openbsd
10 years agodisable IO_NOCACHE B_NOCACHE conversion until it works.
tedu [Fri, 9 May 2014 03:48:58 +0000 (03:48 +0000)]
disable IO_NOCACHE B_NOCACHE conversion until it works.

10 years agoKNF: unify style and reduce indentation in get_line(), no binary change
schwarze [Fri, 9 May 2014 03:13:24 +0000 (03:13 +0000)]
KNF: unify style and reduce indentation in get_line(), no binary change
ok cmp(1)

10 years agoAfter entering an invalid three-letter card name, one letter card names
schwarze [Fri, 9 May 2014 02:47:25 +0000 (02:47 +0000)]
After entering an invalid three-letter card name, one letter card names
stopped working because the third letter remained in the buffer,
incard() skipped the NUL and used the old garbage.
Fix this bug reported by pjanzen@, but in a simpler way than he
suggested, by just clearing any trailing garbage from the buffer.
ok pjanzen@

10 years agoVarious cleanup:
schwarze [Fri, 9 May 2014 00:03:41 +0000 (00:03 +0000)]
Various cleanup:
1. Style: Don't use variables as format strings.
NetBSD rev. 1.26 (May 23, 2011) joerg@NetBSD via maintainer pjanzen@.
2. Style: Avoid needless pointer arithmetics.
NetBSD rev. 1.27 (Oct 13, 2012) dholland@NetBSD via pjanzen@,
who also applied some KNF to the indentation while here.
3. End curses before printing fatal error messages; from pjanzen@.

10 years agoDo not read from index -1 of an array.
schwarze [Thu, 8 May 2014 23:12:40 +0000 (23:12 +0000)]
Do not read from index -1 of an array.
NetBSD rev. 1.16 (Oct 13, 2012) by dholland@NetBSD via maintainer pjanzen@.

10 years agoFormat string fixes for m88k; remove -Wno-format from the m88k kernels.
miod [Thu, 8 May 2014 22:17:33 +0000 (22:17 +0000)]
Format string fixes for m88k; remove -Wno-format from the m88k kernels.

10 years agomove reallocarray() to a seperate file so that -portable applications
deraadt [Thu, 8 May 2014 21:43:49 +0000 (21:43 +0000)]
move reallocarray() to a seperate file so that -portable applications
can avoid reinventing the wheel
ok guenther schwarze

10 years agoFormat string fixes and removal of -Wno-format for landisk kernels
miod [Thu, 8 May 2014 21:43:04 +0000 (21:43 +0000)]
Format string fixes and removal of -Wno-format for landisk kernels

10 years agoFormat string fixes and removal of -Wno-format for hppa kernels.
miod [Thu, 8 May 2014 21:32:45 +0000 (21:32 +0000)]
Format string fixes and removal of -Wno-format for hppa kernels.

10 years agoUpdate #include list after not-so-recent uvm includes cleanups.
miod [Thu, 8 May 2014 21:31:56 +0000 (21:31 +0000)]
Update #include list after not-so-recent uvm includes cleanups.

10 years agoFormat string fixes and removal of -Wno-format for arm kernels.
miod [Thu, 8 May 2014 21:17:00 +0000 (21:17 +0000)]
Format string fixes and removal of -Wno-format for arm kernels.

10 years agoFormat string fixes for alpha kernels, and remove -Wno-format
miod [Thu, 8 May 2014 20:46:49 +0000 (20:46 +0000)]
Format string fixes for alpha kernels, and remove -Wno-format

10 years agoFix some potential integer overflows caused by converting a page number into
kettenis [Thu, 8 May 2014 20:08:50 +0000 (20:08 +0000)]
Fix some potential integer overflows caused by converting a page number into
an offset/size/address by shifting by PAGE_SHIFT.  Make uvm_objwrire/unwire
use voff_t instead of off_t.  The former is the right type here even if it is
equivalent to the latter.

Inspired by a somewhat similar changes in Bitrig.

ok deraadt@, guenther@

10 years agoReplace hand-crafted STRICT_ALIGNMENT with system provided __STRICT_ALIGNMENT.
miod [Thu, 8 May 2014 19:59:19 +0000 (19:59 +0000)]
Replace hand-crafted STRICT_ALIGNMENT with system provided __STRICT_ALIGNMENT.
Forgotten during yesterday's STRICT_ALIGNMENT cleanup commit.

10 years agoNo longer build vax kernels with -Wno-format.
miod [Thu, 8 May 2014 19:06:35 +0000 (19:06 +0000)]
No longer build vax kernels with -Wno-format.

10 years agoFormat string fixes.
miod [Thu, 8 May 2014 19:06:07 +0000 (19:06 +0000)]
Format string fixes.

10 years agoRecognize `t' as a valid format modifier for kprintf-style format strings.
miod [Thu, 8 May 2014 18:58:47 +0000 (18:58 +0000)]
Recognize `t' as a valid format modifier for kprintf-style format strings.

10 years agoFormat string fix in disabled code. Format bus_space_handle_t with %xl,
sf [Thu, 8 May 2014 18:09:22 +0000 (18:09 +0000)]
Format string fix in disabled code. Format bus_space_handle_t with %xl,
cast bus_space_tag_t to (u_long)

10 years agoEnable -Wno-format in the kernel on i386 & amd64
sf [Thu, 8 May 2014 17:59:28 +0000 (17:59 +0000)]
Enable -Wno-format in the kernel on i386 & amd64

ok jsg@
"go for it" kettenis@

10 years agoregen
miod [Thu, 8 May 2014 17:38:40 +0000 (17:38 +0000)]
regen

10 years agoRemove irrelevant devices from the ramdisk target; spotted by deraadt
miod [Thu, 8 May 2014 17:38:12 +0000 (17:38 +0000)]
Remove irrelevant devices from the ramdisk target; spotted by deraadt

10 years ago"Server?" -> "HTTP Server?" to allow unambiguous auto-install
krw [Thu, 8 May 2014 16:46:47 +0000 (16:46 +0000)]
"Server?" -> "HTTP Server?" to allow unambiguous auto-install
handling.

Confusion with "NTP Server?" reported by Xavier Claude via misc@.

ok rpe@ halex@

10 years agofail for unsupported node action/type combinations. Also fail for the
reyk [Thu, 8 May 2014 16:11:06 +0000 (16:11 +0000)]
fail for unsupported node action/type combinations.  Also fail for the
unsupported mark/marked combination in a single rule.

ok andre@

10 years agoMore KNF.
jsing [Thu, 8 May 2014 16:05:38 +0000 (16:05 +0000)]
More KNF.

10 years agoKNF.
jsing [Thu, 8 May 2014 16:01:06 +0000 (16:01 +0000)]
KNF.

10 years agoAdjust this test contents from path keytype to url keytype to match the name of
andre [Thu, 8 May 2014 15:55:32 +0000 (15:55 +0000)]
Adjust this test contents from path keytype to url keytype to match the name of
the test file.

ok reyk

10 years agoprotip: "It's largely bad style to do (int)sizeof"
tedu [Thu, 8 May 2014 15:48:41 +0000 (15:48 +0000)]
protip: "It's largely bad style to do (int)sizeof"

amusingly, theo phrased this with considerably more restraint than i did.

10 years agoRemove KSSL debug code.
jsing [Thu, 8 May 2014 15:46:24 +0000 (15:46 +0000)]
Remove KSSL debug code.

"fire bomb" tedu@

10 years agoKNF.
jsing [Thu, 8 May 2014 15:42:26 +0000 (15:42 +0000)]
KNF.

10 years agoNuke OPENSSL_FIPS - having #ifdefs inside a #ifndef for the same thing
jsing [Thu, 8 May 2014 15:29:00 +0000 (15:29 +0000)]
Nuke OPENSSL_FIPS - having #ifdefs inside a #ifndef for the same thing
is amusing.

ok deraadt@

10 years agoremove debug prints that snuck in; found by reyk@
blambert [Thu, 8 May 2014 15:28:57 +0000 (15:28 +0000)]
remove debug prints that snuck in; found by reyk@

10 years agoKNF.
jsing [Thu, 8 May 2014 15:13:06 +0000 (15:13 +0000)]
KNF.

10 years agoPretty print MIB_ipNetToMediaPhysAddress'es; ok blambert
mikeb [Thu, 8 May 2014 14:48:31 +0000 (14:48 +0000)]
Pretty print MIB_ipNetToMediaPhysAddress'es;  ok blambert

10 years agoPlug one more xfer leak.
mpi [Thu, 8 May 2014 14:00:52 +0000 (14:00 +0000)]
Plug one more xfer leak.

Now that aborting interrupt pipes does not prevent us from freeing
the associated xfer, make sure to flag this xfer as "done" even if
there's no need to abort it in hardware.

10 years agoknfectomie.
jsing [Thu, 8 May 2014 13:48:36 +0000 (13:48 +0000)]
knfectomie.

10 years agoThe pcexmem(4) and pcexio(4) should be attached on LUNA-88K2 only.
aoyama [Thu, 8 May 2014 13:31:00 +0000 (13:31 +0000)]
The pcexmem(4) and pcexio(4) should be attached on LUNA-88K2 only.

Actually, there seems to be no PC-9801 extension board slot on
'original' LUNA-88K.

ok miod@

10 years agoEmergency knfectomie requested by tedu@.
jsing [Thu, 8 May 2014 13:20:49 +0000 (13:20 +0000)]
Emergency knfectomie requested by tedu@.

10 years agomatch iked proc.c infrastructure with proc.c
blambert [Thu, 8 May 2014 13:11:16 +0000 (13:11 +0000)]
match iked proc.c infrastructure with proc.c

ok reyk@

10 years agomatch relayd proc.c infrastructure with snmpd
blambert [Thu, 8 May 2014 13:08:48 +0000 (13:08 +0000)]
match relayd proc.c infrastructure with snmpd

okay reyk@

10 years agoinet(3) -> inet_net(3);
jmc [Thu, 8 May 2014 13:08:43 +0000 (13:08 +0000)]
inet(3) -> inet_net(3);

10 years agoBring back restricted sockets, inadvertantly removed in
blambert [Thu, 8 May 2014 13:04:28 +0000 (13:04 +0000)]
Bring back restricted sockets, inadvertantly removed in
previous proc.c update

ok reyk@

10 years agokill some more VMS ifdefs
giovanni [Thu, 8 May 2014 12:59:17 +0000 (12:59 +0000)]
kill some more VMS ifdefs
ok miod@

10 years agoSync flags with route.h
mpi [Thu, 8 May 2014 09:30:48 +0000 (09:30 +0000)]
Sync flags with route.h

10 years agoIntroduce two new route flags: RTF_LOCAL and RTF_BROADCAST.
mpi [Thu, 8 May 2014 09:28:08 +0000 (09:28 +0000)]
Introduce two new route flags: RTF_LOCAL and RTF_BROADCAST.

Nothing use them for the moment, but here is the plan:

Since a route lookup is always necessary to output a packet it makes
sense to store all the information regarding how the packet should be
sent in the routing entry.  This will save us some expensive lookups
on address lists.

But once we have all the information about our addresses in the routing
table, we can even use it in the input path with the hope that the number
of lookups in the forwarding case can be reduce to one.

ok henning@, chris@

10 years agoSend up and down keys for mouse wheel in alternate screen mode (when it
nicm [Thu, 8 May 2014 07:59:16 +0000 (07:59 +0000)]
Send up and down keys for mouse wheel in alternate screen mode (when it
normally does nothing), from Marcel Partap.

10 years agoInstead of forcing mouse scroll to 1 in choose mode, scale it down
nicm [Thu, 8 May 2014 07:54:47 +0000 (07:54 +0000)]
Instead of forcing mouse scroll to 1 in choose mode, scale it down
instead. Means modifier keys still increase the line count, just not as
much. Based on a diff from Marcel Partap.

10 years agowhack some useless verbiage from DESCRIPTION, and merge some of it into
jmc [Thu, 8 May 2014 06:50:50 +0000 (06:50 +0000)]
whack some useless verbiage from DESCRIPTION, and merge some of it into
the env vars section; tidy up the env vars section and remove the commented
out stuff which relataes mainly to less on other systems;

10 years ago- update DESCRIPTION and COMPATIBILITY WITH MORE
jmc [Thu, 8 May 2014 06:24:14 +0000 (06:24 +0000)]
- update DESCRIPTION and COMPATIBILITY WITH MORE
- Xr more 1

ok millert schwarze

10 years ago+.Xr less 1 ,
jmc [Thu, 8 May 2014 06:22:07 +0000 (06:22 +0000)]
+.Xr less 1 ,

10 years agoPlug a memory leak, from J Raynor.
nicm [Thu, 8 May 2014 06:06:07 +0000 (06:06 +0000)]
Plug a memory leak, from J Raynor.

10 years agoBoth the two previous ways of navigating panes by direction have
nicm [Thu, 8 May 2014 06:03:30 +0000 (06:03 +0000)]
Both the two previous ways of navigating panes by direction have
irritating flaws:

a) The old way of always using the top or left if the choice is
ambiguous is annoying when the layout is unbalanced.

b) The new way of remembering the last used pane is annoying if the
layout is balanced and the leftmost is obvious to the user (because
clearly if we go right from the top-left in a tiled set of four we want
to end up in top-right, even if we were last using the bottom-right).

So instead, use a combination of both: if there is only one possible
pane alongside the current pane, move to it, otherwise choose the most
recently used of the choice.

10 years agoHandle colour 8 properly in the 256 colour palette, from Timothy Allen.
nicm [Thu, 8 May 2014 05:53:29 +0000 (05:53 +0000)]
Handle colour 8 properly in the 256 colour palette, from Timothy Allen.

10 years agoGet __STRICT_ALIGNMENT from <machine/endian.h> and decide upon it, rather
miod [Wed, 7 May 2014 22:05:48 +0000 (22:05 +0000)]
Get __STRICT_ALIGNMENT from <machine/endian.h> and decide upon it, rather
than defining it for not (i386 and amd64 (and sometimes s390)) only.

Compile-time tests remain compile-time tests, and runtime-test remain
runtime-test instead of being converted to compile-time tests, per matthew@'s
explicit demand (rationale: this makes sure the compiler checks your code even
if you won't run it).

No functional change except on s390 (which we don't run on) and vax (which we
run on, but noone cares about)

ok matthew@

10 years agoWhen you want more, you say "more". When you want less, you say "less".
zhuk [Wed, 7 May 2014 21:42:50 +0000 (21:42 +0000)]
When you want more, you say "more".  When you want less, you say "less".
When you don't get what you asked for, you get angry.
When you add a tiny bit, it should be "more" or "less" still,
  or you'll get angry.
So lets make zmore(1) call more(1) and zless(1) call less(1),
  as it's supposed to be.

okay and input from ingo@, no objections from author AKA millert@

10 years agoless than jpake
tedu [Wed, 7 May 2014 21:32:19 +0000 (21:32 +0000)]
less than jpake

10 years agoRepair the termination condition of a write(2) loop.
schwarze [Wed, 7 May 2014 21:20:06 +0000 (21:20 +0000)]
Repair the termination condition of a write(2) loop.

Since _PATH_MASTERPASSWD_LOCK is on a local file system in any sane setup
and written to in blocking mode, i don't see how write(2) could return
before having written everything, so this maybe wasn't an actual bug,
but it should be repaired anyway, if only for clarity and extra safety.

From Ben Cornett <ben at lantern dot is>;
ok millert@.

10 years agodon't free memory unless we allocated it.
tedu [Wed, 7 May 2014 21:18:53 +0000 (21:18 +0000)]
don't free memory unless we allocated it.
reported to openssl by David Ramos (rt 3338)
stealing commit from logan, who's run away. ok miod.

10 years agowith the first bug it uncovered fixed, clear all bignums again.
tedu [Wed, 7 May 2014 21:09:03 +0000 (21:09 +0000)]
with the first bug it uncovered fixed, clear all bignums again.

i've never worked on codebase so resistant to efforts to improve it.

10 years agoin BN_clear_free, don't cleanse the data if the static data flag is set.
tedu [Wed, 7 May 2014 21:06:05 +0000 (21:06 +0000)]
in BN_clear_free, don't cleanse the data if the static data flag is set.
much debugging work done by otto. ok miod otto.

side note: BN_FLG_STATIC_DATA doesn't actually mean the data is static.
it's also used to indicate the data may be secretly shared behind your back
as a sort of poor man's refcounting, but without the refcounting.

10 years agocomment style fix
halex [Wed, 7 May 2014 20:07:59 +0000 (20:07 +0000)]
comment style fix

ok crickets@

10 years agojmc@ points out '-F/-f' are available only in conjunction with
krw [Wed, 7 May 2014 18:55:52 +0000 (18:55 +0000)]
jmc@ points out '-F/-f' are available only in conjunction with
'-E' and '-R'.

10 years agoMatch search_string_def() prototype change, forgotten in previous protector.[ch]
miod [Wed, 7 May 2014 18:51:33 +0000 (18:51 +0000)]
Match search_string_def() prototype change, forgotten in previous protector.[ch]
commit.

10 years agoKNF.
jsing [Wed, 7 May 2014 17:41:41 +0000 (17:41 +0000)]
KNF.

10 years agoRender roff escape sequences contained in manual page descriptions
schwarze [Wed, 7 May 2014 16:18:57 +0000 (16:18 +0000)]
Render roff escape sequences contained in manual page descriptions
before putting them into the mpages table.
Issue found by bentley@ in OpenBSD::Getopt(3p).

10 years agoFix two memory leaks in makewhatis -n:
schwarze [Wed, 7 May 2014 15:10:33 +0000 (15:10 +0000)]
Fix two memory leaks in makewhatis -n:
1. As found by nigel@, names_check() requires database access.
2. Do not leak names and strings in -n mode.

10 years agoconsolidate some code by using reallocarray in all cases.
tedu [Wed, 7 May 2014 14:59:11 +0000 (14:59 +0000)]
consolidate some code by using reallocarray in all cases.
ok deraadt millert

10 years agouse reallocarray. ok guenther
tedu [Wed, 7 May 2014 14:56:57 +0000 (14:56 +0000)]
use reallocarray. ok guenther

10 years agoFix whitespace mistakes.
pirofti [Wed, 7 May 2014 14:44:54 +0000 (14:44 +0000)]
Fix whitespace mistakes.

10 years agoDo not segfault in makewhatis -Q if the next .SH after .SH NAME
schwarze [Wed, 7 May 2014 14:11:24 +0000 (14:11 +0000)]
Do not segfault in makewhatis -Q if the next .SH after .SH NAME
does not have any arguments.  Crash found by nigel@ in kermit(1).

10 years agorevert, thanks sthen
tedu [Wed, 7 May 2014 13:45:29 +0000 (13:45 +0000)]
revert, thanks sthen

10 years agoback out the previous ICMP simplifying diff, it caused livelocks.
pelikan [Wed, 7 May 2014 13:20:47 +0000 (13:20 +0000)]
back out the previous ICMP simplifying diff, it caused livelocks.

reported by Mikolaj Kucharski, thanks!
ok krw

10 years agotry postponed requests first, so we do in-order processing; ok mikeb@
markus [Wed, 7 May 2014 13:09:43 +0000 (13:09 +0000)]
try postponed requests first, so we do in-order processing; ok mikeb@

10 years agoprint msgid for debugging; ok reyk & mikeb
markus [Wed, 7 May 2014 13:04:01 +0000 (13:04 +0000)]
print msgid for debugging; ok reyk & mikeb

10 years agomake authentication work with X509 certificates that don't have a
markus [Wed, 7 May 2014 12:57:13 +0000 (12:57 +0000)]
make authentication work with X509 certificates that don't have a
subject-altname, i.e. support IKEV2_ID_ASN1_DN correctly;
feedback & ok mikeb@

10 years agoRemove the undocumented and ineffective -F option (force hard link
schwarze [Wed, 7 May 2014 12:37:04 +0000 (12:37 +0000)]
Remove the undocumented and ineffective -F option (force hard link
to directory).  This doesn't reduce functionality but merely results
in a better error message when trying to use the option,
and it simplifies the code.

The -f option first appeared in AT&T Version 7 UNIX (1979), and Keith
Bostic renamed it to -F for 4.3BSD-Reno in 1990 because it conflicted
with System V and POSIX.  Meanwhile, NetBSD, FreeBSD, and DragonFly
removed it, too.

From Tristan Le Guern <tleguern at bouledef dot eu>.
OK guenther@ krw@

10 years agofactor out ikev2_ike_auth() (state machine; used multiple times via callbacks)
markus [Wed, 7 May 2014 10:52:47 +0000 (10:52 +0000)]
factor out ikev2_ike_auth() (state machine; used multiple times via callbacks)
from ikev2_ike_auth_recv() code (message parsing; used once); ok mikeb@

10 years agoadd additional includes from eng_cryptodev.c
jsg [Wed, 7 May 2014 08:31:30 +0000 (08:31 +0000)]
add additional includes from eng_cryptodev.c
makes this compile with OPENSSL_NO_DEPRECATED defined.
ok deraadt@

10 years agoRemove the last hacks concerning the global list of IPv4 addresses in the
mpi [Wed, 7 May 2014 08:26:38 +0000 (08:26 +0000)]
Remove the last hacks concerning the global list of IPv4 addresses in the
source address selection logic.

These hacks were only relevant for the NFS diskless boot code in order to
pick the local broadcast address of the only configured interface.  So, be
explicit and set this address directly.

Tested by florian@, ok henning@, beck@, chrisz@

10 years agoMake sure to call uhidev_close() upon detach for every HID driver calling
mpi [Wed, 7 May 2014 08:17:21 +0000 (08:17 +0000)]
Make sure to call uhidev_close() upon detach for every HID driver calling
uhidev_open() at attach time.  This plugs up to 3 xfer leaks and a buffer
one.

ok yuo@

10 years agoKill the {nd6_,}useloopback buttons, using the loopback interface for
mpi [Wed, 7 May 2014 08:14:59 +0000 (08:14 +0000)]
Kill the {nd6_,}useloopback buttons, using the loopback interface for
local traffic is not optional.

ok mikeb@, stsp@, jca@

10 years agoRemove some m->m_pkthdr.rcvif dereference to help for upcoming
mpi [Wed, 7 May 2014 08:09:33 +0000 (08:09 +0000)]
Remove some m->m_pkthdr.rcvif dereference to help for upcoming
receiving pointer -> index conversion.  No functional change.

ok chrisz@, jca@, mikeb@, lteo@

10 years agoInitialize struct iface by zero. Without this, npppd had used random
yasuoka [Wed, 7 May 2014 08:07:53 +0000 (08:07 +0000)]
Initialize struct iface by zero.  Without this, npppd had used random
configuration.

from Anders Berggren.

10 years agoFix ext2fs atime and mtime handling that I broke from an airport
guenther [Wed, 7 May 2014 02:57:41 +0000 (02:57 +0000)]
Fix ext2fs atime and mtime handling that I broke from an airport
lounge after n2k14.

Prototype patch from zhuk@
ok zhuk@ krw@ tedu@

10 years agopass daemon_flags to nsd-control when used to check/reload/stop nsd,
sthen [Wed, 7 May 2014 02:46:05 +0000 (02:46 +0000)]
pass daemon_flags to nsd-control when used to check/reload/stop nsd,
the only useful option here is to specify an alternative config path,
which must be used for these operations as well as for startup.

10 years agomissing word in error message. ok millert
tedu [Wed, 7 May 2014 01:49:36 +0000 (01:49 +0000)]
missing word in error message. ok millert

10 years agodelete some crusty casts
tedu [Wed, 7 May 2014 01:41:18 +0000 (01:41 +0000)]
delete some crusty casts

10 years agoconvert calloc to use reallocarray instead. ok millert
tedu [Wed, 7 May 2014 01:31:25 +0000 (01:31 +0000)]
convert calloc to use reallocarray instead. ok millert

10 years agocalloc, from peter malone
tedu [Wed, 7 May 2014 01:27:42 +0000 (01:27 +0000)]
calloc, from peter malone

10 years agouse calloc. from peter malone
tedu [Wed, 7 May 2014 01:23:52 +0000 (01:23 +0000)]
use calloc. from peter malone

10 years agocalloc is simpler. suggested by Peter Malone and halex
tedu [Wed, 7 May 2014 01:20:53 +0000 (01:20 +0000)]
calloc is simpler. suggested by Peter Malone and halex

10 years agouse calloc. from Peter Malone
tedu [Wed, 7 May 2014 01:14:21 +0000 (01:14 +0000)]
use calloc. from Peter Malone

10 years agomiod forgot about fcrypt_b.c being needed, before he went off to get his
deraadt [Wed, 7 May 2014 00:04:14 +0000 (00:04 +0000)]
miod forgot about fcrypt_b.c being needed, before he went off to get his
beauty sleep.  He's probably having a nightmare about this right now....
ok tedu

10 years agotemp variable for a bit of clarity
tedu [Tue, 6 May 2014 23:50:53 +0000 (23:50 +0000)]
temp variable for a bit of clarity

10 years agoremove unneeded variable
tedu [Tue, 6 May 2014 23:33:04 +0000 (23:33 +0000)]
remove unneeded variable

10 years agoWhen the stack protector heuristics doesn't cover a function, leave
martynas [Tue, 6 May 2014 23:32:34 +0000 (23:32 +0000)]
When the stack protector heuristics doesn't cover a function, leave
a little pointer-sized gap before the return value.  This protects
from common off-by-one type of bugs and costs nothing:  the attacker
won't be able to overwrite return pointer.  Developed at m2k14,
thanks for the hackathon!

10 years agofactor out the safepath
tedu [Tue, 6 May 2014 23:27:50 +0000 (23:27 +0000)]
factor out the safepath

10 years agobe more explicit about what's a string
tedu [Tue, 6 May 2014 23:24:19 +0000 (23:24 +0000)]
be more explicit about what's a string

10 years agoIntroduce -fstack-shuffle, which randomizes local stack variables.
martynas [Tue, 6 May 2014 23:22:33 +0000 (23:22 +0000)]
Introduce -fstack-shuffle, which randomizes local stack variables.
This will make the environment more hostile and help detect bugs
that depend on overrunning one variable into another, with almost
no performance cost.

Discussed with Theo at m2k14 hackathon.  "oh god yes" tedu@, "oh nice" djm@

10 years agobetter alloc idiom
tedu [Tue, 6 May 2014 23:19:46 +0000 (23:19 +0000)]
better alloc idiom

10 years agofactor max message size
tedu [Tue, 6 May 2014 23:17:13 +0000 (23:17 +0000)]
factor max message size