]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
21 years agoDocument the net.inet.ip.portrange.reserved{high,low} sysctl knobs
cjc [Fri, 21 Feb 2003 21:09:35 +0000 (21:09 +0000)]
Document the net.inet.ip.portrange.reserved{high,low} sysctl knobs
added in sys/netinet/in_pcb.c 1.120.

Prodded by: ru

21 years agoMerge the following from the English version:
hrs [Fri, 21 Feb 2003 21:06:28 +0000 (21:06 +0000)]
Merge the following from the English version:

1.496 -> 1.500 relnotes/common/new.sgml

21 years agoLet everyone know when to send the gifts.
das [Fri, 21 Feb 2003 19:25:39 +0000 (19:25 +0000)]
Let everyone know when to send the gifts.

Reviewed by: mike (mentor)

21 years agoHostname specifications must allow commas in the value. They are
thomas [Fri, 21 Feb 2003 19:02:31 +0000 (19:02 +0000)]
Hostname specifications must allow commas in the value. They are
used to separate multiple host names.

Noted by: Dan Nelson <dnelson@allantgroup.com>
Reviewed by: roberto

21 years agoNO_GEOM cleanup:
phk [Fri, 21 Feb 2003 19:00:48 +0000 (19:00 +0000)]
NO_GEOM cleanup:

Retire the "d_dump_t" and use the "dumper_t" type instead.

Dumper_t takes a void * as first arg which is more general than the
dev_t taken by d_dump_t.  (Remember: we could have net-dumpers if
somebody wrote us one!)

Define the convention for GEOM controlled disk devices to be that the
first argument to the dumper function is the struct disk pointer.

Change device drivers accordingly.

21 years agoPush UFS_ACL back for i386. We have 11kbytes of free space on kern.flp.
matusita [Fri, 21 Feb 2003 18:07:24 +0000 (18:07 +0000)]
Push UFS_ACL back for i386.  We have 11kbytes of free space on kern.flp.
If we have no UFS_ACL kernel, users who already uses UFS1/2 attributes
get confused since no access control is performed for an update install.

Still, pc98 and alpha doesn't have UFS_ACL since I don't know about them.

Nyan-san, if kern.flp on tatsu has enough spaces (4k or more spaces),
please back UFS_ACL for pc98 also.

Data collected from: 5.0-CURRENT-20030221-JPSNAP on snapshots.jp.FreeBSD.org

21 years agoSome style fixes.
bmilekic [Fri, 21 Feb 2003 16:24:49 +0000 (16:24 +0000)]
Some style fixes.

Submitted by: Hiten Pandya <hiten@unixdaemons.com>

21 years agoNO_GEOM cleanup:
phk [Fri, 21 Feb 2003 15:13:26 +0000 (15:13 +0000)]
NO_GEOM cleanup:

Change the argument to disk_destroy() to be the same struct disk * as
disk_create() takes.

This enables drivers to ignore the (now) bogus dev_t which disk_create()
returns.

21 years agoMention that dlerror() is also applicable to retrieve error message after
phantom [Fri, 21 Feb 2003 13:43:41 +0000 (13:43 +0000)]
Mention that dlerror() is also applicable to retrieve error message after
dladdr() and dlinfo() functions calls.

21 years agoRespect ``makeoptions NO_MODULES'' specified in the kernel config file.
ru [Fri, 21 Feb 2003 11:19:25 +0000 (11:19 +0000)]
Respect ``makeoptions NO_MODULES'' specified in the kernel config file.

21 years agoOverhaul the targets and hope this is now much more readable.
ru [Fri, 21 Feb 2003 11:02:49 +0000 (11:02 +0000)]
Overhaul the targets and hope this is now much more readable.

This adds some orthodox kernel-* and modules-* targets and
retires harmful ``clobber''.

Now, do we really want to keep the ``modules'' target here?

21 years agoDon't try to deal with vi.recover if the directory doesn't exist
dougb [Fri, 21 Feb 2003 09:37:20 +0000 (09:37 +0000)]
Don't try to deal with vi.recover if the directory doesn't exist

PR: conf/24515
Submitted by: Jeroen C. van Gelderen <jeroen@systemics.com>

21 years agoRemove "sleeper" nonsense.
tjr [Fri, 21 Feb 2003 08:46:44 +0000 (08:46 +0000)]
Remove "sleeper" nonsense.

21 years agoDon't omit --update on 9ball.
des [Fri, 21 Feb 2003 07:33:41 +0000 (07:33 +0000)]
Don't omit --update on 9ball.

21 years agoIf UTS kernel is calling kse_wakeup for itself, do nothing.
davidxu [Fri, 21 Feb 2003 07:11:38 +0000 (07:11 +0000)]
If UTS kernel is calling kse_wakeup for itself, do nothing.

21 years agoFix ATAPI/USB/Firewire CDROM drive handling in cd(4) and hopefully fix
ken [Fri, 21 Feb 2003 06:19:38 +0000 (06:19 +0000)]
Fix ATAPI/USB/Firewire CDROM drive handling in cd(4) and hopefully fix
a number of related problems along the way.

 - Automatically detect CDROM drives that can't handle 6 byte mode
   sense and mode select, and adjust our command size accordingly.
   We have to handle this in the cd(4) driver (where the buffers are
   allocated), since the parameter list length is different for the
   6 and 10 byte mode sense commands.

 - Remove MODE_SENSE and MODE_SELECT translation removed in ATAPICAM
   and in the umass(4) driver, since there's no way for that to work
   properly.

 - Add a quirk entry for CDROM drives that just hang when they get a 6
   byte mode sense or mode select.  The reason for the quirk must be
   documented in a PR, and all quirks must be approved by
   ken@FreeBSD.org.  This is to make sure that we fully understand why
   each quirk is needed.  Once the CAM_NEW_TRAN_CODE is finished, we
   should be able to remove any such quirks, since we'll know what
   protocol the drive speaks (SCSI, ATAPI, etc.) and therefore whether
   we should use 6 or 10 byte mode sense/select commands.

 - Change the way the da(4) handles the no_6_byte sysctl.  There is
   now a per-drive sysctl to set the minimum command size for that
   particular disk.  (Since you could have multiple disks with
   multiple requirements in one system.)

 - Loader tunable support for all the sysctls in the da(4) and cd(4)
   drivers.

 - Add a CDIOCCLOSE ioctl for cd(4) (bde pointed this out a long
   time ago).

 - Add a media validation routine (cdcheckmedia()) to the cd(4)
   driver, to fix some problems bde pointed out a long time ago.  We
   now allow open() to succeed no matter what, but if we don't detect
   valid media, the user can only issue CDIOCCLOSE or CDIOCEJECT
   ioctls.

 - The media validation routine also reads the table of contents off
   the drive.  We use the table of contents to implement the
   CDIOCPLAYTRACKS ioctl using the PLAY AUDIO MSF command.  The
   PLAY AUDIO TRACK INDEX command that we previously used was
   deprecated after SCSI-2.  It works in every SCSI CDROM I've tried,
   but doesn't seem to work on ATAPI CDROM drives.  We still use the
   play audio track index command if we don't have a valid TOC, but
   I suppose it'll fail anyway in that case.

 - Add _len() versions of scsi_mode_sense() and scsi_mode_select() so
   that we can specify the minimum command length.

 - Fix a couple of formatting problems in the sense printing code.

MFC after:  4 weeks

21 years agoThe ancient and outdated concept of "privileged ports" in UNIX-type
cjc [Fri, 21 Feb 2003 05:28:27 +0000 (05:28 +0000)]
The ancient and outdated concept of "privileged ports" in UNIX-type
OSes has probably caused more problems than it ever solved. Allow the
user to retire the old behavior by specifying their own privileged
range with,

  net.inet.ip.portrange.reservedhigh  default = IPPORT_RESERVED - 1
  net.inet.ip.portrange.reservedlo    default = 0

Now you can run that webserver without ever needing root at all. Or
just imagine, an ftpd that can really drop privileges, rather than
just set the euid, and still do PORT data transfers from 20/tcp.

Two edge cases to note,

  # sysctl net.inet.ip.portrange.reservedhigh=0

Opens all ports to everyone, and,

  # sysctl net.inet.ip.portrange.reservedhigh=65535

Locks all network activity to root only (which could actually have
been achieved before with ipfw(8), but is somewhat more
complicated).

For those who stick to the old religion that 0-1023 belong to root and
root alone, don't touch the knobs (or even lock them by raising
securelevel(8)), and nothing changes.

21 years agoSome things don't build for PowerPC yet.
obrien [Fri, 21 Feb 2003 02:30:51 +0000 (02:30 +0000)]
Some things don't build for PowerPC yet.

List from: benno

21 years agoOops, forgot to put back debug level.
simokawa [Fri, 21 Feb 2003 02:27:13 +0000 (02:27 +0000)]
Oops, forgot to put back debug level.

21 years agoDon't try to build devd when NO_CXX is set.
obrien [Fri, 21 Feb 2003 02:16:35 +0000 (02:16 +0000)]
Don't try to build devd when NO_CXX is set.

21 years agoremove unused code.
simokawa [Fri, 21 Feb 2003 02:14:00 +0000 (02:14 +0000)]
remove unused code.

21 years agoFix typo
simokawa [Fri, 21 Feb 2003 02:09:39 +0000 (02:09 +0000)]
Fix typo

Submitted by: Masahiro Ito <m_itoh@mub.biglobe.ne.jp>

21 years agoDon't scan lun by myself while boot process leave it CAM to scan the bus.
simokawa [Fri, 21 Feb 2003 02:09:04 +0000 (02:09 +0000)]
Don't scan lun by myself while boot process leave it CAM to scan the bus.
Some drives seem to be confused by simultaneous probes.

Tested by: marcel

As a side effect, logical units whose lun is greater than 0 might not be
probed correctly if the lun of 0 doesn't exist in the target because
CAM doesn't scan such luns.
I have a SCSI-FireWire bridge which maps SCSI-ID to LUN and it is an
example of such targets.

21 years agoClean up viachan_init.
orion [Fri, 21 Feb 2003 01:47:59 +0000 (01:47 +0000)]
Clean up viachan_init.

21 years agoFix assignment of record sgd_addr and clean up via8233chan_init.
orion [Fri, 21 Feb 2003 01:47:17 +0000 (01:47 +0000)]
Fix assignment of record sgd_addr and clean up via8233chan_init.

21 years agoOn detach, don't remove the child from our list of children unless it has
scottl [Thu, 20 Feb 2003 23:13:48 +0000 (23:13 +0000)]
On detach, don't remove the child from our list of children unless it has
successfully detached.

21 years agoMake camcontrol WARNS=2 clean.
johan [Thu, 20 Feb 2003 21:07:59 +0000 (21:07 +0000)]
Make camcontrol WARNS=2 clean.

Approved by: ken

21 years agoChange the console interface to pass a "struct consdev *" instead of a
phk [Thu, 20 Feb 2003 20:54:45 +0000 (20:54 +0000)]
Change the console interface to pass a "struct consdev *" instead of a
dev_t to the method functions.

The dev_t can still be found at struct consdev *->cn_dev.

Add a void *cn_arg element to struct consdev which the drivers can use
for retrieving their softc.

21 years agomdoc(7) police: tidy up.
ru [Thu, 20 Feb 2003 20:22:20 +0000 (20:22 +0000)]
mdoc(7) police: tidy up.

21 years agoInitialize sign to NULL so that we don't attempt to free() it in case
mikeh [Thu, 20 Feb 2003 20:13:07 +0000 (20:13 +0000)]
Initialize sign to NULL so that we don't attempt to free() it in case
of failure.

MFC after: 1 week

21 years agoFirst round off updates/fixes to the ATA driver.
sos [Thu, 20 Feb 2003 20:02:32 +0000 (20:02 +0000)]
First round off updates/fixes to the ATA driver.

This moves all chipset specific code to a new file 'ata-chipset.c'.
Extensive use of tables and pointers to avoid having the same switch
on chipset type in several places, and to allow substituting various
functions for different HW arch needs.
Added PIO mode setup and all DMA modes.
Support for all known SiS chipsets. Thanks to Christoph Kukulies for
sponsoring a nice ASUS P4S8X SiS648 based board for this work!

Tested on: i386, PC98, alpha and sparc64

21 years agoFirst round off updates/fixes to the ATA driver.
sos [Thu, 20 Feb 2003 19:55:45 +0000 (19:55 +0000)]
First round off updates/fixes to the ATA driver.

This moves all chipset specific code to a new file 'ata-chipset.c'.
Extensive use of tables and pointers to avoid having the same switch
on chipset type in several places, and to allow substituting various
functions for different HW arch needs.
Added PIO mode setup and all DMA modes.
Support for all known SiS chipsets. Thanks to Christoph Kukulies for
sponsoring a nice ASUS P4S8X SiS648 based board for this work!

Tested on: i386, PC98, alpha and sparc64

21 years agoRemove unused variables in the IPSEC case.
jlemon [Thu, 20 Feb 2003 18:22:21 +0000 (18:22 +0000)]
Remove unused variables in the IPSEC case.

Submitted by:  Lars Eggert <larse@ISI.EDU>

21 years agoRemove extraneous word.
ceri [Thu, 20 Feb 2003 17:32:26 +0000 (17:32 +0000)]
Remove extraneous word.

21 years agoImplement a "sndbuf_getbufaddr" function and use it instead of vtophys().
cognet [Thu, 20 Feb 2003 17:31:12 +0000 (17:31 +0000)]
Implement a "sndbuf_getbufaddr" function and use it instead of vtophys().

Reviewed by: orion

21 years agoModified release notes: OpenSSL-0.9.7a.
bmah [Thu, 20 Feb 2003 17:26:11 +0000 (17:26 +0000)]
Modified release notes:  OpenSSL-0.9.7a.

While here, fix a minor markup bogon.

21 years agoUncomment the xl(4) driver since it's now working properly
mux [Thu, 20 Feb 2003 17:08:42 +0000 (17:08 +0000)]
Uncomment the xl(4) driver since it's now working properly
on sparc64.

21 years agoAdd a dead_cdevsw which does its best to return ENXIO if at all possible.
phk [Thu, 20 Feb 2003 15:35:54 +0000 (15:35 +0000)]
Add a dead_cdevsw which does its best to return ENXIO if at all possible.

In devsw() return dead_cdevsw instead of NULL in case the dev_t does not
have a si_devsw.

This may improve our survival chances with devices which go away unexpectedly.

21 years agoRemoved extra parentheses.
ru [Thu, 20 Feb 2003 15:09:52 +0000 (15:09 +0000)]
Removed extra parentheses.

21 years agoFixed comment.
ru [Thu, 20 Feb 2003 15:05:39 +0000 (15:05 +0000)]
Fixed comment.

21 years agoRegen from syscalls.master 1.50.
tjr [Thu, 20 Feb 2003 13:34:15 +0000 (13:34 +0000)]
Regen from syscalls.master 1.50.

21 years agoMark linux_getpid(), linux_getuid() and linux_getgid() as MPSAFE.
tjr [Thu, 20 Feb 2003 13:32:48 +0000 (13:32 +0000)]
Mark linux_getpid(), linux_getuid() and linux_getgid() as MPSAFE.

21 years agoAdd M_WAITOK
ume [Thu, 20 Feb 2003 11:24:55 +0000 (11:24 +0000)]
Add M_WAITOK

21 years agoThe completion queue is no longer used, so nuke its associated code
scottl [Thu, 20 Feb 2003 08:51:16 +0000 (08:51 +0000)]
The completion queue is no longer used, so nuke its associated code
and data structures.

21 years agoAdjust code for new kse_release interface.
davidxu [Thu, 20 Feb 2003 08:24:22 +0000 (08:24 +0000)]
Adjust code for new kse_release interface.

21 years agoForgot to set KU_DOUPCALL in kse_wakeup.
davidxu [Thu, 20 Feb 2003 08:22:04 +0000 (08:22 +0000)]
Forgot to set KU_DOUPCALL in kse_wakeup.

21 years agoAdd a timeout parameter to kse_release.
davidxu [Thu, 20 Feb 2003 08:18:15 +0000 (08:18 +0000)]
Add a timeout parameter to kse_release.

21 years agoSimplify page alignment.
marcel [Thu, 20 Feb 2003 06:47:54 +0000 (06:47 +0000)]
Simplify page alignment.

21 years agoFix fumble in rev 1.525. pmap_kenter()'s second argument is a physical
peter [Thu, 20 Feb 2003 05:35:52 +0000 (05:35 +0000)]
Fix fumble in rev 1.525. pmap_kenter()'s second argument is a physical
address, not a page index.

Laughed at by:  jake

21 years agoUpdate teh aac(4) manpage to note new controllers, the scsi passthru
scottl [Thu, 20 Feb 2003 05:20:05 +0000 (05:20 +0000)]
Update teh aac(4) manpage to note new controllers, the scsi passthru
interface, changes to linux compatibilty, and fewer bugs.

21 years agoFix pathnames: BIN should be BASE for FreeBSD 5.0 and later.
bmah [Thu, 20 Feb 2003 04:32:44 +0000 (04:32 +0000)]
Fix pathnames:  BIN should be BASE for FreeBSD 5.0 and later.

Submitted by: obraun

21 years agoo Allow "buckets" in mb_alloc to be differently sized (according to
bmilekic [Thu, 20 Feb 2003 04:26:58 +0000 (04:26 +0000)]
o Allow "buckets" in mb_alloc to be differently sized (according to
  compile-time constants).  That is, a "bucket" now is not necessarily
  a page-worth of mbufs or clusters, but it is MBUF_BUCK_SZ, CLUS_BUCK_SZ
  worth of mbufs, clusters.
o Rename {mbuf,clust}_limit to {mbuf,clust}_hiwm and introduce
  {mbuf,clust}_lowm, which currently has no effect but will be used
  to set the low watermarks.
o Fix netstat so that it can deal with the differently-sized buckets
  and teach it about the low watermarks too.
o Make sure the per-cpu stats for an absent CPU has mb_active set to 0,
  explicitly.
o Get rid of the allocate refcounts from mbuf map mess.  Instead,
  just malloc() the refcounts in one shot from mbuf_init()
o Clean up / update comments in subr_mbuf.c

21 years agoRemove the PL_SHAREMOD flag from struct plimit, which could have been
tjr [Thu, 20 Feb 2003 04:18:42 +0000 (04:18 +0000)]
Remove the PL_SHAREMOD flag from struct plimit, which could have been
used to share resource limits between rfork threads, but never was.
Removing it makes resource limit locking much simpler -- only the current
process can change the contents of the structure that p_limit points to.

21 years agoAdd a "hw.syscons.bell" sysctl, which can disable the bell at syscons level.
cognet [Thu, 20 Feb 2003 03:27:09 +0000 (03:27 +0000)]
Add a "hw.syscons.bell" sysctl, which can disable the bell at syscons level.

21 years agoRemove duplicate includes.
cognet [Thu, 20 Feb 2003 03:26:11 +0000 (03:26 +0000)]
Remove duplicate includes.

Submitted by: Cyril Nguyen-Huu <cyril@ci0.org>

21 years agoMove trm(4) to the drivers floppy.
cognet [Thu, 20 Feb 2003 03:22:53 +0000 (03:22 +0000)]
Move trm(4) to the drivers floppy.
This should have been done a long time ago.

21 years agoNote we now support the DC395U2W cards.
cognet [Thu, 20 Feb 2003 03:22:15 +0000 (03:22 +0000)]
Note we now support the DC395U2W cards.

21 years agoMerge diff between rev 1.08 and rev 1.11 of Tekram driver, this notably add
cognet [Thu, 20 Feb 2003 03:21:34 +0000 (03:21 +0000)]
Merge diff between rev 1.08 and rev 1.11 of Tekram driver, this notably add
support for Tekram DC395U2W cards.
Add a fix submitted by joerg@ to correctly report some errors to CAM.
Use bus_dma instead of the remaining vtophys().

21 years agoFix a serious bug when computing the index for the
bmilekic [Thu, 20 Feb 2003 03:01:04 +0000 (03:01 +0000)]
Fix a serious bug when computing the index for the
reference counter array for mbuf clusters.  I don't know
how this got past early testing nor how it survived so long
without getting caught.  If anyone was seeing really really
bizarre memory corruption in a few mbufs this would be why.

21 years agoAdjust IRQ count for psim's OpenPIC model - it seems to be
grehan [Thu, 20 Feb 2003 01:59:42 +0000 (01:59 +0000)]
Adjust IRQ count for psim's OpenPIC model - it seems to be
off by 1.

21 years agoCatch up to latest KSE changes
grehan [Thu, 20 Feb 2003 01:57:49 +0000 (01:57 +0000)]
Catch up to latest KSE changes

21 years agoMove thread limits testing code up a bit. This let UPCALLING thread
davidxu [Thu, 20 Feb 2003 01:11:17 +0000 (01:11 +0000)]
Move thread limits testing code up a bit. This let UPCALLING thread
takes possible accumulated contexts away.

21 years agoUnbreak non-IPV6 compilation.
jlemon [Wed, 19 Feb 2003 23:43:04 +0000 (23:43 +0000)]
Unbreak non-IPV6 compilation.

Caught by: phk
Sponsored by: DARPA, NAI Labs

21 years agoFix a 64-bit bogon. The hardware command structure only has one 32 bit
scottl [Wed, 19 Feb 2003 23:33:56 +0000 (23:33 +0000)]
Fix a 64-bit bogon.  The hardware command structure only has one 32 bit
field for holding driver-dependant data.  Instead of putting the pointer
to the driver command struct in there, take advantage of these structs
being a (virtually) contiguous array and just put the array index in the
field.

21 years agoRegenerate man pages after import of OpenSSL 0.9.7a.
nectar [Wed, 19 Feb 2003 23:30:52 +0000 (23:30 +0000)]
Regenerate man pages after import of OpenSSL 0.9.7a.

21 years agoResolve conflicts after import of OpenSSL 0.9.7a.
nectar [Wed, 19 Feb 2003 23:24:16 +0000 (23:24 +0000)]
Resolve conflicts after import of OpenSSL 0.9.7a.

21 years agoThis commit was generated by cvs2svn to compensate for changes in r111147,
nectar [Wed, 19 Feb 2003 23:17:42 +0000 (23:17 +0000)]
This commit was generated by cvs2svn to compensate for changes in r111147,
which included commits to RCS files with non-trunk default branches.

21 years agoVendor import of OpenSSL 0.9.7a.
nectar [Wed, 19 Feb 2003 23:17:42 +0000 (23:17 +0000)]
Vendor import of OpenSSL 0.9.7a.

21 years agoAdd M_WAITOK
phk [Wed, 19 Feb 2003 22:51:33 +0000 (22:51 +0000)]
Add M_WAITOK

21 years agoAdd a TCP TIMEWAIT state which uses less space than a fullblown TCP
jlemon [Wed, 19 Feb 2003 22:32:43 +0000 (22:32 +0000)]
Add a TCP TIMEWAIT state which uses less space than a fullblown TCP
control block.  Allow the socket and tcpcb structures to be freed
earlier than inpcb.  Update code to understand an inp w/o a socket.

Reviewed by: hsu, silby, jayanth
Sponsored by: DARPA, NAI Labs

21 years agoConvert tcp_fillheaders(tp, ...) -> tcpip_fillheaders(inp, ...) so the
jlemon [Wed, 19 Feb 2003 22:18:06 +0000 (22:18 +0000)]
Convert tcp_fillheaders(tp, ...) -> tcpip_fillheaders(inp, ...) so the
routine does not require a tcpcb to operate.  Since we no longer keep
template mbufs around, move pseudo checksum out of this routine, and
merge it with the length update.

Sponsored by: DARPA, NAI Labs

21 years agoMake the aac driver be INTR_MPSAFE. Once the interrupt handler determines
scottl [Wed, 19 Feb 2003 21:58:34 +0000 (21:58 +0000)]
Make the aac driver be INTR_MPSAFE.  Once the interrupt handler determines
that a command completion happened, all further processing is deferred to
a taskqueue.  The taskqueue itself runs implicetely under Giant, but we
already used a taskqueue for the biodone() processing, so this at least
saves the contesting of Giant in the interrupt handler.

21 years agos/to try and retry/to retry/
ceri [Wed, 19 Feb 2003 21:48:28 +0000 (21:48 +0000)]
s/to try and retry/to retry/

PR: misc/48226
Submitted by: Gary W. Swearingen <swear@attbi.com>
MFC After: 2 days
Approved by: murray (mentor)

21 years agoo Move the cleanup of the fib maps into aac_free_commands() so as to
scottl [Wed, 19 Feb 2003 21:38:29 +0000 (21:38 +0000)]
o Move the cleanup of the fib maps into aac_free_commands() so as to
  retain symetry with aac_alloc_commans().  Since aac_alloc_commands()
  allocates fib maps and places them onto the fib lists, aac_free_commands()
  should reverse those operations.

o Combine two ifs with the same body with an ||.

o Switch from uintptr_t to uint32_t for fib map load operations.
  The target is a uint32_t so using this type for the map load call
  avoids an extra cast.  uintptr_t should only be used when you need
  an "int sized the same as the machine's poitner size" which is not
  the case here.

o Removed the commented out M_WAITOK flag in the allocation in
  aac_alloc_commands().  The kernel will only block in the allocator
  if it can grow the size of the kernel.  This usually results in a
  page-out which could involve this aac device.  Thus, sleeping here
  could deadlock the machine.  Assuming this operation is occurring outside
  of attach time, we have enough fibs to operate anyway, so waiting for
  fibs to free up is okay if not optimal.

o In aac_alloc_commands(), if we cannot dmamem_alloc additional fib
  space, free the fib map.

o In aac_alloc_commands(), if we cannot create per-command dmamaps, don't
  lose track of the fib map that is mapping all of the commands that we
  have already released into the free pool.  Instead, just cut out of
  the loop and modify aac_free_commands to not attempt to free maps that
  have not been allocated.

o Don't use a magic number when pre-allocating fibs.

o Use PAGE_SIZE to allocate in page sized chunks instead of an
  architecture specific constant.

Submitted by: gibbs

21 years agoCorrect comments.
jlemon [Wed, 19 Feb 2003 21:33:46 +0000 (21:33 +0000)]
Correct comments.

21 years agoClean up delayed acks and T/TCP interactions:
jlemon [Wed, 19 Feb 2003 21:18:23 +0000 (21:18 +0000)]
Clean up delayed acks and T/TCP interactions:
   - delay acks for T/TCP regardless of delack setting
   - fix bug where a single pass through tcp_input might not delay acks
   - use callout_active() instead of callout_pending()

Sponsored by: DARPA, NAI Labs

21 years ago#include "opt_cpu.h" so we notice our options.
phk [Wed, 19 Feb 2003 20:50:02 +0000 (20:50 +0000)]
#include "opt_cpu.h" so we notice our options.

21 years agoFix definite panic when releasing interrupt resources.
jake [Wed, 19 Feb 2003 19:40:40 +0000 (19:40 +0000)]
Fix definite panic when releasing interrupt resources.

Spotted by: tmm

21 years agoFix panic on sparc64 introduced in my last commit. I really
mux [Wed, 19 Feb 2003 18:33:29 +0000 (18:33 +0000)]
Fix panic on sparc64 introduced in my last commit.  I really
wish the busdma APIs were more consistent accross architectures.

We should probably move all the other DMA map creations in
xl_attach() where we can really handle them failing, since
xl_init() is void and shouldn't fail.

Pointy hat to: mux
Tested by: Anders Andersson <anders@hack.org>

21 years agoMissed a missing M_WAITOK.
jake [Wed, 19 Feb 2003 17:29:07 +0000 (17:29 +0000)]
Missed a missing M_WAITOK.

21 years agoUse M_WAITOK.
jake [Wed, 19 Feb 2003 17:25:58 +0000 (17:25 +0000)]
Use M_WAITOK.

21 years agoDon't spam sys/${MACHINE}/conf/ with _.${MACHINE}.makeLINT.
ru [Wed, 19 Feb 2003 16:57:12 +0000 (16:57 +0000)]
Don't spam sys/${MACHINE}/conf/ with _.${MACHINE}.makeLINT.

Submitted by: johan

21 years agoWe can simplify this Makefile down to a single line now.
obrien [Wed, 19 Feb 2003 16:56:30 +0000 (16:56 +0000)]
We can simplify this Makefile down to a single line now.

21 years agoFixed universe.
ru [Wed, 19 Feb 2003 15:40:19 +0000 (15:40 +0000)]
Fixed universe.

Folded pc98 into the common case.
Retired ${JFLAG} (``make -jX universe'' should work).

21 years agoBack out the removal (here too) of the "custom" version of
ru [Wed, 19 Feb 2003 15:25:07 +0000 (15:25 +0000)]
Back out the removal (here too) of the "custom" version of
<sys/endian.h>.  It is needed to cross-build sparc64 on
RELENG_4 and to build sparc64 on say 5.0-DP1 (on systems
without <sys/endian.h>).  This will be revisited when we
create RELENG_5.

Spotted by: make universe

21 years agoCount non-threaded group.
davidxu [Wed, 19 Feb 2003 13:40:24 +0000 (13:40 +0000)]
Count non-threaded group.

21 years agoUpdate comments to reflect new KSE code.
davidxu [Wed, 19 Feb 2003 13:36:51 +0000 (13:36 +0000)]
Update comments to reflect new KSE code.

21 years agoDo not call smbfs_attr_cacheremove() in the EXDEV case in smbfs_rename().
tjr [Wed, 19 Feb 2003 11:54:35 +0000 (11:54 +0000)]
Do not call smbfs_attr_cacheremove() in the EXDEV case in smbfs_rename().
One of the vnodes is on different mount and is possibly on a different
kind of filesystem; treating it as an smbfs vnode then writing to it
will probably corrupt it.

PR: 48381
MFC after: 1 month

21 years agoThe "m = m->m_next" that was removed in the revision 1.12 was necessary
tjr [Wed, 19 Feb 2003 10:12:42 +0000 (10:12 +0000)]
The "m = m->m_next" that was removed in the revision 1.12 was necessary
for the m->m_next != NULL case to avoid looping infinitely when the first
mbuf in the chain becomes full.

21 years agoM_WAITOK and remove an useless comment.
davidxu [Wed, 19 Feb 2003 09:59:12 +0000 (09:59 +0000)]
M_WAITOK and remove an useless comment.

21 years agoMunge the zs sbus attachment to work with fhc. This is a bit of a hack,
jake [Wed, 19 Feb 2003 08:29:17 +0000 (08:29 +0000)]
Munge the zs sbus attachment to work with fhc.  This is a bit of a hack,
but they would otherwise be almost identical so there's not much point in
splitting it up further.

21 years agoImplement interrupt resource allocation and setup. Set the interrupt
jake [Wed, 19 Feb 2003 08:23:38 +0000 (08:23 +0000)]
Implement interrupt resource allocation and setup.  Set the interrupt
group number properly based on the board id.  Perform dummy reads of
registers after writing to flush the hardware write buffers.

This gets the soon to be committed zs attachment working.

21 years agoEliminate unused KSE symbols.
davidxu [Wed, 19 Feb 2003 06:59:00 +0000 (06:59 +0000)]
Eliminate unused KSE symbols.

21 years agoDon't tell people to include <sys/dkstat.h> any more.
phk [Wed, 19 Feb 2003 06:33:37 +0000 (06:33 +0000)]
Don't tell people to include <sys/dkstat.h> any more.

Pointed out by: kuriyama

21 years agoBump __FreeBSD_version to mark the fact that <sys/dkstat.h> should no longer
phk [Wed, 19 Feb 2003 06:32:52 +0000 (06:32 +0000)]
Bump __FreeBSD_version to mark the fact that <sys/dkstat.h> should no longer
be included.

21 years agoBack out M_* changes, per decision of the TRB.
imp [Wed, 19 Feb 2003 05:47:46 +0000 (05:47 +0000)]
Back out M_* changes, per decision of the TRB.

Approved by: trb

21 years agoFix the description for mkdtemp(), which creates directories, not files.
gshapiro [Wed, 19 Feb 2003 04:40:30 +0000 (04:40 +0000)]
Fix the description for mkdtemp(), which creates directories, not files.

Submitted by: Murray S. Kucherawy <msk@blackops.org>
X-MFC after: re approval

21 years agoUpdate for GCC 3.2.2
kan [Wed, 19 Feb 2003 04:28:12 +0000 (04:28 +0000)]
Update for GCC 3.2.2

21 years agoOptimize the case when max threads number was hit.
davidxu [Wed, 19 Feb 2003 04:01:55 +0000 (04:01 +0000)]
Optimize the case when max threads number was hit.

21 years agoPushing back aacp driver to kern.flp kernel. It can be removed iff
matusita [Wed, 19 Feb 2003 03:49:17 +0000 (03:49 +0000)]
Pushing back aacp driver to kern.flp kernel.  It can be removed iff
aac is not compiled into a kernel.

After this change, kgziped kernel got about 320 bytes; it still fits a floppy.

Pointy hat to: matusita