-fschedule-insns as it wasn't such a big win with 2.95 after all.
Add the *BIG* win "-mpreferred-stack-boundary=2" optimiztion submitted by
Dima. GCC 2.95 ensures the stack frame is always properly [opitimally]
aligned by surrounding every function call by code simular to
"addl $-12, %esp" / "addl $12, %esp". Here we need the reduction in space,
with speed not an issue.
All Makefiles now use MACHINE_ARCH for the target architecture.
Unification is required for cross-building.
Tags added to:
sys/boot/Makefile
sys/boot/arc/loader/Makefile
sys/kern/Makefile
usr.bin/cpp/Makefile
usr.bin/gcore/Makefile
usr.bin/truss/Makefile
usr.bin/gcore/Makefile:
fixed typo: MACHINDE -> MACHINE_ARCH
Historically, the documentation of extended asm was lacking, namely you
should NOT specify the same register as an input, and a clobber.
If the register is clobbered, it should be specified as an output as well,
e.g., by linking input and output through the "number" notation.
(Beware of lvalues, some local variables needed...)
URL:http://egcs.cygnus.com/faq.html
In versions up to egcs-1.1.1, the compiler did not even warn about it,
but it was liable to output bad code. Newer egcs are pickier and simply
refuse to swallow such code.
Note, since *addr changes, it needs to be an output operand.
We might be excessive in saying that all memory has changed.
Obtained from: OpenBSD
w/extra thanks to Marc Espie <Marc.Espie@liafa.jussieu.fr>
Remove some printf() calls, reduce size of buffers, and abbreviate
some strings.
Hopefully the boot people will fix this spamage after the cut over to
Gcc 2.95.2 as the system compiler.
either one gives us an additional 32 bytes of additional space available
when using EGCS 1.1.2. With GCC 2.95.2 -fforce-addr gives us 12 more bytes,
and adding -fschedule-insns gives us an additional 4 bytes.
Note: Previous commit to these files (except coda_vnops and devfs_vnops)
that claimed to remove WILLRELE from VOP_RENAME actually removed it from
VOP_MKNOD.
Without this, ioctl commands for setting formats and speeds were
essentially ignored for simplex devices until the application actually
performed a read or write.
* Make sure that both channels are set in the SB mixer code and provide a
mixer table specifically for the ess18xx which supports the extended
accuracy available on this part.
* Fix a stupid bug in ess_format() which ignored the passed-in format and
changed the hardware based on the value which was set last time. This
meant that the hardware setting was often not set correctly at all.
* Add a custom identify driver for the ESS1888 which automagically detects
and adds the device in a pseudo-PnP way. This driver also emits the magic
sequence which enables the sound hardware after a hard reset, allowing
it to work correctly for the sound hardware of a PWS 433au (and probably
all other PWS class alpha machines).
With these changes, I was able to play back simple sounds on my 433au. I
have not tested recording or any other formats other than 8bit ulaw and
16bit stereo.
Make tuner on French SECAM Hauppauge cards work
Submitted by: Cyrille Lefevre <clefevre@citeweb.net>
Fix clipping bugs ready for Xv support in XFree86 4.0
Submitted by: Thomas Mueller <tmueller@thmu.de>
Fix the 'feature' where /dev/vbi needed to be opened before /dev/bktr
when using Teletext with Alevt and FXTV.
Requested by: Randall Hopper
Rename a few (I wish companies would stop buying each other)
Add a quirk entry for hubs that say they are self powered but are
in fact bus powered (usage in uhub follows shortly).
condition for Short transfers.
Change the scheduling to Depth first. We now transfer as many TD's as
possible from QH before moving to the next queue (Breadth first). It should
still be verified that this does not lead to starvation in a busy system
(in the case were transfers are added to the beginning of the control
or bulk queues).
during autoconfig to support strange hardware (such as the Yamaha DS-1)
which implements 'legacy' ISA devices as well as a PCI device. This will
allow the PCI driver for the YMF724 to add the legacy devices to the ISA
bus and will allow the PnP system to automatically allocate the resources
for those devices.
The old code was spread out through the machdep code and was sloppy about
enabling and disabling the FEN bit (which controls access to the FP
register set). This caused a DIAGNOSTIC warning "DANGER WILL ROBINSON:
FEN SET IN cpu_fork!" sometimes when operating under high loads and could
conceivably lead to processes getting incorrect FP results.
The new code is much more strict about the FEN bit and makes sure that
*only* fpcurproc ever has it enabled. This also allows us to remove a
section of code from the exception_return path which might improve
performance marginally.
Reviewed by: gallatin
instead of duplicating the code. (2) If a wired page is passed
to vm_page_free_toq, panic instead of printing a friendly warning.
(If we don't panic here, we'll just panic later in vm_page_unwire
obscuring the problem.)
Correctly lock vnodes when calling VOP_OPEN() from filesystem mount code.
Unify spec_open() for bdev and cdev cases.
Remove the disabled bdev specific read/write code.
It however posts a bogus button up event once in a while. Whenever
we receive dx=dy=dz=buttons=0 we postpone adding it to the queue for
50msecs with a timeout. If in the meantime something else is posted
the event is ignored.
This avoids the problem Nik Sayer reported. He noticed that X windows
would drop and pick up a window once in a while.
Thanks, Nik, for supplying me with the keyboard to fix the problem!
Try to use a 32bit mask on the IO addresses, this fixes the alpha
and hopefully doesn't break on any i386 machines.
Try to enable both read & write cache on disks, they should be as
default, but better be sure..
when returning an error. Bug fix was extracted from the PR. The PR
is not yet entirely resolved by this commit.
PR: kern/13049
Reviewed by: Matt Dillon <dillon@freebsd.org>
Submitted by: Ian Dowse <iedowse@maths.tcd.ie>
dev nodes with make_dev() in a devfs-compatable way by supplying names.
This doesn't fit into a compile-time scheme like DEV_DRIVER_MODULE,
especially since the cdevsw[] array's days are numbered.
devsw_module_handler() indirectly and not use the chain arguments. To
eliminate this indirection via that function (which does nothing now)
without duplicating a modevent handler into all the routines that don't
presently have one, supply a NOP (do nothing, return OK) routine which
is functionally equivalent to what's there now. This is a hack and is
still wrong, because there doesn't appear to be anything to reclaim
resources on an unload of a module with one of these in it. I'm not
sure whether to make the NOP handler refuse a MOD_UNLOAD event or what.
file object. Also explain some possible directions to re-implement it --
I'm not sure it should be, given the minimal application use. (Other
than having the debugger automatically access the symbols for a process,
the main use I'd found was with some minor accounting ability, but _that_
depends on it being in the filesystem space; an ioctl access method would
be useless in that case.)
This is a code-less change; only a comment has been added.
Document the options available for the ata driver.
Disconnect the atapi devices from the old wd driver to avoid conflicts
(they will go away at some point anyways)
continue doing it despite objections by me (the principal author).
Note that this doesn't fix the real problem -- the real problem is generally
bad setup by ignorant users, and education is the right way to fix it.
So while this doesn't actually solve the prolem mentioned in the complaint
(since it's still possible to do it via other methods, although they mostly
involve a bit more complicity), and there are better methods to do this,
nobody was willing or able to provide me with a real world example that
couldn't be worked around using the existing permissions and group
mechanism. And therefore, security by removing features is the method of
the day.
I only had three applications that used it, in any event. One of them would
have made debugging easier, but I still haven't finished it, and won't
now, so it doesn't really matter.
* GC unused options
* Move options that exist on all architectures to conf/options
* Add missing options to LINT
* Sort undocumented options list in LINT
Reviewed by: green
Fix a bug which could cause panics in ad/atapi-interrupt.
Add support for UDMA66 on Promise Ultra/Fasttrak controllers.
Get rid of ATA_IGNORE_INTR, and introduce ATA_WAIT_INTR instead.
Add a delay in the dump routine in ata-disk.c, some controllers
seem to need this. Also dont use the timeout watchdog when dumping.
Disable DMA on ATAPI devices as default, add option ATA_ENABLE_ATAPI_DMA
for those that has HW that works.
Add support for some not-up-to-spec ATAPI devices that returns data
together with completition status on data moving cmd's.
data is copyin()'ed to a properly aligned buffer.)
Return EFAULT rather than EINVAL if the sigcontext is inaccessible,
as specified by the documentation.
Submitted by: bde
We currently only search SCSI and IDE CDROMs; if there's felt to be a
need for supporting the very old and rare soundcard etc. drives for this
application they can be trivially added.
the soft updates changes: only report the link count to be i_effnlink
in ufs_getattr() for file systems that maintain i_effnlink.
Tested by: Mike Dracopoulos <mdraco@math.uoa.gr>
In order to achieve this, root filesystem mount is moved from
SI_ORDER_FIRST to SI_ORDER_SECOND in the SI_SUB_MOUNT_ROOT sysinit
group. Now, modules which wish to usurp the default root mount
can use SI_ORDER_FIRST.
A compiled-in or preloaded MFS filesystem will become the root
filesystem unless the vfs.root.mountfrom environment variable refers
to a valid bootable device. This will normally only be the case when
the kernel and MFS image have been loaded from a disk which has a
valid /etc/fstab file. In this case, the variable should be manually
overridden in the loader, or the kernel booted with -a. In either
case "mfs:" should be supplied as the new value.
Also fix a typo in one DFLTROOT case that would not have compiled.
can deal with from 1GB to 2GB. I have no way to test with >1GB because I
don't have high enough density memory to get any one box over 1GB. However, I
have tested it on xp1000 & ds10 with < 1GB of memory & have verified that it
does no harm.
filesystem is discovered. Preference is given to using the kernel
environment variable vfs.root.mountfrom, which is set by the loader
according to the contents of /etc/fstab. Changes in the MD code
provide fallback mechanisms for systems not using the loader.
A more robust fallback path is also provided, with the last recourse
being to prompt on the console for a root device.
These changes drastically simplify the machine-dependant parts of
the root configuration process. In addition, support for CDROM root
devices has been removed; it was a nasty hack and didn't work.
Mike says the whole idea of a current device was a bad idea in first place,
and will be doing away with currdev.
Anyway, people are not supposed to even notice this. :-)
be ignored by default by the df(1) program. This is used mostly to
avoid stat()-ing entries that do not represent "real" disk mount
points (such as those made by an automounter such as amd.) It is
also useful not to have to stat() these entries because it takes
longer to report them that for other file systems, being that these
mount points are served by a user-level file server and resulting in
several context switches. Worse, if the automounter is down
unexpectedly, a causal df(1) will hang in an interruptible way.
PR: kern/9764
Submitted by: Erez Zadok <ezk@cs.columbia.edu>
only two conflicts, cdev #98 and cdev #99. These should be fixed.
MAKEDEV should probably be merged as well.
Static majors are (hopefully) going away one day soon.
This file is informational and not machine parsed by anything any more.
to config(8) for static device tables that have not existed for quite
some time. They have been aliases for 'device' for a while, and "tape"
went away entirely as it wasn't used anywhere (except in an example
in LINT.. "fixed").
camcontrol was setup to use the old scheme of going through the xpt(4)
device, which never worked properly (and has been disabled for a while).
camcontrol now sends BDRs through the pass(4) device, and XPT_RESET_DEV
CCBs are put on the device queue in the transport layer, as they should be.
Submitted by: luoqi
Reviewed by: ken
If the ipl is lowered here, a machine can rapidly run out of stack
space when it is under heavy interrupt load. Without this fix,
my test machine would fall over within a few seconds when forwarding
14,000 packets/sec. With the patch, it has been up for over 24 hours
and has built the world at the same time.
submitted by: dfr@freebsd.org
the countdown register.
this should not be necessary but there are broken laptops that
do not restore the countdown register on resume.
when it happnes, it messes up the hardclock interval and system clock,
which leads to the infamous "calcru: negative time" problem.
Submitted by: kjc, iwasaki
Reviewed by: Steve O'Hara-Smith <steveo@eircom.net> and committers.
Obtained from: PAO3
eliminate an extra (useless) level of indirection in half of the page
queue accesses and (2) to use a single name for each queue throughout,
instead of, e.g., "vm_page_queue_active" in some places and
"vm_page_queues[PQ_ACTIVE]" in others.
Reviewed by: dillon
"rw" argument, rather than hijacking B_{READ|WRITE}.
Fix two bugs (physio & cam) resulting by the confusion caused by this.
Submitted by: Tor.Egge@fast.no
Reviewed by: alc, ken (partly)
This is one piece of the grand unified PPP daemon concept, whereby using
netgraph nodes enables PPP data to be handled completely in kernel land,
while leaving negotiation/control to be handled by a single user land
daemon, no matter what the link type(s).
This is a safety checkin only; it compiles, but is utterly untested.
Concept reviewed by: julian, brian
Merge the contents (less some trivial bordering the silly comments)
of <vm/vm_prot.h> and <vm/vm_inherit.h> into <vm/vm.h>. This puts
the #defines for the vm_inherit_t and vm_prot_t types next to their
typedefs.
This paves the road for the commit to follow shortly: change
useracc() to use VM_PROT_{READ|WRITE} rather than B_{READ|WRITE}
as argument.
go to REQ_SENT (and we probably should also log this since it should
only happen in a cross-linked connection).
Submitted by: Mark Tinguely <tinguely@plains.NoDak.edu>
test this support since I don't have the dongle for the card in
question and the dongle I have for my 3C589D doesn't seem to work on
it. I don't know if this is due to the damage I did to the 562 card
in transit from freebsdcon, or a mis-matched dongle...
Update AverMedia GPIO values
Submitted by: AverMedia
Add support for WinTV Theater Dolby Surround Sound DPL3518A chip
Submitted by: Frank Nobis <fn@radio-do.de>
Make PLL mode the default for Bt878s. You no longer need options BKTR_USE_PLL
pccard attachments must activate the resources they want to access.
ep didn't do this, so of course thee eeprom came back as busy. ed and
sio already did this. It was only due to a bug in the logic that the
probe succeeded. These problems have been fixed.
Note to self; when converting a driver to newbus, the foodriver bits
must be removed from sys/i386/isa/isa_compat.h
Reminded gently by: Mitsuru IWASAKI <iwasaki@jp.FreeBSD.org>
- Rip out all the static softc stuff and do softc allocation the right way.
- Rewrite most of the ISA code so that it provides a DEVICE_IDENTIFY
method to enumerate all non-PnP ISA devices.
This has the following consequences:
- No 'ep' devices may be hardwired.
- All hardwired devices will probably be detected twice.
By hardwired I mean:
device ep0 at isa? port 0x300 irq 10
- 'ep' devices are ordered by bus, slot, and then MAC address.
- Make 3c509B cards work in PnP mode. Yes, they really work.
- Convert over to using ifmedia for media selection. No more of this
lame 'linkX' stuff.
- Consolidate a lot of duplicated code.
- Make a stab at not breaking MII based PCCARD devices.
I doubt that the PCCARD stuff works any more than it did before my
changes but theres hope. My PCCARD hardware should arrive in a
week or so.
- Retreive the media settings from the card EEPROM rather than guessing.
I've got a 3c509-TPO that thinks its got an AUI port and if others
can report similar problems I'll write a bit of clever code that will
fix this but right now it works correctly on all but 1 card.
- Clean up a few things and make some cosmetic changes.
- Add myself as the MAINTAINER since nobody else wants to. I'm
in the best position to do this as I've got an example of most
of the cards:
EISA 3c579 bnc/aui
MCA 3c529 tp/aui
ISA 3c509 tpo
ISA-PnP 3c509B combo
If someone wants to send me a any cards I don't have I'd appriciate
it. Also welcome are 3c59x boards since I'll be folding if_vx and
if_ep at some point.
Shift to using the same queueing strategy that the amr driver uses.
Some simple tests indicate that we use about 2% of the CPU at around
500tps with the controller completely saturated with I/O.
the time spent at splbio(). We now avoid it unless we are actually
manipulating the command queues themselves. This doesn't improve
performance noticeably, but should improve concurrency somewhat.
(yet) compile and link. Renamed pcic back to pcic from pcicx, but
conditionalize its inclusion on pccard being included also. card is
the old and pccard is the new, which is a handy way to have both in
the tree at the same time.
Obtained from: newconfig project
More to follow...
