When sending cache flushing IPIs, don't try to IPI the triggering CPU
itself; this causes undefined behaviour on UltraSPARCs. In particular,
the interrupt packet data words will not necessarily be delivered
correctly, which would result in a crash.
This bug also caused the cache-flushing work to be done twice on the
triggering CPU (when it did not cause crashes).
Moved the `distribute' target from bsd.obj.mk to bsd.subdir.mk,
to make it call `install' in the bsd.subdir.mk-driven makefiles
too. (share/examples/Makefile,v 1.29 changed the bsd.prog.mk
to bsd.subdir.mk and many stuff was lost during "make release".
I then merged this change in rev. 1.28.2.2 to work around the
namespace pollution (FILES) in this makefile.)
There was an added complexity here. Both the `distribute' and
`install' targets are recursive (they propagate to SUBDIRs).
So `distribute' first calls `install' in the ${.CURDIR}, then
calls `distribute' in each SUBDIR, etc. The problem is that
`install' (being also recursive) causes the stuff from SUBDIR
to be installed twice, first time thru `install' in ${.CURDIR}
triggered by `distribute', second time by `distribute' run in
the SUBDIR. This problem is not new, but it became apparent
only after I moved the `distribute' target from bsd.obj.mk to
bsd.subdir.mk. My first attempt testing the fix failed due to
this, because the whole world was distributed twice, causing
all the imaginable mess (kerberos5 stuff was installed into both
"base" and "krb5" dists, there was /sbin/init.bak, etc.)
I say the problem is not new because bsd.prog.mk and bsd.lib.mk
makefiles with SUBDIR (even without this fix) had this problem
for years. Try e.g. running ``make distribute DISTDIR=/foo''
from usr.bin/bzip2 or from lib/libcom_err (without the fix) and
watch the output.
So the solution was to make `install' behave non-recursive when
executed by `distribute'. My first attempt in passing SUBDIR=
to the `install' in the `distribute' body failed because of the
way how src/Makefile and src/Makefile.inc1 communicate with each
other. SUBDIR='s assignment precedence on the "make install
SUBDIR=" command line is lowered after src/Makefile wrapper calls
"make ... -f ${.CURDIR}/Makefile.inc1 install" because SUBDIR=
is moved into environment, and Makefile.inc1's assignments now
take higher precedence. This may be fixed someday when we merge
Makefile with Makefile.inc1. For now, this is implemented as a
NO_SUBDIR knob.
Spotted by: Dmitry Pryanishnikov <dmitry@atlantis.dp.ua>
Prodded by: des
MFC after: 3 days
Make dirty, rotten hack really work. As of rev. 1.16, ${BINDIR}
is only defined after <bsd.prog.mk> is included, and .if make(1)
conditionals are evaluted on the first pass.
Spotted by: Michael Bretterklieber <mbretter@jawa.at>
peter [Fri, 12 Jul 2002 07:56:11 +0000 (07:56 +0000)]
Revive backed out pmap related changes from Feb 2002. The highlights are:
- It actually works this time, honest!
- Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive,
so try and optimize things where possible.
- Introduce ranged shootdowns that can be done as a single IPI.
- PG_G support for i386
- Specific-cpu targeted shootdowns. For example, there is no sense in
globally purging the TLB cache for where we are stealing a page from
the local unshared process on the local cpu. Use pm_active to track
this.
- Add some instrumentation for the tlb shootdown code.
- Rip out SMP code from <machine/cpufunc.h>
- Try and fix some very bogus PG_G and PG_PS interactions that were bad
enough to cause vm86 bios calls to break. vm86 depended on our existing
bugs and this was the cause of the VESA panics last time.
- Fix the silly one-line error that caused the 'panic: bad pte' last time.
- Fix a couple of other silly one-line errors that should have caused more
pain than they did.
Some more work is needed:
- pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we
have a hook in cpu_switch.
- The IPI handlers need some cleanup. I have a bogus %ds load that can
be avoided.
- APTD handling is rather bogus and appears to be a large source of
global TLB IPI shootdowns for no really good reason.
I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop.
I expect to see a bigger difference when there is significant pageout
activity or the system otherwise has memory shortages.
I have backed out a few optimizations that I had been using over the last
few days in order to be a little more conservative. I'll revisit these
again over the next few days as the dust settles.
New option: DISABLE_PG_G - In case I missed something.
alfred [Fri, 12 Jul 2002 06:51:57 +0000 (06:51 +0000)]
Create a bug-for-bug FreeBSD4 compatible version of sendfile and move the
fixed sendfile over. This is needed to preserve binary compatibility from
4.x to 5.x.
alfred [Fri, 12 Jul 2002 06:38:34 +0000 (06:38 +0000)]
Introduce syscall.master option 'COMPAT4' which allows one to wrap
syscalls for FreeBSD 4 compatibility.
Add kernel option COMPAT_FREEBSD4 to enable these syscalls.
Add two variables to struct jobqueue, and change the way that getq()
calculates how much space to get for that struct, so it will get the
right amount when new variables are added.
Move prototypes for ctl_readcf and ctl_freeinf from ctlinfo.c to ctlinfo.h,
so the routines can be called by an upcoming change for 'lpc topq/bottomq'.
This is basically a ``C compilation'' of the former whereis.pl file,
employing the same algorithms, and aiming at being mostly
UI-compatible to the old (legally tainted) 4.3BSD whereis(1). In
comparision, the 4.4BSD-Lite version is just another variant of
which(1) only, where in particular the option to search for source
directories is sorely missing.
While i was at it, i added two more options which i contemplated doing
long since. -x will suppress the run of locate(1) to find sources
that could not be found otherwise, potentially saving a lot of time
(but obviously, risking to not find some sources that are well hidden
in the tree). -q will omit the leading name of the query, so in
particular, you can now do something like:
cd `whereis -qs ls`
I'd explicitly like to thank johan for his review which was quite a
bit more than an average review, including sending me a lot of diffs.
Fixed misspelling of "hint." as "hints." in the description of the "hint."
keyword and in the description of rp's hints.
Didn't fix rp's hints being mostly in comments so that they are harder to
use (they don't get linted either way because makeLINT.sh strips them and
there is no compile-time syntax checking of hints anyway).
Set NO_WERROR to ignore the following warning which is emitted on
alphas:
.../elf2aout.c:130: warning: cast increases required alignment of
target type
The warning is about casting ((char *)e + phoff) to a struct pointer,
where e is aligned but phoff might be garbage, so I think the warning
should be emitted on most machines (even on i386's, alignment checking
might be on) and the correct fix would involve validation phoff before
using it.
Fixed 4 printf format errors that were fatal on alphas. %qd is not even
suitable for printing quad_t's since it is equivalent to %lld but quad_t
is unsigned long on alphas. quad_t shouldn't be used anyway.
alfred [Thu, 11 Jul 2002 17:39:50 +0000 (17:39 +0000)]
Fix return values in the sm_notify_1 service routine to return an answer
most of the time (unless fork fails). This should fix the problem where
FreeBSD won't respond to a remote host and therefor the remote hosts
tries indefinitely to contact the FreeBSD hosts thereby irritating the
system administrator.
Uncommented WARNS=0. ipfw2.c is full of printf format errors that are
fatal on alphas.
Fixed setting of WARNS. WARNS should never be set unconditionally, since
this breaks testing of different WARNS values by setting it at a higher
level (e.g., on the command line).
Fixed a printf format error that was fatal on alphas. Adding WFORMAT=0
to the Makefile didn't affect this bug because WFORMAT only controls
higher- level format checking (not the -Wformat that is implicit in
-Wall).
Fixed a nearby printf format error that was benign and 3 nearby style bugs.
Pass -DBOOTSTRAPPING to the kernel's build tool (sys/dev/aic7xxx/aicasm).
This way, it has a chance to be built with gcc 2.95.x (using the bandaid
in share/mk/bsd.cpu.mk,v 1.11).
Do not override the CFLAGS with its standard value. At some point,
different architectures may choose to use different default values
for CFLAGS, for example. (It was added in rev. 1.200 as a measure
to make boot images fit the floppies, and was never reverted.)
Do not override the standard `distribute' target that is currently
available from bsd.obj.mk.
The native version was identical (and pretty much unused except in
the -DMODULES_WITH_WORLD case, which it is not for "make release")
except that the "bin" -> "base" change of the default DISTRIBUTION
name did not propagate here.
Don't assume the shell's controlling terminal is attached to file descriptor
2. Instead, open /dev/tty. This problem stopped commands in subshells from
being executed correctly if standard error was redirected.
Remove support for the "old" tty driver by unifdef -UOLD_TTY_DRIVER;
many other parts of the shell are no longer compatible with this, and it
makes jobs.c quite cluttered with #ifdef's.