]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
17 years agoNew release notes:
Christian Brueffer [Tue, 27 Mar 2007 21:26:45 +0000 (21:26 +0000)]
New release notes:
- priv(9) KPI added
- ipw/iwi firmware in the base system

Updated release notes:
- OpenSSL updated to 0.9.8e
- GNOME updated to 2.8.0
- acpi_dock(4) and acpi_hpet MFC
- msk(4) MFC

17 years agoClarify the definition of the watchdog kernel interface by moving a sentence up
Nick Hibma [Tue, 27 Mar 2007 21:23:20 +0000 (21:23 +0000)]
Clarify the definition of the watchdog kernel interface by moving a sentence up
to the sentence it references.

17 years agoAdd MLINKs for priv(9) functions.
Christian Brueffer [Tue, 27 Mar 2007 21:20:57 +0000 (21:20 +0000)]
Add MLINKs for priv(9) functions.

17 years agoRevisit the watchdogs: Resetting the error to EINVAL after failing to set the
Nick Hibma [Tue, 27 Mar 2007 21:03:37 +0000 (21:03 +0000)]
Revisit the watchdogs: Resetting the error to EINVAL after failing to set the
watchdog might hide the succesful arming of an earlier one. Accept that on
failing to arm any watchdog (because of non-supported timeouts) EOPNOTSUPP is
returned instead of the more appropriate EINVAL.

MFC after: 3 days

17 years agoFix regression in rev. 1.140.
Gleb Smirnoff [Tue, 27 Mar 2007 19:36:12 +0000 (19:36 +0000)]
Fix regression in rev. 1.140.

Reported by: Yuriy Tsibizov <Yuriy.Tsibizov gfk.ru>, bsam

17 years agoSort.
John Baldwin [Tue, 27 Mar 2007 19:32:40 +0000 (19:32 +0000)]
Sort.

17 years agoDon't assume the run bit is still set after a suspend.
Julian Elischer [Tue, 27 Mar 2007 18:55:06 +0000 (18:55 +0000)]
Don't assume the run bit is still set after a suspend.
Submitted by: Andrea Bittau ( adotbittauatcsdotucldotacdotuk)
MFC After: 1 week

17 years agoCorrect HPET header handling.
Takanori Watanabe [Tue, 27 Mar 2007 17:03:52 +0000 (17:03 +0000)]
Correct HPET header handling.
MFC after: 7 days

17 years agoAlign 'struct thread' on 16 byte boundaries so that the lower 4 bits are
John Baldwin [Tue, 27 Mar 2007 16:51:34 +0000 (16:51 +0000)]
Align 'struct thread' on 16 byte boundaries so that the lower 4 bits are
always 0.  Previously we aligned threads on a minimum of 8-byte boundaries.

Note: This changes the uma zone to no longer cache align threads.  We
really want the uma zone to do align threads to MAX(16, cache line size)
but there currently isn't a good way to express that to uma.

Submitted by: attilio

17 years agoFix a case where hardware removal of an interface caused an attempt to
Bruce M Simpson [Tue, 27 Mar 2007 16:11:28 +0000 (16:11 +0000)]
Fix a case where hardware removal of an interface caused an attempt to
announce an ll_ifma which has gone away. Add a KASSERT to catch regressions.

Bug found by: Tom Uffner

17 years agoFix a bug which could lead to receive side lockup when WC is disabled.
Andrew Gallatin [Tue, 27 Mar 2007 15:55:32 +0000 (15:55 +0000)]
Fix a bug which could lead to receive side lockup when WC is disabled.
When submitting rx buffers and not using WC fifo, always replace the
invalid DMA address with the real one, otherwise allocation failures
could lead to the invalid DMA address being given to the NIC, and
that would cause the receive side to lockup.

17 years agoFix a comment grammar nit.
John Baldwin [Tue, 27 Mar 2007 15:09:10 +0000 (15:09 +0000)]
Fix a comment grammar nit.

17 years agoIn account management, verify whether the account has been locked
Yaroslav Tykhiy [Tue, 27 Mar 2007 09:59:15 +0000 (09:59 +0000)]
In account management, verify whether the account has been locked
with `pw lock', so that it's impossible to log into a locked account
using an alternative authentication mechanism, such as an ssh key.
This change affects only accounts locked with pw(8), i.e., having a
`*LOCKED*' prefix in their password hash field, so people still can
use a different pattern to disable password authentication only.

Mention all account management criteria in the manpage.

Approved by: maintainer (timeout)
PR: bin/71147
MFC after: 1 month

17 years agoPrevent a race between vm_object_collapse() and vm_object_split() from
Alan Cox [Tue, 27 Mar 2007 08:55:17 +0000 (08:55 +0000)]
Prevent a race between vm_object_collapse() and vm_object_split() from
causing a crash.

Suppose that we have two objects, obj and backing_obj, where
backing_obj is obj's backing object.  Further, suppose that
backing_obj has a reference count of two.  One being the reference
held by obj and the other by a map entry.  Now, suppose that the map
entry is deallocated and its reference removed by
vm_object_deallocate().  vm_object_deallocate() recognizes that the
only remaining reference is from a shadow object, obj, and calls
vm_object_collapse() on obj.  vm_object_collapse() executes

                if (backing_object->ref_count == 1) {
                        /*
                         * If there is exactly one reference to the backing
                         * object, we can collapse it into the parent.
                         */
                        vm_object_backing_scan(object, OBSC_COLLAPSE_WAIT);

vm_object_backing_scan(OBSC_COLLAPSE_WAIT) executes

        if (op & OBSC_COLLAPSE_WAIT) {
                vm_object_set_flag(backing_object, OBJ_DEAD);
        }

Finally, suppose that either vm_object_backing_scan() or
vm_object_collapse() sleeps releasing its locks.  At this instant,
another thread executes vm_object_split().  It crashes in
vm_object_reference_locked() on the assertion that the object is not
dead.  If, however, assertions are not enabled, it crashes much later,
after the object has been recycled, in vm_object_deallocate() because
the shadow count and shadow list are inconsistent.

Reviewed by: tegge
Reported by: jhb
MFC after: 1 week

17 years agoDon't map mini-data cache page since ARM920T doesn't have it.
Kevin Lo [Tue, 27 Mar 2007 06:29:19 +0000 (06:29 +0000)]
Don't map mini-data cache page since ARM920T doesn't have it.

17 years agoPowerPC is the only architecture with mpsafe_vfs=0. This is now
Marcel Moolenaar [Tue, 27 Mar 2007 05:29:41 +0000 (05:29 +0000)]
PowerPC is the only architecture with mpsafe_vfs=0. This is now
broken. Rudimentary tests show that PowerPC can run with
mpsafe_vfs=1. Make it so...

17 years agoDescribe the contents of the "ar_name" and "ar_rawname" fields of
Joseph Koshy [Tue, 27 Mar 2007 04:47:50 +0000 (04:47 +0000)]
Describe the contents of the "ar_name" and "ar_rawname" fields of
Elf_Arhdr structures better.

17 years agoBug fixes to ar(1) archive handling:
Joseph Koshy [Tue, 27 Mar 2007 04:40:57 +0000 (04:40 +0000)]
Bug fixes to ar(1) archive handling:
 - Correctly retrieve the initial (special) members of an archive after
   an archive descriptor is rewound using elf_rand(SARMAG).
 - Do not strip trailing white space from the 'raw' names retrieved
   using elf_getarhdr().

Reported by: "Hyo geol, Lee" <hyogeollee at gmail dot com>

17 years agoIntroduce the new option -M to allow to set the permissions of
Lukas Ertl [Mon, 26 Mar 2007 23:29:20 +0000 (23:29 +0000)]
Introduce the new option -M to allow to set the permissions of
the user's newly created home directory.  If omitted, it's derived
from the current umask.

PR:      bin/16880, bin/83253 (partially), bin/104248
MFC in:  1 month

17 years ago- Use '*h' instead of 'struct acpi_spinlock' for sizeof[1].
Jung-uk Kim [Mon, 26 Mar 2007 23:04:02 +0000 (23:04 +0000)]
- Use '*h' instead of 'struct acpi_spinlock' for sizeof[1].
- Add a missing 'else' for 'if'[2].

Requested by: njl[1]
Submitted by: njl[2]

17 years agoCheck if the new user already exists right after entering the
Lukas Ertl [Mon, 26 Mar 2007 22:22:10 +0000 (22:22 +0000)]
Check if the new user already exists right after entering the
username instead of watching the final call to pw(8) fail.

17 years agoAnnotate that this giant acqusition is dependent on tty locking.
Kris Kennaway [Mon, 26 Mar 2007 21:56:46 +0000 (21:56 +0000)]
Annotate that this giant acqusition is dependent on tty locking.

17 years agoCorrect ACPI semaphore function parameters.
Jung-uk Kim [Mon, 26 Mar 2007 21:56:35 +0000 (21:56 +0000)]
Correct ACPI semaphore function parameters.

17 years agoFix setting of serial port speed. A junk value was passed in AX when
Thomas Quinot [Mon, 26 Mar 2007 21:56:13 +0000 (21:56 +0000)]
Fix setting of serial port speed. A junk value was passed in AX when
bioscom is called to set up serial port parameters because COMSPEED
was treated as an address instead of an immediate value, causing
serial port parameters to never be set.

PR: i386/110828
Reviewed by: jhb
MFC after: 2 weeks

17 years agomake_dev(9) can be (and is) called without Giant, so there is no need to
Kris Kennaway [Mon, 26 Mar 2007 21:47:03 +0000 (21:47 +0000)]
make_dev(9) can be (and is) called without Giant, so there is no need to
drop the topology lock and acquire Giant around this call.

Reviewed by: phk

17 years agoRemove unnecessary giant acquisition around panic in #ifdef DIAGNOSTIC
Kris Kennaway [Mon, 26 Mar 2007 21:45:44 +0000 (21:45 +0000)]
Remove unnecessary giant acquisition around panic in #ifdef DIAGNOSTIC
code.

# There is some question about whether this code is even relevant any
# longer (it dates back to prehistoric times, i.e. present in r1.1),
# especially on amd64.

Reviewed by: jhb

17 years agoFree the handle, not the lock. Pointy hat to me.
Jung-uk Kim [Mon, 26 Mar 2007 21:36:31 +0000 (21:36 +0000)]
Free the handle, not the lock.  Pointy hat to me.

17 years agoCorrect ACPI spinlock function parameters and use known ACPI spinlock names.
Jung-uk Kim [Mon, 26 Mar 2007 21:23:23 +0000 (21:23 +0000)]
Correct ACPI spinlock function parameters and use known ACPI spinlock names.

17 years agoChange the VPD code to read the VPD data on-demand when a driver asks for
John Baldwin [Mon, 26 Mar 2007 20:18:52 +0000 (20:18 +0000)]
Change the VPD code to read the VPD data on-demand when a driver asks for
it via pci_get_vpd_*() rather than always reading it for each device during
boot.  I've left the tunable so that it can still be turned off if a device
driver causes a lockup via a query to a broken device, but devices whose
drivers do not use VPD (the vast majority) should no longer result in
lockups during boot, and most folks should not need to tweak the tunable
now.

Tested on: bge(4)
Silence from: jmg

17 years agoUse a unique name for each mutex now that acpi-ca is creating more than
Nate Lawson [Mon, 26 Mar 2007 19:38:28 +0000 (19:38 +0000)]
Use a unique name for each mutex now that acpi-ca is creating more than
one (hardware & global lock).  This should address witness complaints that
a duplicate mutex is being acquired.  Be sure to free the mutex to fix a
potential memory leak.

MFC after: 3 days

17 years agoo Update a comment: sonewconn() lives in uipc_socket.c now.
Maxim Konovalov [Mon, 26 Mar 2007 18:17:57 +0000 (18:17 +0000)]
o Update a comment: sonewconn() lives in uipc_socket.c now.

17 years agoBump FreeBSD version for inclusion of CPU frequency change notifiers.
Nate Lawson [Mon, 26 Mar 2007 18:04:41 +0000 (18:04 +0000)]
Bump FreeBSD version for inclusion of CPU frequency change notifiers.

17 years agoAdd an interface for drivers to be notified of changes to CPU frequency.
Nate Lawson [Mon, 26 Mar 2007 18:03:29 +0000 (18:03 +0000)]
Add an interface for drivers to be notified of changes to CPU frequency.
cpufreq_pre_change is called before the change, giving each driver a chance
to revoke the change.  cpufreq_post_change provides the results of the
change (success or failure).  cpufreq_levels_changed gives the unit number
of the cpufreq device whose number of available levels has changed.  Hook
in all the drivers I could find that needed it.

* TSC: update TSC frequency value.  When the available levels change, take the
highest possible level and notify the timecounter set_cputicker() of that
freq.  This gets rid of the "calcru: runtime went backwards" messages.
* identcpu: updates the sysctl hw.clockrate value
* Profiling: if profiling is active when the clock changes, let the user
know the results may be inaccurate.

Reviewed by: bde, phk
MFC after: 1 month

17 years agoAvoid manipulating semu_list outside of the scope of SEMUNDO_LOCK(). This
Ed Maste [Mon, 26 Mar 2007 17:41:14 +0000 (17:41 +0000)]
Avoid manipulating semu_list outside of the scope of SEMUNDO_LOCK().  This
would lead to an occasional hang with a cycle in semu_list.

X-Discussed-On: hackers@

17 years agoClarify memory management rules for pmc_cpuinfo().
Joseph Koshy [Mon, 26 Mar 2007 17:33:41 +0000 (17:33 +0000)]
Clarify memory management rules for pmc_cpuinfo().

Suggested by: "Harald Servat" <redcrash at gmail dot com>

17 years agoFollowing movement of functions from uipc_socket2.c to uipc_socket.c and
Robert Watson [Mon, 26 Mar 2007 17:05:09 +0000 (17:05 +0000)]
Following movement of functions from uipc_socket2.c to uipc_socket.c and
uipc_sockbuf.c, clean up and update comments.

17 years agoDocument the return type of elf_rand(3) correctly.
Joseph Koshy [Mon, 26 Mar 2007 16:31:42 +0000 (16:31 +0000)]
Document the return type of elf_rand(3) correctly.

17 years agoFix a silly bogon that broke ibcs2_rename().
John Baldwin [Mon, 26 Mar 2007 15:39:49 +0000 (15:39 +0000)]
Fix a silly bogon that broke ibcs2_rename().

CID: 1065
Found by: Coverity Prevent (tm)
Reported by: netchild

17 years agoInitialize vfslocked to 0 before nfsm_srvmtofh() so that the variable is
John Baldwin [Mon, 26 Mar 2007 15:14:58 +0000 (15:14 +0000)]
Initialize vfslocked to 0 before nfsm_srvmtofh() so that the variable is
not used uninitialized in 'nfsmout' if nfsm_srvmtofh() gets an internal
error.

CID: 1766
Found by: Coverity Prevent (tm)

17 years ago- Fix typo
Gabor Kovesdan [Mon, 26 Mar 2007 11:44:46 +0000 (11:44 +0000)]
- Fix typo

PR: bin/110648
Submitted by: Seth Hieronymus <shieronymus@speakeasy.net>
Approved by: keramida (mentor)
MFC after: 3 days

17 years agoFix some statements in disc(4) and about it:
Yaroslav Tykhiy [Mon, 26 Mar 2007 09:10:28 +0000 (09:10 +0000)]
Fix some statements in disc(4) and about it:
- ifnet is no more embedded in softc;
- the interface name is `disc', not `ds'.

17 years agoGive a hint that softc can contain many things besides ifp.
Yaroslav Tykhiy [Mon, 26 Mar 2007 09:05:10 +0000 (09:05 +0000)]
Give a hint that softc can contain many things besides ifp.

17 years agoComplete removal of uipc_socket2.c by moving the last few functions to
Robert Watson [Mon, 26 Mar 2007 08:59:03 +0000 (08:59 +0000)]
Complete removal of uipc_socket2.c by moving the last few functions to
other C files:

- Move sbcreatecontrol() and sbtoxsockbuf() to uipc_sockbuf.c.  While
  sbcreatecontrol() is really an mbuf allocation routine, it does its work
  with awareness of the layout of socket buffer memory.

- Move pru_*() protocol switch stubs to uipc_socket.c where the non-stub
  versions of several of these functions live.  Likewise, move socket state
  transition calls (soisconnecting(), etc) to uipc_socket.c.  Moveo
  sodupsockaddr() and sotoxsocket().

17 years agoWe no longer embed ifnet in softc, and the pointer to ifnet
Yaroslav Tykhiy [Mon, 26 Mar 2007 08:58:22 +0000 (08:58 +0000)]
We no longer embed ifnet in softc, and the pointer to ifnet
doesn't need to be first in softc now.  (It was the whole
ifnet structure itself that needed to be first in the good
old days.)  Fix the respective comment accordingly.

Add xrefs to ifnet(9) in some other comments while I'm here.

Pointed out by: thompsa

17 years agoAdding tests for Camellia which will be added to the tree soon.
George V. Neville-Neil [Mon, 26 Mar 2007 07:40:51 +0000 (07:40 +0000)]
Adding tests for Camellia which will be added to the tree soon.
Added tests for aes-ctr which were not present before.

17 years agoIntroduce a new toy interface, edsc(4). It's a discard interface
Yaroslav Tykhiy [Mon, 26 Mar 2007 04:50:06 +0000 (04:50 +0000)]
Introduce a new toy interface, edsc(4).  It's a discard interface
imitating an Ethernet device, so vlan(4) and if_bridge(4) can be
attached to it for testing and benchmarking purposes.  Its source
can be an introduction to the anatomy of a network interface driver
due to its simplicity as well as to a bunch of comments in it.

(The rest of needed changes were in my previous commit, which got
interrupted in the middle.  Alas, CVS commits are not atomic.)

17 years agoIntroduce a new toy interface, edsc(4). It's a discard interface
Yaroslav Tykhiy [Mon, 26 Mar 2007 04:39:18 +0000 (04:39 +0000)]
Introduce a new toy interface, edsc(4).  It's a discard interface
imitating an Ethernet device, so vlan(4) and if_bridge(4) can be
attached to it for testing and benchmarking purposes.  Its source
can be an introduction to the anatomy of a network interface driver
due to its simplicity as well as to a bunch of comments in it.

17 years agoUse "-ne" instead of "!=" for integer comparison.
Brooks Davis [Sun, 25 Mar 2007 23:58:46 +0000 (23:58 +0000)]
Use "-ne" instead of "!=" for integer comparison.

17 years agoIn tcp_sack_doack() remove too tight KASSERT() added in last revision. This
Andre Oppermann [Sun, 25 Mar 2007 23:27:26 +0000 (23:27 +0000)]
In tcp_sack_doack() remove too tight KASSERT() added in last revision.  This
function may be called without any TCP SACK option blocks present.  Protect
iteration over SACK option blocks by checking for SACK options present flag
first.

Bug reported by: wkoszek, keramida, Nicolas Blais

17 years agoReplace a comment about RSVP/mrouting with a different but similar comment
Robert Watson [Sun, 25 Mar 2007 21:49:50 +0000 (21:49 +0000)]
Replace a comment about RSVP/mrouting with a different but similar comment
explaining that some more locking is needed.  The routing pieces are done,
but there is an interlocking issue between optionally compiled code and
mandatory code.

Spotted by: kris

17 years agoReplace GIANT_REQUIRED's present for socket locking with NET_LOCK_GIANT().
Robert Watson [Sun, 25 Mar 2007 21:44:24 +0000 (21:44 +0000)]
Replace GIANT_REQUIRED's present for socket locking with NET_LOCK_GIANT().
If/when someone does the necessary MPSAFEty locking for the NFSv4 client,
the socket code is generally MPSAFE now.

Spotted by: kris

17 years agoRemove ancient preprocessor code. Fix module compilation.
Roman Kurakin [Sun, 25 Mar 2007 20:21:31 +0000 (20:21 +0000)]
Remove ancient preprocessor code. Fix module compilation.

Requested by: n_hibma

17 years agoHaving gone to the trouble of setting up a variable, we should use it.
Mike Makonnen [Sun, 25 Mar 2007 19:45:20 +0000 (19:45 +0000)]
Having gone to the trouble of setting up a variable, we should use it.

17 years agoTwo small changes to vm_map_pmap_enter():
Alan Cox [Sun, 25 Mar 2007 19:33:40 +0000 (19:33 +0000)]
Two small changes to vm_map_pmap_enter():

1) Eliminate an unnecessary check for fictitious pages.  Specifically,
only device-backed objects contain fictitious pages and the object is
not device-backed.

2) Change the types of "psize" and "tmpidx" to vm_pindex_t in order to
prevent possible wrap around with extremely large maps and objects,
respectively.  Observed by: tegge (last summer)

17 years agoo Really commit typo fixes to HEAD.
Maxim Konovalov [Sun, 25 Mar 2007 18:37:59 +0000 (18:37 +0000)]
o Really commit typo fixes to HEAD.

PR: docs/110809
Submitted by: naddy

17 years agoCorrect a comment typo
Kris Kennaway [Sun, 25 Mar 2007 10:07:23 +0000 (10:07 +0000)]
Correct a comment typo

17 years agoUpdate a comment: we usually call exec_vmspace_new with Giant not held,
Kris Kennaway [Sun, 25 Mar 2007 10:05:44 +0000 (10:05 +0000)]
Update a comment: we usually call exec_vmspace_new with Giant not held,
but sometimes it is.

17 years agoo cd9660 code repo-copied, update a comment.
Maxim Konovalov [Sat, 24 Mar 2007 22:40:16 +0000 (22:40 +0000)]
o cd9660 code repo-copied, update a comment.

17 years agoo Remove isofs from the cscope dirs list. Its content was
Maxim Konovalov [Sat, 24 Mar 2007 22:21:01 +0000 (22:21 +0000)]
o Remove isofs from the cscope dirs list.  Its content was
repo-copied to fs/.

17 years agoo Use a define for a buffer size.
Maxim Konovalov [Sat, 24 Mar 2007 22:15:02 +0000 (22:15 +0000)]
o Use a define for a buffer size.

Prodded by: db

o Add missed vars for TCPDEBUG in tcp_do_segment().

Prodded by: tinderbox

17 years agoFix some subtle bugs for posix_memalign() having to do with integer
Jason Evans [Sat, 24 Mar 2007 20:44:06 +0000 (20:44 +0000)]
Fix some subtle bugs for posix_memalign() having to do with integer
rounding and overflow.  Carefully document what the various overflow
tests actually detect.

The bugs mostly canceled out, such that the worst possible failure
cases resulted in non-fatal over-allocations.

17 years agoBack out rev. 1.129 because it breaks the practice of auto-loading
Yaroslav Tykhiy [Sat, 24 Mar 2007 20:26:54 +0000 (20:26 +0000)]
Back out rev. 1.129 because it breaks the practice of auto-loading
hardware drivers.  Unlike pseudo-device drivers, which just attach
to the cloning framework and wait for "ifconfig create", h/w drivers
create interfaces for installed cards as soon as loaded.  The issue
of devd(8) involuntarily reloading modules should be dealt with in a
different way.

17 years agoGeneral style cleanup.
Robert Watson [Sat, 24 Mar 2007 20:19:44 +0000 (20:19 +0000)]
General style cleanup.

Correct spelling errors.

Remove references to M_COPY_PKTHDR -- it was deprecated in 6.x and is not
used (or defined) in our tree.

17 years agoIn order to satisfy ACPI's need for an identity mapping, modify the
Alan Cox [Sat, 24 Mar 2007 19:53:22 +0000 (19:53 +0000)]
In order to satisfy ACPI's need for an identity mapping, modify the
temporary mapping created by locore so that the lowest two to four
megabytes can become a permanent identity mapping.  This implementation
avoids any use of a large page mapping.

17 years agoDrop the ipw softc lock before calling back into net80211, fixing a
Lukas Ertl [Sat, 24 Mar 2007 18:53:33 +0000 (18:53 +0000)]
Drop the ipw softc lock before calling back into net80211, fixing a
LOR/deadlock.

Tested by:   Denis Shaposhnikov <dsh_AT_vlink.ru>, le@
LOR id:      205
MFC in:      3 days

17 years agoUpdate the regression test so that there are actually two independent
George V. Neville-Neil [Sat, 24 Mar 2007 13:47:16 +0000 (13:47 +0000)]
Update the regression test so that there are actually two independent
networks involved, as opposed to two hosts on one network.

MFC after: 1 week

17 years agoo A quirk for Sagem USB-Serial controller.
Maxim Konovalov [Sat, 24 Mar 2007 09:27:58 +0000 (09:27 +0000)]
o A quirk for Sagem USB-Serial controller.

PR: usb/109613
Submitted by: Mayr Gerald
MFC after: 1 month

17 years agoo Add several CDMA-2000 terminals.
Maxim Konovalov [Sat, 24 Mar 2007 09:25:56 +0000 (09:25 +0000)]
o Add several CDMA-2000 terminals.

PR: usb/109838
Submitted by: R.Mahmatkhanov
MFC after: 1 month

17 years agoNotice when mkdir() fails.
Tim Kientzle [Sat, 24 Mar 2007 05:02:16 +0000 (05:02 +0000)]
Notice when mkdir() fails.
Don't change permissions on an existing dir unless _EXTRACT_PERM
is requested.

In particular, bsdtar -x should not edit mode of existing dirs
now; bsdtar -xp will.

17 years agobus_size_t is a bad cross-architectural type with respect to printf, use uint32_t...
Kip Macy [Sat, 24 Mar 2007 04:28:33 +0000 (04:28 +0000)]
bus_size_t is a bad cross-architectural type with respect to printf, use uint32_t instead

17 years agoIssue a warning if there's a non-zero exit value.
Tim Kientzle [Sat, 24 Mar 2007 03:25:49 +0000 (03:25 +0000)]
Issue a warning if there's a non-zero exit value.

17 years agoFix compile error when libbz2 is unavailable.
Tim Kientzle [Sat, 24 Mar 2007 03:23:26 +0000 (03:23 +0000)]
Fix compile error when libbz2 is unavailable.

17 years agoo While the script is correct and works well in sh(1) make it
Maxim Konovalov [Sat, 24 Mar 2007 01:39:24 +0000 (01:39 +0000)]
o While the script is correct and works well in sh(1) make it
works in tcsh(1) which does not support <<- redirection.

PR: docs/110097
Submitted by: Sameh Ghane

17 years agoDefault to booting off the SD card. It is more useful, and a full
Warner Losh [Fri, 23 Mar 2007 23:47:59 +0000 (23:47 +0000)]
Default to booting off the SD card.  It is more useful, and a full
FreeBSD/arm installworld install is only 170MB.  The smallest SD card
I could find at the store today was 512MB (and it was only $10 after
rebate), with a 2GB card for as low as $25.00...

Now that the IIC stuff has been sorted out, include that as well.
Include hints for the icee 16kb 16-bit i2c device.  It should include
info about the temperature sensor as well, but that driver isn't quite
ready.

Add bpf for dhclient happiness.

MFC After: 1 week

17 years agoNew device: icee. Generic i2c eeprom driver.
Warner Losh [Fri, 23 Mar 2007 23:10:35 +0000 (23:10 +0000)]
New device: icee.  Generic i2c eeprom driver.

17 years agoMFp4: Make the iicbus fully hinted. We no longer automatically add
Warner Losh [Fri, 23 Mar 2007 23:08:28 +0000 (23:08 +0000)]
MFp4: Make the iicbus fully hinted.  We no longer automatically add
some devices (and not others).  To get instances onto the iicbus, one
now needs hints or an identify routine.  We also do not probe the bus
for devices because many iic devices cannot be safely probed (and when
they can, the probe order turns out to be somewhat difficult to get
right).

# I'm not 100% sure that the iicsmb removal is right.  Please contact me if
# this causes difficulty.

17 years agoMFp4: Make iicbus_trasnfer_gen suitable for bridge drivers. Use it in the
Warner Losh [Fri, 23 Mar 2007 23:03:54 +0000 (23:03 +0000)]
MFp4: Make iicbus_trasnfer_gen suitable for bridge drivers.  Use it in the
bitbang bridge.

17 years agoMFp4: Create an ivar for each iic device on the iicbus. This ivar
Warner Losh [Fri, 23 Mar 2007 23:02:33 +0000 (23:02 +0000)]
MFp4: Create an ivar for each iic device on the iicbus.  This ivar
holds the device's address.

17 years agoFix posix_memalign() for large objects. Now that runs are extents rather
Jason Evans [Fri, 23 Mar 2007 22:58:15 +0000 (22:58 +0000)]
Fix posix_memalign() for large objects.  Now that runs are extents rather
than binary buddies, the alignment guarantees are weaker, which requires
a more complex aligned allocation algorithm, similar to that used for
alignment greater than the chunk size.

Reported by: matteo

17 years agoMFp4: A bunch of patches from myself and Tisco to improve the
Warner Losh [Fri, 23 Mar 2007 22:57:24 +0000 (22:57 +0000)]
MFp4: A bunch of patches from myself and Tisco to improve the
robustness of IIC transactions when parts aren't present.  This also
removes a bunch of debug.  This also moves this driver to 7-1
addressing rather than 6-0 addressing, which is more inline with all
the other iic drivers in the tree.  I've tested this for about a
million years on the systems at work.

17 years agoSwitch to ANSI function declarations.
Craig Rodrigues [Fri, 23 Mar 2007 22:48:44 +0000 (22:48 +0000)]
Switch to ANSI function declarations.

17 years agoUpdate to FICL 3.03 (the last release before FICL4 rewrite).
Jung-uk Kim [Fri, 23 Mar 2007 22:26:01 +0000 (22:26 +0000)]
Update to FICL 3.03 (the last release before FICL4 rewrite).

The relevant changes for FreeBSD (excerpt from the release note):

  * Newly implemented CORE EXT words: CASE, OF, ENDOF, and ENDCASE. Also
    added FALLTHROUGH, which works like ENDOF but jumps to the instruction
    just after the next OF.
  * Bugfix: John-Hopkins locals syntax now accepts | and -- in the comment
    (between the first -- and the }.)
  * Bugfix: Changed vmGetWord0() to make Purify happier. The resulting
    code is no slower, no larger, and slightly more robust.

17 years agoremove WARNS=2: we're not ready for it yet.
Matteo Riondato [Fri, 23 Mar 2007 22:09:02 +0000 (22:09 +0000)]
remove WARNS=2: we're not ready for it yet.

17 years ago- Increase coalesce_nsecs
Kip Macy [Fri, 23 Mar 2007 22:03:55 +0000 (22:03 +0000)]
- Increase coalesce_nsecs
- commit fixes for the following coverity warnings: 1765, 1760, 1758, 1756

17 years agocommit missed change
Kip Macy [Fri, 23 Mar 2007 22:02:11 +0000 (22:02 +0000)]
commit missed change

17 years agoIf KERNEL_EXTRA is defined, make kernel-all target depend on it.
Warner Losh [Fri, 23 Mar 2007 21:55:59 +0000 (21:55 +0000)]
If KERNEL_EXTRA is defined, make kernel-all target depend on it.
If KERNEL_EXTRA_INSTALL is defined, install it into ${DESTDIR}${KODIR}.

17 years agoExit status should be 1 on error.
Pawel Jakub Dawidek [Fri, 23 Mar 2007 21:05:36 +0000 (21:05 +0000)]
Exit status should be 1 on error.

PR: bin/110705
Reported by: Tom Judge
MFC after: 2 weeks

17 years agoCheck PCI-e link width to avoid foot shooting with 4x links
Kip Macy [Fri, 23 Mar 2007 20:18:07 +0000 (20:18 +0000)]
Check PCI-e link width to avoid foot shooting with 4x links

MFC after: 3 days

17 years agoSplit tcp_input() into its two functional parts:
Andre Oppermann [Fri, 23 Mar 2007 20:16:50 +0000 (20:16 +0000)]
Split tcp_input() into its two functional parts:

 o tcp_input() now handles TCP segment sanity checks and preparations
   including the INPCB lookup and syncache.
 o tcp_do_segment() handles all data and ACK processing and is IPv4/v6
   agnostic.

Change all KASSERT() messages to ("%s: ", __func__).

The changes in this commit are primarily of mechanical nature and no
functional changes besides the function split are made.

Discussed with: rwatson

17 years agoMake sysinstall's code WARNS=2 clean .
Matteo Riondato [Fri, 23 Mar 2007 19:57:27 +0000 (19:57 +0000)]
Make sysinstall's code WARNS=2 clean .

MFC after: 1 week

17 years agoTidy up some code to conform better to surroundings and style(9), 0 = NULL
Andre Oppermann [Fri, 23 Mar 2007 19:11:22 +0000 (19:11 +0000)]
Tidy up some code to conform better to surroundings and style(9), 0 = NULL
and space/tab.

17 years agoBring SACK option handling in tcp_dooptions() in line with all other
Andre Oppermann [Fri, 23 Mar 2007 18:33:21 +0000 (18:33 +0000)]
Bring SACK option handling in tcp_dooptions() in line with all other
options and ajust users accordingly.

17 years ago- Fix exca_(io|mem)_map() to return proper errno values.
John Baldwin [Fri, 23 Mar 2007 17:15:07 +0000 (17:15 +0000)]
- Fix exca_(io|mem)_map() to return proper errno values.
- Change exca_activate_resource() to call BUS_ACTIVATE_RESOURCE() before
  calling exca_(io|mem)_map() since the latter use rman_get_bus(tag|handle)
  and the recent changes to nexus(4) mean that you need to activate a
  resource before reading the bus tag and handle.  This was true before,
  but now the nexus(4) drivers on x86 and ia64 are more forceful about it.

Reviewed by: imp

17 years agoPurge two redundant case labels.
Bruce M Simpson [Fri, 23 Mar 2007 09:43:36 +0000 (09:43 +0000)]
Purge two redundant case labels.

17 years agoDon't destroy a mutex just before we use it, instead,
Xin LI [Fri, 23 Mar 2007 08:52:36 +0000 (08:52 +0000)]
Don't destroy a mutex just before we use it, instead,
destroy it after we have used it.

17 years agovm_page_busy() no longer requires the page queues lock to be held. Reduce
Alan Cox [Fri, 23 Mar 2007 06:11:25 +0000 (06:11 +0000)]
vm_page_busy() no longer requires the page queues lock to be held.  Reduce
the scope of the page queues lock in vm_fault() accordingly.

17 years agoUse extents rather than binary buddies to track free pages within
Jason Evans [Fri, 23 Mar 2007 05:05:48 +0000 (05:05 +0000)]
Use extents rather than binary buddies to track free pages within
chunks.  This allows runs to be any multiple of the page size.  The
primary advantage is that large objects are no longer constrained to be
2^n pages, which can dramatically decrease internal fragmentation for
large objects.  This also allows the sizes for runs that back small
objects to be more finely tuned.

Free runs are searched for linearly using the chunk page map (with the
help of some heuristic optimizations).  This changes the allocation
policy from "first best fit" to "first fit".  A prototype red-black tree
implementation for tracking free runs that implemented "first best fit"
did not cause a measurable speed or memory usage difference for
realistic chunk sizes (though of course it is possible to construct
benchmarks that favor one allocation policy over another).

Refine the handling of fullness constraints for small runs to be more
tunable.

Restructure the per chunk page map to contain only two fields per entry,
rather than four.  Also, increase each entry from 4 to 8 bytes, since it
allows for 32-bit integers, without increasing the number of chunk
header pages.

Relax the maximum chunk size constraint.  This is of no practical
interest; it is merely fallout from the chunk page map restructuring.

Revamp statistics gathering and reporting to be faster, clearer and more
informative.  Statistics gathering is fast enough now to have little
to no impact on application speed, but it still requires approximately
two extra pages of memory per arena (per process).  This memory overhead
may be acceptable for most systems, but we still need to leave
statistics gathering disabled by default in RELENG branches.

Rename NO_MALLOC_EXTRAS to MALLOC_PRODUCTION in order to make its intent
clearer (i.e. it should be defined in RELENG branches).

17 years agoStop setting ki_ocomm (thread name) to the proc name by default, as nothing
Ed Maste [Fri, 23 Mar 2007 04:01:08 +0000 (04:01 +0000)]
Stop setting ki_ocomm (thread name) to the proc name by default, as nothing
in the base system relies on this any longer.

17 years agoReflect ACPI-CA 20070320 import.
Xin LI [Fri, 23 Mar 2007 01:47:04 +0000 (01:47 +0000)]
Reflect ACPI-CA 20070320 import.

17 years agoAdd '-s' option and update the manual page. With this option, it prints
Jung-uk Kim [Fri, 23 Mar 2007 00:00:22 +0000 (00:00 +0000)]
Add '-s' option and update the manual page.  With this option, it prints
little more style(9) friendly output.  For example:

%file2c -n 8 -s -x 'const char data[] = {' '};' < /etc/motd
const char data[] = {
0x46, 0x72, 0x65, 0x65, 0x42, 0x53, 0x44, 0x20,
0x37, 0x2e, 0x30, 0x2d, 0x43, 0x55, 0x52, 0x52,
0x45, 0x4e, 0x54, 0x20, 0x28, 0x42, 0x45, 0x41,
0x53, 0x54, 0x49, 0x45, 0x29, 0x20, 0x23, 0x30,
0x3a, 0x20, 0x57, 0x65, 0x64, 0x20, 0x4d, 0x61,
0x72, 0x20, 0x32, 0x31, 0x20, 0x31, 0x39, 0x3a,
0x30, 0x34, 0x3a, 0x33, 0x36, 0x20, 0x45, 0x44,
0x54, 0x20, 0x32, 0x30, 0x30, 0x37, 0x0a
};