drm: Teach bus_dma_hacks.h about alpha.

Not sure if this is correct! Someone who knows alpha should vet.

drm: Fix index confusion and rename variables to clarify.

drm: Let's go one segment at a time, not two, shall we?

drm: Another pass over i915 and some supporting logic.

This makes a shim around sg_table, which essentially represents two
things:

1. an array of pages (roughly corresponding to bus_dma_segment_t[])
2. an array of DMA addresses stored in a bus_dmamap_t

Both parts are optional; different parts of i915 use sg_tables to
pass around one or both of the two parts. This helps to reduce the
ifdefs by quite a bit, although it's not always clear which part of
an sg_table any particular interface is actually using which is why I
was reluctant to do this before.

drm: Nix use of uvm pglist. Just use arrays of page pointers.

in bus_dmamap_load_pglist() try a 32-element array of
bus_dma_segment_t's before attempting to allocate.

this hopefully avoids hangs i've had in X since updating
from netbsd-8 to netbsd-9 that i've tracked down to this
function failing with ENOMEM.

XXX: maybe can avoid the alloc entirely by batching these
calls in 32 segments each.

XXX pullup-9

Fix PHYS_TO_BUS_MEM and BUS_MEM_TO_PHYS on arm and aarch64

branches: 1.18.2;
Arm DMA tags with _nranges=0 require no address translation. Handle this
in PHYS_TO_BUS_MEM and BUS_TO_PHYS_MEM instead of panicing.

branches: 1.17.4;
Check for DMA-safe addresses before allowing DRM prime import.

Implement BUS_DMA_TO_PHYS/PHYS_TO_BUS_DMA on arm respecting ranges.

At chuq's suggestion, use paddr_t rather than struct vm_page *.

Doesn't make much of a substantive difference.

sparc seems to treat bus/phys addrs in bus dmamem the same.

This is just about bus dmamem -- not about bus dmamap, which rightly
uses an iommu to remap things and which we don't interfere with.

Eliminate now-unused bus_dmamap_load_pgarray.

Rework drm prime guts.

- Our faux struct sg_table is now just an array of _pages_, not of
bus dma segments.

- We handle bus dma segments only in the presence of an adult bus dma
tag.

- All bus dma hacks are once again confined to bus_dma_hacks.h, and
more neatly organized:
. bus_dmamem_export_pages translates bus dma segs to vm_page array
. bus_dmamem_import_pages translates vm_page array to bus dma segs

Use PHYS_TO_BUS_MEM in generic bus_dmamap_load_pglist code.

For arm and x86, this is a noop. For powerpc, it is defined by some
relevant header file.

Kill bus_dmamem_wire_uvm_object, which never made sense!

branches: 1.9.2;
Add initial support for ARMv8 (AARCH64) (by nisimura@ and ryo@)

- sys/arch/evbarm64 is gone and integrated into sys/arch/evbarm. (by skrll@)
- add support fdt. evbarm/conf/GENERIC64 fdt (bcm2837,sunxi,tegra) based generic 64bit kernel config. (by skrll@, jmcneill@)

branches: 1.8.16;
use VM_FREELIST_DEFAULT on arm

branches: 1.7.4; 1.7.6;
Avoid needless #ifdef __i386__ with judicious use of MIN.

Free temporary segs on the way out of bus_dmamap_load_pglist.

First whack at radeon driver.

No hardware to test yet, but it builds.

Make it build and boot on my test machines.

Screen blanks on boot on the Ivy Bridge system with

DRM error in cpt_serr_int_handler: PCH transcoder A FIFO underrun

But after that everything is OK. Appears to be an upstream problem.
To investigate...

I think there's a cache flushing issue somewhere -- there are little
display artefacts on my T60.

branches: 1.3.2; 1.3.4; 1.3.6;
Miscellaneous fixes for drm2 on i386.

Merge riastradh-drm2 to HEAD.

branches: 1.1.2;
file bus_dma_hacks.h was initially added on branch riastradh-drm2.

Delete drmP.h. Good riddance!

Empty out drm_os_netbsd.h.

Give some more declarations a new home.

Move drm_io_mapping_* to its own header file.

Move files which are gone upstream out of dist/.

Rework PCI and AGP conditionalization.

- Push drm_pci_set_unique into driver callback.
- Eliminate drm_pci_set_unique_hook.
- Gather all drm_agp_* functions into struct drm_agp_hooks.
- Replace the nonsensical old atomic garbage by serious locking.
- Make drm_agpsupport.c its own module.
- Eliminate NDRMKMS_PCI.
- Use NAGP from "agp.h" only in drm_module.c for horrible hack.
=> See comment in file for rationale.
- Always define CONFIG_PCI=1 and CONFIG_AGP=1.
- Always go through the drm_agp_* function hooks.
- Ifdef out nouveau agp stuff that doesn't go through drm_agp_*
for reasons that I'm too frustrated to figure out tonight.
- pci_iomap no longer automagically does agp_i810_borrow.
=> Use drm_agp_borrow instead.

Upstream nixed __OS_HAS_AGP; use CONFIG_AGP instead.

branches: 1.6.16; 1.6.18;
dont pull in agp.h when building as a module

only define __OS_HAS_AGP if x86 and NAGP > 0

branches: 1.4.2;
Another round of nouveau whack-a-mole.

branches: 1.3.2; 1.3.4;
Make it build and boot on my test machines.

Screen blanks on boot on the Ivy Bridge system with

DRM error in cpt_serr_int_handler: PCH transcoder A FIFO underrun

But after that everything is OK. Appears to be an upstream problem.
To investigate...

I think there's a cache flushing issue somewhere -- there are little
display artefacts on my T60.

branches: 1.2.2; 1.2.4; 1.2.6;
Merge riastradh-drm2 to HEAD.

branches: 1.1.2;
file drm_agp_netbsd.h was initially added on branch riastradh-drm2.

Reduce diffs by nixing DRM_SUSER macro, deleted upstream long ago.

Instead stub Linux capable(CAP_SYS_ADMIN).

branches: 1.2.4; 1.2.6; 1.2.10; 1.2.36; 1.2.42;
Merge riastradh-drm2 to HEAD.

branches: 1.1.2;
file drm_auth_netbsd.h was initially added on branch riastradh-drm2.

drm: Work around client modeset mutex for entering ddb.

Linux introduced a wait lock (`mutex') around access to the drm
client modesets, even for drm_fb_helper_debug_enter/leave -- which
happen when we're crashing and can't wait for the scheduler to let
whatever thread might hold a mutex continue to run.

This really shouldn't be a wait lock, but disentangling the lock
order and sleeping allocations is difficult, so let's just hope that
it's safe to access unlocked in ddb for now.

Start drm_client.h stubs.

Remove unused definitions.

(These appear to have existed in old-drm)

branches: 1.2.4; 1.2.6; 1.2.10; 1.2.36;
Merge riastradh-drm2 to HEAD.

branches: 1.1.2;
file drm_copy_netbsd.h was initially added on branch riastradh-drm2.

Create stub headers for files included by drm code.


Author: Maya Rashish <maya@NetBSD.org>

Make this const.

(Why did I rewrite drm_encoder_slave.h?)

Tweak includes and forward decls.

branches: 1.1.2; 1.1.4; 1.1.20;
Rewrite drm_encoder_slave.c, as nouveau will need.

Add a missing ';'

drm: Rewrite drm_gem_cma_helper.h from scratch.

The header file was originally part of drm with no licence text,
which I took to mean BSD-licensed like the rest of the Linux drm code
base is by default -- everything in Linux under include/drm and
drivers/gpu/drm not otherwise marked.

In 2017, a tree-wide automated change was applied to Linux which
caused an SPDX GPL header to appear in drm_gem_cma_helper.h. When I
asked the drm developers about this in drm files, they explained that
it was an automated process and many of the SPDX headers were wrong,
so when excluding GPL files in drm updates, I heeded the actual GPL
vs BSD/MIT licence text and revision history rather than the SPDX
headers.

I excluded drm_gem_cma_helper.c because it was clearly originally GPL
as an exception to the default BSD licence under Linux's
drivers/gpu/drm, and Jared reimplemented the API for ues by some
(NetBSD, BSD-licensed) Arm display drivers.

During an update after that, the header file was lost in the merge
shuffle and I copied it back in without noticing the SPDX header
file. In retrospect, I see now that it's not clear what the intent
was: the .c file was obviously GPL, and maybe the .h file was
supposed to be too but it wasn't carved out as an exception among the
drm code.

To keep our story simpler, I just drafted a new .h file that provides
a compatible subset of the interface we use, derived from the .c file
Jared wrote, written according to several deterministic rules based
on KNF that yield a non-creative specification of the compatible
interface:

1. $NetBSD$ RCS id
2. standard TNF copyright statement
3. header guard derived from #include file name
4. minimal #includes, grouped and sorted according to KNF
5. forward declaration of structs, sorted lexicographically
6. definitions of structs, members sorted lexicographically by name
7. downcasting macros, written in KNF and sorted lexicographically
8. minimal extern variable declarations, sorted lexicographically
9. minimal, function prototypes, written in KNF and sorted
lexicographically

`Minimal' means added one by one, according to the errors issued by
the compiler (and in the case of #includes, according to the correct
header file for the missing types or function prototypes being used),
until the compiler is formally satisfied.

Any changes to this will, unless otherwise marked, be covered by the
standard TNF licence now on the file.

Fix up includes of <sys/vmem.h>.

Move files which are gone upstream out of dist/.

branches: 1.1.2; 1.1.6;
Move old DRM_READn definitions into new drm_iomap_netbsd.h

These got deleted from drmP.h.

drm: New IRQ_WAKE_THREAD constant.

Not really wired up to anything but helps reduce diffs -- driver is
responsible for making decisions based on it, corresponding to
Linux's built-in `threaded irq'.

drm: Set IPL_DRM to IPL_VM, not IPL_TTY.

We might want some of this stuff to work at higher IPL when dealing
with ddb during crashes, but...everything else in drm uses IPL_VM,
including Linux spin locks, which means we've been failing to block
interrupts during spin locks for the entire time this code has
existed...which is a little embarrassing.

provide enum irqreturn

merge in old ifdefs
ifdef out code that I'm unsure about right away and turn into panic
call.


Author: coypu <coypu@sdf.org>
Committer: Taylor R Campbell <riastradh@NetBSD.org>

branches: 1.4.16; 1.4.18;
due to hangs seen by several folks, for now revert:
http://mail-index.netbsd.org/source-changes/2014/11/04/msg060120.html

Log Message:
This code should be MP-safe. Use IPL_SCHED in place of IPL_DRM/IPL_VM and set
D_MPSAFE flag in cdevsw.

branches: 1.3.2;
This code should be MP-safe. Use IPL_SCHED in place of IPL_DRM/IPL_VM and set
D_MPSAFE flag in cdevsw.

branches: 1.2.4; 1.2.6; 1.2.10;
Merge riastradh-drm2 to HEAD.

branches: 1.1.2;
file drm_irq_netbsd.h was initially added on branch riastradh-drm2.

Fix typo: list lessees, not list leases.

Declare drm_lease_destroy.

Unfinished local implementation of GPL drm_lease.c.

branches: 1.2.4; 1.2.6; 1.2.10;
Merge riastradh-drm2 to HEAD.

branches: 1.1.2;
file drm_mem_util.h was initially added on branch riastradh-drm2.

Make intel_psr.c build
netbsd-ify
Stub out intel_lpe_audio.c (for now?).
Hammer intel_opregion.c into shape.
Hack up intel_lrc.c
Adapt intel_reset.c
Stub out intel_region_lmem.c
Adapt intel_ring_submission.c
Adapt intel_rps.c
Adapt intel_sdvo.c
Adapt intel_runtime_pm.c
Adapt intel_sideband.c
Adapt intel_sprite.c
Adapt intel_tv.c
Adapt intel_tc.c
Adapt intel_sseu.c
Adapt intel_timeline.c
Adapt intel_uc_fw.c and fix typo
Fix up intel_uncore.c
intel_vga.c
intel_wakeref.c
intel_vdsc.c
intel_wopcm.c
intel_workarounds.c
i915_globals.c
Fix up intelfb.c
Stub out intel_guc_log.c
Stub out intel_dsi_dcs_backlight.c
i915_perf.c
intel_hdmi.c
intel_hdcp.c


Author: Maya Rashish <maya@NetBSD.org>
Committer: Taylor R Campbell <riastradh@NetBSD.org>

i915: another whack at it

branches: 1.1.2; 1.1.6;
Add stub headers to be filled in as needed.

branches: 1.1.2;
file drm_os_linux.h was initially added on branch riastradh-drm2.

drm_os_netbsd.h is no more

Empty out drm_os_netbsd.h.

Fix up includes of <sys/vmem.h>.

nbsd-namespace, and provide CONFIG_LOCKDEP (disabled)


Author: Maya Rashish <maya@NetBSD.org>

Reduce diffs by nixing DRM_SUSER macro, deleted upstream long ago.

Instead stub Linux capable(CAP_SYS_ADMIN).

Disable the use of WC memory on Arm and Arm64. From Linux:

* The DRM driver stack is designed to work with cache coherent devices
* only, but permits an optimization to be enabled in some cases, where
* for some buffers, both the CPU and the GPU use uncached mappings,
* removing the need for DMA snooping and allocation in the CPU caches.
*
* The use of uncached GPU mappings relies on the correct implementation
* of the PCIe NoSnoop TLP attribute by the platform, otherwise the GPU
* will use cached mappings nonetheless. On x86 platforms, this does not
* seem to matter, as uncached CPU mappings will snoop the caches in any
* case. However, on ARM and arm64, enabling this optimization on a
* platform where NoSnoop is ignored results in loss of coherency, which
* breaks correct operation of the device. Since we have no way of
* detecting whether NoSnoop works or not, just disable this
* optimization entirely for ARM and arm64.

branches: 1.15.2;
Add support for DRM GEM/CMA helpers using a driver provided vmem arena.

branches: 1.14.4;
Remove unused definitions.

(These appear to have existed in old-drm)

Rework PCI and AGP conditionalization.

- Push drm_pci_set_unique into driver callback.
- Eliminate drm_pci_set_unique_hook.
- Gather all drm_agp_* functions into struct drm_agp_hooks.
- Replace the nonsensical old atomic garbage by serious locking.
- Make drm_agpsupport.c its own module.
- Eliminate NDRMKMS_PCI.
- Use NAGP from "agp.h" only in drm_module.c for horrible hack.
=> See comment in file for rationale.
- Always define CONFIG_PCI=1 and CONFIG_AGP=1.
- Always go through the drm_agp_* function hooks.
- Ifdef out nouveau agp stuff that doesn't go through drm_agp_*
for reasons that I'm too frustrated to figure out tonight.
- pci_iomap no longer automagically does agp_i810_borrow.
=> Use drm_agp_borrow instead.

Ifdef out pnpbios for now. Presumably needs to use bus_space_alloc.

opt_mtrr.h is x86-only.

Guard #include "pci.h" with _KERNEL_OPT.

Define CONFIG_MTRR and CONFIG_X86_PAT.

Include the legacy DRM_READ/WRITE in drm_legacy.h, not drm_os_*.h.

Define CONFIG_PCI to 1, not empty.

provide CONFIG_PNP if it's defined. i386 only apparently.

XXX I'm not actually building i386 kernels


Author: coypu <coypu@sdf.org>
Committer: Taylor R Campbell <riastradh@NetBSD.org>

bring in changes from OLDER drmP.h to go forward with compilation


Author: coypu <coypu@sdf.org>
Committer: Taylor R Campbell <riastradh@NetBSD.org>

Move old DRM_READn definitions into new drm_iomap_netbsd.h

These got deleted from drmP.h.

branches: 1.3.4; 1.3.28; 1.3.30;
Move CONFIG_X86 from files.i915drmkms to drm_os_netbsd.h.

branches: 1.2.2; 1.2.4; 1.2.6;
Merge riastradh-drm2 to HEAD.

branches: 1.1.2;
file drm_os_netbsd.h was initially added on branch riastradh-drm2.

branches: 1.2.4; 1.2.20;
prettify and add to all the modules that have it.

Add __link_set based code to automatically convert the linux module parameters
into sysctls.

Update drm_sysfs stubs.

drm: Mark variable as used only by trace point.

branches: 1.1.2; 1.1.6;
Forgot to commit drm_trace_netbsd.h. Too late to force an update...

drm: Add drm_vma_offset_exact_lookup_locked.

provide DRM_FILE_PAGE_OFFSET_{START,SIZE}


Author: Maya Rashish <maya@NetBSD.org>

Side-load linux/mm.h from the same headers as linux


Author: Maya Rashish <maya@NetBSD.org>

Linux wants a `readonly' member in struct drm_vma_offset_node.

Constify some prototypes for great constance.

drm_vma_manager deals in struct drm_file, not struct file, now.

Kinda inconvenient because it means we'll need to patch
filp->private_data with filp->f_data in various places...

branches: 1.1.2; 1.1.6;
Make it build and boot on my test machines.

Screen blanks on boot on the Ivy Bridge system with

DRM error in cpt_serr_int_handler: PCH transcoder A FIFO underrun

But after that everything is OK. Appears to be an upstream problem.
To investigate...

I think there's a cache flushing issue somewhere -- there are little
display artefacts on my T60.

drm: Fix comments about wait return values.

drm: Ensure DRM_*WAIT* never counts down from MAX_SCHEDULE_TIMEOUT.

This faux timeout means `wait forever'.

hardclock_ticks -> getticks()

DRM_HZ -> HZ, to reduce diffs.

branches: 1.15.6;
fix various problems i've seen where cv_*wait*() return ERESTART,
which is -3 in netbsd, which we have mapped linux ERESTARTSYS to.

this has a problem because linux code often returns errors and
pointers in the same value, and pointer values between -4095 and
-1 are considered as error returns, but -3 ends up as 3 and thus
is not considered an error, and mayhem ensues.

with this in place my kabylake system seems actually stable, i
have not triggered any of my prior issues in almost 4 weeks now.


Taylor asked me to write up a description and then wrote most of
the text below for me :-)


In Linux code, we always work with ERESTARTSYS so the code meaning
start over is a positive NetBSD errno safe for PTR_ERR/ERR_PTR.
To achieve this:
1. adapt all cv_waits that return to Linux so they map ERESTART to
ERESTARTSYS, and
2. adapt all returns to userland so they convert ERESTARTSYS to
ERESTART.
Leave EINTR and all other error codes alone.

branches: 1.14.10; 1.14.18;
We now use cpu_intr_p() all the time.

Don't break when hardclock_ticks wraps around.

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.

In DRM_SPIN_WAIT_ON, don't stop after waiting only one tick.

Continue the loop to recheck the condition and count the whole
duration.

Actually poll every tick as advertised.

Remove extraneous blank line.

New macro DRM_SPIN_WAIT_ON better reflects DRM_WAIT_ON.

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.

Elaborate on essay about DRM_*WAIT*_UNTIL reutrn convention.

Write an essay to remind myself about (timed) wait return values.

Another attempt to fix the drm timed wait blarf blugh blahhh.

branches: 1.5.2;
Return 0, not uninitialized, if the condition is already true.

branches: 1.4.2; 1.4.4;
First whack at radeon driver.

No hardware to test yet, but it builds.

Make it build and boot on my test machines.

Screen blanks on boot on the Ivy Bridge system with

DRM error in cpt_serr_int_handler: PCH transcoder A FIFO underrun

But after that everything is OK. Appears to be an upstream problem.
To investigate...

I think there's a cache flushing issue somewhere -- there are little
display artefacts on my T60.

branches: 1.2.2; 1.2.4; 1.2.6;
Merge riastradh-drm2 to HEAD.

branches: 1.1.2;
file drm_wait_netbsd.h was initially added on branch riastradh-drm2.

drm_writeback stubs

stub drm_writeback_set_fb


Author: Maya Rashish <maya@NetBSD.org>

provide a stub drm_writeback_prepare_job, knf


Author: Maya Rashish <maya@NetBSD.org>

Remove GPL header for which we have our own code

stub drm_writeback_cleanup_job


Author: Maya Rashish <maya@NetBSD.org>

More drm_writeback stubs.

Disable GPL drm_writeback.c, minimal stub for drm_writeback.h.

amdgpu: Another pass through.

amdgpu_object.c
amdgpu_pci.c
amdgpu_pm.c
amdgpu_polaris10_smumgr.c
amdgpu_pp_psm.c
amdgpu_ppatomctrl.c
amdgpu_ppatomfwctrl.c
amdgpu_process_pptables_v1_0.o
amdgpu_psp.o
amdgpu_psp_v10_0.c
amdgpu_psp_v11_0.c
amdgpu_psp_v12_0.c
amdgpu_psp_v3_1.c
amdgpu_ras.c
amdgpu_ras_eeprom.c
amdgpu_rc_calc.c
amdgpu_rc_calc_dpi.c
amdgpu_ring.c
amdgpu_rlc.c
amdgpu_rn_clk_mgr.c
amdgpu_sdma_v3_0.c
amdgpu_sdma_v4_0.c
amdgpu_sdma_v5_0.c
amdgpu_si.c
amdgpu_si_dma.c
amdgpu_si_dpm.c
amdgpu_smu.c
amdgpu_smu10_hwmgr.c
amdgpu_smu7_hwmgr.c
amdgpu_smu8_hwmgr.c
amdgpu_smu8_smumgr.c
amdgpu_smu_v11_0.c (and a sprinkling of consts)
amdgpu_smu_v11_0_i2c.c
amdgpu_soc15.c
amdgpu_tonga_smumgr.c
amdgpu_ttm.c
amdgpu_ucode.c
amdgpu_uvd.c
amdgpu_uvd_v5_0.c
amdgpu_uvd_v6_0.c
amdgpu_uvd_v7_0.c
amdgpu_vce.c
amdgpu_vce_v2_0.c
amdgpu_vce_v4_0.c
amdgpu_vcn_v1_0.c
amdgpu_vcn_v2_0.c
amdgpu_vcn_v2_5.c
amdgpu_vega10_hwmgr.c
amdgpu_vega10_powertune.c
amdgpu_vega10_processpptables.c
amdgpu_vega10_reg_init.c
amdgpu_vega12_hwmgr.c
amdgpu_vega20_hwmgr.c
amdgpu_vega20_ppt.c
amdgpu_vega20_reg_init.c
amdgpu_vi.c
amdgpu_virt.c
amdgpu_vram_mgr.c
amdgpu_xgmi.c -- disable (for now?)
amdgpufb.c

branches: 1.2.14;
add da_fb_linebytes to drmfb_attach_args and use it to pass linebytes
from nouveau code to drmfb. keep the same linebytes logic for i915.

nvidia hardware needs 256 byte alignment, so aligning to just 64 was
not enough.

fixes broken console with a width of 1440px (PR kern/51181)
ok riastradh

branches: 1.1.2; 1.1.4; 1.1.6;
Factor generic parts of intelfb out into a drmfb.

Adapting radeondrmkmsfb should not be hard.

This will make nouveaufb easier too.

branches: 1.2.2; 1.2.4; 1.2.20;
Factor out intelfb_is_vga_console to drmfb_pci.

Factor generic parts of intelfb out into a drmfb.

Adapting radeondrmkmsfb should not be hard.

This will make nouveaufb easier too.

Make intel_psr.c build
netbsd-ify
Stub out intel_lpe_audio.c (for now?).
Hammer intel_opregion.c into shape.
Hack up intel_lrc.c
Adapt intel_reset.c
Stub out intel_region_lmem.c
Adapt intel_ring_submission.c
Adapt intel_rps.c
Adapt intel_sdvo.c
Adapt intel_runtime_pm.c
Adapt intel_sideband.c
Adapt intel_sprite.c
Adapt intel_tv.c
Adapt intel_tc.c
Adapt intel_sseu.c
Adapt intel_timeline.c
Adapt intel_uc_fw.c and fix typo
Fix up intel_uncore.c
intel_vga.c
intel_wakeref.c
intel_vdsc.c
intel_wopcm.c
intel_workarounds.c
i915_globals.c
Fix up intelfb.c
Stub out intel_guc_log.c
Stub out intel_dsi_dcs_backlight.c
i915_perf.c
intel_hdmi.c
intel_hdcp.c


Author: Maya Rashish <maya@NetBSD.org>
Committer: Taylor R Campbell <riastradh@NetBSD.org>

i915_mei_hdcp_interface.h from openbsd


Author: Maya Rashish <maya@NetBSD.org>
Committer: Taylor R Campbell <riastradh@NetBSD.org>

i915: Make intel_gtt_get prototype match Linux.

i915: ggtt

Update intel gtt API.

TODO: Find where the stolen memory base is in agp_i810.

Linux uses uint64 for these parameters now, so follow suit.

Call it intel_gtt_insert_sg_entries for expedience.

It uses a bus_dmamap, not a Linux scatterlist a.k.a. sg, but
otherwise it functions the same way and this avoids patching
upstream.

branches: 1.5.16; 1.5.18;
Pass cache-related flags through to the GTT on pre-SNB devices.

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!

branches: 1.4.2; 1.4.4; 1.4.6;
Make it build and boot on my test machines.

Screen blanks on boot on the Ivy Bridge system with

DRM error in cpt_serr_int_handler: PCH transcoder A FIFO underrun

But after that everything is OK. Appears to be an upstream problem.
To investigate...

I think there's a cache flushing issue somewhere -- there are little
display artefacts on my T60.

Explain the fields of struct intel_gtt for future reference.

branches: 1.2.2; 1.2.4; 1.2.6;
Merge riastradh-drm2 to HEAD.

branches: 1.1.2;
file intel-gtt.h was initially added on branch riastradh-drm2.

Move drm_io_mapping_* to its own header file.

Reimplement task barrier without semaphores.

match linux side-loading


Author: Maya Rashish <maya@NetBSD.org>

Rework ttm tt swapin/swapout logic.

Rather than handling `swapping in/out' here, per se, we let uvm do
that, we interpret `swap out' as `deactivate pages', and we add
generic ttm operations to wire and unwire pages, for the ttm_tt
driver to use.

This fixes certain graphics buffer eviction logic, which enables
nouveau to suspend/resume on one of my machines. (The machine
doesn't resume overall for other reasons, but the nouveau device
suspends and resumes in isolation.)

XXX pullup to netbsd-7 after a couple weeks

branches: 1.1.2; 1.1.4; 1.1.6; 1.1.8; 1.1.12;
First whack at radeon driver.

No hardware to test yet, but it builds.