rwatson [Sun, 30 Oct 2005 19:44:40 +0000 (19:44 +0000)]
Push the assignment of a new or updated so_qlimit from solisten()
following the protocol pru_listen() call to solisten_proto(), so
that it occurs under the socket lock acquisition that also sets
SO_ACCEPTCONN. This requires passing the new backlog parameter
to the protocol, which also allows the protocol to be aware of
changes in queue limit should it wish to do something about the
new queue limit. This continues a move towards the socket layer
acting as a library for the protocol.
Bump __FreeBSD_version due to a change in the in-kernel protocol
interface. This change has been tested with IPv4 and UNIX domain
sockets, but not other protocols.
marcel [Sun, 30 Oct 2005 17:23:01 +0000 (17:23 +0000)]
Remove a stray return statement in the interrupt dispatch function
that caused a premature exit after calling a fast interrupt handler
and bypassing a much needed critical_exit() and the scheduling of
the interrupt thread for non-fast handlers. In short: unbreak :-)
brooks [Sun, 30 Oct 2005 16:31:05 +0000 (16:31 +0000)]
SSIDs are of length 32-bytes and not NUL terminated. Use the correct
length and bail immediatly if the returned length is wrong rather than
attempting to set an correct value.
This commit differs from the patch in the PR in the use of exit instead
of return and the use of a defined value for the array.
bde [Sun, 30 Oct 2005 12:21:02 +0000 (12:21 +0000)]
Fixed some comments added in rev.1.5.
The log message for 1.5 said that some small (one or two ulp) inaccuracies
were fixed, and a comment implied that the critical change is to switch
the rounding mode to to-nearest, with a switch of the precision to
extended at no extra cost. Actually, the errors are very large (ucbtest
finds ones of several hundred ulps), and it is the switch of the
precision that is critical.
Another comment was wrong about NaNs being handled sloppily.
des [Sun, 30 Oct 2005 10:03:11 +0000 (10:03 +0000)]
Add some safeguards to AIOSFMT:
- Return EINVAL if play_format or rec_format is set but the corresponding
sample rate is 0.
- Don't try to set the playback or recording format to 0. Previously,
issuing an AIOSFMT ioctl with an all-zeroes snd_chan_param would
trigger a KASSERT in chn_fmtchain(); I'm unsure about the effects on
a kernel without INVARIANTS. After this commit, issuing AIOSFMT with
an all-zeroes snd_chan_param is equivalent to issuing AIOGFMT.
grehan [Sun, 30 Oct 2005 07:56:10 +0000 (07:56 +0000)]
In stack_save, stop when a trap-frame is encountered. This prevents
trying to access user-space stack addresses when a user fault
is encountered, as occurs when GEOM KTR code is handling a page fault
and is using stack_save() to capture a trace for debug purposes.
It may be possible to walk beyond the trap-frame if it is a kernel fault,
as db_backtrace() does, but I don't think that complexity is needed in
this routine.
davidxu [Sun, 30 Oct 2005 02:56:08 +0000 (02:56 +0000)]
Let itimer store itimerspec instead of itimerval, so I don't have to
convert to or from timeval frequently.
Introduce function itimer_accept() to ack a timer signal in signal
acceptance code, this allows us to return more fresh overrun counter
than at signal generating time. while POSIX says:
"the value returned by timer_getoverrun() shall apply to the most
recent expiration signal delivery or acceptance for the timer,.."
I prefer returning it at acceptance time.
Introduce SIGEV_THREAD_ID notification mode, it is used by thread
libary to request kernel to deliver signal to a specified thread,
and in turn, the thread library may use the mechanism to implement
SIGEV_THREAD which is required by POSIX.
Timer signal is managed by timer code, so it can not fail even if
signal queue is full filled by sigqueue syscall.
davidxu [Sun, 30 Oct 2005 02:12:49 +0000 (02:12 +0000)]
Fix sigevent's POSIX incompatible problem by adding member fields
sigev_notify_function and sigev_notify_attributes. AIO syscalls
use sigevent, so they have to be adjusted.
stefanf [Sat, 29 Oct 2005 18:41:35 +0000 (18:41 +0000)]
Include disabled options in the output of 'set +o'. POSIX says the output of
set +o can be used to reload previous settings, for this to work disabled
options must be printed as well or otherwise options that were set in the mean
time won't be turned off.
To avoid an excessively long output line I formatted the output to print only
six options per line.
bde [Sat, 29 Oct 2005 16:34:50 +0000 (16:34 +0000)]
Use double precision to simplify and optimize arg reduction for small
and medium size args too: instead of conditionally subtracting a float
17+24, 17+17+24 or 17+17+17+24 bit approximation to pi/2, always
subtract a double 33+53 bit one. The float version is now closer to
the double version than to old versions of itself -- it uses the same
33+53 bit approximation as the simplest cases in the double version,
and where the float version had to switch to the slow general case at
|x| == 2^7*pi/2, it now switches at |x| == 2^19*pi/2 the same as the
double version.
This speeds up arg reduction by a factor of 2 for |x| between 3*pi/4 and
2^7*pi/4, and by a factor of 7 for |x| between 2^7*pi/4 and 2^19*pi/4.
bde [Sat, 29 Oct 2005 08:15:29 +0000 (08:15 +0000)]
Start trying to make the float precision trig functions actually worth
using under FreeBSD. Before this commit, all float precision functions
except exp2f() were implemented using only float precision, apparently
because Cygnus needed this in 1993 for embedded systems with slow or
inefficient double precision. For FreeBSD, except possibly on systems
that do floating point entirely in software (very old i386 and now
arm), this just gives a more complicated implementation, many bugs,
and usually worse performance for float precision than for double
precision. The bugs and worse performance were particulary large in
arg reduction for trig functions. We want to divide by an approximation
to pi/2 which has as many as 1584 bits, so we should use the widest
type that is efficient and/or easy to use, i.e., double. Use fdlibm's
__kernel_rem_pio2() to do this as Sun apparently intended. Cygnus's
k_rem_pio2f.c is now unused. e_rem_pio2f.c still needs to be separate
from e_rem_pio2.c so that it can be optimized for float args. Similarly
for long double precision.
This speeds up cosf(x) on large args by a factor of about 2. Correct
arg reduction on large args is still inherently very slow, so hopefully
these args rarely occur in practice. There is much more efficiency
to be gained by using double precision to speed up arg reduction on
medium and small float args.
imp [Sat, 29 Oct 2005 05:52:17 +0000 (05:52 +0000)]
MFp4: When doing lazy allocation, it turns out that we need to record the
actual resource values we received from the system rather than the range
we requested. Since we request a range starting at 0, we would record
that number. Later, since this == 0, we'd allocate again. However,
we wouldn't write the new resource into the BAR. This resulted in
a resource leak as well as a BAR that couldn't access the resource at
all since rman_get_start, et al, were wrong.
MFC After: 1 week (assuming RELENG_6 is open for business)
imp [Sat, 29 Oct 2005 03:27:43 +0000 (03:27 +0000)]
Shut down the card bus bridge hardware on detach. Before we'd just
free the resoruces w/o actually turning off the interrupts. This lead
to interrupt storms if you were to insert a card after kldunloading
the driver.
wpaul [Sat, 29 Oct 2005 03:01:16 +0000 (03:01 +0000)]
Remove call to txp_set_filter() from txp_attach(). txp_set_filter() needs
the ifp, so you can't call it before doing if_alloc(). Also, there's
really no need to call it here anyway: the code I originally ported from
OpenBSD incorrectly set the station address only once at device attach
time, instead of setting in txp_init(). This meant you couldn't change
the address with ifconfig txp0 ether xx:xx:xx:xx:xx:xx. I added the
call to txp_set_filter() in txp_init() to correct this, but forgot to
remove the call from txp_attach(). Until now, it never mattered.
With this fix, the txp driver tests good:
txp0: <3Com 3cR990-TX-97 Etherlink with 3XP Processor> port 0xb800-0xb87f mem 0xe6800000-0xe683ffff irq 12 at device 10.0 on pci0
txp0: Ethernet address: 00:01:03:d4:91:4f
wpaul [Sat, 29 Oct 2005 02:18:27 +0000 (02:18 +0000)]
Fix ndis_getstate_80211() so that it properly reports the authmode
and channel to ifconfig. Also use the SSID and channel info from
the association info that we already have instead of using ndis_get_info()
to ask the driver for it again.
brooks [Fri, 28 Oct 2005 22:47:40 +0000 (22:47 +0000)]
s/insure/ensure/ in previous commit. My dictionary and m-w.com say they
are synonymous, but ensure seems slightly closer and does not have the
connotation of buying insurance.
Reported by: Jason McIntyre <jmc at kerhand dot co dot uk>
pjd [Fri, 28 Oct 2005 20:25:02 +0000 (20:25 +0000)]
Fix possible live-lock under heavy load where we can't allocate more
memory for request.
I was sure graid3 should handle such situations well, but green@ reported
it is not and we want to fix it before 6.0.
stefanf [Fri, 28 Oct 2005 18:37:09 +0000 (18:37 +0000)]
Add the POSIX options -v and -V to the 'command' builtin. Both describe the
type of their argument, if it is a shell function, an alias, a builtin, etc.
-V is more verbose than -v.
bde [Fri, 28 Oct 2005 13:36:58 +0000 (13:36 +0000)]
Use fairly optimal minimax polynomials for __kernel_cosf() and
__kernel_sinf(). The old ones were the double-precision polynomials
with coefficients truncated to float. Truncation is not a good way
to convert minimax polynomials to lower precision. Optimize for
efficiency and use the lowest-degree polynomials that give a relative
error of less than 1 ulp -- degree 8 instead of 14 for cosf and degree
9 instead of 13 for sinf. For sinf, the degree 8 polynomial happens
to be 6 times more accurate than the old degree 14 one, but this only
gives a tiny amount of extra accuracy in results -- we just need to
use a a degree high enough to give a polynomial whose relative accuracy
in infinite precision (but with float coefficients) is a small fraction
of a float ulp (fdlibm generally uses 1/32 for the small fraction, and
the fraction for our degree 8 polynomial is about 1/600).
The maximum relative errors for cosf() and sinf() are now 0.7719 ulps
and 0.7969 ulps, respectively.
stefanf [Fri, 28 Oct 2005 10:45:19 +0000 (10:45 +0000)]
Protect malloc, realloc and free calls with INT{ON,OFF} directly in chkalloc,
ckrealloc and ckfree (added), respectively. sh jumps out of the signal handler
using longjmp which is obviously a bad idea during malloc calls.
Note: I think there is still a small race here because volatile sig_atomic_t
only guarantees atomic reads and writes while we're doing increments and
decrements.
Protect a setmode call with INT{ON,OFF} as it calls malloc internally.
marcel [Fri, 28 Oct 2005 06:27:53 +0000 (06:27 +0000)]
o Style(9) nits
o Fix typo in comment
o s/-100/BUS_PROBE_GENERIC/
o s/err/error/ for consistency
o Remove non-applicable comment
o Allow uart_bus_probe() to return the predefined BUS_PROBE_*
contants. In this case: explicitly test for error > 0.
wpaul [Fri, 28 Oct 2005 02:17:40 +0000 (02:17 +0000)]
Remove forgotten, no longer needed WB_UNLOCK() from the end wb_ioctl().
With this change, the driver tests good (at least on i386):
wb0: <Winbond W89C840F 10/100BaseTX> port 0xb800-0xb87f mem 0xe6800000-0xe680007f irq 12 at device 10.0 on pci0
miibus1: <MII bus> on wb0
amphy0: <Am79C873 10/100 media interface> on miibus1
amphy0: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, auto
wb0: Ethernet address: 00:00:e8:18:2a:02
wb0: link state changed to DOWN
wb0: link state changed to UP
jhb [Thu, 27 Oct 2005 21:22:58 +0000 (21:22 +0000)]
Fixup locking and mark MPSAFE.
- Add locked variants of init() and start().
- Use callout_*() to manage callout.
- Test IFF_DRV_RUNNING rather than IFF_UP in wb_intr() to see if we are
still active when an interrupt comes in.
I couldn't find any of these cards anywhere to test on myself, and google
turns up references to FreeBSD and OpenBSD manpages for this driver when
trying to locate a card that way. I'm not sure anyone actually uses these
cards with FreeBSD.
jhb [Thu, 27 Oct 2005 21:18:37 +0000 (21:18 +0000)]
- Use if_printf() and device_printf() and remove vr_unit from the softc.
I had to initialize the ifnet a bit earlier in attach so that the
if_printf()'s in vr_reset() didn't explode with a page fault.
- Use M_ZERO with contigmalloc() rather than an explicit bzero.
jhb [Thu, 27 Oct 2005 21:16:17 +0000 (21:16 +0000)]
- Add locking and mark MPSAFE. The driver had a mutex in the softc and
even initialized it, but it never used it.
- Use callout_*() to manage the callout.
- Use m_devget() to copy data out of the rx buffers rather than doing it
all by hand.
- Use m_getcl() to allocate mbuf clusters rather than doing it all by hand.
- Don't free the software descriptor for a rx ring entry if we can't
allocate an mbuf cluster for it. We left a dangling pointer and never
reallocated the entry anyway. OpenBSD's code (from which this was
derived) has the same bug.
Tested by: NO ONE (despite repeated requests)
Reviewed by: wpaul (5)
ru [Thu, 27 Oct 2005 17:39:03 +0000 (17:39 +0000)]
Rename the .dbg extension to .symbols, which matches "symbol-file"
gdb(1) command better, though I must admit it's confusing: these
files have not only [debugging] symbols, but much more than that.
jhb [Thu, 27 Oct 2005 17:34:35 +0000 (17:34 +0000)]
Create a default kernel config for i386 and move 'device isa' and
'device npx' (both of which aren't really optional right now) and
'device io' and 'device mem' (to preserve POLA for 4.x users upgrading
to 6.0) from GENERIC into DEFAULTS.
emaste [Thu, 27 Oct 2005 17:22:47 +0000 (17:22 +0000)]
In watchdog_config enable the software watchdog iff the WD_ACTIVE flag is
set. When watchdogd(1) is terminated intentionally it clears the bit,
which should then disable it in the kernel.
PR: kern/74386
Submitted by: Alex Hoff <ahoff at sandvine dot com>
Approved by: phk, rwatson (mentor)
jhb [Thu, 27 Oct 2005 17:13:23 +0000 (17:13 +0000)]
Optionally include a DEFAULTS config file if it is present in the current
directory before the specified config file. This is implemented by
opening DEFAULTS as stdin if it exists, and if so resetting stdin to the
actual config file when DEFAULTS is fully parsed via yywrap(). In short,
this lets us create DEFAULTS kernel configs in /sys/<arch>/conf that can
enable certain options or devices by default and allow users to disable
them via 'nooptions' or 'nodevice' rather than having to create kludge
NO_FOO options.
jhb [Thu, 27 Oct 2005 16:47:28 +0000 (16:47 +0000)]
Revert most of revision 1.235 and fix the problem a different way. We
can't acquire an sx lock in ttyinfo() because ttyinfo() can be called
from interrupt handlers (such as atkbd_intr()). Instead, go back to
locking the process group while we pick a thread to display information for
and hold that lock until after we drop sched_lock to make sure the
process doesn't exit out from under us. sched_lock ensures that the
specific thread from that process doesn't go away. To protect against
the process exiting after we drop the proc lock but before we dereference
it to lookup the pid and p_comm in the call to ttyprintf(), we now copy
the pid and p_comm to local variables while holding the proc lock.
This problem was found by the recently added TD_NO_SLEEPING assertions for
interrupt handlers.
sos [Thu, 27 Oct 2005 16:32:39 +0000 (16:32 +0000)]
Enclose the delayed attach in Giant so we dont loose the race with other
drivers trying to attach ATA devices like pccard.
Dont clear the delayed flag before we are acutally finished.
ru [Thu, 27 Oct 2005 14:24:45 +0000 (14:24 +0000)]
Installing debug modules was a bad idea -- I bogusly assumed that
our kernel linker will only load PT_LOAD segments, apparently not.
Instead, produce .dbg objects from .debug objects, and install
them together with non-debug objects, as described in objcopy(1).
davidxu [Thu, 27 Oct 2005 03:09:20 +0000 (03:09 +0000)]
Disconnect libc_r from buildworld, it is still kept in the tree to
provide some baseline references, but users are encouraged to use
libpthread or libthr in real world.