Olivier Houchard [Mon, 14 Nov 2005 22:24:00 +0000 (22:24 +0000)]
Add a new sysctl, kern.elf[32|64].can_exec_dyn. When set to 1, one can
execute a ET_DYN binary (shared object).
This does not make much sense, but some linux scripts expect to be able to
execute /lib/ld-linux.so.2 (ldd comes to mind).
The sysctl defaults to 0.
Jung-uk Kim [Mon, 14 Nov 2005 21:54:20 +0000 (21:54 +0000)]
0xb1881106 seems to be an AGP bridge and some BIOSes incorrectly handle
the bridge. Therefore, we give the same treatment as we did for nForce3-250
and ULi chipsets. VIA AGPv3 code was copied from agp_via.c.
Robert Watson [Mon, 14 Nov 2005 19:30:09 +0000 (19:30 +0000)]
In ktr_getrequest(), acquire ktrace_mtx earlier -- while the race
currently present is minor and offers no real semantic issues, it also
doesn't make sense since an earlier lockless check has already
occurred. Also hold the mutex longer, over a manipulation of
per-process ktrace state, which requires synchronization.
Ariff Abdullah [Mon, 14 Nov 2005 18:21:23 +0000 (18:21 +0000)]
From luigi:
This one simply tries to simplify the logic to select the
buffer sizes. I am not sure it is necessary but the code
seems a bit more readable to me. And at least i have tried
to document how the buffer sizes are computed.
Thanks to luigi for deciphering one of the most cryptic part of
sound driver.
Submitted by: luigi
Approved by: netchild (mentor)
Ariff Abdullah [Mon, 14 Nov 2005 18:20:47 +0000 (18:20 +0000)]
From luigi:
In SNDCTL_DSP_SETFRAGMENT, if you specify both read and
write channels, the existing code first acts on the
read channel, but as a side effect it updates the
arguments (maxfrags, fragsz) passed by the caller according
to acceptable values for the read channel, and then uses the
modified values to act on the write channel.
The problem with this approach is that, given a
(maxfrags, fragsz) user-specified value, the actual
values computed by the read and write channels may differ:
e.g. the read channel might want to allocate more fragments
than what the user specified because it has no side-effects
on the delay and it helps in case of slow readers,
whereas the write channel needs to use as few fragments
as possible to keep the audio latency low (very important
with telephony apps).
This patch stores the values computed by the read channel
into temproary variables so the write channel will use
the actual arguments of the ioctl.
This patch is very helpful with telephony apps such as asterisk.
Submitted by: luigi
Approved by: netchild (mentor)
Ruslan Ermilov [Mon, 14 Nov 2005 18:19:57 +0000 (18:19 +0000)]
Unlike the rest of the world, NDIS code can access "struct
ifnet" before is has been fully initialized by if_attach().
Account for that to avoid a null pointer dereference.
Ariff Abdullah [Mon, 14 Nov 2005 18:19:33 +0000 (18:19 +0000)]
ac97.c:
- Added new codec id for CX20468-21 and VIA1617A.
Submitted by: Chen Lihong <lihong.chen@gmail.com>
- Re-enable SOUND_MIXER_IGAIN, but set the default level as 0 (mute)
Suggested by: luigi
mixer.c:
- Set default value for SOUND_MIXER_IGAIN as 0 (mute) to avoid
feedback problems on some laptops (was disabled by jhb during
ac97.c revision 1.42).
Ariff Abdullah [Mon, 14 Nov 2005 18:18:52 +0000 (18:18 +0000)]
Fix a long standing unhandled interrupt bug which can cause
erratic system slowdown (beaten to a pulp) and possible panic. This
issue has bugged me for as long as I could remember, until I
realized that it is possible for register base offset to hold zero
value which is definitely a "FALSE".
Ariff Abdullah [Mon, 14 Nov 2005 18:18:12 +0000 (18:18 +0000)]
- Added few more Intel HDA ids (ICH 6/7) which does have backward
compatible AC97 codec.
- As the driver supports so many variants, create a table ids for
ease of probing and maintenance.
Submitted by: yongari
Reviewed/Tested by: multimedia@
- From luigi:
The code to compute fragment sizes in the ich driver almost
invariably ends up using the full buffer available, no matter
how the user specifies fragment size and number.
With audio telephony (8khz, 16bit-stereo) and the 16k buffer
size this results in an unbearable 500ms delay.
This patch makes sure that we never use more than 4 fragments,
(i don't think we need more unless there are huge interrupt
servicing latencies), and obey to the requested fragment size,
so that latency is acceptable.
Based on this (and after much regression tests), I can conclude
that this driver works best with 2 fragments, thus solving various
long standing issues of ICH driver not capable to flush or play
short files perfectly.
Suggested by: luigi (the idea of smaller fragments)
- MPSAFE conversion.
Ariff Abdullah [Mon, 14 Nov 2005 18:17:31 +0000 (18:17 +0000)]
Use both (enabled by default) DAC1 and DAC2 to provide 2
distinct hardware playback channels. DAC configuration can be
accessed through kernel hint - hint.pcm.<unit>.dac="val" with
following possible values:
0 = Enable both DACs (default)
1 = Enable single DAC (DAC1)
2 = Enable single DAC (DAC2)
3 = Enable both DACs, swap position (DAC2 comes first instead
of DAC1)
Special case for ES1370:
Unlike ES1371,2,3/CT5880, volume for each DAC 1 and 2 can be
controlled indepedently (synth for DAC1, pcm for DAC2). It is
possible that user will confuse by this behaviour, since both
DACs are enabled by default. Thus, provide a knob through sysctl
hw.snd.pcm<unit>.single_pcm_mixer:
0 = each DACs will be controlled separately (synth/pcm).
1 = combine both DACs volume mixer controller into a single
"pcm" (default)
As a side note, fixed rate operation (provided by previous
commit) is not a mandatory if the configuration space does not
involve DAC2 (perhaps disabled by user through the above kernel
hint). Unlike DAC2, DAC1 has its own register / control space,
not affected by the speed settings of ADC.
Marcel Moolenaar [Sun, 13 Nov 2005 22:02:04 +0000 (22:02 +0000)]
Update to match geom_gpt:1.35
o Change the result of gctl(001) now that a bogus verb still requires
a valid geom,
o Insert gctl(024) to test for an appropriate error when a bogus verb
is given that does have a proper geom parameter.
Marcel Moolenaar [Sun, 13 Nov 2005 21:53:55 +0000 (21:53 +0000)]
o Slightly refactor the ctlreq code to maximize code sharing between
verbs. Only the create verb operates on a provider. All other verbs
operate on a GPT geom. Also, the GPT entry oriented verbs require
a non-downgraded GPT.
o Have all verbs take an optional flags parameter. The flags parameter
is a string of single-letter flags. The typical use of these flags
is to enable certain behaviour in support fo the gpt(8) tool.
o Add dummy implementations for the destroy and recover verbs.
This change causes test 2 of the GPT regression test suite to fail.
The presence of a geom parameter is now required even for unknown
verbs.
Andrew Thompson [Sun, 13 Nov 2005 19:36:59 +0000 (19:36 +0000)]
Fix a mbuf and refcnt leak in the broadcast code.
If the packet is rejected from pfil(9) then continue the loop rather than
returning, this means that we can still try to send it out the remaining
interfaces but more importantly the mbuf is freed and refcount decremented on
exit.
Damien Bergamini [Sun, 13 Nov 2005 17:25:21 +0000 (17:25 +0000)]
Be more robust when handling Rx interrupts. If we can't allocate a new mbuf,
just discard the received frame and reuse the old mbuf.
This should prevent the connection from stalling after high network traffic.
Damien Bergamini [Sun, 13 Nov 2005 17:19:12 +0000 (17:19 +0000)]
Be more robust when handling Rx interrupts. If we can't allocate and DMA map
a new mbuf, just discard the received frame and reuse the old mbuf.
This should fix kernel panics on high network traffic.
Damien Bergamini [Sun, 13 Nov 2005 17:17:40 +0000 (17:17 +0000)]
Be more robust when handling Rx interrupts. If we can't allocate and DMA map
a new mbuf, just discard the received frame and reuse the old mbuf.
This should fix kernel panics on high network traffic.
Damien Bergamini [Sun, 13 Nov 2005 17:16:06 +0000 (17:16 +0000)]
Be more robust when handling Rx interrupts. If we can't allocate and DMA map
a new mbuf, just discard the received frame and reuse the old mbuf.
This should fix kernel panics on high network traffic.
Add support for 24/32 bit audio formats/conversion.
It may be the case that you may hear some unwanted noise while
playing back with 24/32 bit. This is a problem in the USB system.
Explanation from Hans Petter Selasky:
---snip---
The current USB sound driver only uses one isochronous
buffer, that is restarted when it is completed. This will lead to a short
period of time, +1ms, where no sound data is sent to the external USB device.
Depending on the load of your computer, this can be as much as 50ms. So the
USB sound driver must use 2 isochronous transfers. At the beginning one will
queue both. Then these are restarted on completion. This will result in a
constant-rate data stream to the external sound device, a minimum sound
buffer equal to the size of the isochronous buffer, and possibly the sound
will reach your ears with less delay. Little delay is a result of constant
data rate. Currently only my USB driver will support that. If one tries that
with the USB driver in *BSD, then it will crash at the first moment one gets
a buffer underrun.
---snip---
Submitted by: Kazuhito HONDA <kazuhito@ph.noda.tus.ac.jp>
Mono-recording still not tested by: julian
Robert Watson [Sun, 13 Nov 2005 14:06:01 +0000 (14:06 +0000)]
Modify netstat -mb to use libmemstat when accessing a core dump or live
kernel memory and not using sysctl. Previously, libmemstat was used
only for the live kernel via sysctl paths.
This results in netstat output becoming both more consistent between
core dumps and the live kernel, and also more information in the core
dump case than previously (i.e., mbuf cache information).
Statistics relating to sfbufs still rely on a kvm descriptor as they
are not currently exposed via libmemstat. netstat -m operating on a
core is still unable to print certain sfbuf stats available on the live
kernel.
Robert Watson [Sun, 13 Nov 2005 13:27:44 +0000 (13:27 +0000)]
Moderate rewrite of kernel ktrace code to attempt to generally improve
reliability when tracing fast-moving processes or writing traces to
slow file systems by avoiding unbounded queueuing and dropped records.
Record loss was previously possible when the global pool of records
become depleted as a result of record generation outstripping record
commit, which occurred quickly in many common situations.
These changes partially restore the 4.x model of committing ktrace
records at the point of trace generation (synchronous), but maintain
the 5.x deferred record commit behavior (asynchronous) for situations
where entering VFS and sleeping is not possible (i.e., in the
scheduler). Records are now queued per-process as opposed to
globally, with processes responsible for committing records from their
own context as required.
- Eliminate the ktrace worker thread and global record queue, as they
are no longer used. Keep the global free record list, as records
are still used.
- Add a per-process record queue, which will hold any asynchronously
generated records, such as from context switches. This replaces the
global queue as the place to submit asynchronous records to.
- When a record is committed asynchronously, simply queue it to the
process.
- When a record is committed synchronously, first drain any pending
per-process records in order to maintain ordering as best we can.
Currently ordering between competing threads is provided via a global
ktrace_sx, but a per-process flag or lock may be desirable in the
future.
- When a process returns to user space following a system call, trap,
signal delivery, etc, flush any pending records.
- When a process exits, flush any pending records.
- Assert on process tear-down that there are no pending records.
- Slightly abstract the notion of being "in ktrace", which is used to
prevent the recursive generation of records, as well as generating
traces for ktrace events.
Future work here might look at changing the set of events marked for
synchronous and asynchronous record generation, re-balancing queue
depth, timeliness of commit to disk, and so on. I.e., performing a
drain every (n) records.
MFC after: 1 month
Discussed with: jhb
Requested by: Marc Olzheim <marcolz at stack dot nl>
Add some cards:
- several TerraTec TValue [1]
- PixelView PlayTV Pro REV-4C [2]
In case you have the PixelView card, please tell us the "pciconf -v -l"
output on multimedia@FreeBSD.org if it works. There are revisions out there
which may not work and we need to know which ones work.
Don't augment the DRIVER_VERSION "v1.12" with __DATE__ and __TIME__.
This is the only file of > 1700 files in a buildkernel here doing that.
It makes reproducible builds (same source => same binary) impossible.
David Xu [Sun, 13 Nov 2005 09:57:44 +0000 (09:57 +0000)]
Define SIGLWP which is an alias for SIGTHR, the reason why I did this
is that gdb knows SIGLWP and will pass it to program, otherwise gdb
will print out "unknown signal" and discard it, and then thread
cancellation won't work for libthr under gdb.
Craig Rodrigues [Sun, 13 Nov 2005 01:27:57 +0000 (01:27 +0000)]
- Make size parameter to build_iovec() a size_t, not an int
- Add build_iovec_argf() helper function, for help converting old
mount options which used the mount_argf() function for the mount() syscall.
Bruce Evans [Sun, 13 Nov 2005 00:41:46 +0000 (00:41 +0000)]
Fixed some magic numbers.
The threshold for not being tiny was too small. Use the usual 2**-12
threshold. This change is not just an optimization, since the general
code that we fell into has accuracy problems even for tiny x. Avoiding
it fixes 2*1366 args with errors of more than 1 ulp, with a maximum
error of 1.167 ulps.
The magic number 22 is log(DBL_EPSILON)/2 plus slop. This is bogus
for float precision. Use 9 (~log(FLT_EPSILON)/2 plus less slop than
for double precision). The code for handling the interval
[2**-28, 9_was_22] has accuracy problems even for [9, 22], so this
change happens to fix errors of more than 1 ulp in about 2*17000
cases. It leaves such errors in about 2*1074000 cases, with a max
error of 1.242 ulps.
The threshold for switching from returning exp(x)/2 to returning
exp(x/2)^2/2 was a little smaller than necessary. As for coshf(),
This was not quite harmless since the exp(x/2)^2/2 case is inaccurate,
and fixing it avoids accuracy problems in 2*6 cases, leaving problems
in 2*19997 cases.
Bruce Evans [Sun, 13 Nov 2005 00:08:23 +0000 (00:08 +0000)]
Fixed some magic numbers.
The threshold for not being tiny was confusing and too small. Use the
usual 2**-12 threshold and simplify the algorithm slightly so that
this threshold works (now use the threshold for sinhf() instead of one
for 1+expm1()). This is just a small optimization.
The magic number 22 is log(DBL_EPSILON)/2 plus slop. This is bogus
for float precision. Use 9 (~log(FLT_EPSILON)/2 plus less slop than
for double precision).
The threshold for switching from returning exp(x)/2 to returning
exp(x/2)^2/2 was a little smaller than necessary. This was not quite
harmless since the exp(x/2)^2/2 case is inaccurate. Fixing it happens
to avoid accuracy problems for 2*6 of the 2*151 args that were handled
by the exp(x)/2 case. This leaves accuracy problems for about 2*19997
args near the overflow threshold (~89); the maximum error there is
2.5029 ulps.
There are also accuracy probles for args in +-[0.5*ln2, 9] -- 2*188885
args with errors of more than 1 ulp, with a maximum error of 1.384 ulps.
Fixed a syntax error and naming errors in pseudo-code in comments.
Marcel Moolenaar [Sat, 12 Nov 2005 20:17:08 +0000 (20:17 +0000)]
Instead of saving the unit number of the md(4) device name, save the
whole name. This does not unnecessarily close the door that in some
future we want to test on something other than md(4) devices.
Also add a "conf" action so that we can check whether a gctl actually
did the right thing or not. It's one thing to check that the result
strings are as expected, but it doesn't tell us if the end result is
correct. This needs a bit more fleshing out, but for now a visual
(i.e. manual) check suffices.
Craig Rodrigues [Sat, 12 Nov 2005 20:12:56 +0000 (20:12 +0000)]
- Minor fixes to raise WARNS level to 6.
- Teach the mount program to call the nmount() syscall directly
- Preserve existing method of calling mount() for UFS, until we clean things
up.
- Preserve existing method of forking and calling external mount programs for
mfs, msdosfs, nfs, nfs4, ntfs, nwfs, nullfs, portalfs, reiserfs, smbfs,
udf, umapfs, unionfs
- devfs, linprocfs, procfs, ext2fs call nmount() syscall directly, since
that is all those external mount programs were doing
Marcel Moolenaar [Sat, 12 Nov 2005 20:02:02 +0000 (20:02 +0000)]
Make the kern.geom.conftxt sysctl more usable by also dumping the
MD class. Previously only the DISK class was dumped. The only
consumer of this sysctl is libdisk (i.e. sysinstall) and it tests
explicitly for instances of the DISK class. Dumping other classes
is therefore harmless.
By also dumping the MD class regression tests can be written that
use the MD class for operations that would normally be done on the
DISK class. The sysctl can now be used to test if those operations
took an effect. An example is partitioning.
Bruce Evans [Sat, 12 Nov 2005 18:20:09 +0000 (18:20 +0000)]
As for the float trig functions, use a minimax polynomial that is
specialized for float precision. The new polynomial has degree 8
instead of 14, and a maximum error of 2**-34.34 (absolute) instead of
2**-30.66. This doesn't affect the final error significantly; the
maximum error was and is about 0.8879 ulps on amd64 -01.
The fdlibm expf() is not used on i386's (the "optimized" asm version
is used), but probably should be since it was already significantly
faster than the asm version on athlons. The asm version has the
advantage of being more accurate, so keep using it for now.
Ian Dowse [Sat, 12 Nov 2005 17:39:31 +0000 (17:39 +0000)]
Fix a > 1 year old typo that caused the ulpt driver to try reading
from the printer and discarding the data even if the ulpt device
was opened for reading. This resulted in crashes because two
conconcurrent read transfers were using the same transfer structure.
PR: usb/88886
Reported By: Alex Pivovarov
MFC after: 1 week
Kirill Ponomarev [Sat, 12 Nov 2005 11:45:01 +0000 (11:45 +0000)]
Add -P flag, it does the same as the -p option, except that the
given prefix is also used recursively for the dependency packages,
if any. If the -P flag appears after any -p flag on the
command line, it overrides it's effect, causing pkg_add to use the
given prefix recursively.
PR: bin/75742
Submitted by: Frerich Raabe <raabe AT kde DOT org>
MFC after: 3 days
Robert Watson [Sat, 12 Nov 2005 10:45:13 +0000 (10:45 +0000)]
Significant refactoring of the accounting code to improve locking and VFS
happiness, as well as correct other bugs:
- Replace notion of current and saved accounting credential/vnode with a
single credential/vnode and an acct_suspended flag. This simplifies the
accounting logic substantially.
- Replace acct_mtx with acct_sx, a sleepable lock held exclusively during
reconfiguration and space polling, but shared during log entry
generation. This avoids holding a mutex over sleepable VFS operations.
- Hold the sx lock over the duration of the I/O so that the vnode I/O
cannot occur after vnode close, which could occur previously if
accounting was disabled as a process exited.
- Write the accounting log entry with Giant conditionally acquired based
on the file system where the log is stored. Previously, the accounting
code relied on the caller acquiring Giant.
- Acquire Giant conditionally in the accounting callout based on the file
system where the accounting log is stored. Run the callout MPSAFE.
- Expose acct_suspended via a read-only sysctl so it is possibly to
programmatically determine whether accounting is suspended or not without
attempting to parse logs.
- Check both acct_vp and acct_suspended lock-free before entering the
accounting sx lock in acct().
- When accounting is disabled due to a VBAD vnode (i.e., forceable unmount),
generate a log message indicating accounting has been disabled.
- Correct a long-standing bug in how free space is calculated and compared
to the required space: generate and compare signed results, not unsigned
results, or negative free space will cause accounting to not be suspended
when required, or worse, incorrectly resumed once negative free space is
reached.
John Baldwin [Fri, 11 Nov 2005 19:57:41 +0000 (19:57 +0000)]
Fix a bug in dlinfo(RTLD_DI_SERINFOSIZE) requests. For each search path
we included the length of the path in the returned size but not the length
of the associated Dl_serpath structure. Without this fix, programs
attempting to allocate a structure to hold the search path information
would allocate too small of a buffer and rtld would overrun the buffer
while filling it via a subsequent RTLD_DI_SERINFO request.
Submitted by: "William K. Josephson" wkj at morphisms dot net
Reviewed by: jdp
MFC after: 2 weeks
Ruslan Ermilov [Fri, 11 Nov 2005 16:04:59 +0000 (16:04 +0000)]
- Store pointer to the link-level address right in "struct ifnet"
rather than in ifindex_table[]; all (except one) accesses are
through ifp anyway. IF_LLADDR() works faster, and all (except
one) ifaddr_byindex() users were converted to use ifp->if_addr.
- Stop storing a (pointer to) Ethernet address in "struct arpcom",
and drop the IFP2ENADDR() macro; all users have been converted
to use IF_LLADDR() instead.