summaryrefslogtreecommitdiffstats
Commit message (Collapse)AuthorAgeFilesLines
* Fix a strict aliasing issue. Also remove an unnecessary pam_get_item()des2003-12-111-8/+6
| | | | | | | call (pam_get_authtok() will return the previous token if try_first_pass or use_first_pass is specified). Incidentally fix an ugly bug where the buffer holding the prompt was freed immediately before use, instead of after.
* More strict aliasing fixes.des2003-12-111-26/+29
| | | | Submitted by: Andreas Hauser <andy-freebsd@splashground.de>
* Fix strict aliasing breakage in PAM modules (except pam_krb5, which needsdes2003-12-119-42/+44
| | | | more work than the others). This should make most modules build with -O2.
* Typo fix.jkoshy2003-12-111-1/+1
|
* PowerPC doesn't need NOLIBC_R and NOFORTH (and hasn't for quite a while now)des2003-12-111-1/+0
|
* Document a few kernel functions and the sysctl tunables thatjkoshy2003-12-115-1/+357
| | | | affect their behaviour.
* Install experimental.rc as well.des2003-12-111-1/+1
|
* This patch fixes two little portability (to !GCC compilers) problems:alfred2003-12-111-3/+3
| | | | | | | | 1. Use C99-style variable argument macros rather than GNU ones. 2. Don't cast id to ident_t. Its type is already ident_t and casting to a union type is a constraint violation. Submitted by: Stefan Farfeleder <stefan@fafoe.narf.at>
* Significantly reduce the "jitter" that is typical for PS/2 micealfred2003-12-112-220/+512
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | when using a KVM. There is no actual solution possible, but this gets us pretty close. Typically when switching back to a FreeBSD box and moving the mouse wild data is produced, because the protocol's validation/checksum system is extremely weak it is impossible to determine that we're out of sync before dropping several bogus packets to user land. The actual solution that appears to offer the best clamping of jitter is to buffer the mouse packets if we've not seen mouse activity for more than .5 seconds. Then waiting to flush that data for 1/20th of a second. If within that 20th of a second we get any packets that do fail the weak test we drop the entire queue and back off accepting data from the mouse for 2 seconds and then repeat the whole deal. You can still get _some_ jitter, notably if you switch to the FreeBSD box, then move the mouse just enough to generate one or two packets. Those packets may be bogus, but may still pass the validity check. One way to finally kill the problem once and for all is to check the initial packets for "wild" values. Typically one sees packets in the +/-60 range during normal operation, however when bogus data is generated it's typically near the outer range of +/-120 or more, those packets would be a good candidate for dropping or clamping. I've been running with this for several weeks now and it has significantly helped me stay sane even with a piece of junk Belkin KVM causing wild jitter each and every time I switch. Lastly I'd like to note that my experience with Windows shows me that somehow the Microsoft PS/2 driver typically avoids this problem, but that may only be possible when running the mouse in a dumb-ed down PS/2 mode that Belkin recommends on their site.
* Fixed misplacement of __FBSDID(). Backed out editing of vendor id lines.bde2003-12-111-6/+7
| | | | Just wrap them in #if 0...#endif.
* Fixed English error in previous commit. Fixed some older English errors.bde2003-12-111-6/+5
| | | | Removed a redundant clause.
* Remove uses of GIANT_LOCK and replace with appropriate threadmtm2003-12-111-7/+12
| | | | and thread list locks.
* Revert part of revision 1.74 after bde reminded me of a detail I'ddes2003-12-111-4/+7
| | | | | forgotten about how sysctl works. This removes a potential (though not very likely) race that 1.74 introduced.
* Only define CACHE_LINE_SIZE in one place..peter2003-12-111-2/+0
|
* CACHE_LINE_SIZE is 64 on athlon and amd64 chips, not 32. This shouldpeter2003-12-111-1/+1
| | | | | probably be 128 since that is what the hardware prefetch fill size is on both the p3, p4 and athlon* cpus.
* - Don't let the pctcpu rate limiter throttle us if we have recorded overjeff2003-12-111-1/+2
| | | | | SCHED_CPU_TICKS ticks. This was allowing processes to display (1/SCHED_CPU_TIME * 100) % more cpu than they had used.
* - In sched_switch(), if a thread has been assigned, don't touch the runqueuesjeff2003-12-111-15/+21
| | | | | or load. These things have already been taken care of in sched_bind() which should be the only place that we're switching in an assigned thread.
* - Add support for CPU groups to ule. All SMT cores on the same physicaljeff2003-12-111-116/+263
| | | | | | | | | | | | | | | | cpu are added to a group. - Don't place a cpu into the kseq_idle bitmask until all cpus in that group have idled. - Prefer idle groups over idle group members in the new kseq_transfer() function. In this way we will prefer to balance load across full cores rather than add further load a partial core. - Before a cpu goes idle, check the other group members for threads. Since SMT cpus may freely share threads, this is cheap. - SMT cores may be individually pinned and bound to now. This contrasts the old mechanism where binding or pinning would have allowed a thread to run on any available cpu. - Remove some unnecessary logic from sched_switch(). Priority propagation should be properly taken care of in sched_prio() now.
* - Call mp_topology() after all CPUs have been probed.jeff2003-12-111-0/+4
|
* - Add the mp_topology() function to mp_machdep.c. This function builds upjeff2003-12-112-3/+44
| | | | | the smp_topology structure to reflect the layout of HTT enabled machines. - Add a prototype for mp_topology() in smp.h
* MFp4:simokawa2003-12-112-129/+310
| | | | | | - Introduce sbp_targ_login instead of sbp_targ_istate. - Implement reconnection and logout. - Freeze simq while bus reset.
* MFi386: remove APIC_IRQ* defines that are no longer used.peter2003-12-111-32/+0
|
* MFi386: (jhb): Deal with MAXCPU etc correctlypeter2003-12-112-4/+4
|
* regenpeter2003-12-114-5/+5
|
* Mark freebsd32_gettimeofday() as mpsafepeter2003-12-111-1/+1
|
* Just implementing a 32 bit version of gettimeofday() was smaller thanpeter2003-12-111-22/+14
| | | | the wrapper code. And it doesn't use the stackgap as a bonus.
* Move the ia32_sigtramp.S file back under amd64/. This interfaces closelypeter2003-12-112-89/+1
| | | | | | with the sendsig code in the MD area. It is not safe to assume that all the register conventions will be the same. Also, the way of producing 32 bit code (.code32 directives) in this file is amd64 specific.
* Assimilate ia64 back into the fold with the common freebsd32/ia32 code.peter2003-12-1115-3763/+96
| | | | | | | | The split-up code is derived from the ia64 code originally. Note that I have only compile-tested this, not actually run-tested it. The ia64 side of the force is missing some significant chunks of signal delivery code.
* Use the correct syscall table limitpeter2003-12-101-1/+2
|
* Finetune last commit, we need to take care of empty channels.sos2003-12-101-3/+7
|
* Fix last second typo.peter2003-12-101-1/+1
|
* Use gcc's superior ffs() builtin.peter2003-12-102-1/+9
|
* Use ffs(x) == popcnt(x ^ (x - 1)) to implement 64 bit ffsl(). gcc'speter2003-12-101-8/+15
| | | | ffs() builtin uses this already but truncates the upper 32 bits.
* Regenpeter2003-12-104-17/+25
|
* Add missing extattr_list_fd(), extattr_list_file(), extattr_list_link()peter2003-12-101-1/+4
| | | | and kse_switchin() syscall slots.
* The osigpending, oaccept, orecvfrom and ogetdirentries entries werepeter2003-12-101-8/+4
| | | | | accidently being compiled in as standard. These are part of the set of unimplemented COMPAT_43 syscall set.
* Whitespace cleanup.des2003-12-101-36/+36
|
* Regenpeter2003-12-105-6/+6
|
* Previous commit also changed the sendmsg prototype to something morepeter2003-12-100-0/+0
| | | | closely matching reality. I did not actually mean to commit that yet.
* Remove debugging printfdes2003-12-101-1/+0
|
* Document the simpler -a semantics.des2003-12-101-3/+0
|
* Fix a couple of issues in the interrupt code:des2003-12-101-21/+13
| | | | | | | | | | | | | | | - Replace overly-complicated (and buggy) -a logic with a much simpler version: -a causes all interrupts to be displayed, otherwise only those that have occurred are displayed. This removes the need for any MD code. - Instead of just making sure intrcnt is large enough, figure out the exact size it needs to be. We derive nintr from this number, and we don't want to risk printing garbage. Note that on sparc64, we end up printing garbage anyway because the names of non-existent interrupts are left uninitialized by the kernel. Tested on: alpha, i386, sparc64
* Update file locations for syscall tables to copy to.peter2003-12-101-3/+4
|
* s/u_(int|long)\>/unsigned \1/des2003-12-101-15/+15
|
* Delete old and unused APIC_IRQ defines.jhb2003-12-101-32/+0
|
* Still-born. These landed in dev/kbtr in the end.obrien2003-12-102-485/+0
|
* The DEV_ACPI option is no longer used, so g/c it.jhb2003-12-101-1/+0
|
* Use NAPICID for the maximum number of local APICs rather than MAXCPU whenjhb2003-12-101-2/+2
| | | | | doing the HTT fixup. This is a step closer to possibly having an apic.ko module someday.
* Correct usage of MAXCPU. The MAXCPU value itself is not a valid CPU IDjhb2003-12-101-2/+2
| | | | | | value as it is a count of maximum values. Reported by: bde
* Fix a couple of warnings (const a function parameter and change somedwmalone2003-12-101-4/+3
| | | | | | | ints to size_ts to better match the types of variables they are used with). Glanced at by: des
OpenPOWER on IntegriCloud