Fix typo in uvm_km(9): there is no struct vm_map_kernel.

(riastradh)

Tweak i915 gen6_gtt_init calculations for clarity.

(riastradh)

Reject 32-bit paddrs on 965.

XXX Doing the check here is wrong; it serves only to report an
earlier problem, which is that there's on way to express constraints
on paddrs to uvm_obj_wirepages.  bus_dmamem_alloc can do this, but it
gives us pages out of thin air, not pages backing a uvm object.  I
was hoping this wouldn't manifest as a real problem, but evidently it
does.

(riastradh)

Fix >40-bit paddr error branch in i915_gem_object_get_pages_gtt.

(riastradh)

Fix error branch in drm_limit_dma_space.

(riastradh)

Cast from uint32_t to bus_addr_t earlier, since it may exceed 32 bits.

(riastradh)

Tweak some DRM GEM page indexing crap.

- Fix order of subtraction in drm_mmap_paddr_locked.
- Address GEM objects' pages from 0, not from the mmap cookie.
- Check page alignment earlier in mmap code paths.
- Sprinkle kasserts throughout.

Still doesn't fix the garbage that is sometimes being scribbled all
over kernel memory!

(riastradh)

Clear only the requested range in gen6_ggtt_clear_range.

Sprinkle kasserts throughout i915_gem_gtt.c.

(riastradh)

Convert pending_flip_lock to spin lock -- interrupt handlers take it.

(riastradh)

Replace cpu_relax() by DELAY(1) or DELAY(1000) to wait us/ms.

(riastradh)

Implement Linux dmi_check_system API.

Use it to get a Thinkpad quirk for i915 graphics.

Patch from nonoka@, PR 48708.

(riastradh)

Fix hot-plug with small patch from upstream before a full update.

(riastradh)

Take the console only from vga or generic genfb.

Adapted from a patch by nonaka@ in PR 48705.

While here, attach i915 genfb earlier as in PR 48706.

(riastradh)

Adapt Linux i915 wait_for to use DELAY so it works while cold.

Adapted from a patch by nonaka@ in PR 48706.

(riastradh)

Add applesmc(4) to {i386,amd64}/ALL.

(riastradh)

Reassign copyright to TNF.

(riastradh)

Fix include order.

(riastradh)

Correct type of i in execve_dovmcmds.  Fixes DEBUG_EXEC build.

(riastradh)

matt@ did the real work for the Beaglebone, so credit him properly.

My memory misled me when I added this entry the other day.

(riastradh)

Avoiding arithmetic on void * need not be NetBSD-specific.

(riastradh)

Note my paltry contributions.

(riastradh)

Fix note on __predict_true: predicts nonzero, not 1 specifically.

(riastradh)

Take vp->v_interlock before vdead_check in lfs_bmapv.

XXX This code is a pile of bodge that needs a serious rototill anyway.

(riastradh)

Don't forget ZFS_EXIT after ZFS_ENTER.

(riastradh)

Try vflush, which may fail, before we mark the file system mounted.

(riastradh)

Remove FFS_NO_SNAPSHOT from x86/ALL so we do compile the snapshot code.

(riastradh)

Add COMPAT_NETBSD32 and COMPAT_LINUX32 to amd64/ALL.

(riastradh)

Enable options VND_COMPRESSION in amd64/GENERIC.

(riastradh)

Move <linux/errno.h> from drm2 to common.

(riastradh)

Restore is_console=1 for i915, even if I keep it off for debugging.

(riastradh)

Add missing declaration from last commit.

(riastradh)

Tweak i915 framebuffer attachment in preparation for detach/reconfig.

Can't do this yet because genfb can't detach or reconfigure...

(riastradh)

Figure drm_debug into the drm_init module kludgerosity.

(riastradh)

Don't ifdef out unmapping of the GTT on NetBSD.

(riastradh)

Back out unintended part of last commit.

(riastradh)

Oops...  Fix sense of test in drmCheckModesettingSupported.

(riastradh)

Copy NO_DRM and DRMKMS kernels for i386.

(riastradh)

Miscellaneous fixes for drm2 on i386.

(riastradh)

Elide broken exists tests for TOOL_CTFCONVERT and TOOL_CTFMERGE.

Either you set MKCTF=yes and they should exist, or you set MKCTF=no
and it doesn't matter.  If MKCTF=yes and they don't exist, something
is broken, and this should fail noisily, not silently.

(riastradh)

By default define CTFCONVERT and CTFMERGE to be no-ops in sys.mk.

Don't test for /usr/bin/ctfconvert or /usr/bin/ctfmerge -- that is
completely bonkers and breaks the build if the host's ctfconvert and
ctfmerge are too old.

Strip the broken conditionals in the .c.o &c. rules.

bsd.own.mk will override CTFCONVERT and CTFMERGE with TOOL_CTFCONVERT
and TOOL_CTFMERGE if MKCTF=yes.

XXX The way CTFCONVERT and CTFMERGE are integrated into the rules is
bogus -- it is not safe to interrupt them.  We need some intermediate
object to do this right.

Please let me know if I broke anything with this -- I am not an elvin
or dwarven wizard, nor a dtrace expert.  But this has been breaking
builds for ages now and nobody else seems to want to do anything
about it.

(riastradh)

ctfconvert/ctfmerge need libdwarf &c., so build them if MKCTF=yes.

(riastradh)

Skip SNA on i386 for now, since it triggers a GCC ICE.

(riastradh)

Correct return type of iic_release_bus.

(riastradh)

Remove ifdef for const fix.

(riastradh)

Need <sys/reboot.h> for AB_DEBUG.

(riastradh)

Nix accidental whitespace to reduce diff from Linux.

(riastradh)

Make i915drmkms grab the console by default.

(riastradh)

Enable drm debug output iff boothowto has AB_DEBUG set.

(riastradh)

Fix up removal of _DRM_SHM mappings and the hardware lock.

(riastradh)

Use drmMap rather than mmap directly, to invoke the ioctl on NetBSD.

(riastradh)

Update reachover makefiles and set lists for xf86-video-intel 2.21.15.

(riastradh)

merge xf86-video-intel-2.21.15

(riastradh)

Link Xorg with -lpthread on x86.

This will be needed by the new xf86-video-intel, which uses pthreads
in SNA (SandyBridge New Acceleration).  It shouldn't hurt, but is not
needed, on any other architecture.

(riastradh)

Install compositeext.h and shmint.h from X.org.

(riastradh)

Remove vestigial modules/apple_smc/apple_smc.ioconf.

(riastradh)

Fix unused variable warning in non-modular apple_smc build.

(riastradh)

Fix apple_smc config in conf/files.

(riastradh)

Fix copypasta error in apple_smc_write_key_4: uint32_t, not uint16_t.

(riastradh)

Polish the Apple SMC code with commentary and little fixes.

(riastradh)

Rework Apple SMC device attachment goo again.

Less bookkeeping at the expense of iteration over all devices when
rescanning applesmc.

(riastradh)

Rework Apple SMC attachment goo and split into multiple modules.

The four modules are apple_smc for the core functions, apple_smc_acpi
for attachment at acpi, and apple_smc_fan & apple_smc_temp for stuff
on the SMC.

Seems like there's a lot more bookkeeping necessary to maintain
children of an applesmc device implemented by loadable modules.
Maybe there's a better way I just haven't figured out yet.

(riastradh)

First draft of drivers for the Apple System Management Controller.

Device interface derived by reading the Linux driver source code and
<http:///www.parhelia.ch/blog/statics/k3_keys.html> as of 2012-12-05.

Includes support for attaching fan and temperature sensors to sysmon.
No accelerometer yet.

Compile-tested only, based on some run-testing of experiments from
userland.  Module attachment is not quite finished, so it won't work
yet.

(riastradh)

Fix volatile qualifiers in drm2 <linux/atomic.h>.

(riastradh)

Explicitly kassert the precondition c_done != 0 in _completion_claim.

(riastradh)

Move <linux/completion.h> from drm2 to external/bsd/common.

(riastradh)

kernel.h and list.h have already been moved to external/bsd/common.

With apologies to future revision control systems for not making it
clear this is effectively a rename.

(riastradh)

Make drmkms find header files from sys/external/bsd/common.

(riastradh)

Merge changes from drm2 <linux/kernel.h> to common <linux/kernel.h>.

(riastradh)

Add some more Linux completion operations.

Adapted from a patch by nick@.

(riastradh)

Move the drmkms i2c dependencies to where they belong.

(riastradh)

Don't leak KVA by forgetting to call uvm_km_free...oops!

From rmind@.

(riastradh)

Remove needless splvm/splx calls.

From rmind@, who explained that the pmap(9) man page is wrong.

(riastradh)

Simplify synchronization between umcs(4) intr and task.

ok martin@, nick@

(riastradh)

Constify list_is_empty.

With apologies to whoever will merge a newer version.  Fortunately,
this file is getting replaced wholesale.

(riastradh)

Use DRM_IOCTL_MODE_GETRESOURCES to detect KMS on any non-Linux.

(riastradh)

Use #ifdef DRM_IOCTL_MMAP, not #ifdef __NetBSD__.

Initialize with `= {0}' rather than a temporary static constant
variable.

Requested by christos@.

(riastradh)

Build xf86drmMode.c into libdrm.

(riastradh)

Switch intel_bufmgr_gem to use drmMap instead of mmap directly.

(riastradh)

Use ioctl, not sysfs, to detect KMS support on NetBSD.

(riastradh)

Use the NetBSD-specific mmapping ioctl for drmMap.

(riastradh)

Make uninitialized_var kludge expand to `x = 0'.

Forgot to commit this the other day.

(riastradh)

Update libfreetype debug data version.  Hi mrg!

(riastradh)

Don't choke on an assertion for an RSDT with a null address entry.

(riastradh)

Don't unmap Rsdp until we've finished using it.

XXX Let me know how this is supposed to be upstreamed if you know and
have time...

(riastradh)

Clarify some wording in getaddrinfo.1.

(riastradh)

Fix uninitialized variable wait in intel_ddi_prepare_link_retrain.

Move a line around, too, to avoid merge conflicts with upstream.

Noted by maxv@.

(riastradh)

Fix unused variable warnings in drm2.

(riastradh)

Merge riastradh-drm2 to HEAD.

(riastradh)

Avoid undefined behaviour in shifts in endian decoding routines.

If int is 32-bit and p is a uint8_t *, then p[0] is promoted to int
and p[0] << 24 can shift a one into the sign bit of an int, which is
nasal demon territory.

(riastradh)

Convert right-recursive rules to left-recursive ones.

This should obviate the need for the workaround of a large stack in
order to handle many rules.

No change in the resulting plists.

ok rmind

(riastradh)

Fix subword atomic_cas.

- Don't loop forever on failure.
- Don't shift away the result and return zero on success.

Evidently we lack automatic tests for these!

(riastradh)

Fix return value of wait_for_completion_interruptible_timeout.

(riastradh)

Make gem mmappings shared, not copy-on-write, like Linux.

(riastradh)

`Controller' takes two l's.

(riastradh)

Take the vmobjlock in i915_gem_release_mmap around pmap_page_protect.

(riastradh)

Fix up i915_gem_gtt for agp-style gen<6 devices.

- Set the DMA mask correctly: gen<4 needs 32-bit; gen4+, 36-bit.
- Use agp_i810_write_gtt_entry, not AGP_BIND_PAGE.
- Do a posting read after changing all the GTT entries.

(riastradh)

Expose agp_i810_write_gtt_entry for drm2.

Also add an agp_i810_post_gtt_entry to do a posting read for that GTT
entry.

(riastradh)

Implement and use non-interruptible DRM_WAIT_* gizmos.

(riastradh)

Kludgey attempt to fill in intel_gtt stubs.

(riastradh)

Oops -- define struct intel_gtt::do_idle_maps.

Missed this in last commit.

(riastradh)

First draft of i915_gem_gtt.c for pre-Sandy Bridge (gen<6) devices.

These go through the legacy AGP cruft to get at the GTT.

(riastradh)

Expose the agp_i810 softc.

This is an Intel-specific kludge so that i915drmkms can get at the
GTT, graphics translation table, for pre-Sandy Bridge Intel graphics
devices.

(riastradh)

Handle attach failure and don't try a framebuffer in i915 on failure.

(riastradh)

Remove drm2 and i915drm2 modules.

(riastradh)

Rework drm2 configuration and modularization.

- Split drm2.kmod into drmkms_linux, drmkms, drmkms_pci.
- Split up drmkms-related files.* similarly.
- Make drm_agp_* routines hooks that the drmkms_pci module installs.
- Reuse drm device major 180 for (old) drm and drmkms.

Now old drm and new drmkms can coexist happily in the same tree, with
or without PCI, and require no changes to userland.

(riastradh)

Kick on-demand entropy sources in rndsinks_distribute.

Partial workaround for indefinite hangs when entropy is scarce or
buffered up.  We need to do more to handle entropy that has been
buffered up -- see the comment for details -- but this will help for
now.

Problem noted by pooka.

(riastradh)

Fix bits/bytes mixup in rnd_getmore.

Remove some needless casts and fix format directives while here.

Bit/byte mixup noticed by pooka.

(riastradh)

__HAVE_ATOMIC_AS_MEMBAR is spelled with two leading underscores.

This underscores the need to replace this error-prone cpp API by
unconditionally defined {pre,post}atomic_membar_*.

This change should only remove unnecessary membar_producers on x86.

(riastradh)

Make wsdisplay_vcons likelier to be modularizable too.

(riastradh)

Redo amd64/NO_DRM config and add amd64/DRMKMS.

(riastradh)

Tweak genfb tweaks so they work.

(riastradh)

Make genfb a little friendlier to modules.

This is not enough to make it actually work as a module itself, but
it's enough to make the i915drm module use genfb.

(riastradh)

Make genfb replay the msgbuf only if it's becoming the console.

(riastradh)

Make genfb_attach configure the wsemuldisplaydev attribute.

(riastradh)

Make i915drm_match return 6 to beat genfb_pci no matter what.

(riastradh)

Give drmkms its own device major so it doesn't compete with old drm.

(riastradh)

First draft of in-kernel drm2 attachment.

The drm2 `drm' device is now called `drmkms' (likewise `i915drmkms',
&c.) so that it can coexist with the old drm code to reduce diffs
from HEAD.  (Can't call a device `drm2'.)

(riastradh)

Defer i915drm_attach_framebuffer until interrupts are running.

The i915 code initialization relies on counting hardclock ticks for a
delay (ugh).

Not an issue for modules, but it will matter when we build drm2 into
the kernel proper.

(riastradh)

Include genfb in NO_DRM kernel rather than modularizing.

Modularizing required a bunch of hacks (wscons/genfb/&c. is not ready
for modules) and failed to pull in rasops dependencies, so rasops
never worked correctly.

(riastradh)

Get SIZE_MAX and OFF_MAX straight...

(riastradh)

Factor out an offtab_compute_window_position routine.

(riastradh)

Fix some more integer overflow/truncation issues.

Arithmetic in C is hard.  Let's go shopping!

(riastradh)

Little reference counting fix in drm_mmap.

(riastradh)

Tidy up and reuse the code I had already written for gem mmapping.

(riastradh)

Implement mmap on a drm device by a new kludgey ioctl.

There is currently no way to set up a custom choice of uvm object and
uvm pager ops for mmap: struct fileops has no fo_mmap, and device
nodes get to say only what paddr each offset maps to, not what uvm
object to use.  For now, to avoid kernel changes outside the drm2
tree, we'll use a non-standard drm ioctl that calls uvm_map.  This
requires source patches to userland programs, but for the most part
they can simply be converted from calling mmap to drmMap, which will
be adapted to use the ioctl.

This is not necessarily a permanent solution.

(riastradh)

Label fail4 is now used, so don't mark it __unused.

(riastradh)

Remove spurious newline to reduce diff from upstream.

(riastradh)

Fix queue_delayed_work edge cases.

- Copy the relevant part of queue_work in-line for ticks == 0, since
  queue_work itself will choke on a delayed work.

- Don't reschedule the callout if it is already scheduled -- let it
  fire when it was already scheduled to fire.

(riastradh)

Take irq_lock before waking up irq_queue in i915_handle_error.

(riastradh)

Fix $NetBSD$ tag.

(riastradh)

Bump date on vndcompress(1) man page.

(riastradh)

Fix vndcompress man page to reflect default window size.

(riastradh)

Change vndcompress to use a default window size of 512.

For vnduncompress on nonseekable input, the window size is as large
as it needs to be by default, as before.  Not clear that this is the
right choice -- by default vnduncompress on nonseekable input will
just use unbounded memory unsolicited.

(riastradh)

Document the new vndcompress -w option and nuke BUGS section.

Perhaps vndcompress and vnduncompress ought by default to choose a
limited window size (say, 8192 entries, i.e. 64k bytes, the default
MAXPHYS), and vnduncompress should fall back to an unlimited window
only if the input is nonseekable.

(riastradh)

Reflect rename of `-s' to `-b' in the vndcompress man page.

(riastradh)

Move err1 & errx1 to the end of vnduncompress.c; add __printflike.

(riastradh)

Rename block size option from `-s' to `-b'.

Makes more sense and makes it consistent with other utilities such as
pax and pigz.  This vndcompress has never gone out in a release, so
changing the name of the option shouldn't cause too many problems...

(riastradh)

Simplify vndcompress offtab_compute_window_size.

(riastradh)

Fix typo in comment.

(riastradh)

Fix up ulimited vndcompress tests and explain what's up with them.

(riastradh)

Add some simple automatic tests for window sizes.

(riastradh)

Remove silly comment in offtab_reset_write.

(riastradh)

Window size is now an option; remove XXX comment to the contrary.

(riastradh)

Add comment explaining why piperestart.cl2part is allowed to fail.

(riastradh)

Split guard in offtab_write_window into offtab_maybe_write_window.

(riastradh)

Seek if necessary at end of offtab_reset_read.

Fixes vnduncompress with a small window, and makes offtab_reset_read
symmetric with offtab_reset_write.

(riastradh)

Add option -w to vnd(un)compress to specify the window size.

(riastradh)

Add some leading zero digits to the flags.  Cosmetic change only.

(riastradh)

Add WARNS=5.

(riastradh)

Judicious (and justified) casts to avoid signed/unsigned comparisons.

(riastradh)

Implement machinery for fixed-size windows into the offset table.

(riastradh)

Write offsets in hexadecimal, not decimal.

(riastradh)

Move block_signals/restore_sigmask to utils.c

(riastradh)

Use write-to-temporary/rename-to-permanent pattern in Makefile.

(riastradh)

Abstract handling of the cloop2 offset table.

Preparation for converting it to use a fixed-size window.

(riastradh)

Use read_block instead of read in vnduncompress.

(riastradh)

Move vndcompress utilities to utils.c.

(riastradh)

Fail if malloc can't allocate offset table.

(riastradh)

Return EINTR too in __wait_seqno.

(riastradh)

Call uvm_vm_page_to_phys before taking a spin lock to ease debugging.

(riastradh)

Call drm_mode_group_init_legacy_group in drm_attach like Linux does.

(riastradh)

Adjust errno from dev->driver->load.

(riastradh)

Initialize the right minor numbers...arghhhh.

(riastradh)

uobj->pgops->pgo_put requires uobj->vmobjlock.

(riastradh)

Remove vestige of wsdisplay XXX from drm_drv.c.

wsdisplay stuff happens in the parent driver now.

(riastradh)

Oops -- unifdef <sys/select.h> and <sys/poll.h> in drm_drv.c.

(riastradh)

Report st_rdev in stat on a drm device.

(riastradh)

Just use kcalloc in drm_malloc_ab.

(riastradh)

Fix order of arguments in DRM_COPY_TO_USER.

...

(riastradh)

Make Linux kmalloc handle a few more gfp flags.

(riastradh)

i915_gem_object_get_page takes a page number, not byte offset.

(riastradh)

Fix i915_genfb_mmap base addresses and offsets.

(riastradh)

Initialize the nsec correctly in ktime_get.

(riastradh)

Use nanouptime, not nanotime, for ktime_get.

As far as I can tell, ktime_get is supposed to provide monotonic
time, not a clock synchronized to TAI.

(riastradh)

Hack up vblank events and make them work.

(riastradh)

Unconditionalize the genfb attachment.

(riastradh)

Pesky ampersands!

(riastradh)

Take master->lock.spinlock to read master->lock.hw_lock.

(riastradh)

Clean up initialization failures in drm2_modcmd.

(riastradh)

Rewrite drm locking support for NetBSD.

(riastradh)

Clean up the still-not-yet-functional genfb attachment cruft a bit.

(riastradh)

Use dev_priv->irq_lock for ring->irq_queue, not dev->struct_mutex.

We need to use a spin lock here, because we need to exclude interrupt
handlers.

(riastradh)

Implement drm_pci_set_busid and drm_pci_set_unique.

(riastradh)

Drop another reference in do_destroy if the context is current.

(riastradh)

Make it a bug to drm_mm_takedown when there remain allocated blocks.

(riastradh)

Destroy the waitqueue ring->irq_queue in intel_cleanup_ring_buffer.

(riastradh)

Add open drm files to the device file list.

(riastradh)

Fix copyoutstr usage in drm_version_string.

(riastradh)

Make DRM_IOCTL_GET_CLIENT work except for pid and uid.

The pid and uid don't make sense because the client is associated
with a file descriptor, which may be held by multiple processes, and
even if only one holds it (which is probably the only sensible state
of affairs), there's no index.

(riastradh)

Lock the right uvmobject for uvm_pagelookup.

(riastradh)

Convert linux_idr(9) to use spin locks.

idr_for_each is unlocked; caller must exclude modifications.

Eventually this should be rewritten with pserialize and a fancy
bitmap algorithm, but this *actually* works for now.

(riastradh)

Revert "Rewrite idr to use a dumber algorithm that admits pserialized use."

This reverts commit 3a389a1cb20777fb73575f0514b96265052ac1ea.

I don't know what I was smoking with this; just need to change the
rwlock to a spin lock and we'll be good!

(riastradh)

Don't put work flush items on the stack; worker kmem_frees them.

(riastradh)

Rewrite idr to use a dumber algorithm that admits pserialized use.

drm2 doesn't use them with RCU, but it does use them under spin locks,
so an rwlock is not kosher.

This algorithm is super-dumb, but the idr API has changed upstream,
and this is not performance-critical, so it's not worth investing
time in a better algorithm at the moment.

(riastradh)

flush_workqueue should work now, so don't work around it i915 unload.

(riastradh)

i915_gem_context_close destroys the idr; i915_driver_postclose needn't.

(riastradh)

Detach framebuffer stuff before drm device in i915drm_detach.

Detaching the drm device calls the i915_driver_unload routine, which
nukes the i915 private data structures, so that we can't use them any
more to detach the framebuffer.

Now i915drm2 unloads!

(riastradh)

Initialize and destroy the pending flip lock.

(riastradh)

Avoid dividing by zero when allocating empty array in kcalloc.

(riastradh)

Tweak drm_fb_helper_single_fb_probe so we needn't reconnect the screen.

(riastradh)