]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
21 years agoForced commit.
markm [Thu, 24 Jul 2003 18:11:27 +0000 (18:11 +0000)]
Forced commit.

The previous commit included some commented-out macro-assignments
with comments documenting their function. These macros were of
considerable use to me in building a test "make release".

21 years agoGet the correct set of cryptographically-infected binaries
markm [Thu, 24 Jul 2003 17:28:52 +0000 (17:28 +0000)]
Get the correct set of cryptographically-infected binaries
automagically from the (secure|kerberos5)/Makefile:[KS]PROG.

21 years agoMake sure that a "make release" (more accurately the bit that makes
markm [Thu, 24 Jul 2003 17:19:15 +0000 (17:19 +0000)]
Make sure that a "make release" (more accurately the bit that makes
the crunched binary) get a non-cryptographic telnet. This is overkill
in that it covers stuff that is not normally used in a crunched binary.

21 years agoThe POSIX spec also requires that kern_sigtimedwait return
mtm [Thu, 24 Jul 2003 17:07:17 +0000 (17:07 +0000)]
The POSIX spec also requires that kern_sigtimedwait return
EINVAL if tv_nsec of the timeout is less than zero.

21 years agoMake the debugging variable that controls dumping of IP over ATM packets
harti [Thu, 24 Jul 2003 15:25:17 +0000 (15:25 +0000)]
Make the debugging variable that controls dumping of IP over ATM packets
accessible as a sysctl.

Submitted by: Vincent Jardin <vjardin@wanadoo.fr>
MFC after: 2 weeks

21 years agoCreate a sysctl that allows to enable/disable printing of SPANS messages.
harti [Thu, 24 Jul 2003 14:37:01 +0000 (14:37 +0000)]
Create a sysctl that allows to enable/disable printing of SPANS messages.
While here delete to sys/types.h includes when sys/param.h is also included.

Submitted by: Vincent Jardin <vjardin@wanadoo.fr>
MFC after: 2 weeks

21 years agoReplaced the hardcoded target names in the touch(1) commands with
ru [Thu, 24 Jul 2003 13:45:05 +0000 (13:45 +0000)]
Replaced the hardcoded target names in the touch(1) commands with
${.TARGET}, to ease maintenance.

21 years agoFree the UNI vcc to the same zone from where it was allocated from.
harti [Thu, 24 Jul 2003 12:24:41 +0000 (12:24 +0000)]
Free the UNI vcc to the same zone from where it was allocated from.
This resulted in a panic when detaching the uni31 signalling manager.

21 years agoNow that we have if_detach() don't try to get rid of all the interface
harti [Thu, 24 Jul 2003 11:17:36 +0000 (11:17 +0000)]
Now that we have if_detach() don't try to get rid of all the interface
stuff (routes, ...) by hand - simply use if_detach().

Submitted by: Vincent Jardin <vjardin@wanadoo.fr>
MFC after: 2 week

21 years agoComplain about extraneous command line arguments.
des [Thu, 24 Jul 2003 10:44:45 +0000 (10:44 +0000)]
Complain about extraneous command line arguments.

21 years agoCreate a subtree 'harp' of the net sysctl tree. This uses a fixed
harti [Thu, 24 Jul 2003 10:33:01 +0000 (10:33 +0000)]
Create a subtree 'harp' of the net sysctl tree. This uses a fixed
OID as the other protocol family sub-trees do, that is equal to the
protocol family identifier. Make the ATM layer debugging flags
available under this tree.

Submitted by: Vincent Jardin <vjardin@wanadoo.fr>
MFC after: 2 weeks

21 years agoForced commit for the MFC.
harti [Thu, 24 Jul 2003 09:14:50 +0000 (09:14 +0000)]
Forced commit for the MFC.

MFC after: 2 weeks

21 years agoConstify the arguments to several pdu_print functions.
harti [Thu, 24 Jul 2003 09:13:03 +0000 (09:13 +0000)]
Constify the arguments to several pdu_print functions.

21 years agoImplement cpu_set_upcall and cpu_set_upcall_kse.
davidxu [Thu, 24 Jul 2003 08:52:44 +0000 (08:52 +0000)]
Implement cpu_set_upcall and cpu_set_upcall_kse.

Reviewed by: peter

21 years agoSet fault address to si_addr.
davidxu [Thu, 24 Jul 2003 08:51:22 +0000 (08:51 +0000)]
Set fault address to si_addr.

Reviewed by: peter

21 years agoAdd BPF support to HARP network interfaces. This allows one to see
harti [Thu, 24 Jul 2003 08:15:20 +0000 (08:15 +0000)]
Add BPF support to HARP network interfaces. This allows one to see
the traffic on LLC multiplexed connections (like CLIP).

PR: kern/51831
Submitted by: Vincent Jardin <vjardin@wanadoo.fr>
MFC after: 2 weeks

21 years agoImplement _get_curthread and _set_curthread. We use GCCs builtin
marcel [Thu, 24 Jul 2003 07:51:49 +0000 (07:51 +0000)]
Implement _get_curthread and _set_curthread. We use GCCs builtin
function this, which expands to PAL calls (rduniq and wruniq).
This needs adjustment when TLS is implemented.

21 years agoImplement casuptr().
marcel [Thu, 24 Jul 2003 07:49:45 +0000 (07:49 +0000)]
Implement casuptr().

21 years agoIn get_mcontext() and set_mcontext() save and restore the current
marcel [Thu, 24 Jul 2003 07:48:11 +0000 (07:48 +0000)]
In get_mcontext() and set_mcontext() save and restore the current
thread pointer.

21 years agoFix rescue build using -j. The problem appears to be make not being able
gordon [Thu, 24 Jul 2003 07:42:42 +0000 (07:42 +0000)]
Fix rescue build using -j. The problem appears to be make not being able
to find the source when the object was specified as <directory>/foo.o.

The fix makes the build go through a make objs before compiling the rest
of the crunchgen. This ensures that the dhclient bits are built in the
correct place where they are picked up for the final compile of rescue.

I'd like to thank dwhite@ and gad@ for helping me track down the problem.

Fast testing box provided by: phk@ (thanks)

21 years agoImplement alpha_pal_wrunique() and alpha_pal_rdunique(). Both are
marcel [Thu, 24 Jul 2003 07:41:08 +0000 (07:41 +0000)]
Implement alpha_pal_wrunique() and alpha_pal_rdunique(). Both are
used to set and get the thread pointer. Note that a context switch
will automaticly save and restore the thread pointer.

21 years agoUse a spare for the thread pointer (mc_thrptr). The thread pointer
marcel [Thu, 24 Jul 2003 07:34:31 +0000 (07:34 +0000)]
Use a spare for the thread pointer (mc_thrptr). The thread pointer
is only read and written by set_mcontext() and get_mcontext() for
use by threading libraries.

21 years agoEnsure that for the cryptographic instances of *telnet*, the "crypto"
markm [Thu, 24 Jul 2003 07:19:55 +0000 (07:19 +0000)]
Ensure that for the cryptographic instances of *telnet*, the "crypto"
distribution is used. This only affects release-building.

21 years agoConnect ncplist, ncplogin, and smbutil to the amd64 build.
peter [Thu, 24 Jul 2003 02:09:19 +0000 (02:09 +0000)]
Connect ncplist, ncplogin, and smbutil to the amd64 build.

21 years agoAdd mount_nwfs/mount_smbfs to the build for amd64.
peter [Thu, 24 Jul 2003 02:07:14 +0000 (02:07 +0000)]
Add mount_nwfs/mount_smbfs to the build for amd64.

21 years agoConnect libncp/libsmb to the build. They compile, but have a couple of
peter [Thu, 24 Jul 2003 02:05:48 +0000 (02:05 +0000)]
Connect libncp/libsmb to the build.  They compile, but have a couple of
silly bugs that probably wont quite make a segfault.  eg: passing a pointer
to an int to sysctl instead of a pointer to a size_t.

21 years agosize_t != int. Make this compile on 64 bit platforms (eg: amd64).
peter [Thu, 24 Jul 2003 01:59:18 +0000 (01:59 +0000)]
size_t != int.  Make this compile on 64 bit platforms (eg: amd64).
Also, "u_short value; if (value > 0xffff)" can never be true.

21 years agoBuild /sbin/gpt on amd64 for good luck as well.
peter [Thu, 24 Jul 2003 01:42:49 +0000 (01:42 +0000)]
Build /sbin/gpt on amd64 for good luck as well.

21 years agoAttach acpi, boot0cfg, lptcontrol, sicontrol, spkrtest, zzz for amd64.
peter [Thu, 24 Jul 2003 01:41:34 +0000 (01:41 +0000)]
Attach acpi, boot0cfg, lptcontrol, sicontrol, spkrtest, zzz  for amd64.

21 years agoFix an annoying bug in tar. When it converted 'tar zcf' to 'tar -z -c -f'
peter [Thu, 24 Jul 2003 00:04:01 +0000 (00:04 +0000)]
Fix an annoying bug in tar.  When it converted 'tar zcf' to 'tar -z -c -f'
it forgot to null terminate the new argv[] array.  If you mixed this
with $TAR_OPTIONS, phkmalloc and a whole bunch of other variables, you
could end up with a segfault.  This isn't strictly a phkmalloc victory
since tar walks off the end of an array rather than use uninitialized
malloc memory, but phkmalloc makes it easier to provoke.

21 years agoSimplistic C comment re is wrong, use more correct one
imp [Wed, 23 Jul 2003 23:50:25 +0000 (23:50 +0000)]
Simplistic C comment re is wrong, use more correct one

21 years agoRemove old workaround
imp [Wed, 23 Jul 2003 23:50:00 +0000 (23:50 +0000)]
Remove old workaround

21 years agoMake the breakpoint instruction trap gate available to users.
peter [Wed, 23 Jul 2003 23:20:20 +0000 (23:20 +0000)]
Make the breakpoint instruction trap gate available to users.
ptrace() needs this.

Submitted by: Mark Kettenis <kettenis@chello.nl>

21 years agoSet the %gs base to pcb_gsbase, not pcb_fsbase. Oops.
peter [Wed, 23 Jul 2003 23:17:15 +0000 (23:17 +0000)]
Set the %gs base to pcb_gsbase, not pcb_fsbase.  Oops.

Discovered by: davidxu

21 years agoTurn -Werror back off again. Leaving out the invariants options causes
peter [Wed, 23 Jul 2003 22:02:30 +0000 (22:02 +0000)]
Turn -Werror back off again.  Leaving out the invariants options causes
more trouble than I expected.

21 years agoDocument an additional error return value. The connect(2) call can also
dds [Wed, 23 Jul 2003 22:00:08 +0000 (22:00 +0000)]
Document an additional error return value.  The connect(2) call can also
return EACCES on non-Unix domain sockets as demonstrated by the
following program:

#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <errno.h>
#include <sys/types.h>
#include <sys/socket.h>
#include <netinet/in.h>
#include <arpa/inet.h>

int
main(int argc, char *argv[])
{
struct sockaddr_in rem_addr;
int sock;

if ((sock = socket(AF_INET, SOCK_STREAM, 0)) < 0) {
perror("socket");
exit(1);
}

bzero((char *)&rem_addr, sizeof(rem_addr));
rem_addr.sin_family = AF_INET;
rem_addr.sin_addr.s_addr = INADDR_NONE;
rem_addr.sin_port = htons(10000);

if (connect(sock, (struct sockaddr *)&rem_addr,
sizeof(rem_addr)) < 0) {
perror("connect");
exit(1);
}
}

The call chain returning this value is probably:

kern/uipc_syscalls.c:connect
kern/uipc_socket.c:soconnect
netinet/tcp_usrreq.c:tcp_usr_connect
netinet/tcp_output.c:tcp_output
netinet/ip_output.c:ip_output

Reviewed by: schweikh (mentor)
MFC after: 2 weeks

21 years agoAdded the (undocumented) EMUL_BOOT variable (for TARGET=i386 only)
ru [Wed, 23 Jul 2003 20:53:37 +0000 (20:53 +0000)]
Added the (undocumented) EMUL_BOOT variable (for TARGET=i386 only)
that causes the bootable ISO images to be created using the floppy
emulation (the old method) as opposed to the new "cdboot" method.

Only copy boot.flp to the 2nd CD-ROM if this variable is defined.

Reviewed by: murray

21 years agoInitialize 'blocked' to NULL. I think this was a real problem, but I
peter [Wed, 23 Jul 2003 20:29:13 +0000 (20:29 +0000)]
Initialize 'blocked' to NULL.  I think this was a real problem, but I
am not sure about that.  The lack of -Werror and the inline noise hid
this for a while.

21 years agoStop GCC from whining when people use a 16 bit port number for inb() and outb()
phk [Wed, 23 Jul 2003 20:28:23 +0000 (20:28 +0000)]
Stop GCC from whining when people use a 16 bit port number for inb() and outb()

21 years agoInstall rc files from ${.CURDIR}, not ${.OBJDIR}.
des [Wed, 23 Jul 2003 20:22:50 +0000 (20:22 +0000)]
Install rc files from ${.CURDIR}, not ${.OBJDIR}.

21 years agoSample configuration files.
des [Wed, 23 Jul 2003 20:16:36 +0000 (20:16 +0000)]
Sample configuration files.

21 years agoRewrite tbmaster to use configuration files instead of a hard-coded hash.
des [Wed, 23 Jul 2003 20:16:11 +0000 (20:16 +0000)]
Rewrite tbmaster to use configuration files instead of a hard-coded hash.

21 years agoImprove the markup a little.
des [Wed, 23 Jul 2003 20:15:49 +0000 (20:15 +0000)]
Improve the markup a little.

21 years agoTurn -Werror back on.
peter [Wed, 23 Jul 2003 20:10:09 +0000 (20:10 +0000)]
Turn -Werror back on.

21 years agoARGH. I *knew* I'd eventually accidently commit this. Change 5.1-XP back
peter [Wed, 23 Jul 2003 20:07:49 +0000 (20:07 +0000)]
ARGH. I *knew* I'd eventually accidently commit this.  Change 5.1-XP back
to 5.1-CURRENT.

21 years agoProvide a knob for raising the inline instruction estimate threshold and
peter [Wed, 23 Jul 2003 20:03:42 +0000 (20:03 +0000)]
Provide a knob for raising the inline instruction estimate threshold and
set an initial value.  This is aimed at getting us closer to being able to
turn -Werror back on and we can adjust the settings later on.  Yes, we
could turn off -Wno-inline instead, but that would hide the effect of
gcc's bogo-estimator ignoring inline (either rightly or wrongly).

21 years agoAnnotate pmap_changebit() as __always_inline. This function was
alc [Wed, 23 Jul 2003 19:49:32 +0000 (19:49 +0000)]
Annotate pmap_changebit() as __always_inline.  This function was
written as a template that when inlined is specialized for the caller
through constant value propagation and dead code elimination.  Thus,
the specialized code that is generated for pmap_clear_reference() et
al. avoids several conditional branches inside of a loop.

21 years agoUse macros from apic.h to when writing to the ICR to send IPIs to startup
jhb [Wed, 23 Jul 2003 19:04:28 +0000 (19:04 +0000)]
Use macros from apic.h to when writing to the ICR to send IPIs to startup
APs rather than magic numbers.

Tested by: scottl

21 years agoAdd a new macro APIC_ICRLO_RESV_MASK that contains all of the reserved
jhb [Wed, 23 Jul 2003 18:59:38 +0000 (18:59 +0000)]
Add a new macro APIC_ICRLO_RESV_MASK that contains all of the reserved
fields in the low 32 bits of the local APIC ICR register.  Use this macro
in place of APIC_RESV2_MASK when masking off existing bits from the ICR
when writing to it to send an IPI.

Tested by: scottl

21 years agoUndo single-intance inlining which is way above the comfort limit for GCC.
phk [Wed, 23 Jul 2003 18:03:22 +0000 (18:03 +0000)]
Undo single-intance inlining which is way above the comfort limit for GCC.

21 years agoUndo agressive inlining which GCC previously wisely ignored.
phk [Wed, 23 Jul 2003 17:58:41 +0000 (17:58 +0000)]
Undo agressive inlining which GCC previously wisely ignored.

The this eliminates 67% of the text segment (relative to respected
inline requests).

21 years agoHandle the new MEDIA definitions.
harti [Wed, 23 Jul 2003 15:04:31 +0000 (15:04 +0000)]
Handle the new MEDIA definitions.

21 years agoAdded the WORLDDIR variable (defaulting to ${.CURDIR}/..) that
ru [Wed, 23 Jul 2003 14:40:51 +0000 (14:40 +0000)]
Added the WORLDDIR variable (defaulting to ${.CURDIR}/..) that
points to a directory where "make buildworld" was run.  Useful
for building 5.x snapshots on 4.x.

21 years agoConvert a lot of uma_zalloc() calls to be NOWAIT instead of WAITOK. All
harti [Wed, 23 Jul 2003 14:28:57 +0000 (14:28 +0000)]
Convert a lot of uma_zalloc() calls to be NOWAIT instead of WAITOK. All
these may be called from contexts where we cannot sleep (callout handlers
for example).

21 years agoGet rid of the zone for network interfaces. We have converted this to
harti [Wed, 23 Jul 2003 14:25:53 +0000 (14:25 +0000)]
Get rid of the zone for network interfaces. We have converted this to
use malloc(9).

21 years agoAdd manual page to document the new rescue utilities in /rescue.
simon [Wed, 23 Jul 2003 13:24:32 +0000 (13:24 +0000)]
Add manual page to document the new rescue utilities in /rescue.

Most text by: Tim Kientzle <kientzle@acm.org>
Reviewed by: ru, doc@
Approved by: ceri (mentor)

21 years agoMake NO_FLOPPIES really mean "no floppies", including not even
ru [Wed, 23 Jul 2003 12:12:53 +0000 (12:12 +0000)]
Make NO_FLOPPIES really mean "no floppies", including not even
creating the "floppies" directory on FTP and CD-ROMs.  Enforce
NO_FLOPPIES on amd64.

Tested for: amd64, i386 with -DNO_FLOPPIES

21 years agoMFi386: revision 1.1090.
nyan [Wed, 23 Jul 2003 12:09:14 +0000 (12:09 +0000)]
MFi386: revision 1.1090.

21 years agoFixed some style bugs (misplacement and misformatting of some commented-out
bde [Wed, 23 Jul 2003 09:24:44 +0000 (09:24 +0000)]
Fixed some style bugs (misplacement and misformatting of some commented-out
code).

21 years agoRevert revision 1.788, and explain a bit the intent of having
ru [Wed, 23 Jul 2003 06:00:56 +0000 (06:00 +0000)]
Revert revision 1.788, and explain a bit the intent of having
the boot.flp image on the second CD-ROM.

Explained by: jhb
Desired by: vendors

21 years agoamd64 will not provide floppies.
ru [Wed, 23 Jul 2003 05:55:02 +0000 (05:55 +0000)]
amd64 will not provide floppies.

21 years agoDo not define BIGBOOTSIZE and the friends for amd64; it serves
ru [Wed, 23 Jul 2003 05:53:52 +0000 (05:53 +0000)]
Do not define BIGBOOTSIZE and the friends for amd64; it serves
no useful purpose other than wasting CPU time in "make release"
creating useless boot.flp.

Desired by: peter

21 years agoOnly provide one copy of the math functions. If we provide a MD function,
peter [Wed, 23 Jul 2003 04:53:47 +0000 (04:53 +0000)]
Only provide one copy of the math functions.  If we provide a MD function,
do not also provide a __generic_XXX version as well.  This is how we
used to runtime select the generic vs i387 versions on the i386 platform.

This saves a pile of #defines in the src/math_private.h file to undo the
__generic_XXX renames in some of the *.c files.

21 years agoRemove _ARCH_INDIRECT, it was the glue to enable having both hardware
peter [Wed, 23 Jul 2003 04:28:51 +0000 (04:28 +0000)]
Remove _ARCH_INDIRECT, it was the glue to enable having both hardware
and software versions of the floating point code in libm.  The runtime
selection was done by reading the hw.floatingpoint sysctl via
__get_hw_float().

21 years agoNo longer need the internal __get_hw_float() function.
peter [Wed, 23 Jul 2003 04:25:04 +0000 (04:25 +0000)]
No longer need the internal __get_hw_float() function.

21 years agoNow that we do not need to do runtime detection for the broken default
peter [Wed, 23 Jul 2003 04:23:36 +0000 (04:23 +0000)]
Now that we do not need to do runtime detection for the broken default
fp emulator, stop doing the runtime selection of hardware or emulated
floating point operations on i386.  Note that I have not suppressed the
duplicate compiles yet.

While here, fix the alpha.  It has provided specific copysign/copysignf
functions since the beginning of time, but they have never been used.

21 years agoWe sloppily created an array for the high FP registers (f32-f127),
marcel [Wed, 23 Jul 2003 03:08:34 +0000 (03:08 +0000)]
We sloppily created an array for the high FP registers (f32-f127),
but this just created a weird inconsistency when porting gdb(1).
Instead, we name each high FP register seperately, like we do for
all the other registers.

21 years agoMove idle kse wakeup to outside of regions where locks are held.
deischen [Wed, 23 Jul 2003 02:11:07 +0000 (02:11 +0000)]
Move idle kse wakeup to outside of regions where locks are held.
This eliminates ping-ponging of locks, where the idle KSE wakes
up only to find the lock it needs is being held.  This gives
little or no gain to M:N mode but greatly speeds up 1:1 mode.

Reviewed & Tested by: davidxu

21 years agoAttempt to preempt any new gcc-ism references to
peter [Tue, 22 Jul 2003 23:50:32 +0000 (23:50 +0000)]
Attempt to preempt any new gcc-ism references to
__attribute__((__always_inline__)) by adding an __always_inline macro
(used like __dead2 etc).  __inline_damnit has also been suggested but we
have a precedent of keeping the names similar so they are easier to find.

21 years agoThere is strong reason to believe that gcc 4 will also support
imp [Tue, 22 Jul 2003 23:14:32 +0000 (23:14 +0000)]
There is strong reason to believe that gcc 4 will also support
__attribute__((__nonnull__(x)), assume it so.

21 years agoAdded entry for Billionton LM5LT-10N
dds [Tue, 22 Jul 2003 21:23:01 +0000 (21:23 +0000)]
Added entry for Billionton LM5LT-10N
Fixed small typo

Reviewed by: imp
Approved by: imp

21 years agoRemove all but one of the inlines here, this reduces the code size by
phk [Tue, 22 Jul 2003 20:54:26 +0000 (20:54 +0000)]
Remove all but one of the inlines here, this reduces the code size by
2032 bytes and has no measurable impact on performance.

21 years agoFixed .Nm element, updated HISTORY section, added AUTHORS section.
dds [Tue, 22 Jul 2003 19:23:45 +0000 (19:23 +0000)]
Fixed .Nm element, updated HISTORY section, added AUTHORS section.

Reviewed by: schweikh (mentor)
MFC after: 2 weeks

21 years agoTest 8.16 in sed.test tests the ability of a sed to handle an empty
dds [Tue, 22 Jul 2003 19:22:08 +0000 (19:22 +0000)]
Test 8.16 in sed.test tests the ability of a sed to handle an empty
regular expression as the first argument to a substitute command. If
used to test a sed which (erroneously) evaluates this at translation
time rather than at execution time, the bugged sed is put into an
infinite loop. This mode of failure seems excessive. Such a failing
sed is the Free Software Foundation's sed 3.02.

The specific test was also not being executed for the BSD sed.

Both problems are now fixed.

PR: misc/25585
Submitted by: Walter Briscoe <w.briscoe@ponl.com>
Approved by: schweikh (mentor)
MFC after: 2 weeks

21 years agocorrect device table a la ath(4)
sam [Tue, 22 Jul 2003 19:08:46 +0000 (19:08 +0000)]
correct device table a la ath(4)

21 years agoadd IPSEC_FILTERGIF suport for FAST_IPSEC
sam [Tue, 22 Jul 2003 18:58:34 +0000 (18:58 +0000)]
add IPSEC_FILTERGIF suport for FAST_IPSEC

PR: kern/51922
Submitted by: Eric Masson <e-masson@kisoft-services.com>
MFC after: 1 week

21 years agoDe-inline some functions. It doesn't gain us anything and bloats
mux [Tue, 22 Jul 2003 15:50:54 +0000 (15:50 +0000)]
De-inline some functions.  It doesn't gain us anything and bloats
code size by 3616 bytes.  Furthemore, it was previously ignored by GCC.

While I'm at it, fix some bogus comments.

21 years agoAllocate network interfaces from malloc() instead of using a zone.
harti [Tue, 22 Jul 2003 15:11:08 +0000 (15:11 +0000)]
Allocate network interfaces from malloc() instead of using a zone.
Usually one needs only a couple of them so using a zone is waste
of memory (esp. on multi-cpu systems).

21 years agoMake sure the crypto versions of libfetch and fetch(1) appear in
ru [Tue, 22 Jul 2003 13:54:31 +0000 (13:54 +0000)]
Make sure the crypto versions of libfetch and fetch(1) appear in
the "crypto" distribution.

Approved by: des

21 years agoRemove the zone limits for all the zones used in the ATM code.
harti [Tue, 22 Jul 2003 12:46:30 +0000 (12:46 +0000)]
Remove the zone limits for all the zones used in the ATM code.
These were a left over from when the private memory pools were
converted to use uma zones. The limit of UMA zones, however,
works differently. When a zone is limited to only one or two pages
than, on multi-cpu systems, processes can get stuck on the zonelimit,
because all remaining free items are in caches of other CPUs.

Also add rudimentary error handling in some places (panic) when a zone
cannot be created.

21 years agoEISA_SLOTS is mandantory to get opt_eisa.h
ticso [Tue, 22 Jul 2003 11:42:45 +0000 (11:42 +0000)]
EISA_SLOTS is mandantory to get opt_eisa.h
Put it into MI files.

21 years agorelocate eisa into MI files.
ticso [Tue, 22 Jul 2003 11:41:15 +0000 (11:41 +0000)]
relocate eisa into MI files.

Suggested by: jhb

21 years agonote rescue issues and fpu emulation removal
imp [Tue, 22 Jul 2003 11:08:41 +0000 (11:08 +0000)]
note rescue issues and fpu emulation removal

21 years agoRevert stuff which accidentally ended up in the previous commit.
phk [Tue, 22 Jul 2003 10:36:36 +0000 (10:36 +0000)]
Revert stuff which accidentally ended up in the previous commit.

21 years agoDon't attempt to inline large functions mb_alloc() and mb_free(),
phk [Tue, 22 Jul 2003 10:24:41 +0000 (10:24 +0000)]
Don't attempt to inline large functions mb_alloc() and mb_free(),
it more than doubles the text size of this file.

GCC has wisely ignored us on this previously

21 years agoMove inlined function ie_ack() up to before first use.
phk [Tue, 22 Jul 2003 09:35:45 +0000 (09:35 +0000)]
Move inlined function ie_ack() up to before first use.

21 years agoDon't inline very large functions.
phk [Tue, 22 Jul 2003 09:27:58 +0000 (09:27 +0000)]
Don't inline very large functions.

Gcc has silently not been doing this for a long time.

21 years agoDon't inline ridiculously very large functions.
phk [Tue, 22 Jul 2003 09:22:00 +0000 (09:22 +0000)]
Don't inline ridiculously very large functions.

Compared to the contents of these functions, an extra function call
is nano-peanuts.

21 years ago/floppies/boot.flp is no longer necessary on bootable i386 CD-ROMs.
ru [Tue, 22 Jul 2003 09:04:09 +0000 (09:04 +0000)]
/floppies/boot.flp is no longer necessary on bootable i386 CD-ROMs.

Forgotten by: jhb

21 years agoDon't complain about inlines for genassym
phk [Tue, 22 Jul 2003 09:02:21 +0000 (09:02 +0000)]
Don't complain about inlines for genassym

21 years agoWhen padding an mbuf chain to have a length that is a multiple
harti [Tue, 22 Jul 2003 08:20:09 +0000 (08:20 +0000)]
When padding an mbuf chain to have a length that is a multiple
of 48 bytes for AAL0, we also need to update the packet header.

Spotted by: Anil Madhavapeddy <anil@recoil.org>

21 years agoUpdate npx status.
peter [Tue, 22 Jul 2003 08:15:29 +0000 (08:15 +0000)]
Update npx status.

Reminded by:  tjr

21 years agoInitiate de-orbit burn for fpu-less operation. 386+387 is still
peter [Tue, 22 Jul 2003 08:11:17 +0000 (08:11 +0000)]
Initiate de-orbit burn for fpu-less operation.  386+387 is still
theoretically supportable, but you'd really be happier with FreeBSD 2.1.8
on it.

21 years agoAdd a note that net.inet.ip.fw.autoinc_step is ipfw2-specific
luigi [Tue, 22 Jul 2003 07:41:24 +0000 (07:41 +0000)]
Add a note that net.inet.ip.fw.autoinc_step is ipfw2-specific

21 years agoFix a 64 bit bug lost in the inline warning noise. sizeof is not always
peter [Tue, 22 Jul 2003 06:58:04 +0000 (06:58 +0000)]
Fix a 64 bit bug lost in the inline warning noise.  sizeof is not always
an int.  Do not printf() it as though it were.

21 years agoswp_pager_hash() was called before it was instantiated inline. This made
peter [Tue, 22 Jul 2003 06:55:48 +0000 (06:55 +0000)]
swp_pager_hash() was called before it was instantiated inline.  This made
gcc (quite rightly) unhappy.  Move it earlier.

21 years agoGo back to 64 bit precision for fadd/fsub/fsqrt etc. This is because on
peter [Tue, 22 Jul 2003 06:50:34 +0000 (06:50 +0000)]
Go back to 64 bit precision for fadd/fsub/fsqrt etc.  This is because on
AMD64, gcc (and the ABI) expects the x87 unit to be running in 80/64
mode (not 64/53) so that it can use it for 'long double' operations.  It
takes the expected precision differences into account when generating
code.

21 years agoInstantiate explicit callable versions of the machine/ieeefp.h inlines
peter [Tue, 22 Jul 2003 06:46:17 +0000 (06:46 +0000)]
Instantiate explicit callable versions of the machine/ieeefp.h inlines
for the use of non-GCC compilers and C++ code.

21 years agoExtend the machine/ieeefp.h that was inherited from i386 to support
peter [Tue, 22 Jul 2003 06:44:54 +0000 (06:44 +0000)]
Extend the machine/ieeefp.h that was inherited from i386 to support
the SSE mxcsr register as well.  Since gcc will intermix SSE2 and x87
FP code, the fpsetround() etc mode had better be the same.

There are hooks to enable these inlines to be instantiated inside libc
for non-gcc or C++ callers. (g++ doesn't like the inlines that tried
to extract an integer and convert it to an enum).

21 years agoTurn off the libc/quad functions since they are not needed for amd64
peter [Tue, 22 Jul 2003 06:34:57 +0000 (06:34 +0000)]
Turn off the libc/quad functions since they are not needed for amd64
and just cause lots of warnings.