]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
24 years agoDon't wakeup threads when there is a process signal and no installed
Daniel Eischen [Tue, 28 Dec 1999 18:08:09 +0000 (18:08 +0000)]
Don't wakeup threads when there is a process signal and no installed
handler.  Thread-to-thread signals (pthread_signal) are treated differently
than process signals; a pthread_signal can wakeup a blocked thread if
a signal handler is not installed for that signal.

Found by: ACE tests

24 years agoBuild openssl properly during make world.
Kris Kennaway [Tue, 28 Dec 1999 16:22:50 +0000 (16:22 +0000)]
Build openssl properly during make world.

24 years agoAvoid to write garbage if uiomove fails.
Boris Popov [Tue, 28 Dec 1999 16:14:54 +0000 (16:14 +0000)]
Avoid to write garbage if uiomove fails.

24 years agoFix an overflow in the msdosfs_read() function which exposed on the files
Boris Popov [Tue, 28 Dec 1999 15:34:23 +0000 (15:34 +0000)]
Fix an overflow in the msdosfs_read() function which exposed on the files
with size > 2GB.

PR: 15639
Submitted by: Tim Kientzle <kientzle@acm.org>
Reviewed by: phk

24 years agoIt is possible that number of sectors specified in the BPB
Boris Popov [Tue, 28 Dec 1999 15:27:39 +0000 (15:27 +0000)]
It is possible that number of sectors specified in the BPB
will exceed FAT capacity. This will lead to kernel panic while other
systems just limit number of clusters.

PR: 4381, 15136
Reviewed by: phk

24 years agoTypo cops.
Tim Vanderhoek [Tue, 28 Dec 1999 15:24:01 +0000 (15:24 +0000)]
Typo cops.

24 years agoAdd history: The reallocf() function first appeared in FreeBSD-3.0.
Tim Vanderhoek [Tue, 28 Dec 1999 15:14:59 +0000 (15:14 +0000)]
Add history: The reallocf() function first appeared in FreeBSD-3.0.

See imp's 199808201619.KAA20970@harmony.village.org in freebsd-hackers (the
reallocf.c cvs history mistakenly refers to freebsd-current).

24 years agoAdd ".Xref tolower 3" since its internal use is inferred in DESCRIPTION.
Tim Vanderhoek [Tue, 28 Dec 1999 14:57:33 +0000 (14:57 +0000)]
Add ".Xref tolower 3" since its internal use is inferred in DESCRIPTION.

24 years agoAvoid the potentially confusing term "a null pointer" and say "the NULL
Tim Vanderhoek [Tue, 28 Dec 1999 14:47:00 +0000 (14:47 +0000)]
Avoid the potentially confusing term "a null pointer" and say "the NULL
pointer" instead.  The potential confusion arises because the string/*.3
pages use the term "null-terminated string" (which is permissable).  Moreover,
this also makes these two manpages more consistent with the other string/*.3
manpages.

24 years agoAdd .Xrefs to tolower.3 and toupper.3, respectively.
Tim Vanderhoek [Tue, 28 Dec 1999 14:10:21 +0000 (14:10 +0000)]
Add .Xrefs to tolower.3 and toupper.3, respectively.

24 years agoBetter handling groff failures. If popen(ROFF_COMMAND) returns zero bytes
Alexey Zelkin [Tue, 28 Dec 1999 12:55:58 +0000 (12:55 +0000)]
Better handling groff failures. If popen(ROFF_COMMAND) returns zero bytes
then handle it as problem. This commit fixes problem with archiving empty files.

PR: gnu/5767
Submitted by: Bill Fenner <fenner@parc.xerox.com>

24 years agoUse the ctype.h version of isascii() - it doesn't loose precision and think
Peter Wemm [Tue, 28 Dec 1999 11:48:23 +0000 (11:48 +0000)]
Use the ctype.h version of isascii() - it doesn't loose precision and think
that 0x100 (int) is an ascii character.

Submitted by: bde

24 years agoTake into account the fact that "[" may be called with a path,
Sheldon Hearn [Tue, 28 Dec 1999 09:34:57 +0000 (09:34 +0000)]
Take into account the fact that "[" may be called with a path,
for example "/bin/[".

Reported by: Vlad Skvortsov <vss@ulstu.ru>
Reported by: Peter Jeremy <peter.jeremy@alcatel.com.au>
Message-Id: 99Dec27.111307est.40321@border.alcanet.com.au

24 years agoMake cloning mask sockaddr (genmask) possible.
Ruslan Ermilov [Tue, 28 Dec 1999 08:38:01 +0000 (08:38 +0000)]
Make cloning mask sockaddr (genmask) possible.

PR: kern/3061
Reviewed by: wollman

24 years agoAdded following modes:
KATO Takenori [Tue, 28 Dec 1999 07:38:38 +0000 (07:38 +0000)]
Added following modes:
 5in     HD 2 heads, 77 cylinders, 8 sectors/track, 1024 bytes/sector
 5/3.5in DD 2 heads, 80 cylinders, 8 sectors/track, 512 bytes/sector

Meanings of the rogrammer-readeble fd name were explained by Brian
Fundakowski Feldman and Peter Wemm in hackers list and NOKUBI
Hirotaka.

Reviewed by: nyan

24 years agoFix a panic when doing non-multiples of PAGE_SIZE or misaligned transfers
Peter Wemm [Tue, 28 Dec 1999 07:32:34 +0000 (07:32 +0000)]
Fix a panic when doing non-multiples of PAGE_SIZE or misaligned transfers
to a swap backed vn device.

OK'ed by: dillon

24 years agoFix the swap backed vn case - this was broken by my rev 1.128 to
Peter Wemm [Tue, 28 Dec 1999 07:30:55 +0000 (07:30 +0000)]
Fix the swap backed vn case - this was broken by my rev 1.128 to
swap_pager.c and related commits.

Essentially swap_pager.c is backed out to before the changes, but
swapdev_vp is converted into a real vnode with just VOP_STRATEGY().
It no longer abuses specfs vnops and no longer needs a dev_t and
/dev/drum (or /dev/swapdev) for the intermediate layer.

This essentially restores the vnode interface as the interface to the
bottom of the swap pager, and vm_swap.c provides a clean vnode interface.

This will need to be revisited when we swap to files (vnodes) - which
is the other reason for keeping the vnode interface between the swap pager
and the swap devices.

OK'ed by: dillon

24 years agoSuppress vast quantities of unneeded warnings spewed by libc's gethostbydns
Robert Watson [Tue, 28 Dec 1999 07:21:08 +0000 (07:21 +0000)]
Suppress vast quantities of unneeded warnings spewed by libc's gethostbydns
on encountering a real-world SIG record during a lookup of another type.

PR: bin/7352
Reviewed by: peter, eivind

24 years agoCorrectly handle a user-requested abort in the middle of displaying a
Mike Smith [Tue, 28 Dec 1999 07:19:22 +0000 (07:19 +0000)]
Correctly handle a user-requested abort in the middle of displaying a
help subtopic.

PR: kern/13196
Submitted by: MIHIRA Sanpei Yoshiro <sanpei@sanpei.org>

24 years agoAdd display of maximum allowed mbuf count to match mbuf cluster count.
Mike Smith [Tue, 28 Dec 1999 06:38:37 +0000 (06:38 +0000)]
Add display of maximum allowed mbuf count to match mbuf cluster count.

Submitted by: Bosko Milekic <bmilekic@dsuper.net>

24 years agoActively limit the allocation of mbufs to NMBUFS/nmbufs and mbuf clusters
Mike Smith [Tue, 28 Dec 1999 06:35:57 +0000 (06:35 +0000)]
Actively limit the allocation of mbufs to NMBUFS/nmbufs and mbuf clusters
to NMBCLUSTERS/nmbclusters/kern.ipc.nmbclusters.

Add a read-only sysctl kern.ipc.nmbufs matching kern.ipc.nmbclusters.

Submitted by: Bosko Milekic <bmilekic@dsuper.net>

24 years agoFix a small typo in the comments.
Bill Fumerola [Tue, 28 Dec 1999 06:04:29 +0000 (06:04 +0000)]
Fix a small typo in the comments.

24 years agoDedocument one of the BUGS listed in the last commit. The bug (needless
Tim Vanderhoek [Tue, 28 Dec 1999 05:56:33 +0000 (05:56 +0000)]
Dedocument one of the BUGS listed in the last commit.  The bug (needless
calculation of line numbers) never existed and the two bugs that made me
think it existed have been fixed (see recent commits about this date to
linenum.c:r.1.3 and ch.c:r.1.8 fixing broken line-number buffering and
braindead algorithms respectively).

24 years agoRather than use an LRU-ordered circular list to store buffered data,
Tim Vanderhoek [Tue, 28 Dec 1999 05:47:24 +0000 (05:47 +0000)]
Rather than use an LRU-ordered circular list to store buffered data,
simply keep an index into the last access on the circular list and begin
searches at that point.  An LRU list is inappropriate here since the
vast majority of accesses will occur in the same order that the list
is created in.  The only case where an LRU is remotely useful here is when
reading from a file and the user is jumping to randomish positions and
constantly returning to some central position.  Even for this case it is
such a small optimization as not to be noticed in an interactive program
such as more(1).

This change results in a _tremendously_ noticable speed-up when reading long
files through a pipe (where long = ~200k, machine = ~2.5h single-disk
worldstone).

24 years agoSmall bug fix and improvements
Yoshinobu Inoue [Tue, 28 Dec 1999 05:37:39 +0000 (05:37 +0000)]
Small bug fix and improvements
  (1)added error check of if_nameindex() return value at getaddrinfo().
  (2)print out more detailed information when getaddrinfo() error value
     is EAI_SYSTEM.(in this case system error num is kept in errno)

(1) is Discovered by: jinmei@kame.net in KAME environment.

24 years agoUpgrade to the pam_ssh module, version 1.1..
Brian Feldman [Tue, 28 Dec 1999 05:32:54 +0000 (05:32 +0000)]
Upgrade to the pam_ssh module, version 1.1..

(From the author:)
Primarily, I have added built-in functions for manipulating the
environment, so putenv() is no longer used.  XDM and its variants
should now work without modification.  Note that the new code uses
the macros in <sys/queue.h>.

Submitted by: Andrew J. Korty <ajk@iu.edu>

24 years agoWork around an assert failure in the dynamic linker's default thread
John Polstra [Tue, 28 Dec 1999 04:38:17 +0000 (04:38 +0000)]
Work around an assert failure in the dynamic linker's default thread
locking functions.  If an application loads a shared object with
dlopen() and the shared object has an init function which requires
lazy binding, then _rtld_bind is called when the thread is already
inside the dynamic linker.  This leads to a recursive acquisition
of the lock, which I was not expecting -- hence the assert failure.

This work-around makes the default locking functions handle recursive
locking.  It is NOT the correct fix -- that should be implemented
at the generic locking level rather than in the default locking
functions.  I will implement the correct fix in a future commit.

Since the dllockinit() interface will likely need to change, warn
about that in both the man page and the header file.

24 years agoDelete dead code and clean comments a little.
Tim Vanderhoek [Tue, 28 Dec 1999 03:01:20 +0000 (03:01 +0000)]
Delete dead code and clean comments a little.

24 years agoGetaddrinfo(), getnameinfo(), and etc support in libc/net.
Yoshinobu Inoue [Tue, 28 Dec 1999 02:37:14 +0000 (02:37 +0000)]
Getaddrinfo(), getnameinfo(), and etc support in libc/net.
Several udp and raw apps IPv6 support.

Reviewed by: freebsd-arch, cvs-committers
Obtained from: KAME project

24 years agoThis commit adds device driver support for the ADMtek AN986 Pegasus
Bill Paul [Tue, 28 Dec 1999 02:01:18 +0000 (02:01 +0000)]
This commit adds device driver support for the ADMtek AN986 Pegasus
USB ethernet chip. Adapters that use this chip include the LinkSys
USB100TX. There are a few others, but I'm not certain of their
availability in the U.S. I used an ADMtek eval board for development.
Note that while the ADMtek chip is a 100Mbps device, you can't really
get 100Mbps speeds over USB. Regardless, this driver uses miibus to
allow speed and duplex mode selection as well as autonegotiation.
Building and kldloading the driver as a module is also supported.

Note that in order to make this driver work, I had to make what some
may consider an ugly hack to sys/dev/usb/usbdi.c. The usbd_transfer()
function will use tsleep() for synchronous transfers that don't complete
right away. This is a problem since there are times when we need to
do sync transfers from an interrupt context (i.e. when reading registers
from the MAC via the control endpoint), where tsleep() us a no-no.
My hack allows the driver to have the code poll for transfer completion
subject to the xfer->timeout timeout rather that calling tsleep().
This hack is controlled by a quirk entry and is only enabled for the
ADMtek device.

Now, I'm sure there are a few of you out there ready to jump on me
and suggest some other approach that doesn't involve a busy wait. The
only solution that might work is to handle the interrupts in a kernel
thread, where you may have something resembling a process context that
makes it okay to tsleep(). This is lovely, except we don't have any
mechanism like that now, and I'm not about to implement such a thing
myself since it's beyond the scope of driver development. (Translation:
I'll be damned if I know how to do it.) If FreeBSD ever aquires such
a mechanism, I'll be glad to revisit the driver to take advantage of
it. In the meantime, I settled for what I perceived to be the solution
that involved the least amount of code changes. In general, the hit
is pretty light.

Also note that my only USB test box has a UHCI controller: I haven't
I don't have a machine with an OHCI controller available.

Highlights:

- Updated usb_quirks.* to add UQ_NO_TSLEEP quirk for ADMtek part.
- Updated usbdevs and regenerated generated files
- Updated HARDWARE.TXT and RELNOTES.TXT files
- Updated sysinstall/device.c and userconfig.c
- Updated kernel configs -- device aue0 is commented out by default
- Updated /sys/conf/files
- Added new kld module directory

24 years agoTwo nits and disable isa probe due to its overly agressive claiming of
Warner Losh [Tue, 28 Dec 1999 01:07:16 +0000 (01:07 +0000)]
Two nits and disable isa probe due to its overly agressive claiming of
devices.
o Return ENXIO from sn_isa_probe
o Fix SN_DEBUG printf
o Use IFQ_MAXLEN rather than 8

I'll fix the isa probe when I get access to a real isa attachment
device to test against here in a few days.

Overly agressive snagging behavior noticed by: phk

24 years agoCorrectly maintain state when manipulating linked lists. This fixes a
Tim Vanderhoek [Mon, 27 Dec 1999 20:53:35 +0000 (20:53 +0000)]
Correctly maintain state when manipulating linked lists.  This fixes a
bug that prevented the line-number buffer from working correctly.  AFAIK
the bug is still present in other derivatives of more(1).

24 years agoAdd ipsec_esp option to files which depend on crypto.
Yoshinobu Inoue [Mon, 27 Dec 1999 18:53:24 +0000 (18:53 +0000)]
Add ipsec_esp option to files which depend on crypto.
Now you can build a kernel which support IPsec message authentication
but don't support message encryption, by defining IPSEC in your kernel
config file and not defining IPSEC_ESP.

24 years agoInitialize unitialized variable from prev. commit.
Tim Vanderhoek [Mon, 27 Dec 1999 17:33:25 +0000 (17:33 +0000)]
Initialize unitialized variable from prev. commit.

24 years agoFixed breakage of static linkage of rlogind. Someone added -lutil and
Bruce Evans [Mon, 27 Dec 1999 17:25:40 +0000 (17:25 +0000)]
Fixed breakage of static linkage of rlogind.  Someone added -lutil and
-lcom_err to some libpam modules without updating LIBPAM here.

24 years agoFixed some bugs (one serious one: timeouts were claimed to be executed at
Bruce Evans [Mon, 27 Dec 1999 16:22:20 +0000 (16:22 +0000)]
Fixed some bugs (one serious one: timeouts were claimed to be executed at
spl0) and some bitrot (the not-so-new callout_init/stop/reset functions
were not mentioned; the callout_activate/deactivate/pending macros are
still not mentioned).

Submitted by: mostly by jlemon

24 years agoFix typo "," vs ";"
Peter Wemm [Mon, 27 Dec 1999 16:03:38 +0000 (16:03 +0000)]
Fix typo "," vs ";"

PR: 15696
Submitted by: Takashi Okumura <taka@cs.pitt.edu>

24 years agoFixed namespace pollution in rev.1.24 (don't implement <sys/signal.h> here).
Bruce Evans [Mon, 27 Dec 1999 15:54:52 +0000 (15:54 +0000)]
Fixed namespace pollution in rev.1.24 (don't implement <sys/signal.h> here).
Fixed long lines.

24 years agoSynced with sys/isa/sio.c rev 1.282.
KATO Takenori [Mon, 27 Dec 1999 14:01:06 +0000 (14:01 +0000)]
Synced with sys/isa/sio.c rev 1.282.

24 years agoSynced with sys/i386/isa/clock.c rev 1.148. This is a cosmetic change
KATO Takenori [Mon, 27 Dec 1999 13:56:54 +0000 (13:56 +0000)]
Synced with sys/i386/isa/clock.c rev 1.148.  This is a cosmetic change
because PC-98 doesn't have RTC and RTC related code is included by
`#ifndef PC98' and `#endif'.

24 years agoSynced with sys/i386/conf/Makefile.i386 rev 1.169.
KATO Takenori [Mon, 27 Dec 1999 13:53:43 +0000 (13:53 +0000)]
Synced with sys/i386/conf/Makefile.i386 rev 1.169.

24 years agoSynced with Makefile.i386. The following cleanups in Makefile.i386
Bruce Evans [Mon, 27 Dec 1999 12:06:11 +0000 (12:06 +0000)]
Synced with Makefile.i386.  The following cleanups in Makefile.i386
rev.1.168 should have been committed concurrently:

Fixed some style bugs (always use precisely 1 space after `:' in
dependency specifications).
Removed bogus dependency of ${FULLKERNEL} on ${BEFORE_DEPEND}.

Reminded by: peter

24 years agoAdd a bunch of `const's and fix a typo.
Brian Somers [Mon, 27 Dec 1999 11:54:57 +0000 (11:54 +0000)]
Add a bunch of `const's and fix a typo.

Submitted by: Rich Neswold <rneswold@MCS.Net>

24 years agoDon't allowt '#' as a comment when it's embedded in quotes:
Brian Somers [Mon, 27 Dec 1999 11:43:31 +0000 (11:43 +0000)]
Don't allowt '#' as a comment when it's embedded in quotes:

  set something "xxx yyy # zzz" aaa

shouldn't be interpreted as

  set something "xxx yyy" aaa

24 years agoFix some typo's.
Jeroen Ruigrok van der Werven [Mon, 27 Dec 1999 11:02:43 +0000 (11:02 +0000)]
Fix some typo's.

PR: 15678
Submitted by: Christian Weisgerber <naddy@mips.rhein-neckar.de>

24 years agoThis should have been committed with related changes to .c files.
Bruce Evans [Mon, 27 Dec 1999 10:47:32 +0000 (10:47 +0000)]
This should have been committed with related changes to .c files.

Changed the type used to represent the user stack pointer from `long *'
to `register_t *'.  This fixes bugs like misplacement of argc and argv
on the user stack on i386's with 64-bit longs.  We still use longs to
represent "words" like argc and argv, and assume that they are on the
stack (and that there is stack).  The suword() and fuword() families
should also use register_t.

24 years agoChanged the type used to represent the user stack pointer from `long *'
Bruce Evans [Mon, 27 Dec 1999 10:42:55 +0000 (10:42 +0000)]
Changed the type used to represent the user stack pointer from `long *'
to `register_t *'.  This fixes bugs like misplacement of argc and argv
on the user stack on i386's with 64-bit longs.  We still use longs to
represent "words" like argc and argv, and assume that they are on the
stack (and that there is stack).  The suword() and fuword() families
should also use register_t.

24 years agoFixed some type mismatches. p_retval[0] in struct proc has type
Bruce Evans [Mon, 27 Dec 1999 10:22:09 +0000 (10:22 +0000)]
Fixed some type mismatches.  p_retval[0] in struct proc has type
register_t, so pointers to it must be passed around as `register_t *',
not as `int *'.  The type mismatches were non-benign on alphas, but
the broken code is normally only configured by LINT.

24 years agoConnect up the bootforth glue and compile it, but don't initialize it
Peter Wemm [Mon, 27 Dec 1999 09:20:33 +0000 (09:20 +0000)]
Connect up the bootforth glue and compile it, but don't initialize it
at runtime as it has a nasty habit of crashing on the Alpha :-(.
This is being done this way so we have a common starting point for
debugging.

24 years agoAdd in missing ENABLE TARGET MODE opcode.
Matt Jacob [Mon, 27 Dec 1999 08:47:11 +0000 (08:47 +0000)]
Add in missing ENABLE TARGET MODE opcode.

24 years agoTidy up some loose ends. nullfs_read/write were returning the wrong value.
Peter Wemm [Mon, 27 Dec 1999 08:45:14 +0000 (08:45 +0000)]
Tidy up some loose ends.  nullfs_read/write were returning the wrong value.
Fix some ctype problems - isascii() caused a warning if fed an unsigned
char - it's always > 0 and libstand is compiled with -Wall.
Missing prototype/include in printf.c

24 years agoMake this compile with -Wall -Werror
Peter Wemm [Mon, 27 Dec 1999 08:40:40 +0000 (08:40 +0000)]
Make this compile with -Wall -Werror

24 years agoReset the Zoom 56K modem when starting to use it.
Julian Elischer [Mon, 27 Dec 1999 08:39:29 +0000 (08:39 +0000)]
Reset the Zoom 56K modem when starting to use it.
Failing to do so freezes your system.

Submitted by:Sean O'Connell <sean@stat.Duke.EDU>  (my hero)

24 years agokvm_mkdb(8) no longer exists (in .Xr) and kvm_nlist(8) has even less to do
Peter Wemm [Mon, 27 Dec 1999 08:01:11 +0000 (08:01 +0000)]
kvm_mkdb(8) no longer exists (in .Xr) and kvm_nlist(8) has even less to do
with dev_mkdb(8) than it did before.

24 years agoInclude opt_nfs.h
Peter Wemm [Mon, 27 Dec 1999 07:51:05 +0000 (07:51 +0000)]
Include opt_nfs.h

PR: 15711
Submitted by: Wilko Bulte <wilko@yedi.iaf.nl>

24 years agoRemove kvm_mkdb as it isn't installed anymore. libkvm asks directly
Peter Wemm [Mon, 27 Dec 1999 07:43:07 +0000 (07:43 +0000)]
Remove kvm_mkdb as it isn't installed anymore.  libkvm asks directly
via kldsym(2), and crash dumps use nlist directly (and always have).

24 years agoZap kvm_mkdb, it was for kvm_nlist's benefit, but now it goes direct
Peter Wemm [Mon, 27 Dec 1999 07:27:50 +0000 (07:27 +0000)]
Zap kvm_mkdb, it was for kvm_nlist's benefit, but now it goes direct
to the in-kernel hashed symbol tables (including modules).

24 years agoZap the kvm_kernel.db juggling at 'make install' time, it isn't needed
Peter Wemm [Mon, 27 Dec 1999 07:19:45 +0000 (07:19 +0000)]
Zap the kvm_kernel.db juggling at 'make install' time, it isn't needed
any more.

24 years agoZap kvm_kernel.db stuff now that libkvm gets it directly from the running
Peter Wemm [Mon, 27 Dec 1999 07:17:51 +0000 (07:17 +0000)]
Zap kvm_kernel.db stuff now that libkvm gets it directly from the running
kernel.

24 years agoUse kldsym(2) to lookup symbol values. This avoids the kvm_mkdb juggling
Peter Wemm [Mon, 27 Dec 1999 07:14:58 +0000 (07:14 +0000)]
Use kldsym(2) to lookup symbol values.  This avoids the kvm_mkdb juggling
and is module aware.  Yes, this means that kvm_nlist(3) will find symbols
in loaded modules.  The emulation of the nlist struct is pretty crude but
seems to work well enough for all the users in the tree that I found.

24 years agoCorrect an uninitialized variable use, which, unlike most times, is
Brian Feldman [Mon, 27 Dec 1999 06:31:53 +0000 (06:31 +0000)]
Correct an uninitialized variable use, which, unlike most times, is
actually a bug this time.

Submitted by: bde
Reviewed by: bde

24 years agoRecognize the GVC0505 (GVC 56k Faxmodem) as a sio device.
Peter Wemm [Mon, 27 Dec 1999 05:02:25 +0000 (05:02 +0000)]
Recognize the GVC0505 (GVC 56k Faxmodem) as a sio device.

Obtained from: Dan J Fraser <dfraser@capybara.org> (for NetBSD)

24 years agoAdd a new function dllockinit() for registering thread locking
John Polstra [Mon, 27 Dec 1999 04:44:04 +0000 (04:44 +0000)]
Add a new function dllockinit() for registering thread locking
functions to be used by the dynamic linker.  This can be called by
threads packages at start-up time.  I will add the call to libc_r
soon.

Also add a default locking method that is used up until dllockinit()
is called.  The default method works by blocking SIGVTALRM, SIGPROF,
and SIGALRM in critical sections.  It is based on the observation
that most user-space threads packages implement thread preemption
with one of these signals (usually SIGVTALRM).

The dynamic linker has never been reentrant, but it became less
reentrant in revision 1.34 of "src/libexec/rtld-elf/rtld.c".
Starting with that revision, multiple threads each doing lazy
binding could interfere with each other.  The usual symptom was
that a symbol was falsely reported as undefined at start-up time.
It was rare but not unseen.  This commit fixes it.

24 years agoAdd support of SB for PC98 into VoxWare 3.5, and more $FreeBSD$.
Seigo Tanimura [Mon, 27 Dec 1999 04:37:19 +0000 (04:37 +0000)]
Add support of SB for PC98 into VoxWare 3.5, and more $FreeBSD$.

Submitted by: T.Yamaoka <taka@windows.squares.net>
Pressed to review by: nyan

24 years agoFixed stripping of aout debugging kernels.
Bruce Evans [Sun, 26 Dec 1999 17:07:15 +0000 (17:07 +0000)]
Fixed stripping of aout debugging kernels.
Fixed some style bugs (always use precisely 1 space after `:' in
dependency specifications).
Removed bogus dependency of ${FULLKERNEL} on ${BEFORE_DEPEND}.

24 years agoFixed spelling error in document title.
Bruce Evans [Sun, 26 Dec 1999 16:58:31 +0000 (16:58 +0000)]
Fixed spelling error in document title.
Fixed syntax error in synopsis.

24 years agoFix tab completion mounts (like /cdrom/)
Eivind Eklund [Sun, 26 Dec 1999 16:34:53 +0000 (16:34 +0000)]
Fix tab completion mounts (like /cdrom/)

Submitted by: Martin Blapp <mb@imp.ch>

24 years agoDon't include <isa/isavar.h> or compile code depending on it when isa
Bruce Evans [Sun, 26 Dec 1999 16:21:19 +0000 (16:21 +0000)]
Don't include <isa/isavar.h> or compile code depending on it when isa
is not configured.  Including <isa/isavar.h> when it is not used is
harmful as well as bogus, since it includes "isa_if.h" which is not
generated when isa is not configured.

24 years agoMoved $FreeBSD$ to its usual place after the CSRG id.
Bruce Evans [Sun, 26 Dec 1999 15:24:58 +0000 (15:24 +0000)]
Moved $FreeBSD$ to its usual place after the CSRG id.

Fixed some style bugs for cam (superflous umask and missing newline).

Fixed bogons for apm.  The pattern "apm*" matched too many things; apm
and apmctl were both made twice by `MAKEDEV all'.  Hopefully no one
depends on `MAKEDEV apm0' making apm or on `MAKEDEV apm' making apmctl.

24 years agoUpdate the cvs pserver example so that it gives some more obvious clues
Peter Wemm [Sun, 26 Dec 1999 15:18:58 +0000 (15:18 +0000)]
Update the cvs pserver example so that it gives some more obvious clues
about the --allow-root switch.

PR: 14463

24 years agoUpdated a comment to match code.
Bruce Evans [Sun, 26 Dec 1999 14:16:00 +0000 (14:16 +0000)]
Updated a comment to match code.

24 years agoRemoved unused includes.
Bruce Evans [Sun, 26 Dec 1999 14:07:43 +0000 (14:07 +0000)]
Removed unused includes.

Rumoved unused compatibility cruft for dup().  Using it today would just
break dup() on fd's >= 64.

Fixed some style bugs.

24 years agoUse vfs_timestamp() instead of getnanotime() to set timestamps. This
Bruce Evans [Sun, 26 Dec 1999 13:04:52 +0000 (13:04 +0000)]
Use vfs_timestamp() instead of getnanotime() to set timestamps.  This
fixee incoherency of pipe timestamps relative to file timestamps in
the usual case where getnanotime() is not used for the latter.  (File
and pipe timestamps are still incoherent relative to real time unless
the vfs_timestamp_precision sysctl is set to 2 or 3).

24 years agoReplaced the INTRMASK and INTRUNMASK macros by "|" and "&~" operations.
Bruce Evans [Sun, 26 Dec 1999 12:43:47 +0000 (12:43 +0000)]
Replaced the INTRMASK and INTRUNMASK macros by "|" and "&~" operations.
Some interface botches went away, leaving the macros unused outside of
the implementation of interrupt masking, and it was silly for the
implementation to use the macros in only one place each.

24 years agoFixed breakage of read-only opening of /dev/*mem at securelevel > 0 in
Bruce Evans [Sun, 26 Dec 1999 11:44:45 +0000 (11:44 +0000)]
Fixed breakage of read-only opening of /dev/*mem at securelevel > 0 in
previous pair of commits.

Spell the "securelevel > 0" check consistently.

Use the proc arg instead of curproc in mmopen() and mmclose().

24 years agoFix a mistake in the PNP EISA-encoding of the Avance ALS120 id.
Peter Wemm [Sun, 26 Dec 1999 10:51:47 +0000 (10:51 +0000)]
Fix a mistake in the PNP EISA-encoding of the Avance ALS120 id.

Submitted by: Bryan Liesner <bleez@netaxs.com>

24 years agoRemoved ex script left over from incomplete backout in previous commit.
Bruce Evans [Sun, 26 Dec 1999 10:31:29 +0000 (10:31 +0000)]
Removed ex script left over from incomplete backout in previous commit.

Use SYMLINKS instead of an ad hoc rule for installing words -> web2.
Don't override the install target; doing so just breaks things like
SYMLINKS.

Don't override the correct defaults for the all, clean, depend, lint
and tags targets.  Don't add a null rule to the cleandepend target.

24 years agoOops, deactivate ed drivers because of undefiend references from
KATO Takenori [Sun, 26 Dec 1999 05:26:43 +0000 (05:26 +0000)]
Oops, deactivate ed drivers because of undefiend references from
if_ed_pci.o.

24 years agoAdded Allied Telesis SIU-98-D support.
KATO Takenori [Sun, 26 Dec 1999 05:10:14 +0000 (05:10 +0000)]
Added Allied Telesis SIU-98-D support.

Submitted by: Isizu Takaaki <isizu-t01@aso-group.co.jp>
chi@bd.mbn.or.jp (Chiharu Shibata)
(w/ minor change by kato)

24 years agoA missing verb to a sentence from last commit.
Tim Vanderhoek [Sun, 26 Dec 1999 04:41:29 +0000 (04:41 +0000)]
A missing verb to a sentence from last commit.

24 years ago- Cut down amount of memory in 64MB when BIOS tells the amount of
KATO Takenori [Sun, 26 Dec 1999 04:22:27 +0000 (04:22 +0000)]
- Cut down amount of memory in 64MB when BIOS tells the amount of
  memory >= 64MB.
- Don't perform destructive memory inspection for 15 - 16MB system
  area.

Submitted by: NOKUBI Hirotaka <hnokubi@yyy.or.jp>
chi@bd.mbn.or.jp (Chiharu Shibata)

24 years agoFor some reason the HOME key wasn't documented. Document it. Expand
Tim Vanderhoek [Sun, 26 Dec 1999 04:17:27 +0000 (04:17 +0000)]
For some reason the HOME key wasn't documented.  Document it.  Expand
the docs on a couple other keys.  While I'm here, document another ~3 bugs
that have been around for all eternity in the hope that I'll someday bother
to fix them.

24 years agoAllow excessive backspacing to correctly abort an input (most significantly
Tim Vanderhoek [Sun, 26 Dec 1999 03:03:04 +0000 (03:03 +0000)]
Allow excessive backspacing to correctly abort an input (most significantly
a search string input).

24 years agoExpand source comments a little -=> increased readability.
Tim Vanderhoek [Sun, 26 Dec 1999 02:39:26 +0000 (02:39 +0000)]
Expand source comments a little -=> increased readability.

24 years agoThis commit was generated by cvs2svn to compensate for changes in r55099,
Kris Kennaway [Sat, 25 Dec 1999 16:37:36 +0000 (16:37 +0000)]
This commit was generated by cvs2svn to compensate for changes in r55099,
which included commits to RCS files with non-trunk default branches.

24 years agoInitial import of OpenSSL v0.9.4
Kris Kennaway [Sat, 25 Dec 1999 16:37:36 +0000 (16:37 +0000)]
Initial import of OpenSSL v0.9.4

24 years agoFixed races accessing the RTC. The races apparently caused
Bruce Evans [Sat, 25 Dec 1999 15:30:31 +0000 (15:30 +0000)]
Fixed races accessing the RTC.  The races apparently caused
apm_default_resume() to sometimes set a very wrong time.
(1) Accesses to the RTC index and data registers were not atomic enough.
    Interrupts were not masked.  This was only good enough until an
    interrupt handler (rtcintr()) started accessing the RTC in FreeBSD-2.0.
(2) Access to the block of time registers in inittodr() was not atomic
    enough.  inittodr() has 244us to read the time registers.  Interrupts
    were not masked.  This was only good enough until something (apm)
    started calling inittodr() after boot time in FreeBSD-2.0.
The fix for (2) also makes the timecounter update more atomic, although
this is currently unimportant due to the low resolution of the RTC.

Problem reported by: mckay

24 years agoAdd directories: preparation for OpenSSL import.
Kris Kennaway [Sat, 25 Dec 1999 14:12:09 +0000 (14:12 +0000)]
Add directories: preparation for OpenSSL import.

24 years agoSacrifice, for now, NFS and NETGRAPH on the alpha boot floppy
Jordan K. Hubbard [Sat, 25 Dec 1999 07:22:10 +0000 (07:22 +0000)]
Sacrifice, for now, NFS and NETGRAPH on the alpha boot floppy
so that we can get it back onto 1.44MB boot media again.  This
does not affect the i386 class build.

24 years agoBackout rev1.7, as it broke adventure(6) (const'ing a non-r/o variable).
Bill Fumerola [Sat, 25 Dec 1999 03:50:42 +0000 (03:50 +0000)]
Backout rev1.7, as it broke adventure(6) (const'ing a non-r/o variable).
As wpaul notes, I was trying to fix something that wasn't broke.

24 years agoFix the firmware build number output (again).
Mike Smith [Sat, 25 Dec 1999 01:45:22 +0000 (01:45 +0000)]
Fix the firmware build number output (again).

24 years agoDon't pass u_int32_t pointers to BUS_READ_IVAR since it tends to make
Doug Rabson [Fri, 24 Dec 1999 16:22:45 +0000 (16:22 +0000)]
Don't pass u_int32_t pointers to BUS_READ_IVAR since it tends to make
alphas panic.

24 years ago* Set the devclass of a device before calling the probe method. This allows
Doug Rabson [Fri, 24 Dec 1999 16:21:15 +0000 (16:21 +0000)]
* Set the devclass of a device before calling the probe method. This allows
  device_printf() etc. to print something intelligible.
* Allow device_set_devclass(dev, 0) for clearing the devclass.

24 years agoFixed a cast of a pointer to an integer of a possibly different size.
Bruce Evans [Fri, 24 Dec 1999 15:33:36 +0000 (15:33 +0000)]
Fixed a cast of a pointer to an integer of a possibly different size.
Fixed casts of non-`void *' pointers to uintptr_t.  Fixed related
style bugs.  This file uses perfectly non-KNF formatting for casts.

24 years agoSince we build and install gasp(1) in /usr/libexec/${OBJFORMAT}/,
Brian Feldman [Fri, 24 Dec 1999 15:18:19 +0000 (15:18 +0000)]
Since we build and install gasp(1) in /usr/libexec/${OBJFORMAT}/,
and it is a user-run utility, it should be in objformat, so gasp
is in the ${PATH} :)

24 years agoRemoved unnecessary const poisoning (redundant casting) which was added
Bruce Evans [Fri, 24 Dec 1999 15:16:10 +0000 (15:16 +0000)]
Removed unnecessary const poisoning (redundant casting) which was added
in rev.1.4.

24 years agoMerge from sys/i386/conf/files.i386 rev 1.293.
KATO Takenori [Fri, 24 Dec 1999 11:20:14 +0000 (11:20 +0000)]
Merge from sys/i386/conf/files.i386 rev 1.293.

24 years agoMerge from sys/i386/conf/GENERIC rev 1.218 & 1.219.
KATO Takenori [Fri, 24 Dec 1999 11:12:43 +0000 (11:12 +0000)]
Merge from sys/i386/conf/GENERIC rev 1.218 & 1.219.

24 years agoRemoved -mno-486 from CFLAGS.
KATO Takenori [Fri, 24 Dec 1999 11:09:40 +0000 (11:09 +0000)]
Removed -mno-486 from CFLAGS.

24 years agoMandate the inclusion of <ieeefp.h> instead of <floatingpoint.h>. The
Sheldon Hearn [Fri, 24 Dec 1999 08:43:19 +0000 (08:43 +0000)]
Mandate the inclusion of <ieeefp.h> instead of <floatingpoint.h>.  The
latter is just wrong.

There were many other problems spotted by bde, none of which are
addressed here.

Reviewed by: bde