]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
20 years agoInsert "arm" between "amd64" and "powerpc"
cognet [Fri, 14 May 2004 23:08:52 +0000 (23:08 +0000)]
Insert "arm" between "amd64" and "powerpc"

Requested by: bde

20 years agoChange ffs_realloccg() to set the valid bits for the extended part of the
kensmith [Fri, 14 May 2004 22:00:08 +0000 (22:00 +0000)]
Change ffs_realloccg() to set the valid bits for the extended part of the
fragment to zero the valid parts of a VM_IO buffer.

RE would like this to be part of 4.10-RC3 so this will be MFC-ed immediately.

Reviewed by: alc, tegge

20 years agoFixed total disorder of MD includes.
bde [Fri, 14 May 2004 21:05:35 +0000 (21:05 +0000)]
Fixed total disorder of MD includes.

Disorder buffer overflowed by: arm commit

20 years agoFix breakage caused by alphabetically sorting SRCS: rtld_start.S must come first!
se [Fri, 14 May 2004 21:01:52 +0000 (21:01 +0000)]
Fix breakage caused by alphabetically sorting SRCS: rtld_start.S must come first!
The previous version made all shared binaries dump core.

20 years agoFixed some common printf format errors. Don't assume that "struct foo *"
bde [Fri, 14 May 2004 20:51:42 +0000 (20:51 +0000)]
Fixed some common printf format errors.  Don't assume that "struct foo *"
is "void *" (it isn't) or that the default promotion of pid_t is int.
Instead, assume that casting "struct foo *" to "void *" and printing the
result with %p is useful, and that all pid_t's are representable as longs.

Fixed some minor style bugs (mainly spelling errors in comments).

20 years agoBack out last revision that unnecessarily changed valid assembler
ru [Fri, 14 May 2004 20:29:30 +0000 (20:29 +0000)]
Back out last revision that unnecessarily changed valid assembler
line comments and damaged the CVS history.

Prompted by: bde, jhb

20 years agoPartly back out last revision that unnecessarily changed valid
ru [Fri, 14 May 2004 20:28:31 +0000 (20:28 +0000)]
Partly back out last revision that unnecessarily changed valid
assembler line comments and damaged the CVS history.

Prompted by: bde, jhb

20 years agoCorrect the misuse of \- (mdoc(7) mathematical minus) when - (mdoc(7)
simon [Fri, 14 May 2004 20:07:49 +0000 (20:07 +0000)]
Correct the misuse of \- (mdoc(7) mathematical minus) when - (mdoc(7)
hyphen) is meant.

Mdoc(7) clue by: ru
Reviewed by: ru

20 years agoStyle fixes:
bde [Fri, 14 May 2004 19:52:35 +0000 (19:52 +0000)]
Style fixes:
Main ones: mostly use conditional expressions in ifdefs instead of a
mixture of conditional expressions and nested ifdefs.
Nearby ones:
- don't do less than echo the code in the comment about libc_r
- fixed some internal insertion sort errors and indentation errors.

20 years agoFixed some insertion sort errors (external ones only).
bde [Fri, 14 May 2004 19:31:20 +0000 (19:31 +0000)]
Fixed some insertion sort errors (external ones only).

20 years agoBump the .Dd value.
joerg [Fri, 14 May 2004 19:29:39 +0000 (19:29 +0000)]
Bump the .Dd value.

Update xrefs to more contemporary items.

Requested by: ru, green
MFC after: 1 week

20 years agoWhen I wrote this man page more than 5 years ago, I simply didn't
joerg [Fri, 14 May 2004 18:46:16 +0000 (18:46 +0000)]
When I wrote this man page more than 5 years ago, I simply didn't
understand the true symmetric nature of Enigma, so my description of
``automatically detects that the input is encrypted'' was simply
wrong.  Replace that by a more accurate description of why feeding the
ciphertext again into the engine will decrypt it.

MFC after: 1 week

20 years agomdoc(7) police:
hmp [Fri, 14 May 2004 16:57:27 +0000 (16:57 +0000)]
mdoc(7) police:

* remove an extra comma at the end of the .Nm list

20 years agoInstead of scanning the entire lower 1 MB of RAM, only scan locations
njl [Fri, 14 May 2004 16:52:39 +0000 (16:52 +0000)]
Instead of scanning the entire lower 1 MB of RAM, only scan locations
where the RSD PTR can actually occur.  According to section 5.2.2
of the ACPI spec, we only consider two regions for the base address:

    1. EBDA (0x0 - 0x3FF)
    2. High memory (0xE0000 - 0xFFFFF)

I don't know whether this fixes any actual problems but is more correct.

20 years agoUse a simpler, faster buffering scheme for partial characters in mbrtowc().
tjr [Fri, 14 May 2004 15:40:47 +0000 (15:40 +0000)]
Use a simpler, faster buffering scheme for partial characters in mbrtowc().

20 years agoMFNetBSD (umodem.c 1.46):
sanpei [Fri, 14 May 2004 15:16:09 +0000 (15:16 +0000)]
MFNetBSD (umodem.c 1.46):
URL updates

20 years agoThis is FreeBSD/arm, not FreeBSD/i386
cognet [Fri, 14 May 2004 14:52:32 +0000 (14:52 +0000)]
This is FreeBSD/arm, not FreeBSD/i386

Submitted by: ru

20 years agoUse four spaces instead of tab for second level indentation.
cognet [Fri, 14 May 2004 14:50:55 +0000 (14:50 +0000)]
Use four spaces instead of tab for second level indentation.

Requested by: ru

20 years agoDo not attempt to build libdisk, libthr and libc_r for arm.
cognet [Fri, 14 May 2004 13:45:56 +0000 (13:45 +0000)]
Do not attempt to build libdisk, libthr and libc_r for arm.

20 years agoDo not build sysintall for arm.
cognet [Fri, 14 May 2004 13:44:55 +0000 (13:44 +0000)]
Do not build sysintall for arm.

20 years agoDefine PTRDIFF_IS_LONG and SIZEOF_IS_ULONG for arm.
cognet [Fri, 14 May 2004 13:44:36 +0000 (13:44 +0000)]
Define PTRDIFF_IS_LONG and SIZEOF_IS_ULONG for arm.

20 years agoSome arm bits for gprof.
cognet [Fri, 14 May 2004 13:44:04 +0000 (13:44 +0000)]
Some arm bits for gprof.

20 years agoDo not build truss on arm.
cognet [Fri, 14 May 2004 13:43:18 +0000 (13:43 +0000)]
Do not build truss on arm.

20 years agoDefine uart_sa1110_ops and uart_sa1110_classe in .c files instead of spamming
cognet [Fri, 14 May 2004 13:42:50 +0000 (13:42 +0000)]
Define uart_sa1110_ops and uart_sa1110_classe in .c files instead of spamming
uart_cpu.h

20 years agoDefine INLINE_LIMIT for arm.
cognet [Fri, 14 May 2004 13:35:46 +0000 (13:35 +0000)]
Define INLINE_LIMIT for arm.

20 years agoDo not attempt to build anything in sys/boot for arm.
cognet [Fri, 14 May 2004 13:34:53 +0000 (13:34 +0000)]
Do not attempt to build anything in sys/boot for arm.

20 years agoDefine MACHINE_CPU for arm.
cognet [Fri, 14 May 2004 13:34:15 +0000 (13:34 +0000)]
Define MACHINE_CPU for arm.

20 years agoDefine iaddr_t and saddr_t for arm.
cognet [Fri, 14 May 2004 13:32:13 +0000 (13:32 +0000)]
Define iaddr_t and saddr_t for arm.

20 years agoUse WARNS?=3 for these in the arm case for now, due to toolchain issues.
cognet [Fri, 14 May 2004 13:31:21 +0000 (13:31 +0000)]
Use WARNS?=3 for these in the arm case for now, due to toolchain issues.

20 years agoAdd a list of all local changed files. All these files have been taken
mbr [Fri, 14 May 2004 13:29:48 +0000 (13:29 +0000)]
Add a list of all local changed files. All these files have been taken
of the vendor tree.

20 years agoWe don't need these files. Add the to the x-list.
mbr [Fri, 14 May 2004 13:28:37 +0000 (13:28 +0000)]
We don't need these files. Add the to the x-list.

20 years agoImplement enough of an uart driver to get serial console working.
cognet [Fri, 14 May 2004 13:26:52 +0000 (13:26 +0000)]
Implement enough of an uart driver to get serial console working.

20 years ago-maxdepth and -mindepth are global variables
eik [Fri, 14 May 2004 12:58:13 +0000 (12:58 +0000)]
-maxdepth and -mindepth are global variables

PR: 66613
Approved by: ru
MFC after: 3 days

20 years agoImport libkern arm specific bits.
cognet [Fri, 14 May 2004 12:28:31 +0000 (12:28 +0000)]
Import libkern arm specific bits.

20 years agoImport the openssl conf for arm.
cognet [Fri, 14 May 2004 12:26:51 +0000 (12:26 +0000)]
Import the openssl conf for arm.

20 years agoImport _setjmp.S for arm in libstand.
cognet [Fri, 14 May 2004 12:24:51 +0000 (12:24 +0000)]
Import _setjmp.S for arm in libstand.

20 years agoImport libkvm MD file for arm.
cognet [Fri, 14 May 2004 12:24:03 +0000 (12:24 +0000)]
Import libkvm MD file for arm.

20 years agoArm bits for libpthread. It has no chances to work and should be considered
cognet [Fri, 14 May 2004 12:21:29 +0000 (12:21 +0000)]
Arm bits for libpthread. It has no chances to work and should be considered
as stubs.

20 years agoC runtime support for FreeBSD/arm.
cognet [Fri, 14 May 2004 12:19:04 +0000 (12:19 +0000)]
C runtime support for FreeBSD/arm.

20 years agoImport arm bits for rtld-elf.
cognet [Fri, 14 May 2004 12:15:51 +0000 (12:15 +0000)]
Import arm bits for rtld-elf.

Obtained from: NetBSD

20 years agoImport the softfloat emulation library, needed for FreeBSD/arm right now.
cognet [Fri, 14 May 2004 12:13:06 +0000 (12:13 +0000)]
Import the softfloat emulation library, needed for FreeBSD/arm right now.
It should become useless when gcc 3.4 will be imported, as libgcc from
gcc 3.4 contains this bits for arm.

20 years agoImport the FreeBSD/arm libc bits.
cognet [Fri, 14 May 2004 12:04:31 +0000 (12:04 +0000)]
Import the FreeBSD/arm libc bits.

Obtained from: NetBSD

20 years agoWe use __arm__, not __arm32__.
cognet [Fri, 14 May 2004 11:51:22 +0000 (11:51 +0000)]
We use __arm__, not __arm32__.

20 years agoDefine malloc_pageshift and malloc_minsize for arm.
cognet [Fri, 14 May 2004 11:50:51 +0000 (11:50 +0000)]
Define malloc_pageshift and malloc_minsize for arm.

20 years agoAdd config magic for arm.
cognet [Fri, 14 May 2004 11:49:40 +0000 (11:49 +0000)]
Add config magic for arm.

20 years agoImport FreeBSD/arm kernel bits.
cognet [Fri, 14 May 2004 11:46:45 +0000 (11:46 +0000)]
Import FreeBSD/arm kernel bits.
It only supports sa1110 (on simics) right now, but xscale support should come
soon.
Some of the initial work has been provided by :
Stephane Potvin <sepotvin at videotron.ca>
Most of this comes from NetBSD.

20 years agoAdd support for GPE being a package of { reference, gpe bit }.
njl [Fri, 14 May 2004 04:17:56 +0000 (04:17 +0000)]
Add support for GPE being a package of { reference, gpe bit }.
Rework the ECDT probe to pass all the parameters in a temporary struct.
Note why we are mostly ok evaluating _GLK so early.

20 years agoFix a bug which I discovered recently while doing IPv6 testing at
wpaul [Fri, 14 May 2004 03:57:17 +0000 (03:57 +0000)]
Fix a bug which I discovered recently while doing IPv6 testing at
Wind River. In the IPv4 output path, one of the tests in ip_output()
checks how many slots are actually available in the interface output
queue before attempting to send a packet. If, for example, we need
to transmit a packet of 32K bytes over an interface with an MTU of
1500, we know it's going to take about 21 fragments to do it. If
there's less than 21 slots left in the output queue, there's no point
in transmitting anything at all: IP does not do retransmission, so
sending only some of the fragments would just be a waste of bandwidth.
(In an extreme case, if you're sending a heavy stream of fragmented
packets, you might find yourself sending nothing by the first fragment
of all your packets.) So if ip_output() notices there's not enough
room in the output queue to send the frame, it just dumps the packet
and returns ENOBUFS to the app.

It turns out ip6_output() lacks this code. Consequently, this caused
the netperf UDPIPV6_STREAM test to produce very poor results with large
write sizes. This commit adds code to check the remaining space in the
output queue and junk fragmented packets if they're too big to be
sent, just like with IPv4. (I can't imagine anyone's running an NFS
server using UDP over IPv6, but if they are, this will likely make them
a lot happier. :)

20 years agoFix a potential stack buffer overflow on systems whose ACPI OEMID
jdp [Fri, 14 May 2004 01:29:21 +0000 (01:29 +0000)]
Fix a potential stack buffer overflow on systems whose ACPI OEMID
fills its field (6 characters).  In that case the OEMID is not
null-terminated, and the sprintf that was used would copy up to the
next null byte, which could be pretty far away.

20 years agoDocument the "files" directive.
cognet [Thu, 13 May 2004 21:55:53 +0000 (21:55 +0000)]
Document the "files" directive.

Reminded by: jmg

20 years agoAdd some verbage over the difference between BUS_ADD_CHILD and
imp [Thu, 13 May 2004 20:54:49 +0000 (20:54 +0000)]
Add some verbage over the difference between BUS_ADD_CHILD and
device_add_child.  The former calls the latter.

Pointed out by: jhb

20 years agoo Add cautions about multiple adds of the device
imp [Thu, 13 May 2004 20:31:05 +0000 (20:31 +0000)]
o Add cautions about multiple adds of the device
o Remove setting the description of the device (that is more properly done in
  probe).
o Remove setting the driver, as it is unnecessary
o Add pseudo code to ensure that the device isn't added to the tree already.
o remove some now unreferenced routines

20 years agoAdd caveats about adding devices in the IDENTIFY context.
imp [Thu, 13 May 2004 20:29:40 +0000 (20:29 +0000)]
Add caveats about adding devices in the IDENTIFY context.

20 years agoThe DEVICE_IDENTIFY man page refers to this, create a dummy one
imp [Thu, 13 May 2004 20:28:18 +0000 (20:28 +0000)]
The DEVICE_IDENTIFY man page refers to this, create a dummy one
directing people to device_add_child.

20 years agoDifferent VIA host bridges use different offsets to their AGP config
jhb [Thu, 13 May 2004 20:05:42 +0000 (20:05 +0000)]
Different VIA host bridges use different offsets to their AGP config
registers, so add a register offset array to the softc.  We key off the
device ID to determine which set of register offsets.  Currently the 8385
host bridge used on amd64 is the only bridge to use the AGP3_VIA_*
register offsets and all other bridges use the AGP_VIA_* offsets.  It is
currently unclear if the AGP3_VIA_* offsets are for VIA bridges that
implement AGP 3.0 bridges or just for amd64 bridges.

Submitted by: Kenneth Culver culverk at sweetdreamsracing dot biz

20 years agoSplit sleepq_wakeup_thread() into two functions. sleepq_remove_thread()
jhb [Thu, 13 May 2004 20:00:43 +0000 (20:00 +0000)]
Split sleepq_wakeup_thread() into two functions.  sleepq_remove_thread()
removes a specific thread from a sleep queue.  sleepq_resume_thread()
resumes scheduling of a thread that has been previously removed from a
sleep queue.
- sleepq_catch_signals() just removes a thread from the queue it was just
  added to when a pending signal is found.
- sleepq_signal() and sleepq_broadcast() remove threads from a queue,
  drop the queue lock, and then resume all the previously removed threads.
  This doesn't completely fix the sched_lock <-> sleepq chain LOR, but it
  makes it a little better as we no longer call setrunnble() with a sleep
  queue lock held meaning if setrunnable() tries to wakeup the swapper we
  don't try to lock two sleep queue chains at the same time.

20 years agoAdd a reference to the URL of the SMBus specification, so the
joerg [Thu, 13 May 2004 19:31:31 +0000 (19:31 +0000)]
Add a reference to the URL of the SMBus specification, so the
terminology used in these drivers will become more transparent.

MFC after: 1 week

20 years agoUnbreak the build.
des [Thu, 13 May 2004 19:15:21 +0000 (19:15 +0000)]
Unbreak the build.

20 years agoEnable INTR_ENTROPY if the interrupt is not set to INTR_FAST. Remove the
scottl [Thu, 13 May 2004 16:02:18 +0000 (16:02 +0000)]
Enable INTR_ENTROPY if the interrupt is not set to INTR_FAST.  Remove the
testing and setting of the INTR_ENTROPY macro as it is not needed in
FreeBSD 5.x.

Submitted by: Alex Vasylenko

20 years agoFixed some minor style bugs.
bde [Thu, 13 May 2004 15:59:38 +0000 (15:59 +0000)]
Fixed some minor style bugs.

20 years agocheck if the null encryption is supported or not.
ume [Thu, 13 May 2004 15:46:28 +0000 (15:46 +0000)]
check if the null encryption is supported or not.

Requested by: bms
Obtained from: KAME

20 years agoSome mdoc fixes.
kensmith [Thu, 13 May 2004 13:09:47 +0000 (13:09 +0000)]
Some mdoc fixes.

mdoc help from: ru
MFC after: 5 days

20 years agoRemove extraneous spaces.
truckman [Thu, 13 May 2004 11:33:44 +0000 (11:33 +0000)]
Remove extraneous spaces.

20 years agoImplement sbc_lockassert() and sb_lockassert() functions to allow
truckman [Thu, 13 May 2004 11:32:54 +0000 (11:32 +0000)]
Implement sbc_lockassert() and sb_lockassert() functions to allow
proper locking to be checked at runtime.

Remove sb_lock() and sb_unlock() calls from sb_reset_dsp() because the
latter is called from sb_setup() with the lock already held.  Add a
call to sb_lockassert().

Surround the call to sb_reset_dsp() in sb16_attach() with sb_lock()
and sb_unlock() calls.

Tested by: Bartek Marcinkiewicz <junior AT p233.if.pwr.wroc.pl>

20 years agoAllow encoding modules to override the default implementations of
tjr [Thu, 13 May 2004 11:20:27 +0000 (11:20 +0000)]
Allow encoding modules to override the default implementations of
mbsrtowcs() and wcsrtombs(). Provide a fast implementation for the
trivial "NONE" encoding.

20 years agoMFi386: revision 1.489
nyan [Thu, 13 May 2004 11:17:07 +0000 (11:17 +0000)]
MFi386: revision 1.489

20 years agoMove the ichwd to i386 only.
nyan [Thu, 13 May 2004 11:13:55 +0000 (11:13 +0000)]
Move the ichwd to i386 only.

20 years agoMake getprotobynumber() calls in FilterCheck conditional on the log
dds [Thu, 13 May 2004 09:03:00 +0000 (09:03 +0000)]
Make getprotobynumber() calls in FilterCheck conditional on the log
levels by which they are used.  On a typical production setting (no
debug or filter logging) this will save an open/read/close system
call sequence per packet, approximately halving the system overhead
and reducing the overall overhead by 38%.

dd bs=1k count=512 if=/usr/share/dict/web2 |
ssh ppp-linked-host dd of=/dev/null

# time original-ppp -nat -foreground connection
Working in foreground mode
Using interface: tun0
2.822u 2.404s 2:00.31 4.3%    392+496k 8+18io 3pf+0w

# time new-ppp  -nat -foreground connection
Working in foreground mode
Using interface: tun0
2.082u 1.173s 1:26.06 3.7%    379+450k 0+18io 0pf+0w

MFC after: 3 weeks

20 years agoRecord the libssl.so dependency on libcrypto.so. This should
ru [Thu, 13 May 2004 07:51:47 +0000 (07:51 +0000)]
Record the libssl.so dependency on libcrypto.so.  This should
help some ports that depend on libradius that recently gained
the dependency on libssl.  This is also how the stock OpenSSL
build would link libssl.so on FreeBSD.

Prompted by: kris
OK'ed by: markm, nectar

20 years agoUpdate raw byte count statistic correctly.
kientzle [Thu, 13 May 2004 06:38:30 +0000 (06:38 +0000)]
Update raw byte count statistic correctly.

20 years agoAdd MLINK for newly-added archive_read_extract_set_progress_callback(3).
kientzle [Thu, 13 May 2004 06:04:21 +0000 (06:04 +0000)]
Add MLINK for newly-added archive_read_extract_set_progress_callback(3).

20 years agoAdd hook for a client-provided progress callback to be invoked
kientzle [Thu, 13 May 2004 06:01:14 +0000 (06:01 +0000)]
Add hook for a client-provided progress callback to be invoked
during lengthy extract operations.

20 years agoHandle variable argument lists correctly in reply() and lreply().
tjr [Thu, 13 May 2004 05:36:38 +0000 (05:36 +0000)]
Handle variable argument lists correctly in reply() and lreply().
In particular, do not pass the same va_list to both vprintf() and
vsyslog() without first reinitializing it. This fixes ftpd -d
on amd64.

20 years agoAdd support for retrieving _GLK in the ECDT probe. Now we no longer always
njl [Thu, 13 May 2004 03:17:08 +0000 (03:17 +0000)]
Add support for retrieving _GLK in the ECDT probe.  Now we no longer always
use the global lock at the beginning of the ECDT probe.  Instead, we use
the handle from the ECDT to call _GLK.  Also, unify the device description.

20 years agoExpose USBVERBOSE as a first-class option. It will be needed soon as
imp [Thu, 13 May 2004 03:15:04 +0000 (03:15 +0000)]
Expose USBVERBOSE as a first-class option.  It will be needed soon as
an option.  Note that this option doesn't follow the normal USB_ or
Uxxx_ convention.  That's because it is this way in the upstream
provider and I didn't want to change that.

20 years agoFix braino in previous: check that the second byte in the character
tjr [Thu, 13 May 2004 03:08:28 +0000 (03:08 +0000)]
Fix braino in previous: check that the second byte in the character
buffer is non-null when the character is two bytes long, not when
the buffer is two bytes long.

20 years agoSync to pccarddevs 1.84
imp [Thu, 13 May 2004 01:24:48 +0000 (01:24 +0000)]
Sync to pccarddevs 1.84

20 years agoAdd Intel EtherExpress PRO/100 LAN Modem. This will eventually be
imp [Thu, 13 May 2004 01:24:26 +0000 (01:24 +0000)]
Add Intel EtherExpress PRO/100 LAN Modem.  This will eventually be
added to xe once the pccard issues are resolved...

20 years agoMoved FreeBSD build pollution (/etc/make.conf) and zombie aout support
bde [Wed, 12 May 2004 22:53:52 +0000 (22:53 +0000)]
Moved FreeBSD build pollution (/etc/make.conf) and zombie aout support
(OBJFORMAT) into a non-POSIX section.

20 years agoRemoved conditional include of /etc/make.conf.local and the error for the
bde [Wed, 12 May 2004 21:51:11 +0000 (21:51 +0000)]
Removed conditional include of /etc/make.conf.local and the error for the
existence of this file.  This stopped working more than 4 years ago when
the generation of the error was added in rev.1.44.  The .error directive
gives fatal errors, so stale /etc/make.conf.local files must have been
gone away more than 4 years on systems where make(1) works.

20 years agoFix some^Wseveral style bugs from last commit.
peadar [Wed, 12 May 2004 21:38:39 +0000 (21:38 +0000)]
Fix some^Wseveral style bugs from last commit.

Remove "sys/types.h" as "sys/param.h" is already included

Use cast rather than back-pointer to convert from public to private
version of FTS data, and so avoid littering fts.h with any of the
details.

Pointed out By: bde, kientzle

20 years agoFix typo.
josef [Wed, 12 May 2004 20:45:51 +0000 (20:45 +0000)]
Fix typo.

Submitted by: Michel Lavondes <fox@vader.aacc.cc.md.us>
PR: docs/66576

20 years agoFix typo.
josef [Wed, 12 May 2004 20:40:09 +0000 (20:40 +0000)]
Fix typo.

Submitted by: Michel Lavondès <fox@vader.aacc.cc.md.us>
PR: docs/66538

20 years agoFix typo.
josef [Wed, 12 May 2004 20:33:25 +0000 (20:33 +0000)]
Fix typo.

Submitted by: Michel Lavondès <fox@vader.aacc.cc.md.us>
PR: docs/66536

20 years agoRestore source code compatibility with 5.2-RELEASE.
wpaul [Wed, 12 May 2004 15:58:42 +0000 (15:58 +0000)]
Restore source code compatibility with 5.2-RELEASE.

20 years agoReduce overhead by calling internal versions of the multibyte conversion
tjr [Wed, 12 May 2004 14:26:54 +0000 (14:26 +0000)]
Reduce overhead by calling internal versions of the multibyte conversion
functions directly wherever possible.

20 years agoMove prototypes of various encoding-related functions into a new header
tjr [Wed, 12 May 2004 14:09:04 +0000 (14:09 +0000)]
Move prototypes of various encoding-related functions into a new header
file to avoid extern'ing them all over the place.

20 years agoConvert the #if 0 magic to #if SCAN_IICBUS, and make it actually compile
joerg [Wed, 12 May 2004 13:43:41 +0000 (13:43 +0000)]
Convert the #if 0 magic to #if SCAN_IICBUS, and make it actually compile
again.  While it's not generally recommended anymore, it might still prove
useful for debugging purposes.

20 years agoLink radixsort(3) to sradixsort(3), make the latter appear in
ru [Wed, 12 May 2004 08:13:40 +0000 (08:13 +0000)]
Link radixsort(3) to sradixsort(3), make the latter appear in
the whatis(1) output.

20 years agoOutput style nit.
ru [Wed, 12 May 2004 07:36:59 +0000 (07:36 +0000)]
Output style nit.

20 years agoPolish the mdoc(7) markup.
ru [Wed, 12 May 2004 07:34:05 +0000 (07:34 +0000)]
Polish the mdoc(7) markup.

20 years agoCatch up with the current output of tools/make_libdeps.sh:
ru [Wed, 12 May 2004 07:02:17 +0000 (07:02 +0000)]
Catch up with the current output of tools/make_libdeps.sh:

- Removed the stale dependency of libypclnt on librpcsvc.
- Fixed the dependency graph of libssh.

20 years agoPut crypto-aware version of the library into the right distribution.
ru [Wed, 12 May 2004 06:31:36 +0000 (06:31 +0000)]
Put crypto-aware version of the library into the right distribution.

20 years agoRestructure vm_page_select_cache() so that adding assertions is easy.
alc [Wed, 12 May 2004 04:27:18 +0000 (04:27 +0000)]
Restructure vm_page_select_cache() so that adding assertions is easy.

Some of the conditions that caused vm_page_select_cache() to deactivate a
page were wrong.  For example, deactivating an unmanaged or wired page is a
nop.  Thus, if vm_page_select_cache() had ever encountered an unmanaged or
wired page, it would have looped forever.  Now, we assert that the page is
neither unmanaged nor wired.

20 years ago1. Fixed potential problem that would cause out-of-order requests in twe_startio.
vkashyap [Wed, 12 May 2004 04:10:37 +0000 (04:10 +0000)]
1. Fixed potential problem that would cause out-of-order requests in twe_startio.
2. Changed version.

Submitted by: scottl
Reviewed by: vkashyap
Approved by: re

20 years agoCache queue pages are not mapped. Thus, the pmap_remove_all() by
alc [Wed, 12 May 2004 04:10:35 +0000 (04:10 +0000)]
Cache queue pages are not mapped.  Thus, the pmap_remove_all() by
vm_pageout_scan()'s loop for freeing cache queue pages is unnecessary.

20 years agomdoc police.
des [Tue, 11 May 2004 20:47:17 +0000 (20:47 +0000)]
mdoc police.

Submitted by: ru

20 years agoaic79xx.c:
gibbs [Tue, 11 May 2004 20:46:05 +0000 (20:46 +0000)]
aic79xx.c:
Allow 500us between pauses in ahd_pause_and_flushwork().
The maximum we will wait is now 500ms.

In the same routine, remove any attempt to clear ENSELO.
Let the firmware do it once the current selection has
completed.  This avoids some race conditions having to
do with non-packetized completions and the auto-clearing
of ENSELO on packetized completions.

Also avoid attempts to clear critical sections when
interrups are pending.  We are going to loop again
anyway, so clearing critical sections is a waste of
time.  It also may not be possible to clear a critical
section if the source of the interrupt was a SEQINT.

aic79xx_pci.c:
Use the Generic 9005 mask when looking for generic 7901B
parts.  This allows the driver to attach to 7901B parts
on motherboards using a non-Adaptec subvendor ID.

aic79xx_inline.h:
Test for the SCBRAM_RD_BUG against the bugs
field, not the flags field in the softc.

aic79xx.c:
Cancel pending transactions on devices that
respond with a selection timeout.  This decreases
the duration of timeout recovery when a device
disappears.

aic79xx.c:
Don't bother forcing renegotiation on a selection
timeout now that we use the device reset handler
to abort any pending commands on the target.
The device reset handler already takes us down
to async narrow and forces a renegotiation.

In the device reset handlers, only send a
BDR sent async event if the status is not
CAM_SEL_TIMEOUT.  This avoids sending this
event in the selection timeout case

aic79xx.c:
Modify the Core timeout handler to verify that another
command has the potential to timeout before passing off
a command timeout as due to some other command.  This
safety measure is added in response to a timeout recovery
failure on H2B where it appears that incoming reselection
status was lost during a drive pull test.  In that case,
the recovery handler continued to wait for the command
that was active on the bus indefinetly.  While the root
cause of the above issue is still being determined seems
a prudent safeguard.

aic79xx_pci.c:
Add a specific probe entry for the Dell OEM 39320(B).

aic79xx.c:
aic79xx.h:
aic79xx.reg:
aic79xx.seq:
Modify the aic79xx firmware to never cross a cacheline or
ADB boundary when DMA'ing completion entries to the host.
In PCI mode, at least in 32/33 configurations, the SCB
DMA engine may lose its place in the data-stream should
the target force a retry on something other than an
8byte aligned boundary. In PCI-X mode, we do this to
avoid split transactions since many chipsets seem to be
unable to format proper split completions to continue
the data transfer.

The above change allows us to drop our completion entries
from 8 bytes to 4.  We were using 8 byte entries to ensure
that PCI retries could only occur on an 8byte aligned
boundary.  Now that the sequencer guarantees this by splitting
up completions, we can safely drop the size to 4 bytes (2
byte tag, one byte SG_RESID, one byte pad).

Both the split-completion and PCI retry problems only show
up under high tag load when interrupt coalescing is being
especially effective.  The switch from a 2byte completion
entry to an 8 byte entry to solve the PCI problem increased
the chance of incurring a split in PCI-X mode when multiple
transactions were completed at once.  Dropping the completion
size to 4 bytes also means that we can complete more commands
in a single DMA (128byte FIFO -> 32 commands instead of 16).

aic79xx.c:
Modify the SCSIINT handler to defer clearing
sequencer critical sections to the individual
interrupt handlers.  This allows us to
immediately disable any outgoing selections in
the case of an unexpected busfree so we don't
inadvertantly clear ENSELO *after* a new selection
has started.  Doing so may cause the sequencer
to miss a successful selection.

In ahd_update_pending_scbs(), only clear ENSELO if
the bus is currently busy and a selection is not
already in progress or the sequencer has yet to
handle a pending selection.  While we want to ensure
that the selection for the SCB at the head of the
selection queue is restarted so that any change in
negotiation request can take effect, we can't clobber
pending selection state without confusing the sequencer
into missing a selection.

20 years agoo When restarting the sequencer, clear any pending
gibbs [Tue, 11 May 2004 20:39:46 +0000 (20:39 +0000)]
o When restarting the sequencer, clear any pending
  sequencer interrupt codes.  These codes are only
  relevant to the code that was last being executed
  and that context is cleared when we reset the
  program counter.  This addresses a race condition
  between a sequencer interrupt and any SCSI event
  that causes us to restart the sequencer.

o When running the untagged-Q, we must start the
  timer for any transaction we queue.

o Give the firmware half a millisecond between
  pauses to flush work out.  This should give us
  around half a second of total delay before flagging
  an issue with pausing and flushing controller work.

  Only attempt to clear critical sections if there
  are no pending interrupts in the pause and flush
  loop.  If the sequencer has issued an INTSTAT, we
  may not be able to step out of the critical section.

o Cancel pending transactions on devices that
  respond with a selection timeout.  This decreases
  the duration of timeout recovery when a device
  disappears.

  Don't bother forcing renegotiation on a selection
  timeout now that we use the device reset handler
  to abort any pending commands on the target.
  The device reset handler already takes us down
  to async narrow and forces a renegotiation.

o In the device reset handlers, only send a
  BDR sent async event if the status is not
  CAM_SEL_TIMEOUT.  This avoids sending this
  event in the selection timeout case.

o Modify the Core timeout handler to verify that another
  command has the potential to timeout before passing off
  a command timeout as due to some other command.

20 years agoAdd aic_scb_timer_start() which will be used in the aic7xxx driver to
gibbs [Tue, 11 May 2004 20:33:42 +0000 (20:33 +0000)]
Add aic_scb_timer_start() which will be used in the aic7xxx driver to
start the timer on SCBs queued in the untagged SCB queue.

The core treats timeouts in usecs to satisfy Linux requirements.  Multiply
accordingly.