]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
21 years agoMFi386 rev 1.51 by scottl: make dflt_lock() always panic.
Peter Wemm [Mon, 22 Sep 2003 23:11:42 +0000 (23:11 +0000)]
MFi386 rev 1.51 by scottl:  make dflt_lock() always panic.

21 years agoMFi386 rev 1.53 by scottl: Allocate the S/G list in the tag, not on
Peter Wemm [Mon, 22 Sep 2003 23:10:24 +0000 (23:10 +0000)]
MFi386 rev 1.53 by scottl:  Allocate the S/G list in the tag, not on
the stack.  This means that s/g lists can be arbitrarily long.

21 years agoMFi386 machdep.c rev 1.201, clock.c 1.201, clock.h 1.45 by phk: Dont
Peter Wemm [Mon, 22 Sep 2003 23:02:24 +0000 (23:02 +0000)]
MFi386 machdep.c rev 1.201, clock.c 1.201, clock.h 1.45 by phk:  Dont
initialize a TSC timecounter until we know if it is broke or not.

XXX I think there is a bug in the i386 code here.  init_TSC_tc() comes
after:
  if (statclock_disable)
    return;

ie: if you turn off the statclock interrupt, you dont get the TSC either.

21 years agoMFi386 rev 1.105 by jhb: fix comment typo
Peter Wemm [Mon, 22 Sep 2003 22:54:14 +0000 (22:54 +0000)]
MFi386 rev 1.105 by jhb: fix comment typo

21 years agoMFi386 rev 1.256 by jhb: remove redundant #include <sys/sysctl.h>
Peter Wemm [Mon, 22 Sep 2003 22:52:46 +0000 (22:52 +0000)]
MFi386 rev 1.256 by jhb:  remove redundant #include <sys/sysctl.h>

21 years agoMFi386 rev 1.25 by jhb: add new MSR's and some missing older ones and
Peter Wemm [Mon, 22 Sep 2003 22:51:46 +0000 (22:51 +0000)]
MFi386 rev 1.25 by jhb: add new MSR's and some missing older ones and
APICBASE MSR constants.

21 years agoMFi386 rev 1.55 by sam: remove unused #define BUS_DMAMAP_NSEGS
Peter Wemm [Mon, 22 Sep 2003 22:43:21 +0000 (22:43 +0000)]
MFi386 rev 1.55 by sam: remove unused #define BUS_DMAMAP_NSEGS

21 years agoMFi386 rev 1.37: constant-friendly bswap macros
Peter Wemm [Mon, 22 Sep 2003 22:37:49 +0000 (22:37 +0000)]
MFi386 rev 1.37: constant-friendly bswap macros

21 years agoMFi386: pci_cfgreg.h rev 1.10 by jhb/des/njl. Fix CONF1_ENABLE_MSK.
Peter Wemm [Mon, 22 Sep 2003 22:21:21 +0000 (22:21 +0000)]
MFi386: pci_cfgreg.h rev 1.10 by jhb/des/njl.  Fix CONF1_ENABLE_MSK.

21 years agoMFCi386: trap.c rev 1.257 by bde. Don't forget to reenable interrupts
Peter Wemm [Mon, 22 Sep 2003 22:19:59 +0000 (22:19 +0000)]
MFCi386: trap.c rev 1.257 by bde.  Don't forget to reenable interrupts
for breakpoint and trace traps from usermode.  Although all the setidt
entries are interrupt gates on amd64, all but the trace and bpt trap
entry handlers reenable interrupts after the swapgs instruction in order
to simulate the trap/interrupt gate distinction.  In other words, the
amd64 code behaves the same way that i386 does here.

21 years agoMFi386 by jhb: add acpi_SetDefaultIntrModel();
Peter Wemm [Mon, 22 Sep 2003 22:12:46 +0000 (22:12 +0000)]
MFi386 by jhb:  add acpi_SetDefaultIntrModel();

21 years agoMFi386 by jhb: use symbolic constants for the IDT entries.
Peter Wemm [Mon, 22 Sep 2003 22:09:02 +0000 (22:09 +0000)]
MFi386 by jhb: use symbolic constants for the IDT entries.

21 years agoMFi386: machdep.c:1.570 clock.c:1.204 by bde: Quick fix for calling DELAY
Peter Wemm [Mon, 22 Sep 2003 21:56:48 +0000 (21:56 +0000)]
MFi386: machdep.c:1.570 clock.c:1.204 by bde:  Quick fix for calling DELAY
for ddb input in some atkbd-based console drivers.  ddb must not use any
normal locks but DELAY() normally calls getit() which needs clock_lock.
This also removes the need for recursion on clock_lock.

21 years agoMicrooptimization to allow the compiler to evaluate ntohl() etc on
Peter Wemm [Mon, 22 Sep 2003 21:46:47 +0000 (21:46 +0000)]
Microoptimization to allow the compiler to evaluate ntohl() etc on
known constants at compile time rather than at run time.  We have a number
of nasty hacks around the place to cache ntohl() of constants (eg: nfs).
This change allows the compiler to compile-time evaluate ntohl(1) as
0x01000000 rather than having to emit assembler code to do it.  This
has other smaller flow-on effects because the compiler can see that
ntohl(constant) itself has a constant value now and can propagate the
compile time evaluation.

Obtained from:  Ideas from NetBSD and Linux, and some code from NetBSD

21 years agoSimplify the KOBJOPLOOKUP macro for the non-debug case so that gcc's
Peter Wemm [Mon, 22 Sep 2003 21:32:49 +0000 (21:32 +0000)]
Simplify the KOBJOPLOOKUP macro for the non-debug case so that gcc's
heuristics do not overestimate the code size quite so much.

21 years agoFix inverted BURN_BRIDGES and GONE_IN_5 #ifdefs.
Thomas Moestl [Mon, 22 Sep 2003 17:58:40 +0000 (17:58 +0000)]
Fix inverted BURN_BRIDGES and GONE_IN_5 #ifdefs.

21 years agoPrevious commit contained too-smart-for-its-own-good code that might
Dag-Erling Smørgrav [Mon, 22 Sep 2003 16:05:11 +0000 (16:05 +0000)]
Previous commit contained too-smart-for-its-own-good code that might
produce incorrect (though harmless) output on single-CPU systems.

21 years agoFake multi-cpu statistics for proc/stat by dividing the totals by the
Dag-Erling Smørgrav [Mon, 22 Sep 2003 15:52:32 +0000 (15:52 +0000)]
Fake multi-cpu statistics for proc/stat by dividing the totals by the
number of CPUs.

PR: kern/27522

21 years agoSave and restore timeout field for signal frame just like what we did
David Xu [Mon, 22 Sep 2003 14:40:36 +0000 (14:40 +0000)]
Save and restore timeout field for signal frame just like what we did
for interrupted field.
Also in _thr_sig_handler, retrieve current signal mask from kernel not
from ucp, the later is pre-unioned mask, not current signal mask.

21 years agoFix FPU state restoring bug by jumping to right position.
David Xu [Mon, 22 Sep 2003 14:34:02 +0000 (14:34 +0000)]
Fix FPU state restoring bug by jumping to right position.

21 years agoSoften assert in pmap_remove_all.
Peter Grehan [Mon, 22 Sep 2003 11:59:05 +0000 (11:59 +0000)]
Soften assert in pmap_remove_all.
Introduct pmap_extract_and_hold.

Stolen from: sparc64

21 years agoATAng requires <sys/taskqueue.h>
Peter Grehan [Mon, 22 Sep 2003 11:55:02 +0000 (11:55 +0000)]
ATAng requires <sys/taskqueue.h>

21 years agodocument Sony PEGA-WL110
Warner Losh [Mon, 22 Sep 2003 05:36:32 +0000 (05:36 +0000)]
document Sony PEGA-WL110

# I think this is prism 3 based, but I could be wrong

21 years agoSony PEGA-WL110 Wireless LAN
Warner Losh [Mon, 22 Sep 2003 05:33:22 +0000 (05:33 +0000)]
Sony PEGA-WL110 Wireless LAN

Obtained from: NetBSD (onoe-san)

21 years agosync to 1.72
Warner Losh [Mon, 22 Sep 2003 05:27:33 +0000 (05:27 +0000)]
sync to 1.72

21 years agoMerge from NetBSD:
Warner Losh [Mon, 22 Sep 2003 05:27:03 +0000 (05:27 +0000)]
Merge from NetBSD:
1.186: onoe; Sony's PEGA-WL110 CF WLAN (which strangely has fujitsu's
   vendor id)
1.185: ichiro; Quatech Inc, PCMCIA Enhanced Parallel Port Card

Also:
o update $NetBSD$
o minor tweaks to FUJITSU.  We've tried to keep the CIS only entries seprate
  from vendor id/product id.

21 years agoAdd an off_t cast to prevent overflow on media size. This should
Nate Lawson [Mon, 22 Sep 2003 04:59:07 +0000 (04:59 +0000)]
Add an off_t cast to prevent overflow on media size.  This should
fix issues with geometry not matching drive size.

Submitted by: Clive Lin <clive@tongi.org>

21 years ago - vm_hold_free_pages() should lock the kernel object. (The pages being
Alan Cox [Mon, 22 Sep 2003 04:58:09 +0000 (04:58 +0000)]
 - vm_hold_free_pages() should lock the kernel object.  (The pages being
   freed belong to the kernel object.)
 - Increase the granularity of the vm object locking in vm_hold_load_pages()
   in order to reduce the number of times that we acquire and release the
   same lock.

21 years agoFix the ACPI_DEBUG case.
Nate Lawson [Mon, 22 Sep 2003 04:50:29 +0000 (04:50 +0000)]
Fix the ACPI_DEBUG case.

Reported by: Larry Rosenman <ler@lerctr.org>
PR:
Submitted by:
Reviewed by:
Approved by:
Obtained from:
MFC after:

21 years agoInitialize the page's pindex field even for VM_ALLOC_NOOBJ allocations.
Alan Cox [Mon, 22 Sep 2003 00:56:13 +0000 (00:56 +0000)]
Initialize the page's pindex field even for VM_ALLOC_NOOBJ allocations.
(This field is useful for implementing sanity checks even if the page does
not belong to an object.)

21 years agoPrint waitset correctly.
David Xu [Mon, 22 Sep 2003 00:40:23 +0000 (00:40 +0000)]
Print waitset correctly.

21 years agochange bridge mib usage to reflect new world order
Sam Leffler [Sun, 21 Sep 2003 23:16:02 +0000 (23:16 +0000)]
change bridge mib usage to reflect new world order

Prodded by: Ruslan Ermilov <ru@FreeBSD.org>

21 years agorevert to version 1.25 and use va_copy to obtain another copy of the
Paul Saab [Sun, 21 Sep 2003 22:14:49 +0000 (22:14 +0000)]
revert to version 1.25 and use va_copy to obtain another copy of the
variable arguments. version 1.26 incorrectly truncated the message if
the buffer was too long.

Requested by: bde

21 years agoAllocate the page table directory page(s) as "no object" pages. (This
Alan Cox [Sun, 21 Sep 2003 21:36:13 +0000 (21:36 +0000)]
Allocate the page table directory page(s) as "no object" pages.  (This
leaves one explicit use of the pte object.)

21 years agoPut x11-wm/sawfish2 and irc/xchat on CD 1 instead of sawfish and xchat.
Joe Marcus Clarke [Sun, 21 Sep 2003 19:07:23 +0000 (19:07 +0000)]
Put x11-wm/sawfish2 and irc/xchat on CD 1 instead of sawfish and xchat.
Sawfish is gone, and xchat2 is more widely used these days since it is the
current stable release.

Approved by: re (murray)

21 years agoDocument the -x and -y options.
Poul-Henning Kamp [Sun, 21 Sep 2003 19:05:35 +0000 (19:05 +0000)]
Document the -x and -y options.

21 years agoMake the clone handler BURN_BRIDGES but the actual 'a' and 'c' partition
Poul-Henning Kamp [Sun, 21 Sep 2003 18:57:12 +0000 (18:57 +0000)]
Make the clone handler BURN_BRIDGES but the actual 'a' and 'c' partition
check GONE_IN_5:  We need the clone handler for root filesystem case.

Once under GEOM, we can remove the clone handler as GEOM provides one.

21 years agoMake this compile.
Poul-Henning Kamp [Sun, 21 Sep 2003 18:54:52 +0000 (18:54 +0000)]
Make this compile.

21 years ago(atapi_cb): Reorganize error handling circuitry.
Thomas Quinot [Sun, 21 Sep 2003 16:49:53 +0000 (16:49 +0000)]
(atapi_cb): Reorganize error handling circuitry.
 Fix to the messages output under CAM_DEBUG_CCB: the summary sense
 information (error bits and sense key) is in the error field, not
 in the result field, of struct ata_request. No other functional change.

21 years ago(atapi_cb): Disable autosense for now as it causes a missed ATA interrupt.
Thomas Quinot [Sun, 21 Sep 2003 08:53:05 +0000 (08:53 +0000)]
(atapi_cb): Disable autosense for now as it causes a missed ATA interrupt.
Minor reformatting.

21 years ago(scsi_request_sense): Set allocation length in REQUEST_SENSE CCB.
Thomas Quinot [Sun, 21 Sep 2003 08:45:36 +0000 (08:45 +0000)]
(scsi_request_sense): Set allocation length in REQUEST_SENSE CCB.

Reviewed by: ken

21 years ago(camperiphdone): When the cam_periph layer performs sense recovery,
Thomas Quinot [Sun, 21 Sep 2003 08:42:32 +0000 (08:42 +0000)]
(camperiphdone): When the cam_periph layer performs sense recovery,
 completion of recovery is indicated by positioning the CAM_AUTOSNS_VALID
 bit in the status field of the CCB, not in the flags field.
This fixes an endless loop of sense recovery actions.

Reviewed by: ken

21 years agoThe ports freeze may take longer than anticipated. Instead of
Daniel Eischen [Sun, 21 Sep 2003 07:59:16 +0000 (07:59 +0000)]
The ports freeze may take longer than anticipated.  Instead of
waiting for it to be delayed, temporarily back out the -pthread
removal until the freeze is lifted.

Freeze possibly taking longer than necessary: will
Requested by: kris

21 years ago - Fix MD_SMALL_ALLOC on architectures that support it. Define a new alloc
Jeff Roberson [Sun, 21 Sep 2003 07:39:16 +0000 (07:39 +0000)]
 - Fix MD_SMALL_ALLOC on architectures that support it.  Define a new alloc
   function, startup_alloc(), that is used for single page allocations prior
   to the VM starting up.  If it is used after the VM startups up, it
   replaces the zone's allocf pointer with either page_alloc() or
   uma_small_alloc() where appropriate.

Pointy hat to: me
Tested by: phk/amd64, me/x86

21 years agoMake LOCAL_PATCHES accept multiple patch files, as its name suggests.
Ruslan Ermilov [Sun, 21 Sep 2003 06:33:14 +0000 (06:33 +0000)]
Make LOCAL_PATCHES accept multiple patch files, as its name suggests.

21 years agoOnly print an end '}' if the field was non-zero (i.e. there were some flags
Nate Lawson [Sun, 21 Sep 2003 03:51:48 +0000 (03:51 +0000)]
Only print an end '}' if the field was non-zero (i.e. there were some flags
to print).

21 years agoAdd support for fixed event buttons defined in the DSDT (HID "ACPI_FSB"
Nate Lawson [Sun, 21 Sep 2003 02:49:59 +0000 (02:49 +0000)]
Add support for fixed event buttons defined in the DSDT (HID "ACPI_FSB"
and "ACPI_FPB").

Pointed out by: Linux

21 years agoReimplement pmap_release() such that it uses the page table rather than the
Alan Cox [Sat, 20 Sep 2003 23:54:36 +0000 (23:54 +0000)]
Reimplement pmap_release() such that it uses the page table rather than the
pte object to locate the page table directory pages.  (This is another step
toward the elimination of the pte object.)

21 years agoFix improper use of varargs.
Paul Saab [Sat, 20 Sep 2003 23:35:37 +0000 (23:35 +0000)]
Fix improper use of varargs.

Reviewed by: peter

21 years agoBad Jeffr! No cookie!
Peter Wemm [Sat, 20 Sep 2003 23:35:33 +0000 (23:35 +0000)]
Bad Jeffr! No cookie!

Temporarily disable the UMA_MD_SMALL_ALLOC stuff since recent commits
break sparc64, amd64, ia64 and alpha.  It appears only i386 and maybe
powerpc were not broken.

21 years agoFix copyright (Wind River has no rights to this file).
Bill Paul [Sat, 20 Sep 2003 21:42:56 +0000 (21:42 +0000)]
Fix copyright (Wind River has no rights to this file).

Add Gigabyte 7N400 Pro2 motherboard integrated gigE to list of
supported devices.

21 years agoUpdate the copyright year.
Ruslan Ermilov [Sat, 20 Sep 2003 21:32:26 +0000 (21:32 +0000)]
Update the copyright year.

21 years agoKeep up with recent mdoc(7) changes.
Ruslan Ermilov [Sat, 20 Sep 2003 21:23:53 +0000 (21:23 +0000)]
Keep up with recent mdoc(7) changes.

21 years agoComment out and mark all non-translated library strings.
Ruslan Ermilov [Sat, 20 Sep 2003 21:22:19 +0000 (21:22 +0000)]
Comment out and mark all non-translated library strings.

21 years agoSync with FSF branch.
Ruslan Ermilov [Sat, 20 Sep 2003 21:20:01 +0000 (21:20 +0000)]
Sync with FSF branch.

21 years agoRemove the dual-address cycle stuff. DAC is used to allow a bus master
Bill Paul [Sat, 20 Sep 2003 21:18:27 +0000 (21:18 +0000)]
Remove the dual-address cycle stuff. DAC is used to allow a bus master
device to access 64-bit addresses from a 32-bit PCI bus. While the
RealTek manual says you can set this bit and the chip will perform
DAC only if you give it a DMA address with any of the upper 32
bits set, this appears not to be the case. If I turn on the DAC
bit, the chip sets the 'system error' bit in the status register
when I to do a DMA on my Athlon test box with 32-bit PCI bus (VIA
chipset) even though I only have 128MB of physical memory, and thus
can never give the chip a 64-bit address.

Obviously, I can't just set it and forget it, so until I figure
out the right rule for when it's safe/necessary to enable it, keep
it turned off.

21 years agoThis commit was generated by cvs2svn to compensate for changes in r120297,
Ruslan Ermilov [Sat, 20 Sep 2003 21:13:43 +0000 (21:13 +0000)]
This commit was generated by cvs2svn to compensate for changes in r120297,
which included commits to RCS files with non-trunk default branches.

21 years agoKeep up with recent mdoc(7) changes, that include:
Ruslan Ermilov [Sat, 20 Sep 2003 21:13:43 +0000 (21:13 +0000)]
Keep up with recent mdoc(7) changes, that include:

- If no .Os call is made, don't print the OS info at all.
- Allow for multiple .Lb calls in the LIBRARY section.
- Improved documentation on punctuation characters.

21 years agoFix the last remaining problem encountered by KSE: apparently it is
Marcel Moolenaar [Sat, 20 Sep 2003 20:34:58 +0000 (20:34 +0000)]
Fix the last remaining problem encountered by KSE: apparently it is
not guaranteed that the RSE writes the NaT collection immediately,
sort of atomically, to the backing store when it writes the register
immediately prior to the NaT collection point. This means that we
cannot assume that the low 9 bits of the backingstore pointer do not
point to the NaT collection. This is rather a surprise and I don't
know at this time if it's a bug in the Merced or that it's actually
a valid condition of the architecture. A quick scan over the sources
does not indicate that we depend on the false assumption elsewhere,
but it's something to keep in mind.

The fix is to write the saved contents of the ar.rnat register to
the backingstore prior to entering the loop that copies the dirty
registers from the kernel stack to the user stack.

21 years agoAdd vendor Samsung and device Samsung ML-6060.
Stefan Eßer [Sat, 20 Sep 2003 20:01:08 +0000 (20:01 +0000)]
Add vendor Samsung and device Samsung ML-6060.

21 years agoMove uma_small_alloc() and uma_small_free() to uma_machdep.c. These
Marcel Moolenaar [Sat, 20 Sep 2003 19:27:48 +0000 (19:27 +0000)]
Move uma_small_alloc() and uma_small_free() to uma_machdep.c. These
functions reference UMA internals from <vm/uma_int.h>, which makes
them highly unwanted in non-UMA specific files.

While here, prune the includes in pmap.c and use __FBSDID(). Move
the includes above the descriptive comment.

The copyright of uma_machdep.c is assigned to the project and can
be reassigned to the foundation if and when when such is preferrable.

21 years agoEXPsys PCMCIA Ethernet Combo
Warner Losh [Sat, 20 Sep 2003 18:03:07 +0000 (18:03 +0000)]
EXPsys PCMCIA Ethernet Combo

Submitted by: Ludwig Pummer

21 years agoupdate to 1.71
Warner Losh [Sat, 20 Sep 2003 18:01:51 +0000 (18:01 +0000)]
update to 1.71

21 years agoEXPsys PCMCIA Ethernet Combo
Warner Losh [Sat, 20 Sep 2003 17:54:30 +0000 (17:54 +0000)]
EXPsys PCMCIA Ethernet Combo

Submitted by: Ludwig Pummer

21 years agoRemove unused #defines.
Warner Losh [Sat, 20 Sep 2003 17:46:39 +0000 (17:46 +0000)]
Remove unused #defines.

21 years agoThe method link_preload_finish is not static.
Doug Rabson [Sat, 20 Sep 2003 17:39:32 +0000 (17:39 +0000)]
The method link_preload_finish is not static.

21 years agoRemove an invalid KASSERT. Apparently pmap_remove_all gets called on
Jake Burkholder [Sat, 20 Sep 2003 17:00:59 +0000 (17:00 +0000)]
Remove an invalid KASSERT.  Apparently pmap_remove_all gets called on
unmanaged pages.

21 years agoAdd missing file, it is sorely needed to make if_sk.c compile again.
Wilko Bulte [Sat, 20 Sep 2003 15:49:22 +0000 (15:49 +0000)]
Add missing file, it is sorely needed to make if_sk.c compile again.

21 years agoFix width of first column in FILES table to account for longer device name.
Stefan Eßer [Sat, 20 Sep 2003 15:16:56 +0000 (15:16 +0000)]
Fix width of first column in FILES table to account for longer device name.

21 years agoFix name of no-reset device (ulpn -> unlpt).
Stefan Eßer [Sat, 20 Sep 2003 15:13:35 +0000 (15:13 +0000)]
Fix name of no-reset device (ulpn -> unlpt).

21 years agoCorrect card type of Netgear WG311 and add WAG311 and WGT511T
Stefan Eßer [Sat, 20 Sep 2003 15:10:38 +0000 (15:10 +0000)]
Correct card type of Netgear WG311 and add WAG311 and WGT511T
which are based on the AR5212 and should just work (not verified).
Add Proxim Skyline 4032, the PCI version of th e4030.
Add revision suffix 'B' to D-Link DWL-G520/G650 entries, in order
to indicate that revision A1 cards are not supported by this driver
(both A1 and B1/B2 cards are sold in identical boxes).
Explicitly point out the existence of unsupported DWL-G520/G650
(rev. A1) cards in the CAVEATS section.
Approved by: sam

21 years agoNote added support for 3C940 and Marvell Semiconductor 88E1000* gigabit PHY
Wilko Bulte [Sat, 20 Sep 2003 11:10:51 +0000 (11:10 +0000)]
Note added support for 3C940 and Marvell Semiconductor 88E1000* gigabit PHY

Submitted by: Jung-uk Kim <jkim@niksun.com>
Based on: Nathan L. Binkert's OpenBSD patch for sk(4)

21 years agoAdd support for SK-9521 V2.0 and 3COM 3C940.
Wilko Bulte [Sat, 20 Sep 2003 10:53:08 +0000 (10:53 +0000)]
Add support for SK-9521 V2.0 and 3COM 3C940.

Tested at 100Mbit only, using Asus P4P800 onboard 3C940.
The -stable version of this patch I have in use for ~2 weeks now, and works
just fine for me.

Based on: Nathan L. Binkert's patch for OpenBSD
Patch submitted by and thanks to: Jung-uk Kim <jkim@niksun.com>
MFC after: 2 weeks

21 years agofixed memory leak.
SUZUKI Shinsuke [Sat, 20 Sep 2003 09:41:13 +0000 (09:41 +0000)]
fixed memory leak.

Obtained From: KAME

21 years agoDont reject a master device if it said slave failed.
Søren Schmidt [Sat, 20 Sep 2003 08:38:33 +0000 (08:38 +0000)]
Dont reject a master device if it said slave failed.

21 years agoIn umass_bbb_get_max_lun() move maxlun to before the DPRINTF so that
Gary Jennejohn [Sat, 20 Sep 2003 08:18:16 +0000 (08:18 +0000)]
In umass_bbb_get_max_lun() move maxlun to before the DPRINTF so that
the actual number of LUNs is printed.

21 years agoNew Proxim Harmony OEM card.
Warner Losh [Sat, 20 Sep 2003 05:27:18 +0000 (05:27 +0000)]
New Proxim Harmony OEM card.

Submitted by: Jeremy Bingham

21 years agosync to 1.70
Warner Losh [Sat, 20 Sep 2003 05:24:50 +0000 (05:24 +0000)]
sync to 1.70

21 years agoOEM version of Proxim Harmony card
Warner Losh [Sat, 20 Sep 2003 05:23:50 +0000 (05:23 +0000)]
OEM version of Proxim Harmony card

Submitted by: Jeremy Bingham

21 years agoA new proxim harmony oem card spotted in the field.
Warner Losh [Sat, 20 Sep 2003 05:21:58 +0000 (05:21 +0000)]
A new proxim harmony oem card spotted in the field.

Submitted by: Jeremy Bingham

21 years agoo Properly spell my last name. Bad Hiten, no cookie :-)
Marcel Moolenaar [Sat, 20 Sep 2003 04:13:16 +0000 (04:13 +0000)]
o  Properly spell my last name. Bad Hiten, no cookie :-)
o  Use the email address I put in my signature.

21 years ago - Somewhere along the line I stupidly removed critical logic from
Jeff Roberson [Sat, 20 Sep 2003 02:05:58 +0000 (02:05 +0000)]
 - Somewhere along the line I stupidly removed critical logic from
   sched_ptcpu_update().  This caused erroneous cpu times in TOP for
   processes that were asleep.  Replace the code that was removed.

21 years ago - In reassignbuf() don't unlock vp and lock newvp if they are the same.
Jeff Roberson [Sat, 20 Sep 2003 00:21:48 +0000 (00:21 +0000)]
 - In reassignbuf() don't unlock vp and lock newvp if they are the same.
   Doing so creates a race where the buf is on neither list.
 - Only vfree() in an error case in vclean() if VSHOULDFREE() thinks we
   should.
 - Convert the error case in vclean() to INVARIANTS from DIAGNOSTIC as this
   really should not happen and is fast to check.

21 years agoNew errata: SA-03:13.
Bruce A. Mah [Sat, 20 Sep 2003 00:17:52 +0000 (00:17 +0000)]
New errata:  SA-03:13.

21 years agoNew release note: SA-03:13.
Bruce A. Mah [Sat, 20 Sep 2003 00:01:29 +0000 (00:01 +0000)]
New release note:  SA-03:13.

Modified release notes:  GNOME 2.4, KDE 3.1.4.

21 years ago - Remove spls(). The locking that has replaced them is in place and they
Jeff Roberson [Fri, 19 Sep 2003 23:52:06 +0000 (23:52 +0000)]
 - Remove spls().  The locking that has replaced them is in place and they
   no longer serve as guidelines for future work.

21 years ago - Remove interlock protection around VI_XLOCK. The interlock is not
Jeff Roberson [Fri, 19 Sep 2003 23:37:49 +0000 (23:37 +0000)]
 - Remove interlock protection around VI_XLOCK.  The interlock is not
   sufficient to guarantee that this race is not hit.  The XLOCK will likely
   have to be redesigned due to the way reference counting and mutexes work
   in FreeBSD.  We currently can not be guaranteed that xlock was not set
   and cleared while we were blocked on the interlock while waiting to check
   for XLOCK.  This would lead us to reference a vnode which was not the
   vnode we requested.
 - Add a backtrace() call inside of INVARIANTS in the hopes of finding out if
   this condition is ever hit.  It should not, since we should be retaining
   a reference to the vnode in these cases.  The reference would be sufficient
   to block recycling.

21 years agoMake KSE_STACKSIZE machine dependent by moving it from thr_kern.c to
Marcel Moolenaar [Fri, 19 Sep 2003 23:28:13 +0000 (23:28 +0000)]
Make KSE_STACKSIZE machine dependent by moving it from thr_kern.c to
pthread_md.h. This commit only moves the definition; it does not
change it for any of the platforms. This more easily allows 64-bit
architectures (in particular) to pick a slightly larger stack size.

21 years ago - Remove the working-set algorithm. Instead, use the per cpu buckets as the
Jeff Roberson [Fri, 19 Sep 2003 23:27:46 +0000 (23:27 +0000)]
 - Remove the working-set algorithm.  Instead, use the per cpu buckets as the
   working set cache.  This has several advantages.  Firstly, we never touch
   the per cpu queues now in the timeout handler.  This removes one more
   reason for having per cpu locks.  Secondly, it reduces the size of the zone
   by 8 bytes, bringing it under 200 bytes for a single proc x86 box.  This
   tidies up other logic as well.
 - The 'destroy' flag no longer needs to be passed to zone_drain() since it
   always frees everything in the zone's slabs.
 - cache_drain() is now only called from zone_dtor() and so it destroys by
   default.  It also does not need the destroy parameter now.

21 years agoAdd a release note for the sendmail 8.12.10 import
Gregory Neil Shapiro [Fri, 19 Sep 2003 23:23:55 +0000 (23:23 +0000)]
Add a release note for the sendmail 8.12.10 import

21 years agoUpdate for 8.12.10 import
Gregory Neil Shapiro [Fri, 19 Sep 2003 23:20:46 +0000 (23:20 +0000)]
Update for 8.12.10 import

21 years agoFix sendmail 8.12.10 import conflicts
Gregory Neil Shapiro [Fri, 19 Sep 2003 23:14:57 +0000 (23:14 +0000)]
Fix sendmail 8.12.10 import conflicts

21 years agoThis commit was generated by cvs2svn to compensate for changes in r120256,
Gregory Neil Shapiro [Fri, 19 Sep 2003 23:11:30 +0000 (23:11 +0000)]
This commit was generated by cvs2svn to compensate for changes in r120256,
which included commits to RCS files with non-trunk default branches.

21 years agoImport sendmail 8.12.10
Gregory Neil Shapiro [Fri, 19 Sep 2003 23:11:30 +0000 (23:11 +0000)]
Import sendmail 8.12.10

21 years ago - Remove the cache colorization code. We can't use it due to all of the
Jeff Roberson [Fri, 19 Sep 2003 23:04:44 +0000 (23:04 +0000)]
 - Remove the cache colorization code.  We can't use it due to all of the
   broken consumers of the malloc interface who assume that the allocated
   address will be an even multiple of the size.
 - Remove disabled time delay code on uma_reclaim().  The comment there said
   it all.  It was not an effective strategy and it should not be left in
   #if 0'd for all eternity.

21 years ago_ia64_break_setcontext() now takes a mcontext_t. While here, define
Marcel Moolenaar [Fri, 19 Sep 2003 23:00:28 +0000 (23:00 +0000)]
_ia64_break_setcontext() now takes a mcontext_t. While here, define
THR_SETCONTEXT as PANIC(). The THR_SETCONTEXT macro is currently not
used, which means that the definition we had could be wrong, overly
pessimistic or unknowingly right. I don't like the odds...

The new _ia64_break_setcontext() and corresponding kernel fixes make
KSE mostly usable. There's still a case where we don't properly
restore a context and end up with a NaT consumption fault (typically
an indication for not handling NaT collection points correctly),
but at least now mutex_d works...

21 years agoStop using the setcontext() syscall to restore an async context.
Marcel Moolenaar [Fri, 19 Sep 2003 22:54:05 +0000 (22:54 +0000)]
Stop using the setcontext() syscall to restore an async context.
Instead use the break instruction with an immediate specially
created for us.

21 years agoFix the most significant KSE breakage caused by not restoring the
Marcel Moolenaar [Fri, 19 Sep 2003 22:51:26 +0000 (22:51 +0000)]
Fix the most significant KSE breakage caused by not restoring the
restart instruction bits in the PSR. As such, we were returning
from interrupt to the instruction in the bundle that caused us
to enter the kernel, only now we're returning to a completely
different bundle.

While close here: add two KASSERTs to make sure that we restore
sync contexts only when entered the kernel through a syscall and
restore an async context only when entered the kernel through an
interrupt, trap or fault.

While not exactly here, but close enough: use suword64() when we
copy the dirty registers from the kernel stack to the user stack.
The code was intended to be be replaced shortly after being added,
but that was a couple of weeks ago. I might as well avoid that it
is a source for panics until it's replaced.

21 years agoAdd HP 4400C, Belkin F5U208 VideoBusII
John Birrell [Fri, 19 Sep 2003 22:42:51 +0000 (22:42 +0000)]
Add HP 4400C, Belkin F5U208 VideoBusII

21 years agoRevamp trap(): make it more explicit which kinds of traps/faults we
Marcel Moolenaar [Fri, 19 Sep 2003 22:41:52 +0000 (22:41 +0000)]
Revamp trap(): make it more explicit which kinds of traps/faults we
can get (or not) and what we do with them. This fixes the behaviour
for NaT consumption and speculation faults in that we now don't panic
for user faults.

Remove the dopanic label and move the code to a function. This makes
it easier in the simulator to set a breakpoint.

While here, remove the special handling of the old break-based syscall
path and move it to where we handle the break vector. While here,
reserve a new break immediate for KSE. We currently use the old break-
based syscall to deal with restoring async contexts. However, it has
the side-effect of also setting the signal mask and callong ast() on
the way out. The new break immediate simply restores the context and
returns without calling ast().