]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
20 years ago/etc/rc.serial -> /etc/rc.d/serial.
Tom Rhodes [Tue, 23 Dec 2003 07:16:38 +0000 (07:16 +0000)]
/etc/rc.serial -> /etc/rc.d/serial.

20 years agoRe-do the handling of ndis_buffers. The NDIS_BUFFER structure is
Bill Paul [Tue, 23 Dec 2003 04:08:22 +0000 (04:08 +0000)]
Re-do the handling of ndis_buffers. The NDIS_BUFFER structure is
supposed to be opaque to the driver, however it is exposed through
several macros which expect certain behavior. In my original
implementation, I used the mappedsystemva member of the structure
to hold a pointer to the buffer and bytecount to hold the length.
It turns out you must use the startva pointer to point to the
page containing the start of the buffer and set byteoffset to
the offset within the page where the buffer starts. So, for a buffer
with address 'baseva,' startva is baseva & ~(PAGE_SIZE -1) and
byteoffset is baseva & (PAGE_SIZE -1). We have to maintain this
convention everywhere that ndis_buffers are used.

Fortunately, Microsoft defines some macros for initializing and
manipulating NDIS_BUFFER structures in ntddk.h. I adapted some
of them for use here and used them where appropriate.

This fixes the discrepancy I observed between how RX'ed packet sizes
were being reported in the Broadcom wireless driver and the sample
ethernet drivers that I've tested. This should also help the
Intel Centrino wireless driver work.

Also try to properly initialize the 802.11 BSS and IBSS channels.
(Sadly, the channel value is meaningless since there's no way
in the existing NDIS API to get/set the channel, but this should
take care of any 'invalid channel (NULL)' messages printed on
the console.

20 years agoRegen (should be a NOP except for rcsid)
Peter Wemm [Tue, 23 Dec 2003 04:07:47 +0000 (04:07 +0000)]
Regen (should be a NOP except for rcsid)

20 years agoGC unused namespc column.
Peter Wemm [Tue, 23 Dec 2003 04:07:22 +0000 (04:07 +0000)]
GC unused namespc column.

20 years agoRegen (should be a NOP except for rcsid changes)
Peter Wemm [Tue, 23 Dec 2003 03:55:06 +0000 (03:55 +0000)]
Regen (should be a NOP except for rcsid changes)

20 years agoGC unused third namespace column.
Peter Wemm [Tue, 23 Dec 2003 03:54:40 +0000 (03:54 +0000)]
GC unused third namespace column.

20 years agoGC unused namespace column. Unwrap some long lines that now fit.
Peter Wemm [Tue, 23 Dec 2003 03:53:21 +0000 (03:53 +0000)]
GC unused namespace column.  Unwrap some long lines that now fit.

20 years agoRegen - this should be essentially a NOP, except for rcsid changes.
Peter Wemm [Tue, 23 Dec 2003 03:52:14 +0000 (03:52 +0000)]
Regen - this should be essentially a NOP, except for rcsid changes.

20 years agoRemove namespc column and attempt to un-fold some of the longer lines
Peter Wemm [Tue, 23 Dec 2003 03:51:36 +0000 (03:51 +0000)]
Remove namespc column and attempt to un-fold some of the longer lines
that now fit.

20 years agoRemove the namespace column from the syscalls tables. We don't actually
Peter Wemm [Tue, 23 Dec 2003 03:50:43 +0000 (03:50 +0000)]
Remove the namespace column from the syscalls tables.  We don't actually
use it, if we ever did.  They have been been VERY poorly maintained for
some time, possibly because they were a NOP.  FWIW, This brings our table
formats back closer to the other *BSD's.

20 years agoRegen
Peter Wemm [Tue, 23 Dec 2003 03:21:49 +0000 (03:21 +0000)]
Regen

20 years agofreebsd32_fstat(2) is now MPSAFE
Peter Wemm [Tue, 23 Dec 2003 03:21:06 +0000 (03:21 +0000)]
freebsd32_fstat(2) is now MPSAFE

20 years agoRather than screw around with the (unsafe) stackgap, call vn_stat/fo_stat
Peter Wemm [Tue, 23 Dec 2003 03:20:49 +0000 (03:20 +0000)]
Rather than screw around with the (unsafe) stackgap, call vn_stat/fo_stat
directly for stat/fstat/lstat syscall emulation.  It turns out not only
safer, but the code is smaller this way too.

20 years agoRegen
Peter Wemm [Tue, 23 Dec 2003 02:48:58 +0000 (02:48 +0000)]
Regen

20 years agoEliminate stackgap usage for the (woefully incomplete) path translations
Peter Wemm [Tue, 23 Dec 2003 02:48:11 +0000 (02:48 +0000)]
Eliminate stackgap usage for the (woefully incomplete) path translations
since it isn't needed here anymore.
Use standard open(2)/access(2) and chflags(2) syscalls now.

20 years agoForced commit; previous commit also included:
Peter Wemm [Tue, 23 Dec 2003 02:45:24 +0000 (02:45 +0000)]
Forced commit; previous commit also included:
- eliminate a malloc()/snprintf()/free() in the native exec(2) case and
  in the easy emulation environments.
- Allow the brand emul_path (ie: /compat/xxx) to be NULL rather than
  needing it to be an empty string that is always referenced.

20 years agoAdd an additional field to the elf brandinfo structure to support
Peter Wemm [Tue, 23 Dec 2003 02:42:39 +0000 (02:42 +0000)]
Add an additional field to the elf brandinfo structure to support
quicker exec-time replacement of the elf interpreter on an emulation
environment where an entire /compat/* tree isn't really warranted.

20 years agoAdd a reminder note about removing the amd64 test here once the gcc33 port
Peter Wemm [Tue, 23 Dec 2003 02:38:55 +0000 (02:38 +0000)]
Add a reminder note about removing the amd64 test here once the gcc33 port
has been updated.

20 years agoCatch a few places where NULL (pointer) was used where 0 (integer) was
Peter Wemm [Tue, 23 Dec 2003 02:36:43 +0000 (02:36 +0000)]
Catch a few places where NULL (pointer) was used where 0 (integer) was
expected.

20 years agoDon peril sensitive sunglasses and set NULL to an actual pointer type,
Peter Wemm [Tue, 23 Dec 2003 02:34:25 +0000 (02:34 +0000)]
Don peril sensitive sunglasses and set NULL to an actual pointer type,
but *only* for the kernel.  We can do this because the kernel is not a
standard C application environment.  This would have stopped the recent
mtx_* arg NULL/MTX_DEF mixups from going unnoticed for so long.

20 years agoDon't use NULL (pointer) when we're testing for a count of 0 (integer).
Peter Wemm [Tue, 23 Dec 2003 02:29:46 +0000 (02:29 +0000)]
Don't use NULL (pointer) when we're testing for a count of 0 (integer).

20 years agoDon't use NULL (pointer) when we mean 0 (integer) for the number of ticks
Peter Wemm [Tue, 23 Dec 2003 02:28:42 +0000 (02:28 +0000)]
Don't use NULL (pointer) when we mean 0 (integer) for the number of ticks
in msleep.

20 years agoRemove the Minolta item, it already exists in the uscanner(4) manual page.
Tom Rhodes [Mon, 22 Dec 2003 23:20:58 +0000 (23:20 +0000)]
Remove the Minolta item, it already exists in the uscanner(4) manual page.

Noticed by: simon

20 years agoUpdate the supported device list.
Tom Rhodes [Mon, 22 Dec 2003 23:01:25 +0000 (23:01 +0000)]
Update the supported device list.

Discussed with: sanpei

20 years agoFix a couple of stylistic issues
Maksim Yevmenkin [Mon, 22 Dec 2003 22:50:21 +0000 (22:50 +0000)]
Fix a couple of stylistic issues

Reviewed by: imp (mentor), ru
Approved by: imp (mentor)

20 years agoAdd Epson Perfection 1250 and Minolta 5400 to the list of supported
Simon L. B. Nielsen [Mon, 22 Dec 2003 20:23:43 +0000 (20:23 +0000)]
Add Epson Perfection 1250 and Minolta 5400 to the list of supported
devices.

20 years agoresort device list
MIHIRA Sanpei Yoshiro [Mon, 22 Dec 2003 19:58:27 +0000 (19:58 +0000)]
resort device list

20 years agoFixed incomplete initialization in some ohci controllers with
Takeshi Shibagaki [Mon, 22 Dec 2003 15:40:10 +0000 (15:40 +0000)]
Fixed incomplete initialization in some ohci controllers with
broken BIOS. Separate ohci_controller_init() from ohci_init(),
and call ohci_controller_init() at resume process once more.

Discussed on [bsd-nomads:16737] - [bsd-nomads:16746].

Submitted by Hiroyuki Aizu <eyes@navi.org> [bsd-nomads:16741]

20 years agoPut the em and vx drivers back into the kernel so that the mfsroot stops
Scott Long [Mon, 22 Dec 2003 15:19:45 +0000 (15:19 +0000)]
Put the em and vx drivers back into the kernel so that the mfsroot stops
overflowing.  We are one again on the brink of the floppies having major
overflow problems.

20 years agoEnable support for DEVICE_SUSPEND, DEVICE_RESUME and DEVICE_SHUTDOWN
Takeshi Shibagaki [Mon, 22 Dec 2003 15:18:46 +0000 (15:18 +0000)]
Enable support for DEVICE_SUSPEND, DEVICE_RESUME and DEVICE_SHUTDOWN
methods for USB devices in the same way of uhci driver. But this change
is not complete because some ohci controlers are not initialized completely.
So "kernel: usb0: 1 scheduling overruns" interrupt will generate many times.

This change will be same one in PR kern/60099.

Discussed on [bsd-nomads:16737] - [bsd-nomads:16746].

20 years agoMake oldsize in smbfs_getattr() 64 bits wide instead of 32 to avoid
Tim J. Robbins [Mon, 22 Dec 2003 12:33:31 +0000 (12:33 +0000)]
Make oldsize in smbfs_getattr() 64 bits wide instead of 32 to avoid
truncation when files are larger than 4GB.

20 years agoSome minor touchups:
Bill Paul [Mon, 22 Dec 2003 10:22:57 +0000 (10:22 +0000)]
Some minor touchups:

In NdisQueryBuffer() and NdisQueryBufferSafe(), the vaddr argument is
optional, so test it before trying to dereference it.

Also correct NdisGetFirstBufferFromPacket()/NdisGetFirstBufferFromPacketSafe():
we need to use nb_mappedsystemva from the buffer, not nb_systemva.

20 years agoNow that I finally have power back, implement a couple more NDIS API
Bill Paul [Mon, 22 Dec 2003 08:24:32 +0000 (08:24 +0000)]
Now that I finally have power back, implement a couple more NDIS API
routines: NdisUnchainBufferAtBack(), NdisGetFirstBufferFromPacketSafe()
and NdisGetFirstBufferFromPacket(). This should bring us a little
closer to getting the Intel centrino wireless NIC to work.

Note: I have not actually tested these additions since I don't
have a driver that calls them, however they're pretty simple, and
one of them is taken pretty much directly from the Windows ndis.h
header file, so I'm fairly confident they work, but disclaimers
apply.

20 years agoFix typo in ENE CB710 description. It isn't a 720.
Warner Losh [Mon, 22 Dec 2003 06:09:35 +0000 (06:09 +0000)]
Fix typo in ENE CB710 description.  It isn't a 720.

20 years agoAdd more bridges, based on reading the pccbb.c file.
Warner Losh [Mon, 22 Dec 2003 06:08:14 +0000 (06:08 +0000)]
Add more bridges, based on reading the pccbb.c file.

20 years agoFix ASM_OUTPUT_LABELREF to deal with TSL model prefixes GCC
Alexander Kabaev [Mon, 22 Dec 2003 04:27:17 +0000 (04:27 +0000)]
Fix ASM_OUTPUT_LABELREF to deal with TSL model prefixes GCC
started to use internally. They need to be stripped before
the name of the label or symbol is printed.

20 years agofixed a bug that 'ip6addrctl delete' command does not work
SUZUKI Shinsuke [Mon, 22 Dec 2003 03:13:50 +0000 (03:13 +0000)]
fixed a bug that 'ip6addrctl delete' command does not work

obtained from: KAME

20 years agofixed a bug that IPv6 routing header does not work properly if specified from userlan...
SUZUKI Shinsuke [Mon, 22 Dec 2003 03:12:13 +0000 (03:12 +0000)]
fixed a bug that IPv6 routing header does not work properly if specified from userland application

reviewed by: ume

20 years ago - Create an unmapped guard page to trap access to vm_page_array[-1].
Alan Cox [Mon, 22 Dec 2003 02:04:08 +0000 (02:04 +0000)]
 - Create an unmapped guard page to trap access to vm_page_array[-1].
   This guard page would have trapped the problems with the MFC of the PAE
   support to RELENG_4 at an earlier point in the sequence of events.

Submitted by: tegge

20 years ago - Significantly reduce the number of preallocated pv entries in
Alan Cox [Mon, 22 Dec 2003 01:01:32 +0000 (01:01 +0000)]
 - Significantly reduce the number of preallocated pv entries in
   pmap_init().  Such a large preallocation is unnecessary and wastes
   nearly eight megabytes of kernel virtual address space per gigabyte
   of managed physical memory.
 - Increase UMA_BOOT_PAGES by two.  This enables the removal of
   pmap_pv_allocf().  (Note: this function was only used during
   initialization, specifically, after pmap_init() but before
   pmap_init2().  During pmap_init2(), a new allocator is installed.)

20 years agoExplain what all this is about.
Greg Lehey [Mon, 22 Dec 2003 00:01:53 +0000 (00:01 +0000)]
Explain what all this is about.

20 years agoTeach the Fixit environment how to deal with a dynamic root. Symlink
Scott Long [Sun, 21 Dec 2003 17:16:44 +0000 (17:16 +0000)]
Teach the Fixit environment how to deal with a dynamic root.  Symlink
/libexec to /mnt2/libexec, and execute /mnt2/rescue/ldconfig to add
the /mnt2/lib and /mnt2/usr/lib library directories.  Thanks to John Baldwin
for working to track this down.

Submitted by: jhb

20 years agoFix a typo.
Hideyuki KURASHINA [Sun, 21 Dec 2003 14:30:26 +0000 (14:30 +0000)]
Fix a typo.

Submitted by: Lukas Ertl <l.ertl@univie.ac.at>
PR: docs/60469

20 years agoCorrect URI to USB specs.
Hideyuki KURASHINA [Sun, 21 Dec 2003 14:30:23 +0000 (14:30 +0000)]
Correct URI to USB specs.

Submitted by: Lukas Ertl <l.ertl@univie.ac.at>
PR: docs/60457

20 years ago - Cleanup some garbage left by KSE. There is still much garbage left to be
Jeff Roberson [Sun, 21 Dec 2003 13:47:42 +0000 (13:47 +0000)]
 - Cleanup some garbage left by KSE.  There is still much garbage left to be
   removed, see the 110 instances of "XXXKSE" in src/sys for examples.

20 years ago- Add ':' as a separator between the OpenFirmware device space and
Peter Grehan [Sun, 21 Dec 2003 12:38:25 +0000 (12:38 +0000)]
- Add ':' as a separator between the OpenFirmware device space and
the file path. Commonly used on Macs e.g. "hd:9".
- Update the ofw_setcurrdev routine to match libstand setenv prototype

Not objected to by: sparc64

20 years agoMake __elfN(ofw_loadfile) match parameter declaration for file_format
Peter Grehan [Sun, 21 Dec 2003 12:27:01 +0000 (12:27 +0000)]
Make __elfN(ofw_loadfile) match parameter declaration for file_format
in boot/common/bootstrap.h. Having a 32-bit size when a 64-bit param
is declared wreaks havoc on PPC.

Not objected to by: sparc64

20 years agoOnly print out an error if returned data size is < 0. A value of 0
Peter Grehan [Sun, 21 Dec 2003 12:19:38 +0000 (12:19 +0000)]
Only print out an error if returned data size is < 0. A value of 0
happens almost every time at the end of a file when using NFS.

No objection by: sparc64

20 years agoUse daddr_t instead of u_long for byte offset in strategy
Peter Grehan [Sun, 21 Dec 2003 12:16:58 +0000 (12:16 +0000)]
Use daddr_t instead of u_long for byte offset in strategy
routine to avoid >4G truncation on 32-bit systems.

no objection by: sparc64

20 years ago- use correct pointer arithmetic in heapsize calculation
Peter Grehan [Sun, 21 Dec 2003 12:11:31 +0000 (12:11 +0000)]
- use correct pointer arithmetic in heapsize calculation
- handle multiple Ofw memory regions when determining mem size
- allow currdev to be set as a loader command-line option.
  parse() is used to allow future options to be processed.

20 years ago - Correct an error in mincore(2) that has existed since its introduction:
Alan Cox [Sun, 21 Dec 2003 06:03:40 +0000 (06:03 +0000)]
 - Correct an error in mincore(2) that has existed since its introduction:
   mincore(2) should check that the page is valid, not just allocated.
   Otherwise, it can return a false positive for a page that is not yet
   resident because it is being read from disk.

20 years ago - Add new FTP mirror site in Turkey (first/only one).
Ken Smith [Sun, 21 Dec 2003 05:42:00 +0000 (05:42 +0000)]
- Add new FTP mirror site in Turkey (first/only one).

Approved by: murray (re@)

20 years agoBig round of updates:
Bill Paul [Sun, 21 Dec 2003 00:00:08 +0000 (00:00 +0000)]
Big round of updates:

- Make ndis_get_info()/ndis_set_info() sleep on the setdone/getdone
  routines if they get back NDIS_STATUS_PENDING.

- Add a bunch of net80211 support so that 802.11 cards can be twiddled
  with ifconfig. This still needs more work and is not guaranteed to
  work for everyone. It works on my 802.11b/g card anyway.

The problem here is Microsoft doesn't provide a good way to a) learn
all the rates that a card supports (if it has more than 8, you're
kinda hosed) and b) doesn't provide a good way to distinguish between
802.11b, 802.11b/g an 802.11a/b/g cards, so you sort of have to guess.

Setting the SSID and switching between infrastructure/adhoc modes
should work. WEP still needs to be implemented. I can't find any API
for getting/setting the channel other than the registry/sysctl keys.

20 years ago - Make our transfer decisions based on load and not transferable load. A
Jeff Roberson [Sat, 20 Dec 2003 22:35:20 +0000 (22:35 +0000)]
 - Make our transfer decisions based on load and not transferable load.  A
   cpu could have been bogged down with non-transferable load and still not
   migrated a new thread to an idle cpu.  This required some benchmarking and
   tuning to get right as the comment above it suggests.

20 years ago - Enable ithread migration on x86. This is done to work around a bug in the
Jeff Roberson [Sat, 20 Dec 2003 20:36:19 +0000 (20:36 +0000)]
 - Enable ithread migration on x86.  This is done to work around a bug in the
   IO APIC on Xeons that prevents round-robin interrupt assignment from
   working.

20 years agoSince we have additional kernel virtual address space, allow the buffer
Alan Cox [Sat, 20 Dec 2003 20:03:10 +0000 (20:03 +0000)]
Since we have additional kernel virtual address space, allow the buffer
cache to grow to 400M bytes.

20 years agoRemove a variable that has been initialized but otherwise unused since
Alan Cox [Sat, 20 Dec 2003 19:46:21 +0000 (19:46 +0000)]
Remove a variable that has been initialized but otherwise unused since
revision 1.315.

20 years agoo move mutex init/destroy logic to the module load/unload hooks;
Sam Leffler [Sat, 20 Dec 2003 18:32:48 +0000 (18:32 +0000)]
o move mutex init/destroy logic to the module load/unload hooks;
  otherwise they are initialized twice when the code is statically
  configured in the kernel because the module load method gets
  invoked before the user application calls ip_mrouter_init
o add a mutex to synchronize the module init/done operations; this
  sort of was done using the value of ip_mroute but X_ip_mrouter_done
  sets it to NULL very early on which can lead to a race against
  ip_mrouter_init--using the additional mutex means this is safe now
o don't call ip_mrouter_reset from ip_mrouter_init; this now happens
  once at module load and X_ip_mrouter_done does the appropriate
  cleanup work to insure the data structures are in a consistent
  state so that a subsequent init operation inherits good state

Reviewed by: juli

20 years agoMake the multiple include guard correct for this file location.
David E. O'Brien [Sat, 20 Dec 2003 17:12:25 +0000 (17:12 +0000)]
Make the multiple include guard correct for this file location.
This fixes a bug where the guard conflicted with machine/ioctl_*.

20 years agoDo an update mount operation to mount the mfsroot as read/write rather
John Baldwin [Sat, 20 Dec 2003 16:34:45 +0000 (16:34 +0000)]
Do an update mount operation to mount the mfsroot as read/write rather
than read/only when sysinstall is running as init.  This fixes several
install issues.

20 years ago - In kseq_transfer() return if smp has not been started.
Jeff Roberson [Sat, 20 Dec 2003 14:03:14 +0000 (14:03 +0000)]
 - In kseq_transfer() return if smp has not been started.
 - In sched_add(), do the idle check prior to the transfer check so that we
   don't try to transfer load from an idle cpu.  This fixes panics caused by
   IPIs on UP machines running SMP kernels.

Reported/Debugged by: seanc

20 years ago - Running interactive tasks with the minimum time-slice is fine for vi and
Jeff Roberson [Sat, 20 Dec 2003 12:54:35 +0000 (12:54 +0000)]
 - Running interactive tasks with the minimum time-slice is fine for vi and
   sh, but not so great for mozilla, X, etc.  Add a fixed define for the slice
   size granted to interactive KSEs.

20 years agoAdd map for CP1131
Andrey A. Chernov [Sat, 20 Dec 2003 11:20:05 +0000 (11:20 +0000)]
Add map for CP1131

Submitted by:   Yury Tarasievich <grog@grsu.by>

20 years agoAdd be_BY.*
Andrey A. Chernov [Sat, 20 Dec 2003 11:18:43 +0000 (11:18 +0000)]
Add be_BY.*

Submitted by:   Yury Tarasievich <grog@grsu.by>

20 years agoAdd be_BY.*
Andrey A. Chernov [Sat, 20 Dec 2003 11:05:34 +0000 (11:05 +0000)]
Add be_BY.*

Submitted by:   Yury Tarasievich <grog@grsu.by>

20 years agoMove all of the recovery thread routines next
Justin T. Gibbs [Fri, 19 Dec 2003 18:34:30 +0000 (18:34 +0000)]
Move all of the recovery thread routines next
to each other.

Correct the recovery thread's loop so that it
will terminate properly on shutdown.  We also
clear the recovery_thread proc pointer so that
any additional calls to aic_terminate_recovery_thread()
will not attempt to kill a thread that doesn't
exist.  Lastly, code the loop so that termination
will still be successfull even if the termination
request occurs just prior to us entering the loop
or while the recovery thread is off recovering
commands.

20 years agoUse __cxa_atexit, rather than atexit, to register C++ destructors for local
Alexander Kabaev [Fri, 19 Dec 2003 18:20:48 +0000 (18:20 +0000)]
Use __cxa_atexit, rather than atexit, to register C++ destructors for local
statics and global objects.  This is essential for fully standards-compliant
handling of destructors, and requires __cxa_atexit in libc.

20 years agoBump __FreeBSD_version to indicate __cxa_atexit/__cxa_finalize presence.
Alexander Kabaev [Fri, 19 Dec 2003 18:17:13 +0000 (18:17 +0000)]
Bump __FreeBSD_version to indicate __cxa_atexit/__cxa_finalize presence.

20 years agoFix uncontrolled access to the buffer in rfcomm_sppd(1).
Maksim Yevmenkin [Fri, 19 Dec 2003 18:15:56 +0000 (18:15 +0000)]
Fix uncontrolled access to the buffer in rfcomm_sppd(1).
Fix typo in hcsecd(8) man page.

Submitted by: Guido Falsi <mad@madpilot.net>
Reviewed by: imp (mentor)
Approved by: imp (mentor)

20 years agoWe only need to terminate our recovery thread once.
Justin T. Gibbs [Fri, 19 Dec 2003 18:10:59 +0000 (18:10 +0000)]
We only need to terminate our recovery thread once.

20 years agoImplement __cxa_atexit/__cxa_finalize as specified by the cross-vendor
Alexander Kabaev [Fri, 19 Dec 2003 17:11:21 +0000 (17:11 +0000)]
Implement __cxa_atexit/__cxa_finalize as specified by the cross-vendor
ó++ ABI document at http://www.codesourcery.com/cxx-abi/abi.html#dso-dtor

The ABI was initially defined for ia64, but GCC3 and Intel compilers
have adopted it on other platforms.

This is the patch from PR bin/59552 with a number of changes by
me.

PR: bin/59552
Submitted by: Bradley T Hughes (bhughes at trolltech dot com)

20 years agoImplement __cxa_atexit/__cxa_finalize as specified by the cross-vendor
Alexander Kabaev [Fri, 19 Dec 2003 17:11:20 +0000 (17:11 +0000)]
Implement __cxa_atexit/__cxa_finalize as specified by the cross-vendor
C++ ABI document at http://www.codesourcery.com/cxx-abi/abi.html#dso-dtor

The ABI was initially defined for ia64, but GCC3 and Intel compilers
have adopted it on other platforms.

This is the patch from PR bin/59552 with a number of changes by
me.

PR: bin/59552
Submitted by: Bradley T Hughes (bhughes at trolltech dot com)

20 years agoBring the description of the sysctl(8) variable
Yaroslav Tykhiy [Fri, 19 Dec 2003 16:42:35 +0000 (16:42 +0000)]
Bring the description of the sysctl(8) variable
net.graph.nonstandard_pppoe into accord with the reality.

MFC after: 1 week

20 years agoThe default value of net.graph.nonstandard_pppoe is changed to -1,
Yaroslav Tykhiy [Fri, 19 Dec 2003 16:03:28 +0000 (16:03 +0000)]
The default value of net.graph.nonstandard_pppoe is changed to -1,
which means "always stay in the standard mode of PPPoE operation
regardless of any junk floating around."

As the referenced PR stated clearly, the old default setting of 0
was extremely dangerous because it opened a possibility for a
spurious frame not only to put down a single PPPoE node running
FreeBSD, but to plague *every* FreeBSD node in a PPPoE network in
such a way that those nodes would keep poisoning each other until
rebooted simultaneously.

PR: kern/47920
Reviewed by: Gleb Smirnoff <glebius <at> cell.sick.ru>
MFC after: 1 week

20 years agoFixed panic on hook disconnection that previous revision has introduced.
Ruslan Ermilov [Fri, 19 Dec 2003 15:09:12 +0000 (15:09 +0000)]
Fixed panic on hook disconnection that previous revision has introduced.

20 years agoReplace a comment with more accurated one, memory heap is now protected by
David Xu [Fri, 19 Dec 2003 13:24:54 +0000 (13:24 +0000)]
Replace a comment with more accurated one, memory heap is now protected by
new fork() wrapper.

20 years agoReduce the overhead of semop() by using the kernel stack instead of
Tim J. Robbins [Fri, 19 Dec 2003 13:07:17 +0000 (13:07 +0000)]
Reduce the overhead of semop() by using the kernel stack instead of
malloc'd memory to store the operations array if it is small enough
to fit.

20 years agoCode clean up, remove unused MACROS and function prototypes.
David Xu [Fri, 19 Dec 2003 12:57:08 +0000 (12:57 +0000)]
Code clean up, remove unused MACROS and function prototypes.

20 years agoFirst byte of GBK-like sequences is 0x81, not 0x80
Andrey A. Chernov [Fri, 19 Dec 2003 12:54:42 +0000 (12:54 +0000)]
First byte of GBK-like sequences is 0x81, not 0x80

20 years agoChange encoding to GBK to get correct first byte range
Andrey A. Chernov [Fri, 19 Dec 2003 12:51:40 +0000 (12:51 +0000)]
Change encoding to GBK to get correct first byte range

20 years agoSync to 1.149 of usbdevs
MIHIRA Sanpei Yoshiro [Fri, 19 Dec 2003 12:21:11 +0000 (12:21 +0000)]
Sync to 1.149 of usbdevs

20 years agoAdd support Panasonic KXL-CB35AN(DVD-ROM & CD-R/RW)
MIHIRA Sanpei Yoshiro [Fri, 19 Dec 2003 12:19:12 +0000 (12:19 +0000)]
Add support Panasonic KXL-CB35AN(DVD-ROM & CD-R/RW)

Submitted by: OISHI Masakuni <yamasa@bsdhouse.org> [FreeBSD-users-jp 77672]

20 years agoChange the select timeout from 100ms to 2 seconds now that SIGCHILD is
Scott Long [Fri, 19 Dec 2003 11:18:37 +0000 (11:18 +0000)]
Change the select timeout from 100ms to 2 seconds now that SIGCHILD is
handled.

20 years agoFixed compilation on 64-bit platforms.
Ruslan Ermilov [Fri, 19 Dec 2003 09:34:37 +0000 (09:34 +0000)]
Fixed compilation on 64-bit platforms.

20 years agoAdd zh_HK.Big5HKSCS
Andrey A. Chernov [Fri, 19 Dec 2003 07:04:56 +0000 (07:04 +0000)]
Add zh_HK.Big5HKSCS

PR:             59799
Submitted by:   Statue <statue@freebsd.sinica.edu.tw>

20 years agoAdd zh_HK.Big5HKSCS
Andrey A. Chernov [Fri, 19 Dec 2003 06:57:57 +0000 (06:57 +0000)]
Add zh_HK.Big5HKSCS

20 years agoIn ahd_run_qoutfifos, correct a !=/== logic bug
Justin T. Gibbs [Fri, 19 Dec 2003 04:17:43 +0000 (04:17 +0000)]
In ahd_run_qoutfifos, correct a !=/== logic bug
that would cause an infinite loop any time we
manually flush the good status FIFO.  Also make
our loop delay unconditional to ensure we don't
miss any FIFO allocations by the hardware.

20 years agoMake ndiscvt(8) emit the binary image array as inline assembly code rather
Bill Paul [Thu, 18 Dec 2003 21:47:14 +0000 (21:47 +0000)]
Make ndiscvt(8) emit the binary image array as inline assembly code rather
than a char array. Emitting the data as a big char array works fine in
the typical case, where a .sys file may be ~50K in size. Unfortunately,
some .sys files can be several hundred Kbytes in size, or even several
megabytes in size. One extreme case is the Intel centrino wireless
driver, which is 2.4MB. This causes us to emit an ndis_driver_data.h
file that's on the order of 15MB in size, and gcc consumes enormous
amounts of virtual memory while trying to compile it. On my laptop,
with 128MB of RAM and 256MB of swap space, gcc consumed all available
VM and crashed without being able to compile if_ndis.o.

By emitting the array as assembler, we bypass the C compiler and consume
much less memory. I was able to easily test compile if_ndis.ko with the
centrino driver on my laptop after this change.

This is merely a convenience, and should not have any operational effect
on the NDISulator itself.

20 years agoMFamd64: Remove i386_protection_init() and the protection_codes[] array
John Baldwin [Thu, 18 Dec 2003 21:15:18 +0000 (21:15 +0000)]
MFamd64: Remove i386_protection_init() and the protection_codes[] array
and replace them with a simple if test to turn on PG_RW.  i386 != vax.

20 years agoFix PANASONIC KXLCB20AN Protocol.
MIHIRA Sanpei Yoshiro [Thu, 18 Dec 2003 19:59:32 +0000 (19:59 +0000)]
Fix PANASONIC KXLCB20AN Protocol.

Submitted by: OISHI Masakuni <yamasa@bsdhouse.org> [FreeBSD-users-jp 77672]

20 years agoFix the register timings for AMD/VIA/nVidia chipsets.
Søren Schmidt [Thu, 18 Dec 2003 17:36:41 +0000 (17:36 +0000)]
Fix the register timings for AMD/VIA/nVidia chipsets.

20 years agomdoc(7) style: Start each sentence on a new line.
Yaroslav Tykhiy [Thu, 18 Dec 2003 17:31:44 +0000 (17:31 +0000)]
mdoc(7) style: Start each sentence on a new line.

20 years agoCapitalize PPPoE, PPP, and Ethernet in a consistent way.
Yaroslav Tykhiy [Thu, 18 Dec 2003 17:27:49 +0000 (17:27 +0000)]
Capitalize PPPoE, PPP, and Ethernet in a consistent way.

20 years agoInitialize acpi buffer structs early in order to avoid freeing
Alexander Kabaev [Thu, 18 Dec 2003 17:04:11 +0000 (17:04 +0000)]
Initialize acpi buffer structs early in order to avoid freeing
unallocated pointers later in done: section.

20 years agoThe politically incorrect sysctl "stupid_isp" had its name
Yaroslav Tykhiy [Thu, 18 Dec 2003 16:55:09 +0000 (16:55 +0000)]
The politically incorrect sysctl "stupid_isp" had its name
changed long ago.  It has been "nonstandard_pppoe" for quite
a while.

20 years agoThere are two modes of ng_pppoe operation, standard and
Yaroslav Tykhiy [Thu, 18 Dec 2003 16:38:35 +0000 (16:38 +0000)]
There are two modes of ng_pppoe operation, standard and
nonstandard.  They differ in the values of certain fields in
the PPPoE frame.  Previously, ng_pppoe would start in standard
mode, yet switch to nonstandard one upon reception of a single
nonstandard frame.  After having done so, ng_pppoe would be unable
to interact with standard PPPoE peers.  Thus, a DoS condition
existed that could be triggered by a buggy peer or malicious party.

Since few people have expressed their displeasure WRT this problem,
the default operation of ng_pppoe is left untouched for now.  However,
a new value for the sysctl net.graph.nonstandard_pppoe is introduced,
-1, which will force ng_pppoe stay in standard mode regardless of any
bogus frames floating around.

PR: kern/47920
Submitted by: Gleb Smirnoff <glebius <at> cell.sick.ru>
MFC after: 1 week

20 years agoAdd the Solaris x86 boot partition type. This is used in Solaris 10
Hartmut Brandt [Thu, 18 Dec 2003 13:13:02 +0000 (13:13 +0000)]
Add the Solaris x86 boot partition type. This is used in Solaris 10
(and perhaps earlier).

Submitted by: Joerg Schilling <schilling@fokus.fraunhofer.de>

20 years agoNow I understand what Bruce was getting at - -1 can be parsed as two
Jordan K. Hubbard [Thu, 18 Dec 2003 10:41:39 +0000 (10:41 +0000)]
Now I understand what Bruce was getting at - -1 can be parsed as two
tokens, so it does indeed need to be parenthesized.  Duh.  Sometimes
it can stare you right and the face and you still don't see it.  Thanks, bde.

20 years agoRemoved an outdated comment.
Ruslan Ermilov [Thu, 18 Dec 2003 09:16:40 +0000 (09:16 +0000)]
Removed an outdated comment.

Submitted by: archie

20 years agoSync to 1.148 of usbdevs
MIHIRA Sanpei Yoshiro [Thu, 18 Dec 2003 08:20:02 +0000 (08:20 +0000)]
Sync to 1.148 of usbdevs