openbsd
10 years agoWe still need to cast the resulting pointer in NEW and NEW2 as it
millert [Wed, 8 Jan 2014 23:12:57 +0000 (23:12 +0000)]
We still need to cast the resulting pointer in NEW and NEW2 as it
is used in pointer arithmetic.  Otherwise we do arithmetic on void
* and corrupt the heap.

10 years agoAdd casts to unsigned char for ctype functions and other places
millert [Wed, 8 Jan 2014 22:55:59 +0000 (22:55 +0000)]
Add casts to unsigned char for ctype functions and other places
where a char is stored in an int.  Joint work with and OK deraadt@

10 years agoregen
brad [Wed, 8 Jan 2014 22:52:54 +0000 (22:52 +0000)]
regen

10 years agoUse the same naming scheme on both sides for the Attansic entries and
brad [Wed, 8 Jan 2014 22:52:05 +0000 (22:52 +0000)]
Use the same naming scheme on both sides for the Attansic entries and
replace some tabs with spaces in between product and Intel.

10 years agoName the local variables for struct ifaddr consistently "ifa".
bluhm [Wed, 8 Jan 2014 22:38:29 +0000 (22:38 +0000)]
Name the local variables for struct ifaddr consistently "ifa".
OK mikeb@

10 years agoUse calloc() instead of malloc() + memset. Based on a diff from
millert [Wed, 8 Jan 2014 22:36:37 +0000 (22:36 +0000)]
Use calloc() instead of malloc() + memset.  Based on a diff from
Michael W. Bombardieri.  OK deraadt@

10 years agoMake allocate() take size_t and return void *. This lets us drop
millert [Wed, 8 Jan 2014 22:30:32 +0000 (22:30 +0000)]
Make allocate() take size_t and return void *.  This lets us drop
some more useless casts.  Also add missing arguments to a couple
of prototypes while here.  OK matthew@ pelikan@

10 years agoFix typo recieve -> receive.
bluhm [Wed, 8 Jan 2014 22:24:35 +0000 (22:24 +0000)]
Fix typo recieve -> receive.
OK sf@

10 years agoTake restore off the floppy media for now. Sorry.
deraadt [Wed, 8 Jan 2014 22:17:52 +0000 (22:17 +0000)]
Take restore off the floppy media for now.  Sorry.

10 years agoSwitch to using the base sha256 command, rather than some customized
deraadt [Wed, 8 Jan 2014 21:53:33 +0000 (21:53 +0000)]
Switch to using the base sha256 command, rather than some customized
abomination.
install.sub chunk tested by rpe

10 years agosync
deraadt [Wed, 8 Jan 2014 21:52:28 +0000 (21:52 +0000)]
sync

10 years agocreate the /etc/signify directory
deraadt [Wed, 8 Jan 2014 21:52:22 +0000 (21:52 +0000)]
create the /etc/signify directory

10 years agoRemove CALLOC, MALLOC, FREE and REALLOC macros and just call calloc(),
millert [Wed, 8 Jan 2014 21:40:25 +0000 (21:40 +0000)]
Remove CALLOC, MALLOC, FREE and REALLOC macros and just call calloc(),
nalloc(), free() and realloc() directly.  The macros were casting
to the wrong (pre-C89) types and there is no need for them in a C89
world.  OK matthew@

10 years agoregen
jcs [Wed, 8 Jan 2014 21:16:38 +0000 (21:16 +0000)]
regen

10 years agoadd a bunch of newer apple wellspring keyboard/trackpad ids
jcs [Wed, 8 Jan 2014 21:15:29 +0000 (21:15 +0000)]
add a bunch of newer apple wellspring keyboard/trackpad ids

10 years agoIf the file size is zero, there's nothing to transmit so avoid spamming
jca [Wed, 8 Jan 2014 17:31:36 +0000 (17:31 +0000)]
If the file size is zero, there's nothing to transmit so avoid spamming
syslog with mmap failures.  Noticed by and ok henning@, "looks right"
deraadt@

10 years agoBe sure to have pmap_extract() return FALSE for existing, but non-valid user
miod [Wed, 8 Jan 2014 17:12:18 +0000 (17:12 +0000)]
Be sure to have pmap_extract() return FALSE for existing, but non-valid user
pmap ptes.  Found the hard way by tobiasu@.

10 years agoFix wrapping of usage message for cksum.
millert [Wed, 8 Jan 2014 16:23:21 +0000 (16:23 +0000)]
Fix wrapping of usage message for cksum.

10 years agoflesh out STANDARDS somewhat; help/ok sobrado
jmc [Wed, 8 Jan 2014 16:19:24 +0000 (16:19 +0000)]
flesh out STANDARDS somewhat; help/ok sobrado
this was also ok millert some time ago

10 years agoerr() not errx() when fopen fails.
millert [Wed, 8 Jan 2014 16:16:44 +0000 (16:16 +0000)]
err() not errx() when fopen fails.

10 years agozap trailing whitespace;
jmc [Wed, 8 Jan 2014 16:15:45 +0000 (16:15 +0000)]
zap trailing whitespace;

10 years agoRemove NHASHES, it is not needed and is incompatible with #ifdef SMALL.
millert [Wed, 8 Jan 2014 16:13:11 +0000 (16:13 +0000)]
Remove NHASHES, it is not needed and is incompatible with #ifdef SMALL.
Remove style_hash and style_cksum, adjust the style of the selected
algorithms based on rflag and qflag instead.

10 years agosort options;
jmc [Wed, 8 Jan 2014 16:12:44 +0000 (16:12 +0000)]
sort options;

10 years agoRemove pmode, fix hfile open mode, check return value when writing
millert [Wed, 8 Jan 2014 15:54:09 +0000 (15:54 +0000)]
Remove pmode, fix hfile open mode, check return value when writing
file name to stdout.

10 years agoctype cleanup; ok gilles
deraadt [Wed, 8 Jan 2014 15:30:49 +0000 (15:30 +0000)]
ctype cleanup; ok gilles

10 years agoenter md5 directory
deraadt [Wed, 8 Jan 2014 14:20:52 +0000 (14:20 +0000)]
enter md5 directory

10 years agoBuild a -DSMALL verison of md5 for the install media (will be installed
deraadt [Wed, 8 Jan 2014 14:20:11 +0000 (14:20 +0000)]
Build a -DSMALL verison of md5 for the install media (will be installed
as sha256 and sha512, primarily)

10 years agoAdd -DSMALL support to kill a lot of the unncecessary hashes and checksums
deraadt [Wed, 8 Jan 2014 14:19:25 +0000 (14:19 +0000)]
Add -DSMALL support to kill a lot of the unncecessary hashes and checksums
that the install media won't need
ok millert, but he used the word butcher!

10 years agoAdd support for -h hashfile. This is very useful with with -p, in a pipeline,
deraadt [Wed, 8 Jan 2014 14:15:54 +0000 (14:15 +0000)]
Add support for -h hashfile.  This is very useful with with -p, in a pipeline,
to deliver the hash output to a different location.
ok millert

10 years agounsigned char for ctype. ok deraadt
okan [Wed, 8 Jan 2014 13:23:55 +0000 (13:23 +0000)]
unsigned char for ctype. ok deraadt

10 years agofix logic/usage in VERIFYONLY case
espie [Wed, 8 Jan 2014 07:04:29 +0000 (07:04 +0000)]
fix logic/usage in VERIFYONLY case
okay deraadt@

10 years agoDelete struct definitions that have been obsolete for a dozen years
guenther [Wed, 8 Jan 2014 06:50:57 +0000 (06:50 +0000)]
Delete struct definitions that have been obsolete for a dozen years

ok deraadt@

10 years agoZap leading whitespace
deraadt [Wed, 8 Jan 2014 06:43:34 +0000 (06:43 +0000)]
Zap leading whitespace

10 years agoZap trailing whitespace
guenther [Wed, 8 Jan 2014 06:41:49 +0000 (06:41 +0000)]
Zap trailing whitespace

10 years agosynch with signify(1) paranoia
espie [Wed, 8 Jan 2014 06:40:56 +0000 (06:40 +0000)]
synch with signify(1) paranoia

10 years agoalways include dd
deraadt [Wed, 8 Jan 2014 06:17:21 +0000 (06:17 +0000)]
always include dd

10 years agocalling HashFinal with a null digest should crash, not be silently ignored
tedu [Wed, 8 Jan 2014 06:14:56 +0000 (06:14 +0000)]
calling HashFinal with a null digest should crash, not be silently ignored

10 years agoReplace strtoq(), QUAD_MAX, and %qu with their long long equivalent
guenther [Wed, 8 Jan 2014 05:52:47 +0000 (05:52 +0000)]
Replace strtoq(), QUAD_MAX, and %qu with their long long equivalent

10 years agoa few more blocks can use VERIFYONLY
deraadt [Wed, 8 Jan 2014 05:51:35 +0000 (05:51 +0000)]
a few more blocks can use VERIFYONLY

10 years agoallow building a VERIFYONLY version that's considerably smaller.
tedu [Wed, 8 Jan 2014 05:00:01 +0000 (05:00 +0000)]
allow building a VERIFYONLY version that's considerably smaller.
most of the savings comes from eliminating the large constant table
in ge25519.c. ok deraadt

10 years agoEliminate poisonous LONG_OFF_T conditional
guenther [Wed, 8 Jan 2014 04:58:36 +0000 (04:58 +0000)]
Eliminate poisonous LONG_OFF_T conditional

adjective suggested by deraadt@

10 years agoDelete a bogus extern declaration and move around a couple others
guenther [Wed, 8 Jan 2014 04:48:29 +0000 (04:48 +0000)]
Delete a bogus extern declaration and move around a couple others

10 years agoUse size_t where appropriate
guenther [Wed, 8 Jan 2014 04:43:48 +0000 (04:43 +0000)]
Use size_t where appropriate

ok fgsch@

10 years agoRename arguments to avoid shadowing global variables
guenther [Wed, 8 Jan 2014 04:42:24 +0000 (04:42 +0000)]
Rename arguments to avoid shadowing global variables

ok fgsch@

10 years agoMark some functions as printf-like and fix a bogus format string
guenther [Wed, 8 Jan 2014 04:41:40 +0000 (04:41 +0000)]
Mark some functions as printf-like and fix a bogus format string

ok fgsch@

10 years agorename SHA256_ONLY to SHA2_SMALL; changing things so that sha512 support
deraadt [Wed, 8 Jan 2014 04:35:34 +0000 (04:35 +0000)]
rename SHA256_ONLY to SHA2_SMALL; changing things so that sha512 support
is also pulled in

10 years agoenable building of signify in the ramdisks
deraadt [Wed, 8 Jan 2014 04:33:23 +0000 (04:33 +0000)]
enable building of signify in the ramdisks

10 years agoreduce size by using equivalent code from libc instead of nacl
tedu [Wed, 8 Jan 2014 03:59:46 +0000 (03:59 +0000)]
reduce size by using equivalent code from libc instead of nacl
ok deraadt

10 years agono tomfoolery: mark the comment as untrusted, and enforce it.
tedu [Wed, 8 Jan 2014 03:57:57 +0000 (03:57 +0000)]
no tomfoolery: mark the comment as untrusted, and enforce it.
usability: add random key fingerprints to detect sig/pubkey mismatch
feedback espie naddy

10 years agoPerform manual malloc/free of a large object in pfkeyv2_send() to
deraadt [Wed, 8 Jan 2014 02:39:02 +0000 (02:39 +0000)]
Perform manual malloc/free of a large object in pfkeyv2_send() to
reduce stack space requirements.
ok mikeb mpi

10 years agoMake skipping the root fs DUID aware.
rpe [Tue, 7 Jan 2014 23:58:44 +0000 (23:58 +0000)]
Make skipping the root fs DUID aware.

If ROOTDEV is either the device from fstab or the converted DUID
device name with the partition added, skip the current fstab entry

ok halex@

10 years agoAsa Yeamans points out that we need to CMSG_SPACE[] for the 2nd message
deraadt [Tue, 7 Jan 2014 19:23:13 +0000 (19:23 +0000)]
Asa Yeamans points out that we need to CMSG_SPACE[] for the 2nd message
as well.
ok guenther millert

10 years agoseverely restrict what characters can be in a signer, to make
espie [Tue, 7 Jan 2014 18:35:23 +0000 (18:35 +0000)]
severely restrict what characters can be in a signer, to make
certain it's not possible to go ../  from /etc/signify

10 years agoeven if Ustar handles fifo and stuff gracefully, pkg contents should be more
espie [Tue, 7 Jan 2014 17:33:53 +0000 (17:33 +0000)]
even if Ustar handles fifo and stuff gracefully, pkg contents should be more
limited...

10 years agoPropagate an rdomain number to the nd6_lookup independently from
mikeb [Tue, 7 Jan 2014 17:07:45 +0000 (17:07 +0000)]
Propagate an rdomain number to the nd6_lookup independently from
the ifp pointer which can be NULL.  This prevents a crash reported
by David Hill <dhill at mindcry ! org>.  OK bluhm

10 years agoSome follow-up fixes for IFID collision handling in IPv6CP.
stsp [Tue, 7 Jan 2014 16:34:05 +0000 (16:34 +0000)]
Some follow-up fixes for IFID collision handling in IPv6CP.

Really change the link-local address in the unlikely event of an IFID
collision, instead of going into an infinite conf-nak loop with the peer.

To make the netinet6 code use the IPv6CP IFID in a new link-local address,
in6_ifattach_linklocal() must accept a provided IFID.  Replace the unused
'altifp' parameter with a new 'ifid' parameter for this purpose.

Always use the latest suggested address in IPv6CP replies, even if
the task to update the interface's address hasn't run yet.
Also, clear the ifindex (KAME hack) in addresses sent during IPv6CP.

ok mpi

10 years agouse IF_NAMESIZE as size of buffer for interface name
gsoares [Tue, 7 Jan 2014 14:58:40 +0000 (14:58 +0000)]
use IF_NAMESIZE as size of buffer for interface name

OK mpi@ deraadt@

10 years agoctype cleanups; ok okan
deraadt [Tue, 7 Jan 2014 14:08:16 +0000 (14:08 +0000)]
ctype cleanups; ok okan

10 years agosync
deraadt [Tue, 7 Jan 2014 14:06:16 +0000 (14:06 +0000)]
sync

10 years agosome more devices; Alexey E. Suslikov
deraadt [Tue, 7 Jan 2014 14:06:10 +0000 (14:06 +0000)]
some more devices; Alexey E. Suslikov

10 years agosync
deraadt [Tue, 7 Jan 2014 14:00:54 +0000 (14:00 +0000)]
sync

10 years agoFTX support; from Chris Hettrick
deraadt [Tue, 7 Jan 2014 14:00:50 +0000 (14:00 +0000)]
FTX support; from Chris Hettrick

10 years agoscrape support for old +COMMENT file.
espie [Tue, 7 Jan 2014 13:52:43 +0000 (13:52 +0000)]
scrape support for old +COMMENT file.
8 years of backward compatibility is long enough... :)

10 years agorecognize special file +META
espie [Tue, 7 Jan 2014 13:37:23 +0000 (13:37 +0000)]
recognize special file +META

10 years agotedu some very old checks: we no longer have INSTALL/UNINSTALL scripts,
espie [Tue, 7 Jan 2014 13:31:35 +0000 (13:31 +0000)]
tedu some very old checks: we no longer have INSTALL/UNINSTALL scripts,
so we don't care whether /var/db/pkg is mounted noexec.

10 years agolet's verify all checksums.
espie [Tue, 7 Jan 2014 13:26:17 +0000 (13:26 +0000)]
let's verify all checksums.
I forgot about special files, since they've been extracted early,
but they can be checked nonetheless.

(prevents people from tampering with DESC and the like)

10 years agointegrate signer logic and fix resign with -DSIGNER
espie [Tue, 7 Jan 2014 11:51:15 +0000 (11:51 +0000)]
integrate signer logic and fix resign with -DSIGNER

10 years agodocument new stuff
espie [Tue, 7 Jan 2014 10:44:07 +0000 (10:44 +0000)]
document new stuff

10 years agoif HISTORY_DIR is defined, create a new "permanent" file during build.
espie [Tue, 7 Jan 2014 10:38:12 +0000 (10:38 +0000)]
if HISTORY_DIR is defined, create a new "permanent" file during build.
this file records the sha256 of all elements in the archive, and keeps
track of changes (quite simply: sha256 that were already present end
up at the end of the file).

Start recording these *now*, so that we have enough info to shuffle
packages later.

10 years agosimplify set_modes, there's no need to stat(2) stuff, chown -1 takes care
espie [Tue, 7 Jan 2014 10:22:05 +0000 (10:22 +0000)]
simplify set_modes, there's no need to stat(2) stuff, chown -1 takes care
of things.

10 years agoRemove unused fields.
mpi [Tue, 7 Jan 2014 09:54:18 +0000 (09:54 +0000)]
Remove unused fields.

ok sasano@

10 years agoCache the result of uname(3) such that we don't need to call it
schwarze [Tue, 7 Jan 2014 09:10:58 +0000 (09:10 +0000)]
Cache the result of uname(3) such that we don't need to call it
over and over again for each manual; found with gprof(1).
Speeds up mandocdb(8) -Q by 3%, now at 39.5% of makewhatis(8).

10 years agoIncrease the size of the bge_flags field to allow for more flags.
brad [Tue, 7 Jan 2014 06:25:33 +0000 (06:25 +0000)]
Increase the size of the bge_flags field to allow for more flags.

ok mikeb@

10 years agoDelay checking the vnode locking of the target vnodes in tmpfs_rename()
guenther [Tue, 7 Jan 2014 04:44:56 +0000 (04:44 +0000)]
Delay checking the vnode locking of the target vnodes in tmpfs_rename()
until we've confirmed they're on the same filesystem as the source:
not all filesystems do locking the same.

problem report and testing chrisz@
ok espie@

10 years agoSRCDRS usr.bin usr.sbin is a bad idea (because of PIE). The places we
deraadt [Tue, 7 Jan 2014 02:27:53 +0000 (02:27 +0000)]
SRCDRS usr.bin usr.sbin is a bad idea (because of PIE).  The places we
pick binaries up must be compiled -static or at least in some safe
model for reuse.  Since we now use distrib/special for those, we can
safely remove usr.bin and usr.sbin from this list.

10 years agomore stuff than cannot end in .libs nor partial- packages
espie [Tue, 7 Jan 2014 01:38:58 +0000 (01:38 +0000)]
more stuff than cannot end in .libs nor partial- packages

10 years ago... and let sign_list(\@ARGV... work too)
espie [Tue, 7 Jan 2014 01:35:56 +0000 (01:35 +0000)]
... and let sign_list(\@ARGV... work too)

10 years agoallow processing of package lists in parallel.
espie [Tue, 7 Jan 2014 01:30:28 +0000 (01:30 +0000)]
allow processing of package lists in parallel.
Turns out re-gzipping a signed package is really expensive, so turn
it into several jobs with a -j option.

10 years agonew method: reinitialize, for repositories with a sequential connection
espie [Tue, 7 Jan 2014 01:29:17 +0000 (01:29 +0000)]
new method: reinitialize, for repositories with a sequential connection
that need a new connection if they're ever called from separate processes
simultaneously

10 years agoa few mechanical style changes
espie [Tue, 7 Jan 2014 00:48:24 +0000 (00:48 +0000)]
a few mechanical style changes

10 years agothe return from asprintf does NOT include the space used by the
dlg [Tue, 7 Jan 2014 00:11:11 +0000 (00:11 +0000)]
the return from asprintf does NOT include the space used by the
terminating \0 character. if i want to use it as a c string i have
to take that into account.

found by simon kuhnle who supplied a good bug report. prodded by deraadt@

10 years agoGprof(1) is fun. You should use it more often.
schwarze [Mon, 6 Jan 2014 23:46:01 +0000 (23:46 +0000)]
Gprof(1) is fun.  You should use it more often.

Another 10% speedup for mandocdb(8) -Q, and even 3% without -Q.
With -Q, we are now at 41% of the time required by makewhatis(8).

Do not copy predefined strings into the dynamic string table, just
leave them in their own static table and use that one as a fallback
at lookup time.  This saves us copying and deleting them for each manual.
No functional change.

10 years agoAnother 18% speedup for mandocdb(8) -Q, found by gprof(1).
schwarze [Mon, 6 Jan 2014 22:39:19 +0000 (22:39 +0000)]
Another 18% speedup for mandocdb(8) -Q, found by gprof(1).
In -Q mode, refrain form validating and normalizing the format
of the date given in .Dd or .TH, as it won't be used anyway.

For /usr/share/man, mandocdb -Q now takes 45% of the time of makewhatis(8).

10 years agodocument random buffer must be multiple of 4 in size
deraadt [Mon, 6 Jan 2014 22:34:22 +0000 (22:34 +0000)]
document random buffer must be multiple of 4 in size

10 years agoIn the new PIE world, all dynamic binaries (for instance those found
deraadt [Mon, 6 Jan 2014 22:28:45 +0000 (22:28 +0000)]
In the new PIE world, all dynamic binaries (for instance those found
in usr.bin and usr.sbin on the install media) must be re-compiled non-PIE
(here, in the special directory) so that crunchgen can reuse them.  This
issue is most acute on i386, which has register pressure, and does weird
things for PIE.
thanks for matthew for helping figure this out.

10 years agowe do not need apm or apmd on the install media, that is silly
deraadt [Mon, 6 Jan 2014 22:22:31 +0000 (22:22 +0000)]
we do not need apm or apmd on the install media, that is silly

10 years agoDisable the code disabling memory beyond 1.5GB on Octane temporarily, but
miod [Mon, 6 Jan 2014 21:41:15 +0000 (21:41 +0000)]
Disable the code disabling memory beyond 1.5GB on Octane temporarily, but
hopefully for good; I believe the kernel panics with large memory fast MP
systems have been fixed in mutex.c 1.13

10 years agoUse the dedicated BCD macros from sys/time.h instead of handrolling
pirofti [Mon, 6 Jan 2014 21:38:46 +0000 (21:38 +0000)]
Use the dedicated BCD macros from sys/time.h instead of handrolling
our own for the m41t8x clocks.

Tested by me and jasper@.
Okay jasper@, guenther@, deraadt@

10 years agoAnother 25% speedup for mandocdb(8) -Q mode, found with gprof(1).
schwarze [Mon, 6 Jan 2014 21:33:00 +0000 (21:33 +0000)]
Another 25% speedup for mandocdb(8) -Q mode, found with gprof(1).
For /usr/share/man, we only need 56% of the time of makewhatis(8) now.

In groff, user-defined macros clashing with mdoc(7) or man(7)
standard macros are cleared when parsing the .Dd or .TH macro,
respectively.  Of course, we continue doing that in standard mode
to assure full groff bug compatibility.

However, in -Q mode, full groff bug compatibility makes no sense
when it's unreasonably expensive, so skip this step in -Q mode.
Real-world manuals hardly ever redefine standard macros,
that's terrible style, and if they do, it's pointless to do so
before .Dd or .TH because it has no effect.  Even if someone does,
it's extremely unlikely to break mandocdb(8) -Q parsing because we
abort the parse sequence after the NAME section, anyway.

So if you manually redefine .Sh, .Nm, .Nd, or .SH in a way that doesn't
work at all and rely on .Dd or .TH to fix it up for you, your broken
manual will no longer get a perfect apropos(1) entry until you re-run
mandocdb(8) without -Q.  It think that consequence is acceptable
in order to get a 25% speedup for everyone else.

10 years agotypo
miod [Mon, 6 Jan 2014 21:28:53 +0000 (21:28 +0000)]
typo

10 years agoFix sparc64 kernel profiling, add the etext symbol back.
tobiasu [Mon, 6 Jan 2014 21:16:31 +0000 (21:16 +0000)]
Fix sparc64 kernel profiling, add the etext symbol back.

ok miod@

10 years agoRemove mention of installXX not bootable on IP27/35.
miod [Mon, 6 Jan 2014 21:05:32 +0000 (21:05 +0000)]
Remove mention of installXX not bootable on IP27/35.

10 years agoRework the setup of the bootable installation cd-rom (installXX.iso) to
miod [Mon, 6 Jan 2014 21:03:13 +0000 (21:03 +0000)]
Rework the setup of the bootable installation cd-rom (installXX.iso) to
contain both a 2048-byte sector ffs filesystem, and a 512-byte volume header,
so that the IP27 boot magic^Wuglyness recently added to the boot-only cd-rom
(cdXX.iso) can be applied as well.

The full-blown installation iso can now boot on IP27/28/30/32/35.

10 years agoIn readsgilabel(), do not override the driver-provided initial geometry with
miod [Mon, 6 Jan 2014 21:00:55 +0000 (21:00 +0000)]
In readsgilabel(), do not override the driver-provided initial geometry with
the volume header geometry, if one is found. The driver knows better.

This will allow a 2048-byte sector media to use a volume header using fake
512-byte sectors, and still be able to read the native label off it.

10 years agoDo not sync to disk after each individual manual page (duh!),
schwarze [Mon, 6 Jan 2014 20:53:36 +0000 (20:53 +0000)]
Do not sync to disk after each individual manual page (duh!),
only sync to disk one single time when all data is ready.

Rebuild times for /usr/share/man/mandoc.db shrink on my notebook:
In standard mode from 45 seconds to 11 seconds   (75% reduction)
In -Q mode       from 25 seconds to 3.1 seconds  (87% reduction)
For comparison: makewhatis(8):      4.2 seconds

That is, in -Q mode, we are now *faster* than the existing makewhatis(8),
and careful profiling shows there is still a lot of room for improval.

10 years agosync
miod [Mon, 6 Jan 2014 20:37:18 +0000 (20:37 +0000)]
sync

10 years agosync
miod [Mon, 6 Jan 2014 20:31:12 +0000 (20:31 +0000)]
sync

10 years agoPut PG_EXEC and PG_FOE into the PG_PROT mask, and make sure the default
miod [Mon, 6 Jan 2014 20:27:44 +0000 (20:27 +0000)]
Put PG_EXEC and PG_FOE into the PG_PROT mask, and make sure the default
pte protection masks, as initialized in alpha_protection_init(), set PG_FOE
by default when VM_PROT_EXECUTE is not set.

Also, change pmap_emulate_reference() to only clear PG_FOE if the affected
pte has executable permission.

This allows various pmap_pte_exec() checks (added to explicitely set PG_FOE)
to be removed.

All tests of regress/sys/kern/noexec now reliably pass on EV5. EV6 systems
still see spurious (but no longer 100% reproduceable) failures of the `catch
a signal' tests, which is likely caused by the effect of mprotect() removing
execute permission not taking effect correctly, despite PAL IMB being issued
(and no, this is not caused by the previous pmap_changebit() change), to be
investigated.

10 years agoRemove I-sync stuff from pmap_changebit(). The AARM says that we
miod [Mon, 6 Jan 2014 20:21:32 +0000 (20:21 +0000)]
Remove I-sync stuff from pmap_changebit().  The AARM says that we
only have to sync the I-stream when the mapping is removed or changed,
and since the I-stream is fetch-only, changing protection bits does
not constitute changing the mapping (the VA->PA translation is still
the same).

From NetBSD

10 years agosync
deraadt [Mon, 6 Jan 2014 20:17:02 +0000 (20:17 +0000)]
sync