* Allocate the correct buffer for the VGA BIOS image.
* Abort if buffer allocation failed, not if buffer allocation succeeded.
Not directly tested, but I can't imagine this makes anything worse.
It'll make life simpler. No need to distinguish drm from drm2 here
because it's just an interface attribute, no content, and we use it
compatibly: parent provides pci_attach_args.
Missing part: something needs to issue config_found_ia("drmums_pci"),
and config(5) needs to be told that it will do that. The sensible
approach is for vga_pci to do so. But config(5) doesn't seem to
understand that *only* vga_pci will do that, not all vga.
The old scheme was to tell config(5) that vga would always attach
drm, and actually attach it in vga_pci, and hope that nobody would
ever consider trying to attach non-PCI devices but quietly leave the
token `pci' out of the name of the drm interface attribute so it
didn't look wrong to have it outside dev/pci/files.pci.
Maybe that's still the right thing, as far as anything UMS can be
right.
my PCI 9250 card, and hoses the colours in X when running.
unfortunately, my RV250 does need this to avoid the console being
very very dark on black.
i'm curious what other earlier radeon card (anything matching any
of the card marked as: CHIP_R100, CHIP_RV100, CHIP_RS100,
CHIP_RV200, CHIP_RS200, CHIP_R200, CHIP_RV250, and CHIP_RS300.)
the RV280 now works again properly, but the RV250 is still having
a problem waiting for commands to complete and it appears as if
X is hung, when infact it is asking the kernel quite frequently
if the job it asked for is complete, each time sleeping for a
short time and thus only appearing to take almost no CPU.
- call pmf_device_deregister on detach.
i've kept the "resume = true" for radeon_resume_kms() call as it
seems to work for me (indeed, code inspection shows it is unused
on netbsd :-)
my old nforce4 box that can resume old drm (or could, last i tried
several years ago) while X and GL apps were running, can at least
survive a resume if X hasn't started. my one attempt so far with
X exited, but having run, did not work.
10 April 2015. Summary of changes for version 20150410:
Reverted a change introduced in version 20150408 that caused
a regression in the disassembler where incorrect operator
symbols could be emitted.
----------------------------------------
08 April 2015. Summary of changes for version 20150408:
1) ACPICA kernel-resident subsystem:
Permanently set the return value for the _REV predefined name. It now
returns 2 (was 5). This matches other ACPI implementations. _REV will be
deprecated in the future, and is now defined to be 1 for ACPI 1.0, and 2
for ACPI 2.0 and later. It should never be used to differentiate or
identify operating systems.
Added the "Windows 2015" string to the _OSI support. ACPICA will now
return TRUE to a query with this string.
Fixed several issues with the local version of the printf function.
Added the C99 compiler option (-std=c99) to the Unix makefiles.
Current Release:
Non-Debug Version: 99.9K Code, 27.4K Data, 127.3K Total
Debug Version: 195.2K Code, 80.7K Data, 275.9K Total
Previous Release:
Non-Debug Version: 98.8K Code, 27.3K Data, 126.1K Total
Debug Version: 192.1K Code, 79.8K Data, 271.9K Total
2) iASL Compiler/Disassembler and Tools:
iASL: Implemented an enhancement to the constant folding feature to
transform the parse tree to a simple Store operation whenever possible:
Add (2, 3, X) ==> is converted to: Store (5, X)
X = 2 + 3 ==> is converted to: Store (5, X)
Updated support for the SLIC table (Software Licensing Description Table)
in both the Data Table compiler and the disassembler. The SLIC table
support now conforms to "Microsoft Software Licensing Tables (SLIC and
MSDM). November 29, 2011. Copyright 2011 Microsoft". Note: Any SLIC data
following the ACPI header is now defined to be "Proprietary Data", and as
such, can only be entered or displayed as a hex data block.
Implemented full support for the MSDM table as described in the document
above. Note: The format of MSDM is similar to SLIC. Any MSDM data
following the ACPI header is defined to be "Proprietary Data", and can
only be entered or displayed as a hex data block.
Implemented the -Pn option for the iASL Table Compiler (was only
implemented for the ASL compiler). This option disables the iASL
preprocessor.
Disassembler: For disassembly of Data Tables, added a comment field
around the Ascii equivalent data that is emitted as part of the "Raw
Table Data" block. This prevents the iASL Preprocessor from possible
confusion if/when the table is compiled.
Disassembler: Added an option (-df) to force the disassembler to assume
that the table being disassembled contains valid AML. This feature is
useful for disassembling AML files that contain ACPI signatures other
than DSDT or SSDT (such as OEMx or other signatures).
Changes for the EFI version of the tools:
1) Fixed a build error/issue
2) Fixed a cast warning
iASL: Fixed a path issue with the __FILE__ operator by making the
directory prefix optional within the internal SplitInputFilename
function.
Debugger: Removed some unused global variables.
Tests: Updated the makefile for proper generation of the AAPITS suite.
----------------------------------------
04 February 2015. Summary of changes for version 20150204:
ACPICA kernel-resident subsystem:
Updated all ACPICA copyrights and signons to 2014. Added the 2014
copyright to all module headers and signons, including the standard Linux
header. This affects virtually every file in the ACPICA core subsystem,
iASL compiler, all ACPICA utilities, and the test suites.
Events: Introduce ACPI_GPE_DISPATCH_RAW_HANDLER to fix GPE storm issues.
A raw gpe handling mechanism was created to allow better handling of GPE
storms that aren't easily managed by the normal handler. The raw handler
allows disabling/renabling of the the GPE so that interrupt storms can be
avoided in cases where events cannot be timely serviced. In this
scenario, handlers should use the AcpiSetGpe() API to disable/enable the
GPE. This API will leave the reference counts undisturbed, thereby
preventing unintentional clearing of the GPE when the intent in only to
temporarily disable it. Raw handlers allow enabling and disabling of a
GPE by removing GPE register locking. As such, raw handlers much provide
their own locks while using GPE API's to protect access to GPE data
structures.
Lv Zheng
Events: Always modify GPE registers under the GPE lock.
Applies GPE lock around AcpiFinishGpe() to protect access to GPE register
values. Reported as bug by joe.liu@apple.com.
Unix makefiles: Separate option to disable optimizations and
_FORTIFY_SOURCE. This change removes the _FORTIFY_SOURCE flag from the
NOOPT disable option and creates a separate flag (NOFORTIFY) for this
purpose. Some toolchains may define _FORTIFY_SOURCE which leads redefined
errors when building ACPICA. This allows disabling the option without
also having to disable optimazations.
David Box
Current Release:
Non-Debug Version: 101.7K Code, 27.9K Data, 129.6K Total
Debug Version: 199.2K Code, 82.4K Data, 281.6K Total
----------------------------------------
07 November 2014. Summary of changes for version 20141107:
This release is available at https://acpica.org/downloads
This release introduces and implements language extensions to ASL that
provide support for symbolic ("C-style") operators and expressions. These
language extensions are known collectively as ASL+.
1) iASL Compiler/Disassembler and Tools:
Disassembler: Fixed a problem with disassembly of the UartSerialBus
macro. Changed "StopBitsNone" to the correct "StopBitsZero". David E.
Box.
Disassembler: Fixed the Unicode macro support to add escape sequences.
All non-printable ASCII values are emitted as escape sequences, as well
as the standard escapes for quote and backslash. Ensures that the
disassembled macro can be correctly recompiled.
iASL: Added Printf/Fprintf macros for formatted output. These macros are
translated to existing AML Concatenate and Store operations. Printf
writes to the ASL Debug object. Fprintf allows the specification of an
ASL name as the target. Only a single format specifier is required, %o,
since the AML interpreter dynamically converts objects to the required
type. David E. Box.
(old) Store (Concatenate (Concatenate (Concatenate (Concatenate
(Concatenate (Concatenate (Concatenate ("", Arg0),
": Unexpected value for "), Arg1), ", "), Arg2),
" at line "), Arg3), Debug)
(new) Printf ("%o: Unexpected value for %o, %o at line %o",
Arg0, Arg1, Arg2, Arg3)
(old) Store (Concatenate (Concatenate (Concatenate (Concatenate
("", Arg1), ": "), Arg0), " Successful"), STR1)
(new) Fprintf (STR1, "%o: %o Successful", Arg1, Arg0)
iASL: Added debug options (-bp, -bt) to dynamically prune levels of the
ASL parse tree before the AML code is generated. This allows blocks of
ASL code to be removed in order to help locate and identify problem
devices and/or code. David E. Box.
AcpiExec: Added support (-fi) for an optional namespace object
initialization file. This file specifies initial values for namespace
objects as necessary for debugging and testing different ASL code paths
that may be taken as a result of BIOS options.
2) Overview of symbolic operator support for ASL (ASL+)
-------------------------------------------------------
As an extension to the ASL language, iASL implements support for symbolic
(C-style) operators for math and logical expressions. This can greatly
simplify ASL code as well as improve both readability and
maintainability. These language extensions can exist concurrently with
all legacy ASL code and expressions.
The symbolic extensions are 100% compatible with existing AML
interpreters, since no new AML opcodes are created. To implement the
extensions, the iASL compiler transforms the symbolic expressions into
the legacy ASL/AML equivalents at compile time.
Full symbolic expressions are supported, along with the standard C
precedence and associativity rules.
Full disassembler support for the symbolic expressions is provided, and
creates an automatic migration path for existing ASL code to ASL+ code
via the disassembly process. By default, the disassembler now emits ASL+
code with symbolic expressions. An option (-dl) is provided to force the
disassembler to emit legacy ASL code if desired.
Below is the complete list of the currently supported symbolic operators
with examples. See the iASL User Guide for additional information.
ASL+ Syntax Legacy ASL Equivalent
----------- ---------------------
// Math operators
Z = X + Y Add (X, Y, Z)
Z = X - Y Subtract (X, Y, Z)
Z = X * Y Multiply (X, Y, Z)
Z = X / Y Divide (X, Y, , Z)
Z = X % Y Mod (X, Y, Z)
Z = X << Y ShiftLeft (X, Y, Z)
Z = X >> Y ShiftRight (X, Y, Z)
Z = X & Y And (X, Y, Z)
Z = X | Y Or (X, Y, Z)
Z = X ^ Y Xor (X, Y, Z)
Z = ~X Not (X, Z)
X++ Increment (X)
X-- Decrement (X)
// Logical operators
(X == Y) LEqual (X, Y)
(X != Y) LNotEqual (X, Y)
(X < Y) LLess (X, Y)
(X > Y) LGreater (X, Y)
(X <= Y) LLessEqual (X, Y)
(X >= Y) LGreaterEqual (X, Y)
(X && Y) LAnd (X, Y)
(X || Y) LOr (X, Y)
(!X) LNot (X)
// Assignment and compound assignment operations
X = Y Store (Y, X)
X += Y Add (X, Y, X)
X -= Y Subtract (X, Y, X)
X *= Y Multiply (X, Y, X)
X /= Y Divide (X, Y, , X)
X %= Y Mod (X, Y, X)
X <<= Y ShiftLeft (X, Y, X)
X >>= Y ShiftRight (X, Y, X)
X &= Y And (X, Y, X)
X |= Y Or (X, Y, X)
X ^= Y Xor (X, Y, X)
3) ASL+ Examples:
-----------------
Legacy ASL:
If (LOr (LOr (LEqual (And (R510, 0x03FB), 0x02E0), LEqual (
And (R520, 0x03FB), 0x02E0)), LOr (LEqual (And (R530,
0x03FB),
0x02E0), LEqual (And (R540, 0x03FB), 0x02E0))))
{
And (MEMB, 0xFFFFFFF0, SRMB)
Store (MEMB, Local2)
Store (PDBM, Local1)
And (PDBM, 0xFFFFFFFFFFFFFFF9, PDBM)
Store (SRMB, MEMB)
Or (PDBM, 0x02, PDBM)
}
ASL+ version:
If (((R510 & 0x03FB) == 0x02E0) ||
((R520 & 0x03FB) == 0x02E0) ||
((R530 & 0x03FB) == 0x02E0) ||
((R540 & 0x03FB) == 0x02E0))
{
SRMB = (MEMB & 0xFFFFFFF0)
Local2 = MEMB
Local1 = PDBM
PDBM &= 0xFFFFFFFFFFFFFFF9
MEMB = SRMB
PDBM |= 0x02
}
Legacy ASL:
Store (0x1234, Local1)
Multiply (Add (Add (Local1, TEST), 0x20), Local2, Local3)
Multiply (Local2, Add (Add (Local1, TEST), 0x20), Local3)
Add (Local1, Add (TEST, Multiply (0x20, Local2)), Local3)
Store (Index (PKG1, 0x03), Local6)
Store (Add (Local3, Local2), Debug)
Add (Local1, 0x0F, Local2)
Add (Local1, Multiply (Local2, Local3), Local2)
Multiply (Add (Add (Local1, TEST), 0x20), ToBCD (Local1), Local3)
ASL+ version:
Local1 = 0x1234
Local3 = (((Local1 + TEST) + 0x20) * Local2)
Local3 = (Local2 * ((Local1 + TEST) + 0x20))
Local3 = (Local1 + (TEST + (0x20 * Local2)))
Local6 = Index (PKG1, 0x03)
Debug = (Local3 + Local2)
Local2 = (Local1 + 0x0F)
Local2 = (Local1 + (Local2 * Local3))
Local3 = (((Local1 + TEST) + 0x20) * ToBCD (Local1))
Since we now only count time spent in wait, rather than determining
the end time and checking whether we've passed it, timeouts might be
marginally longer in effect. Unlikely to be an issue.
As provided by Takahiro HAYASHI in PR kern/48109. Additional error
registration in ipf(8) by myself. Changes tested with GENERIC and
XEN3_DOM0. Thanks!
XXX pull-up netbsd-7
Leave the CLUT alone on ancient cards. At least this leaves us with a
semi working console (red and blue are flipped). Leave an example of what
seems to be happening but disable it because colors are better than 444 bit
greyscale.
I had assumed for ages this would increase the amount of caching and
thereby increase the chance of stale caches leading to rendering
glitches. But apparently I was wrong, and failing to pass these
through was causing all sorts of problems!
Otherwise GCC helpfully realizes we're invoking undefined behaviour
and optimizes away the possibility that this routine will return
success, without saying a word about it.
We still need to adapt all waits from upstream to use an interlock,
so we can't implement DRM_WAIT_ON verbatim, but this more closely
reflects the API of DRM_WAIT_ON than DRM_*WAIT*_UNTIL do.
Major difference is that this polls every tick, like DRM_WAIT_ON,
unlike DRM_*WAIT*_UNTIL. So it will mask missing wakeups, but it
wouldn't surprise me if there were such things upstream.
MAX(ret, 0) is 0 if ret is negative, but if ret is negative we want
to return that negative value, meaning error. Should've been
MIN(ret, 0), but I'll just rewrite it to clarify a wee bit.
If the GPU reset, call i915_gem_check_wedge and always return failure
like Linux does. Caller must retry in that case.
commit 95468892fdfeef6d1004b524e35957629efdbe00
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date: Thu Aug 7 15:39:54 2014 +0100
drm/i915: Reset the HEAD pointer for the ring after writing START
Ville found an old w/a documented for g4x that suggested that we need to
reset the HEAD after writing START. This is a useful fixup for some of
the g4x ring initialisation woes, but as usual, not all.
v2: Do the rewrite unconditionally anyway
commit ece4a17d237a79f63fbfaf3f724a12b6d500555c
Author: Jiri Kosina <jkosina@suse.cz>
Date: Thu Aug 7 16:29:53 2014 +0200
drm/i915: read HEAD register back in init_ring_common() to enforce ordering
Withtout this, ring initialization fails reliabily during resume with
[drm:init_ring_common] *ERROR* render ring initialization failed ctl 000
1f001 head ffffff8804 tail 00000000 start 000e4000
This is not a complete fix, but it is verified to make the ring
initialization failures during resume much less likely.
We were not able to root-cause this bug (likely HW-specific to Gen4 chips)
yet. This is therefore used as a ducttape before problem is fully
understood and proper fix created, so that people don't suffer from
completely unusable systems in the meantime.
The discussion and debugging is happening at
https://bugs.freedesktop.org/show_bug.cgi?id=76554
radeondrmkms on i386. the problem is that the pcdisplay@isa
and vga@isa drivers rely on some other driver having already
mapped the vga registers to know they should not attach, but
radeondrmkms attaches late as it needs rootfs access, and it
does not map these registers, and either of vga or pcdisplay
will attach, and then attach wsdisplay0, and then getty and
X11 will fail and you shake a tiny fist at the computer.
so, for i386, map and unmap the VGA register space between
the normal attach and the mountroot attach.
that earlier revision itself reverted an upstream bugfix,
and that change fixed some systems but broke others
(in particular, it broke my HP Z800 workstation.)
this latest revision brings us back in sync with upstream.
the current upstream (and this latest flip-flop of our copy)
is clearly correct, since it zeros a buffer which is intended
to be used as scratch space for a single acpica operation.
agreed by christos and ryoon back in april 2014.
