openbsd
10 years agorearrange/correct timeout conditionals to work better.
tedu [Tue, 24 Dec 2013 23:29:38 +0000 (23:29 +0000)]
rearrange/correct timeout conditionals to work better.
fixes negative timeout panics. tested by sthen.

10 years agoDelete the unused flag TERMP_IGNDELIM
schwarze [Tue, 24 Dec 2013 23:04:29 +0000 (23:04 +0000)]
Delete the unused flag TERMP_IGNDELIM
and the empty callback termp_igndelim_pre().
Sort the remaining termp flags.

10 years agomore fixing after circleq conversion. a better fix to check the prev
tedu [Tue, 24 Dec 2013 22:26:19 +0000 (22:26 +0000)]
more fixing after circleq conversion. a better fix to check the prev
pointer is forthcoming.

10 years agoImplement SYNOPSIS .Fn indentation for -Tman.
schwarze [Tue, 24 Dec 2013 22:08:23 +0000 (22:08 +0000)]
Implement SYNOPSIS .Fn indentation for -Tman.

10 years agoIt turns out SYNOPSIS mode does not imply .Bk in general,
schwarze [Tue, 24 Dec 2013 20:45:21 +0000 (20:45 +0000)]
It turns out SYNOPSIS mode does not imply .Bk in general,
but only within .Nm blocks.  Simplify the code accordingly.
This reduces groff-mandoc differences in base by about 2%.

Triggered by research done by Franco Fichtner.

10 years agoWhen deciding whether two consecutive macros are on the same input line,
schwarze [Tue, 24 Dec 2013 19:10:34 +0000 (19:10 +0000)]
When deciding whether two consecutive macros are on the same input line,
we have to compare the line where the first one *ends* (not where it begins)
to the line where the second one starts.
This fixes the bug that .Bk allowed output line breaks right after block
macros spanning more than one input line, even when the next macro follows
on the same line.

10 years agoMore .Bk tests to lessen the risk of upcoming work.
schwarze [Tue, 24 Dec 2013 16:01:15 +0000 (16:01 +0000)]
More .Bk tests to lessen the risk of upcoming work.

10 years agoWhen i replaced the very ugly .in: suffix rule in my previous commit,
schwarze [Tue, 24 Dec 2013 15:58:07 +0000 (15:58 +0000)]
When i replaced the very ugly .in: suffix rule in my previous commit,
that inadvertently disabled the -Tman tests.  Oops.
Enable them again.
Fortunately, in the meantime, nothing slipped.

10 years agosync
deraadt [Tue, 24 Dec 2013 14:17:02 +0000 (14:17 +0000)]
sync

10 years agoIf the FADT has its SMI_CMD set to zero, assume we're only ACPI-only hardware
kettenis [Tue, 24 Dec 2013 13:23:21 +0000 (13:23 +0000)]
If the FADT has its SMI_CMD set to zero, assume we're only ACPI-only hardware
and don't need to disable SMI ownership of the ACPI hardware registers.

ok mlarkin@

10 years agoAdd support for SSL/TLS server certificate validation, enabled by
jca [Tue, 24 Dec 2013 13:00:59 +0000 (13:00 +0000)]
Add support for SSL/TLS server certificate validation, enabled by
default.  See the documentation for the `-S' switch.  This also allows
setting the preferred ciphers for the communication.  Documentation bits
ok'ed by jmc@, ok beck@ sthen@.

10 years agoRemove some leftover bits that are no longer necessary now that we no
kettenis [Tue, 24 Dec 2013 10:53:37 +0000 (10:53 +0000)]
Remove some leftover bits that are no longer necessary now that we no
longer call atexit(4) directly on hppa.

10 years agosync
deraadt [Tue, 24 Dec 2013 05:31:48 +0000 (05:31 +0000)]
sync

10 years agoyup, we do ELF
deraadt [Tue, 24 Dec 2013 05:31:13 +0000 (05:31 +0000)]
yup, we do ELF

10 years agoget rid of if (timeout_pending()) timeout_del(). this is racy. any
dlg [Tue, 24 Dec 2013 01:11:04 +0000 (01:11 +0000)]
get rid of if (timeout_pending()) timeout_del(). this is racy. any
conditionals you did on timeout_pending can now be done on timeout_del
now that it returns what it did.

ok and a very good fix from kettenis@

10 years agorename local ticks to nticks to avoid aliasing global. ok krw
tedu [Tue, 24 Dec 2013 01:11:00 +0000 (01:11 +0000)]
rename local ticks to nticks to avoid aliasing global. ok krw

10 years agofix error checking oddity in msdosfs code, as noted by kettenis@
halex [Tue, 24 Dec 2013 00:18:46 +0000 (00:18 +0000)]
fix error checking oddity in msdosfs code, as noted by kettenis@

ok mikeb@ espie@

10 years agoremove junk
deraadt [Mon, 23 Dec 2013 23:32:40 +0000 (23:32 +0000)]
remove junk

10 years agospelling
deraadt [Mon, 23 Dec 2013 23:23:22 +0000 (23:23 +0000)]
spelling

10 years agoinstall a link as sha512. simplify program internals; there are only
tedu [Mon, 23 Dec 2013 23:00:38 +0000 (23:00 +0000)]
install a link as sha512. simplify program internals; there are only
two modes. ok deraadt gilles jca

10 years agofix circleq tailq conversion
tedu [Mon, 23 Dec 2013 22:39:50 +0000 (22:39 +0000)]
fix circleq tailq conversion

10 years agoUse slightly diffrerent code to get the global offset table address. This
kettenis [Mon, 23 Dec 2013 22:24:37 +0000 (22:24 +0000)]
Use slightly diffrerent code to get the global offset table address.  This
version will match the (upcoming) 32-bit version (for sparc) and allegedly
is slightly faster.

10 years agoMake ld.so pass its cleanup handler in %rdx as required by the AMD64 System V
kettenis [Mon, 23 Dec 2013 22:13:03 +0000 (22:13 +0000)]
Make ld.so pass its cleanup handler in %rdx as required by the AMD64 System V
ABI, and stop calling atexit(4) directly from ld.so on amd64.  Note that this
means that binaries built before the middle of november may no longer call
destructors in shared libraries implemented in C++.  Update your packages!

ok guenther@

10 years agoMake ld.so pass its cleanup handler in r7 as required by the PowerPC System V
kettenis [Mon, 23 Dec 2013 22:03:52 +0000 (22:03 +0000)]
Make ld.so pass its cleanup handler in r7 as required by the PowerPC System V
ABI, and stop calling atexit(4) directly from ld.so on powerpc.

10 years agoFix indentation.
kettenis [Mon, 23 Dec 2013 21:49:27 +0000 (21:49 +0000)]
Fix indentation.

10 years agoFix two dotline bugs for the price of one.
florian [Mon, 23 Dec 2013 21:47:32 +0000 (21:47 +0000)]
Fix two dotline bugs for the price of one.
1) When answering "No newline at end of file, add one?" with y and
   dot is not at the end of buffer dotline is off by one
2) When displaying the same buffer in two windows, dot in the
   second window is below dot in the first and a newline is entered in
   the first window dotline is not updated in the second.

OK lum

10 years ago- fix gamma ramp (now white is real white instead of light grey)
miod [Mon, 23 Dec 2013 21:32:30 +0000 (21:32 +0000)]
- fix gamma ramp (now white is real white instead of light grey)
- speed-up the integer to single precision float routine by omitting the
  sign check and starting the log2 computation at 12 instead of 32, since
  all the values passed to this routine are actually on-screen coordinates.

10 years agoMake ld.so pass its cleanup handler in v0 and stop calling atexit(4) directly
kettenis [Mon, 23 Dec 2013 21:18:57 +0000 (21:18 +0000)]
Make ld.so pass its cleanup handler in v0 and stop calling atexit(4) directly
from ld.so on alpha.  Note that this means that binaries built before early
december may no longer call destructors in shared libraries implemented in
C++.  Update your packages!

10 years agosync specvops with ffs (fixing spec_open along the way)
tedu [Mon, 23 Dec 2013 20:35:19 +0000 (20:35 +0000)]
sync specvops with ffs (fixing spec_open along the way)
ok espie

10 years agoAdd an instruction barrier to bus_space_barrier(). From David Hill.
kettenis [Mon, 23 Dec 2013 20:32:31 +0000 (20:32 +0000)]
Add an instruction barrier to bus_space_barrier().  From David Hill.
ok deraadt@

10 years agoAdd assembly code to set up the arguments for the common C startup code. This
kettenis [Mon, 23 Dec 2013 18:16:39 +0000 (18:16 +0000)]
Add assembly code to set up the arguments for the common C startup code.  This
means we no longer rely on how the C compiler sets up the initial stack frame
and allows us to get rid of MD_NO_CLEANUP.

ok deraadt@, guenther@

10 years agoMake ld.so pass its cleanup handler in %g1 as required by the SPARC System V
kettenis [Mon, 23 Dec 2013 17:29:15 +0000 (17:29 +0000)]
Make ld.so pass its cleanup handler in %g1 as required by the SPARC System V
ABI, and stop calling atexit(4) directly from ld.so on sparc64

10 years agoMove KASSERT a bit further down because it can apparently be triggered by
kettenis [Mon, 23 Dec 2013 17:23:51 +0000 (17:23 +0000)]
Move KASSERT a bit further down because it can apparently be triggered by
trying to hardlink a directory.  While this is not allowed on tmpfs, this
allows us to reach the code that checks for this such that we can return
the appropriate error.

Cluestick from Pedro Martelletto.
ok espie@

10 years agoCope with the change in ftp(1) warning messages since last commit.
jca [Mon, 23 Dec 2013 16:55:00 +0000 (16:55 +0000)]
Cope with the change in ftp(1) warning messages since last commit.
Issue reported by Maxime Villard, ok espie@.

10 years agomove ntogo code, so it can be used while copying packages in pkg_create.
espie [Mon, 23 Dec 2013 16:50:29 +0000 (16:50 +0000)]
move ntogo code, so it can be used while copying packages in pkg_create.
add the few lines necessary for that to work.

10 years agoget rid of "copy_long", we don't really need to recheck names while copying
espie [Mon, 23 Dec 2013 16:34:51 +0000 (16:34 +0000)]
get rid of "copy_long", we don't really need to recheck names while copying
a package contents.

Allow -o/-S for output/signature dirs for batch operations.
Fix copying of packages while signing, add progressmeter.

10 years agosimpler logic: if one package is signed, check all digests
espie [Mon, 23 Dec 2013 15:40:24 +0000 (15:40 +0000)]
simpler logic: if one package is signed, check all digests

10 years agoreal fix: turns out visit_with_* are transparent, e.g., pass all their
espie [Mon, 23 Dec 2013 15:04:37 +0000 (15:04 +0000)]
real fix: turns out visit_with_* are transparent, e.g., pass all their
parameters to the end consumer.  Make sure the !Term version obeys that too.

10 years agoMove findbuffer() to buffer.c.
lum [Mon, 23 Dec 2013 14:58:16 +0000 (14:58 +0000)]
Move findbuffer() to buffer.c.
ok florian@

10 years agoSync Makefile with what other platforms do; makes dependency tracking actually
kettenis [Mon, 23 Dec 2013 14:22:12 +0000 (14:22 +0000)]
Sync Makefile with what other platforms do; makes dependency tracking actually
work.

ok deraadt@, guenther@

10 years ago- clarify that sio_close() stops the device first, then
ratchov [Mon, 23 Dec 2013 14:14:31 +0000 (14:14 +0000)]
- clarify that sio_close() stops the device first, then
  frees the handle and not the opposite
- clarify that sio_close() and sio_stop() dont discard
  samples in the play buffer.

suggested by Reimar Doeffinger <Reimar.Doeffinger at gmx.de>

ok jmc@

10 years agoCall __cxa_atexit(3) instead of atexit(3). Preparation for upcoming changes to
kettenis [Mon, 23 Dec 2013 14:13:03 +0000 (14:13 +0000)]
Call __cxa_atexit(3) instead of atexit(3).  Preparation for upcoming changes to
make it possible to call atexit from dynamically loaded shared objects.

10 years agoand I fucked up that one too...
espie [Mon, 23 Dec 2013 14:03:58 +0000 (14:03 +0000)]
and I fucked up that one too...

10 years agoMake ld.so pass its cleanup handler in %arg1 and stop calling atiexit(4)
kettenis [Mon, 23 Dec 2013 13:57:44 +0000 (13:57 +0000)]
Make ld.so pass its cleanup handler in %arg1 and stop calling atiexit(4)
directly from ld.so on hppa.

10 years agoThe xbase set requires the comp set. If xbase is in the selection,
rpe [Mon, 23 Dec 2013 13:57:05 +0000 (13:57 +0000)]
The xbase set requires the comp set. If xbase is in the selection,
but not comp, put comp back in automatically.

suggested by and ok deraadt@
ok halex@

10 years agointerface bug: visit_with_size has a state parameter.
espie [Mon, 23 Dec 2013 13:29:41 +0000 (13:29 +0000)]
interface bug: visit_with_size has a state parameter.
I've been lucky that I've not invoked it with more parameters so far...

10 years agoWhen no address for the current AF was specified with -a generate pass
florian [Mon, 23 Dec 2013 13:08:25 +0000 (13:08 +0000)]
When no address for the current AF was specified with -a generate pass
rules instead of erroring out.
ok dlg, benno, camield

10 years agoMake it possible to give -a multiple times on the command line.
florian [Mon, 23 Dec 2013 13:07:47 +0000 (13:07 +0000)]
Make it possible to give -a multiple times on the command line.
Initial diff by camield, tweaked by me.
ok dlg, benno, camield
manpage input/ok jmc

10 years agoImplement -a option in tftp-proxy to make it work on a NAT gateway.
florian [Mon, 23 Dec 2013 13:06:53 +0000 (13:06 +0000)]
Implement -a option in tftp-proxy to make it work on a NAT gateway.
input/ok dlg, benno, camield
manpage input/ok jmc

10 years agoregen
brad [Mon, 23 Dec 2013 12:46:46 +0000 (12:46 +0000)]
regen

10 years agoAdd Atheros AR9462 and AR9565
brad [Mon, 23 Dec 2013 12:18:37 +0000 (12:18 +0000)]
Add Atheros AR9462 and AR9565

10 years agofix IPv6 in proxy_dst6()
benno [Mon, 23 Dec 2013 11:45:39 +0000 (11:45 +0000)]
fix IPv6 in proxy_dst6()
ok florian camield dlg

10 years agoRevert previous commit. It makes resume hang on some MP systems running
kettenis [Mon, 23 Dec 2013 10:48:43 +0000 (10:48 +0000)]
Revert previous commit.  It makes resume hang on some MP systems running
GENERIC.MP.

10 years agoMake ld.so pass its cleanup handler in %edx as required by the i386 System V
kettenis [Mon, 23 Dec 2013 10:39:10 +0000 (10:39 +0000)]
Make ld.so pass its cleanup handler in %edx as required by the i386 System V
ABI, and stop calling atexit(4) directly from ld.so on i386.

ok guenther@

10 years agomlink TASK_INITIALIZER
dlg [Mon, 23 Dec 2013 04:25:42 +0000 (04:25 +0000)]
mlink TASK_INITIALIZER

10 years agodocument TASK_INITIALIZER()
dlg [Mon, 23 Dec 2013 04:24:43 +0000 (04:24 +0000)]
document TASK_INITIALIZER()

10 years agoprovide a TASK_INITIALIZER() macro.
dlg [Mon, 23 Dec 2013 04:20:47 +0000 (04:20 +0000)]
provide a TASK_INITIALIZER() macro.

10 years agoImplement a long-standing desideratum,
schwarze [Mon, 23 Dec 2013 02:19:57 +0000 (02:19 +0000)]
Implement a long-standing desideratum,
hanging indentation for .Fn in SYNOPSIS mode,
exploiting the new trailspace feature
by deliberately *NOT* using it.

10 years agosync
deraadt [Mon, 23 Dec 2013 00:23:15 +0000 (00:23 +0000)]
sync

10 years agoInstall a bunch more of OpenSSL manpages. ok deraadt@
jca [Mon, 23 Dec 2013 00:09:22 +0000 (00:09 +0000)]
Install a bunch more of OpenSSL manpages.  ok deraadt@

10 years agoPolishing the worms in my favourite can, term_flushln().
schwarze [Sun, 22 Dec 2013 23:33:52 +0000 (23:33 +0000)]
Polishing the worms in my favourite can, term_flushln().

The TERMP_TWOSPACE flag i introduced in August 2009 was idiosyncratic
and served only a very narrow purpose.  Replace it by a more intuitive
and more general termp attribute "trailspace", to be used together
with TERMP_NOBREAK, to request a minimum amount of whitespace at
the end of the current column.  Adapt all code to the new interface.

No functional change intended;
code reviews to confirm that are welcome *eg*.

10 years agoremove unneeded check for null context. ok deraadt gilles millert
tedu [Sun, 22 Dec 2013 22:55:51 +0000 (22:55 +0000)]
remove unneeded check for null context. ok deraadt gilles millert

10 years agoTry to keep the framebuffer console layout that was set up by the firmware
kettenis [Sun, 22 Dec 2013 21:03:45 +0000 (21:03 +0000)]
Try to keep the framebuffer console layout that was set up by the firmware
on sparc64.

10 years agoAdd support for BCM57786,which seems to be almost indistinguishable from the
kettenis [Sun, 22 Dec 2013 20:39:23 +0000 (20:39 +0000)]
Add support for BCM57786,which seems to be almost indistinguishable from the
BCM57785.

Tested by Mark Rowland.

10 years agoUse the rasops_info struct as the accesscookie for wsdisplay. Simplifies the
kettenis [Sun, 22 Dec 2013 19:49:23 +0000 (19:49 +0000)]
Use the rasops_info struct as the accesscookie for wsdisplay.  Simplifies the
code as it allows us to use some of the rasops_xxx functions directly.

10 years agoBail out early if the PCI bus number is -1. This means the hardware isn't
kettenis [Sun, 22 Dec 2013 18:55:25 +0000 (18:55 +0000)]
Bail out early if the PCI bus number is -1.  This means the hardware isn't
there and the AML might reference stuff that isn't there.

10 years agoOnly attach cpu0 on single-processor kernels, to match the expected behaviour.
miod [Sun, 22 Dec 2013 18:53:14 +0000 (18:53 +0000)]
Only attach cpu0 on single-processor kernels, to match the expected behaviour.

10 years agoCompute ncpusfound correctly. The logic was there but the count variable was
miod [Sun, 22 Dec 2013 18:52:34 +0000 (18:52 +0000)]
Compute ncpusfound correctly. The logic was there but the count variable was
kept MD.

10 years agoMP-safe rw_cas() implementation for rwlocks.
miod [Sun, 22 Dec 2013 18:32:40 +0000 (18:32 +0000)]
MP-safe rw_cas() implementation for rwlocks.

10 years agoSet the primary CPU's PAL revision to the OSF/1 PAL revision after switching
miod [Sun, 22 Dec 2013 18:31:29 +0000 (18:31 +0000)]
Set the primary CPU's PAL revision to the OSF/1 PAL revision after switching
to it, per the Green Book; from NetBSD.

It turns out this is necessary on some (if not all) systems to properly spin
up secondary processors.

Bump version.

10 years agoctype cleanups, now that the kernel has u_char strings (that is the
deraadt [Sun, 22 Dec 2013 16:50:15 +0000 (16:50 +0000)]
ctype cleanups, now that the kernel has u_char strings (that is the
easy road, otherwise this would be a righteous mess)
ok florian

10 years agochange pf_osfp_entry strings to unsigned, to help userland
deraadt [Sun, 22 Dec 2013 16:49:29 +0000 (16:49 +0000)]
change pf_osfp_entry strings to unsigned, to help userland
ok florian

10 years agospacing; Loganaden Velvindron
deraadt [Sun, 22 Dec 2013 16:06:12 +0000 (16:06 +0000)]
spacing; Loganaden Velvindron

10 years agoFix end-of-sentence spacing after the DragonFly text production macro.
schwarze [Sun, 22 Dec 2013 14:06:29 +0000 (14:06 +0000)]
Fix end-of-sentence spacing after the DragonFly text production macro.
Patch from Franco Fichtner <franco at lastsummer dot de>.

10 years agoImplement end-of-sentence spacing at the end of man(7) macro lines.
schwarze [Sun, 22 Dec 2013 13:24:00 +0000 (13:24 +0000)]
Implement end-of-sentence spacing at the end of man(7) macro lines.

Improves cvs(1) pcap-filter(3) readline(3) termcap(5) cvsbug(8)
mkhybrid(8) nsd-zonec(8) in multiple places and c++filt(1) com_err(3)
nsd_conf(5) term(5) nsdc(8) in one place each.
Reduces overall groff-mandoc-differences in base by about 1.5%.

Patch from Franco Fichtner <franco at lastsummer dot de> (DragonFly).

10 years agoSeems gcc4 whines about an empty file as well, so just unconditionally do
kettenis [Sun, 22 Dec 2013 12:29:25 +0000 (12:29 +0000)]
Seems gcc4 whines about an empty file as well, so just unconditionally do
what we did for gcc3.

10 years agoUse a bigger timeout, otherwise this test fails on some (not really) slower
kettenis [Sun, 22 Dec 2013 11:08:31 +0000 (11:08 +0000)]
Use a bigger timeout, otherwise this test fails on some (not really) slower
hardware.

10 years agoformat string fix: %08x instead of %p for uint32_t
sf [Sun, 22 Dec 2013 11:05:58 +0000 (11:05 +0000)]
format string fix: %08x instead of %p for uint32_t

10 years agoformat string fix: %lx for register_t
sf [Sun, 22 Dec 2013 11:00:13 +0000 (11:00 +0000)]
format string fix: %lx for register_t

10 years agosize_t format string fixes
sf [Sun, 22 Dec 2013 10:49:40 +0000 (10:49 +0000)]
size_t format string fixes

10 years agoSetting format id to 't' when it's already 't' is a waste of
krw [Sun, 22 Dec 2013 04:47:57 +0000 (04:47 +0000)]
Setting format id to 't' when it's already 't' is a waste of
time.

10 years agotweak comment
tedu [Sun, 22 Dec 2013 03:29:07 +0000 (03:29 +0000)]
tweak comment

10 years agoNo need to change format id 'X' (ascii or hex) to 'x' (hex) while
krw [Sun, 22 Dec 2013 03:20:05 +0000 (03:20 +0000)]
No need to change format id 'X' (ascii or hex) to 'x' (hex) while
pretty printing, when 'X' is not used in the next switch statement.
Just keep 'X'. Makes the code a touch less mysterious.

10 years agoDon't look for two semi-colons when encountering an unrecognized
krw [Sun, 22 Dec 2013 02:37:25 +0000 (02:37 +0000)]
Don't look for two semi-colons when encountering an unrecognized
statement in dhclient.conf.

10 years ago- replaces "audio stream" by "audio device", as sndio is a device
ratchov [Sun, 22 Dec 2013 01:13:31 +0000 (01:13 +0000)]
- replaces "audio stream" by "audio device", as sndio is a device
  abstraction
- remove useless statements.

ok jmc@

10 years agoDocument that the "user" socket check is not reliable with wildcard
camield [Sat, 21 Dec 2013 20:57:01 +0000 (20:57 +0000)]
Document that the "user" socket check is not reliable with wildcard
listeners.  Update the example to reflect this.

Deprecate usage of user "unknown" too, same reason.

Noted by Maxim Khitrov on bugs

ok henning, jmc

10 years agoDon't assign a literal number to a variable that otherwise holds an errno.
guenther [Sat, 21 Dec 2013 20:46:20 +0000 (20:46 +0000)]
Don't assign a literal number to a variable that otherwise holds an errno.

ok jsing@

10 years agoAdvertise that we provide monotonic timestamps.
kettenis [Sat, 21 Dec 2013 19:36:41 +0000 (19:36 +0000)]
Advertise that we provide monotonic timestamps.

10 years agoIntroduce pretty_print_string() and use for printing both text
krw [Sat, 21 Dec 2013 18:23:10 +0000 (18:23 +0000)]
Introduce pretty_print_string() and use for printing both text
valued options and previously snprintf()'d filename and servername
lease attributes.

Should fix "string constant too long" errors when reading back a
lease with filename or servername attributes with escaped characters.

Reported by Rivo Nurges.

10 years agorearrange DVACT_RESUME chunk to avoid implying something ridiculous
deraadt [Sat, 21 Dec 2013 17:10:08 +0000 (17:10 +0000)]
rearrange DVACT_RESUME chunk to avoid implying something ridiculous
ok stsp

10 years agosave/restore hpet configuration. Perhaps a little more than we
deraadt [Sat, 21 Dec 2013 16:41:01 +0000 (16:41 +0000)]
save/restore hpet configuration.  Perhaps a little more than we
need to do, since we do not use many hpet features.
ok kettenis millert

10 years agoi915_gem_mmap_ioctl() should return a Linux-style negative errno value.
kettenis [Sat, 21 Dec 2013 15:09:23 +0000 (15:09 +0000)]
i915_gem_mmap_ioctl() should return a Linux-style negative errno value.

10 years agosync
deraadt [Sat, 21 Dec 2013 13:35:15 +0000 (13:35 +0000)]
sync

10 years agoFix locking in the page fault handler. A (somewhat malicious) userland
kettenis [Sat, 21 Dec 2013 12:21:23 +0000 (12:21 +0000)]
Fix locking in the page fault handler.  A (somewhat malicious) userland
program could force a copyin/copyout from/to memory mapped through the GTT,
forcing a "locking against myself" panic.  The intel-gpu-tools "package" has
a test for this.  The problem can be circumvented by making the fault handler
fail if we already hold the (DRM) lock.  This will make the copyin/copyout
return with EFAULT making the caller fall back on a "slow path".

This makes it obvious that using a shared (read) lock here doesn't make any
sense.  So use an exclusive (write) lock like everywhere else in the inteldrm
code.

10 years agoRecognize itimer and ktrace facility names to {get,set}itimer() and ktrace()
guenther [Sat, 21 Dec 2013 07:32:35 +0000 (07:32 +0000)]
Recognize itimer and ktrace facility names to {get,set}itimer() and ktrace()

ok otto@

10 years agoRefer to the kernel headers consistently via ../../sys, even for mkioctl
guenther [Sat, 21 Dec 2013 07:30:00 +0000 (07:30 +0000)]
Refer to the kernel headers consistently via ../../sys, even for mkioctl
and mksubr

ok otto@

10 years agosmall typo
tedu [Sat, 21 Dec 2013 07:10:47 +0000 (07:10 +0000)]
small typo

10 years agoDon't use a non-constant format string with printf() that gcc can't
guenther [Sat, 21 Dec 2013 06:54:53 +0000 (06:54 +0000)]
Don't use a non-constant format string with printf() that gcc can't
figure out how to trust.  Shuts up gcc -Wformat=2

ok deraadt@

10 years agoConstipate some functions
guenther [Sat, 21 Dec 2013 06:29:17 +0000 (06:29 +0000)]
Constipate some functions

10 years agoMatch the type to the calls
guenther [Sat, 21 Dec 2013 05:45:04 +0000 (05:45 +0000)]
Match the type to the calls