]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
20 years agoNext step in making usb more newbus:
imp [Sat, 14 Aug 2004 22:10:26 +0000 (22:10 +0000)]
Next step in making usb more newbus:
o reprobe children when a new driver is added to uhub
o fix the usbd_probe_and_attach to set the ivars to a malloc'd area, as well
  as freeing the ivars on child destruction.
o Don't delete children that don't attach. Evidentally, the need to do this
  is a common misconception.
o minor formatting foo that may violate style(9) at the moment, but keeps the
  diffs against my p4 tree smaller.

This does not solve the ugen gobbling things up problem, but the fixes
I have for that expose bugs in other parts of the tree...

20 years ago- Capitalize Ethernet and Fast Ethernet.
marius [Sat, 14 Aug 2004 21:44:35 +0000 (21:44 +0000)]
- Capitalize Ethernet and Fast Ethernet.
- Add a NOTES section with information regarding the "local-mac-address?"
  system configuration variable on sparc64 and add a reference to eeprom(8)
  for using it. Dump document date for this.
- In dc.4, add the on-board DM9102A on Sun Netra X1 and Sun Fire V100 to
  the list of known working devices.

20 years ago- Make OF_getetheraddr() honour the "local-mac-address?" system config
marius [Sat, 14 Aug 2004 21:43:37 +0000 (21:43 +0000)]
- Make OF_getetheraddr() honour the "local-mac-address?" system config
  variable. If set to "true" OF_getetheraddr() will now return the unique
  MAC address stored in the "local-mac-address" property of the device's
  OFW node if present and the host address/system default MAC address if
  the node doesn't doesn't have such a property. If set to "false" the
  host address will be returned for all devices like before this change.
  This brings the behaviour of device drivers for NICs with OFW support/
  FCode, i.e. dc(4) for on-board DM9102A on Sun machines, gem(4) and hme(4),
  regarding "local-mac-address?" in line with NetBSD and Solaris.
  The man pages of the respective drivers will be updated separately to
  reflect this change.
- Remove OF_getetheraddr2() which was used as a stopgap in dc(4). Its
  functionality is now part of OF_getetheraddr().

20 years agoRemove confused comment.
phk [Sat, 14 Aug 2004 21:40:28 +0000 (21:40 +0000)]
Remove confused comment.

20 years agoRemove spl calls.
alc [Sat, 14 Aug 2004 18:57:41 +0000 (18:57 +0000)]
Remove spl calls.

20 years agoStop defining '_Complex' in a C99 environment, it is supposed to be a keyword.
stefanf [Sat, 14 Aug 2004 18:03:21 +0000 (18:03 +0000)]
Stop defining '_Complex' in a C99 environment, it is supposed to be a keyword.

20 years agoUse tabulators after '#define'.
stefanf [Sat, 14 Aug 2004 17:55:15 +0000 (17:55 +0000)]
Use tabulators after '#define'.

20 years agoAvoid using void pointers in additive expressions.
stefanf [Sat, 14 Aug 2004 17:46:10 +0000 (17:46 +0000)]
Avoid using void pointers in additive expressions.

PR: 56653

20 years agoCause pfind() not to return processes in the PRS_NEW state. As a result,
rwatson [Sat, 14 Aug 2004 17:15:16 +0000 (17:15 +0000)]
Cause pfind() not to return processes in the PRS_NEW state.  As a result,
threads consuming the result of pfind() will not need to check for a NULL
credential pointer or other signs of an incompletely created process.
However, this also means that pfind() cannot be used to test for the
existence or find such a process.  Annotate pfind() to indicate that this
is the case.  A review of curent consumers seems to indicate that this is
not a problem for any of them.  This closes a number of race conditions
that could result in NULL pointer dereferences and related failure modes.
Other related races continue to exist, especially during iteration of the
allproc list without due caution.

Discussed with: tjr, green

20 years agoAdd a note about RANDOM_IP_ID.
dwmalone [Sat, 14 Aug 2004 16:16:01 +0000 (16:16 +0000)]
Add a note about RANDOM_IP_ID.

20 years agoGet rid of the RANDOM_IP_ID option and make it a sysctl. NetBSD
dwmalone [Sat, 14 Aug 2004 15:32:40 +0000 (15:32 +0000)]
Get rid of the RANDOM_IP_ID option and make it a sysctl.  NetBSD
have already done this, so I have styled the patch on their work:

        1) introduce a ip_newid() static inline function that checks
        the sysctl and then decides if it should return a sequential
        or random IP ID.

        2) named the sysctl net.inet.ip.random_id

        3) IPv6 flow IDs and fragment IDs are now always random.
        Flow IDs and frag IDs are significantly less common in the
        IPv6 world (ie. rarely generated per-packet), so there should
        be smaller performance concerns.

The sysctl defaults to 0 (sequential IP IDs).

Reviewed by: andre, silby, mlaier, ume
Based on: NetBSD
MFC after: 2 months

20 years agoFix outgoing ICMP on global instance.
phk [Sat, 14 Aug 2004 14:21:09 +0000 (14:21 +0000)]
Fix outgoing ICMP on global instance.

20 years agoAdd support for C3 Nehemiah ACE ("Padlock") AES crypto. This comes
markm [Sat, 14 Aug 2004 13:38:35 +0000 (13:38 +0000)]
Add support for C3 Nehemiah ACE ("Padlock") AES crypto. This comes
from OpenSSL 0.9.5 (yet to be released), and is pretty complete.

20 years agoMake informational output look less like an accident.
le [Sat, 14 Aug 2004 09:56:17 +0000 (09:56 +0000)]
Make informational output look less like an accident.

20 years agoAdd XXX comment about findcdev() misuse.
phk [Sat, 14 Aug 2004 08:38:17 +0000 (08:38 +0000)]
Add XXX comment about findcdev() misuse.

20 years agoDon't declare everything we find on a loopback interface for passive:
phk [Sat, 14 Aug 2004 08:36:35 +0000 (08:36 +0000)]
Don't declare everything we find on a loopback interface for passive:

Only the actual loopback address should be declared passive, other
addresses are very likely to be desirable to announce.

Check for IFF_LOOPBACK instead of IFF_PASSIVE to determine if we have
an unknown interface type.

20 years agoAdd some KASSERTS.
phk [Sat, 14 Aug 2004 08:33:49 +0000 (08:33 +0000)]
Add some KASSERTS.

20 years agoWhitespace nit.
julian [Sat, 14 Aug 2004 07:21:20 +0000 (07:21 +0000)]
Whitespace nit.

20 years agoCorrect grammo.
schweikh [Sat, 14 Aug 2004 06:29:17 +0000 (06:29 +0000)]
Correct grammo.

20 years agoAllocate memory in the unwinder with M_NOWAIT. We may need to provide
marcel [Sat, 14 Aug 2004 05:00:37 +0000 (05:00 +0000)]
Allocate memory in the unwinder with M_NOWAIT. We may need to provide
backtraces with locks held.

20 years agoEliminate reliance on non-portable <err.h> by implementing a very
kientzle [Sat, 14 Aug 2004 03:45:45 +0000 (03:45 +0000)]
Eliminate reliance on non-portable <err.h> by implementing a very
simple errx() function.
Improve behavior when bzlib/zlib are missing by detecting and
issuing an error message on attempts to read gzip/bzip2 compressed
archives.

20 years agoAfter completing a name lookup for a target UNIX domain socket to
rwatson [Sat, 14 Aug 2004 03:43:49 +0000 (03:43 +0000)]
After completing a name lookup for a target UNIX domain socket to
connect to, re-check that the local UNIX domain socket hasn't been
closed while we slept, and if so, return EINVAL.  This affects the
system running both with and without Giant over the network stack,
and recent ULE changes appear to cause it to trigger more
frequently than previously under load.  While here, improve catching
of possibly closed UNIX domain sockets in one or two additional
circumstances.  I have a much larger set of related changes in
Perforce, but they require more testing before they can be merged.

One debugging printf is left in place to indicate when such a race
takes place: this is typically triggered by a buggy application
that simultaenously connect()'s and close()'s a UNIX domain socket
file descriptor.  I'll remove this at some point in the future, but
am interested in seeing how frequently this is reported.  In the
case of Martin's reported problem, it appears to be a result of a
non-thread safe syslog() implementation in the C library, which
does not synchronize access to its logging file descriptor.

Reported by: mbr

20 years agoWe don't need <paths.h>, so don't bother including it.
kientzle [Sat, 14 Aug 2004 03:43:35 +0000 (03:43 +0000)]
We don't need <paths.h>, so don't bother including it.

20 years agoAdd libgcov, a runtime support library for binaries compiled for basic
kan [Sat, 14 Aug 2004 03:23:24 +0000 (03:23 +0000)]
Add libgcov, a runtime support library for binaries compiled for basic
block profiling.

20 years agoAdd new PCI device ID for PERC4/DI.
ambrisko [Sat, 14 Aug 2004 02:48:13 +0000 (02:48 +0000)]
Add new PCI device ID for PERC4/DI.

20 years agoSince if_oltr doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Sat, 14 Aug 2004 00:19:07 +0000 (00:19 +0000)]
Since if_oltr doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

20 years agoSince if_ixgb doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Sat, 14 Aug 2004 00:17:04 +0000 (00:17 +0000)]
Since if_ixgb doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

20 years agoSince if_xe doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Sat, 14 Aug 2004 00:15:26 +0000 (00:15 +0000)]
Since if_xe doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

20 years agoSince if_vx doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Sat, 14 Aug 2004 00:12:42 +0000 (00:12 +0000)]
Since if_vx doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

20 years agoSince if_txp doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Fri, 13 Aug 2004 23:53:36 +0000 (23:53 +0000)]
Since if_txp doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

Note: mutexes are initialized in the softc for this driver, but the
locking appears inadequate to allow Giant-free operation.

20 years agoSince if_tx doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Fri, 13 Aug 2004 23:52:33 +0000 (23:52 +0000)]
Since if_tx doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

20 years agoSince if_sr doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Fri, 13 Aug 2004 23:49:48 +0000 (23:49 +0000)]
Since if_sr doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

20 years agoSince if_snc doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Fri, 13 Aug 2004 23:47:01 +0000 (23:47 +0000)]
Since if_snc doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

20 years agoSince if_sbni doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Fri, 13 Aug 2004 23:41:00 +0000 (23:41 +0000)]
Since if_sbni doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

20 years agoSince if_ray doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Fri, 13 Aug 2004 23:39:07 +0000 (23:39 +0000)]
Since if_ray doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

20 years agoSince if_plip doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Fri, 13 Aug 2004 23:32:17 +0000 (23:32 +0000)]
Since if_plip doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

20 years agoSince if_nge doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Fri, 13 Aug 2004 23:22:53 +0000 (23:22 +0000)]
Since if_nge doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

Note: this driver does declare and occasionally reference mutexes,
but I believe not nearly enough to provide safety.

20 years agoAvoid using void pointers in additive expressions.
stefanf [Fri, 13 Aug 2004 23:22:38 +0000 (23:22 +0000)]
Avoid using void pointers in additive expressions.

PR: 56653 (libc/rpc bits)
Approved by: alfred

20 years agoSince if_lnc doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Fri, 13 Aug 2004 23:20:50 +0000 (23:20 +0000)]
Since if_lnc doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

20 years agoSince if_lge doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Fri, 13 Aug 2004 23:18:01 +0000 (23:18 +0000)]
Since if_lge doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

20 years agoSince if_ic doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Fri, 13 Aug 2004 23:16:44 +0000 (23:16 +0000)]
Since if_ic doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

20 years agoSince if_ie doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Fri, 13 Aug 2004 23:15:44 +0000 (23:15 +0000)]
Since if_ie doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

20 years agoSince if_hme doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Fri, 13 Aug 2004 23:14:50 +0000 (23:14 +0000)]
Since if_hme doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

20 years agoSince if_gem doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Fri, 13 Aug 2004 23:11:24 +0000 (23:11 +0000)]
Since if_gem doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

20 years agoSince if_fwip doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Fri, 13 Aug 2004 23:09:41 +0000 (23:09 +0000)]
Since if_fwip doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

Approved by: dfr

20 years agoSince if_fe doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Fri, 13 Aug 2004 23:08:08 +0000 (23:08 +0000)]
Since if_fe doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

20 years agoSince if_ex doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Fri, 13 Aug 2004 23:06:55 +0000 (23:06 +0000)]
Since if_ex doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

Note: there are locking macros in if_exreg.h, but they appear to be
unused.

20 years agoSince if_ed doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Fri, 13 Aug 2004 23:04:23 +0000 (23:04 +0000)]
Since if_ed doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

20 years agoWordsmithing on the GEOM_STRIPE, MSDOSFS_LARGE, and portaudit notes.
bmah [Fri, 13 Aug 2004 23:04:14 +0000 (23:04 +0000)]
Wordsmithing on the GEOM_STRIPE, MSDOSFS_LARGE, and portaudit notes.

20 years agoSince if_cs doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Fri, 13 Aug 2004 23:03:11 +0000 (23:03 +0000)]
Since if_cs doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

20 years agoSince if_cp doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Fri, 13 Aug 2004 23:02:17 +0000 (23:02 +0000)]
Since if_cp doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

20 years agoAdd support for SSDT tables. Dumping or disassembling the DSDT will
marcel [Fri, 13 Aug 2004 22:59:09 +0000 (22:59 +0000)]
Add support for SSDT tables. Dumping or disassembling the DSDT will
now include the contents if any SSDT table as well. This makes use
of the property that one can concatenate the body of SSDT tables to
the DSDT, updating the DSDT header (length and checksum) and end up
with a larger and valid DSDT table. Hence, this also works with -f.

Reviewed by: njl@

20 years agoSince if_cm doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Fri, 13 Aug 2004 22:57:44 +0000 (22:57 +0000)]
Since if_cm doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

20 years agoSince if_awi doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Fri, 13 Aug 2004 22:55:25 +0000 (22:55 +0000)]
Since if_awi doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

20 years agoSince if_arl doesn't contain locking or run with INTR_MPSAFE, mark
rwatson [Fri, 13 Aug 2004 22:54:19 +0000 (22:54 +0000)]
Since if_arl doesn't contain locking or run with INTR_MPSAFE, mark
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.

20 years agoAs the if_ar driver doesn't contain locking or run its interrupt
rwatson [Fri, 13 Aug 2004 22:52:11 +0000 (22:52 +0000)]
As the if_ar driver doesn't contain locking or run its interrupt
MPSAFE, mark it as IFF_NEEDSGIANT so that its if_start routine is
run holding Giant.

20 years agoSince the if_de driver doesn't contain locking, mark it as
rwatson [Fri, 13 Aug 2004 22:48:05 +0000 (22:48 +0000)]
Since the if_de driver doesn't contain locking, mark it as
IFF_NEEDSGIANT so that ifp->if_start won't be called without Giant
when running debug.mpsafenet=1.

20 years agoAdd a man page for tgmath.h.
stefanf [Fri, 13 Aug 2004 22:43:12 +0000 (22:43 +0000)]
Add a man page for tgmath.h.

Reviewed by: keramida

20 years agoFix the memory scaling bug when basemem was converted to Kbytes from
ambrisko [Fri, 13 Aug 2004 22:30:55 +0000 (22:30 +0000)]
Fix the memory scaling bug when basemem was converted to Kbytes from
bytes for AMD64.  Otherwise the AP will be started at 640K which
won't work.  Bug found on a Xeon 64bit system.

20 years agoUse IFQ_SET_MAXLEN() to set the maximum queue depth of the routing
rwatson [Fri, 13 Aug 2004 22:23:21 +0000 (22:23 +0000)]
Use IFQ_SET_MAXLEN() to set the maximum queue depth of the routing
socket netisr queue.

Pointed out by: winter

20 years agoExtend critical section protection around portions of selection processing
gibbs [Fri, 13 Aug 2004 21:41:23 +0000 (21:41 +0000)]
Extend critical section protection around portions of selection processing
that cannot tolerate changes to the waiting for selection queue by the
host or the host canceling an active selection.

20 years agoFix an off by one in the critical section clearing code. The
gibbs [Fri, 13 Aug 2004 21:39:14 +0000 (21:39 +0000)]
Fix an off by one in the critical section clearing code.  The
code was adjusting twice for the instruction pointer indicating
the *next* instruction to execute.  The aic79xx driver had a similar
bug, but was fixed some time ago.

20 years agoCommit a work-around for a more general bug involving process state:
rwatson [Fri, 13 Aug 2004 20:27:56 +0000 (20:27 +0000)]
Commit a work-around for a more general bug involving process state:
check whether p_ucred is NULL or not in pfs_getattr() before
dereferencing the credential, and return ENOENT if there wasn't one.

This is a symptom of a larger problem, wherein pfind() can return
references to incompletely initialized processes, and we instead ought
to not return them, or check the process state before acting on the
process.

Reported by: kris
Discussed with: tjr, others

20 years agoThis commit was generated by cvs2svn to compensate for changes in r133665,
markm [Fri, 13 Aug 2004 19:37:23 +0000 (19:37 +0000)]
This commit was generated by cvs2svn to compensate for changes in r133665,
which included commits to RCS files with non-trunk default branches.

20 years agoBring in support for VIA C3 Nehemiah Padlock crypto support (AES).
markm [Fri, 13 Aug 2004 19:37:23 +0000 (19:37 +0000)]
Bring in support for VIA C3 Nehemiah Padlock crypto support (AES).
This is from the upcoming OpenSSL 0.9.8 release.

20 years agoFix building for the ACPI_DEBUG case.
njl [Fri, 13 Aug 2004 19:27:21 +0000 (19:27 +0000)]
Fix building for the ACPI_DEBUG case.

20 years agoAdd pmap locking to pmap_remove_all().
alc [Fri, 13 Aug 2004 18:54:21 +0000 (18:54 +0000)]
Add pmap locking to pmap_remove_all().

20 years agoWhen printing out an unknown sense code we should print it in hex, not
mjacob [Fri, 13 Aug 2004 18:45:04 +0000 (18:45 +0000)]
When printing out an unknown sense code we should print it in hex, not
decimal.

Reviewed by: gibbs, nate, kdm

20 years agoRecord the new status after checking if it has changed, not before. This
njl [Fri, 13 Aug 2004 17:47:40 +0000 (17:47 +0000)]
Record the new status after checking if it has changed, not before.  This
fixes lost AC line transition events.

Bug report: Kevin Oberman

20 years agoclean up whitespace...
jmg [Fri, 13 Aug 2004 17:43:53 +0000 (17:43 +0000)]
clean up whitespace...

20 years agoMiscellaneous edits. Mostly style, grammar, and punctuation fixes in
bmah [Fri, 13 Aug 2004 15:09:20 +0000 (15:09 +0000)]
Miscellaneous edits.  Mostly style, grammar, and punctuation fixes in
the vein of my previous, recent commits.

Updated release notes: Augmented sha(1)/rmd160(1) note, updated pf(4)
pseudo-users, retweaked doscmd note to reinstate a missing word I
deleted by accident [1] and add some more info.

MFCs noted:  ppp(8) LQM, ppp(8) set rad_alive.

Pointed out by: "Eagle Eyes" ru [1]

20 years agoDocument the MNT_SNAPSHOT mount flag with a cross-reference
roam [Fri, 13 Aug 2004 14:33:03 +0000 (14:33 +0000)]
Document the MNT_SNAPSHOT mount flag with a cross-reference
to mksnap_ffs(8).

PR: 70402
Submitted by: James Raftery <james@now.ie>

20 years agoRemoved COPTS support from kmod.mk and kern.pre.mk.
ru [Fri, 13 Aug 2004 14:30:26 +0000 (14:30 +0000)]
Removed COPTS support from kmod.mk and kern.pre.mk.
COPTS support in bsd.prog.mk is preserved but discouraged.

20 years agoReplaced COPTS by equivalent CFLAGS.
ru [Fri, 13 Aug 2004 14:21:49 +0000 (14:21 +0000)]
Replaced COPTS by equivalent CFLAGS.

20 years agoRemoved COPTS with equivalent CFLAGS.
ru [Fri, 13 Aug 2004 14:19:12 +0000 (14:19 +0000)]
Removed COPTS with equivalent CFLAGS.
While here, fixed setting of BINDIR.

20 years agoRemoved commented out bitrot.
ru [Fri, 13 Aug 2004 14:18:24 +0000 (14:18 +0000)]
Removed commented out bitrot.

20 years agoProofreading the 4.x -> 5.x upgrade procedure.
ru [Fri, 13 Aug 2004 13:14:05 +0000 (13:14 +0000)]
Proofreading the 4.x -> 5.x upgrade procedure.

20 years agoReplaced a way outdated and unsafe sequence of upgrading the system
ru [Fri, 13 Aug 2004 12:59:46 +0000 (12:59 +0000)]
Replaced a way outdated and unsafe sequence of upgrading the system
from sources with the link to the relevant section in src/UPDATING.

20 years agoFix resource check while autodetection.
rik [Fri, 13 Aug 2004 12:35:52 +0000 (12:35 +0000)]
Fix resource check while autodetection.

20 years agoWhite space cleanup.
rik [Fri, 13 Aug 2004 12:27:24 +0000 (12:27 +0000)]
White space cleanup.

20 years agoWhite space cleanup.
rik [Fri, 13 Aug 2004 12:22:01 +0000 (12:22 +0000)]
White space cleanup.

20 years agoWhite space cleanup.
rik [Fri, 13 Aug 2004 12:14:39 +0000 (12:14 +0000)]
White space cleanup.

20 years agoRegen.
fjoe [Fri, 13 Aug 2004 09:56:22 +0000 (09:56 +0000)]
Regen.

20 years agoI was born in USSR.
fjoe [Fri, 13 Aug 2004 09:54:29 +0000 (09:54 +0000)]
I was born in USSR.

20 years agoRegression test for geom_uzip.
fjoe [Fri, 13 Aug 2004 09:53:52 +0000 (09:53 +0000)]
Regression test for geom_uzip.

20 years agoAdd geom_uzip -- geom class that implements read-only compressed disks.
fjoe [Fri, 13 Aug 2004 09:40:58 +0000 (09:40 +0000)]
Add geom_uzip -- geom class that implements read-only compressed disks.
Currently supports cloop V2.0 disk compression format.
May support more formats in future.

20 years agoAdd a regression test for the ATM call control stuff.
harti [Fri, 13 Aug 2004 09:27:21 +0000 (09:27 +0000)]
Add a regression test for the ATM call control stuff.

20 years agontptimeset is not build anymore by default so remove it for the time being.
roberto [Fri, 13 Aug 2004 09:10:38 +0000 (09:10 +0000)]
ntptimeset is not build anymore by default so remove it for the time being.

Prodded by: imp

20 years agoAllow the use of a supplied function to set the PRD table. This is
sos [Fri, 13 Aug 2004 08:14:27 +0000 (08:14 +0000)]
Allow the use of a supplied function to set the PRD table. This is
needed for new chips that supports 64bit addressing.

20 years agoReplace the linear search in vm_map_findspace() with an O(log n)
alc [Fri, 13 Aug 2004 08:06:34 +0000 (08:06 +0000)]
Replace the linear search in vm_map_findspace() with an O(log n)
algorithm built into the map entry splay tree.  This replaces the
first_free hint in struct vm_map with two fields in vm_map_entry:
adj_free, the amount of free space following a map entry, and
max_free, the maximum amount of free space in the entry's subtree.
These fields make it possible to find a first-fit free region of a
given size in one pass down the tree, so O(log n) amortized using
splay trees.

This significantly reduces the overhead in vm_map_findspace() for
applications that mmap() many hundreds or thousands of regions, and
has a negligible slowdown (0.1%) on buildworld.  See, for example, the
discussion of a micro-benchmark titled "Some mmap observations
compared to Linux 2.6/OpenBSD" on -hackers in late October 2003.

OpenBSD adopted this approach in March 2002, and NetBSD added it in
November 2003, both with Red-Black trees.

Submitted by: Mark W. Krentel

20 years agolooks like rwatson forgot tabs... :)
jmg [Fri, 13 Aug 2004 07:38:58 +0000 (07:38 +0000)]
looks like rwatson forgot tabs... :)

20 years agoMake -I and --files-from be synonyms for -T
kientzle [Fri, 13 Aug 2004 07:23:02 +0000 (07:23 +0000)]
Make -I and --files-from be synonyms for -T

20 years agoMove PNP IDs back into oldcard files
imp [Fri, 13 Aug 2004 06:57:31 +0000 (06:57 +0000)]
Move PNP IDs back into oldcard files

20 years ago1. Add missing functions: libthr_dbresume,libthr_dbsuspend.
davidxu [Fri, 13 Aug 2004 06:47:33 +0000 (06:47 +0000)]
1. Add missing functions: libthr_dbresume,libthr_dbsuspend.
2. Implement functions: libthr_db_thr_setfpregs, libthr_db_thr_setregs,
   libthr_db_ta_map_id2thr.
3. simplify libthr_db_thr_getfpregs, libthr_db_thr_getgregs.

20 years agoMPSAFE locking
njl [Fri, 13 Aug 2004 06:22:35 +0000 (06:22 +0000)]
MPSAFE locking

* Serialize access to the sysctl routines and the notify handler
* Assert that the sx lock is held in any functions they call.
* Note that recursively calling to re-enable the hotkeys is sub-optimal.

20 years agoMPSAFE locking
njl [Fri, 13 Aug 2004 06:22:31 +0000 (06:22 +0000)]
MPSAFE locking

* Serialize access to the sysctl routines and the notify handler
* Assert that the sx lock is held in any functions they call.

20 years agoMPSAFE locking
njl [Fri, 13 Aug 2004 06:22:29 +0000 (06:22 +0000)]
MPSAFE locking

* Serialize access to the sysctl routines and the notify handler.

20 years agoMPSAFE locking
njl [Fri, 13 Aug 2004 06:22:26 +0000 (06:22 +0000)]
MPSAFE locking

* Simplify taskqueue locking.
* Don't acquire Giant around calls to the taskqueue function.
* Remove 4.x compatibility routines.

20 years agoMPSAFE locking
njl [Fri, 13 Aug 2004 06:22:24 +0000 (06:22 +0000)]
MPSAFE locking

* Remove the interrupt wrapper that locked Giant and call the handler
  directly.  Mark the handler as MPSAFE.
* Don't attempt to detect if a handler is installed.  Leave that to the
  bus_alloc_resource() function.

20 years agoMPSAFE locking
njl [Fri, 13 Aug 2004 06:22:20 +0000 (06:22 +0000)]
MPSAFE locking

* Serialize operations in acpi_video_bind_outputs(), acpi_video_detach(),
  acpi_video_notify_handler(), acpi_video_power_profile(), and the sysctls.
  The main goal is to protect the shared device list and prevent conflicting
  settings.
* Add assertions that the sx lock is held in the leaf functions.