]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
23 years ago-pthread -> -lc_r
deischen [Wed, 24 Jan 2001 13:41:04 +0000 (13:41 +0000)]
-pthread -> -lc_r

23 years agoOops, when converting if (foo) panic() to a KASSERT(), you have to invert
jhb [Wed, 24 Jan 2001 13:10:17 +0000 (13:10 +0000)]
Oops, when converting if (foo) panic() to a KASSERT(), you have to invert
the test case.

Spotted by: peter, jasone

23 years agoBump FreeBSD version to reflect changes to libc and libc_r.
deischen [Wed, 24 Jan 2001 13:08:12 +0000 (13:08 +0000)]
Bump FreeBSD version to reflect changes to libc and libc_r.

23 years ago-pthread -> -lc_r
deischen [Wed, 24 Jan 2001 13:04:10 +0000 (13:04 +0000)]
-pthread -> -lc_r

23 years agoAdd weak definitions for wrapped system calls. In general:
deischen [Wed, 24 Jan 2001 13:03:38 +0000 (13:03 +0000)]
Add weak definitions for wrapped system calls.  In general:

_foo - wrapped system call
foo - weak definition to _foo

and for cancellation points:

_foo - wrapped system call
__foo - enter cancellation point, call _foo(), leave
        cancellation point
foo - weak definition to __foo

Change use of global _thread_run to call a function to get the
currently running thread.

Make all pthread_foo functions weak definitions to _pthread_foo,
where _pthread_foo is the implementation.  This allows an application
to provide its own pthread functions.

Provide slightly different versions of pthread_mutex_lock and
pthread_mutex_init so that we can tell the difference between
a libc mutex and an application mutex.  Threads holding mutexes
internal to libc should never be allowed to exit, call signal
handlers, or cancel.

Approved by: -arch

23 years agoAdd a lock to DIR to make telldir and friends MT-safe.
deischen [Wed, 24 Jan 2001 13:01:47 +0000 (13:01 +0000)]
Add a lock to DIR to make telldir and friends MT-safe.

Clean up stdio.h a bit and remove _THREAD_SAFE.  Some of the
usual macros getc, putc, getchar, putchar are no longer macros.

Approved by: -arch

23 years agoRemove _THREAD_SAFE and make libc thread-safe by default by
deischen [Wed, 24 Jan 2001 13:01:12 +0000 (13:01 +0000)]
Remove _THREAD_SAFE and make libc thread-safe by default by
adding (weak definitions to) stubs for some of the pthread
functions.  If the threads library is linked in, the real
pthread functions will pulled in.

Use the following convention for system calls wrapped by the
threads library:
__sys_foo - actual system call
_foo - weak definition to __sys_foo
foo - weak definition to __sys_foo

Change all libc uses of system calls wrapped by the threads
library from foo to _foo.  In order to define the prototypes
for _foo(), we introduce namespace.h and un-namespace.h
(suggested by bde).  All files that need to reference these
system calls, should include namespace.h before any standard
includes, then include un-namespace.h after the standard
includes and before any local includes.  <db.h> is an exception
and shouldn't be included in between namespace.h and
un-namespace.h  namespace.h will define foo to _foo, and
un-namespace.h will undefine foo.

Try to eliminate some of the recursive calls to MT-safe
functions in libc/stdio in preparation for adding a mutex
to FILE.  We have recursive mutexes, but would like to avoid
using them if possible.

Remove uneeded includes of <errno.h> from a few files.

Add $FreeBSD$ to a few files in order to pass commitprep.

Approved by: -arch

23 years ago- Catch up to new proc flags.
jhb [Wed, 24 Jan 2001 12:59:50 +0000 (12:59 +0000)]
- Catch up to new proc flags.

23 years agoAdd a new item to kinfo_proc: ki_sflag to mirror p_sflag.
jhb [Wed, 24 Jan 2001 12:49:52 +0000 (12:49 +0000)]
Add a new item to kinfo_proc: ki_sflag to mirror p_sflag.

23 years agoConvert all simplelocks to mutexes and remove the simplelock implementations.
jasone [Wed, 24 Jan 2001 12:35:55 +0000 (12:35 +0000)]
Convert all simplelocks to mutexes and remove the simplelock implementations.

23 years agoArgh, I didn't get this test right when I converted it. Break this up
jhb [Wed, 24 Jan 2001 12:23:17 +0000 (12:23 +0000)]
Argh, I didn't get this test right when I converted it.  Break this up
into two separate if's instead of nested if's.  Also, reorder things
slightly to avoid unnecessary mutex operations.

23 years agomdoc(7) police: Remove the debug line that prevents hyphenation
ru [Wed, 24 Jan 2001 12:22:39 +0000 (12:22 +0000)]
mdoc(7) police: Remove the debug line that prevents hyphenation
for the purposes of spelling checking that left in after my last
visit to this file.

23 years ago- Catch up to proc flag changes.
jhb [Wed, 24 Jan 2001 11:28:36 +0000 (11:28 +0000)]
- Catch up to proc flag changes.
- Minimal proc locking.
- Use queue macros.

23 years agoAdd mtx_assert()'s to verify that kmem_alloc() and kmem_free() are called
jhb [Wed, 24 Jan 2001 11:27:29 +0000 (11:27 +0000)]
Add mtx_assert()'s to verify that kmem_alloc() and kmem_free() are called
with Giant held.

23 years ago- Catch up to proc flag changes.
jhb [Wed, 24 Jan 2001 11:25:56 +0000 (11:25 +0000)]
- Catch up to proc flag changes.
- Proc locking in a few places.
- faultin() now must be called with the proc lock held.
- Split up swappable() into a couple of tests so that it can be locke in
  swapout_procs().
- Use queue macros.

23 years ago- Catch up to proc flag changes.
jhb [Wed, 24 Jan 2001 11:20:05 +0000 (11:20 +0000)]
- Catch up to proc flag changes.

23 years ago- Catch up to proc flag changes.
jhb [Wed, 24 Jan 2001 11:15:59 +0000 (11:15 +0000)]
- Catch up to proc flag changes.
- Assert sched_lock is held in proc_compare.

23 years ago- Catch up to proc flag changes.
jhb [Wed, 24 Jan 2001 11:15:24 +0000 (11:15 +0000)]
- Catch up to proc flag changes.
- Update stopevent() to assert that the proc lock is held when it is
  held and is not recursed.  Note that the STOPEVENT() macro obtains
  the proc lock when calling this function.

23 years ago- Catch up to proc flag changes.
jhb [Wed, 24 Jan 2001 11:12:37 +0000 (11:12 +0000)]
- Catch up to proc flag changes.
- Add proc locking for selwakeup() and selrecord().

23 years ago- Catch up to proc flag changes.
jhb [Wed, 24 Jan 2001 11:11:35 +0000 (11:11 +0000)]
- Catch up to proc flag changes.

23 years ago- Catch up to proc flag changes.
jhb [Wed, 24 Jan 2001 11:10:55 +0000 (11:10 +0000)]
- Catch up to proc flag changes.
- Add in some locking ops that might fix SIGXCPU, but don't enable them
  yet.
- Assert that sched_lock is not recursed when mi_switch() is called.

23 years ago- Proc locking.
jhb [Wed, 24 Jan 2001 11:08:02 +0000 (11:08 +0000)]
- Proc locking.
- Catch up to proc flag changes.

23 years ago- Add a mtx_assert() for sched_lock in calcru().
jhb [Wed, 24 Jan 2001 11:06:39 +0000 (11:06 +0000)]
- Add a mtx_assert() for sched_lock in calcru().
- Protect calcru() with sched_lock later on in the file when it is called.

23 years ago- Proc locking.
jhb [Wed, 24 Jan 2001 11:05:50 +0000 (11:05 +0000)]
- Proc locking.
- Catch up to proc flag changes.
- Reorder the way we get things in fill_kinfoproc() to minimize the
  number of locking operations.

23 years ago- Don't use a union and fun tricks to shave one extra pointer off of struct
jhb [Wed, 24 Jan 2001 10:57:01 +0000 (10:57 +0000)]
- Don't use a union and fun tricks to shave one extra pointer off of struct
  mtx right now as it makes debugging harder.  When we are in optimizing
  mode, we can revisit this.
- Fix the KTR trace messages to use %p rather than 0x%p to avoid duplicate
  0x's in KTR output.
- During witness_fixup, release Giant so that witness doesn't get confused.
  Also, grab all_mtx while walking the list of mutexes.
- Remove w_sleep and w_recurse.  Instead, perform checks on mutexes using
  the mutex's mtx_flags field.
- Allow debug.witness_ddb and debug.witness_skipspin to be set from the
  loader.
- Add Giant to the front of existing order_list entries to help ensure
  Giant is always first.
- Add an order entry for the various proc locks.  Note that this only
  helps keep proc in order mostly as the allproc and proctree mutexes are
  only obtained during a lockmgr operation on the specified mutex.

23 years ago- Catch up to proc flag changes.
jhb [Wed, 24 Jan 2001 10:47:50 +0000 (10:47 +0000)]
- Catch up to proc flag changes.
- Set the new P_KTHREAD flag for kthreads during kthread_create.

23 years ago- Catch up to proc flag changes.
jhb [Wed, 24 Jan 2001 10:47:14 +0000 (10:47 +0000)]
- Catch up to proc flag changes.
- Add new fork_exit() and fork_return() MI C functions.

23 years agoCatch up to P_FOO -> PS_FOO changes in proc flags.
jhb [Wed, 24 Jan 2001 10:44:01 +0000 (10:44 +0000)]
Catch up to P_FOO -> PS_FOO changes in proc flags.

23 years ago- Proc locking.
jhb [Wed, 24 Jan 2001 10:43:25 +0000 (10:43 +0000)]
- Proc locking.
- P_FOO -> PS_FOO.

23 years ago- Catch up to p_sflag changes.
jhb [Wed, 24 Jan 2001 10:40:56 +0000 (10:40 +0000)]
- Catch up to p_sflag changes.
- The MD code now initializes proc0.p_heldmtx, proc0.p_contested, and
  curproc.
- The MD code calls here with Giant already held.
- Proc locking.

23 years ago- Proc locking.
jhb [Wed, 24 Jan 2001 10:38:58 +0000 (10:38 +0000)]
- Proc locking.
- P_OWEUPC -> PS_OWEUPC.

23 years ago- Proc locking.
jhb [Wed, 24 Jan 2001 10:38:13 +0000 (10:38 +0000)]
- Proc locking.
- Update userret() to take a struct trapframe * as a second argument.
- Axe have_giant and use mtx_owned(&Giant) where appropriate.

23 years ago- Proc locking.
jhb [Wed, 24 Jan 2001 10:36:47 +0000 (10:36 +0000)]
- Proc locking.
- P_FOO -> PS_FOO.

23 years ago- Proc locking.
jhb [Wed, 24 Jan 2001 10:36:21 +0000 (10:36 +0000)]
- Proc locking.
- Bring across forwarded_statclock() fixes from i386 and alpha.

23 years ago- Proc locking around the vinumdaemon dinking with its flags.
jhb [Wed, 24 Jan 2001 10:28:19 +0000 (10:28 +0000)]
- Proc locking around the vinumdaemon dinking with its flags.
- P_INMEM -> PS_INMEM.

23 years agoMake mp_machdep.c be an optional file conditional on SMP.
jhb [Wed, 24 Jan 2001 10:27:37 +0000 (10:27 +0000)]
Make mp_machdep.c be an optional file conditional on SMP.

23 years agoProc locking.
jhb [Wed, 24 Jan 2001 10:27:11 +0000 (10:27 +0000)]
Proc locking.

23 years agoUse queue macros.
jhb [Wed, 24 Jan 2001 10:26:51 +0000 (10:26 +0000)]
Use queue macros.

23 years agoAdd a prototype for isa_irq_mask() to quiet a warning.
jhb [Wed, 24 Jan 2001 10:25:44 +0000 (10:25 +0000)]
Add a prototype for isa_irq_mask() to quiet a warning.

23 years agoWrap the IPI definitions and function prototypes in #ifdef SMP.
jhb [Wed, 24 Jan 2001 10:25:09 +0000 (10:25 +0000)]
Wrap the IPI definitions and function prototypes in #ifdef SMP.

23 years ago- Rename the gd_cpuno member of struct globaldata to gd_cpuid.
jhb [Wed, 24 Jan 2001 10:24:49 +0000 (10:24 +0000)]
- Rename the gd_cpuno member of struct globaldata to gd_cpuid.
- Add a globaldata_register() prototype in the SMP case.

23 years ago- Proc locking.
jhb [Wed, 24 Jan 2001 10:23:54 +0000 (10:23 +0000)]
- Proc locking.
- P_OWEUPC -> PS_OWEUPC.

23 years ago- Change userret() to take a struct trapframe * as its second argument and
jhb [Wed, 24 Jan 2001 10:23:21 +0000 (10:23 +0000)]
- Change userret() to take a struct trapframe * as its second argument and
  to extract the PC from that to send to addupc_task() so that it can be
  called from MI code.
- Remove all traces of have_giant with extreme prejudice and use
  mtx_owned(&Giant) instead where appropriate.
- Proc locking.
- P_FOO -> PS_FOO.
- Don't grab Giant just to look in curproc's p_addr during a trap since we
  may choose to immediately exit.  Instead, delay grabbing Giant a bit
  until we actually need it.
- Don't reset 'p' to 'curproc' in syscall() to handle the case of a child
  returning from fork1() since children don't return via syscall().
- Remove an XXX comment in ast() that questions the correctness of the
  userland check.  The code is correct.

23 years ago- Proc locking.
jhb [Wed, 24 Jan 2001 10:16:23 +0000 (10:16 +0000)]
- Proc locking.
- P_INMEM -> PS_INMEM.

23 years ago- Proc locking.
jhb [Wed, 24 Jan 2001 10:16:01 +0000 (10:16 +0000)]
- Proc locking.
- Don't send IPIs for pmap_invalidate_page() or pmap_invalidate_all()
  in the UP case.
- Catch up to cpuno -> cpuid.
- Convert some sanity checks that were #ifdef DIAGNOSTIC to KASSERT()'s.

23 years ago- Adjust some whitespace to reduce diffs with the i386 version.
jhb [Wed, 24 Jan 2001 10:13:13 +0000 (10:13 +0000)]
- Adjust some whitespace to reduce diffs with the i386 version.
- Rename the per-CPU variable 'cpuno' to 'cpuid'.  This was done so that
  there is one consistent name across all architectures for a logical
  CPU id.
- Remove all traces of IRQ forwarding.
- Add globaldata_register() hook called by globaldata_init() to register
  globaldata structures in the cpuid_to_globaldata array.
- Catch up to P_FOO -> PS_FOO.
- Bring across some fixes for forwarded_statclock() from the i386 version
  to handle ithreads and idleproc properly.
- Rename addugd_intr_forwarded() to addupc_intr_forwarded() so that it is
  the same name on all architectures.
- Set flags in p_sflag instead of calling psignal() from
  forward_hardclock().
- Proc locking.
- When we handle an IPI, turn off its bit in the mask of IPI's we are
  currently handling so that an IPI doesn't send a CPU into an infinite
  loop.

23 years ago- Initialize curproc, proc0.p_heldmtx, and proc0.p_contested earlier so
jhb [Wed, 24 Jan 2001 10:07:42 +0000 (10:07 +0000)]
- Initialize curproc, proc0.p_heldmtx, and proc0.p_contested earlier so
  that mutex operations work.
- Enter Giant earlier so we hold it during boot.
- Proc locking.
- Move globaldata_init() into here from mp_machdep.c so that UP kernels
  don't depend on mp_machdep.c.  Use a callout in the SMP case to register
  the boot processor's globaldata in the cpuid_to_globaldata array.

23 years ago- Wrap the IPI interrupt handler in #ifdef SMP.
jhb [Wed, 24 Jan 2001 10:05:24 +0000 (10:05 +0000)]
- Wrap the IPI interrupt handler in #ifdef SMP.
- Catch up to cpuno -> cpuid change.
- Add parens around a subexpression to quiet a warning.

23 years agocpuno -> cpuid.
jhb [Wed, 24 Jan 2001 10:04:32 +0000 (10:04 +0000)]
cpuno -> cpuid.

23 years agoDon't import the nonexistent astpending variable.
jhb [Wed, 24 Jan 2001 10:03:05 +0000 (10:03 +0000)]
Don't import the nonexistent astpending variable.

23 years agoWrap the startup code used by secondary processors in #ifdef SMP.
jhb [Wed, 24 Jan 2001 10:01:53 +0000 (10:01 +0000)]
Wrap the startup code used by secondary processors in #ifdef SMP.

23 years agoRemove the Xforward_irq IPI.
jhb [Wed, 24 Jan 2001 10:01:13 +0000 (10:01 +0000)]
Remove the Xforward_irq IPI.

23 years ago- Remove all the #if 0'd code that used to implement IRQ forwarding.
jhb [Wed, 24 Jan 2001 10:00:07 +0000 (10:00 +0000)]
- Remove all the #if 0'd code that used to implement IRQ forwarding.
- Remove #if 0'd lazy interrupt mask.

23 years agoRemove unused locks: cpl, fast_intr, intr, mpintr.
jhb [Wed, 24 Jan 2001 09:58:56 +0000 (09:58 +0000)]
Remove unused locks: cpl, fast_intr, intr, mpintr.

23 years ago- Proc locking.
jhb [Wed, 24 Jan 2001 09:56:49 +0000 (09:56 +0000)]
- Proc locking.
- P_OWEUPC -> PS_OWEUPC.
- Remove obsolete prototype for MD fork_return().

23 years ago- Remove Xforward_irq, cpl_lock, and fast_intr_lock.
jhb [Wed, 24 Jan 2001 09:55:39 +0000 (09:55 +0000)]
- Remove Xforward_irq, cpl_lock, and fast_intr_lock.
- Add fork_exit.

23 years agoSetup the return values for a child process in the trapframe when we setup
jhb [Wed, 24 Jan 2001 09:54:44 +0000 (09:54 +0000)]
Setup the return values for a child process in the trapframe when we setup
the rest of the trapframe instead of doing it in fork_return().

23 years ago- Kill the have_giant parameter to userret() along with all instances of
jhb [Wed, 24 Jan 2001 09:53:49 +0000 (09:53 +0000)]
- Kill the have_giant parameter to userret() along with all instances of
  that name as a variable.  Use mtx_owned(&Giant) where appropriate
  instead.
- Proc locking.
- P_FOO -> PS_FOO.
- Update comments about enable interrupts during trap and why this may be
  bad if we trap while holding a spin mutex.
- Don't bother resetting p to curproc in syscall() in case we are the child
  returning from fork.  The child hasn't returned from fork through syscall
  in a while.
- Remove fork_return() as it has been superseded by the MI version.

23 years ago- Proc locking.
jhb [Wed, 24 Jan 2001 09:49:49 +0000 (09:49 +0000)]
- Proc locking.
- P_INMEM -> PS_INMEM.

23 years ago- Relocate portions of this file to get it into an order closer to that of
jhb [Wed, 24 Jan 2001 09:48:52 +0000 (09:48 +0000)]
- Relocate portions of this file to get it into an order closer to that of
  the alpha mp_machdep.c.
- Proc locking.
- Catch up to the P_FOO -> PS_FOO proc flags changes.
- Stick ap_init()'s prototype with the other prototypes.
- Remove the Xforwardirq IPI.
- Remove unused simplelocks.
- Don't try to psignal() from forward_statclock(), but set the appropriate
  signal pending flag in p_sflag instead.
- Add in KTR_SMP tracepoints for various SMP functions.   (Brought over
  from the alpha port)

23 years ago- Proc locking.
jhb [Wed, 24 Jan 2001 09:45:20 +0000 (09:45 +0000)]
- Proc locking.
- Setup proc0.p_heldmtx, proc0.contested, and curproc earlier so that we
  can use mutexes.
- Initialize sched_lock and Giant earlier and enter Giant during init386.
- Use suser(9) instead of checking cr_uid directly.

23 years agoMdoc(7)ify the previous revision.
ru [Wed, 24 Jan 2001 09:45:19 +0000 (09:45 +0000)]
Mdoc(7)ify the previous revision.

23 years agoCall fork_exit() now instead of futzing around in assembly during a fork
jhb [Wed, 24 Jan 2001 09:43:26 +0000 (09:43 +0000)]
Call fork_exit() now instead of futzing around in assembly during a fork
return.

23 years agomdoc(7) police: fixed some minor formatting/spelling errors.
ru [Wed, 24 Jan 2001 09:41:17 +0000 (09:41 +0000)]
mdoc(7) police: fixed some minor formatting/spelling errors.

23 years ago- Split p_flag up into two fields. p_flag keeps most of the previous flags
jhb [Wed, 24 Jan 2001 09:41:03 +0000 (09:41 +0000)]
- Split p_flag up into two fields.  p_flag keeps most of the previous flags
  and is protected by the proc lock.  p_sflag is protected by sched_lock
  and holds the following flags: PS_INMEM, PS_OWEUPC, PS_PROFIL, PS_SINTR,
  PS_TIMEOUT, PS_ALRMPEND, PS_PROFPEND, PS_CVWAITQ, PS_SWAPINREQ, and
  PS_SWAPPING.
- p_klist is definitely locked now by the proc lock.
- p_runtime, p_[usi]u are locked by sched_lock.
- Add a new P_KTHREAD flag set for kernel threads created via
  kthread_create(9).
- STOPEVENT() only needs the proc lock, it does not need Giant.
- faultin() already checks PS_INMEM, so simplify the check in PHOLD() so
  that we only need to grab the proc lock and let faultin() perform the
  PS_INMEM check.
- Add a prototype for zpfind().
- Add prototypes for the new fork_exit() and fork_return() MI functions
  that manage the fork return path.
- Add a prototype for the MD function userret() so that it can be called
  from fork_return().
- Add needed include of <machine/frame.h> in the kernel.

23 years agoProtect against multiple inclusion.
sobomax [Wed, 24 Jan 2001 09:06:42 +0000 (09:06 +0000)]
Protect against multiple inclusion.

23 years agoMinor fixes:
sobomax [Wed, 24 Jan 2001 09:05:35 +0000 (09:05 +0000)]
Minor fixes:
- Use "OPOST | ONLCR" as a c_oflag in the raw mode instead of 0, which should
  make debug output mode readable;
- flush input buffer when switching to/from raw mode. This should prevent
  leaking raw scancodes into your shell prompt when program called
  VGLKeyboardEnd() w/o processing all scancodes in the buffer.

23 years agoFix a typo.
jhb [Wed, 24 Jan 2001 08:42:39 +0000 (08:42 +0000)]
Fix a typo.

Reported by: albert

23 years agoremove redundant definitions of card types:
hm [Wed, 24 Jan 2001 08:41:52 +0000 (08:41 +0000)]
remove redundant definitions of card types:
 - remove all occurances of FLAG_XXX values and replace with CARD_TYPEP_XXX
 - remove sc_flag from isic softc and replace it with sc_cardtyp
remove some left over redundant definitions from isic and i4b_l1.h
remove left over cvs id from i4b cvs repository

23 years agoMention the -F in the usage().
ru [Wed, 24 Jan 2001 08:27:59 +0000 (08:27 +0000)]
Mention the -F in the usage().

PR: 24586
Submitted by: Cyrille Lefevre <clefevre@citeweb.net>

23 years agoMake `-y' option actually work.
sobomax [Wed, 24 Jan 2001 08:22:53 +0000 (08:22 +0000)]
Make `-y' option actually work.

Submitted by: reg

23 years agoAdd missing include.
jhb [Wed, 24 Jan 2001 06:54:24 +0000 (06:54 +0000)]
Add missing include.

23 years agoNever reuse AUTO_OID values.
mckusick [Wed, 24 Jan 2001 04:35:13 +0000 (04:35 +0000)]
Never reuse AUTO_OID values.

Approved by: Alfred Perlstein <bright@wintelcom.net>

23 years agoThe lock being destroyed was misnamed, not unused. Add the lockdestroy()
jhb [Wed, 24 Jan 2001 02:18:54 +0000 (02:18 +0000)]
The lock being destroyed was misnamed, not unused.  Add the lockdestroy()
back in but with the proper name so that this compiles.

Submitted by: jasone

23 years ago#ifdef a local variable only used in debugging code.
jhb [Wed, 24 Jan 2001 01:46:57 +0000 (01:46 +0000)]
#ifdef a local variable only used in debugging code.

Reviewed by: -scsi

23 years agoConst'ify cam_sim.sim_name to quiet warnings.
jhb [Wed, 24 Jan 2001 01:46:18 +0000 (01:46 +0000)]
Const'ify cam_sim.sim_name to quiet warnings.

Reviewed by: -scsi

23 years agodon't grope around inside snd_dbuf structures as they will change in future
cg [Wed, 24 Jan 2001 01:36:15 +0000 (01:36 +0000)]
don't grope around inside snd_dbuf structures as they will change in future

23 years agofix certain cards failing to attach
cg [Wed, 24 Jan 2001 01:27:31 +0000 (01:27 +0000)]
fix certain cards failing to attach

Submitted by: Russell Cattelan <cattelan@thebarn.com>

23 years agofix blocksizing for playback, implement blocksizing for record
cg [Wed, 24 Jan 2001 01:22:53 +0000 (01:22 +0000)]
fix blocksizing for playback, implement blocksizing for record

23 years agosome cosmetics, changed channel setup and revamped irq handling - ignore
cg [Wed, 24 Jan 2001 01:20:04 +0000 (01:20 +0000)]
some cosmetics, changed channel setup and revamped irq handling - ignore
repeated interrupts

23 years agoprevent excessively small buffer sizes resulting in hundreds of irqs per
cg [Wed, 24 Jan 2001 01:15:50 +0000 (01:15 +0000)]
prevent excessively small buffer sizes resulting in hundreds of irqs per
second

23 years agoDon't grab Giant when calling kmem_alloc/kmem_free as this is just
jhb [Wed, 24 Jan 2001 00:36:03 +0000 (00:36 +0000)]
Don't grab Giant when calling kmem_alloc/kmem_free as this is just
encouraging other people to follow the same practice.  If this is going
to be done, then it should be done inside of those two functions instead.

23 years agoProc locking.
jhb [Wed, 24 Jan 2001 00:35:12 +0000 (00:35 +0000)]
Proc locking.

23 years ago- Proc locking.
jhb [Wed, 24 Jan 2001 00:33:44 +0000 (00:33 +0000)]
- Proc locking.
- Protect calcru() with sched_lock.

23 years ago- Proc locking.
jhb [Wed, 24 Jan 2001 00:28:07 +0000 (00:28 +0000)]
- Proc locking.
- Protect calcru() with sched_lock.

23 years agoProc locking.
jhb [Wed, 24 Jan 2001 00:27:28 +0000 (00:27 +0000)]
Proc locking.

23 years agoProc locking to protect p_ucred while we obtain additional references.
jhb [Wed, 24 Jan 2001 00:26:19 +0000 (00:26 +0000)]
Proc locking to protect p_ucred while we obtain additional references.

23 years ago- Proc locking.
jhb [Wed, 24 Jan 2001 00:24:57 +0000 (00:24 +0000)]
- Proc locking.
- Use FreeBSD stackgap_init() semantics.
- Adjust some #ifdef's to hide unused variables.

23 years ago- Proc locking.
jhb [Wed, 24 Jan 2001 00:23:30 +0000 (00:23 +0000)]
- Proc locking.
- Use NULL instead of 0.

23 years ago"Rejected with status" is better than "Rejected for status".
mjacob [Wed, 24 Jan 2001 00:14:37 +0000 (00:14 +0000)]
"Rejected with status" is better than "Rejected for status".

23 years agoUse selrecord() instead of doing the work ourselves.
jhb [Wed, 24 Jan 2001 00:06:26 +0000 (00:06 +0000)]
Use selrecord() instead of doing the work ourselves.

23 years agoMake the device name in an isa_device a const char * instead of a char *
jhb [Wed, 24 Jan 2001 00:05:13 +0000 (00:05 +0000)]
Make the device name in an isa_device a const char * instead of a char *
to silence cast-qual warnings.

23 years agoUse queue macros.
jhb [Wed, 24 Jan 2001 00:01:42 +0000 (00:01 +0000)]
Use queue macros.

23 years agoProc locking.
jhb [Tue, 23 Jan 2001 23:59:38 +0000 (23:59 +0000)]
Proc locking.

23 years ago- Use 'p' instead of 'curproc' for the namei lookup as this is what
jhb [Tue, 23 Jan 2001 23:59:30 +0000 (23:59 +0000)]
- Use 'p' instead of 'curproc' for the namei lookup as this is what
  other image activators use.
- Proc locking.

23 years agoProc locking.
jhb [Tue, 23 Jan 2001 23:51:56 +0000 (23:51 +0000)]
Proc locking.

23 years agoHmm. This must be a first. The man page was updated *before* the source.
mjacob [Tue, 23 Jan 2001 23:24:48 +0000 (23:24 +0000)]
Hmm. This must be a first. The man page was updated *before* the source.
Correct some slight inaccuracies in the man page.

23 years agoAllow fxp to configure in I/O space if the user wants it and specifies
mjacob [Tue, 23 Jan 2001 23:22:17 +0000 (23:22 +0000)]
Allow fxp to configure in I/O space if the user wants it and specifies
an override as a loader settable variable (fxp_iomap). fxp_iomap is
a bitmap of fxp units that should be configured to use PCI I/O space
in stead of PCI Memory space.

Reviewed by: Kees Jan Koster <dutchman@tccn.cs.kun.nl>, dg@freebsd.org

23 years agoProc locking, mostly protecting p_ucred while obtaining additional
jhb [Tue, 23 Jan 2001 22:41:15 +0000 (22:41 +0000)]
Proc locking, mostly protecting p_ucred while obtaining additional
references.

23 years ago- Remove unused header include.
jhb [Tue, 23 Jan 2001 22:38:38 +0000 (22:38 +0000)]
- Remove unused header include.
- Use queue macros.