piso [Thu, 31 May 2007 19:25:35 +0000 (19:25 +0000)]
In some particular cases (like in pccard and pccbb), the real device
handler is wrapped in a couple of functions - a filter wrapper and an
ithread wrapper. In this case (and just in this case), the filter
wrapper could ask the system to schedule the ithread and mask the
interrupt source if the wrapped handler is composed of just an ithread
handler: modify the "old" interrupt code to make it support
this situation, while the "new" interrupt code is already ok.
ariff [Thu, 31 May 2007 18:43:33 +0000 (18:43 +0000)]
Last major commit and updates for RELENG_7:
- Rework the entire pcm_channel structure:
* Remove rarely used link placeholder, instead, make each pcm_channel
as head/link of each own/each other. Unlock - Lock sequence due to
sleep malloc has been reduced.
* Implement "busy" queue which will contain list of busy/active
channels. This greatly reduce locking contention for example while
servicing interrupt for hardware with many channels or when virtual
channels reach its 256 peak channels.
- So I heard you like v chan ... O RLY?
Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for
recording, Rec-Chan, you decide), the ultimate solutions for your
nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing
single record channel causing EBUSY. Vrec works exactly like Vchans
(or, should I rename it to "Vplay" :) , except that it operates on the
opposite direction (recording). Up to 256 vrecs (like vchans) are
possible.
Notes:
* Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its
respective node/direction:
dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d)
dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d)
* Don't expect that it will magically give you ability to split
"recording source" (eg: 1 channel for cdrom, 1 channel for mic,
etc). Just admit that you only have a *single* recording source /
channel. Please bug your hardware vendor instead :)
- Bump maxautovchans from 4 to 16. For a full-fledged multimedia
desktop/workstation with too many soundservers installed (esound,
artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh,
etc), 4 seems inadequate. There will be no memory penalty here, since
virtual channels are allocate only by demand.
- Nuke/Rework the entire statically created cdev entries. Everything is
clonable through snd own clone manager which designed to withstand many
kind of abusive devfs droids such as:
* while : ; do /bin/test -e /dev/dsp ; done
* jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done
* hundreds (could be thousands) concurrent threads/process opening
"/dev/dsp" (previously, this might result EBUSY even with just
3 contesting threads/procs).
o Reusable clone objects (instead of creating new one like there's no
tomorrow) after certain expiration deadline. The clone allocator will
decide whether to reuse, share, or creating new clone.
o Automatic garbage collector.
- Dynamic unit magic allocator. Maximum attached soundcards can be tuned
using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and
maximum is 2048.
- ..other fixes, mostly related to concurrency issues.
deischen [Thu, 31 May 2007 13:07:37 +0000 (13:07 +0000)]
Export quad symbols. They were previously commented out. These symbols
really shouldn't be exported since they should be pulled from libgcc, but
the build of some applications is broken and they expect to see them in
libc. glibc exports these symbols, although Solaris doesn't appear to,
so export them for compatibility's sake.
kib [Thu, 31 May 2007 11:51:53 +0000 (11:51 +0000)]
Revert UF_OPENING workaround for CURRENT.
Change the VOP_OPEN(), vn_open() vnode operation and d_fdopen() cdev operation
argument from being file descriptor index into the pointer to struct file.
Proposed and reviewed by: jhb
Reviewed by: daichi (unionfs)
Approved by: re (kensmith)
rwatson [Thu, 31 May 2007 11:51:22 +0000 (11:51 +0000)]
Now that sx(9) locks support an interruptible lock acquire primitive,
properly observe the SB_NOINTR flag in sblock. This restores the
required behavior that lock acquisition be interruptible on the socket
buffer I/O serialization lock to allow threads waiting for I/O to be
signaled even if they aren't the thread currently holding the I/O lock.
With this change, the sblock regression test is again passed.
attilio [Thu, 31 May 2007 09:14:48 +0000 (09:14 +0000)]
Add functions sx_xlock_sig() and sx_slock_sig().
These functions are intended to do the same actions of sx_xlock() and
sx_slock() but with the difference to perform an interruptible sleep, so
that sleep can be interrupted by external events.
In order to support these new featueres, some code renstruction is needed,
but external API won't be affected at all.
Note: use "void" cast for "int" returning functions in order to avoid tools
like Coverity prevents to whine.
imp [Thu, 31 May 2007 03:40:29 +0000 (03:40 +0000)]
Add information for FreeBSD 1.0.2 from November 1993. I have a
January 1994 Trans-Ameritech Release 1 CD-ROM that has this directory
named "update_bsd_1.0.2" dated November 14, 1993.
I didn't add it into the main-line sequence of releases because that
would disrupt things too much...
njl [Thu, 31 May 2007 00:52:32 +0000 (00:52 +0000)]
Remove "acpi_bus_number: can't get _ADR" message. It usually appears as
we traverse \_SB and \ in the namespace, which won't have _ADR anyway.
Use a proper extern instead of our own private copy.
jfv [Wed, 30 May 2007 23:32:21 +0000 (23:32 +0000)]
A few small but significant fixes:
- Coverity Prevent(tm) CID 1906 a bogus use of bzero where unneeded.
- ICH8 systems autoneg to 100 rather than 1000, this can also be
seen in 82573, the logic was backwards.
- On new 82575 quadports half duplex tx speed is slow... this was due
to overwriting TCTL reg rather than adding bits.
rrs [Wed, 30 May 2007 22:34:21 +0000 (22:34 +0000)]
- Fixed (Apple) compiler warnings in sctp_input.c, sctputil.c, sctp_output.c
- Fixed a LOR in handling a cookie. Turns out create lock is applied.
And if we abort processing, this causes LOR. Changed to force the
timer to clean up, that way create lock is released.
rrs [Wed, 30 May 2007 17:39:45 +0000 (17:39 +0000)]
- Fix a memory overwrite when the mapping array
is expanded, size of expansion was not taken int consideration.
- Fix so vtag hash is 1 bigger so that it modulo's out
correctly, avoids a panic when restart with right modulo happens.
- do not dereference stcb when control->do_not_ref_stcb is set
- Fix up packet logging to not often use a lock and also to
add to options.
- Fix some logging option duplication in the sctputil.h
bz [Tue, 29 May 2007 22:32:12 +0000 (22:32 +0000)]
Implement ICMPv6 support in ipsec6_get_ulp().
This is needed to make security policies work correctly if ICMPv6 type
and/or code are given. See setkey(8) 'upperspec' para. for details.
imp [Tue, 29 May 2007 20:05:13 +0000 (20:05 +0000)]
Add more devices from the hps tree. These appear to have come from
OpenBSD's if_ral.c.
I didn't make the LINKSYS4 -> CISCOLINKSYS name change, nor did I
include the RALINK RT2573 that's supported by the rum(4) driver. I
didn't merge any code changes either.
attilio [Tue, 29 May 2007 19:34:32 +0000 (19:34 +0000)]
Add a small fix for lock profiling in sx locks.
"0" cannot be a correct value since when the function is entered at least
one shared holder must be present and since we want the last one "1" is
the correct value.
Note that lock_profiling for sx locks is far from being perfect.
Expect further fixes for that.
attilio [Tue, 29 May 2007 18:55:41 +0000 (18:55 +0000)]
Fix some problems introduced with the last descriptors tables locking
patch:
- Do the correct test for ldt allocation
- Drop dt_lock just before to call kmem_free (since it acquires blocking
locks inside)
- Solve a deadlock with smp_rendezvous() where other CPU will wait
undefinitively for dt_lock acquisition.
- Add dt_lock in the WITNESS list of spinlocks
While applying these modifies, change the requirement for user_ldt_free()
making that returning without dt_lock held.
jhay [Tue, 29 May 2007 18:10:42 +0000 (18:10 +0000)]
Remove the hardcoded IXP425_UART?_VBASE values in the
uart_ixp425_probe() and uart_cpu_getdev(). Change
uart_cpu_getdev() to use hints to find the console.
bmah [Tue, 29 May 2007 16:14:06 +0000 (16:14 +0000)]
Add a few sentences about updating. Some older text on this subject
went away around the time of 5.0, when doing binary upgrades from
within sysinstall (at least from 4.X) was deemed hazardous due to the
huge differences between these two major versions.
It's much less risky now, and it's also high time that we mentioned
freebsd-update(8) too.
sam [Tue, 29 May 2007 16:13:59 +0000 (16:13 +0000)]
Drain task q items when transitioning to INIT state; this closes a
race seen on smp laptops when suspending where the rx task can be
entered after the interface is detach'd.
NB: use of taskqueue_drain while holding the softc mutex is problematic
simokawa [Tue, 29 May 2007 14:35:57 +0000 (14:35 +0000)]
MFp4: add FireWire/dcons support in loader for i386/amd64.
It is disabled by default. You need to put
LOADER_FIREWIRE_SUPPORT=yes in /etc/make.conf
and rebuild loader to enable it.
(cd /sys/boot/i386 && make clean && make && make install)
You can find a short introduction of dcons at
http://wiki.freebsd.org/DebugWithDcons
yar [Tue, 29 May 2007 12:40:45 +0000 (12:40 +0000)]
Sync ether_ioctl() with ioctl(2) and ifnet.if_ioctl
as to the type of the command argument: int -> u_long.
These types have different widths in the 64-bit world.
Add a note to UPDATING because the change breaks KBI
on 64-bit platforms.
rrs [Tue, 29 May 2007 09:29:03 +0000 (09:29 +0000)]
- Fixes so we won't try to start a timer when we
hold a wq lock for the iterator. Panda uses a
silly recursive lock they hold through the timer.
- Add poor mans wireshark compile option..
- Allocate and start using SCTP_M_XXX for all SCTP_MALLOC() calls.
- sysctl now will get back the refcnt for viewing by onlookers.
bde [Tue, 29 May 2007 07:13:07 +0000 (07:13 +0000)]
Merge the relevant part of rev.1.14 of s_cbrt.c (a micro-optimization
involving moving the check for x == 0). The savings in cycles are
smaller for cbrtf() than for cbrt(), and positive in all measured cases
with gcc-3.4.4, but still very machine/compiler-dependent.
yongari [Tue, 29 May 2007 06:30:26 +0000 (06:30 +0000)]
Honor maxsegsz of less than a page size in a DMA tag. Previously it
used to return PAGE_SIZE without respect to restrictions of a DMA tag.
This affected all of the busdma load functions that use
_bus_dmamap_loader_buffer() as their back-end.
dougb [Tue, 29 May 2007 06:22:14 +0000 (06:22 +0000)]
Now that a separate /usr/X11R6 directory is no longer in fashion,
stop looking there for things like rc.d and periodic. This avoids
duplicating effort when /usr/X11R6 is a symlink to /usr/local,
which it is by default now.
It is not anticipated at this time that we will MFC this change, since
we'd like to avoid breaking legacy systems. However, there is a fix for
/etc/rc.subr in the works to avoid running any rc.d scripts twice which
we should be able to MFC.
kientzle [Tue, 29 May 2007 05:39:10 +0000 (05:39 +0000)]
bsdtar 2.2.3:
* Implement --use-compress-program using new libarchive feature.
* Minor portability improvement by adjusting casts used to
print out uids, gids, and device numbers.
Thanks to: Joerg Sonnenberger for the --use-compress-program implementation.
MFC after: 15 days
marcel [Tue, 29 May 2007 03:49:00 +0000 (03:49 +0000)]
"make delete-old" needs more than 64KB worth of arguments. Both
arm and powerpc have 64KB as the maximum argument size, so one
cannot run "make delete-old" on arm or powerpc anymore. Stop
special-casing powerpc and give it 256KB of arguments like all
other platforms, but keep arm on 64KB for now. There may be a
purpose to it that doesn't exist for powerpc.
kientzle [Tue, 29 May 2007 01:00:21 +0000 (01:00 +0000)]
libarchive 2.2.3
* "compression_program" support uses an external program
* Portability: no longer uses "struct stat" as a primary
data interchange structure internally
* Part of the above: refactor archive_entry to separate
out copy_stat() and stat() functions
* More complete tests for archive_entry
* Finish archive_entry_clone()
* Isolate major()/minor()/makedev() in archive_entry; remove
these from everywhere else.
* Bug fix: properly handle decompression look-ahead at end-of-data
* Bug fixes to 'ar' support
* Fix memory leak in ZIP reader
* Portability: better timegm() emulation in iso9660 reader
* New write_disk flags to suppress auto dir creation and not
overwrite newer files (for future cpio front-end)
* Simplify trailing-'/' fixup when writing tar and pax
* Test enhancements: fix various compiler warnings, improve
portability, add lots of new tests.
* Documentation: document new functions, first draft of
libarchive_internals.3
MFC after: 14 days
Thanks to: Joerg Sonnenberger (compression_program)
Thanks to: Kai Wang (ar)
Thanks to: Colin Percival (many small fixes)
Thanks to: Many others who sent me various patches and problem reports.
kan [Mon, 28 May 2007 23:02:56 +0000 (23:02 +0000)]
Disable shared libgcc use for profiled builds. GNU linker pulls symbols
out of libgcc_s.so.1 somehow and fails to record DT_NEEED dependency
when linking with profiled libraries despite --ld-as-needed option.
kmacy [Mon, 28 May 2007 22:57:27 +0000 (22:57 +0000)]
When building cxgb as a module make include paths relative to the driver's root.
This will make it possible to build the module out of tree against an older src tree.
emaste [Mon, 28 May 2007 19:51:12 +0000 (19:51 +0000)]
Eliminate explicit kdb_enter in the software watchdog handler (which
produced incorrect behaviour with the KDB_UNATTENDED option) and call
panic in both the KDB and non-KDB cases. This change is consistent
with rwatson's current kdb/ddb work.
jhay [Mon, 28 May 2007 18:54:08 +0000 (18:54 +0000)]
We do not need to get the irq out of ivars in ixp425_setup_intr(). By
this time they have already been set. In fact trying to set it here too
breaks irqs for pci devices.
rwatson [Mon, 28 May 2007 18:20:15 +0000 (18:20 +0000)]
Don't save SYSCTL_ADD_*() results in a local variable just to throw them
away; preserve the ones that are needed for further calls in the init
function and ignore the rest entirely.
ariff [Mon, 28 May 2007 16:22:07 +0000 (16:22 +0000)]
- Enable soft pcm volume flag early to ensure it not being clobbered
by the subsequent mix_setdevs() and friends.
- Minor style(9) declaration arrangement nit.
ariff [Mon, 28 May 2007 14:09:06 +0000 (14:09 +0000)]
Fix broken "rec" and "igain" introduced by previous commit. Convert
reg to a full blown int since there's not much gain compacting it,
and we do need its signess.
rrs [Mon, 28 May 2007 11:17:24 +0000 (11:17 +0000)]
- fixed autclose to not allow setting on 1-2-1 model.
- bounded cookie-life to 1 second minimum in socket option set.
- Delayed_ack_time becomes delayed_ack per new socket api document.
- Improve port number selection, we now use low/high bounds and
no chance of a endless loop. Only one call to random per bind
as well.
- fixes so set_peer_primary pre-screens addresses to be
valid to this host.
- maxseg did not allow setting on an assoc basis. We needed
to thus track and use an association value instead of a inp value.
- Fixed ep get of HB status to report back properly.
- use settings flag to tell if assoc level hb is on off not
the timer.. since the timer may still run if unconf address
are present.
- check for crazy ENABLE/DISABLE conditions.
- set and get of pmtud (fixed path mtu) not always taking into account ovh.
- Getting PMTU info on stcb only needs to return PMTUD_ENABLED if
any net is doing PMTU discovery.
- Panic or warning fixed to not do so when a valid ip frag is
taking place.
- sndrcvinfo appearing in both inp and stcb was full size, instead
of the non-pad version. This saves about 92 bytes from each struct
by carefully converting to use the smaller version.
- one-2-one model get(maxseg) would always get ep value, never the
tcb's value.
- The delayed ack time could be under a tick, this fixes so
it bounds it to at least 1 tick for platforms whos tick
is more than a ms.
- Fragment interleave level set to wrong default value.
- Fragment interleave could not set level 0.
- Defered stream reset was broken due to a guard check and ntohl issue.
- Found two lock order reversals and fixed.
- Tighten up address checking, if the user gives an address the sa_len
had better be set properly.
- Get asoc by assoc-id would return a locked tcb when it was asked
not to if the tcb was in the restart hash.
- sysctl to dig down and get more association details
andre [Mon, 28 May 2007 11:03:53 +0000 (11:03 +0000)]
Refactor and rewrite in parts the SYN handling code on listen sockets
in tcp_input():
o tighten the checks on allowed TCP flags to be RFC793 and
tcp-secure conform
o log check failures to syslog at LOG_DEBUG level
o rearrange the code flow to be easier to follow
o add KASSERTs to validate assumptions of the code flow
Add sysctl net.inet.tcp.syncache.rst_on_sock_fail defaulting to enable
that controls the behavior on socket creation failure for a otherwise
successful 3-way handshake. The socket creation can fail due to global
memory shortage, listen queue limits and file descriptor limits. The
sysctl allows to chose between two options to deal with this. One is
to send a reset to the other endpoint to notify it about the failure
(default). The other one is to ignore and treat the failure as a
transient error and have the other endpoint retransmit for another try.
pjd [Mon, 28 May 2007 02:37:43 +0000 (02:37 +0000)]
Adjust va_mask for setattr. FreeBSD doesn't have va_mask, so we initialize it
based on individual fields beeing set. This doesn't work for setattr replay,
because va_type is set there, so we add AT_TYPE flag to va_mask, which won't
be accepted by zfs_setattr().
kmacy [Sun, 27 May 2007 22:07:47 +0000 (22:07 +0000)]
Tuning for small packet handling
- Double the number of descriptors that a single call to send can use
- Quadruple the number of descriptors that can be reclaimed per pass
- only run reclaim twice per second
- increase coalesce timer from 3.5us to 5us