeout wait in kern_sigtimedwait, also allow it
davidxu [Sat, 19 Feb 2005 06:05:49 +0000 (06:05 +0000)]
Don't restart a timeout wait in kern_sigtimedwait, also allow it
to wait longer than a single integer can represent.

19 years agoFix for a potential NFS client race where shared data is updated from
ps [Fri, 18 Feb 2005 23:41:39 +0000 (23:41 +0000)]
Fix for a potential NFS client race where shared data is updated from
base context as well as the socket callback.

Submitted by: Mohan Srinivasan

19 years agoAllocate the M_VLANTAG m_pkthdr flag, and use it to indicate that
ru [Fri, 18 Feb 2005 22:31:19 +0000 (22:31 +0000)]
Allocate the M_VLANTAG m_pkthdr flag, and use it to indicate that
a packet has VLAN mbuf tag attached.  This is faster to check than
m_tag_locate(), and allows us to use the tags in non-vlan(4) VLAN
producers.

The first argument to VLAN_OUTPUT_TAG() is now unused but retained
for backward compatibility.

While here, embellish a fix in rev. 1.174 of if_ethersubr.c -- it
now checks for packets with VLAN (mbuf) tags, and it should now
be possible to bridge(4) on vlan(4)'s whose parent interfaces
support VLAN decapsulation in hardware.

Reviewed by: sam

19 years agoSwap the arguments for CP so we copy the correct source and
ps [Fri, 18 Feb 2005 22:14:40 +0000 (22:14 +0000)]
Swap the arguments for CP so we copy the correct source and
destination.

19 years agoSupport correct programming environment name for amd64.
wollman [Fri, 18 Feb 2005 21:53:05 +0000 (21:53 +0000)]
Support correct programming environment name for amd64.

MFC after: 3 days

19 years agoRemove now unused 'int s' from spl().
rwatson [Fri, 18 Feb 2005 21:39:55 +0000 (21:39 +0000)]
Remove now unused 'int s' from spl().

MFC after: 3 days

19 years agoDe-spl kern_connect().
rwatson [Fri, 18 Feb 2005 19:37:36 +0000 (19:37 +0000)]
De-spl kern_connect().

MFC after: 3 days

19 years agoDocument the --null, --one-file-system, and -I options.
kientzle [Fri, 18 Feb 2005 19:19:45 +0000 (19:19 +0000)]
Document the --null, --one-file-system, and -I options.
Clarify that -T can be used in 'x' mode as well as 'c' mode.

19 years agoCorrect a typo in the comment describing soreceive_rcvoob().
rwatson [Fri, 18 Feb 2005 19:15:22 +0000 (19:15 +0000)]
Correct a typo in the comment describing soreceive_rcvoob().

MFC after: 3 days

19 years agoIn soconnect(), when resetting so->so_error, the socket lock is not
rwatson [Fri, 18 Feb 2005 19:13:51 +0000 (19:13 +0000)]
In soconnect(), when resetting so->so_error, the socket lock is not
required due to a straight integer write in which minor races are not
a problem.

19 years agoRe-style do_setopt_accept_filter() to match uipc_accf.c style, and fix
rwatson [Fri, 18 Feb 2005 19:01:22 +0000 (19:01 +0000)]
Re-style do_setopt_accept_filter() to match uipc_accf.c style, and fix
one other style nit in the file.

MFC after: 3 days

19 years ago- Add a custom version of exec_copyin_args() to deal with the 32-bit
jhb [Fri, 18 Feb 2005 18:56:04 +0000 (18:56 +0000)]
- Add a custom version of exec_copyin_args() to deal with the 32-bit
  pointers in argv and envv in userland and use that together with
  kern_execve() and exec_free_args() to implement freebsd32_execve()
  without using the stackgap.
- Fix freebsd32_adjtime() to call adjtime() rather than utimes().  Still
  uses stackgap for now.
- Use kern_setitimer(), kern_getitimer(), kern_select(), kern_utimes(),
  kern_statfs(), kern_fstatfs(), kern_fhstatfs(), kern_stat(),
  kern_fstat(), and kern_lstat().

Tested by: cokane (amd64)
Silence on: amd64, ia64

19 years agoMove do_setopt_accept_filter() from uipc_socket.c to uipc_accf.c, where
rwatson [Fri, 18 Feb 2005 18:54:42 +0000 (18:54 +0000)]
Move do_setopt_accept_filter() from uipc_socket.c to uipc_accf.c, where
the rest of the accept filter code currently lives.

MFC after: 3 days

19 years ago- Add a custom version of exec_copyin_args() to deal with the 32-bit
jhb [Fri, 18 Feb 2005 18:51:59 +0000 (18:51 +0000)]
- Add a custom version of exec_copyin_args() to deal with the 32-bit
  pointers in argv and envv in userland and use that together with
  kern_execve() and exec_free_args() to implement linux_execve() for the
  amd64/linux32 ABI without using the stackgap.
- Implement linux_nanosleep() using the recently added kern_nanosleep().
- Use linux_emul_convpath() instead of linux_emul_find() in
  exec_linux_imgact_try().

Tested by: cokane
Silence on: amd64

19 years agoMinor style tweaks: line wrap comments and lines more consistently.
rwatson [Fri, 18 Feb 2005 18:49:44 +0000 (18:49 +0000)]
Minor style tweaks: line wrap comments and lines more consistently.

MFC after: 3 days

19 years agoRe-order checks in socheckuid() so that we check all deny cases before
rwatson [Fri, 18 Feb 2005 18:43:33 +0000 (18:43 +0000)]
Re-order checks in socheckuid() so that we check all deny cases before
returning accept.

MFC after: 3 days

19 years ago- Implement osf1_emul_find() using kern_alternate_path(). This changes
jhb [Fri, 18 Feb 2005 18:37:26 +0000 (18:37 +0000)]
- Implement osf1_emul_find() using kern_alternate_path().  This changes
  the semantics in that the returned filename to use is now a kernel
  pointer rather than a user space pointer.  This required changing the
  arguments to the CHECKALT*() macros some and changing the various system
  calls that used pathnames to use the kern_foo() functions that can accept
  kernel space filename pointers instead of calling the system call
  directly.
- Use kern_open(), kern_stat(), kern_lstat(), kern_fstat(), kern_access(),
  kern_truncate(), kern_pathconf(), kern_execve(), kern_select(),
  kern_setitimer(), kern_getitimer(), kern_statfs(), and kern_fstatfs().

Silence on: alpha@

19 years agoUse LCONVPATHEXIST() rather than CHECKALTEXIST() and use
jhb [Fri, 18 Feb 2005 18:32:32 +0000 (18:32 +0000)]
Use LCONVPATHEXIST() rather than CHECKALTEXIST() and use
exec_copyin_args(), kern_execve(), and exec_free_args() rather than
execve() to eliminate stackgap use from Alpha's linux_execve().

Silence on: alpha@

19 years agoFor the I/O port case, we need to set ok to 1 if we have what looks
imp [Fri, 18 Feb 2005 17:35:03 +0000 (17:35 +0000)]
For the I/O port case, we need to set ok to 1 if we have what looks
like a valid range.  We already do this in the memory case (although
the code there is somewhat different than the I/o case because we have
to deal with different kinds of memory).  Since most laptops don't
have non-subtractive bridges, this wasn't seen in practice.

Evidentally the Compaq R3000 hits this problem with PC Cards.

Some minor style fixes while I'm here.

Submitted by: Jung-uk Kim

19 years agoSupport high vectors for arm9.
cognet [Fri, 18 Feb 2005 17:29:12 +0000 (17:29 +0000)]
Support high vectors for arm9.

Obtained from: NetBSD

19 years agoSomewhere along the line, tick accumulation for SA threads was
deischen [Fri, 18 Feb 2005 16:07:05 +0000 (16:07 +0000)]
Somewhere along the line, tick accumulation for SA threads was
changed to use the statclock.  Make sure we calculate the value
of a tick correctly in userland.

Noticed by: Kazuaki Oda <kaakun at highway dot ne dot jp>

19 years agoSilence witness warnings about duplicate pmap lock emitted since
marius [Fri, 18 Feb 2005 15:37:34 +0000 (15:37 +0000)]
Silence witness warnings about duplicate pmap lock emitted since
rev. 1.145 of sys/sparc64/sparc64/pmap.c.

Submitted by: alc

19 years agoPush the assignments of some variables down into the sub-blocks where
harti [Fri, 18 Feb 2005 15:32:06 +0000 (15:32 +0000)]
Push the assignments of some variables down into the sub-blocks where
it is actually needed. This makes clear in which subblocks the variables
are not needed and which can easier be split out.

Submitted by: Max Okumoto <okumoto@ucsd.edu>

19 years agoRemove basically unused root_vp pointer in udfmount.
rwatson [Fri, 18 Feb 2005 11:47:51 +0000 (11:47 +0000)]
Remove basically unused root_vp pointer in udfmount.

MFC after: 1 week
Discussed with: scottl

19 years agoMake sure to drop the VI_LOCK in vgonel();
phk [Fri, 18 Feb 2005 11:13:56 +0000 (11:13 +0000)]
Make sure to drop the VI_LOCK in vgonel();

Spotted by: Taku YAMAMOTO <taku@tackymt.homeip.net>

19 years agoRun the netatalk netisrs without Giant.
rwatson [Fri, 18 Feb 2005 10:53:00 +0000 (10:53 +0000)]
Run the netatalk netisrs without Giant.

MFC after: 1 week

19 years agoConditionalize cd9660 chattiness regarding the nature of the file system
rwatson [Fri, 18 Feb 2005 10:49:55 +0000 (10:49 +0000)]
Conditionalize cd9660 chattiness regarding the nature of the file system
mounted (is it Joliet, RockRidge, High Sierra) based on bootverbose.
Most file systems don't generate log messages based on details of the
file system superblock, and these log messages disrupt sysinstall output
during a new install from CD.  We may want to explore exposing this
status information using nmount() at some point.

MFC after: 3 days

19 years agoFor non-embedded platforms, increase the size of the argument list.
obrien [Fri, 18 Feb 2005 08:01:59 +0000 (08:01 +0000)]
For non-embedded platforms, increase the size of the argument list.
Note that this results in more kernel virtual memory being reserved for
temporary storage of the args.  The args temporary space is allocated out
of exec_map (a submap of kernel_map).  This will use roughly 4MB of KVM.

OK'ed by: dg

19 years agoPush down variables into local scope in Var_Parse to make their
harti [Fri, 18 Feb 2005 07:33:43 +0000 (07:33 +0000)]
Push down variables into local scope in Var_Parse to make their
liveness clear for splitting up this monster function.

Submitted by: Max Okumoto <okumoto@ucsd.edu>

19 years agoFix a couple of u_int_foos that should have been uint_foos.
wpaul [Fri, 18 Feb 2005 04:33:34 +0000 (04:33 +0000)]
Fix a couple of u_int_foos that should have been uint_foos.

19 years agoMake the Win64 -> ELF64 template a little smaller by using a string
wpaul [Fri, 18 Feb 2005 03:22:37 +0000 (03:22 +0000)]
Make the Win64 -> ELF64 template a little smaller by using a string
copy op to shift arguments on the stack instead of transfering each
argument one by one through a register. Probably doesn't affect overall
operation, but makes the code a little less grotty and easier to update
later if I choose to make the wrapper handle more args. Also add
comments.

19 years agoIn solisten(), unconditionally set the SO_ACCEPTCONN option in
rwatson [Fri, 18 Feb 2005 00:52:17 +0000 (00:52 +0000)]
In solisten(), unconditionally set the SO_ACCEPTCONN option in
so->so_options when solisten() will succeed, rather than setting it
conditionally based on there not being queued sockets in the completed
socket queue.  Otherwise, if the protocol exposes new sockets via the
completed queue before solisten() completes, the listen() system call
will succeed, but the socket and protocol state will be out of sync.
For TCP, this didn't happen in practice, as the TCP code will panic if
a new connection comes in after the tcpcb has been transitioned to a
listening state but the socket doesn't have SO_ACCEPTCONN set.

This is historical behavior resulting from bitrot since 4.3BSD, in which
that line of code was associated with the conditional NULL'ing of the
connection queue pointers (one-time initialization to be performed
during the transition to a listening socket), which are now initialized
separately.

Discussed with: fenner, gnn
MFC after: 3 days

19 years agoDocument the cpufreq_drv_type() method and info-only flag.
njl [Fri, 18 Feb 2005 00:28:52 +0000 (00:28 +0000)]
Document the cpufreq_drv_type() method and info-only flag.

19 years agoIntroduce a new method, cpufreq_drv_type(), that returns the type of the
njl [Fri, 18 Feb 2005 00:23:36 +0000 (00:23 +0000)]
Introduce a new method, cpufreq_drv_type(), that returns the type of the
driver.  This used to be handled by cpufreq_drv_settings() but it's
useful to get the type/flags separately from getting the settings.
(For example, you don't have to pass an array of cf_setting just to find
the driver type.)

Use this new method in our in-tree drivers to detect reliably if acpi_perf
is present and owns the hardware.  This simplifies logic in drivers as well
as fixing a bug introduced in my last commit where too many drivers attached.

19 years agoRemove 2 (SACK) fields from the tcpcb. These are only used by a
ps [Thu, 17 Feb 2005 23:04:56 +0000 (23:04 +0000)]
Remove 2 (SACK) fields from the tcpcb. These are only used by a
function that is called from tcp_input(), so they oughta be passed on
the stack instead of stuck in the tcpcb.

Submitted by: Mohan Srinivasan

19 years agoFix a small style problem and add a __packed pragma for safety.
scottl [Thu, 17 Feb 2005 21:09:26 +0000 (21:09 +0000)]
Fix a small style problem and add a __packed pragma for safety.

19 years agomemspace is set to some value by masking off bits. When these bits
imp [Thu, 17 Feb 2005 21:05:04 +0000 (21:05 +0000)]
memspace is set to some value by masking off bits.  When these bits
are equal to PCCARD_TPCE_FS_MEMSPACE_NONE, memspace will be zero, so
testing for this case inside of the if statement results in dead code.
We'd fail to set a value to zero that's already zero (since it is
initialized to 0 indirectly) with this code being there.  Well, except
in the very rare case that we have a card that has a defualt entry
that includes a memory space followed by one that has no memory space
(these are extremely rare, I don't recall ever having seen one :-).

Fix this by setting num_memspace to 0 in a more appropriate place.

Submitted by: Coverity Prevent analysis tool

19 years agoUndoing recent changes to make 3ware's i386 tools work on amd64, since there are
vkashyap [Thu, 17 Feb 2005 19:05:42 +0000 (19:05 +0000)]
Undoing recent changes to make 3ware's i386 tools work on amd64, since there are
now amd64 versions of CLI and 3DM2 available.

19 years agoCheck for the address space type first before validating it. In particular,
njl [Thu, 17 Feb 2005 19:00:14 +0000 (19:00 +0000)]
Check for the address space type first before validating it.  In particular,
we want to return EOPNOTSUPP for FFixedHW no matter what the address.

Submitted by: Bruno Ducrot

19 years agoFix the support for the ax88[17]90 to print the right type, rather
imp [Thu, 17 Feb 2005 18:15:13 +0000 (18:15 +0000)]
Fix the support for the ax88[17]90 to print the right type, rather
than the generic ne-2000 string.  This should have no effect on the
actual support of the parts, just reporting what the part was.

Also, rename a few functins and symbols to reflect a more generic
part support that grew out of the early specific support.

19 years agoMake the on-disk format of .dat files produced by strfile(8)
ru [Thu, 17 Feb 2005 18:06:37 +0000 (18:06 +0000)]
Make the on-disk format of .dat files produced by strfile(8)
architecture independent.  Besides the fixed-width types in
the header, the offsets are now stored as 64-bit off_t (also
in big endian format).

Tested on: i386, amd64, sparc64, ia64

19 years agoAdd myself to the calendar
hq [Thu, 17 Feb 2005 16:38:56 +0000 (16:38 +0000)]
Add myself to the calendar

19 years agoIn case of drive errors, don't close the associated consumer and
le [Thu, 17 Feb 2005 16:08:36 +0000 (16:08 +0000)]
In case of drive errors, don't close the associated consumer and
detach it, but instead let the geom wither away.

Bump copyright year.

19 years agoRemove dublicated device entry from the synopsis.
brueffer [Thu, 17 Feb 2005 16:01:20 +0000 (16:01 +0000)]
Remove dublicated device entry from the synopsis.

Submitted by: Ulrich Spoerlein <q@uni.de>
MFC after: 3 days

19 years agoMark netatm and netnatm explicitly as requiring Giant, as they still do.
rwatson [Thu, 17 Feb 2005 14:21:22 +0000 (14:21 +0000)]
Mark netatm and netnatm explicitly as requiring Giant, as they still do.

MFC after: 3 days

19 years agoIn accept1(), extend coverage of the socket lock from just covering
rwatson [Thu, 17 Feb 2005 13:00:23 +0000 (13:00 +0000)]
In accept1(), extend coverage of the socket lock from just covering
soref() to also covering the update of so_state.  While no other user
threads can update the socket state here as it's not yet hooked up to
the file descriptor array yet, the protocol could also frob the
socket state here, leading to a lost update to the so_state field.
No reported instances of this bug (as yet).

MFC after:      3 days

19 years agoIn sonewconn(), set the new socket's state to show the protocol-provided
rwatson [Thu, 17 Feb 2005 12:53:45 +0000 (12:53 +0000)]
In sonewconn(), set the new socket's state to show the protocol-provided
connection status before inserting the new socket into the listen
socket's accept queue, or there might be a race in which another thread
wakes up when the accept lock is released, and sees the socket before its
state is set correctly.  The wakeup still occurs after the accept lock is
released.  There have been no diagnoses of this bug in real-world systems
(as yet).

MFC after: 3 days

19 years agoWhitespace cleanup: substitute mixed tabs and spaces by canonical
harti [Thu, 17 Feb 2005 12:35:32 +0000 (12:35 +0000)]
Whitespace cleanup: substitute mixed tabs and spaces by canonical
whitespace and line up some variable definitions.

19 years agoMove error case to begin of if-else chain. Do not needless initialize
harti [Thu, 17 Feb 2005 12:31:53 +0000 (12:31 +0000)]
Move error case to begin of if-else chain. Do not needless initialize
startc, but only at the place where the initialisation is needed. Remove
a needless else.

Submitted by: Max Okumoto <okumoto@ucsd.edu> (partly)

19 years agoDocument the new default init_path.
des [Thu, 17 Feb 2005 11:14:45 +0000 (11:14 +0000)]
Document the new default init_path.

Reminded by: ru
MFC after: 2 weeks

19 years agoIntroduce vx_wait{l}() and use it instead of home-rolled versions.
phk [Thu, 17 Feb 2005 10:49:51 +0000 (10:49 +0000)]
Introduce vx_wait{l}() and use it instead of home-rolled versions.

19 years agoConvert KASSERTS to VNASSERTS
phk [Thu, 17 Feb 2005 10:28:58 +0000 (10:28 +0000)]
Convert KASSERTS to VNASSERTS

19 years agoAdd /rescue/init to the default init_path, before /stand/sysinstall.
des [Thu, 17 Feb 2005 10:00:10 +0000 (10:00 +0000)]
Add /rescue/init to the default init_path, before /stand/sysinstall.

MFC after: 2 weeks

19 years agoFix two typos in comments.
harti [Thu, 17 Feb 2005 09:09:34 +0000 (09:09 +0000)]
Fix two typos in comments.

Submitted by: ru & Max Okumoto <okumoto@ucsd.edu>

19 years agoSort out the error case that a single '$' was passed a little bit
harti [Thu, 17 Feb 2005 09:01:19 +0000 (09:01 +0000)]
Sort out the error case that a single '$' was passed a little bit
earlier instead of mixing its handling with other cases.

Submitted by: Max Okumoto <okumoto@ucsd.edu>

19 years agoFix inteface clear time. pf printed "Thu Jan 1 09:00:01 1970"
yongari [Thu, 17 Feb 2005 03:36:31 +0000 (03:36 +0000)]
Fix inteface clear time. pf printed "Thu Jan  1 09:00:01 1970"
in "pfctl -vvsI" output when pf was statically linked to kernel.

Discussed with: mlaier

19 years agoThe correct error value for not having enough storage is E2BIG, not
njl [Thu, 17 Feb 2005 01:02:58 +0000 (01:02 +0000)]
The correct error value for not having enough storage is E2BIG, not
ENOMEM.  The manpage and ichss(4) are correct.

19 years agoFix the check for acpi_perf(4) so that we verify if it is fully attached
njl [Thu, 17 Feb 2005 01:01:40 +0000 (01:01 +0000)]
Fix the check for acpi_perf(4) so that we verify if it is fully attached
or just offering info.  In the former case, we don't probe/attach to allow
the ACPI driver precedence.  A refinement of this would be to actually
use the info provided by acpi_perf(4) to get the real CPU clock rates
instead of estimating them but since all systems that support both
acpi_perf(4) and ichss(4) export the control registers to acpi_perf(4),
it can just handle the registers on its own.