unix-history/.git
30 years agoFixed up a compiler warning.
Andreas Schulz [Sat, 4 Dec 1993 01:53:16 +0000 (01:53 +0000)]
Fixed up a compiler warning.

30 years agoMore changes to bring FreeBSD in sync with Paul K's latest.
Jordan K. Hubbard [Sat, 4 Dec 1993 00:53:16 +0000 (00:53 +0000)]
More changes to bring FreeBSD in sync with Paul K's latest.

30 years agoMore changes to bring FreBSD in sync with Paul K's latest.
Jordan K. Hubbard [Sat, 4 Dec 1993 00:53:02 +0000 (00:53 +0000)]
More changes to bring FreBSD in sync with Paul K's latest.

30 years agoFixed up a compiler warning about a misplaced external reference.
Andreas Schulz [Fri, 3 Dec 1993 23:16:59 +0000 (23:16 +0000)]
Fixed up a compiler warning about a misplaced external reference.

30 years agoFrom: Jeffrey Hsu <hsu@soda.berkeley.edu>
Andrew Moore [Fri, 3 Dec 1993 05:10:08 +0000 (05:10 +0000)]
From: Jeffrey Hsu <hsu@soda.berkeley.edu>

The following patch adds the addr argument to signal handlers.

The kernel with the patch is no more and no less in compliance or in
violation of POSIX and ANSI C than the kernel before the patch.

The added functionality this addr argument provides is quite useful.  It
enables an entire class of algorithms which use mprotect to trace memory
references.  Beside garbage collectors, I have heard of this technique being
applied to debuggers and profilers.  The only benchmarking I've performed is
using akcl to compile maxima:  without the kernel patch, it takes 7 hours to
compile maxima, while with stratified garbage collection, it only takes 50
minutes.

Basically, I can't think of a reason not to add the addr argument and there
is a compelling need for it.

If you find the patch acceptable, please let me know so I can send my
FreeBSD akcl config files to wfs for inclusion in the core akcl release.
The old 386BSD config files there won't work on either NetBSD or FreeBSD.

30 years agoReset fdc during probe.
Andrew Moore [Fri, 3 Dec 1993 05:01:40 +0000 (05:01 +0000)]
Reset fdc during probe.

From: <dec@lazarus.nrtc.northrop.com>
   Changes between EPSILON and RELEASE of FreeBSD have again caused
the kernel to not see my floppy disk drives.  I don't know what happened,
'cause I don't see any changes to fd.c, but here is an old fix that
I have applied to the probe routine which will solve the problem (at
least for me).  Since this is a rather brute-force solution - I understand
if you want to ignore it...

[Upgrading to pre-Beta FreeBSD caused this on my system. -AM]

30 years agoBacked out the change from read to fread. lex fails with this change
Andreas Schulz [Fri, 3 Dec 1993 00:05:27 +0000 (00:05 +0000)]
Backed out the change from read to fread. lex fails with this change
in nslookup. nslookup simply stops working and i don't find the
error for now. So, i think, it is better in the moment to go to
the previous version until someone corrects nslookup. Our
sourcetree should be compilable and runnable with our utilities.

30 years agoFixed up all compiler warnings.
Andreas Schulz [Thu, 2 Dec 1993 16:24:44 +0000 (16:24 +0000)]
Fixed up all compiler warnings.

30 years agoSorry, Jordan, but I restore previous version again.
Andrey Chernov [Thu, 2 Dec 1993 09:56:30 +0000 (09:56 +0000)]
Sorry, Jordan, but I restore previous version again.

You write that
LDFLAGS+= -Xlinker -Bstatic
no more needed, but you have
CFLAGS += -static -I$(.CURDIR) -I$(.CURDIR)/$(MACHINE)
This -static doesn't make any sense in your version,
because it not passed to linker's state (LDFLAGS),
so we have SHARED /usr/bin/ld in this case.
(Older Makefile produce non-shared ld).
I suppose, that -static in CFLAGS was introduced to
make non-shared ld, so I restore previous LDFLAGS
to have non-shared ld. If we want to have shared ld
we need to remove -static from CFLAGS too, not only LDFLAGS,
but this need special issue and corresponding
commit log. Your current version hang into intermediate
state (beetween two sides), so I move it to one side.

Second, I restore NOPIC dependance again from older Makefile:
.if !defined(NOPIC)
SUBDIR+= rtld
.endif
We don't need ld.so, if NOPIC

P.S. I don't see any purpose to commit new makefile, old version
is better.

30 years ago#ifdef DEBUG unwanted message, third time after Jordan...
Andrey Chernov [Thu, 2 Dec 1993 09:32:26 +0000 (09:32 +0000)]
#ifdef DEBUG unwanted message, third time after Jordan...

30 years agoMoved DEBUG conditional where it belonged to remove DEBUG code out of
Nate Willams [Thu, 2 Dec 1993 05:09:52 +0000 (05:09 +0000)]
Moved DEBUG conditional where it belonged to remove DEBUG code out of
default compilation.

30 years agoPatch from Andy Heffernan <ahh@netcom.com>
David Greenman [Thu, 2 Dec 1993 02:48:15 +0000 (02:48 +0000)]
Patch from Andy Heffernan <ahh@netcom.com>
Description from Sean Eric Fagan:

This has never bitten me, but someone (whose name I forget, but I do
thank) pointed out it was wrong.  He was right, so here is a fix.
(Basicly, the kernel's address of the temporary page gets set by
one of the vm* routines, so I didn't bother initializing it, but it
does get checked by the vm* routine before it sets it, so it should be
initialized.  Zero seems a reasonable value.)  (Oh, and if whomever-

30 years agoSecond attempt to integrate Paul K's changes.
Jordan K. Hubbard [Thu, 2 Dec 1993 01:03:47 +0000 (01:03 +0000)]
Second attempt to integrate Paul K's changes.

30 years agoSecond attempt to integrate Paul K's changes.
Jordan K. Hubbard [Thu, 2 Dec 1993 00:56:40 +0000 (00:56 +0000)]
Second attempt to integrate Paul K's changes.

30 years agoFixed two forward declarations.
Andreas Schulz [Wed, 1 Dec 1993 19:18:34 +0000 (19:18 +0000)]
Fixed two forward declarations.

30 years agoAdded a definition for -p1003.2-92, that was used in some man pages
Andreas Schulz [Wed, 1 Dec 1993 18:38:33 +0000 (18:38 +0000)]
Added a definition for -p1003.2-92, that was used in some man pages
and had resulted in no output.

30 years agodeleted all references to rindex and included string.h instead.
Andreas Schulz [Wed, 1 Dec 1993 16:45:42 +0000 (16:45 +0000)]
deleted all references to rindex and included string.h instead.

30 years agochanged the order of the includes and deleted a reference to index
Andreas Schulz [Wed, 1 Dec 1993 16:44:43 +0000 (16:44 +0000)]
changed the order of the includes and deleted a reference to index
nd added an include to string.h instead

30 years agoAll -static and -Xlinker -Bstatic flags missed in
Andrey Chernov [Wed, 1 Dec 1993 15:58:22 +0000 (15:58 +0000)]
All -static and -Xlinker -Bstatic flags missed in
newly commited Makefile.
NOPIC dependance missed too!
Oh it is really pain to check all that broken stuff,
Please, check it *before* commit!

30 years ago#ifdef DEBUG missed by several lines!
Andrey Chernov [Wed, 1 Dec 1993 15:34:01 +0000 (15:34 +0000)]
#ifdef DEBUG missed by several lines!
!!! Please, COMPILE *BEFORE* COMMIT!

30 years agoAdd DEBUG #ifdef, (second time!)
Andrey Chernov [Wed, 1 Dec 1993 15:13:42 +0000 (15:13 +0000)]
Add DEBUG #ifdef, (second time!)
!!! Please, anybody who update ld to NetBSD current,
!!! look at old cvs logs!

30 years agoRemove hanging #else part (with second sbrk()) without #if & #endif
Andrey Chernov [Wed, 1 Dec 1993 15:05:31 +0000 (15:05 +0000)]
Remove hanging #else part (with second sbrk()) without #if & #endif
Add missing argument to findshlib, I am not shure,
but it seems that it is 1.
!!! Does anybody compile it before commit?

30 years agoRemove -g option (second time!)
Andrey Chernov [Wed, 1 Dec 1993 14:45:35 +0000 (14:45 +0000)]
Remove -g option (second time!)
!!! Please anybody who port Makefiles from NetBSD,
!!! remove -g option!

30 years agoFrom: Julian Howard Stacey <stacey@guug.de>
Rod Grimes [Wed, 1 Dec 1993 08:44:39 +0000 (08:44 +0000)]
From: Julian Howard Stacey <stacey@guug.de>
Subject: Bug & Fix for etc/Makefile cpio-floppy: re /tmp creation.
Date: Fri, 26 Nov 1993 11:35:04 +0100

Editors Note:  tmp was listed in the CPIO_FILES section and thus the
entire contents of ${DESTDIR}/tmp would end up on the cpio floppy.  This fix
moves tmp to CPIO_DIRS so that no longer happens.

30 years agoFixed the error, that makes g++ fail in compiling groff/libbib/index.cc.
Andreas Schulz [Tue, 30 Nov 1993 21:54:25 +0000 (21:54 +0000)]
Fixed the error, that makes g++ fail in compiling groff/libbib/index.cc.
I don't understand why g++ uses the struct buf definition from types.h
instead of the local char *buf definition, but it seems to happen. Error
in scope handling in g++ ?

30 years agoMore recent changes from Paul K.
Jordan K. Hubbard [Tue, 30 Nov 1993 21:04:34 +0000 (21:04 +0000)]
More recent changes from Paul K.

30 years agoLatest from Paul K. for better checking of PIC code.
Jordan K. Hubbard [Tue, 30 Nov 1993 20:57:41 +0000 (20:57 +0000)]
Latest from Paul K. for better checking of PIC code.

30 years agoMany recent fixes from Paul K, add support for chaining of shared lib deps.
Jordan K. Hubbard [Tue, 30 Nov 1993 20:47:54 +0000 (20:47 +0000)]
Many recent fixes from Paul K, add support for chaining of shared lib deps.

30 years agoForgot half of the changes for CPUSTATES and DK_NDRIVE, sorry.
Andreas Schulz [Tue, 30 Nov 1993 20:31:41 +0000 (20:31 +0000)]
Forgot half of the changes for CPUSTATES and DK_NDRIVE, sorry.

30 years agoShut up compiler warnings. Fixed some lazy declarations.
Andreas Schulz [Tue, 30 Nov 1993 19:52:11 +0000 (19:52 +0000)]
Shut up compiler warnings. Fixed some lazy declarations.

30 years agoChanged the definition of CPUSTATES and DK_NDRIVE to RSTAT_CPUSTATES and
Andreas Schulz [Tue, 30 Nov 1993 19:40:06 +0000 (19:40 +0000)]
Changed the definition of CPUSTATES and DK_NDRIVE to RSTAT_CPUSTATES and
RSTAT_DK_NDRIVE to avoid nameclashes.

30 years agoallow spaces in $sendmail_flags.
Paul Richards [Tue, 30 Nov 1993 16:59:27 +0000 (16:59 +0000)]
allow spaces in $sendmail_flags.

30 years agoDeleted the prototype of time() and included <time.h> instead.
Andreas Schulz [Tue, 30 Nov 1993 14:51:31 +0000 (14:51 +0000)]
Deleted the prototype of time() and included <time.h> instead.

30 years agoFixed a prototype of a function to match its actual declaration.
Andreas Schulz [Tue, 30 Nov 1993 14:37:01 +0000 (14:37 +0000)]
Fixed a prototype of a function to match its actual declaration.

30 years agoChanged a prototype to match the actual declaration.
Andreas Schulz [Tue, 30 Nov 1993 14:31:44 +0000 (14:31 +0000)]
Changed a prototype to match the actual declaration.

30 years agoFixed a typo. Changed the "include <stdlib> to include <stdlib.h>.
Andreas Schulz [Tue, 30 Nov 1993 14:28:01 +0000 (14:28 +0000)]
Fixed a typo. Changed the "include <stdlib> to include <stdlib.h>.

30 years agoDeleted a wrong reference to random() and included <stdlib.h> instead.
Andreas Schulz [Tue, 30 Nov 1993 11:27:40 +0000 (11:27 +0000)]
Deleted a wrong reference to random() and included <stdlib.h> instead.

30 years agoDeleted a misplaced reference to time() and inserted an include <time.h>.
Andreas Schulz [Tue, 30 Nov 1993 10:34:39 +0000 (10:34 +0000)]
Deleted a misplaced reference to time() and inserted an include <time.h>.

30 years agoAdd support for ~ in a regexp pattern and fix handling of ~ in
Andrew Moore [Tue, 30 Nov 1993 09:39:38 +0000 (09:39 +0000)]
Add support for  ~ in a regexp pattern and fix handling of ~ in
a replacement.  In both cases, ~ expands to the replacement text of the
last substitution as per POSIX.

30 years agowmesg was too long (>7), fixed
Andrey Chernov [Mon, 29 Nov 1993 19:26:32 +0000 (19:26 +0000)]
wmesg was too long (>7), fixed

30 years agoMake fill_eproc more robust:
Andrey Chernov [Mon, 29 Nov 1993 19:22:18 +0000 (19:22 +0000)]
Make fill_eproc more robust:
always zero last byte of e_wmesg to prevent fault on
long (>7) p_wmesg

30 years agoWeird kern_kinfo bug, fix continued:
Andrey Chernov [Mon, 29 Nov 1993 18:01:12 +0000 (18:01 +0000)]
Weird kern_kinfo bug, fix continued:
p_wmesg left in undefined state after fork, so first
getkerninfo before first tsleep fails.
Fix: zero p_wmesg after copy proc structure.

30 years ago * Revision 2.16 1993/11/29 16:55:56 davidg
David Greenman [Mon, 29 Nov 1993 17:07:33 +0000 (17:07 +0000)]
 * Revision 2.16  1993/11/29  16:55:56  davidg
 * merged in Garrett Wollman's strict prototype changes
 *
 * Revision 2.15  1993/11/29  16:32:58  davidg
 * From Thomas Sandford <t.d.g.sandford@comp.brad.ac.uk>
 * Add support for the 8013W board type

30 years agoRemoving another null pointer reference.
Andreas Schulz [Mon, 29 Nov 1993 15:54:49 +0000 (15:54 +0000)]
Removing another null pointer reference.

30 years agoUse slightly less disgusting double cast to quiet compiler. This will be
David Greenman [Mon, 29 Nov 1993 15:45:44 +0000 (15:45 +0000)]
Use slightly less disgusting double cast to quiet compiler. This will be
removed when type of handle is changed.

30 years agoAdded disgusting kludge to get rid of compiler warning until 'handle'
David Greenman [Mon, 29 Nov 1993 15:30:25 +0000 (15:30 +0000)]
Added disgusting kludge to get rid of compiler warning until 'handle'
type is changed.

30 years agoWeird bug with kern_kinfo now fixed! (applauds)
Andrey Chernov [Mon, 29 Nov 1993 14:50:27 +0000 (14:50 +0000)]
Weird bug with kern_kinfo now fixed!  (applauds)
Missing argument to thread_block() caused this.
Added vm.h for prototype and add string argument to thread_block

30 years ago Fixed some bugs I introduced when integrating Mike Hibler's
David Greenman [Mon, 29 Nov 1993 14:30:24 +0000 (14:30 +0000)]
   Fixed some bugs I introduced when integrating Mike Hibler's
device_pager. The biggest problem was that no reference was being
held on the vnode associated with the device special file, and as
a result, the vp that was the handle was reused - leaving no way
to get to the major/minor anymore. Rather than gaining a reference,
I decided to go back to the previous way of doing things (i.e.
using the major/minor as the handle); this makes more sense anyway,
but is kinda kludgy. Additionally, I removed one of the #ifdef
DIAGNOSTIC conditionals so that the code that checks for a valid
device mmap routine always does the check. This can be re-ifdef'd
later if necessary, but the check is relatively low overhead and
would catch some potentially disasterous bugs.
   Also found that several fields in the vm_page structure that
is allocated in the getfake routine weren't being initialized.
This was fixed by bzeroing the allocated struct first, and then
properly initializing all of the non-zero elements.

30 years agoThe prototype for strerror is not in stdio.h , but in string.h. Checked
Andreas Schulz [Mon, 29 Nov 1993 10:13:16 +0000 (10:13 +0000)]
The prototype for strerror is not in stdio.h , but in string.h. Checked
that against solaris.

30 years agoDeleted a definition of strerror and included string.h instead.
Andreas Schulz [Mon, 29 Nov 1993 09:56:16 +0000 (09:56 +0000)]
Deleted a definition of strerror and included string.h instead.

30 years agoMoved an external function definition to a more global place, to shut
Andreas Schulz [Mon, 29 Nov 1993 09:31:46 +0000 (09:31 +0000)]
Moved an external function definition to a more global place, to shut
up the compiler.

30 years agoFixed a typo in the manpage of strerror.3 .
Andreas Schulz [Mon, 29 Nov 1993 09:05:49 +0000 (09:05 +0000)]
Fixed a typo in the manpage of strerror.3 .

30 years agoSee, what i do to keep Garett happy :-).
Andreas Schulz [Sun, 28 Nov 1993 23:43:29 +0000 (23:43 +0000)]
See, what i do to keep Garett happy :-).
Deleted the lseek definition and included unistd.h instead.

30 years agoChanged the forward definition of tskip and tdecode to match the
Andreas Schulz [Sun, 28 Nov 1993 23:31:53 +0000 (23:31 +0000)]
Changed the forward definition of tskip and tdecode to match the
actual function.

30 years agoChanged the declaration of a signal handler routine to void.
Andreas Schulz [Sun, 28 Nov 1993 23:12:25 +0000 (23:12 +0000)]
Changed the declaration of a signal handler routine to void.

30 years agocommented out a forward declaration to get rid of a compiler warning.
Andreas Schulz [Sun, 28 Nov 1993 23:00:17 +0000 (23:00 +0000)]
commented out a forward declaration to get rid of a compiler warning.

30 years agoCorrected a forward declaration to get rid of a compiler warning.
Andreas Schulz [Sun, 28 Nov 1993 22:46:26 +0000 (22:46 +0000)]
Corrected a forward declaration to get rid of a compiler warning.

30 years agoMoved a forward definition from a function into a more global place, to
Andreas Schulz [Sun, 28 Nov 1993 22:39:50 +0000 (22:39 +0000)]
Moved a forward definition from a function into a more global place, to
delete a compiler warning.

30 years agoChanged a forward declaration to match the actual function.
Andreas Schulz [Sun, 28 Nov 1993 22:28:29 +0000 (22:28 +0000)]
Changed a forward declaration to match the actual function.

30 years agoDeleted a wrong definition of calloc and included stdlib.h instead.
Andreas Schulz [Sun, 28 Nov 1993 22:21:39 +0000 (22:21 +0000)]
Deleted a wrong definition of calloc and included stdlib.h instead.

30 years agodeleted a register definition from a "buf[20]" declaration.
Andreas Schulz [Sun, 28 Nov 1993 22:07:02 +0000 (22:07 +0000)]
deleted a register definition from a "buf[20]" declaration.

30 years agoDeleted the extern wrong definition of inet_ntoa and included
Andreas Schulz [Sun, 28 Nov 1993 15:06:50 +0000 (15:06 +0000)]
Deleted the extern wrong definition of inet_ntoa and included
arpa/inet.h for the definition . Eliminates a compiler warning.

30 years agoFixed the typo of return(news) to return(new).
Andreas Schulz [Sun, 28 Nov 1993 14:37:58 +0000 (14:37 +0000)]
Fixed the typo of return(news) to return(new).

30 years agoMoved the definition from extern lseek from a local function to a more
Andreas Schulz [Sun, 28 Nov 1993 14:27:44 +0000 (14:27 +0000)]
Moved the definition from extern lseek from a local function to a more
global place to get rid of compiler warnings.

30 years agoChanged types.h that the common struct definitions are also available
Andreas Schulz [Sun, 28 Nov 1993 14:23:11 +0000 (14:23 +0000)]
Changed types.h that the common struct definitions are also available
outside of the kernel. They are needed in conf.h. Added an include of
cdefs.h in disklabel.h for the non-kernel case for the definition of __P.

30 years agoChanged the location of an external declaration to get rid of a compiler
Andreas Schulz [Sun, 28 Nov 1993 10:56:06 +0000 (10:56 +0000)]
Changed the location of an external declaration to get rid of a compiler
warning.

30 years agoDirect #undefs removed to allow flexible manipulation
Andrey Chernov [Sun, 28 Nov 1993 10:55:59 +0000 (10:55 +0000)]
Direct #undefs removed to allow flexible manipulation
via config options. Unneded #defines removed.
Improved dependance of "snd.h" and NSND.

30 years agoGet rid of some compilation warnings.
Andreas Schulz [Sun, 28 Nov 1993 10:24:52 +0000 (10:24 +0000)]
Get rid of some compilation warnings.

30 years agoPatch from Gene Stark:
David Greenman [Sun, 28 Nov 1993 09:28:54 +0000 (09:28 +0000)]
Patch from Gene Stark:

Subject: Page fault in PTE area fails in copyout
Index: sys/i386/i386/trap.c FreeBSD-1.0.2

Description:
Reading files of several megabytes into Emacs, or many small
files all at once, would fail with "IO error - bad address".

Repeat-By:
The bug can be exercised by a test program that malloc()'s
a 5MB chunk of memory, and then, without accessing the memory
first, filling it with data from a file using read().
(I read 64k chunks from /dev/wd0d into successive 64k regions
of the 5MB chunk.)  The read() will fail with EFAULT at the first
virtual address boundary that is a multiple of 0x400000.

Fix:
The problem was code in sys/i386/i386/trap.c that tries to
figure out what kind of trap occurred and to handle it appropriately.
It was interpreting any page fault with virtual address
>= vm->vm_maxsaddr as being a user stack segment fault.
In fact, addresses >= USRSTACK are in the user structure/PTE area,
and if they are handled as stack faults, the proper PTE will
not be paged in when it is supposed to be.  This situation comes
up in copyout() and copyoutstr(), if PTE's are accessed for the
first time ever.  The page fault on accessing the nonexistent PTE
is mishandled as a stack fault, and then the fault that occurs on
the subsequent access to the page itself causes copyout to fail
with EFAULT.

30 years agoPatch from Gene Stark:
David Greenman [Sun, 28 Nov 1993 09:16:07 +0000 (09:16 +0000)]
Patch from Gene Stark:
Description:
Running "strip -x 386bsd" (the last step in building a kernel)
over NFS hangs in uninterruptible sleep.

Repeat-By:
Copy an unstripped kernel to an NFS-mounted directory and
run "strip -x" on it.

Fix:
Apply the patch below to /sys/nfs/nfs_bio.c.  The problem is
that vnode_pager_io sets up a uio structure with a null
process pointer, and then nfs_write uses that pointer to
check whether the file size resource limit is exceeded.
If so, nfs_write returns EFBIG, which is ultimately voided
by the call to vm_pager_put in vm_object_page_clean.
The process thus hangs forever waiting for pageout to complete.
The code for nfs_write was evidently cribbed from ufs_write
in ufs_vnops.c.  The check for vp->v_type == VREG is actually
already done earlier in nfs_write, but why not be a little
defensive?  The problem evidently does not exist in isofs
(no writing), or pcfs (done correctly).

(Note that the check for VREG is indeed redundant, and I removed it -DG)

30 years agoMove soundcard.h & ultrasound.h to proper location <machine/...>
Andrey Chernov [Sat, 27 Nov 1993 22:16:44 +0000 (22:16 +0000)]
Move soundcard.h & ultrasound.h to proper location <machine/...>
to allow application access it.

30 years agoFrom: hsu@soda.berkeley.edu
Andrey Chernov [Sat, 27 Nov 1993 19:42:31 +0000 (19:42 +0000)]
From: hsu@soda.berkeley.edu

The problem is that fp->_w is not getting set properly, which causes the
stdio buffer to overflow later on.  Setvbuf() defers this setting, among
other things, to __srefill or __swsetup, neither of which gets called in
this case.  Simply setting fp->_w will fix the bus error, but does not
address the questions of whether any of the other operations in __srefill
and __swsetup are needed or how to correctly force invocation of __srefill
or __swetup, depending on whether the first operation is a read or write.

30 years agoDeclare cnopen, cnclose, and other console routines.
Garrett Wollman [Sat, 27 Nov 1993 19:19:19 +0000 (19:19 +0000)]
Declare cnopen, cnclose, and other console routines.

30 years agoDon't set splbio() in vfs_update, and don't set PCATCH on the sleep.
David Greenman [Sat, 27 Nov 1993 17:07:05 +0000 (17:07 +0000)]
Don't set splbio() in vfs_update, and don't set PCATCH on the sleep.
Ignore all signals sent to system (internal) daemons.

30 years agoAdded mkdep to list of things built in bootstrapld.
Paul Richards [Sat, 27 Nov 1993 16:12:55 +0000 (16:12 +0000)]
Added mkdep to list of things built in bootstrapld.

Not really a shlibs thing but here's as good a place as any to put it.
It needs to be made before libg++ becasue the old mkdep fails.

30 years agoFix some pointer casting that Garrett missed.
David Greenman [Sat, 27 Nov 1993 16:04:36 +0000 (16:04 +0000)]
Fix some pointer casting that Garrett missed.

30 years agoDeclared cn{open,close,read,write,ioctl,select} extern.
Paul Richards [Sat, 27 Nov 1993 06:50:45 +0000 (06:50 +0000)]
Declared cn{open,close,read,write,ioctl,select} extern.

30 years agoFix conflicting prototypes and return values.
Paul Richards [Sat, 27 Nov 1993 06:32:41 +0000 (06:32 +0000)]
Fix conflicting prototypes and return values.

30 years agoFix d_write_t -> d_rdwr_t (typing error).
Garrett Wollman [Fri, 26 Nov 1993 22:44:17 +0000 (22:44 +0000)]
Fix d_write_t -> d_rdwr_t (typing error).

30 years agoUse #include "kernel.h" to get hz variable, rather than declaring it
Garrett Wollman [Fri, 26 Nov 1993 19:08:26 +0000 (19:08 +0000)]
Use #include "kernel.h" to get hz variable, rather than declaring it
extern.

30 years agoRemove 'update' completely; it's an internal daemon in our kernel.
David Greenman [Thu, 25 Nov 1993 13:35:54 +0000 (13:35 +0000)]
Remove 'update' completely; it's an internal daemon in our kernel.

30 years agoDon't start update daemon; it's now internel to the kernel.
David Greenman [Thu, 25 Nov 1993 13:23:46 +0000 (13:23 +0000)]
Don't start update daemon; it's now internel to the kernel.

30 years agomade update an internel daemon (code from John Dyson)
David Greenman [Thu, 25 Nov 1993 13:16:09 +0000 (13:16 +0000)]
made update an internel daemon (code from John Dyson)
fixed broken biodone (specifically, the handling of B_CALL) while I was
at it. Note that the handling of B_CALL in the swap pager is broken; this
is only half of the fix.

30 years agofix from John Dyson for swap space "leak"
David Greenman [Thu, 25 Nov 1993 12:07:42 +0000 (12:07 +0000)]
fix from John Dyson for swap space "leak"
Description:

The function vm_object_collapse merges the objects shadow
into the object if it is possible.  This can have the positive
effect of eliminating objects that are created by vm_map_copy
at perhaps unoportune times when the source object might be
shadowed and collapse cannot occur.  Once a collapse does not
occur and a pager is created, then that object with a new pager
cannot ever be collapsed again.  Therefore shadows that would
have been created will never be collapsed into the 'parent'
object.  The major thing that keeps vm_object_collapse from
doing its job is if the shadow object has the paging_in_progress
flag set.  Now, what happens is that when vm_map_copy is called
and the system is paging heavily, the vm_object_collapse does not
work well in both vm_map_copy and vm_pageout.  Also vm_map_copy
creates new shadow objects a bit too often, and it has an optimization
to increase the reference count to an existing shadow if a pager
does not exist on the source object.  Additionally, the case of using
the default(swap) pager is ok.

In vm_pageout, a check has been added to defer creating a pager for an
object until all of its shadows have no paging in progress, then the
collapse will work very well.  It is problematical to make collapse
work when paging is occuring, so the solution to defer is signicantly
easier.

The end effect of these changes is to minimize the creation of new
shadow objects and their associated pagers and 'lost' paging space.

30 years agoPatch from Julian Elischer:
David Greenman [Thu, 25 Nov 1993 06:30:58 +0000 (06:30 +0000)]
Patch from Julian Elischer:

Here is the fix for the 'hanging' bug.

This bug happenned whenever two operations were already underway on the
disk and a third (non-IO) command was requested..
in this case the process submitting the NON-IO command was requested to wait,
and a flag set so that on completion of the IO commands, the Non-io command
was given priority over any pending IO commands. (the queue is not
allowed to drain while there are pending "special" ops).

The flag that indicated this was not being reset, so further IO commands
were prohibited from that moment on.

30 years agoMake the LINT kernel compile with -W -Wreturn-type -Wcomment -Werror, and
Garrett Wollman [Thu, 25 Nov 1993 01:39:19 +0000 (01:39 +0000)]
Make the LINT kernel compile with -W -Wreturn-type -Wcomment -Werror, and
add same (sans -Werror) to Makefile for future compilations.

30 years agoAdded -lgcc_pic back again.
Jordan K. Hubbard [Thu, 25 Nov 1993 01:06:42 +0000 (01:06 +0000)]
Added -lgcc_pic back again.

30 years agoChanged the header title from SETPGRP to SETPGID in the manual page
Andreas Schulz [Wed, 24 Nov 1993 18:19:50 +0000 (18:19 +0000)]
Changed the header title from SETPGRP to SETPGID in the manual page
setpgid.2 . ATS

30 years ago/var/run is cleared on boot. Rerun 'ldconfig /usr/X386/lib' after
Paul Richards [Wed, 24 Nov 1993 05:17:14 +0000 (05:17 +0000)]
/var/run is cleared on boot. Rerun 'ldconfig /usr/X386/lib' after
clearing /var/run if /sbin/ldconfig exists.

30 years agoAdded an MLINK command for the setpgrp system call. It is documented
Andreas Schulz [Tue, 23 Nov 1993 23:46:00 +0000 (23:46 +0000)]
Added an MLINK command for the setpgrp system call. It is documented
in the setpgid manual page. Bug report from phk@data.fls.dk
 (Poul-Henning Kamp/P-HK).

30 years agoSubject: Panic - can't mount route (Soren's changes)
Nate Willams [Tue, 23 Nov 1993 21:36:37 +0000 (21:36 +0000)]
Subject: Panic - can't mount route (Soren's changes)
From: Geoff Rehmet  <g89r4222@braae.ru.ac.za>

Description:
On bootup, probe of wd drives fails (CP30104), and kernel panics
- cannot mount root
It appears that the device probe just times out.
Increasing the timeout back to its old value fixes the problem.
Repeat-By:
SUP FreBSD-current,  Find a CP30104 .....  (ok ok ok)
Basically - Soren's changes barf my disk.

30 years agoFixed a null ptr reference, found by apg@demos.su (Paul Antonov).
Andreas Schulz [Tue, 23 Nov 1993 20:07:31 +0000 (20:07 +0000)]
Fixed a null ptr reference, found by apg@demos.su (Paul Antonov).

30 years agoFrom: Andrew Valencia <vandys@cisco.com>
Christoph Robitschko [Tue, 23 Nov 1993 18:51:43 +0000 (18:51 +0000)]
From: Andrew Valencia <vandys@cisco.com>
backgammon will core dump if the optimizer is used to compile the game.
Add the following patch to /usr/src/backgammon/backgammon/move.c to fix this.
From: chmr
Another small fix (remove new element from the free list).

30 years agoChanged return(-1) in switch_scr to return(EINVAL), because -1 is
Christoph Robitschko [Tue, 23 Nov 1993 18:20:52 +0000 (18:20 +0000)]
Changed return(-1) in switch_scr to return(EINVAL), because -1 is
ERESTART on return from a system call.

30 years agoSync with NetBSD version.
J.T. Conklin [Tue, 23 Nov 1993 00:34:05 +0000 (00:34 +0000)]
Sync with NetBSD version.

30 years agoPOSIX.2ification, from my work with NetBSD.
J.T. Conklin [Tue, 23 Nov 1993 00:20:04 +0000 (00:20 +0000)]
POSIX.2ification, from my work with NetBSD.

30 years agoPOSIX.2ifciation, from my work with NetBSD.
J.T. Conklin [Tue, 23 Nov 1993 00:17:17 +0000 (00:17 +0000)]
POSIX.2ifciation, from my work with NetBSD.

30 years agoPOSIX.2ification, from my work with NetBSD.
J.T. Conklin [Tue, 23 Nov 1993 00:13:55 +0000 (00:13 +0000)]
POSIX.2ification, from my work with NetBSD.

30 years agoPOSIX.2ification, from my work on NetBSD.
J.T. Conklin [Tue, 23 Nov 1993 00:10:09 +0000 (00:10 +0000)]
POSIX.2ification, from my work on NetBSD.

30 years agoPOSIX.2ifciations, from my work with NetBSD.
J.T. Conklin [Tue, 23 Nov 1993 00:06:21 +0000 (00:06 +0000)]
POSIX.2ifciations, from my work with NetBSD.