]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
20 years agoSync with sys/boot/forth/frames.4th revision 1.2 (PC98 support).
nyan [Sun, 18 Jan 2004 15:16:12 +0000 (15:16 +0000)]
Sync with sys/boot/forth/frames.4th revision 1.2 (PC98 support).

Suggested by: matusita

20 years agoDeal better with the crypto version of the PAM library that goes
ru [Sun, 18 Jan 2004 14:58:07 +0000 (14:58 +0000)]
Deal better with the crypto version of the PAM library that goes
on the release media -- only put what is different in the crypto
version compared to the base version.  This reduces PAM entries
in /usr/lib in the "crypto" distribution to:

libpam.a
libpam.so@
libpam.so.2
pam_krb5.so@
pam_krb5.so.2
pam_ksu.so@
pam_ksu.so.2
pam_ssh.so@
pam_ssh.so.2

The libpam.so* is still redundant (it is identical to the "base"
version), but we can't set DISTRIBUTION differently for libpam.a
and libpam.so.

(The removal of libpam.so* from the crypto distribution could be
addressed by the release/scripts/crypto-make.sh script, but then
we'd also need to remove redundant PAM headers, and I'm not sure
this is worth a hassle.)

20 years agoSince "m" is not part of the "mp" chain, need to free() it.
ru [Sun, 18 Jan 2004 14:02:53 +0000 (14:02 +0000)]
Since "m" is not part of the "mp" chain, need to free() it.

Reported by: Stanford Metacompilation research group

20 years agoApparently there's a good reason why M_WAITOK malloc() is done before
ru [Sun, 18 Jan 2004 13:09:38 +0000 (13:09 +0000)]
Apparently there's a good reason why M_WAITOK malloc() is done before
xpt_create_path().

20 years agoFixed a memory leak.
ru [Sun, 18 Jan 2004 12:49:36 +0000 (12:49 +0000)]
Fixed a memory leak.

Reported by: Stanford Metacompilation research group

20 years agoEmpty vendor string overrides knowndevs
sanpei [Sun, 18 Jan 2004 12:46:19 +0000 (12:46 +0000)]
Empty vendor string overrides knowndevs

ubd_devinfo_vp() is getting an empty string from its  usbd_get_string()
call on the vendor, instead of NULL.  This means usb_knowndevs in not
consulted.

Add lines between grabbing those char *s and the USBVERBOSE ifdef to
set vendor to NULL if it is the empty string (similarly for product).

This causes vendor to be filled-out, although the product name read
overrules usb_knowndevs (this appears to be a conscience decision made
by the NetBSD folks):

PR: kern/56097
Submitted by: Hal Burch <hburch@lumeta.com>
MFC after: 1 week

20 years agoFixed a memory leak.
ru [Sun, 18 Jan 2004 12:32:06 +0000 (12:32 +0000)]
Fixed a memory leak.

Submitted by: Stanford Metacompilation research group

20 years agoFixed a memory leak.
ru [Sun, 18 Jan 2004 12:26:33 +0000 (12:26 +0000)]
Fixed a memory leak.

Submitted by: Stanford Metacompilation research group

20 years agoAdd missing free's.
sos [Sun, 18 Jan 2004 10:50:40 +0000 (10:50 +0000)]
Add missing free's.

20 years agoImplement __hdtoa() and __hldtoa() and enable printf() support for %a
das [Sun, 18 Jan 2004 10:32:49 +0000 (10:32 +0000)]
Implement __hdtoa() and __hldtoa() and enable printf() support for %a
and %A, which print floating-point numbers in hexadecimal.

20 years agoAvoid overwriting capability bits marked earlier
yar [Sun, 18 Jan 2004 10:15:48 +0000 (10:15 +0000)]
Avoid overwriting capability bits marked earlier
when setting HW checksum offload bits.

Enable available capabilities properly.

Reviewed by: sam

20 years agoDISTRIBUTION is normally single-valued.
ru [Sun, 18 Jan 2004 09:32:52 +0000 (09:32 +0000)]
DISTRIBUTION is normally single-valued.

20 years agoPut chkey(1), newkey(8), and keyserv(8) into the crypto distribution.
ru [Sun, 18 Jan 2004 09:29:47 +0000 (09:29 +0000)]
Put chkey(1), newkey(8), and keyserv(8) into the crypto distribution.

20 years agoPay attention to the timeout value passed down by the upper layer.
sanpei [Sun, 18 Jan 2004 09:19:53 +0000 (09:19 +0000)]
Pay attention to the timeout value passed down by the upper layer.
(NetBSD src/sys/dev/usb/umass.c rev.1.67)

PR: kern/58649 (Problem 1)
Submitted by: SAKIYAMA Nobuo <sakichan@sakichan.org>
MFC after: 1 week

20 years agoWe also want the *_p.a symlinks to be part of the proflibs distribution.
ru [Sun, 18 Jan 2004 09:06:40 +0000 (09:06 +0000)]
We also want the *_p.a symlinks to be part of the proflibs distribution.

20 years agoSync to 1.158 of usbdevs
sanpei [Sun, 18 Jan 2004 09:04:57 +0000 (09:04 +0000)]
Sync to 1.158 of usbdevs

20 years agoPrototype __hdtoa() and __hldtoa().
das [Sun, 18 Jan 2004 08:28:47 +0000 (08:28 +0000)]
Prototype __hdtoa() and __hldtoa().

20 years agoFix some bugs affecting the %a and %A format specifiers. Since
das [Sun, 18 Jan 2004 08:28:32 +0000 (08:28 +0000)]
Fix some bugs affecting the %a and %A format specifiers.  Since
these are not fully implemented and ifdef'd out, the bugs have
never manifested themselves.  Specifically:

- Fix a memory leak in the case where %a follows another
  floating-point format.
- Make the %a/%A code behave like %e/%E with respect to
  precision.
- It is no longer valid to assume that '-' and '0x' are
  mutually exclusive.
- Address other minor issues.

20 years agoAdd a delta accidentally omitted from the previous commit:
das [Sun, 18 Jan 2004 08:05:21 +0000 (08:05 +0000)]
Add a delta accidentally omitted from the previous commit:
Define DBL_MANH_SIZE and DBL_MANL_SIZE to be the sizes of the
high and low words of the mantissa in bits, respectively.

20 years agoadd device id for PANASONIC KXLRW32AN(USB CD-R/RW,not yet support)
sanpei [Sun, 18 Jan 2004 08:02:34 +0000 (08:02 +0000)]
add device id for PANASONIC KXLRW32AN(USB CD-R/RW,not yet support)

20 years agoDefine LDBL_MANH_SIZE and LDBL_MANL_SIZE to be the sizes of the
das [Sun, 18 Jan 2004 07:57:02 +0000 (07:57 +0000)]
Define LDBL_MANH_SIZE and LDBL_MANL_SIZE to be the sizes of the
high and low words of the mantissa in bits, respectively.

20 years agoFix a bug that caused long double subnormals to be printed
das [Sun, 18 Jan 2004 07:53:49 +0000 (07:53 +0000)]
Fix a bug that caused long double subnormals to be printed
incorrectly on architectures without an explicit normalization
bit (sparc64, powerpc).

20 years agoAdded two utility targets "secure" and "insecure", analogous to
ru [Sun, 18 Jan 2004 07:44:53 +0000 (07:44 +0000)]
Added two utility targets "secure" and "insecure", analogous to
"kerberize" and "dekerberize" in kerberos5/Makefile.  These can
be used to recompile bits with optional crypto support with and
without crypto, respectively.

Reviewed by: markm

20 years agoWork around a recently-introduced gcc bug. The compiler no longer
das [Sun, 18 Jan 2004 07:06:21 +0000 (07:06 +0000)]
Work around a recently-introduced gcc bug.  The compiler no longer
accepts certain floating point constant representations that are
legal in C99.

20 years agoIncrease UMA_BOOT_PAGES because of changes to pv entry initialization in
alc [Sun, 18 Jan 2004 05:51:06 +0000 (05:51 +0000)]
Increase UMA_BOOT_PAGES because of changes to pv entry initialization in
revision 1.457 of i386/i386/pmap.c.

20 years agoPC98 uses the different frame code.
nyan [Sun, 18 Jan 2004 04:13:27 +0000 (04:13 +0000)]
PC98 uses the different frame code.

20 years agoFix to support KANA and graphics characters which code are over 0x80.
nyan [Sun, 18 Jan 2004 04:10:45 +0000 (04:10 +0000)]
Fix to support KANA and graphics characters which code are over 0x80.

20 years agoDon't acquire Giant in vm_object_deallocate() unless the object is vnode-
alc [Sun, 18 Jan 2004 03:44:14 +0000 (03:44 +0000)]
Don't acquire Giant in vm_object_deallocate() unless the object is vnode-
backed.

20 years agoFix type in comment, 's/writtent/written/'
johan [Sun, 18 Jan 2004 00:17:14 +0000 (00:17 +0000)]
Fix type in comment, 's/writtent/written/'

PR: 56954
Submitted by: Dan Langille <dan@langille.org>

20 years agoDo not install SYMLINKS if both NO_MAILWRAPPER and
johan [Sat, 17 Jan 2004 23:58:23 +0000 (23:58 +0000)]
Do not install SYMLINKS if both NO_MAILWRAPPER and
NO_SENDMAIL is defined.

PR: 57058
Reported by: Henri Hennebert <hlh@cocoon.cercle.be>
Melvyn Sopacua on current@
Submitted by: ru@
MFC after: 2 weeks

20 years agoAdd support for the Silicon Image 3114 4 channel SATA controller.
sos [Sat, 17 Jan 2004 23:34:13 +0000 (23:34 +0000)]
Add support for the Silicon Image 3114 4 channel SATA controller.

20 years agoIt appears that the changes in the resources allocated is causing much
imp [Sat, 17 Jan 2004 21:54:04 +0000 (21:54 +0000)]
It appears that the changes in the resources allocated is causing much
pain and suffering.  Attempt to back it out by removing the 'if the
requested range is larger than the window, clip to the window' code.
This is a band-aide until the issues are better understood and the
issues with the lazy allocation patches are resolved.

20 years agoDon't panic because of RPC proto mismatches. Whitespace cleanup.
alfred [Sat, 17 Jan 2004 21:25:05 +0000 (21:25 +0000)]
Don't panic because of RPC proto mismatches.  Whitespace cleanup.

Submitted by: Jim Rees <rees@umich.edu>

20 years agoHandle sf_buf_alloc() returning null. This can happen if the
gallatin [Sat, 17 Jan 2004 21:16:51 +0000 (21:16 +0000)]
Handle sf_buf_alloc() returning null.  This can happen if the
process takes a signal while waiting for an sf_buf to become available.

Reviewed by: alc

20 years agoOnce upon a time we had both "crypto" and "krb5" distributions,
ru [Sat, 17 Jan 2004 19:22:36 +0000 (19:22 +0000)]
Once upon a time we had both "crypto" and "krb5" distributions,
and rebuilt some bits with crypto but without Kerberos support
(most notably SSH) during "make release", to put them into the
"crypto" distribution.

Now that we don't ship the separate "krb5" distribution anymore
(it's now part of the "crypto" distribuion), don't waste time
recompiling SSH bits without crypto and without Kerberos support
in an attempt to put them in the "base" distribution -- it just
doesn't work as SSH always uses crypto code.

We avoid this by not rebuilding KPROGS from kerberos5/Makefile in
release/Makefile and adding "libpam" to SPROGS in secure/Makefile
to ensure it's still rebuilt without crypto support for the "base"
distribution.  (Disabling crypto (NOCRYPT) also disables building
of Kerberos-related PAM modules, and it's OK to depend on this.)

This should be a no-op change saving some "make release" time.

20 years agoGuess better the source for object files in case .depend file is
ru [Sat, 17 Jan 2004 18:51:55 +0000 (18:51 +0000)]
Guess better the source for object files in case .depend file is
missing and there are multiple choices using multiple inference
(suffix transformation) rules.

This is known to fix compilation of s_log1p.o in lib/msun on i386,
as otherwise it attempted to use s_log1p.S as the source (which is
marked broken) instead of legal s_log1p.c which is in CFLAGS.  The
normal case where .depend file exists is not affected.

Reviewed by: bde

20 years agoMFRELENG_5_2: Default the snapshot versioning to 5.2.
ru [Sat, 17 Jan 2004 18:25:43 +0000 (18:25 +0000)]
MFRELENG_5_2: Default the snapshot versioning to 5.2.

PR: misc/61402
Submitted by: Chris Knight

20 years ago- Properly build both crypto and non-crypto versions of the
ru [Sat, 17 Jan 2004 13:41:16 +0000 (13:41 +0000)]
- Properly build both crypto and non-crypto versions of the
  package management tools.

- Drop redundant dependency of pkg_create(1) and pkg_delete(1)
  on crypto libraries now that they do not link with libfetch.

20 years agoReplaced black magic with white magic.
ru [Sat, 17 Jan 2004 13:27:27 +0000 (13:27 +0000)]
Replaced black magic with white magic.

20 years agoremove elan_mmcr, I'm not sure I understand what it did here in the
phk [Sat, 17 Jan 2004 13:13:48 +0000 (13:13 +0000)]
remove elan_mmcr, I'm not sure I understand what it did here in the
first place.

20 years agoRemove an extraneous semicolon (;;).
mtm [Sat, 17 Jan 2004 11:52:37 +0000 (11:52 +0000)]
Remove an extraneous semicolon (;;).

20 years agoAdd Canon N676U and Canon N1220U to the list of supported devices.
simon [Sat, 17 Jan 2004 11:51:23 +0000 (11:51 +0000)]
Add Canon N676U and Canon N1220U to the list of supported devices.

MFC after: 1 week

20 years agoUse the checkyesno routine instead of a case statement to
mtm [Sat, 17 Jan 2004 11:31:15 +0000 (11:31 +0000)]
Use the checkyesno routine instead of a case statement to
check an rc.conf(5) variable. This is in line with rc.d style.

20 years agoRemove checks for the FreeBSD OS.
mtm [Sat, 17 Jan 2004 11:25:16 +0000 (11:25 +0000)]
Remove checks for the FreeBSD OS.

20 years agoMake conform to general rc.d style.
mtm [Sat, 17 Jan 2004 11:19:01 +0000 (11:19 +0000)]
Make conform to general rc.d style.

20 years agoo Rename the start command to conform to rc.d style.
mtm [Sat, 17 Jan 2004 11:15:04 +0000 (11:15 +0000)]
o Rename the start command to conform to rc.d style.
o Move the additional commands list to the end of the
  variable initialization to conform to rc.d style

20 years agoSimplify setting of securelevel.
mtm [Sat, 17 Jan 2004 11:11:18 +0000 (11:11 +0000)]
Simplify setting of securelevel.

20 years agoRemove variables whose sole purpose was to make easier to
mtm [Sat, 17 Jan 2004 11:06:41 +0000 (11:06 +0000)]
Remove variables whose sole purpose was to make easier to
work with the NetBSD specific stuff.

20 years agoLuke Mewburn has indicated that they (NetBSD) are not interested
mtm [Sat, 17 Jan 2004 10:59:43 +0000 (10:59 +0000)]
Luke Mewburn has indicated that they (NetBSD) are not interested
in keeping the scripts under rc.d in sync with us. So, remove
NetBSD specific stuff (which made our scripts more complicated
than necessary).

The NetBSD ident string will be left intact, both for history and
also incase we wish to pull in future versions.

20 years agoMostly mechanical rework of libalias:
phk [Sat, 17 Jan 2004 10:52:21 +0000 (10:52 +0000)]
Mostly mechanical rework of libalias:

Makes it possible to have multiple packet aliasing instances in a
single process by moving all static and global variables into an
instance structure called "struct libalias".

Redefine a new API based on s/PacketAlias/LibAlias/g

Add new "instance" argument to all functions in the new API.

Implement old API in terms of the new API.

20 years agoMake sure set the media type in the phy, we cannot trust it to have chosen
phk [Sat, 17 Jan 2004 10:44:39 +0000 (10:44 +0000)]
Make sure set the media type in the phy, we cannot trust it to have chosen
the type we happen to want.

Bug triggered by net-booting soekris hardware which comes up in 10/hdx mode
by default.

20 years agoDrop a hint about '+' being a potential wildcard.
phk [Sat, 17 Jan 2004 10:41:14 +0000 (10:41 +0000)]
Drop a hint about '+' being a potential wildcard.

20 years agoLuke Mewburn has indicated that they (NetBSD) are not interested
mtm [Sat, 17 Jan 2004 10:40:45 +0000 (10:40 +0000)]
Luke Mewburn has indicated that they (NetBSD) are not interested
in keeping the scripts under rc.d in sync with us. So, remove
NetBSD specific stuff (which made our scripts more complicated
than necessary).

The NetBSD ident string will be left intact, both for history and
also incase we wish to pull in future versions.

20 years agoAs previously announced: discontinue use of makedev() call in soundcode.
phk [Sat, 17 Jan 2004 10:37:11 +0000 (10:37 +0000)]
As previously announced:  discontinue use of makedev() call in soundcode.

This takes us a lot closer to refcounting dev_t.

This patch originally by cg@ with a few minor changes by me.

It is largely untested, but has been HEADSUP'ed twice, so presumably
people have not found any issues with it.

Submitted by: cg@

20 years agoLuke Mewburn has indicated that they (NetBSD) are not interested
mtm [Sat, 17 Jan 2004 10:16:38 +0000 (10:16 +0000)]
Luke Mewburn has indicated that they (NetBSD) are not interested
in keeping the scripts under rc.d in sync with us. So, begin removal
of NetBSD specific stuff (which made our scripts more complicated
than necessary), starting with the NetBSD KEYWORD.

20 years agoRemove crossref to pam.conf(5) which never existed.
schweikh [Sat, 17 Jan 2004 09:46:49 +0000 (09:46 +0000)]
Remove crossref to pam.conf(5) which never existed.

20 years agoBackout revision 1.6, because some stub functions not in libc, and
davidxu [Sat, 17 Jan 2004 07:15:06 +0000 (07:15 +0000)]
Backout revision 1.6, because some stub functions not in libc, and
non-threaded won't build. The cancellation point support should be
further discussed.

20 years ago- Handle failure of cam_sim_alloc().
mdodd [Sat, 17 Jan 2004 05:57:52 +0000 (05:57 +0000)]
- Handle failure of cam_sim_alloc().
  This prevents xpt_bus_register() from dereferencing NULL.
- Assign pointer to NULL after cam_sim_free().

Submitted by:  Paul Twohey <twohey@CS.Stanford.EDU>

20 years agoReplaced outdated comment.
ru [Sat, 17 Jan 2004 03:28:27 +0000 (03:28 +0000)]
Replaced outdated comment.

20 years agoImprove translation.
hrs [Sat, 17 Jan 2004 03:20:48 +0000 (03:20 +0000)]
Improve translation.

20 years agoRemoved well outdated comment.
ru [Sat, 17 Jan 2004 03:12:46 +0000 (03:12 +0000)]
Removed well outdated comment.

20 years agoReturn EPERM if mutex owner is not current thread but it tries to
davidxu [Sat, 17 Jan 2004 03:09:57 +0000 (03:09 +0000)]
Return EPERM if mutex owner is not current thread but it tries to
unlock the mutex, old code confuses some programs when it returns EINVAL.

Noticed by: bland

20 years agoEnable cancellation point in sem_wait, it is required by POSIX.
davidxu [Sat, 17 Jan 2004 02:45:37 +0000 (02:45 +0000)]
Enable cancellation point in sem_wait, it is required by POSIX.
For pshared semaphore, this commit still does not enable cancellation
point, I think there should be a pthread_enter_cancellation_point_np
for libc to implement a safe cancellation point.

20 years agoFixed the fatal bug: CFLAGS should be added to, not overridden.
ru [Sat, 17 Jan 2004 02:27:55 +0000 (02:27 +0000)]
Fixed the fatal bug: CFLAGS should be added to, not overridden.
This prevented CPU and optimization flags to be added here.

20 years agoCreate ${WORLDTMP}/legacy/usr/include/c++/3.3 for usr.bin/lex.
ru [Sat, 17 Jan 2004 01:24:13 +0000 (01:24 +0000)]
Create ${WORLDTMP}/legacy/usr/include/c++/3.3 for usr.bin/lex.

20 years agoRestore correct semantics for F_DUPFD fcntl. This should fix the errors
des [Sat, 17 Jan 2004 00:59:04 +0000 (00:59 +0000)]
Restore correct semantics for F_DUPFD fcntl.  This should fix the errors
people have been getting with configure scripts.

20 years agoFixed spacing in previous revision.
ru [Sat, 17 Jan 2004 00:31:38 +0000 (00:31 +0000)]
Fixed spacing in previous revision.

20 years agoDo not allow operations which cause known file-system corruption.
green [Fri, 16 Jan 2004 22:55:25 +0000 (22:55 +0000)]
Do not allow operations which cause known file-system corruption.

20 years agoRemove a warning.
green [Fri, 16 Jan 2004 22:54:33 +0000 (22:54 +0000)]
Remove a warning.

20 years agoWITNESS won't let us hold two filedesc locks at the same time, so juggle
des [Fri, 16 Jan 2004 21:54:56 +0000 (21:54 +0000)]
WITNESS won't let us hold two filedesc locks at the same time, so juggle
fdp and newfdp around a bit.

20 years agoKASSERT() that initproc->p_pid is 1. Very bad things happen if init's
rwatson [Fri, 16 Jan 2004 20:29:23 +0000 (20:29 +0000)]
KASSERT() that initproc->p_pid is 1.  Very bad things happen if init's
pid isn't 1, and it can actually occur if kthread_create() is called
before SUB_SI_CREATE_INIT without RFHIGHPID.

Discussed with: jhb

20 years ago_OS and _OS_ are the same object (due to _ padding for spaces).
njl [Fri, 16 Jan 2004 18:10:42 +0000 (18:10 +0000)]
_OS and _OS_ are the same object (due to _ padding for spaces).

Pointed out by: Ducrot Bruno (Linux)

20 years agoSplice in cryptodev (this man page originated on openbsd where there is
sam [Fri, 16 Jan 2004 17:21:02 +0000 (17:21 +0000)]
Splice in cryptodev (this man page originated on openbsd where there is
no cryptodev module.)  Should probably create a separate cryptodev man
page (or a link to this one) but for now add enough so folks understand
the relationship between the two.

Prodded by: phk

20 years agoFix an upper-vnode leak created in revision 1.52. When an upper-layer
green [Fri, 16 Jan 2004 16:31:01 +0000 (16:31 +0000)]
Fix an upper-vnode leak created in revision 1.52.  When an upper-layer
file has been removed, it should be purged from the cache, but it need
not be removed from the directory stack causing corruption; instead,
it will simply be removed once the last references and holds on it
are dropped at the end of the unlink/rmdir system calls, and the
normal !UN_CACHED VOP_INACTIVE() handler for unionfs finishes it off.

This is easily reproduced by repeated "echo >file; rm file" on a
unionfs mount.  Strangely, "echo -n >file; rm file" didn't make
it happen.

20 years agolex(1) prior to flex.skl,v 1.8 and gen.c,v 1.7 requires bootstrapping.
ru [Fri, 16 Jan 2004 16:06:54 +0000 (16:06 +0000)]
lex(1) prior to flex.skl,v 1.8 and gen.c,v 1.7 requires bootstrapping.

Prodded by: nectar

20 years ago- Build things in pure alphabetical dictionary order.
ru [Fri, 16 Jan 2004 15:55:29 +0000 (15:55 +0000)]
- Build things in pure alphabetical dictionary order.

- Unify the conditional assignments section so that architectural
  exclusions come first, sorted, then options and !options, sorted
  by the option name, also in directory order, then architecture
  specific sections, sorted by the architecture name, with i386
  being a traditional exception.

Prodded by: bde

20 years agoSwitch manpath for FreeBSD manpages to 5.2-current.
bmah [Fri, 16 Jan 2004 15:39:56 +0000 (15:39 +0000)]
Switch manpath for FreeBSD manpages to 5.2-current.

20 years agoSync to 1.157 of usbdevs
sanpei [Fri, 16 Jan 2004 15:31:20 +0000 (15:31 +0000)]
Sync to 1.157 of usbdevs

20 years agoAdd support SUNTAC U-Cable type A3(AS64LX)
sanpei [Fri, 16 Jan 2004 15:29:27 +0000 (15:29 +0000)]
Add support SUNTAC U-Cable type A3(AS64LX)

Submitted by: TAKAHASHI Daisuke <takaha@mue.biglobe.ne.jp> [FreeBSD-users-jp 72475]
and MIZUTANI Masaki <m-saki@rr.iij4u.or.jp>
MFC after: 1 week

20 years ago- Build things in pure dictionary order (see sort(1)).
ru [Fri, 16 Jan 2004 15:23:19 +0000 (15:23 +0000)]
- Build things in pure dictionary order (see sort(1)).

- Unify the conditional assignments section so that architectural
  exclusions come first, then options and !options, sorted by the
  option name, also in directory order, then architecture specific
  sections, sorted by the architecture name, with i386 being a
  traditional exception.

Prodded by: bde

20 years agoAdd an implementation of pthread_rwlock_timed{rd,wr}lock() to libthr with
mtm [Fri, 16 Jan 2004 10:52:10 +0000 (10:52 +0000)]
Add an implementation of pthread_rwlock_timed{rd,wr}lock() to libthr with
attendant documentation.

20 years agoRemove two KASSERTs which were overly paranoid.
des [Fri, 16 Jan 2004 08:45:56 +0000 (08:45 +0000)]
Remove two KASSERTs which were overly paranoid.

20 years agoo We are not required to initialize an invalid rwlock. So axe all that
mtm [Fri, 16 Jan 2004 07:10:30 +0000 (07:10 +0000)]
o We are not required to initialize an invalid rwlock. So axe all that
  code and simply return EINVAL (which is allowed by the standard) in
  all those pthread functions that previously initialized it.

o Refactor the pthread_rwlock_[try]rdlock() and pthread_rwlock_[try]wrlock()
  functions. They are now completeley condensed into rwlock_rdlock_common()
  and rwlock_wrlock_common(), respectively.

o If the application tries to destroy an rwlock that is currently
  held by a thread return EBUSY where it previously went ahead and
  freed all resources associated with the lock.

o Refactor _pthread_rwlock_init() to make it look (relatively) sane.

o When obtaining a read lock on an rwlock the check for whether it
  would exceed the maximum allowed read locks should happen *before*
  we obtain the lock.

o The pthread_rwlock_* functions shall *never* return EINTR, so make
  sure to requeue/resuspend the thread if it encounters such an error.

o Make a note that pthread_rwlock_unlock() needs to ensure it holds a
  lock on an rwlock it tries to unlock. It will be implemented in a
  separate commit because it requires some additional rwlock infrastructure.

20 years agoThe ndis_kspin_lock type is called KSPIN_LOCK in MS-Windows.
obrien [Fri, 16 Jan 2004 02:07:04 +0000 (02:07 +0000)]
The ndis_kspin_lock type is called KSPIN_LOCK in MS-Windows.
According to the Windows DDK header files, KSPIN_LOCK is defined like this:
typedef ULONG_PTR KSPIN_LOCK;

From basetsd.h (SDK, Feb. 2003):
typedef [public] unsigned __int3264 ULONG_PTR, *PULONG_PTR;
typedef unsigned __int64 ULONG_PTR, *PULONG_PTR;
typedef _W64 unsigned long ULONG_PTR, *PULONG_PTR;

The keyword __int3264 specifies an integral type that has the following
properties:
 + It is 32-bit on 32-bit platforms
 + It is 64-bit on 64-bit platforms
 + It is 32-bit on the wire for backward compatibility.
   It gets truncated on the sending side and extended appropriately
   (signed or unsigned) on the receiving side.

Thus register_t seems the proper mapping onto FreeBSD for spin locks.

20 years agoCatch up with ATA UMA changes
grehan [Thu, 15 Jan 2004 23:52:32 +0000 (23:52 +0000)]
Catch up with ATA UMA changes

20 years agoConnect tx.4 to the build and MLINK it to if_tx.4
brueffer [Thu, 15 Jan 2004 23:05:22 +0000 (23:05 +0000)]
Connect tx.4 to the build and MLINK it to if_tx.4

20 years agoDon't claim this file is i386 specific
brueffer [Thu, 15 Jan 2004 23:03:17 +0000 (23:03 +0000)]
Don't claim this file is i386 specific

20 years agotx.4 has been repocopied to src/share/man/man4
brueffer [Thu, 15 Jan 2004 23:01:12 +0000 (23:01 +0000)]
tx.4 has been repocopied to src/share/man/man4

Repo-surgery by: joe

20 years agoThe definition for __stdcall logically belongs in pe_var.h, but
wpaul [Thu, 15 Jan 2004 21:31:49 +0000 (21:31 +0000)]
The definition for __stdcall logically belongs in pe_var.h, but
the definitions for NDIS_BUS_SPACE_IO and NDIS_BUS_SPACE_MEM logically
belong in hal_var.h. At least, that's my story, and I'm sticking to it.

Also, remove definition of __stdcall from if_ndis.c now that it's pulled
in from pe_var.h.

20 years agoProvide an escape to our "savestring" compatibility with ancient readline
obrien [Thu, 15 Jan 2004 19:37:45 +0000 (19:37 +0000)]
Provide an escape to our "savestring" compatibility with ancient readline
versions madness.  Define 'RL_NO_COMPAT' and be worry free.

20 years agoCreate NDIS_BUS_SPACE_{IO,MEM} to abstract MD BUS_SPACE macros.
obrien [Thu, 15 Jan 2004 19:34:56 +0000 (19:34 +0000)]
Create NDIS_BUS_SPACE_{IO,MEM} to abstract MD BUS_SPACE macros.
Provide appropriate definitions for i386 and AMD64.

20 years agoTake care to drop locks when calling malloc()
des [Thu, 15 Jan 2004 18:50:11 +0000 (18:50 +0000)]
Take care to drop locks when calling malloc()

20 years ago- Move the code to try to open a single chunk file and prompt for the
jhb [Thu, 15 Jan 2004 18:38:15 +0000 (18:38 +0000)]
- Move the code to try to open a single chunk file and prompt for the
  associated floppy if needed into a static split_openfile() function.
- Use this function in splitfs_open() to open the first chunk rather
  than using open() directly.  This allows the first chunk to be located
  on a different disk than the actual foo.split file.

20 years agoWhitespace tweaks to make indentation consistent within this file and even
jhb [Thu, 15 Jan 2004 18:36:48 +0000 (18:36 +0000)]
Whitespace tweaks to make indentation consistent within this file and even
within a single function.

20 years ago- Print out line number in a panic message using %d rather than %p. Line
jhb [Thu, 15 Jan 2004 18:35:32 +0000 (18:35 +0000)]
- Print out line number in a panic message using %d rather than %p.  Line
  numbers aren't pointers.
- Add a __printflike() attribute to libstand's panic() prototype to catch
  such bogons in the future.

20 years agoRemoved outdated text about libc_r replacing libc (it now provides
ru [Thu, 15 Jan 2004 17:58:26 +0000 (17:58 +0000)]
Removed outdated text about libc_r replacing libc (it now provides
just libc functions wrappers), and updated text to match reality:
there are three threading libraries in FreeBSD these days.

Removed instructions of how not to build libc_r, it's documented in
the make.conf(5) manpage already.

Removed description of the FreeBSD-specific gcc(1) option, -pthread.
While it's still provided (for backwards compatibility reasons),
its usefulness is questionable.

20 years agoFixed style of previous commit.
ru [Thu, 15 Jan 2004 17:27:28 +0000 (17:27 +0000)]
Fixed style of previous commit.

Submitted by: bde

20 years agoawi(4) needs 'device wlan' now
brueffer [Thu, 15 Jan 2004 17:21:06 +0000 (17:21 +0000)]
awi(4) needs 'device wlan' now

20 years agoRemoved the IMPLEMENTATION NOTES section made redundant by not
ru [Thu, 15 Jan 2004 16:13:49 +0000 (16:13 +0000)]
Removed the IMPLEMENTATION NOTES section made redundant by not
putting libc_r into the LIBRARY section.

20 years agopthread_multi_np() is implemented in all threading libraries.
ru [Thu, 15 Jan 2004 16:12:11 +0000 (16:12 +0000)]
pthread_multi_np() is implemented in all threading libraries.
pthread_single_np() is not implemented in libthr.