]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
19 years agoAdd a missing dev_relthread() call.
phk [Mon, 29 Aug 2005 11:14:18 +0000 (11:14 +0000)]
Add a missing dev_relthread() call.

Remove unused variable.

Spotted by: Hans Petter Selasky <hselasky@c2i.net>

19 years ago- Document the fact that the real length of listen queue is 1.5 more
glebius [Mon, 29 Aug 2005 10:23:46 +0000 (10:23 +0000)]
- Document the fact that the real length of listen queue is 1.5 more
  than the value of backlog argument.
- Document the fact that a subsequent listen(2) calls on the listening
  socket change the backlog argument.
- Note that current listen queue lengths can be queried using netstat(1).

Submitted by: Igor Sysoev <is rambler-co.ru>
Wording by: gnn

19 years agoFix unhandled malloc errors.
philip [Mon, 29 Aug 2005 10:08:04 +0000 (10:08 +0000)]
Fix unhandled malloc errors.

PR: bin/83467
Submitted by: Dan Lukes <dan -at- obluda.cz>
Prodded by: glebius
MFC after: 3 days

19 years agoFix support for the VIA 6421 controller.
sos [Mon, 29 Aug 2005 09:01:57 +0000 (09:01 +0000)]
Fix support for the VIA 6421 controller.

19 years agoo Discuss the ways to escape an alias.
jcamou [Mon, 29 Aug 2005 06:47:43 +0000 (06:47 +0000)]
o Discuss the ways to escape an alias.
o Fix the alias syntax lines.
o Refer to the Aliases subsection.

PR: docs/84914
Submitted by: garys
Approved by: trhodes (mentor)
MFC after: 3 days

19 years agoGlobal Village appears to be using 0x018c as its vendor ID, but this
imp [Mon, 29 Aug 2005 05:46:05 +0000 (05:46 +0000)]
Global Village appears to be using 0x018c as its vendor ID, but this
doesn't appear in the official lists, so make a note of that.

19 years agoSimplify names.
imp [Mon, 29 Aug 2005 00:05:04 +0000 (00:05 +0000)]
Simplify names.

19 years agoFinish conversion to bus space and make ed MP safe.
imp [Sun, 28 Aug 2005 23:56:25 +0000 (23:56 +0000)]
Finish conversion to bus space and make ed MP safe.

o Lock ed
o Fix extra newline in probe messages
o Eliminate gone.
o Make detach less-racy.
o Eliminate spl*
o Switch from timeout/untimeout to callout interface.
o Read/write card memory using bus_space calls.
o generalize readmem so that we don't need ifs in the code.
o Fix memory stuff to be consistant.
o Remove OLDCARD compat stuff.
o Mark interrupt as MPSAFE.

# sic, hpp not tested at all
# ISA and PCI attachments lightly tested

19 years agoFix a typo in vop_rename_pre() where we ended up using vholdl()
ssouhlal [Sun, 28 Aug 2005 23:00:11 +0000 (23:00 +0000)]
Fix a typo in vop_rename_pre() where we ended up using vholdl()
instead of vhold(), even though the vnode interlock is unlocked.

MFC after: 3 days

19 years agoFix minor typo in a comment.
bmah [Sun, 28 Aug 2005 18:48:04 +0000 (18:48 +0000)]
Fix minor typo in a comment.

19 years agoPrevent that sync operations can be started when they are already
le [Sun, 28 Aug 2005 18:16:31 +0000 (18:16 +0000)]
Prevent that sync operations can be started when they are already
in progress, and be a bit more user friendly in terms of error
messages returned from the kernel.

19 years agoAdd gem(4) to the list of devices supporting VLAN-sized frames natively
marius [Sun, 28 Aug 2005 15:08:57 +0000 (15:08 +0000)]
Add gem(4) to the list of devices supporting VLAN-sized frames natively
(since sys/dev/gem/if_gem.c rev. 1.34).

MFC after: 1 week

19 years ago- Allow for VLAN-sized frames and set IFCAP_VLAN_MTU.
marius [Sun, 28 Aug 2005 15:07:30 +0000 (15:07 +0000)]
- Allow for VLAN-sized frames and set IFCAP_VLAN_MTU.
- On resume all registers have to be initialized again like after
  power-on so reset sc_inited in gem_suspend() in order get all of
  the registers set next time gem_init_regs() is called.
- On at least some ERI and GEM revisions GEM_MAC_RX_OVERFLOW happen
  often due to a silicon bug and re-initializing is all we can do
  about these errors so make handling them non-verbose.
- Remove a superfluous memset(3) call in gem_meminit(), all elements
  are initialized to 0 anyway.

MFC after: 1 week

19 years agoFix kernel build.
delphij [Sun, 28 Aug 2005 13:11:08 +0000 (13:11 +0000)]
Fix kernel build.

Reported by: tinderbox

19 years agoRemove the el(4) device, which has been retired recently.
simon [Sun, 28 Aug 2005 08:20:41 +0000 (08:20 +0000)]
Remove the el(4) device, which has been retired recently.

19 years agoHandle vm_map_wire()'s failure.
alc [Sun, 28 Aug 2005 05:38:40 +0000 (05:38 +0000)]
Handle vm_map_wire()'s failure.

19 years agoCorrectly handle vm_map_wire()'s failure. (See also revisions 1.81 and
alc [Sun, 28 Aug 2005 04:50:11 +0000 (04:50 +0000)]
Correctly handle vm_map_wire()'s failure.  (See also revisions 1.81 and
1.82.)

Reviewed by: tegge

19 years agoI must have missed the release note for the sendmail 8.13.4 import.
gshapiro [Sun, 28 Aug 2005 04:43:48 +0000 (04:43 +0000)]
I must have missed the release note for the sendmail 8.13.4 import.

19 years agoUpdate number and size of Ports Collection.
murray [Sun, 28 Aug 2005 02:30:50 +0000 (02:30 +0000)]
Update number and size of Ports Collection.

MFC after: 1 day

19 years agoEliminate an unneeded reference on a vm object. If, in fact, the nearby
alc [Sun, 28 Aug 2005 00:24:58 +0000 (00:24 +0000)]
Eliminate an unneeded reference on a vm object.  If, in fact, the nearby
vm_map_find() fails, then the excess reference causes the vm object to be
leaked.

Reviewed by: tegge

19 years agoRevert the previous change for two reasons: (1) If vm_map_find() succeeds
alc [Sun, 28 Aug 2005 00:19:54 +0000 (00:19 +0000)]
Revert the previous change for two reasons: (1) If vm_map_find() succeeds
but vm_map_wire() fails, then a vm object, vm map entries, and kernel_map
free space is leaked and (2) unwiring is handled automatically by
vm_map_remove().

Suggested by:   tegge

19 years agoVerify length of the data to read as well.
pjd [Sun, 28 Aug 2005 00:14:21 +0000 (00:14 +0000)]
Verify length of the data to read as well.

19 years agoMake it clear that the .ko extention of a module to be loaded
keramida [Sat, 27 Aug 2005 22:25:51 +0000 (22:25 +0000)]
Make it clear that the .ko extention of a module to be loaded
is optional.

PR: docs/85356
Submitted by: Julien Gabel <jpeg@thilelli.net>
MFC after: 3 days

19 years agoCorrect a comment to match the configuration parameter it describes.
simon [Sat, 27 Aug 2005 20:57:06 +0000 (20:57 +0000)]
Correct a comment to match the configuration parameter it describes.

MFC after: 3 days

19 years agoAs with NO_CRYPT, don't try to compile geli(8) when NO_OPENSSL is defined
marius [Sat, 27 Aug 2005 20:51:12 +0000 (20:51 +0000)]
As with NO_CRYPT, don't try to compile geli(8) when NO_OPENSSL is defined
either.

MFC after: 1 week

19 years agoMFamd64 revision 1.526
alc [Sat, 27 Aug 2005 19:51:14 +0000 (19:51 +0000)]
MFamd64 revision 1.526
 When pmap_allocpte() destroys a 2/4MB "superpage" mapping it does not
 reduce the pmap's resident count accordingly.  It should.

19 years agoRemove the man page of the recently axed el(4).
marius [Sat, 27 Aug 2005 18:14:53 +0000 (18:14 +0000)]
Remove the man page of the recently axed el(4).

19 years agoRe-enable sampling on the AMD64.
jkoshy [Sat, 27 Aug 2005 16:07:12 +0000 (16:07 +0000)]
Re-enable sampling on the AMD64.

19 years ago- Special-case NMI handling on the AMD64.
jkoshy [Sat, 27 Aug 2005 16:03:40 +0000 (16:03 +0000)]
- Special-case NMI handling on the AMD64.

  On entry or exit from the kernel the 'alltraps' and 'doreti' code
  used taken by normal traps disables interrupts to protect the
  critical sections where it is setting up %gs.

  This protection is insufficient in the presence of NMIs since NMIs
  can be taken even when the processor has disabled normal interrupts.
  Thus the NMI handler needs to actually read MSR_GBASE on entry to
  the kernel to determine whether a swap of %gs using 'swapgs' is
  needed.  However, reads of MSRs are expensive and integrating this
  check into the 'alltraps'/'doreti' path would penalize normal
  interrupts.

- Teach DDB about the 'nmi_calltrap' symbol.

Reviewed by: bde, peter (older versions of this change)

19 years agoRewrite linux_ifconf() to be more like ifconf() in net/if.c
rodrigc [Sat, 27 Aug 2005 14:44:10 +0000 (14:44 +0000)]
Rewrite linux_ifconf() to be more like ifconf() in net/if.c
so that we do not call uiomove() while IFNET_RLOCK() is held.
This eliminates the witness warning:

Calling uiomove() with the following non-sleepable locks held:
exclusive sleep mutex ifnet r = 0 (0xc096dd60) locked @
/usr/src/sys/modules/linux/../../compat/linux/linux_ioctl.c:2170

MFC after: 2 days

19 years agoFix a panic in softclock() if the interface is destroyed with a bpf consumer
thompsa [Sat, 27 Aug 2005 01:17:42 +0000 (01:17 +0000)]
Fix a panic in softclock() if the interface is destroyed with a bpf consumer
attached.

This is caused by bpf_detachd clearing IFF_PROMISC on the interface which does
a SIOCSIFFLAGS ioctl. The problem here is that while the interface has been
stopped, IFF_UP has not been cleared so IFF_UP != IFF_DRV_RUNNING, this causes
the ioctl function to init() the interface which resets the callouts.

The destroy then completes and frees the softc but softclock will panic on a
dead callout pointer.

Ensure ifp->if_flags matches reality by clearing IFF_UP when we destroy.

Silence from: rwatson
Approved by: mlaier (mentor)
MFC after: 3 days

19 years agofix CardBus issue for Compaq R3000 series laptop
jkim [Fri, 26 Aug 2005 23:39:44 +0000 (23:39 +0000)]
fix CardBus issue for Compaq R3000 series laptop

Approved by: anholt (mentor)

19 years agoThe $medium string often contains quoted values with spaces in them (ssids,
brooks [Fri, 26 Aug 2005 20:31:04 +0000 (20:31 +0000)]
The $medium string often contains quoted values with spaces in them (ssids,
for example).  Follow the example of the ISC script and wrap ifconfig
calls using $medium in eval "..." so this works.

Reported by: iedowse

19 years agoAllow one to override the endian flags for make distribution. This
imp [Fri, 26 Aug 2005 18:54:06 +0000 (18:54 +0000)]
Allow one to override the endian flags for make distribution.  This
can be useful for when you know that you are doing something that
won't work with the standard settings and different settings are more
appropriate.
This allows 5.3 tools to build a 6.x userland when these
values are set to null.

19 years agoMerge tzdata2005l update from vendor branch.
wollman [Fri, 26 Aug 2005 18:46:27 +0000 (18:46 +0000)]
Merge tzdata2005l update from vendor branch.

MT6: after release
MT5: immediately

19 years agoThis commit was generated by cvs2svn to compensate for changes in r149511,
wollman [Fri, 26 Aug 2005 18:39:31 +0000 (18:39 +0000)]
This commit was generated by cvs2svn to compensate for changes in r149511,
which included commits to RCS files with non-trunk default branches.

19 years agoVendor update of timezone database.
wollman [Fri, 26 Aug 2005 18:39:31 +0000 (18:39 +0000)]
Vendor update of timezone database.

Changes to historical dates in: Azerbaijan, B.I.O.T., China, Denmark,
Libya, and Poland.

Changes to current dates in: East Timor, Haiti, Israel, Kazakhstan,
Nicaragua, Paraguay, Samoa, Tunisia, Uruguay

Changes to future dates in: Australia, Iran, United States

Obtained from: ftp://elsie.nci.nih.gov/pub/tzdata2005l.tar.gz

19 years agoClarify the description of the bus_dmasync_op_t operation flags. This has
scottl [Fri, 26 Aug 2005 18:19:01 +0000 (18:19 +0000)]
Clarify the description of the bus_dmasync_op_t operation flags.  This has
been misleading and even wrong since the import of the page.

19 years agoPretty print multipath routes.
andre [Fri, 26 Aug 2005 18:08:24 +0000 (18:08 +0000)]
Pretty print multipath routes.

Before (backslash in c syntax meaning):
 6  p16-2-0-0.r21.sttlwa01.us.bb.verio.net (129.250.2.180)  71.027 ms \
p16-1-1-3.r20.sttlwa01.us.bb.verio.net (129.250.2.6)  66.730 ms  66.535 ms
 7  xe-0-2-0.r20.sttlwa01.us.bb.verio.net (129.250.4.16)  71.092 ms \
xe-3-1.r00.sttlwa01.us.bb.verio.net (129.250.2.205)  66.598 ms \
xe-0-2-0.r20.sttlwa01.us.bb.verio.net (129.250.4.16)  71.024 ms

After:
 6  p16-2-0-0.r21.sttlwa01.us.bb.verio.net (129.250.2.180)  71.027 ms
    p16-1-1-3.r20.sttlwa01.us.bb.verio.net (129.250.2.6)  66.730 ms  66.535 ms
 7  xe-0-2-0.r20.sttlwa01.us.bb.verio.net (129.250.4.16)  71.092 ms
    xe-3-1.r00.sttlwa01.us.bb.verio.net (129.250.2.205)  66.598 ms
    xe-0-2-0.r20.sttlwa01.us.bb.verio.net (129.250.4.16)  71.024 ms

Submitted by: Richard A Steenbergen <ras at e-gerbil.net>
MFC after: 3 days

19 years agoAdd newline to debuging printf.
glebius [Fri, 26 Aug 2005 15:27:18 +0000 (15:27 +0000)]
Add newline to debuging printf.

PR: kern/85271
Submitted by: Simon Morgan

19 years ago Cleanup the reader/writer policy in netgraph(4). Assign
glebius [Fri, 26 Aug 2005 15:14:33 +0000 (15:14 +0000)]
  Cleanup the reader/writer policy in netgraph(4). Assign
either reader or writer flag on item in the function, that
allocates the item. Do not modify these flags when item is
applied or queued.
  The only exceptions are node and hook overrides - they can
change item flags to writer.

19 years agoShuffle around the order in which the components are compiled.
le [Fri, 26 Aug 2005 14:40:32 +0000 (14:40 +0000)]
Shuffle around the order in which the components are compiled.

This way, the VINUMDRIVE class is loaded before the VINUM class,
but since geom does the tasting for newly arrived classes
last-in-first-out, the VINUM class tastes first.

This removes the need to call gv_parse_config() in the drive
taste path.

19 years ago- Use m_defrag() instead of homerolling our own variant
jhb [Fri, 26 Aug 2005 14:27:38 +0000 (14:27 +0000)]
- Use m_defrag() instead of homerolling our own variant
  tulip_mbuf_compress().  If we fail to allocate a new mbuf to copy the
  data into, put the mbuf back in the driver's send queue so that we can
  retry it later rather than throwing the packet away.
- Use m_devget() instead of doing it inline ourselves in the
  TULIP_COPY_RXDATA case.  If we fail to allocate an mbuf to copy the data
  into, don't forget about the original mbuf cluster.  The old code would
  lose the pointer and leak the cluster in that case.  Now it doesn't lose
  it but always sticks the original rx buffer back into the receive ring
  after trying to copy the data out and send it up the stack.  Also, if we
  fail to allocate a new mbuf to copy the data into, log an input error.
  Also, don't combine the priming case with the received-a-packet case to
  make the code flow a bit clearer and easier to follow.

19 years agoRemove the el(4) driver for 3Com 3c501 ISA NICs from HEAD as threatened
jhb [Fri, 26 Aug 2005 13:42:04 +0000 (13:42 +0000)]
Remove the el(4) driver for 3Com 3c501 ISA NICs from HEAD as threatened
earlier as no one has stepped up to test recent changes to the driver.
Oddly, the module was actually turned on on ia64 though I'm fairly certain
that no ia64 machine has ever had or will ever have an ISA slot.

Axe borrowed from: phk

19 years agoVerify offset before reading.
pjd [Fri, 26 Aug 2005 12:50:08 +0000 (12:50 +0000)]
Verify offset before reading.

MFC after: 2 days

19 years agoOops, I forget to add item in files .
takawata [Fri, 26 Aug 2005 12:48:34 +0000 (12:48 +0000)]
Oops, I forget to add item in files .

Pointed out by: pjd

19 years agoUpdate Document.
takawata [Fri, 26 Aug 2005 11:39:38 +0000 (11:39 +0000)]
Update Document.

19 years agoAdd NTFS labeling function.
takawata [Fri, 26 Aug 2005 11:35:10 +0000 (11:35 +0000)]
Add NTFS labeling function.

Reviewed by:pjd

19 years agoTwo minor optimizations of fdalloc():
des [Fri, 26 Aug 2005 11:16:39 +0000 (11:16 +0000)]
Two minor optimizations of fdalloc():

 - if minfd < fd_freefile (as is most often the case, since minfd is
   usually 0), set it to fd_freefile.

 - remove a call to fd_first_free() which duplicates work already done
   by fdused().

This change results in a small but measurable speedup for processes
with large numbers (several thousands) of open files.

PR: kern/85176
Submitted by: Divacky Roman <xdivac02@stud.fit.vutbr.cz>
MFC after: 3 weeks

19 years agoFix a one byte off error which causes "less" to crash if you view a
tjr [Fri, 26 Aug 2005 10:05:59 +0000 (10:05 +0000)]
Fix a one byte off error which causes "less" to crash if you view a
large file, jump to the end, scroll backwards with "b" and try to search
backwards afterwards.

Submitted by: Matthias Scheler

19 years agoUnbreak tail -f on non-local filesystems.
ps [Fri, 26 Aug 2005 08:15:57 +0000 (08:15 +0000)]
Unbreak tail -f on non-local filesystems.

19 years agoRemedy the following three problems:
alc [Fri, 26 Aug 2005 05:18:46 +0000 (05:18 +0000)]
Remedy the following three problems:

1. The amd64 pmap, unlike the i386 pmap, maintains a reference count
   for each page directory (PD) page.  However, in the transformation
   of the i386 pmap into the amd64 pmap, operations, such as
   pmap_copy() and pmap_object_init_pt(), that create 2MB "superpage"
   mappings by setting the PG_PS bit in a PD entry were not modified
   to adjust the underlying PD page's reference count.  Consequently,
   superpage mappings could disappear prematurely.

2. pmap_object_init_pt() could crash or corrupt memory if either the
   virtual address range being mapped crosses a 1GB boundary in the
   virtual address space or nothing is mapped in the 1GB area.

3. When pmap_allocpte() destroys a 2MB "superpage" mapping it does not
   reduce the pmap's resident count accordingly.  It should.  (This
   bug is inherited from i386.)

Discussed with: peter
Reviewed by:    tegge

19 years agoSupport ifconfig_<ifn> variables containing quoted variables with spaces
brooks [Fri, 26 Aug 2005 04:06:17 +0000 (04:06 +0000)]
Support ifconfig_<ifn> variables containing quoted variables with spaces
in them by wrapping the ifconfig command with eval "...".

For example, this allows:

ifconfig_iwi0="DHCP ssid 'foo bar baz'"

19 years agoUnbreak compiles with ACPI_DEBUG.
kan [Fri, 26 Aug 2005 02:21:02 +0000 (02:21 +0000)]
Unbreak compiles with ACPI_DEBUG.

19 years agoIn read_string(), when the last character was a backslash, unincrement
brooks [Fri, 26 Aug 2005 01:25:59 +0000 (01:25 +0000)]
In read_string(), when the last character was a backslash, unincrement
the output index instead of keeping what ever trash was in the buffer.

Reported by: iedowse

19 years agoUse a more robust, grep-free command to get the interface of the current
brooks [Fri, 26 Aug 2005 01:07:51 +0000 (01:07 +0000)]
Use a more robust, grep-free command to get the interface of the current
default route.

Submitted by: Rostislav Krasny <rosti dot bsd at gmail dot com>

19 years agoDon't and/remove a route to our assigned IP through 127.0.0.1. It
brooks [Fri, 26 Aug 2005 01:02:38 +0000 (01:02 +0000)]
Don't and/remove a route to our assigned IP through 127.0.0.1.  It
serves no apparent purpose (we commented this out ages ago in the ISC
scripts) and cases problems with some ADSL setups.

Reported by: Rostislav Krasny <rosti dot bsd at gmail dot com>

19 years agoFor FreeBSD 4 binaries, when trying to read from a device that does
ps [Fri, 26 Aug 2005 01:00:19 +0000 (01:00 +0000)]
For FreeBSD 4 binaries, when trying to read from a device that does
not exsist, do not have ioctl return an error, but instead set -1
in the data returned to the user.  This allows the HP bios flash
utilities to work without requiring changes to their code.

Reviewed by: jhb

19 years agoClarify the confusion over Farallon cards taht differ only by arrow color
imp [Thu, 25 Aug 2005 23:11:10 +0000 (23:11 +0000)]
Clarify the confusion over Farallon cards taht differ only by arrow color

19 years ago- Remove non-bus-dma code.
jhb [Thu, 25 Aug 2005 21:06:56 +0000 (21:06 +0000)]
- Remove non-bus-dma code.
- Remove form feed characters.
- Fixup style of function declarations.
- Assume that an mbuf cluster is big enough to hold an ethernet frame.
  (This should really be using m_defrag(), but this diff is just simple
  changes for now.)

19 years agoNMI handler should not enable interrupts.
ups [Thu, 25 Aug 2005 20:33:43 +0000 (20:33 +0000)]
NMI handler should not enable interrupts.

Tested by: kris@
MFC after: 3 weeks

19 years agoAdd regression tests for '-L' option.
pjd [Thu, 25 Aug 2005 20:13:58 +0000 (20:13 +0000)]
Add regression tests for '-L' option.

19 years agoMajor rototill of this driver to add FreeBSD bus-dma support:
jhb [Thu, 25 Aug 2005 20:12:43 +0000 (20:12 +0000)]
Major rototill of this driver to add FreeBSD bus-dma support:
- Allocate arrays of metadata for the descriptors in the rx and tx rings
  and change the ring pointers to walk the metadata array rather than the
  actual descriptor rings.  Each metadata object contains a pointer to its
  descriptor, a pointer to any associated mbuf, and a pointer to the
  associated bus_dmamap_t in the bus_dma case.  The mbuf pointers replace
  the tulip_txq and tulip_rxq local ifqueue's in the softc.
- Add lots of KTR trace entries using a local KTR_TULIP level which
  defaults to 0, but can be changed to KTR_DEV at the top of the file
  when debugging.
- Rename tulip_init(), tulip_start(), tulip_ifinit(), and tulip_ifstart()
  to tulip_init_locked(), tulip_start_locked(), tulip_init(), and
  tulip_start(), respectively, to match the convention in other drivers.
- Add a TULIP_SP_MAC() macro to encode two bytes of the MAC address into
  the setup buffer and use that in place of lots of BYTE_ORDER #ifdef's.
  Also, remove an incorrect XXX comment I added earlier, the driver was
  correct (at least it does the same thing dc(4) does).  TULIP_SP_MAC
  was shamelessly copied from DC_SP_MAC() in dc(4).
- Remove the #ifdef'd NetBSD bus-dma code and replace it with FreeBSD
  bus-dma code that not only compiles but even works at runtime.
- Use callout_init_mtx() instead of just callout_init().
- Correct the various wrapper macros for bus_dmamap_sync() for the rx
  and tx buffers to only ask for the sync ops that they actually need.
- Tidy the #ifdef TULIP_COPY_RXDATA code by expanding an #ifdef a bit
  so it becomes easier to read at the expense of a couple of duplicated
  lines of code.  Also, use m_getcl() to get an mbuf cluster rather than
  MGETHDR() followed by MCLGET().
- Maintain the ring free (ri_free) count for the rx ring metadata since
  we no longer have tulip_rxq.ifq_len around to indicate how many mbuf's
  are currently in the rx ring.
- Add code to teardown bus_dma resources when attach fails and generally
  fixup attach to do a better job of cleaning up when it fails.  This
  gets us a good bit closer to possibly having a detach method someday
  and making this driver an unloadable module.
- Add some functions that can be called from ddb to dump the state of
  a descriptor ring and to dump the state of an individual descriptor.
- Various comment grammer and spelling fixes.

I have bus-dma turned on by default, but I've left the non-bus-dma code
around so that it can be turned off to aid in debugging should any problems
turn up later on.  I'll be removing the non-bus-dma code in a subsequent
commit.

19 years ago'-F' option behaviour was reverted, so revert previous commit.
pjd [Thu, 25 Aug 2005 20:11:39 +0000 (20:11 +0000)]
'-F' option behaviour was reverted, so revert previous commit.

19 years agoRevert previous behaviour of '-F' option and add '-L' option, which will
pjd [Thu, 25 Aug 2005 20:10:47 +0000 (20:10 +0000)]
Revert previous behaviour of '-F' option and add '-L' option, which will
tell pkill(1)/pgrep(1) to try to flock(2) pidfile before reading PID from
there.

Discussed with: jhb, gad

19 years agoBump __FreeBSD_version for memmem(3).
andre [Thu, 25 Aug 2005 19:49:53 +0000 (19:49 +0000)]
Bump __FreeBSD_version for memmem(3).

19 years agoAppropriate namespace protection for memmem(3).
andre [Thu, 25 Aug 2005 19:46:38 +0000 (19:46 +0000)]
Appropriate namespace protection for memmem(3).

Submitted by: wollman
MFC after: 3 days

19 years agoAdd the function memmem(3) as found in glibc and others.
andre [Thu, 25 Aug 2005 18:26:58 +0000 (18:26 +0000)]
Add the function memmem(3) as found in glibc and others.
It is the binary equivalent to strstr(3).

 void *memmem(const void *big, size_t big_len,
const void *little, size_t little_len);

Submitted by: Pascal Gloor <pascal.gloor at spale.com>
MFC after: 3 days

19 years agoAdd myself as a maintainer of Bluetooth code
emax [Thu, 25 Aug 2005 17:14:39 +0000 (17:14 +0000)]
Add myself as a maintainer of Bluetooth code

19 years agoFix a bug introduced in revision 1.9 which causes chkgrp to coredump on
cperciva [Thu, 25 Aug 2005 17:01:06 +0000 (17:01 +0000)]
Fix a bug introduced in revision 1.9 which causes chkgrp to coredump on
certain corrupt files.

Submitted by: liamfoy@dragonflybsd.org
Reported by: Nikolay Kalev, via freebsd-current
MFC after: 3 days

19 years agoMake sure ng_fec_init() uses the same calling convention as the rest of
emax [Thu, 25 Aug 2005 17:00:02 +0000 (17:00 +0000)]
Make sure ng_fec_init() uses the same calling convention as the rest of
the code, i.e. ng_fec_init() is called with the ifp->if_softc pointer and
NOT with the ifp pointer.

PR: kern/85239
Reviewed by: brooks
MFC after: 1 day

19 years agoDont set default mode to ATA_DMA_MAX on devices not capable of DMA.
sos [Thu, 25 Aug 2005 15:45:03 +0000 (15:45 +0000)]
Dont set default mode to ATA_DMA_MAX on devices not capable of DMA.

19 years agoFixes for gcc 4.x.
roberto [Thu, 25 Aug 2005 14:09:35 +0000 (14:09 +0000)]
Fixes for gcc 4.x.

PR:             bin/84995
Submitted by:   Divacky Roman <xdivac02@stud.fit.vutbr.cz>

19 years ago- Refuse hashsize of 0, since it is invalid.
glebius [Thu, 25 Aug 2005 13:57:00 +0000 (13:57 +0000)]
- Refuse hashsize of 0, since it is invalid.
- Use defined constant instead of 512.

19 years agoRemove all references to T/TCP, which was removed several months ago.
glebius [Thu, 25 Aug 2005 13:47:41 +0000 (13:47 +0000)]
Remove all references to T/TCP, which was removed several months ago.

Approved by: andre

19 years agoFixes for gcc4.x.
roberto [Thu, 25 Aug 2005 13:44:02 +0000 (13:44 +0000)]
Fixes for gcc4.x.

Submitted by:   Divacky Roman <xdivac02@stud.fit.vutbr.cz>
PR:             bin/84991

19 years agoDe-spl parts of the routing socket code now generally protected
rwatson [Thu, 25 Aug 2005 13:30:04 +0000 (13:30 +0000)]
De-spl parts of the routing socket code now generally protected
through locking; leave some spl references around code where there
are open questions about global variable references.  Also, add
an XXX regarding locking in sysctl.

MFC after: 3 days

19 years agoWhen we have a published ARP entry for some IP address, do reply on
glebius [Thu, 25 Aug 2005 13:25:57 +0000 (13:25 +0000)]
When we have a published ARP entry for some IP address, do reply on
ARP requests only on the network where this IP address belong, to.

Before this change we did replied on all interfaces. This could
lead to an IP address conflict with host we are doing ARP proxy
for.

PR: kern/75634
Reviewed by: andre

19 years agoget current temperature from _TMP during passive cooling is active.
ume [Thu, 25 Aug 2005 11:31:30 +0000 (11:31 +0000)]
get current temperature from _TMP during passive cooling is active.
it makes CPU freq transition smooth.

19 years agoinitialize only ACPI_BUFFER to avoid race condition with passive
ume [Thu, 25 Aug 2005 10:50:36 +0000 (10:50 +0000)]
initialize only ACPI_BUFFER to avoid race condition with passive
cooling thread which refers psv, tc1, tc2 and tsp.  The previous
code made the period where sc->tz_zone.tsp was zero, and it caused
panic at msleep().

Reported by: keramida
Tested by: keramida

19 years agoHook getino.3 up to the build and link it to putino.3.
trhodes [Thu, 25 Aug 2005 10:22:29 +0000 (10:22 +0000)]
Hook getino.3 up to the build and link it to putino.3.

PR: 83820

19 years agoBackout revision 1.54, because it exposes a worse problem, than
glebius [Thu, 25 Aug 2005 07:21:15 +0000 (07:21 +0000)]
Backout revision 1.54, because it exposes a worse problem, than
it fixes. I believe the problem lives somewhere outside ng_ksocket,
but until it is found, let the node be working.

PR: kern/84952
PR: kern/82413
MFC after: 3 days

19 years agoTrack all lock relationships instead of pruning direct relationships
truckman [Thu, 25 Aug 2005 03:47:37 +0000 (03:47 +0000)]
Track all lock relationships instead of pruning direct relationships
if an indirect relationship exists (keep both A->B->C and A->C).
This allows witness_checkorder() to use isitmychild() instead of
the much more expensive isitmydescendant() to check for valid lock
ordering.

Don't do an expensive tree walk to update the w_level values when
the tree is updated.  Only update the w_level values when using the
debugger to display the tree.

Nuke the experimental "witness_watch > 1" mode that only compared
w_level for the two locks.  This information is no longer maintained
at run time, and the use of isitmychild() in witness_checkorder
should bring performance close enough to the acceptable level that
this hack is not needed.

Report witness data structure allocation statistics under the
debug.witness sysctl.

Reviewed by: jhb
MFC after: 30 days

19 years ago- Remove redundant assertions that the driver lock is not held in attach()
jhb [Wed, 24 Aug 2005 20:28:56 +0000 (20:28 +0000)]
- Remove redundant assertions that the driver lock is not held in attach()
  and detach() since mtx_lock() will assert that already since the driver
  lock is not recursive.
- Move the call to callout_init_mtx() before hme_stop() so that the
  callout_stop() in hme_stop() doesn't operate on an uninitialized callout
  structure during attach.

Reported by: yongari (2)
MFC after: 3 days

19 years agoTest last minute changes, as gcc isn't too smart.
pjd [Wed, 24 Aug 2005 19:42:40 +0000 (19:42 +0000)]
Test last minute changes, as gcc isn't too smart.

Reported by: Steve Kargl <sgk@troutmask.apl.washington.edu>

19 years ago'-F' option now needs locked pidfile. Modify regression tests respectively.
pjd [Wed, 24 Aug 2005 19:39:29 +0000 (19:39 +0000)]
'-F' option now needs locked pidfile. Modify regression tests respectively.

19 years agoModify '-F' option to work nicely with pidfile(3) - a pidfile given as
pjd [Wed, 24 Aug 2005 19:38:28 +0000 (19:38 +0000)]
Modify '-F' option to work nicely with pidfile(3) - a pidfile given as
an argument has to be locked.

19 years agoUse pidfile(3) in watchdogd(8).
pjd [Wed, 24 Aug 2005 19:28:33 +0000 (19:28 +0000)]
Use pidfile(3) in watchdogd(8).

19 years agoUse pidfile(3) in mountd(8). There is no need to use /var/run/mountd.lock
pjd [Wed, 24 Aug 2005 19:17:06 +0000 (19:17 +0000)]
Use pidfile(3) in mountd(8). There is no need to use /var/run/mountd.lock
anymore.

19 years agoUse pidfile(3) in inetd(8).
pjd [Wed, 24 Aug 2005 19:04:21 +0000 (19:04 +0000)]
Use pidfile(3) in inetd(8).

19 years agoUse pidfile(3) in cron(8).
pjd [Wed, 24 Aug 2005 17:51:36 +0000 (17:51 +0000)]
Use pidfile(3) in cron(8).

Note, that when cron(8) cannot create pidfile, it'll exit. I didn't
changed this behaviour, but its better to ignore errors other than
EEXIST, so daemon can be started on systems where /var/ file system
doesn't support locking (like NFS without rpc.lockd(8)).

19 years agoAbout life
wilko [Wed, 24 Aug 2005 17:45:09 +0000 (17:45 +0000)]
About life

19 years agoAdd '-P' option which allows to specify pidfile.
pjd [Wed, 24 Aug 2005 17:32:41 +0000 (17:32 +0000)]
Add '-P' option which allows to specify pidfile.

19 years agoUse pidfile(3) in moused(8).
pjd [Wed, 24 Aug 2005 17:27:06 +0000 (17:27 +0000)]
Use pidfile(3) in moused(8).

19 years agoUse pidfile(3) in syslogd(8).
pjd [Wed, 24 Aug 2005 17:26:26 +0000 (17:26 +0000)]
Use pidfile(3) in syslogd(8).

19 years agoTeach daemon(8) how to use pidfile(3).
pjd [Wed, 24 Aug 2005 17:24:39 +0000 (17:24 +0000)]
Teach daemon(8) how to use pidfile(3).

19 years agoAdd a family of functions for reliable pidfiles handling.
pjd [Wed, 24 Aug 2005 17:21:38 +0000 (17:21 +0000)]
Add a family of functions for reliable pidfiles handling.

Idea from: jmg
Discussed on: arch@

19 years agoStop hard-coding an -M flag to mdmfs(8) in /etc/rc.subr.
yar [Wed, 24 Aug 2005 16:25:47 +0000 (16:25 +0000)]
Stop hard-coding an -M flag to mdmfs(8) in /etc/rc.subr.
Now this flag can be set, or not set, for memory-backed
file systems on individual basis, as illustrated by the
rc.conf(5) variables tmpmfs_flags and varmfs_flags.  The
flag is set for those FS'en by default, in /etc/defaults/rc.conf,
in order to stay compatible with the old rc.subr behaviour.

Submitted by: marck
MFC after: 3 days

19 years ago- Fix checking range of strings of struct iconv_add_in in libsmb and libkiconv,
imura [Wed, 24 Aug 2005 12:38:26 +0000 (12:38 +0000)]
- Fix checking range of strings of struct iconv_add_in in libsmb and libkiconv,
- Add checking range of strings to iconv_sysctl_add().

Submitted by: Rudolf Cejka

19 years agoIgnore HTTP_PROXY if it is defined but empty. This was already handled
des [Wed, 24 Aug 2005 12:28:05 +0000 (12:28 +0000)]
Ignore HTTP_PROXY if it is defined but empty.  This was already handled
correctly in the case of FTP_PROXY, because an empty FTP_PROXY has a
specific meaning ("don't use any proxy at all for ftp, even if HTTP_PROXY
is defined"), while an empty HTTP_PROXY has no meaning at all.

PR: bin/85185
Submitted by: Conall O'Brien <conallob=freebsd@maths.tcd.ie>
MFC after: 2 weeks