]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
21 years agoThis commit was generated by cvs2svn to compensate for changes in r118624,
Nate Lawson [Thu, 7 Aug 2003 18:19:12 +0000 (18:19 +0000)]
This commit was generated by cvs2svn to compensate for changes in r118624,
which included commits to RCS files with non-trunk default branches.

21 years agoFix a bug that would result in a very long sleep. Change has been
Nate Lawson [Thu, 7 Aug 2003 18:19:12 +0000 (18:19 +0000)]
Fix a bug that would result in a very long sleep.  Change has been
submitted to the vendor.

Obtained from: iwasaki

21 years agoNew PIM header files.
Jeffrey Hsu [Thu, 7 Aug 2003 18:17:43 +0000 (18:17 +0000)]
New PIM header files.

Submitted by: Pavlin Radoslavov <pavlin@icir.org>

21 years ago1. Basic PIM kernel support
Jeffrey Hsu [Thu, 7 Aug 2003 18:16:59 +0000 (18:16 +0000)]
1. Basic PIM kernel support
Disabled by default. To enable it, the new "options PIM" must be
added to the kernel configuration file (in addition to MROUTING):

options MROUTING # Multicast routing
options PIM # Protocol Independent Multicast

2. Add support for advanced multicast API setup/configuration and
extensibility.

3. Add support for kernel-level PIM Register encapsulation.
Disabled by default.  Can be enabled by the advanced multicast API.

4. Implement a mechanism for "multicast bandwidth monitoring and upcalls".

Submitted by: Pavlin Radoslavov <pavlin@icir.org>

21 years agoAdd man pages for acpidb and iasl
Nate Lawson [Thu, 7 Aug 2003 18:09:17 +0000 (18:09 +0000)]
Add man pages for acpidb and iasl

21 years agotd_dupfd just needs to be less than 0, it does not have to hold the
John Baldwin [Thu, 7 Aug 2003 17:08:26 +0000 (17:08 +0000)]
td_dupfd just needs to be less than 0, it does not have to hold the
negative value of the index of the new file, so just use -1.

21 years agoImport the userland interface from acpica 0619
Nate Lawson [Thu, 7 Aug 2003 16:59:44 +0000 (16:59 +0000)]
Import the userland interface from acpica 0619

21 years agoThis commit was generated by cvs2svn to compensate for changes in r118617,
Nate Lawson [Thu, 7 Aug 2003 16:59:44 +0000 (16:59 +0000)]
This commit was generated by cvs2svn to compensate for changes in r118617,
which included commits to RCS files with non-trunk default branches.

21 years agoAdd the acpidb and iasl utilities. This is not yet hooked up to the build.
Nate Lawson [Thu, 7 Aug 2003 16:51:50 +0000 (16:51 +0000)]
Add the acpidb and iasl utilities.  This is not yet hooked up to the build.

Obtained from: acpicatools port by Iwasaki-san

21 years agoUpdate some argument-documenting comments to match reality.
Jacques Vidrine [Thu, 7 Aug 2003 16:42:27 +0000 (16:42 +0000)]
Update some argument-documenting comments to match reality.

Add an explicit range check to those same arguments to reduce risk of
cardiac arrest in future code readers.

21 years agoImport common/ files from acpica 0619
Nate Lawson [Thu, 7 Aug 2003 16:38:54 +0000 (16:38 +0000)]
Import common/ files from acpica 0619

21 years agoThis commit was generated by cvs2svn to compensate for changes in r118613,
Nate Lawson [Thu, 7 Aug 2003 16:38:54 +0000 (16:38 +0000)]
This commit was generated by cvs2svn to compensate for changes in r118613,
which included commits to RCS files with non-trunk default branches.

21 years agoImport compiler/ files from acpica 0619
Nate Lawson [Thu, 7 Aug 2003 16:38:31 +0000 (16:38 +0000)]
Import compiler/ files from acpica 0619

21 years agoThis commit was generated by cvs2svn to compensate for changes in r118611,
Nate Lawson [Thu, 7 Aug 2003 16:38:31 +0000 (16:38 +0000)]
This commit was generated by cvs2svn to compensate for changes in r118611,
which included commits to RCS files with non-trunk default branches.

21 years agoRemove the user include of ctype.h and instead include it for each utility
Nate Lawson [Thu, 7 Aug 2003 15:52:42 +0000 (15:52 +0000)]
Remove the user include of ctype.h and instead include it for each utility
that uses strupr().

21 years agoRemove nowerror from some files that no longer generate warnings.
Nate Lawson [Thu, 7 Aug 2003 15:51:35 +0000 (15:51 +0000)]
Remove nowerror from some files that no longer generate warnings.

21 years agoConsistently use the BSD u_int and u_short instead of the SYSV uint and
John Baldwin [Thu, 7 Aug 2003 15:04:27 +0000 (15:04 +0000)]
Consistently use the BSD u_int and u_short instead of the SYSV uint and
ushort.  In most of these files, there was a mixture of both styles and
this change just makes them self-consistent.

Requested by: bde (kern_ktrace.c)

21 years agoAdd all interface polling patches.
Martin Blapp [Thu, 7 Aug 2003 15:00:55 +0000 (15:00 +0000)]
Add all interface polling patches.

21 years agoDo not poll during discover. Add a polling tunable to the interface
Martin Blapp [Thu, 7 Aug 2003 14:58:46 +0000 (14:58 +0000)]
Do not poll during discover. Add a polling tunable to the interface
struct to be able to turn polling on/off.

Tested by: Larry Rosenman <ler@lerctr.org>

21 years agoNull terminate the OEM hint. This rids my laptop of the smiley face that
Nate Lawson [Thu, 7 Aug 2003 14:53:14 +0000 (14:53 +0000)]
Null terminate the OEM hint.  This rids my laptop of the smiley face that
would follow the 6 valid chars of the table entry.

21 years agoAdd additional documentation to setfacl(1) regarding the behavior of
Robert Watson [Thu, 7 Aug 2003 14:52:17 +0000 (14:52 +0000)]
Add additional documentation to setfacl(1) regarding the behavior of
tools such as chmod(1) and ls(1) when it comes to acting on objects
that have POSIX.1e extended ACLs.  Specifically, discuss the
substitution of the mask entry for the group entry in the mode
representation of the ACL.  Differently worded from the submission,
and could probably use further refinement.

PR: 55319
Submitted by: Grzegorz Czaplinski <G.Czaplinski@prioris.mini.pw.edu.pl>

21 years ago-v no longer a valid argument to setfacl(1) -- remove from usage().
Robert Watson [Thu, 7 Aug 2003 14:43:43 +0000 (14:43 +0000)]
-v no longer a valid argument to setfacl(1) -- remove from usage().

PR: 55318
Submitted by: Grzegorz Czaplinski <G.Czaplinski@prioris.mini.pw.edu.pl>

21 years agoMake open channels persist across ifconfig down and up. All channels
Hartmut Brandt [Thu, 7 Aug 2003 14:30:58 +0000 (14:30 +0000)]
Make open channels persist across ifconfig down and up. All channels
that are not currently closing when the interface is configured down
will be brough up as soon as the interface is configured up.

21 years agoComment out ctype.h. It's not ok to always include in the !_KERNEL case
Nate Lawson [Thu, 7 Aug 2003 14:17:23 +0000 (14:17 +0000)]
Comment out ctype.h.  It's not ok to always include in the !_KERNEL case
since the bootblocks have their own local includes.  Sorry for breaking
world.

21 years agoThe ktrace mutex does not need to be locked around the post of the ktrace
John Baldwin [Thu, 7 Aug 2003 13:58:13 +0000 (13:58 +0000)]
The ktrace mutex does not need to be locked around the post of the ktrace
semaphore and doing so can lead to a possible reversal.  WITNESS would have
caught this if semaphores were used more often in the kernel.

Submitted by: Ted Unangst <tedu@stanford.edu>, Dawson Engler

21 years agoMake the driver preserve open connections accross ifconfig down
Hartmut Brandt [Thu, 7 Aug 2003 13:42:31 +0000 (13:42 +0000)]
Make the driver preserve open connections accross ifconfig down
and up commands. When configuring the interface down only the
connections that are currently closing are deleted from the connection
table. When the interface is configured up, all connections that
are in the table are re-opened.

21 years agoWhen configuring the interface down and up again try to re-open all
Hartmut Brandt [Thu, 7 Aug 2003 10:40:24 +0000 (10:40 +0000)]
When configuring the interface down and up again try to re-open all
connections that have been open (and were not closing) when
the interface was stopped. This makes the behaviour of fatm(4) more like
the behaviour of en(4).

21 years agoImplement mblen(s, n) as mbtowc(NULL, s, n) to avoid calling sgetrune()
Tim J. Robbins [Thu, 7 Aug 2003 09:34:51 +0000 (09:34 +0000)]
Implement mblen(s, n) as mbtowc(NULL, s, n) to avoid calling sgetrune()
and to simplify things. This is only valid until we start supporting
state-dependent encodings.

21 years agoWe donot need `\n' for panic().
Noriaki Mitsunaga [Thu, 7 Aug 2003 08:13:37 +0000 (08:13 +0000)]
We donot need `\n' for panic().

21 years agoImplement mbstowcs() as a wrapper around mbsrtowcs(), and wcstombs()
Tim J. Robbins [Thu, 7 Aug 2003 08:04:01 +0000 (08:04 +0000)]
Implement mbstowcs() as a wrapper around mbsrtowcs(), and wcstombs()
as a wrapper around wcsrtombs().

21 years agoGrok async contexts. When a thread is interrupted and an upcall
Marcel Moolenaar [Thu, 7 Aug 2003 08:03:05 +0000 (08:03 +0000)]
Grok async contexts. When a thread is interrupted and an upcall
happens, the context of the interrupted thread is exported to
userland. Unlike most contexts, it will be an async context and
we cannot easily use our existing functions to set such a
context.
To avoid a lot of complexity that may possibly interfere with
the common case, we simply let the kernel deal with it. However,
we don't use the EPC based syscall path to invoke setcontext(2).
No, we use the break-based syscall path. That way the trapframe
will be compatible with the context we're trying to restore and
we save the kernel a lot of trouble. The kind of trouble we did
not want to go though ourselves...

However, we also need to set the threads mailbox and there's no
syscall to help us out. To avoid creating a new syscall, we use
the context itself to pass the information to the kernel so that
the kernel can update the mailbox. This involves setting a flag
(_MC_FLAGS_KSE_SET_MBOX) and setting ifa (the address) and isr
(the value).

21 years agoImplement mbtowc() in terms of mbrtowc(), and wctomb() in terms of wcrtomb().
Tim J. Robbins [Thu, 7 Aug 2003 07:59:36 +0000 (07:59 +0000)]
Implement mbtowc() in terms of mbrtowc(), and wctomb() in terms of wcrtomb().

21 years agoBetter define the flags in the mcontext_t and properly set the flags
Marcel Moolenaar [Thu, 7 Aug 2003 07:52:39 +0000 (07:52 +0000)]
Better define the flags in the mcontext_t and properly set the flags
when we create contexts. The meaning of the flags are documented in
<machine/ucontext.h>. I only list them here to help browsing the
commit logs:
_MC_FLAGS_ASYNC_CONTEXT
_MC_FLAGS_HIGHFP_VALID
_MC_FLAGS_KSE_SET_MBOX
_MC_FLAGS_RETURN_VALID
_MC_FLAGS_SCRATCH_VALID

Yes, _MC_FLAGS_KSE_SET_MBOX is a hack and I'm proud of it :-)

21 years agoImplement btowc() in terms of mbrtowc() instead of sgetrune(), and
Tim J. Robbins [Thu, 7 Aug 2003 07:45:35 +0000 (07:45 +0000)]
Implement btowc() in terms of mbrtowc() instead of sgetrune(), and
wctob() in terms of wcrtomb() instead of sputrune(). There should be
no functional differences, but there may be a small performance hit
because we make an extra function call.

The aim here is to have as few functions as possible calling
s{get,put}rune() to make it easier to remove them in the future.

21 years agoo Fix cut-n-paste whitespace corruption in previous commit
Marcel Moolenaar [Thu, 7 Aug 2003 07:40:19 +0000 (07:40 +0000)]
o  Fix cut-n-paste whitespace corruption in previous commit
o  For trap-based upcalls the argument (the kse_mailbox) to
   the UTS must be written onto the kernel stack, not the
   user stack. While here, deal with the fact that we may
   be at a NaT collection point.

21 years ago- Fix the polling code to work with media settings in dhclient.conf.
Martin Blapp [Thu, 7 Aug 2003 07:27:14 +0000 (07:27 +0000)]
- Fix the polling code to work with media settings in dhclient.conf.

- Rename linkstatus to linkstate which is more correct.

- Sort any leases each time we loose and regain link. With the first
  version we spammed the dhclient.leases file.

- When there was no link on a interface, polling has been done twice.
  This has been fixed with a bigger sleep interval.

- interface_active() now returns only a assumed state of the link.
  If we are using media settings, we cannot be sure that the link
  actually works. That means we assume that the link is working and
  continue to send requests as usual on the interface and loop over
  all possible media options. Polling is still done. If your interface
  gets suddenly link, dhclient will send a dhcp discover request.

After all media settings have been tried a few times, dhclient will
sleep. If one does change networks very often, it can help to set the
"retry timeout" to a low value like 100 seconds. The default is over
six minutes.

Tested by: Larry Rosenman <ler@lerctr.org>, imp

21 years agoAdd preliminary support for the RealTek 8169 gigE chip. Changes:
Bill Paul [Thu, 7 Aug 2003 07:00:30 +0000 (07:00 +0000)]
Add preliminary support for the RealTek 8169 gigE chip. Changes:

- Fix a bug in rl_dma_map_desc(): set the 'end of ring' bit in the
  right descriptor (DESC_CNT - 1, not DESC_CNT). The 8139C+ is limited
  to 64 descriptors and automatically wraps at 64 descriptors even
  if the EOR bit isn't set, but the 8169 NIC can have up to 1024
  descriptors per ring, so we must set the wrap point in the right
  place.

- RealTek moved the RL_TIMERINT register from offset 0x54 to 0x58 in
  the 8169 -- account for this.

- Added rl_gmii_readreg() and rl_gmii_writereg() routines.

- Fix rl_probe() to deal with the case where the base type is
  not RL_8139.

The next step is to add jumbo buffer support.

Tested with the Xterasys XN-152 NIC (hard to beat $29 for a gigE NIC).

21 years agoNo need to define externs for getopt interface.
Warner Losh [Thu, 7 Aug 2003 05:47:49 +0000 (05:47 +0000)]
No need to define externs for getopt interface.

21 years agoPrefer PATH_MAX to MAXPATHLEN. Both contain the trailing NUL, so
Warner Losh [Thu, 7 Aug 2003 05:38:56 +0000 (05:38 +0000)]
Prefer PATH_MAX to MAXPATHLEN.  Both contain the trailing NUL, so
remove the unneeded +1.

21 years agoNo need to define optind as an extern. stdlib.h does that for us.
Warner Losh [Thu, 7 Aug 2003 04:53:48 +0000 (04:53 +0000)]
No need to define optind as an extern.  stdlib.h does that for us.

21 years agoPrefer PATH_MAX to MAXPATHLEN. PATH_MAX has the trailing NUL.
Warner Losh [Thu, 7 Aug 2003 04:51:41 +0000 (04:51 +0000)]
Prefer PATH_MAX to MAXPATHLEN.  PATH_MAX has the trailing NUL.

21 years agoMAXPATHLEN includes the trailing NUL, so no need to add 1 here.
Warner Losh [Thu, 7 Aug 2003 04:50:29 +0000 (04:50 +0000)]
MAXPATHLEN includes the trailing NUL, so no need to add 1 here.

21 years agoDocument hw.acpi.os_name. Also update the credits section.
Nate Lawson [Thu, 7 Aug 2003 04:46:53 +0000 (04:46 +0000)]
Document hw.acpi.os_name.  Also update the credits section.

21 years agoAdd includes to fix user-compilation of acpica. Also clarify a comment
Nate Lawson [Thu, 7 Aug 2003 04:46:17 +0000 (04:46 +0000)]
Add includes to fix user-compilation of acpica.  Also clarify a comment
about overriding the OS name.

21 years agogetopt returns -1 not EOF at the end of args. Compare against that.
Warner Losh [Thu, 7 Aug 2003 04:40:54 +0000 (04:40 +0000)]
getopt returns -1 not EOF at the end of args.  Compare against that.

21 years agoAdd entries for NETGEAR MA401RA, IO DATA PCET10CL and Panasonic KXL-CB10AN.
Warner Losh [Thu, 7 Aug 2003 04:38:01 +0000 (04:38 +0000)]
Add entries for NETGEAR MA401RA, IO DATA PCET10CL and Panasonic KXL-CB10AN.

Appologies to: those that submitted these to me.

21 years agoRemove trailing newlines (from the right branch this time)
Warner Losh [Thu, 7 Aug 2003 04:33:47 +0000 (04:33 +0000)]
Remove trailing newlines (from the right branch this time)

21 years ago - Remove GIANT_REQUIRED from pipe_free_kmem().
Alan Cox [Thu, 7 Aug 2003 04:32:40 +0000 (04:32 +0000)]
 - Remove GIANT_REQUIRED from pipe_free_kmem().
 - Remove the acquisition and release of Giant around pipe_kmem_free() and
   uma_zfree() in pipeclose().

21 years agoIncrease SMALL_PIPE_SIZE from 4096 to PAGE_SIZE.
Mike Silbersack [Thu, 7 Aug 2003 00:03:29 +0000 (00:03 +0000)]
Increase SMALL_PIPE_SIZE from 4096 to PAGE_SIZE.

Sorry people with 8K pages, this could be have been causing some of the
resent panics under heavy pipe load.

21 years agoIn cpu_set_upcall_kse(), create the upcall according to the entry
Marcel Moolenaar [Wed, 6 Aug 2003 23:28:19 +0000 (23:28 +0000)]
In cpu_set_upcall_kse(), create the upcall according to the entry
path into the kernel. Normally it's due to a syscall, but one can
also be created as the result of a clock interrupt (for example).
This now even more looks like exec_setregs().

While here, add an assert that we don't expect more than 8KB of
dirty registers on the kernel stack.

21 years agoAdd a reference to config(5) in the SEE ALSO section.
Simon L. B. Nielsen [Wed, 6 Aug 2003 21:41:35 +0000 (21:41 +0000)]
Add a reference to config(5) in the SEE ALSO section.

Suggested by: dcs
Approved by: ceri (mentor)

21 years agoo In revision 1.45 of exception.S we changed exception_restore to
Marcel Moolenaar [Wed, 6 Aug 2003 21:32:38 +0000 (21:32 +0000)]
o  In revision 1.45 of exception.S we changed exception_restore to
   unconditionally restore ar.k7 (kernel memory stack) and ar.k6
   (kernel register stack). I don't know what I was smoking then,
   but if you unconditionally restore ar.k6, you also want to
   compute its value unconditionally. By having the computation
   predicated and dependent on whether we return to user mode, we
   would end up writing junk (= invalid value for ar.bspstore) if
   we would return to kernel mode. But the whole point of the
   unconditional restoration was that there is a grey area where
   we still need to have ar.k6 restored. If we restore with a junk
   value, we would end up wedging the machine on the next interrupt.
   So, unconditionally calculate the value we unconditionally write
   to ar.k6.

o  The previous braino was found while making the following change:
   We used to clear the lower 9 bits of the value we write to ar.k6.
   The meaning being that we know that the kernel register stack is
   at least 512 byte aligned and simply clearing the lower 9 bits
   allows us to return to a context of which we don't have dirty
   registers on the kernel stack, even though the context that
   entered the kernel does have dirty registers on the kernel stack.
   By masking-off the lower bits, we correctly obtain the base of
   the register stack without having to worry that we didn't actually
   reached the base while unwinding it.
   The change is to mask off the lower 13 bits, knowing that the
   kernel register stack is always 8KB aligned. The advantage is that
   we don't have to worry anymore if there's more than 512 bytes of
   dirty registers on the kernel stack. A situation that frequently
   occurs. In exec_setregs() in machdep.c:1.147 or older, we had to
   deal with that situation by copying the active portion of the
   register stack down in multiples of 512 bytes. Now that we mask off
   the lower 13 bits we don't have to do that at all. Contemporary
   IPF processors have a register file that can hold up to 96 stacked
   registers (=784 bytes [incl. 2 NaT collections]). With no indication
   that register files grow beyond a couple of hundred registers, we
   should not have to worry about it anymore... and yes, 640KB is
   enough for everybody :-)
   This change helps setcontext(2) and cpu_set_upcall_kse() in that
   they can return to completely different contexts without having to
   mess with the kernel stack. Of course exec_setregs() doesn't need
   to do that anymore as well.

21 years agoCorrect a mistake in the previous revision: Reduce the scope of the page
Alan Cox [Wed, 6 Aug 2003 19:18:08 +0000 (19:18 +0000)]
Correct a mistake in the previous revision: Reduce the scope of the page
queues lock such that it isn't held around the call to get_pv_entry(),
which calls uma_zalloc().  At the point of the call to get_pv_entry(), the
lock isn't necessary and holding it could lead to recursive acquisition,
which isn't allowed.

21 years agoAcquire the page queues lock in pmap_insert_entry(). (I used to believe
Alan Cox [Wed, 6 Aug 2003 18:46:47 +0000 (18:46 +0000)]
Acquire the page queues lock in pmap_insert_entry().  (I used to believe
that the page's busy flag could be relied upon to synchronize access to the
pv list.  I don't any longer.  See, for example, the call to
pmap_insert_entry() from pmap_copy().)

21 years agoRemove dangling extern reference to swap_pager_full
Poul-Henning Kamp [Wed, 6 Aug 2003 18:40:49 +0000 (18:40 +0000)]
Remove dangling extern reference to swap_pager_full

21 years agoBacked out previous commit. This restores the warning about pessimized
Bruce Evans [Wed, 6 Aug 2003 18:21:27 +0000 (18:21 +0000)]
Backed out previous commit.  This restores the warning about pessimized
(short) types for the port arg of inb() (rev.1.56).  The warning started
working for u_short types with gcc-3.3.  The pessimizations exposed
by this been fixed except for the cx and oltr drivers where the breakage
of the warning has been pushed to the drivers.

21 years agoRFC 3561 (ad-hoc on-demand distance vector routing) was ratified
Daniel C. Sobral [Wed, 6 Aug 2003 18:15:10 +0000 (18:15 +0000)]
RFC 3561 (ad-hoc on-demand distance vector routing) was ratified
and the ports assigned by IANA.

PR: conf/55125
Submitted by: Bruce M Simpson <bms@spc.org>
MFC after: 2 days

21 years agoDon't use pessimal (u_short) types for i/o ports. This is mainly for
Bruce Evans [Wed, 6 Aug 2003 18:06:30 +0000 (18:06 +0000)]
Don't use pessimal (u_short) types for i/o ports.  This is mainly for
completenss.  The pessimization is tiny compared with i/o port slowness
except on very old machines, but code that used signed short types for
i/o ports was unpessimized long ago, and the macro that detected it
recently started working for u_short types too.  Use of bus space
should have made this moot long ago.

Not tested at runtime by:  bde

21 years agoRemove all USB quirks that only specify DA_Q_NO_6_BYTE. They are no longer
Nate Lawson [Wed, 6 Aug 2003 17:30:03 +0000 (17:30 +0000)]
Remove all USB quirks that only specify DA_Q_NO_6_BYTE.  They are no longer
needed.  Other USB quirks remain in hopes that user testing will reveal
which ones are also no longer needed.

Courtesy of: USENIX lunch break

21 years agoDon't use pessimal (u_short) types for i/o ports. This is mainly for
Bruce Evans [Wed, 6 Aug 2003 17:28:33 +0000 (17:28 +0000)]
Don't use pessimal (u_short) types for i/o ports.  This is mainly for
completenss.  The pessimization is tiny compared with i/o port slowness
except on very old machines, but code that used signed short types for
i/o ports was unpessimized long ago, and the macro that detected it
recently started working for u_short types too.  Use of bus space
should have made this moot long ago.

Not tested at runtime by:  bde

21 years agoRemove the 6-10 byte translation from UFI and ATAPI command sets. It is
Nate Lawson [Wed, 6 Aug 2003 17:19:06 +0000 (17:19 +0000)]
Remove the 6-10 byte translation from UFI and ATAPI command sets.  It is
no longer needed now that we have PIM_NO_6_BYTE.

Requested by: jhb

21 years agoWhen compiling the module with INVARIANTS we also need INVARIANT_SUPPORT.
Hartmut Brandt [Wed, 6 Aug 2003 16:01:32 +0000 (16:01 +0000)]
When compiling the module with INVARIANTS we also need INVARIANT_SUPPORT.

21 years agoUps. I forgot this one in the SIOCATMENA/SIOCATMDIS removal commit.
Hartmut Brandt [Wed, 6 Aug 2003 15:56:37 +0000 (15:56 +0000)]
Ups. I forgot this one in the SIOCATMENA/SIOCATMDIS removal commit.

This change allows one to specify almost the complete traffic parameters
for IPoverATM channels through the routing table. Up to now we used
4 byte DL addresses (flag, vpi, vciH, vciL). This format is still allowed.
If the address is longer, however, the 5th byte is interpreted as the
traffic class (UBR, CBR, VBR or ABR) and the remaining bytes are the
parameters for this traffic class:

  UBR: 0 byte or 3 byte PCR
  CBR: 3 byte PCR
  VBR: 3 byte PCR, 3 byte SCR, 3 byte MBS
  ABR: 3 byte PCR, 3 byte MCR, 3 byte ICR, 3 byte TBE, 1 byte NRM,
       1 byte TRM, 2 bytes ADTF, 1 byte RIF, 1 byte RDF and 1 byte CDF

A script to generate the corresponding 'route add' arguments will follow soon.

21 years agoChange ieee802 to ieee80211.
Martin Blapp [Wed, 6 Aug 2003 15:19:52 +0000 (15:19 +0000)]
Change ieee802 to ieee80211.
Add bogospaces to match the ISC style.

Suggested by: eivind

21 years agoDont initialize a TSC timecounter until we know if it is broken or not.
Poul-Henning Kamp [Wed, 6 Aug 2003 15:05:27 +0000 (15:05 +0000)]
Dont initialize a TSC timecounter until we know if it is broken or not.

21 years agoUpdate to recognize Geode and note that the TSC seems broken.
Poul-Henning Kamp [Wed, 6 Aug 2003 15:03:47 +0000 (15:03 +0000)]
Update to recognize Geode and note that the TSC seems broken.

21 years agoRemove the ATMIOCENA and ATMIOCDIS ioctl. Everyting has been converted
Hartmut Brandt [Wed, 6 Aug 2003 14:53:27 +0000 (14:53 +0000)]
Remove the ATMIOCENA and ATMIOCDIS ioctl. Everyting has been converted
to use the new OPENVCC and CLOSEVCC calls that allow the sepcification
of traffic parameters for the connections.

21 years agoRemove the last vestiges of ATM raw mode. This has not been useful for a
Hartmut Brandt [Wed, 6 Aug 2003 14:34:38 +0000 (14:34 +0000)]
Remove the last vestiges of ATM raw mode. This has not been useful for a
long time and has already been removed from the only driver that supported
it (en(4)) and from the man page.

21 years agoRemove unused cmd6workaround() calls.
Nate Lawson [Wed, 6 Aug 2003 14:24:16 +0000 (14:24 +0000)]
Remove unused cmd6workaround() calls.

MFC after: 1 day

21 years agoDocument that connect(2) can return EINTR, and that it
Yaroslav Tykhiy [Wed, 6 Aug 2003 14:21:06 +0000 (14:21 +0000)]
Document that connect(2) can return EINTR, and that it
can return EALREADY for a socket in blocking mode as well.

21 years agoMake the first two pages magic to protect the BSD labels rather than
Poul-Henning Kamp [Wed, 6 Aug 2003 14:13:38 +0000 (14:13 +0000)]
Make the first two pages magic to protect the BSD labels rather than
only one.

21 years agoUse the new OPENVCC and CLOSEVCC ioctls to open and close the NATM
Hartmut Brandt [Wed, 6 Aug 2003 14:07:54 +0000 (14:07 +0000)]
Use the new OPENVCC and CLOSEVCC ioctls to open and close the NATM
channel. These have been the last consumers of the old ioctls.

21 years agoIf connect(2) has been interrupted by a signal and therefore the
Yaroslav Tykhiy [Wed, 6 Aug 2003 14:04:47 +0000 (14:04 +0000)]
If connect(2) has been interrupted by a signal and therefore the
connection is to be established asynchronously, behave as in the
case of non-blocking mode:

- keep the SS_ISCONNECTING bit set thus indicating that
  the connection establishment is in progress, which is the case
  (clearing the bit in this case was just a bug);

- return EALREADY, instead of the confusing and unreasonable
  EADDRINUSE, upon further connect(2) attempts on this socket
  until the connection is established (this also brings our
  connect(2) into accord with IEEE Std 1003.1.)

21 years agoStyle(9): remove unneccesary space and blank lines, indentation,
Hartmut Brandt [Wed, 6 Aug 2003 13:46:15 +0000 (13:46 +0000)]
Style(9): remove unneccesary space and blank lines, indentation,
function prototypes. Use LIST_FOREACH instead of explicit loops.

The indentation of functions indendet by 4 space have been left alone.
2-space indented functions have been re-indented.

21 years agoMake the driver honor the ATMIO_FLAG_ASYNC that requests
Hartmut Brandt [Wed, 6 Aug 2003 13:16:51 +0000 (13:16 +0000)]
Make the driver honor the ATMIO_FLAG_ASYNC that requests
asynchronuous open/close operations.

21 years agoHonor the ATMIO_FLAG_ASYNC for asynchronuous open/close of VCs.
Hartmut Brandt [Wed, 6 Aug 2003 13:09:36 +0000 (13:09 +0000)]
Honor the ATMIO_FLAG_ASYNC for asynchronuous open/close of VCs.

21 years agoHonor the new ATMIO_FLAG_ASYNC for asynchronuous open/close operations
Hartmut Brandt [Wed, 6 Aug 2003 12:37:50 +0000 (12:37 +0000)]
Honor the new ATMIO_FLAG_ASYNC for asynchronuous open/close operations
on a VC.

21 years agoRemove an unused variable.
Poul-Henning Kamp [Wed, 6 Aug 2003 12:09:34 +0000 (12:09 +0000)]
Remove an unused variable.

21 years agoStaticize swap_pager_putpages()
Poul-Henning Kamp [Wed, 6 Aug 2003 12:08:27 +0000 (12:08 +0000)]
Staticize swap_pager_putpages()

Eliminate a lot of checkes to make sure requests are not cross-device
which is unnecessary with the new layout.  We know a sequential request
cannot possibly be cross-device because there is a reserved page between
the devices.

Remove a couple of comments which no longer are relevant.

21 years agoAccess the swap_pagers' ->putpages() through swappagerops instead
Poul-Henning Kamp [Wed, 6 Aug 2003 12:05:48 +0000 (12:05 +0000)]
Access the swap_pagers' ->putpages() through swappagerops instead
of directly, this is a cleaner way to do it.

21 years agoHonor the new ATMIO_FLAG_ASYNC that requests asynchronuous open/close
Hartmut Brandt [Wed, 6 Aug 2003 11:53:53 +0000 (11:53 +0000)]
Honor the new ATMIO_FLAG_ASYNC that requests asynchronuous open/close
operations on VCs.

21 years agoWhen sending check that the channel is really open. Get the information
Hartmut Brandt [Wed, 6 Aug 2003 11:35:41 +0000 (11:35 +0000)]
When sending check that the channel is really open. Get the information
whether we use AAL5 or not from the aal field of the channel parameters
instead of the flag in the pseudo header. This flag will go away soon.

21 years agoPrint an array index that is computed as ptrdiff_t with %tu.
Hartmut Brandt [Wed, 6 Aug 2003 11:30:53 +0000 (11:30 +0000)]
Print an array index that is computed as ptrdiff_t with %tu.

21 years ago- Don't print "elf" before printing "make world" (aout support has been
Ruslan Ermilov [Wed, 6 Aug 2003 11:06:38 +0000 (11:06 +0000)]
- Don't print "elf" before printing "make world" (aout support has been
  removed).

- Avoid using printf(1) where a simple shell's built-in echo is enough.

21 years agoBroke the warning that the cx driver (cronyx.c, cx,c and if_cx.c) uses
Bruce Evans [Wed, 6 Aug 2003 11:05:42 +0000 (11:05 +0000)]
Broke the warning that the cx driver (cronyx.c, cx,c and if_cx.c) uses
pessimal (u_short) types for i/o ports.  I don't understand the data
structures in this driver well enough to unpessimize the types.

21 years agoBroke the warning that this driver uses pessimal (u_short) types for
Bruce Evans [Wed, 6 Aug 2003 10:52:16 +0000 (10:52 +0000)]
Broke the warning that this driver uses pessimal (u_short) types for
i/o ports by calling the implementation-detail level below inb() and
outb() instead of inb() and outb().  Unpessimizing the types is too
hard since they are mainly used in microcode.

21 years agoAdd XXX: comment to vm_pager_unswapped().
Poul-Henning Kamp [Wed, 6 Aug 2003 10:51:40 +0000 (10:51 +0000)]
Add XXX: comment to vm_pager_unswapped().

21 years agoExplicitly set B_PAGING
Poul-Henning Kamp [Wed, 6 Aug 2003 09:22:47 +0000 (09:22 +0000)]
Explicitly set B_PAGING

21 years agoLocalize 'ls' output
Andrey A. Chernov [Wed, 6 Aug 2003 08:46:21 +0000 (08:46 +0000)]
Localize 'ls' output
Don't set 8bit in quote processing

21 years ago- Unbreak Alpha snapshots (we were still attempting to build a
Ruslan Ermilov [Wed, 6 Aug 2003 08:09:40 +0000 (08:09 +0000)]
- Unbreak Alpha snapshots (we were still attempting to build a
  no longer existing "fixit" crunched binary).

- Simplify the CRUNCH_TARGETS variable's initialization.

- Simplify the release.5 logic a bit by accounting for the fact
  that release/*_crunch.conf files were removed two years ago.

21 years agoChange the format type to unbreak LINT on 64-bit platforms.
David E. O'Brien [Wed, 6 Aug 2003 07:18:42 +0000 (07:18 +0000)]
Change the format type to unbreak LINT on 64-bit platforms.

21 years agoRetire the B_KEEPGIANT flag, we are nowhere near ready.
Poul-Henning Kamp [Wed, 6 Aug 2003 06:53:54 +0000 (06:53 +0000)]
Retire the B_KEEPGIANT flag, we are nowhere near ready.

21 years agoRip out the totally bogos vnode swapdev_vp with extreeme prejudice.
Poul-Henning Kamp [Wed, 6 Aug 2003 06:53:31 +0000 (06:53 +0000)]
Rip out the totally bogos vnode swapdev_vp with extreeme prejudice.

Don't mark buffers with B_KEEPGIANT, we don't drop giant in strategy
at this point in time.

21 years agoDon't drop giant around ->d_strategy(), too much code explodes.
Poul-Henning Kamp [Wed, 6 Aug 2003 06:49:18 +0000 (06:49 +0000)]
Don't drop giant around ->d_strategy(), too much code explodes.

21 years agoFix a typo. s/Line/Like/
Daniel Eischen [Wed, 6 Aug 2003 06:12:54 +0000 (06:12 +0000)]
Fix a typo.  s/Line/Like/

21 years agoAvoid a level of indirection to get from the thread pointer to the
Marcel Moolenaar [Wed, 6 Aug 2003 04:17:42 +0000 (04:17 +0000)]
Avoid a level of indirection to get from the thread pointer to the
TCB. We know that the thread pointer points to &tcb->tcb_tp, so all
we have to do is subtract offsetof(struct tcb, tcb_tp) from the
thread pointer to get to the TCB. Any reasonably smart compiler will
translate accesses to fields in the TCB as negative offsets from TP.

In _tcb_set() make sure the fake TCB gets a pointer to the current
KCB, just like any other TCB. This fixes a NULL-pointer dereference
in _thr_ref_add() when it tried to get the current KSE.

21 years agoAdd Jonathan Postel's year of birth.
Greg Lehey [Wed, 6 Aug 2003 03:38:12 +0000 (03:38 +0000)]
Add Jonathan Postel's year of birth.

21 years agoDon't call kse_set_curthread() when scheduling a new bound
Daniel Eischen [Wed, 6 Aug 2003 00:43:28 +0000 (00:43 +0000)]
Don't call kse_set_curthread() when scheduling a new bound
thread.  It should only be called by the current kse and
never by a KSE on behalf of another.

Submitted by: davidxu

21 years agoRename localdaemons to localpkg.
Mike Makonnen [Wed, 6 Aug 2003 00:35:13 +0000 (00:35 +0000)]
Rename localdaemons to localpkg.
The original name was really a mistake since
/usr/local/etc/rc.d scripts can (and usually do) start
more than just daemons. Even the output in the script
uses 'local packages.' Also, the term 'local daemons' is
used by rc.d/local, which was etc/rc.local of rcOG fame.
No repo-copy because there isn't much history to save.
I will remove localdaemons shortly with all the other
files that don't belong in rc.d anymore.

Discussed with: dougb, freebsd-rc@yahoogroups.com

21 years agoFix an off by one error in the number of arguments passed to
Marcel Moolenaar [Wed, 6 Aug 2003 00:23:40 +0000 (00:23 +0000)]
Fix an off by one error in the number of arguments passed to
makecontext(). We only supply 3, not 4. This is mostly harmless,
except that on ia64 the garbage can include NaT bits, resulting
in NaT consumption faults.