]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
20 years agoReplace the if_name and if_unit members of struct ifnet with new members
Brooks Davis [Fri, 31 Oct 2003 18:32:15 +0000 (18:32 +0000)]
Replace the if_name and if_unit members of struct ifnet with new members
if_xname, if_dname, and if_dunit. if_xname is the name of the interface
and if_dname/unit are the driver name and instance.

This change paves the way for interface renaming and enhanced pseudo
device creation and configuration symantics.

Approved By: re (in principle)
Reviewed By: njl, imp
Tested On: i386, amd64, sparc64
Obtained From: NetBSD (if_xname)

20 years ago - Take advantage of the swap pager locking: Eliminate the use of Giant
Alan Cox [Fri, 31 Oct 2003 18:32:03 +0000 (18:32 +0000)]
 - Take advantage of the swap pager locking: Eliminate the use of Giant
   from vm_object_madvise().
 - Remove excessive blank lines from vm_object_madvise().

20 years agocorrect stat to increment.
Hajimu UMEMOTO [Fri, 31 Oct 2003 17:51:54 +0000 (17:51 +0000)]
correct stat to increment.

Obtained from: KAME

20 years agoA grab bag of new (but overdue) release note items: dcons(4) and
Bruce A. Mah [Fri, 31 Oct 2003 16:58:43 +0000 (16:58 +0000)]
A grab bag of new (but overdue) release note items: dcons(4) and
friends, Pentium 4 and related workarounds, some MPSAFE network
drivers, IPv6 supports RFC 3542 (Advanced Sockets API), ifconfig(8)
staticarp, libkse default threading library for ia64.

Updates:  New DRM import.

20 years agodo not insert a dest option header (even specified by a user) that
Hajimu UMEMOTO [Fri, 31 Oct 2003 16:32:12 +0000 (16:32 +0000)]
do not insert a dest option header (even specified by a user) that
should be placed before a routing header, unless a routing header
really exists.

Obtained from: KAME

20 years agoAdd support for the Broadcom BCM5788 (just another entry to the PCI
Bill Paul [Fri, 31 Oct 2003 16:25:39 +0000 (16:25 +0000)]
Add support for the Broadcom BCM5788 (just another entry to the PCI
device ID list, probably a 5705 ASIC).

Submitted by: Marcel Prisi <marcel@virtua.ch>

20 years ago(icmp6_rip6_input) if the received data is small enough but in an
Hajimu UMEMOTO [Fri, 31 Oct 2003 16:21:26 +0000 (16:21 +0000)]
(icmp6_rip6_input) if the received data is small enough but in an
mbuf cluster, copy the data to a separate mbuf that do not use a
cluster.  this change will reduce the possiblity of packet loss
in the socket layer.

Obtained from: KAME

20 years agorename MLD6_* to MLD_*.
Hajimu UMEMOTO [Fri, 31 Oct 2003 16:07:15 +0000 (16:07 +0000)]
rename MLD6_* to MLD_*.

Obtained from: KAME

20 years agouse arc4random.
Hajimu UMEMOTO [Fri, 31 Oct 2003 16:06:05 +0000 (16:06 +0000)]
use arc4random.

Obtained from: KAME

20 years agoinitialize in6_tmpaddrtimer_ch.
Hajimu UMEMOTO [Fri, 31 Oct 2003 15:57:02 +0000 (15:57 +0000)]
initialize in6_tmpaddrtimer_ch.

Obtained from: KAME

20 years agonuku unused functions in6_nigroup_attach() and
Hajimu UMEMOTO [Fri, 31 Oct 2003 15:51:28 +0000 (15:51 +0000)]
nuku unused functions in6_nigroup_attach() and
in6_nigroup_detach().

Obtained from: KAME

20 years agoMFi386: revision 1.579.
Yoshihiro Takahashi [Fri, 31 Oct 2003 14:54:33 +0000 (14:54 +0000)]
MFi386: revision 1.579.

20 years ago- Finish externing of r_idt in the f00f hack code.
John Baldwin [Fri, 31 Oct 2003 14:24:28 +0000 (14:24 +0000)]
- Finish externing of r_idt in the f00f hack code.
- Miscellaneous style fixes in the f00f hack code and some nearby code.

Submitted by: bde

20 years agoMFi386: revision 1.198.
Yoshihiro Takahashi [Fri, 31 Oct 2003 13:50:09 +0000 (13:50 +0000)]
MFi386: revision 1.198.

20 years agoMerged from sys/isa/syscons_isa.c revision 1.24.
Yoshihiro Takahashi [Fri, 31 Oct 2003 13:48:24 +0000 (13:48 +0000)]
Merged from sys/isa/syscons_isa.c revision 1.24.

20 years agoMFi386: revision 1.578.
Yoshihiro Takahashi [Fri, 31 Oct 2003 13:47:43 +0000 (13:47 +0000)]
MFi386: revision 1.578.

20 years agoMake mtree's update function update the modification time too.
Poul-Henning Kamp [Fri, 31 Oct 2003 13:39:19 +0000 (13:39 +0000)]
Make mtree's update function update the modification time too.

Submitted by: Dan Nelson <dnelson@allantgroup.com>
PR: 53063

20 years agoFix breakage on ia64: PAGE_SIZE on ia64 is large enough that more than
Hartmut Brandt [Fri, 31 Oct 2003 13:35:22 +0000 (13:35 +0000)]
Fix breakage on ia64: PAGE_SIZE on ia64 is large enough that more than
256 raw receive buffers (96 byte each) fit into one page. This breaks the
limit imposed by the usage of an uint8_t for the buffer number. Restrict
the allocation size for buffers to a maximum of 8192.

20 years agoDon't bother passing a freshly-zeroed mbstate to mbsrtowcs() etc.
Tim J. Robbins [Fri, 31 Oct 2003 13:29:00 +0000 (13:29 +0000)]
Don't bother passing a freshly-zeroed mbstate to mbsrtowcs() etc.
when the current implementation won't use it, anyway. Just pass NULL.
This will need to be changed when state-dependent encodings are
supported, but there's no need to take the performance hit
in the meantime.

20 years agoAttached is a small patch to ministat that separates the
Poul-Henning Kamp [Fri, 31 Oct 2003 13:25:43 +0000 (13:25 +0000)]
Attached is a small patch to ministat that separates the
avg/median/stddev bars onto two lines.  Useful for datasets that
overlap.

Submitted by:    Dan Nelson <dnelson@allantgroup.com>

20 years agoRemove the code for parsing octal modes, since setmode(3) already
Anton Berezin [Fri, 31 Oct 2003 13:20:09 +0000 (13:20 +0000)]
Remove the code for parsing octal modes, since setmode(3) already
handles them.

Reviewed by: audit
MFC after: 2 weeks

20 years agoCombining "all" and "install" in one invocation is a very bad idea as
Ruslan Ermilov [Fri, 31 Oct 2003 13:15:07 +0000 (13:15 +0000)]
Combining "all" and "install" in one invocation is a very bad idea as
conditional evaluation is done only once in this case, and this may
not account for the fact that some conditionals would be evaluated
differently if "all" has already been run, like is the case with the
loader.help existance check in sys/boot/alpha/common/Makefile.common,
beforeinstall target.  This should fix Alpha snapshot builds.

20 years agoAdd a new sysctl MIB to control exclusive login.
Hidetoshi Shimokawa [Fri, 31 Oct 2003 12:05:55 +0000 (12:05 +0000)]
Add a new sysctl MIB to control exclusive login.

20 years ago - Add static to local functions and data where it was missing.
Jeff Roberson [Fri, 31 Oct 2003 11:16:04 +0000 (11:16 +0000)]
 - Add static to local functions and data where it was missing.
 - Add an IPI based mechanism for migrating kses.  This mechanism is
   broken down into several components.  This is intended to reduce cache
   thrashing by eliminating most cases where one cpu touches another's
   run queues.
 - kseq_notify() appends a kse to a lockless singly linked list and
   conditionally sends an IPI to the target processor.  Right now this is
   protected by sched_lock but at some point I'd like to get rid of the
   global lock.  This is why I used something more complicated than a
   standard queue.
 - kseq_assign() processes our list of kses that have been assigned to us
   by other processors.  This simply calls sched_add() for each item on the
   list after clearing the new KEF_ASSIGNED flag.  This flag is used to
   indicate that we have been appeneded to the assigned queue but not
   added to the run queue yet.
 - In sched_add(), instead of adding a KSE to another processor's queue we
   use kse_notify() so that we don't touch their queue.  Also in sched_add(),
   if KEF_ASSIGNED is already set return immediately.  This can happen if
   a thread is removed and readded so that the priority is recorded properly.
 - In sched_rem() return immediately if KEF_ASSIGNED is set.  All callers
   immediately readd simply to adjust priorites etc.
 - In sched_choose(), if we're running an IDLE task or the per cpu idle thread
   set our cpumask bit in 'kseq_idle' so that other processors may know that
   we are idle.  Before this, make a single pass through the run queues of
   other processors so that we may find work more immediately if it is
   available.
 - In sched_runnable(), don't scan each processor's run queue, they will IPI
   us if they have work for us to do.
 - In sched_add(), if we're adding a thread that can be migrated and we have
   plenty of work to do, try to migrate the thread to an idle kseq.
 - Simplify the logic in sched_prio() and take the KEF_ASSIGNED flag into
   consideration.
 - No longer use kseq_choose() to steal threads, it can lose it's last
   argument.
 - Create a new function runq_steal() which operates like runq_choose() but
   skips threads based on some criteria.  Currently it will not steal
   PRI_ITHD threads.  In the future this will be used for CPU binding.
 - Create a kseq_steal() that checks each run queue with runq_steal(), use
   kseq_steal() in the places where we used kseq_choose() to steal with
   before.

20 years ago - Add 4 kse flags for use in the schedulers.
Jeff Roberson [Fri, 31 Oct 2003 11:02:16 +0000 (11:02 +0000)]
 - Add 4 kse flags for use in the schedulers.

20 years agoImplement fgetrune(), fungetrune() and fputrune() as wrappers around
Tim J. Robbins [Fri, 31 Oct 2003 10:55:19 +0000 (10:55 +0000)]
Implement fgetrune(), fungetrune() and fputrune() as wrappers around
fgetwc(), ungetwc() and fputwc().

20 years agoAdd OpenBSD 3.4
Wolfram Schneider [Fri, 31 Oct 2003 10:24:02 +0000 (10:24 +0000)]
Add OpenBSD 3.4
fix footer.

20 years agoFix two bugs introduced with the rstack functionality and specific to
Marcel Moolenaar [Fri, 31 Oct 2003 07:29:28 +0000 (07:29 +0000)]
Fix two bugs introduced with the rstack functionality and specific to
the rstack functionality:
1. Fix a KASSERT that tests for the address to be above the upward
   growable stack. Typically for rstack, the faulting address can be
   identical to the record end of the upward growable entry, and
   very likely is on ia64. The KASSERT tested for greater than, not
   greater equal, so whenever the register stack had to be grown
   the assertion fired.
2. When we grow the upward growable stack entry and adjust the
   unlying object, don't forget to adjust the size of the VM map.
   Not doing so would trigger an assert in vm_mapzdtor().

Pointy hat: marcel (for not testing with INVARIANTS).

20 years agoTweak the calculation of minbfree in ffs_dirpref() so that only
Don Lewis [Fri, 31 Oct 2003 07:25:06 +0000 (07:25 +0000)]
Tweak the calculation of minbfree in ffs_dirpref() so that only
those cylinder groups that have at least 75% of the average free
space per cylinder group for that file system are considered as
candidates for the creation of a new directory.  The previous formula
for minbfree would set it to zero if the file system was more than
75% full, which allowed cylinder groups with no free space at all
to be chosen as candidates for directory creation, which resulted
in an expensive search for free blocks for each file that was
subsequently created in that directory.

Modify the calculation of minifree in the same way.

Decrease maxcontigdirs as the file system fills to decrease the
likelyhood that a cluster of directories will overflow the available
space in a cylinder group.

Reviewed by: mckusick
Tested by: kmarx@vicor.com
MFC after: 2 weeks

20 years agoRename logf --> chat_logf to avoid naming conflicts
Tim Kientzle [Fri, 31 Oct 2003 06:22:03 +0000 (06:22 +0000)]
Rename logf --> chat_logf  to avoid naming conflicts
with GCC built-in (and with C90 reserved library name).

Approved by: gordon (Mentor)

20 years agoWhile not illegal, attempt to pacify gcc -Wundef. It just so happens
Peter Wemm [Fri, 31 Oct 2003 05:42:53 +0000 (05:42 +0000)]
While not illegal, attempt to pacify gcc -Wundef.  It just so happens
that libtool-using packages seem to love using this flag.

/usr/include/sys/cdefs.h:184:5: warning: "__STDC_VERSION__" is not defined
/usr/include/sys/cdefs.h:372:5: warning: "_POSIX_C_SOURCE" is not defined
/usr/include/sys/cdefs.h:378:5: warning: "_POSIX_C_SOURCE" is not defined

20 years ago - Synchronize access to the swdevt's sw_flags with sw_dev_mtx.
Alan Cox [Fri, 31 Oct 2003 05:18:45 +0000 (05:18 +0000)]
 - Synchronize access to the swdevt's sw_flags with sw_dev_mtx.
 - Remove several instances of GIANT_REQUIRED.

20 years agoReject too large OHCI version number.
Hidetoshi Shimokawa [Fri, 31 Oct 2003 05:05:12 +0000 (05:05 +0000)]
Reject too large OHCI version number.

Invalid OHCI version indicates OHCI registers are not mapped
correctly in PCI or CardBus layer.

20 years agoAdd a compatibility macro for FreeBSD-4.
Hidetoshi Shimokawa [Fri, 31 Oct 2003 04:58:03 +0000 (04:58 +0000)]
Add a compatibility macro for FreeBSD-4.

20 years agofixe remote panic on short packet
SUZUKI Shinsuke [Fri, 31 Oct 2003 03:18:45 +0000 (03:18 +0000)]
fixe remote panic on short packet

Obtained from: KAME

20 years agoMake TUNDEBUG use if_printf instead of printf.
Brooks Davis [Fri, 31 Oct 2003 02:48:12 +0000 (02:48 +0000)]
Make TUNDEBUG use if_printf instead of printf.

20 years agoReplace a couple printfs with if_printfs.
Brooks Davis [Fri, 31 Oct 2003 01:35:07 +0000 (01:35 +0000)]
Replace a couple printfs with if_printfs.

20 years agoInitialize if_softc and use it in favor of &ipr_softc[ifp->if_unit] to
Brooks Davis [Fri, 31 Oct 2003 00:44:56 +0000 (00:44 +0000)]
Initialize if_softc and use it in favor of &ipr_softc[ifp->if_unit] to
get the softc.

20 years agoRemove unused lp_unit member from softc.
Brooks Davis [Thu, 30 Oct 2003 23:30:35 +0000 (23:30 +0000)]
Remove unused lp_unit member from softc.

20 years agoRemove unused FXP_UNIT() macro.
Brooks Davis [Thu, 30 Oct 2003 23:12:24 +0000 (23:12 +0000)]
Remove unused FXP_UNIT() macro.

20 years agoifp->if_name can't be null when we just assigned a static string to it.
Brooks Davis [Thu, 30 Oct 2003 23:05:31 +0000 (23:05 +0000)]
ifp->if_name can't be null when we just assigned a static string to it.

20 years agoOverhaul routing table entry cleanup by introducing a new rtexpunge
Sam Leffler [Thu, 30 Oct 2003 23:02:51 +0000 (23:02 +0000)]
Overhaul routing table entry cleanup by introducing a new rtexpunge
routine that takes a locked routing table reference and removes all
references to the entry in the various data structures. This
eliminates instances of recursive locking and also closes races
where the lock on the entry had to be dropped prior to calling
rtrequest(RTM_DELETE).  This also cleans up confusion where the
caller held a reference to an entry that might have been reclaimed
(and in some cases used that reference).

Supported by: FreeBSD Foundation

20 years agostyle: strcmp() does not return a bool.
Brooks Davis [Thu, 30 Oct 2003 23:00:15 +0000 (23:00 +0000)]
style: strcmp() does not return a bool.

20 years agoUse -D to rename the log() function (extensively used) rather than
Peter Wemm [Thu, 30 Oct 2003 22:59:41 +0000 (22:59 +0000)]
Use -D to rename the log() function (extensively used) rather than
using the gcc-specific -fno-builtin-log switch.

20 years agoRename the 'log' variable rather than -fno-builtin-log
Peter Wemm [Thu, 30 Oct 2003 22:57:43 +0000 (22:57 +0000)]
Rename the 'log' variable rather than -fno-builtin-log

20 years agoRather than use the gcc -fno-builtin-log flag, just rename the 'int log'
Peter Wemm [Thu, 30 Oct 2003 22:56:44 +0000 (22:56 +0000)]
Rather than use the gcc -fno-builtin-log flag, just rename the 'int log'
variable.

20 years agouse a local variable to avoid holding a lock across a call out of view
Sam Leffler [Thu, 30 Oct 2003 22:56:13 +0000 (22:56 +0000)]
use a local variable to avoid holding a lock across a call out of view

Supported by: FreeBSD Foundation

20 years agoReplace -fno-builtin-logf with a -D argument to change the function name.
Peter Wemm [Thu, 30 Oct 2003 22:53:56 +0000 (22:53 +0000)]
Replace -fno-builtin-logf with a -D argument to change the function name.

20 years agoChange the pmap_invalidate_xxx() functions so they test against
Peter Wemm [Thu, 30 Oct 2003 22:15:22 +0000 (22:15 +0000)]
Change the pmap_invalidate_xxx() functions so they test against
pmap == kernel_pmap rather than pmap->pm_active == -1.  gcc's inliner
can remove more code that way.  Only kernel_pmap has a pm_active of -1.

20 years agoFixup the sorting of some of the options. DISABLE_P* are still out of
John Baldwin [Thu, 30 Oct 2003 21:45:34 +0000 (21:45 +0000)]
Fixup the sorting of some of the options.  DISABLE_P* are still out of
order to keep all of the opt_pmap.h options together.

20 years agoEnsure that mp_ncpus is set to 1 if mp_cpu_probe() fails.
John Baldwin [Thu, 30 Oct 2003 21:44:01 +0000 (21:44 +0000)]
Ensure that mp_ncpus is set to 1 if mp_cpu_probe() fails.

20 years agoInclude "opt_pmap.h" so that the DISABLE_P* options are honored.
John Baldwin [Thu, 30 Oct 2003 21:42:44 +0000 (21:42 +0000)]
Include "opt_pmap.h" so that the DISABLE_P* options are honored.

20 years agoAlways export r_gdt and r_idt and give them extern declarations in
John Baldwin [Thu, 30 Oct 2003 21:42:17 +0000 (21:42 +0000)]
Always export r_gdt and r_idt and give them extern declarations in
machine/segments.h.

20 years agoRemove /usr/libdata/msdosfs description.
Max Khon [Thu, 30 Oct 2003 21:25:37 +0000 (21:25 +0000)]
Remove /usr/libdata/msdosfs description.

Noticed by: ru

20 years agoThere is no way to enter the attach routine twice with the same softc
Brooks Davis [Thu, 30 Oct 2003 19:45:58 +0000 (19:45 +0000)]
There is no way to enter the attach routine twice with the same softc
without a detach call in between so don't try to deal with that
possiability.

This is a diff-reduction commit for the upcoming if_xname conversion.

20 years agoMFi386: thread specific fpu state optimizations
Peter Wemm [Thu, 30 Oct 2003 19:04:58 +0000 (19:04 +0000)]
MFi386: thread specific fpu state optimizations

20 years ago- unlock on error.
Hajimu UMEMOTO [Thu, 30 Oct 2003 18:42:25 +0000 (18:42 +0000)]
- unlock on error.
- don't call malloc with M_WAITOK within lock context.

20 years agoinstall address selection policy into kernel when there is
Hajimu UMEMOTO [Thu, 30 Oct 2003 17:55:52 +0000 (17:55 +0000)]
install address selection policy into kernel when there is
/etc/ip6addrctl.conf.
following sample of ip6addrctl.conf is a good default:

#Prefix       Precedence Label
::1/128               50     0
::/0                  40     1
2002::/16             30     2
::/96                 20     3
::ffff:0:0/96         10     4

20 years agoadd ip6addrctl(8) which is a command to configure address
Hajimu UMEMOTO [Thu, 30 Oct 2003 17:46:40 +0000 (17:46 +0000)]
add ip6addrctl(8) which is a command to configure address
selection policy for IPv6 and IPv4 described in RFC3484.
source address selection part of RFC3484 is not merged
from KAME, yet.

Obtained from: KAME

20 years agoadd destination address selection described in RFC3484.
Hajimu UMEMOTO [Thu, 30 Oct 2003 17:36:53 +0000 (17:36 +0000)]
add destination address selection described in RFC3484.
in KAME implementation, even when no policy is installed
into kernel, getaddrinfo(3) sorts addresses.  Since it
causes POLA violation, I modified to don't sort addresses
when no policy is installed into kernel,

Obtained from: KAME

20 years agoCosmetic fix on some TIME related columns.
Mitsuru IWASAKI [Thu, 30 Oct 2003 16:52:14 +0000 (16:52 +0000)]
Cosmetic fix on some TIME related columns.

20 years agoRelock mntvnode_mtx if vget fails in vfs_stdsync. The loop is
Alexander Kabaev [Thu, 30 Oct 2003 16:22:51 +0000 (16:22 +0000)]
Relock mntvnode_mtx if vget fails in vfs_stdsync. The loop is
always shoould entered with mutex locked.

20 years agoExplain why the lock-free allocation algorithm is safe in our case
Hartmut Brandt [Thu, 30 Oct 2003 16:19:50 +0000 (16:19 +0000)]
Explain why the lock-free allocation algorithm is safe in our case
while beeing not safe in the general case. Thanks to David Schultz
<das@freebsd.org> for help.

20 years agoAdd sysctl MIB and loader tunable `hw.acpi.no_reset_video' as
Mitsuru IWASAKI [Thu, 30 Oct 2003 16:14:55 +0000 (16:14 +0000)]
Add sysctl MIB and loader tunable `hw.acpi.no_reset_video' as
the same effect as ACPI_NO_RESET_VIDEO kernel option.

20 years agoadd management part of address selection policy described in
Hajimu UMEMOTO [Thu, 30 Oct 2003 15:29:17 +0000 (15:29 +0000)]
add management part of address selection policy described in
RFC3484.

Obtained from: KAME

20 years agoUse (char *)NULL to terminated to argument list for execl().
Hartmut Brandt [Thu, 30 Oct 2003 15:14:34 +0000 (15:14 +0000)]
Use (char *)NULL to terminated to argument list for execl().
Without this cast the compiler cannot know that it has to convert
the null pointer constant NULL to a null pointer.

20 years agoUse (char *)NULL to terminate the argument list for execlp().
Hartmut Brandt [Thu, 30 Oct 2003 15:04:37 +0000 (15:04 +0000)]
Use (char *)NULL to terminate the argument list for execlp().
Without this cast the compiler cannot know that it has to convert the
null pointer constant NULL to a null pointer.

20 years agoAdd support for the VIA 8237 (both PATA and SATA part).
Søren Schmidt [Thu, 30 Oct 2003 13:16:21 +0000 (13:16 +0000)]
Add support for the VIA 8237 (both PATA and SATA part).
Cleanup the SATA support a bit now we are here anyways.

20 years agoAdd '#' to the characters VIS_GLOB encodes. This fixes a bug in mtree.
Poul-Henning Kamp [Thu, 30 Oct 2003 12:41:50 +0000 (12:41 +0000)]
Add '#' to the characters VIS_GLOB encodes.  This fixes a bug in mtree.

20 years agoUpdate test-case to also fail on filenames containing '#'
Poul-Henning Kamp [Thu, 30 Oct 2003 12:40:54 +0000 (12:40 +0000)]
Update test-case to also fail on filenames containing '#'

20 years agoRemove MD bits for unsupported architectures.
Tim J. Robbins [Thu, 30 Oct 2003 12:21:31 +0000 (12:21 +0000)]
Remove MD bits for unsupported architectures.

20 years agoUse vis(3) to encode/decode the target for symlinks.
Poul-Henning Kamp [Thu, 30 Oct 2003 12:03:25 +0000 (12:03 +0000)]
Use vis(3) to encode/decode the target for symlinks.

Make vis(3) decode errors fatal for both symlink targets and fienames.

20 years agoAdd another test-script for mtree(8): Check symlinks with space in them.
Poul-Henning Kamp [Thu, 30 Oct 2003 12:01:32 +0000 (12:01 +0000)]
Add another test-script for mtree(8):  Check symlinks with space in them.

20 years agoAdd regression tests for sgetrune() and sputrune().
Tim J. Robbins [Thu, 30 Oct 2003 11:51:36 +0000 (11:51 +0000)]
Add regression tests for sgetrune() and sputrune().

20 years agoFix bug in mtree(8)'s create mode where files or directories with had
Poul-Henning Kamp [Thu, 30 Oct 2003 10:48:43 +0000 (10:48 +0000)]
Fix bug in mtree(8)'s create mode where files or directories with had
names containing glob(3) expressions would appear verbatim in the
output.

If such an mtree file were used by mtree in update mode, wrong things
would happen.

20 years agoAdd a test script which Mtree fails.
Poul-Henning Kamp [Thu, 30 Oct 2003 10:46:23 +0000 (10:46 +0000)]
Add a test script which Mtree fails.

20 years agoReinvent the used-flag for external mbufs. This and the on-card flag
Hartmut Brandt [Thu, 30 Oct 2003 10:43:52 +0000 (10:43 +0000)]
Reinvent the used-flag for external mbufs. This and the on-card flag
are now in the header of the external buffer itself which allows us
to manipulate them in the free routine without having to lock the softc
structure or the free list. To get space for these flags the chunk number
is reduced to 8 bit which amounts to a maximum of 256 chunks per allocated
page. This restriction is now enforced by a CTASSERT.

20 years agoAdd a new flag to vis(3): VIS_GLOB which encodes the glob(3) magic
Poul-Henning Kamp [Thu, 30 Oct 2003 10:40:49 +0000 (10:40 +0000)]
Add a new flag to vis(3): VIS_GLOB which encodes the glob(3) magic
characters '*', '?' and '['.

20 years ago - Synchronize access to the swdevt's sw_blist with sw_dev_mtx.
Alan Cox [Thu, 30 Oct 2003 09:12:43 +0000 (09:12 +0000)]
 - Synchronize access to the swdevt's sw_blist with sw_dev_mtx.
 - Remove several instances of GIANT_REQUIRED.

20 years agoIn check.c:
Tom Rhodes [Thu, 30 Oct 2003 09:08:09 +0000 (09:08 +0000)]
In check.c:
Avoid shadowing declarations.
Avoid compairing signed and unsigned types.

20 years ago - Synchronize access to swdevhd using sw_dev_mtx.
Alan Cox [Thu, 30 Oct 2003 07:11:06 +0000 (07:11 +0000)]
 - Synchronize access to swdevhd using sw_dev_mtx.
 - Use swp_sizecheck() rather than assignment to swap_pager_full in
   swaponsomething().

20 years agoRemove a few unused variables.
Tom Rhodes [Thu, 30 Oct 2003 05:43:56 +0000 (05:43 +0000)]
Remove a few unused variables.

20 years agoMFi386: rev 1.451 (jhb): call pmap_kremove() rather than duplicate it
Peter Wemm [Thu, 30 Oct 2003 04:08:22 +0000 (04:08 +0000)]
MFi386: rev 1.451 (jhb): call pmap_kremove() rather than duplicate it

20 years agoMFi386: trap.c rev 1.259: fetch thread mailbox address in page fault trap
Peter Wemm [Thu, 30 Oct 2003 04:06:28 +0000 (04:06 +0000)]
MFi386: trap.c rev 1.259: fetch thread mailbox address in page fault trap

20 years agoTry to fetch thread mailbox address in page fault trap, so when thread
David Xu [Thu, 30 Oct 2003 02:55:43 +0000 (02:55 +0000)]
Try to fetch thread mailbox address in page fault trap, so when thread
blocks in page fault hanlder, and upcall thread can be scheduled. It is
useful if process is doing lots of mmap based I/O.

20 years agoOops, forgot to save these in the editor. Add CTASSERTS for signal and
Peter Wemm [Thu, 30 Oct 2003 02:43:19 +0000 (02:43 +0000)]
Oops, forgot to save these in the editor.  Add CTASSERTS for signal and
context related things.

20 years agoAdd CTASSERT()'s to check that the sizes of our replicas of the 32 bit
Peter Wemm [Thu, 30 Oct 2003 02:40:30 +0000 (02:40 +0000)]
Add CTASSERT()'s to check that the sizes of our replicas of the 32 bit
structures come out the right size.

Fix the ones that broke.  stat32 had some missing fields from the end
and statfs32 was broken due to the strange definition of MNAMELEN
(which is dependent on sizeof(long))

I'm not sure if this fixes any actual problems or not.

20 years agoavoid recursive lock panic by unlocking before calling rtrequest;
Sam Leffler [Wed, 29 Oct 2003 23:01:37 +0000 (23:01 +0000)]
avoid recursive lock panic by unlocking before calling rtrequest;
this is consistent with other places but will be replaced
shortly by a "proper fix"

Supported by: FreeBSD Foundation
Pain felt by: Jiri Mikulas

20 years agocorrect LOR by using a local variable to hold result
Sam Leffler [Wed, 29 Oct 2003 22:59:12 +0000 (22:59 +0000)]
correct LOR by using a local variable to hold result
instead of holding a lock while calling out of view

Supported by: FreeBSD Foundation

20 years agoDon't attempt to destroy the driver mutex when it hasn't yet
Daniel Eischen [Wed, 29 Oct 2003 21:54:37 +0000 (21:54 +0000)]
Don't attempt to destroy the driver mutex when it hasn't yet
been created.  This has been sitting in my local tree for
far too long; I can't believe noone else has come across this
yet.

20 years agostyle(9): sort functions.
Tom Rhodes [Wed, 29 Oct 2003 21:23:44 +0000 (21:23 +0000)]
style(9): sort functions.

20 years agoProperly prototype C function usage().
Tom Rhodes [Wed, 29 Oct 2003 21:21:09 +0000 (21:21 +0000)]
Properly prototype C function usage().
Sync usage() with the manual page: s/file/snapshot_name/g.

20 years agoFix a panic that occurs when resuming. For some reason, sc->cur_scp is
Nate Lawson [Wed, 29 Oct 2003 20:48:13 +0000 (20:48 +0000)]
Fix a panic that occurs when resuming.  For some reason, sc->cur_scp is
NULL.

Submitted by: Andrew Thompson <andy@fud.org.nz>

20 years agoBasic PAM authentication support.
Ruslan Ermilov [Wed, 29 Oct 2003 20:32:19 +0000 (20:32 +0000)]
Basic PAM authentication support.

20 years agoDisclaim ownership of pppd/chat. It is vendor code, but has been badly
Peter Wemm [Wed, 29 Oct 2003 19:24:04 +0000 (19:24 +0000)]
Disclaim ownership of pppd/chat.  It is vendor code, but has been badly
neglected here.

20 years agoPotential fix for races shutting down callouts when unloading
Sam Leffler [Wed, 29 Oct 2003 19:15:00 +0000 (19:15 +0000)]
Potential fix for races shutting down callouts when unloading
the module.  Previously we grabbed the mutex used by the callouts,
then stopped the callout with callout_stop, but if the callout
was already active and blocked by the mutex then it would continue
later and reference the mutex after it was destroyed.  Instead
stop the callout first then lock.

Supported by: FreeBSD Foundation

20 years agoo add locking to protect routing table refcnt manipulations
Sam Leffler [Wed, 29 Oct 2003 19:03:58 +0000 (19:03 +0000)]
o add locking to protect routing table refcnt manipulations
o add some more debugging help for figuring out why folks are
  getting complaints about releasing routing table entries with
  a zero refcnt
o fix comment that talked about spl's
o remove duplicate define of DUMMYNET_DEBUG

Supported by: FreeBSD Foundation

20 years agoAlways queue looped back packets (rather than potentially using
Sam Leffler [Wed, 29 Oct 2003 18:37:47 +0000 (18:37 +0000)]
Always queue looped back packets (rather than potentially using
direct dispatch) to avoid extensive kernel stack usage and to
avoid directly re-entering the network stack.  The latter causes
locking problems when, for example, a complete TCP handshake`
happens w/o a context switch.

20 years agomark interrupt handlers MPSAFE
Sam Leffler [Wed, 29 Oct 2003 18:32:14 +0000 (18:32 +0000)]
mark interrupt handlers MPSAFE

20 years agoAdd a temporary mechanism to disble INTR_MPSAFE from network interface
Sam Leffler [Wed, 29 Oct 2003 18:29:50 +0000 (18:29 +0000)]
Add a temporary mechanism to disble INTR_MPSAFE from network interface
drivers.  This is prepatory to running more parts of the network system
w/o Giant.