]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
18 years agoDon't save and restore the ELCR register across suspend and resume for
John Baldwin [Thu, 29 Sep 2005 15:00:09 +0000 (15:00 +0000)]
Don't save and restore the ELCR register across suspend and resume for
the Intel 82371AB PCI-ISA bridge.  We now do this all the time for the
!APIC case in the atpic driver.  This cuts the raw line count for this
driver by about 40%.

MFC after: 1 week

18 years agoI added the device IDs to the header, but not to the table. Add them
Warner Losh [Thu, 29 Sep 2005 14:57:51 +0000 (14:57 +0000)]
I added the device IDs to the header, but not to the table.  Add them
to the table.

MFC After: 3 days

18 years ago- Use PCIR_BAR macro rather than hardcoding 0x10.
John Baldwin [Thu, 29 Sep 2005 14:56:30 +0000 (14:56 +0000)]
- Use PCIR_BAR macro rather than hardcoding 0x10.
- Use pci_enable_busmaster() rather than doing it by hand.
- Remove duplicate if_free().

18 years agoRemove checks for BOOTSIG[23] from FAT32 bootblocks.
Peter Edwards [Thu, 29 Sep 2005 14:09:46 +0000 (14:09 +0000)]
Remove checks for BOOTSIG[23] from FAT32 bootblocks.

There seems to be very little documentary evidence outside this
implementation to suggest a these checks are neccessary, and more
than one camera-formatted flash disk fails the check, but mounts
successfully on most other systems.

Reviewed By: bde@

18 years agoIn em_process_receive_interrupts() store and clear adapter->fmt. This
Gleb Smirnoff [Thu, 29 Sep 2005 13:23:34 +0000 (13:23 +0000)]
In em_process_receive_interrupts() store and clear adapter->fmt. This
make function reenterable. In the runtime the race is masked by serializing
of em_process_receive_interrupts() either by interrupt thread, or by
polling. The race can be triggered when polling is switched on or off.

18 years agoerr() -> errx() where appropriate.
Ruslan Ermilov [Thu, 29 Sep 2005 13:09:04 +0000 (13:09 +0000)]
err() -> errx() where appropriate.

18 years agoReturn zero when disabling watchdog, unless any of the drivers complain.
Poul-Henning Kamp [Thu, 29 Sep 2005 12:31:44 +0000 (12:31 +0000)]
Return zero when disabling watchdog, unless any of the drivers complain.

18 years agoAdd -DINVARIANT_SUPPORT -DINVARIANTS
Max Khon [Thu, 29 Sep 2005 11:56:16 +0000 (11:56 +0000)]
Add -DINVARIANT_SUPPORT -DINVARIANTS
(to commented out CFLAGS, used for debugging).

18 years agoClose a race in biodone(), whereby the bio_done field of the passed
Peter Edwards [Thu, 29 Sep 2005 10:37:20 +0000 (10:37 +0000)]
Close a race in biodone(), whereby the bio_done field of the passed
bio may have been freed and reassigned by the wakeup before being
tested after releasing the bdonelock.

There's a non-zero chance this is the cause of a few of the crashes
knocking around with biodone() sitting in the stack backtrace.

Reviewed By: phk@

18 years agoAdd a note in example as well, that last sector is used for metadata,
Pawel Jakub Dawidek [Thu, 29 Sep 2005 08:56:15 +0000 (08:56 +0000)]
Add a note in example as well, that last sector is used for metadata,
so it don't provoke confusions.

Noticed by: Victor Sudakov <sudakov@sibptus.tomsk.ru>
MFC after: 2 days

18 years agoThe 'intr_bufferfull' driver statistic actually counts the number
Joseph Koshy [Thu, 29 Sep 2005 01:40:03 +0000 (01:40 +0000)]
The 'intr_bufferfull' driver statistic actually counts the number
of times sampling was stopped due to a space shortage; change its
description in the output of `pmcstat -s` to match reality.

MFC after: 3 days

18 years agoBe less IPv4 centric. When checking if the interface is already
Brooks Davis [Wed, 28 Sep 2005 19:59:18 +0000 (19:59 +0000)]
Be less IPv4 centric.  When checking if the interface is already
configured, check if the UP flag is set instead of checking for the
netmask keyword.

18 years agoadd product ID for Linux Ethernet/RNDIS gadget on pxa210/25x/26x.
Hajimu UMEMOTO [Wed, 28 Sep 2005 19:41:25 +0000 (19:41 +0000)]
add product ID for Linux Ethernet/RNDIS gadget on pxa210/25x/26x.

Submitted by: Keiichi SHIMA <keiichi__at__iijlab.net>
MFC after: 2 days

18 years agoImplement suspend/resume methods to be more ACPI friendly.
Pawel Jakub Dawidek [Wed, 28 Sep 2005 19:20:49 +0000 (19:20 +0000)]
Implement suspend/resume methods to be more ACPI friendly.
I'm able to suspend/resume my laptop without this change, but then I need
to wait for the watchdog to reset the card.
With this change, it is ready immediately.

Glanced at by: glebius

18 years agoNuke duplicates found via fuzzy logic:
Jens Schweikhardt [Wed, 28 Sep 2005 18:18:29 +0000 (18:18 +0000)]
Nuke duplicates found via fuzzy logic:
- typos
- different spelling, punctuation, whitespace
- phonetically similar names
- words rearranged ("was once" vs "once was" etc)

If a limerick appeared as a single one and as part of a
double or triple, the singleton was removed.

With a little help from: sort limerick|uniq -d
This still turns up 20 lines being repeated, but the respective
limericks are sufficiently unique to leave them in (i.e. most differ
in at least two lines).

Nuke spaces in front of colons while I'm here.

18 years agopuc(4) does strange things to resources in order to fool the
Poul-Henning Kamp [Wed, 28 Sep 2005 18:06:25 +0000 (18:06 +0000)]
puc(4) does strange things to resources in order to fool the
subdrivers to hook up.

It should probably be rewritten to implement a simple bus to which
the sub drivers attach using some kind of hint.

Until then, provide a couple of crutch functions with big warning
signs so it can survive the recent changes to struct resource.

18 years agoAdd interrupt counters for IPIs. By default they are disabled, but they
John Baldwin [Wed, 28 Sep 2005 18:04:11 +0000 (18:04 +0000)]
Add interrupt counters for IPIs.  By default they are disabled, but they
can be enabled by enabling COUNT_IPIS in smptests.h.  When enabled, each
CPU provides an interrupt counter for nearly all of the IPIs it receives
(IPI_STOP currently doesn't have a counter) that can be examined using
vmstat -i, etc.

MFC after: 3 days
Requested by: rwatson

18 years agoRename the lapic timer interrupt counters from lapicX: timer to cpuX: timer
John Baldwin [Wed, 28 Sep 2005 18:01:41 +0000 (18:01 +0000)]
Rename the lapic timer interrupt counters from lapicX: timer to cpuX: timer
since it's not always obvious that lapic == cpu.

MFC after: 3 days

18 years agoMake it a good-mannered rcNG script respectful to the command line.
Yaroslav Tykhiy [Wed, 28 Sep 2005 16:24:47 +0000 (16:24 +0000)]
Make it a good-mannered rcNG script respectful to the command line.

18 years agosetsockopt() will return -1 on error, not 0.
Olivier Houchard [Wed, 28 Sep 2005 16:14:43 +0000 (16:14 +0000)]
setsockopt() will return -1 on error, not 0.

Pointy hat to: Jeremie Le Hen

18 years agoUse SO_REUSEADDR on the socket, to make the use of syslogd in jails easier.
Olivier Houchard [Wed, 28 Sep 2005 16:05:13 +0000 (16:05 +0000)]
Use SO_REUSEADDR on the socket, to make the use of syslogd in jails easier.

Submitted by:   Jeremie Le Hen <jeremie le-hen org>

18 years agoo Follow the trend and update a host-based wireless bridge example.
Maxim Konovalov [Wed, 28 Sep 2005 15:22:17 +0000 (15:22 +0000)]
o Follow the trend and update a host-based wireless bridge example.

18 years agoFix an endianness issue in pnp_eisaformat(). This corrects printing PnP IDs
Marius Strobl [Wed, 28 Sep 2005 15:01:58 +0000 (15:01 +0000)]
Fix an endianness issue in pnp_eisaformat(). This corrects printing PnP IDs
on big-endian archs like sparc64, e.g.:
uart0: <16550 or compatible> at port 0x3f8-0x3ff irq 43 pnpid @HEd041 on isa0
is now correctly printed as:
uart0: <16550 or compatible> at port 0x3f8-0x3ff irq 43 pnpid PNP0501 on isa0

There are probably other endianness issues lurking in the PnP code which
however aren't exhibited on sparc64 as the PnP devices there are sort of
PnP BIOS devices rather than ISA PnP devices.

Tested on: i386, sparc64
MFC after: 1 week

18 years agoAdd a font width argument to vi_load_font_t, vi_save_font_t and vi_putm_t
Marius Strobl [Wed, 28 Sep 2005 14:54:07 +0000 (14:54 +0000)]
Add a font width argument to vi_load_font_t, vi_save_font_t and vi_putm_t
and do some preparations for handling 12x22 fonts (currently lots of code
implies and/or hardcodes a font width of 8 pixels). This will be required
on sparc64 which uses a default font size of 12x22 in order to add font
loading and saving support as well as to use a syscons(4)-supplied mouse
pointer image.
This API breakage is committed now so it can be MFC'ed in time for 6.0
and later on upcoming framebuffer drivers destined for use on sparc64
and which are expected to rely on using font loading internally and on
a syscons(4)-supplied mouse pointer image can be easily MFC'ed to
RELENG_6 rather than requiring a backport.

Tested on: i386, sparc64, make universe
MFC after: 1 week

18 years agoCommenting out WARNS actually brought it up to 4.
Ruslan Ermilov [Wed, 28 Sep 2005 14:36:16 +0000 (14:36 +0000)]
Commenting out WARNS actually brought it up to 4.

18 years agoNote the requirement for 'device apic' on i386 systems. [1]
Joseph Koshy [Wed, 28 Sep 2005 14:19:31 +0000 (14:19 +0000)]
Note the requirement for 'device apic' on i386 systems. [1]

Consistently use 'x86' when referring to behaviour common to the
i386 and amd64.

Submitted by: Niklas Sorensson <nik@cs.chalmers.se> [1]

18 years agoMinor clarifications and fixes to wording, grammar, and markup
Yaroslav Tykhiy [Wed, 28 Sep 2005 13:07:43 +0000 (13:07 +0000)]
Minor clarifications and fixes to wording, grammar, and markup
around ${network_interfaces}.

18 years agoAdd SHA256 checksums to the release build.
Ken Smith [Wed, 28 Sep 2005 12:39:00 +0000 (12:39 +0000)]
Add SHA256 checksums to the release build.

Requested by: cperciva (so)
Reviewed/improved by: ru
MFC after: 2 days

18 years agoIntroduce "route del" as an alias to "route delete".
Anton Berezin [Wed, 28 Sep 2005 12:12:15 +0000 (12:12 +0000)]
Introduce "route del" as an alias to "route delete".

Reviewed by: arch

18 years agoFix "taskqueue_drain with the following non-sleepable locks held"
Ruslan Ermilov [Wed, 28 Sep 2005 09:27:46 +0000 (09:27 +0000)]
Fix "taskqueue_drain with the following non-sleepable locks held"
witness(9) warning on detach.

18 years agoDo not commit before the first cup of coffee. Other entries don't have a
Max Laier [Wed, 28 Sep 2005 09:19:17 +0000 (09:19 +0000)]
Do not commit before the first cup of coffee. Other entries don't have a
preceeding slash.

Pointed out by: thompsa

18 years agoDocument bridge(4) leftovers.
Max Laier [Wed, 28 Sep 2005 08:25:48 +0000 (08:25 +0000)]
Document bridge(4) leftovers.

Submitted by: maxim

18 years agoRedirect bridge(4) to if_bridge(4) and rename sysctl accordingly.
Max Laier [Wed, 28 Sep 2005 08:18:55 +0000 (08:18 +0000)]
Redirect bridge(4) to if_bridge(4) and rename sysctl accordingly.

Reminded by: ru

18 years agoWe bootstrap make(1) if necessary during an upgrade, so checking
Ruslan Ermilov [Wed, 28 Sep 2005 08:17:30 +0000 (08:17 +0000)]
We bootstrap make(1) if necessary during an upgrade, so checking
if MACHINE_ARCH is defined is no longer needed.

18 years agoRedirect bridge(4) to if_bridge(4). These should have pointed to if_bridge
Max Laier [Wed, 28 Sep 2005 08:11:15 +0000 (08:11 +0000)]
Redirect bridge(4) to if_bridge(4).  These should have pointed to if_bridge
from the begining.

Reminded by: ru

18 years agoOops, revert last commit (the manpage is still built in objdir).
Ruslan Ermilov [Wed, 28 Sep 2005 07:54:18 +0000 (07:54 +0000)]
Oops, revert last commit (the manpage is still built in objdir).

18 years agoAdd NO_OBJ, remove redundant SCRIPTSNAME.
Ruslan Ermilov [Wed, 28 Sep 2005 07:52:27 +0000 (07:52 +0000)]
Add NO_OBJ, remove redundant SCRIPTSNAME.

18 years agoAdd NO_OBJ.
Ruslan Ermilov [Wed, 28 Sep 2005 07:49:30 +0000 (07:49 +0000)]
Add NO_OBJ.

18 years agoUnbreak - remove leftover reference to pcb_rflags.
Peter Wemm [Wed, 28 Sep 2005 07:40:27 +0000 (07:40 +0000)]
Unbreak - remove leftover reference to pcb_rflags.

18 years agoSort LOCALES.
Ruslan Ermilov [Wed, 28 Sep 2005 07:40:03 +0000 (07:40 +0000)]
Sort LOCALES.

18 years agoSort MLINKS.
Ruslan Ermilov [Wed, 28 Sep 2005 07:32:43 +0000 (07:32 +0000)]
Sort MLINKS.

18 years agoSort MAN.
Ruslan Ermilov [Wed, 28 Sep 2005 07:31:18 +0000 (07:31 +0000)]
Sort MAN.

18 years agoSort SUBDIR.
Ruslan Ermilov [Wed, 28 Sep 2005 07:20:46 +0000 (07:20 +0000)]
Sort SUBDIR.

18 years agoBack out alpha/alpha/trap.c:1.124, osf1_ioctl.c:1.14, osf1_misc.c:1.57,
Robert Watson [Wed, 28 Sep 2005 07:03:03 +0000 (07:03 +0000)]
Back out alpha/alpha/trap.c:1.124, osf1_ioctl.c:1.14, osf1_misc.c:1.57,
osf1_signal.c:1.41, amd64/amd64/trap.c:1.291, linux_socket.c:1.60,
svr4_fcntl.c:1.36, svr4_ioctl.c:1.23, svr4_ipc.c:1.18, svr4_misc.c:1.81,
svr4_signal.c:1.34, svr4_stat.c:1.21, svr4_stream.c:1.55,
svr4_termios.c:1.13, svr4_ttold.c:1.15, svr4_util.h:1.10,
ext2_alloc.c:1.43, i386/i386/trap.c:1.279, vm86.c:1.58,
unaligned.c:1.12, imgact_elf.c:1.164, ffs_alloc.c:1.133:

Now that Giant is acquired in uprintf() and tprintf(), the caller no
longer leads to acquire Giant unless it also holds another mutex that
would generate a lock order reversal when calling into these functions.
Specifically not backed out is the acquisition of Giant in nfs_socket.c
and rpcclnt.c, where local mutexes are held and would otherwise violate
the lock order with Giant.

This aligns this code more with the eventual locking of ttys.

Suggested by: bde

18 years agoSimplified markup.
Ruslan Ermilov [Wed, 28 Sep 2005 06:57:56 +0000 (06:57 +0000)]
Simplified markup.

18 years agoComment out WARNS, the OpenSSL headers don't compile cleanly on some platforms.
Dag-Erling Smørgrav [Wed, 28 Sep 2005 06:23:47 +0000 (06:23 +0000)]
Comment out WARNS, the OpenSSL headers don't compile cleanly on some platforms.

18 years agoRemove more OLDCARD references.
Warner Losh [Wed, 28 Sep 2005 06:05:45 +0000 (06:05 +0000)]
Remove more OLDCARD references.

18 years agoCorrect a typo.
Joseph Koshy [Wed, 28 Sep 2005 03:02:16 +0000 (03:02 +0000)]
Correct a typo.

MFC after: 3 days

18 years agoPush Giant down in jails. Pass the MPSAFE flag to NDINIT, and keep track
Christian S.J. Peron [Wed, 28 Sep 2005 00:30:56 +0000 (00:30 +0000)]
Push Giant down in jails. Pass the MPSAFE flag to NDINIT, and keep track
of whether or not Giant was picked up by the filesystem. Add VFS_LOCK_GIANT
macros around vrele as it's possible that this can call in the VOP_INACTIVE
filesystem specific code. Also while we are here, remove the Giant assertion.
from the sysctl handler,  we do not actually require Giant here so we
shouldn't assert it. Doing so will just complicate things when Giant is removed
from the sysctl framework.

18 years agoIf KDB_STOP_NMI is compiled into the kernel, default
Robert Watson [Tue, 27 Sep 2005 21:12:05 +0000 (21:12 +0000)]
If KDB_STOP_NMI is compiled into the kernel, default
debug.kdb.stop_cpus_with_nmi to 1 rather than 0.

MFC after: 3 days

18 years agoI believe the stack underflows during early development that caused me to
Peter Wemm [Tue, 27 Sep 2005 21:11:35 +0000 (21:11 +0000)]
I believe the stack underflows during early development that caused me to
add spare padding at the beginning of the pcb are long gone.  Remove the
padding fields.

18 years agoKill pcb_rflags. It served no purpose.
Peter Wemm [Tue, 27 Sep 2005 21:10:10 +0000 (21:10 +0000)]
Kill pcb_rflags.  It served no purpose.

Reported by:  bde

18 years agoIn lockstatus(), don't lock and unlock the interlock when testing the
Robert Watson [Tue, 27 Sep 2005 21:02:59 +0000 (21:02 +0000)]
In lockstatus(), don't lock and unlock the interlock when testing the
sleep lock status while kdb_active, or we risk contending with the
mutex on another CPU, resulting in a panic when using "show
lockedvnods" while in DDB.

MFC after: 3 days
Reviewed by: jhb
Reported by: kris

18 years ago- Add a work-around for nForce3-250. Aperture base address encoded in misc.
Jung-uk Kim [Tue, 27 Sep 2005 20:57:50 +0000 (20:57 +0000)]
- Add a work-around for nForce3-250.  Aperture base address encoded in misc.
control register and AGP bridge seems to be inconsistent with some BIOS.
Instead of relying on BIOS settings, we just take the initial aperture size
and encode them for both miscellaneous control register and AGP bridge.
Some idea was borrowed from agp_nvidia.c.

- Add preliminary ULi M1689 chipset support.  The idea was taken from Linux
because hardware and documentation are unavailable.  Not tested.

- Add more VIA chipset PCI IDs taken from Linux driver.

Approved by: anholt (mentor)
Tested by: Adam Gregoire <ebola at psychoholics dot org>
Ganael Laplanche <ganael.laplanche at martymac dot com>
K Wieland <kwieland at wustl dot edu>

18 years agoNo longer maintain mbstat statistics for the mbuf allocator, UMA
Robert Watson [Tue, 27 Sep 2005 20:28:43 +0000 (20:28 +0000)]
No longer maintain mbstat statistics for the mbuf allocator, UMA
statistics and libmemstat(3) are now used to track mbuf statistics.

MFC after: 1 month

18 years agos/routnes/routines/ in help message.
Brooks Davis [Tue, 27 Sep 2005 18:52:17 +0000 (18:52 +0000)]
s/routnes/routines/ in help message.

This file is already off the vendor branch and there hasn't been a bc
release in more than 4 years so I can't see any harm in fixing this.

Submitted by: Arne Woerner <arne_woerner at yahoo dot com>
PR: gnu/86627

18 years agoFix a minor nit that has been bugging me for a while. Fix the obvious
Peter Wemm [Tue, 27 Sep 2005 18:32:46 +0000 (18:32 +0000)]
Fix a minor nit that has been bugging me for a while.  Fix the obvious
cases of using a 64 bit operation to zero a register.  32 bit opcodes are
smaller and supposedly faster, and clear the upper 32 bits for free.

18 years agoAdd a bare minimum (but wrong) R_X86_64_JMP_SLOT relocation type for
Peter Wemm [Tue, 27 Sep 2005 18:18:23 +0000 (18:18 +0000)]
Add a bare minimum (but wrong) R_X86_64_JMP_SLOT relocation type for
kernel modules.  We actually need to include any addends and the symbol
offset value, but for gcc/binutils didn't set it anywhere I've found on
'cc -fpic -shared' kernel modules.

18 years agoDon't report Maxmem as 'real memory'. It is really the highest address
Peter Wemm [Tue, 27 Sep 2005 18:15:57 +0000 (18:15 +0000)]
Don't report Maxmem as 'real memory'.  It is really the highest address
available and can give the wrong impression when there are memory holes.
Report the total amount of usable memory that we detected instead of the
highest address.

18 years agoMFi386: If we take a trap with interrupts disabled while in a critical
Peter Wemm [Tue, 27 Sep 2005 18:13:07 +0000 (18:13 +0000)]
MFi386:  If we take a trap with interrupts disabled while in a critical
section, don't enable them if we're servicing an NMI.

18 years agoRemove bridge(4) from the tree. if_bridge(4) is a full functional
Max Laier [Tue, 27 Sep 2005 18:10:43 +0000 (18:10 +0000)]
Remove bridge(4) from the tree.  if_bridge(4) is a full functional
replacement and has additional features which make it superior.

Discussed on: -arch
Reviewed by: thompsa
X-MFC-after: never (RELENG_6 as transition period)

18 years agoDon't let the upper bits of %dr6/%dr7 get set.
Peter Wemm [Tue, 27 Sep 2005 18:10:26 +0000 (18:10 +0000)]
Don't let the upper bits of %dr6/%dr7 get set.

Submitted by:  Nate Eldredge <neldredge@math.ucsd.edu>

18 years agoUse the refcount API to manage the reference count for user credentials
John Baldwin [Tue, 27 Sep 2005 18:09:42 +0000 (18:09 +0000)]
Use the refcount API to manage the reference count for user credentials
rather than using pool mutexes.

Tested on: i386, alpha, sparc64

18 years agoUse the reference count API to manage the reference counts for process
John Baldwin [Tue, 27 Sep 2005 18:07:05 +0000 (18:07 +0000)]
Use the reference count API to manage the reference counts for process
limit structures rather than using pool mutexes to protect the reference
counts.

Tested on: i386, alpha, sparc64

18 years agoRegenerate
Peter Wemm [Tue, 27 Sep 2005 18:04:52 +0000 (18:04 +0000)]
Regenerate

18 years agoImplement 32 bit getcontext/setcontext/swapcontext on amd64. I've added
Peter Wemm [Tue, 27 Sep 2005 18:04:20 +0000 (18:04 +0000)]
Implement 32 bit getcontext/setcontext/swapcontext on amd64.  I've added
stubs for ia64 to keep it compiling.  These are used by 32 bit apps such
as gdb.

18 years agoUse the refcount API to implement reference counts on process argument
John Baldwin [Tue, 27 Sep 2005 18:03:15 +0000 (18:03 +0000)]
Use the refcount API to implement reference counts on process argument
structures rather than using a global mutex to protect the reference
counts.

Tested on: i386, alpha, sparc64

18 years agoAdd a simple reference count API that is simply a thin wrapper API around
John Baldwin [Tue, 27 Sep 2005 18:01:33 +0000 (18:01 +0000)]
Add a simple reference count API that is simply a thin wrapper API around
atomic operations on ints.

Reviewed by: arch@
MFC after: 1 week

18 years ago- Document atomic_fetchadd(9) and add a MLINK.
John Baldwin [Tue, 27 Sep 2005 17:59:29 +0000 (17:59 +0000)]
- Document atomic_fetchadd(9) and add a MLINK.
- Add arm and ppc to the list of archs not supporting operations on 64-bit
  integers.
- Update the sample code for acquiring a mutex to be more recent and to
  take into account the recent atomic_foo_ptr() changes.

MFC after: 1 week

18 years agoAdd a new atomic_fetchadd() primitive that atomically adds a value to a
John Baldwin [Tue, 27 Sep 2005 17:39:11 +0000 (17:39 +0000)]
Add a new atomic_fetchadd() primitive that atomically adds a value to a
variable and returns the previous value of the variable.

Tested on: i386, alpha, sparc64, arm (cognet)
Reviewed by: arch@
Submitted by: cognet (arm)
MFC after: 1 week

18 years agoRestore if_cp.c 1.27
Roman Kurakin [Tue, 27 Sep 2005 16:57:44 +0000 (16:57 +0000)]
Restore if_cp.c 1.27
----------------------------
revision 1.27
date: 2005/09/19 03:10:16;  author: imp;  state: Exp;  lines: +3 -2
Make sure that we call if_free(ifp) after bus_teardown_intr.  Since we
could get an interrupt after we free the ifp, and the interrupt
handler depended on the ifp being still alive, this could, in theory,
cause a crash.  Eliminate this possibility by moving the if_free to
after the bus_teardown_intr() call.

In fact, this change do nothing for this driver. It is protected from
this by cp_destroy variable. This variable also protects driver from initiation
of any activity from network stack with disabled intr handler with this change
applied.

18 years agoBack out fifo_vnops.c:1.127, which introduced an sx lock around I/O on
Robert Watson [Tue, 27 Sep 2005 16:45:22 +0000 (16:45 +0000)]
Back out fifo_vnops.c:1.127, which introduced an sx lock around I/O on
a fifo.  While this did indeed close the race, confirming suspicions
about the nature of the problem, it causes difficulties with blocking
I/O on fifos.

Discussed with: ups
Also spotted by: Peter Holm <peter at holm dot cc>

18 years agoBackout if_cp 1.26, if_ct 1.27, if_cx 1.47 by obrien:
Roman Kurakin [Tue, 27 Sep 2005 16:12:49 +0000 (16:12 +0000)]
Backout if_cp 1.26, if_ct 1.27, if_cx 1.47 by obrien:
----------------------------
revision 1.26
date: 2005/09/07 09:53:35;  author: obrien;  state: Exp;  lines: +1452 -1453
Reorder code to not depend on an ISO-C illegal forward extern declaration.
----------------------------

Reason: do not move large functions location without serious reason. The same
could be done by forward function declaration. Please do not enlarge diff
without a reason any more.

Backout if_cp 1.27
----------------------------
revision 1.27
date: 2005/09/19 03:10:16;  author: imp;  state: Exp;  lines: +3 -2
Make sure that we call if_free(ifp) after bus_teardown_intr.  Since we
could get an interrupt after we free the ifp, and the interrupt
handler depended on the ifp being still alive, this could, in theory,
cause a crash.  Eliminate this possibility by moving the if_free to
after the bus_teardown_intr() call.

Reason: bad previous commit. Would be restored by next commit.

18 years agoUpdate the "created from" section to reflect the most recent version of
Christian S.J. Peron [Tue, 27 Sep 2005 14:36:59 +0000 (14:36 +0000)]
Update the "created from" section to reflect the most recent version of
syscalls.master

Requested by: jhb

18 years agoMark the extended attribute syscalls as being MP safe.
Christian S.J. Peron [Tue, 27 Sep 2005 14:32:04 +0000 (14:32 +0000)]
Mark the extended attribute syscalls as being MP safe.

Requested by: jhb

18 years agoCalling rman_get_start() after bus_release_resource() is evil.
Ruslan Ermilov [Tue, 27 Sep 2005 13:33:46 +0000 (13:33 +0000)]
Calling rman_get_start() after bus_release_resource() is evil.
It became fatal after a recent "struct resource" split change.

18 years agoSwitch from OLDCARD to NEWCARD on pc98.
Yoshihiro Takahashi [Tue, 27 Sep 2005 13:10:24 +0000 (13:10 +0000)]
Switch from OLDCARD to NEWCARD on pc98.

18 years ago- Print space character in show_stat(). Remove a lot of priuntf(" ").
Gleb Smirnoff [Tue, 27 Sep 2005 10:42:02 +0000 (10:42 +0000)]
- Print space character in show_stat(). Remove a lot of priuntf(" ").
- Utilize show_stat() in sidewaysintpr() loop. This makes periodic
  statistics to honor -h flag.

18 years agoFix build.
Gleb Smirnoff [Tue, 27 Sep 2005 09:11:44 +0000 (09:11 +0000)]
Fix build.

18 years agoConvert fxp(4) to use the new bus_alloc_resources() API, it simplifies
Maxime Henrion [Tue, 27 Sep 2005 09:01:11 +0000 (09:01 +0000)]
Convert fxp(4) to use the new bus_alloc_resources() API, it simplifies
the resource allocation code significantly.

18 years agoRemove redundant `\&' escapes.
Ruslan Ermilov [Tue, 27 Sep 2005 08:06:21 +0000 (08:06 +0000)]
Remove redundant `\&' escapes.

18 years agoFixup previous commit.
Ruslan Ermilov [Tue, 27 Sep 2005 08:02:03 +0000 (08:02 +0000)]
Fixup previous commit.

18 years agoDon't print anything if we can't do any localpkg shutdown (start already
Brooks Davis [Tue, 27 Sep 2005 02:05:55 +0000 (02:05 +0000)]
Don't print anything if we can't do any localpkg shutdown (start already
does this).

Submitted by: Andre Albsmeier <Andre dot Albsmeier at siemens dot com>
PR: conf/86606

18 years agoReport status in hex, not decimal.
Scott Long [Tue, 27 Sep 2005 01:59:32 +0000 (01:59 +0000)]
Report status in hex, not decimal.

18 years agoPlease ship the pointy hat to its home position.
Poul-Henning Kamp [Mon, 26 Sep 2005 22:07:59 +0000 (22:07 +0000)]
Please ship the pointy hat to its home position.

18 years agoDocument the IP_DONTFRAG IP socket option.
Andre Oppermann [Mon, 26 Sep 2005 20:58:49 +0000 (20:58 +0000)]
Document the IP_DONTFRAG IP socket option.

Sponsored by:   TCP/IP Optimization Fundraise 2005

18 years agoReplace custom mbuf writeability test with generic M_WRITABLE() test
Andre Oppermann [Mon, 26 Sep 2005 20:35:45 +0000 (20:35 +0000)]
Replace custom mbuf writeability test with generic M_WRITABLE() test
covering all edge cases too.

Sponsored by: TCP/IP Optimization Fundraise 2005

18 years agoUpdate coordinates for "mph" hash generator now that it lives in ports.
Poul-Henning Kamp [Mon, 26 Sep 2005 20:34:21 +0000 (20:34 +0000)]
Update coordinates for "mph" hash generator now that it lives in ports.

18 years agoIncrease WARNS.
Dag-Erling Smørgrav [Mon, 26 Sep 2005 20:34:09 +0000 (20:34 +0000)]
Increase WARNS.

18 years agoCorrect the logic for determining whether the user has already entered
Dag-Erling Smørgrav [Mon, 26 Sep 2005 20:33:53 +0000 (20:33 +0000)]
Correct the logic for determining whether the user has already entered
a password.  Also, work around some harmless type pun warnings.

MFC after: 3 days

18 years agoAdd a .sinclude directive which does the exact same as .include, except
Poul-Henning Kamp [Mon, 26 Sep 2005 20:31:00 +0000 (20:31 +0000)]
Add a .sinclude directive which does the exact same as .include, except
whine when the file cannot be found and opened.

18 years agoImplement IP_DONTFRAG IP socket option enabling the Don't Fragment
Andre Oppermann [Mon, 26 Sep 2005 20:25:16 +0000 (20:25 +0000)]
Implement IP_DONTFRAG IP socket option enabling the Don't Fragment
flag on IP packets.  Currently this option is only repected on udp
and raw ip sockets.  On tcp sockets the DF flag is controlled by the
path MTU discovery option.

Sending a packet larger than the MTU size of the egress interface
returns an EMSGSIZE error.

Discussed with: rwatson
Sponsored by: TCP/IP Optimization Fundraise 2005

18 years agoSmall fixes to sis(4):
John Baldwin [Mon, 26 Sep 2005 18:42:27 +0000 (18:42 +0000)]
Small fixes to sis(4):
- Remove sis_unit and use device_printf() and if_printf() instead.
- Use callout_init_mtx() for the callout.
- Remove spls.
- Fix locking for ifmedia to happen in the ifmedia handlers rather than in
  sis_ioctl().
- Log an error message if we fail to allocate any resources.  Perform
  cleanup if we fail to allocate any resources so that we don't leave
  a mutex hanging around.

Tested by: Jason Tsai jason dot tsai at newcyberian dot com (1-4)
MFC after: 3 days

18 years agoAdd the spin lock used by the binary nvidia driver to the static lock
John Baldwin [Mon, 26 Sep 2005 18:30:12 +0000 (18:30 +0000)]
Add the spin lock used by the binary nvidia driver to the static lock
order list so that WITNESS and the driver play together nicely.

Tested by: Harald Schmalzbauer
MFC after: 3 days

18 years agoEmit a warning when a card matches multiple entries in our table. A
Warner Losh [Mon, 26 Sep 2005 18:27:13 +0000 (18:27 +0000)]
Emit a warning when a card matches multiple entries in our table.  A
number of cards have been discovered to be matching on the strings of
the cis rather than manufacturer/product id for cards we already had a
prod id for.  This is a result of getting the list from the NetBSD
driver which also includes the OID for the cards where such a
distinction mattered (since it was tested against the MAC address we
got from the card).  Since we do not try to match OIDs, we do not need
the extra entries and they just waste space.

I'm guessing that some of the dlink entires (DE-660, DE-660+) and many
of the corega cards may fall into this boat and can safely be removed.

18 years agoGo ahead and detach our children in our detach routine. I'm undecided
Warner Losh [Mon, 26 Sep 2005 18:22:24 +0000 (18:22 +0000)]
Go ahead and detach our children in our detach routine.  I'm undecided
if we should delete them also or not, but have decided not to do so
for the moment.

18 years agoAdd "show allpcpu" to DDB, which prints the current CPU id followed by
Robert Watson [Mon, 26 Sep 2005 16:55:11 +0000 (16:55 +0000)]
Add "show allpcpu" to DDB, which prints the current CPU id followed by
the per-cpu data for all CPUs.  This is easier to ask users to do than
"figure out how many CPUs you have, now run show pcpu, then run it
once for each CPU you have".

MFC after: 3 days

18 years agoThis one goes as well.
Poul-Henning Kamp [Mon, 26 Sep 2005 14:58:58 +0000 (14:58 +0000)]
This one goes as well.

18 years agoRemove the old Makefile based nanobsd build method, the new shell script
Poul-Henning Kamp [Mon, 26 Sep 2005 14:57:27 +0000 (14:57 +0000)]
Remove the old Makefile based nanobsd build method, the new shell script
based is far better and the only one supported from now on.

18 years agoReorder statements to avoid accessing unknown memory.
David Xu [Mon, 26 Sep 2005 14:14:55 +0000 (14:14 +0000)]
Reorder statements to avoid accessing unknown memory.
In theory, invoking kenv with very long string can panic
kernel.