]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
19 years agoDo install BSM include files (such as they are) when installing system
Robert Watson [Sun, 29 May 2005 16:17:00 +0000 (16:17 +0000)]
Do install BSM include files (such as they are) when installing system
includes.

Submitted by: wsalamon
Obtained from: TrustedBSD Project

19 years agoFor consistency with more system include files, add a trailing '_' to
Robert Watson [Sun, 29 May 2005 16:11:34 +0000 (16:11 +0000)]
For consistency with more system include files, add a trailing '_' to
the define guards in audit_kevents.h.

19 years agoAdd place-holder audit.h that defines only au_event_t, which is needed
Robert Watson [Sun, 29 May 2005 16:10:33 +0000 (16:10 +0000)]
Add place-holder audit.h that defines only au_event_t, which is needed
in order to modify the system call table to include event identifiers.
The full audit.h will be merged at a later date.

Obtained from: TrustedBSD Project

19 years agoGive variable an initial value. Use errx() instead of fprintf().
Philippe Charnier [Sun, 29 May 2005 16:07:10 +0000 (16:07 +0000)]
Give variable an initial value. Use errx() instead of fprintf().

19 years agoReduce compiler warning: variable might be used uninitialized, by giving
Philippe Charnier [Sun, 29 May 2005 16:04:46 +0000 (16:04 +0000)]
Reduce compiler warning: variable might be used uninitialized, by giving
an initial value.

19 years agoEven if variable is never used uninitialized by the semantic, reduce compiler
Philippe Charnier [Sun, 29 May 2005 16:01:12 +0000 (16:01 +0000)]
Even if variable is never used uninitialized by the semantic, reduce compiler
warning by giving an initial value in all cases.

19 years agorscid -> __FBSDID. Mark parameter as __unused when necessary.
Philippe Charnier [Sun, 29 May 2005 15:57:00 +0000 (15:57 +0000)]
rscid -> __FBSDID. Mark parameter as __unused when necessary.

19 years agoMove variable initialization to reduce compiler warning.
Philippe Charnier [Sun, 29 May 2005 15:52:48 +0000 (15:52 +0000)]
Move variable initialization to reduce compiler warning.

19 years agoReduce compiler warning: variable might be used uninitialized, by giving
Philippe Charnier [Sun, 29 May 2005 15:49:53 +0000 (15:49 +0000)]
Reduce compiler warning: variable might be used uninitialized, by giving
an initial value.

19 years agoMove FreeBSD Id outside of copyright. Initialize variable.
Philippe Charnier [Sun, 29 May 2005 15:47:31 +0000 (15:47 +0000)]
Move FreeBSD Id outside of copyright. Initialize variable.

19 years agoTypo correction.
Tai-hwa Liang [Sun, 29 May 2005 14:56:51 +0000 (14:56 +0000)]
Typo correction.

19 years agoModify vmstat(8)'s domem() routine, which is responsible for extracting
Robert Watson [Sun, 29 May 2005 13:40:00 +0000 (13:40 +0000)]
Modify vmstat(8)'s domem() routine, which is responsible for extracting
malloc(9) statistics from kernel memory or a kernel coredump, to catch
up with recent changes to adopt per-CPU malloc(9) statistics.  The new
routines walk the per-CPU statistics pools and coalesce them for
presentation to the user.

19 years agoKernel malloc layers malloc_type allocation over one of two underlying
Robert Watson [Sun, 29 May 2005 13:38:07 +0000 (13:38 +0000)]
Kernel malloc layers malloc_type allocation over one of two underlying
allocators: a set of power-of-two UMA zones for small allocations, and the
VM page allocator for large allocations.  In order to maintain unified
statistics for specific malloc types, kernel malloc maintains a separate
per-type statistics pool, which can be monitored using vmstat -m.  Prior
to this commit, each pool of per-type statistics was protected using a
per-type mutex associated with the malloc type.

This change modifies kernel malloc to maintain per-CPU statistics pools
for each malloc type, and protects writing those statistics using critical
sections.  It also moves to unsynchronized reads of per-CPU statistics
when generating coalesced statistics.  To do this, several changes are
implemented:

- In the previous world order, the statistics memory was allocated by
  the owner of the malloc type structure, allocated statically using
  MALLOC_DEFINE().  This embedded the definition of the malloc_type
  structure into all kernel modules.  Move to a model in which a pointer
  within struct malloc_type points at a UMA-allocated
  malloc_type_internal data structure owned and maintained by
  kern_malloc.c, and not part of the exported ABI/API to the rest of
  the kernel.  For the purposes of easing a possible MFC, re-use an
  existing pointer in 'struct malloc_type', and maintain the current
  malloc_type structure size, as well as layout with respect to the
  fields reused outside of the malloc subsystem (such as ks_shortdesc).
  There are several unused fields as a result of no longer requiring
  the mutex in malloc_type.

- Struct malloc_type_internal contains an array of malloc_type_stats,
  of size MAXCPU.  The structure defined above avoids hard-coding a
  kernel compile-time value of MAXCPU into kernel modules that interact
  with malloc.

- When accessing per-cpu statistics for a malloc type, surround read -
  modify - update requests with critical_enter()/critical_exit() in
  order to avoid races during write.  The per-CPU fields are written
  only from the CPU that owns them.

- Per-CPU stats now maintained "allocated" and "freed" counters for
  number of allocations/frees and bytes allocated/freed, since there is
  no longer a coherent global notion of the totals.  When coalescing
  malloc stats, accept a slight race between reading stats across CPUs,
  and avoid showing the user a negative allocation count for the type
  in the event of a race.  The global high watermark is no longer
  maintained for a malloc type, as there is no global notion of the
  number of allocations.

- While tearing up the sysctl() path, also switch to using sbufs.  The
  current "export as text" sysctl format is retained with the same
  syntax.  We may want to change this in the future to export more
  per-CPU information, such as how allocations and frees are balanced
  across CPUs.

This change results in a substantial speedup of kernel malloc and free
paths on SMP, as critical sections (where usable) out-perform mutexes
due to avoiding atomic/bus-locked operations.  There is also a minor
improvement on UP due to the slightly lower cost of critical sections
there.  The cost of the change to this approach is the loss of a
continuous notion of total allocations that can be exploited to track
per-type high watermarks, as well as increased complexity when
monitoring statistics.

Due to carefully avoiding changing the ABI, as well as hardening the ABI
against future changes, it is not necessary to recompile kernel modules
for this change.  However, MFC'ing this change to RELENG_5 will require
also MFC'ing optimizations for soft critical sections, which may modify
exposed kernel ABIs.  The internal malloc API is changed, and
modifications to vmstat in order to restore "vmstat -m" on core dumps will
follow shortly.

Several improvements from: bde
Statistics approach discussed with: ups
Tested by: scottl, others

19 years agoAdd vr_init_t member to sc_rndr_sw_t instances in order to unbreak
Marius Strobl [Sun, 29 May 2005 12:47:39 +0000 (12:47 +0000)]
Add vr_init_t member to sc_rndr_sw_t instances in order to unbreak
compilation after sys/dev/syscons/syscons.h rev. 1.83.

19 years agoFix check for leading zero, so that it does not block two zeroes
Gleb Smirnoff [Sun, 29 May 2005 12:20:41 +0000 (12:20 +0000)]
Fix check for leading zero, so that it does not block two zeroes
in hook name.

19 years agoSync with syscons update (Add new member to struct sc_rndr_sw).
Yoshihiro Takahashi [Sun, 29 May 2005 11:53:14 +0000 (11:53 +0000)]
Sync with syscons update (Add new member to struct sc_rndr_sw).

19 years agoThe end values passed to rman_manage_region() for PCI i/o and mem
Peter Grehan [Sun, 29 May 2005 08:51:21 +0000 (08:51 +0000)]
The end values passed to rman_manage_region() for PCI i/o and mem
spaces were 1 too large. This resulted in the rman list not being
sorted correctly, and USB ports not being discovered on older
TiBooks.

Detective work by:   Andreas Tobler <toa at pop dot agri dot ch>

19 years agoAdd VESA mode support for syscons, which enables the support of 15, 16,
Xin LI [Sun, 29 May 2005 08:43:44 +0000 (08:43 +0000)]
Add VESA mode support for syscons, which enables the support of 15, 16,
24, and 32 bit modes.  To use that, syscons(4) must be built with
the compile time option 'options SC_PIXEL_MODE', and VESA support (a.k.a.
vesa.ko) must be either loaded, or be compiled into the kernel.

Do not return EINVAL when the mouse state is changed to what it already is,
which seems to cause problems when you have two mice attached, and
applications are not likely obtain useful information through the EINVAL
caused by showing the mouse pointer twice.

Teach vidcontrol(8) about mode names like MODE_<NUMBER>, where <NUMBER> is
the video mode number from the vidcontrol -i mode output.  Also, revert the
video mode if something fails.

Obtained from: DragonFlyBSD
Discussed at: current@ with patch attached [1]
PR: kern/71142 [2]
Submitted by: Xuefeng DENG <dsnofe at msn com> [1],
Cyrille Lefevre <cyrille dot lefevre at laposte dot net> [2]

19 years agoMove information about exit status into a DIAGNOSTICS section.
Tim J. Robbins [Sun, 29 May 2005 08:18:48 +0000 (08:18 +0000)]
Move information about exit status into a DIAGNOSTICS section.

19 years agoRemove bus_{mem,p}io.h and related code for a micro-optimization on i386
Yoshihiro Takahashi [Sun, 29 May 2005 04:42:30 +0000 (04:42 +0000)]
Remove bus_{mem,p}io.h and related code for a micro-optimization on i386
and amd64.  The optimization is a trivial on recent machines.

Reviewed by: -arch (imp, marcel, dfr)

19 years agoFix panic when module is compiled in and it is loaded from loader.conf.
Pawel Jakub Dawidek [Sat, 28 May 2005 23:20:05 +0000 (23:20 +0000)]
Fix panic when module is compiled in and it is loaded from loader.conf.
Only panic is fixed, module will be still listed in kldstat(8) output.
Not sure what is correct fix, because adding unloading code in case of
failure to linker_init_kernel_modules() doesn't work.

19 years agoProvide info on the incompatible change in v1.33 of sys/kern/imgact_shell.c
Garance A Drosehn [Sat, 28 May 2005 22:45:31 +0000 (22:45 +0000)]
Provide info on the incompatible change in v1.33 of sys/kern/imgact_shell.c

Discussed with: imp

19 years agoChange the way options are parsed on the `#!'-line of a shell-script. Instead
Garance A Drosehn [Sat, 28 May 2005 22:42:41 +0000 (22:42 +0000)]
Change the way options are parsed on the `#!'-line of a shell-script.  Instead
of having the kernel parse that line and add an entry to the argument list for
each 'separate word' it finds, have it add only one entry which holds all
the words found on that line.  The old behavior is useful in some situations,
but it does not match the way any other operating system will parse that line.

This has been discussed in the thread "Bug in #! processing - One More Time"
on the freebsd-arch mailing list (starting back on Feb 24, 2005).  The first
few messages in that thread provide the background in much detail.

PR: 16393
Reviewed by: freebsd-arch

19 years agoPrevent loading modules with are compiled into the kernel.
Pawel Jakub Dawidek [Sat, 28 May 2005 22:29:44 +0000 (22:29 +0000)]
Prevent loading modules with are compiled into the kernel.

PR: kern/48759
Submitted by: Pawe³ Ma³achowski <pawmal@unia.3lo.lublin.pl>
Patch from: demon
MFC after: 2 weeks

19 years agointegrate changes from libpcap-0.9.1-096
Sam Leffler [Sat, 28 May 2005 21:56:41 +0000 (21:56 +0000)]
integrate changes from libpcap-0.9.1-096

Reviewed by: bms

19 years agoUpdate some comments to reflect the change from spl-based to lock-based
Alan Cox [Sat, 28 May 2005 17:56:18 +0000 (17:56 +0000)]
Update some comments to reflect the change from spl-based to lock-based
synchronization.

19 years agopmap_enter() no longer requires Giant. Therefore, stop acquiring and
Alan Cox [Sat, 28 May 2005 17:13:36 +0000 (17:13 +0000)]
pmap_enter() no longer requires Giant.  Therefore, stop acquiring and
releasing it in pmap_enter_quick().

MFC after: 3 weeks

19 years agoDocument 'jid' keyword for ps(1) and '-j' option for pgrep(1)/pkill(1).
Pawel Jakub Dawidek [Sat, 28 May 2005 16:23:29 +0000 (16:23 +0000)]
Document 'jid' keyword for ps(1) and '-j' option for pgrep(1)/pkill(1).

19 years agoRegenerate from syscalls.master.
Robert Watson [Sat, 28 May 2005 14:35:43 +0000 (14:35 +0000)]
Regenerate from syscalls.master.

19 years agoMark ntp_gettime() as MSTD, since its system call path will acquire
Robert Watson [Sat, 28 May 2005 14:35:05 +0000 (14:35 +0000)]
Mark ntp_gettime() as MSTD, since its system call path will acquire
Giant if required.

19 years agoExplicitly acquire Giant around the ntp_gettime() and assert it in the
Robert Watson [Sat, 28 May 2005 14:34:41 +0000 (14:34 +0000)]
Explicitly acquire Giant around the ntp_gettime() and assert it in the
sysctl path.  While this code is close to MPSAFE, it may require some
additional locking.  Mark ntp_gettime1() as GIANT_REQUIRED for now.

Suggested by: phk

19 years agoChange the spkr_set_pitch() function to a macro to fix low level profiling.
Yoshihiro Takahashi [Sat, 28 May 2005 13:40:27 +0000 (13:40 +0000)]
Change the spkr_set_pitch() function to a macro to fix low level profiling.

19 years agoRegenerate for updated syscalls.master.
Robert Watson [Sat, 28 May 2005 13:24:05 +0000 (13:24 +0000)]
Regenerate for updated syscalls.master.

19 years agoMark the following compatability system calls as MCOMPAT or MCOMPAT4 based
Robert Watson [Sat, 28 May 2005 13:23:42 +0000 (13:23 +0000)]
Mark the following compatability system calls as MCOMPAT or MCOMPAT4 based
on the their simply wrapping MPSAFE implementations of existing MPSAFE
system calls:

  getfsstat()
  lseek()
  stat()
  lstat()
  truncate()
  ftruncate()
  statfs()
  fstatfs()

Note that ogetdirentries() is not marked MPSAFE because it does not share
the MPSAFE implementation used for getdirentries(), and requires separate
locking to be implemented.

19 years agoFix use of uninitialized variable len in ngd_send.
Bjoern A. Zeeb [Sat, 28 May 2005 13:15:44 +0000 (13:15 +0000)]
Fix use of uninitialized variable len in ngd_send.

Note: len gets intialized to 0 for sap == NULL case only to
make compiler on amd64 happy. This has nothing todo with the
former uninitialized use of len in sap != NULL case.

Reviewed by: glebius
Approved by: pjd (mentor)

19 years agoRegenerate from syscalls.master.
Robert Watson [Sat, 28 May 2005 13:13:01 +0000 (13:13 +0000)]
Regenerate from syscalls.master.

19 years agoMark quotactl() as MSTD.
Robert Watson [Sat, 28 May 2005 13:12:04 +0000 (13:12 +0000)]
Mark quotactl() as MSTD.

19 years agoAcquire Giant explicitly in quotactl() so that the syscalls.master
Robert Watson [Sat, 28 May 2005 13:11:35 +0000 (13:11 +0000)]
Acquire Giant explicitly in quotactl() so that the syscalls.master
entry can become MSTD.

19 years agoRegenerate from updated syscalls.master.
Robert Watson [Sat, 28 May 2005 13:09:56 +0000 (13:09 +0000)]
Regenerate from updated syscalls.master.

19 years agoMark kenv(2) as MPSAFE, since it appears to be properly locked down.
Robert Watson [Sat, 28 May 2005 13:09:41 +0000 (13:09 +0000)]
Mark kenv(2) as MPSAFE, since it appears to be properly locked down.

19 years agoRegenerate system call tables from syscalls.master.
Robert Watson [Sat, 28 May 2005 13:08:26 +0000 (13:08 +0000)]
Regenerate system call tables from syscalls.master.

19 years agoAlso mark the COMPAT4 version of fhstatfs() as MPSAFE.
Robert Watson [Sat, 28 May 2005 13:07:43 +0000 (13:07 +0000)]
Also mark the COMPAT4 version of fhstatfs() as MPSAFE.

19 years agoMark fhopen(), fhstat(), and fhstatfs() as MSTD, since they now
Robert Watson [Sat, 28 May 2005 12:59:33 +0000 (12:59 +0000)]
Mark fhopen(), fhstat(), and fhstatfs() as MSTD, since they now
acquire Giant themselves.

19 years agoAcquire Giant explicitly in fhopen(), fhstat(), and kern_fhstatfs(),
Robert Watson [Sat, 28 May 2005 12:58:54 +0000 (12:58 +0000)]
Acquire Giant explicitly in fhopen(), fhstat(), and kern_fhstatfs(),
so that we can start to eliminate the presence of non-MPSAFE system
call entries in syscalls.master.

19 years agoEnable ReiserFS note on pc98.
Yoshihiro Takahashi [Sat, 28 May 2005 12:41:55 +0000 (12:41 +0000)]
Enable ReiserFS note on pc98.

19 years agoAdd 6300ESB, which should be treated as ICH4.
Seigo Tanimura [Sat, 28 May 2005 09:32:43 +0000 (09:32 +0000)]
Add 6300ESB, which should be treated as ICH4.

PR: kern/81573
Submitted by: OOTOMO Hiroyuki <ootomo@za.wakwak.com>
MFC after: 1 week

19 years agoLet OSPFv3 go through ipfw. Some more additional checks would be
Seigo Tanimura [Sat, 28 May 2005 07:46:44 +0000 (07:46 +0000)]
Let OSPFv3 go through ipfw.  Some more additional checks would be
desirable, though.

19 years agoUpdate refrenced URL for SNMP list of ifTypes to refer to iana.org
Brooks Davis [Sat, 28 May 2005 06:11:38 +0000 (06:11 +0000)]
Update refrenced URL for SNMP list of ifTypes to refer to iana.org
instead of a dead location on ftp.isi.edu.

19 years agoMove AVM USB Bluetooth-Adapter BlueFritz! from "broken" devices list
Maksim Yevmenkin [Sat, 28 May 2005 00:48:42 +0000 (00:48 +0000)]
Move AVM USB Bluetooth-Adapter BlueFritz! from "broken" devices list
(where I incorrectly put it initially) to "ignored" devices list (where
it should be). Pointy hat goes to me.

MFC after: 3 days

19 years agoreduce cast.
Hajimu UMEMOTO [Fri, 27 May 2005 20:44:57 +0000 (20:44 +0000)]
reduce cast.

MFC after: 1 week

19 years ago- Add support to the loader for multiple consoles.
John Baldwin [Fri, 27 May 2005 19:31:00 +0000 (19:31 +0000)]
- Add support to the loader for multiple consoles.
- Teach the i386 and pc98 loaders to honor multiple console requests from
  their respective boot2 binaries so that the same console(s) are used in
  both boot2 and the loader.
- Since the kernel doesn't support multiple consoles, whichever console is
  listed first is treated as the "primary" console and is passed to the
  kernel in the boot_howto flags.

PR: kern/66425
Submitted by: Gavin Atkinson gavin at ury dot york dot ac dot uk
MFC after: 1 week

19 years agoFix a warning by adding a missing 'const'.
John Baldwin [Fri, 27 May 2005 19:28:04 +0000 (19:28 +0000)]
Fix a warning by adding a missing 'const'.

MFC after: 1 week

19 years agoPrint out the commands from /boot.config after parsing them so that they
John Baldwin [Fri, 27 May 2005 19:26:11 +0000 (19:26 +0000)]
Print out the commands from /boot.config after parsing them so that they
output is sent to the correct console(s).

PR: kern/66425
Submitted by: Gavin Atkinson gavin at ury dot york dot ac dot uk
MFC after: 1 week

19 years agoRemove (now) unused argument 'td' from bsd_to_linux_statfs().
Pawel Jakub Dawidek [Fri, 27 May 2005 19:25:39 +0000 (19:25 +0000)]
Remove (now) unused argument 'td' from bsd_to_linux_statfs().

19 years agoRemove (now) unused argument 'td' from cvtstatfs().
Pawel Jakub Dawidek [Fri, 27 May 2005 19:23:48 +0000 (19:23 +0000)]
Remove (now) unused argument 'td' from cvtstatfs().

19 years agoSync locking in freebsd4_getfsstat() with getfsstat().
Pawel Jakub Dawidek [Fri, 27 May 2005 19:21:08 +0000 (19:21 +0000)]
Sync locking in freebsd4_getfsstat() with getfsstat().
Giant is probably also needed in kern_fhstatfs().

19 years agoUse consistent style in functions I want to modify in the near future.
Pawel Jakub Dawidek [Fri, 27 May 2005 19:15:46 +0000 (19:15 +0000)]
Use consistent style in functions I want to modify in the near future.

19 years agoDo not tread 128-bit UUID as int128. Provide separate macros to get/put
Maksim Yevmenkin [Fri, 27 May 2005 19:11:33 +0000 (19:11 +0000)]
Do not tread 128-bit UUID as int128. Provide separate macros to get/put
128-bit UUID libsdp(3). Fix 128-bit UUID printing in sdpcontrol(8).

MFC after: 3 days

19 years agodisable defining NI_WITHSCOPEID. It was obsoleted, and was exist
Hajimu UMEMOTO [Fri, 27 May 2005 19:02:12 +0000 (19:02 +0000)]
disable defining NI_WITHSCOPEID.  It was obsoleted, and was exist
only for backward compatibility since 5.2-RELEASE.

19 years agodd a '-n' option to ministat, which causes it to display only summary
Robert Watson [Fri, 27 May 2005 17:52:56 +0000 (17:52 +0000)]
dd a '-n' option to ministat, which causes it to display only summary
statistics, not graph and statistical test output.  Useful for automated
processing.

19 years agoIn the current world order, each socket has two mutexes: a mutex
Robert Watson [Fri, 27 May 2005 17:16:43 +0000 (17:16 +0000)]
In the current world order, each socket has two mutexes: a mutex
that protects socket and receive socket buffer state, and a second
mutex to protect send socket buffer state.  In some places, the
mutex shared between the socket and receive socket buffer will be
acquired twice, once by each layer, resulting in some
inconsistency, but providing the abstraction benefit of being able
to more easily separate the two mutexes in the future if desired.

When transitioning a socket to the SS_ISDISCONNECTING or
SS_ISDISCONNECTED states, grab the socket/receive socket buffer lock
once rather than grabbing it as the socket lock, modifying socket
state, then grabbing a second time as the receive lock in order to
modify the socket buffer state to indicate no further data can be
read.  This change is believed to close a race between the change in
socket state and the change in socket buffer state, which for a
remotely initiated close on a UNIX domain socket, resulted in
soreceive() returning ENOTCONN rather than an EOF condition.

A similar race still exists in the case of send, however, and is
harder to fix as the socket and send socket buffer mutexes are not
the same, and we would like to avoid holding combinations of socket
mutexes over sb_upcall until we've finished clarifying the locking
protocol for upcalls.

This change has the side affect of reducing the number of mutex
operations to initiate disconnect or perform disconnect on a
socket by two.

PR: 78824
Rerported by: Marc Olzheim <marcolz@stack.nl>
MFC after: 2 weeks

19 years agoRemove thread_upcall_check, it was used to avoid race bug in earlier
David Xu [Fri, 27 May 2005 15:57:27 +0000 (15:57 +0000)]
Remove thread_upcall_check, it was used to avoid race bug in earlier
day's sleep queue code, today the bug no longer exists.
please see 04/25/2004 freebsd-threads@ mailing list archive.

19 years agoFix for 64-bit platforms. random() returns values between 0 and RAND_MAX,
John Baldwin [Fri, 27 May 2005 15:29:01 +0000 (15:29 +0000)]
Fix for 64-bit platforms.  random() returns values between 0 and RAND_MAX,
and RAND_MAX != LONG_MAX on 64-bit platforms.

PR: amd64/81279
Submitted by: Vivek Khera vivek at khera dot org
Submitted by: Adriaan de Groot groot at kde dot org
MFC after: 1 week

19 years agoBack out ipx.h:1.18, which introduced a Linux API compatibility field in
Robert Watson [Fri, 27 May 2005 12:25:42 +0000 (12:25 +0000)]
Back out ipx.h:1.18, which introduced a Linux API compatibility field in
the ipx_net data structure.  Doing so introduced a stronger alignment
requirement for the address structure, which in turn propagated into
other dependent data structures, which turns out not to be suported by
the available IPX source code.  As a result, a number of user space
applications, such as IPX routing components, failed to operate
correctly.

RELENG_5_3 candidate?

PRs: 74059, 80266
Pointy hat to: bms
Fix by: bde
Tested by: Keith White <Keith dot White at site dot uottawa dot ca>
MFC after: 1 week
Suffering: great

19 years agoFix long (and long long) to long double, unsigned to long double and unsigned
Stefan Farfeleder [Fri, 27 May 2005 10:00:22 +0000 (10:00 +0000)]
Fix long (and long long) to long double, unsigned to long double and unsigned
long (and unsigned long long) to long double conversions.
- Add a parameter that specifies the position of the sign bit to the _QP_TTOQ
  macro, previously it always looked at bit 31.  Pass a negative number to
  disable sign inspection for unsigned types.  This fixes _Qp_xtoq(),
  _Qp_uitoq() and _Qp_uxtoq().
- In the functions __fpu_itof() and __fpu_xtof(), look at the sign bit to
  decide whether we're doing a conversion from an unsigned type.  If so, don't
  negate the mantissa if the integer exceeds the biggest signed number.

PR: 55773
Patch by: Stephen Paskaluk (based upon)
MFC after: 2 weeks

19 years agoFixup of last commit: Use the name X instead of XFree86 for the server binary,
Eivind Eklund [Fri, 27 May 2005 06:07:21 +0000 (06:07 +0000)]
Fixup of last commit: Use the name X instead of XFree86 for the server binary,
thus being compatible with both XFree86 and X.org.

Noticed by: danfe

19 years agoDocument support for the 82573 chip.
Christian Brueffer [Fri, 27 May 2005 04:52:21 +0000 (04:52 +0000)]
Document support for the 82573 chip.

19 years agoRemove sleep queue hack, it is no longer needed with current sleep queue.
David Xu [Fri, 27 May 2005 04:27:22 +0000 (04:27 +0000)]
Remove sleep queue hack, it is no longer needed with current sleep queue.
Actually, it causes process to hang when it is being debugged.

PR: gnu/77818

19 years agoSince this is already off the vendor branch: Our kernel is now in
Eivind Eklund [Fri, 27 May 2005 01:09:42 +0000 (01:09 +0000)]
Since this is already off the vendor branch: Our kernel is now in
/boot/kernel/kernel, not plain /kernel

19 years agoRemove pmap_deactivate(), we do not use it.
Olivier Houchard [Fri, 27 May 2005 00:45:39 +0000 (00:45 +0000)]
Remove pmap_deactivate(), we do not use it.

19 years agoMake the example for using xterm as a login manager to match the X of the
Eivind Eklund [Fri, 27 May 2005 00:39:11 +0000 (00:39 +0000)]
Make the example for using xterm as a login manager to match the X of the
noughties, not eighties.

19 years agoAdd cross-references to iostat, systat and vmstat.
Giorgos Keramidas [Fri, 27 May 2005 00:21:12 +0000 (00:21 +0000)]
Add cross-references to iostat, systat and vmstat.

19 years agoUse clnt_create_timed() instead of clnt_create(). The former has an
Maxime Henrion [Fri, 27 May 2005 00:05:16 +0000 (00:05 +0000)]
Use clnt_create_timed() instead of clnt_create().  The former has an
additional argument that allows us to specify a timeout, like we do for
the subsequent clnt_call() calls.

Submitted by: Jeremie Le Hen <jeremie@le-hen.org>
MFC after: 3 weeks

19 years agoLatest README to correspond to latest Intel version 2.1.7
Tony Ackerman [Thu, 26 May 2005 23:33:24 +0000 (23:33 +0000)]
Latest README to correspond to latest Intel version 2.1.7

19 years agoChanges to update driver with latest Intel driver version 2.1.7
Tony Ackerman [Thu, 26 May 2005 23:32:02 +0000 (23:32 +0000)]
Changes to update driver with latest Intel driver version 2.1.7
- Changed from using explicit devices id to using descriptive labels.
- Added support for 82573 and 82546 Quad adapters.
- Corrected support for 82547EI and 82541ER (mac_type was not assigned)
- Removed #ifdef DBG_STATS and extraneous code.

if_em_hw.c/if_em_hw.h
- Added support for 82573 and 82546 Quad adapters.
- Brought forward Intel's most current mac and phy changes.

19 years agoBaby, we are not in Kansas anymore. Nor are we in 1996 or FreeBSD 2.1.
Eivind Eklund [Thu, 26 May 2005 23:01:30 +0000 (23:01 +0000)]
Baby, we are not in Kansas anymore.  Nor are we in 1996 or FreeBSD 2.1.
Note that these papers are mostly quite old, and add a pointer to
more recent docs.

19 years agoRemove an errno reset that became unnecessary.
Eivind Eklund [Thu, 26 May 2005 22:49:08 +0000 (22:49 +0000)]
Remove an errno reset that became unnecessary.

Noticed by: juli

19 years agodecode utimes, lutimes, futimes, chflags, lchflags.
Alfred Perlstein [Thu, 26 May 2005 22:49:06 +0000 (22:49 +0000)]
decode utimes, lutimes, futimes, chflags, lchflags.

19 years agoDocument a couple of gotchas.
Eivind Eklund [Thu, 26 May 2005 22:30:12 +0000 (22:30 +0000)]
Document a couple of gotchas.

19 years agoWe are past 4.4BSD - use our new-found stat flags for pipes and fifos.
Eivind Eklund [Thu, 26 May 2005 22:14:37 +0000 (22:14 +0000)]
We are past 4.4BSD - use our new-found stat flags for pipes and fifos.

19 years agodecode mkdir args.
Alfred Perlstein [Thu, 26 May 2005 20:06:57 +0000 (20:06 +0000)]
decode mkdir args.

19 years agoInterlink systat(1), iostat(8) and vmstat(8) through their SEE ALSO
Giorgos Keramidas [Thu, 26 May 2005 17:54:16 +0000 (17:54 +0000)]
Interlink systat(1), iostat(8) and vmstat(8) through their SEE ALSO
sections, so that users of one can learn about the others easily.

19 years agos/_KLD_MODULE/KLD_MODULE/
Olivier Houchard [Thu, 26 May 2005 16:05:22 +0000 (16:05 +0000)]
s/_KLD_MODULE/KLD_MODULE/

19 years agoDon't enable interrupts in the dispatcher, there's no need to do so.
Olivier Houchard [Thu, 26 May 2005 15:02:47 +0000 (15:02 +0000)]
Don't enable interrupts in the dispatcher, there's no need to do so.

19 years agoDon't call vm_page_dirty() in pmap_nuke_pv(), it's not the place to do so, and
Olivier Houchard [Thu, 26 May 2005 15:01:13 +0000 (15:01 +0000)]
Don't call vm_page_dirty() in pmap_nuke_pv(), it's not the place to do so, and
it leads to funny things, such as pmap_remove_all() marking the page as dirty.

19 years agoMove sed(1) -l release note to the right section. Don't know what I
Bruce A. Mah [Thu, 26 May 2005 14:41:20 +0000 (14:41 +0000)]
Move sed(1) -l release note to the right section.  Don't know what I
was thinking (was I thinking?) when I wrote this.

19 years agoNew release notes: autoboot_delay -1 (+MFC), NgATM 1.2, texinfo 4.8,
Bruce A. Mah [Thu, 26 May 2005 14:38:16 +0000 (14:38 +0000)]
New release notes:  autoboot_delay -1 (+MFC), NgATM 1.2, texinfo 4.8,
pkg_version -I.

Modified release notes:  bsnmp 1.9, new manpages (devfs.conf(5),
devfs.rules(5), pthread_atfork(3)).

Other MFCs noted:  sed(1) -l.

19 years ago- Add further functionality to check for invalid characters
Ollivier Robert [Thu, 26 May 2005 10:57:03 +0000 (10:57 +0000)]
- Add further functionality to check for invalid characters
- Remove keyword 'continue' for more indepth error reporting
  on each line
- WARNS 6 Clean

Submitted by: Liam J. Foy <liamfoy@dragonflybsd.org>
MFC after:     1 week

19 years agoNew release note: Read-only support for ReiserFS.
Christian Brueffer [Thu, 26 May 2005 10:48:35 +0000 (10:48 +0000)]
New release note:  Read-only support for ReiserFS.

19 years agoFix: printed output flags (onocr) and (onlret) same as oxtabs
Andrey A. Chernov [Thu, 26 May 2005 06:57:57 +0000 (06:57 +0000)]
Fix: printed output flags (onocr) and (onlret) same as oxtabs

PR:             81256
Submitted by:   Arseny Nasokin <tarc@tarc.po.cs.msu.su>

19 years agoPlug mbuf leak, that I have introduced in 1.85. Also restore important comment
Gleb Smirnoff [Thu, 26 May 2005 06:50:00 +0000 (06:50 +0000)]
Plug mbuf leak, that I have introduced in 1.85. Also restore important comment
from if_ethersubr.c:1.178. While here adjust formatting, to make code more
readable.

Reported by: Alexey Kamyshev, rwatson

19 years agoAdd myself to the calendar.
Andrew Thompson [Thu, 26 May 2005 03:21:16 +0000 (03:21 +0000)]
Add myself to the calendar.

Approved by: bms (mentor)

19 years agoAdd a note how to use nextboot(8) to test a kernel only once.
Alexander Leidinger [Wed, 25 May 2005 21:03:13 +0000 (21:03 +0000)]
Add a note how to use nextboot(8) to test a kernel only once.

Approved by: mentor (joerg)
Discussed with: imp

19 years agoThis is conform with the terminology in
Paul Saab [Wed, 25 May 2005 17:55:27 +0000 (17:55 +0000)]
This is conform with the terminology in

  M.Mathis and J.Mahdavi,
  "Forward Acknowledgement: Refining TCP Congestion Control"
  SIGCOMM'96, August 1996.

Submitted by:   Noritoshi Demizu, Raja Mukerji

19 years agoLooking just at Makefiles was slightly too narrow to catch all
Jens Schweikhardt [Wed, 25 May 2005 17:37:57 +0000 (17:37 +0000)]
Looking just at Makefiles was slightly too narrow to catch all
"inofficial" declarations of maintainership. Grep all plain files,
and insert the actual command the list was generated with, so
future updates avoid that pitfall.

Removed sheldonh@ who just released his maintainership of ntp/doc
after I informed him of this effort.

19 years agoRelease maintainership. More ambitious minds have plans for the ntp
Sheldon Hearn [Wed, 25 May 2005 16:30:43 +0000 (16:30 +0000)]
Release maintainership.  More ambitious minds have plans for the ntp
docs.  Last I heard, Harlan Stenn was considering using FreeBSD's
pages as a starting point for the ISC NTP distribution's own pages.
If that happens, everyone wins and these can go away, to be replaced
by imported versions in contrib/ntp.

19 years agoUnder certain conditions the condition parser would go one past end of
Hartmut Brandt [Wed, 25 May 2005 16:06:14 +0000 (16:06 +0000)]
Under certain conditions the condition parser would go one past end of
the string. Until now this caused no harm, because the buffer code used
to tack two NULs onto buffers. With the new, soon to come, parsing code
this isn't the case anymore in all cases, so fix this.

19 years agoAfter provider creation!!
Pawel Jakub Dawidek [Wed, 25 May 2005 15:54:17 +0000 (15:54 +0000)]
After provider creation!!

19 years agoDocument support for some Dell PERC adapters.
Christian Brueffer [Wed, 25 May 2005 15:35:31 +0000 (15:35 +0000)]
Document support for some Dell PERC adapters.

Submitted by: Muthu_T@Dell.com
MFC after: 3 days

19 years agoSeparate out address-detaching part of if_detach into if_purgeaddrs,
Peter Edwards [Wed, 25 May 2005 13:52:03 +0000 (13:52 +0000)]
Separate out address-detaching part of if_detach into if_purgeaddrs,
so if_tap doesn't need to rely on locally-rolled code to do same.

The observable symptom of if_tap's bzero'ing the address details
was a crash in "ifconfig tap0" after an if_tap device was closed.

Reported By: Matti Saarinen (mjsaarin at cc dot helsinki dot fi)