There are too many oustanding issues:

- Fence handling in the current code is broken. There's a patch series
  from me, but it's blocked on and extended review (which includes
  writing the testcases).

- IOMMU mapping handling is broken, we need to properly refcount it -
  currently it gets destroyed when the first vma is unbound, so way
  too early.

- There's a pending reset issue on snb. Since Mika's reset work and
  full ppgtt have been pulled in in separate branches and ended up
  intermittingly breaking each another it's unclear who's the exact
  culprit here.

- We still have persistent evidince of crazy recursion bugs through
  vma_unbind and ppgtt_relase, e.g.

  https://bugs.freedesktop.org/show_bug.cgi?id=73383

  This issue (and a few others meanwhile resolved) have blocked our
  performance measuring/tuning group since 3 months.

- Secure batch dispatching is broken. This is blocking Brad Volkin's
  command checker work since 3 months.

All these issues are confirmed to only happen when full ppgtt is
enabled, falling back to aliasing ppgtt resolves them. But even
aliasing ppgtt itself still has a regression:

- We currently unconditionally bind objects into the aliasing ppgtt,
  which means all priviledged objects like ringbuffers are visible to
  unpriviledged access again. On top of that this also breaks the
  command checker for aliasing ppgtt, since it can't hide the
  validated batch any more.

Furthermore topic/full-ppgtt has never been reviewed:

- Lifetime rules around vma unbinding/release are unclear, resulting
  into this awesome hack called ppgtt_release. Which seems to take the
  blame for most of the recursion fallout.

- Context/ring init works different on gpu reset than anywhere else.
  Such differeneces have in the past always lead to really hard to
  track down bugs.

- Aliasing ppgtt is treated in a bunch of places as a real address
  space, but it isn't - the real address space is always the global
  gtt in that case. This results in a bit a mess between contexts and
  ppgtt object, further complication the context/ppgtt/vma lifetime
  rules.

- We don't have any docs describing the overall concepts introduced
  with full ppgtt. A short, concise overview describing vmas and some
  of the strange bits around them (like the unbound vmas used by
  execbuf, or the new binding rules) really is needed.

Note that a lot of the post topic/full-ppgtt merge fallout has already
been addressed, this entire list here of 10 issues really only contains
the still outstanding issues.

Finally the 3.15 merge window is approaching and I think we need to
use the remaining time to ensure that our fallback option of using
aliasing ppgtt is in solid shape. Hence I think it's time to throw the
switch. While at it demote the helper from static inline status
because really.

Cc: Ben Widawsky <ben@bwidawsk.net>
Cc: Dave Airlie <airlied@gmail.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

With the original PPGTT implementation if the number of PDPs was not a
power of two, the number of pages for the page tables would end up being
rounded up. The code actually had a bug here afaict, but this is a
theoretical bug as I don't believe this can actually occur with the
current code/HW..

With the rework of the page table allocations, there is no longer a
distinction between number of page table pages, and number of page
directory entries. To avoid confusion, kill the redundant (and newer)
struct member.

Cc: Imre Deak <imre.deak@intel.com>
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Reviewed-by: NImre Deak <imre.deak@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Simply to match the GEN8 style of PPGTT initialization, split up the
allocations and mappings. Unlike GEN8, we skip a separate dma_addr_t
allocation function, as it is much simpler pre-gen8.

With this code it would be easy to make a more general PPGTT
initialization function with per GEN alloc/map/etc. or use a common
helper, similar to the ringbuffer code. I don't see a benefit to doing
this just yet, but who knows...
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This cleanup is similar to the GEN8 cleanup (though less necessary).
Having everything split will make cleaning the initialization path error
paths easier to understand.
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

I keep meaning to do this... by now almost the entire file has been
written by an Intel employee (including Daniel post-2010).
Reviewed-by: NImre Deak <imre.deak@intel.com>
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This reverts commit 3a2ffb65.

Now that the code is fixed to use smaller allocations, it should be safe
to let the full GGTT be used on BDW.

The testcase for this is anything which uses more than half of the GTT,
thus eclipsing the old limit.
Reviewed-by: NImre Deak <imre.deak@intel.com>
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The previous allocation mechanism would get 2 contiguous allocations,
one for the page directories, and one for the page tables. As each page
table is 1 page, and there are 512 of these per page directory, this
goes to 2MB. An unfriendly request at best. Worse still, our HW now
supports 4 page directories, and a 2MB allocation is not allowed.

In order to fix this, this patch attempts to split up each page table
allocation into a single, discrete allocation. There is nothing really
fancy about the patch itself, it just has to manage an extra pointer
indirection, and have a fancier bit of logic to free up the pages.

To accommodate some of the added complexity, two new helpers are
introduced to allocate, and free the page table pages.

NOTE: I really wanted to split the way we do allocations, and the way in
which we identify the page table/page directory being used. I found
splitting this functionality up to be too unwieldy. I apologize in
advance to the reviewer. I'd recommend looking at the result, rather
than the diff.

v2/NOTE2: This patch predated commit:
6f1cc993
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Tue Dec 31 15:50:31 2013 +0000

    drm/i915: Avoid dereference past end of page arr

It fixed the same issue as that patch, but because of the limbo state of
PPGTT, Chris patch was merged instead. The excess churn is a result of
my using my original patch, which has my preferred naming. Primarily
act_* is changed to which_*, but it's mostly the same otherwise. I've
kept the convention Chris used for the pte wrap (I had something
slightly different, and broken - but fixable)

v3: Rename which_p[..]e to drop which_ (Chris)
Remove BUG_ON in inner loop (Chris)
Redo the pde/pdpe wrap logic (Chris)

v4: s/1MB/2MB in commit message (Imre)
Plug leaking gen8_pt_pages in both the error path, as well as general
free case (Imre)

v5: Rename leftover "which_" variables (Imre)
Add the pde = 0 wrap that was missed from v3 (Imre)
Reviewed-by: NImre Deak <imre.deak@intel.com>
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
[danvet: Squash in fixup from Ben.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This patch converts insert_entries and clear_range, both functions which
are specific to the VM. These functions tend to encapsulate the gen
specific PTE writes. Passing absolute addresses to the insert_entries,
and clear_range will help make the logic clearer within the functions as
to what's going on. Currently, all callers simply do the appropriate
page shift, which IMO, ends up looking weird with an upcoming change for
the gen8 page table allocations.

Up until now, the PPGTT was a funky 2 level page table. GEN8 changes
this to look more like a 3 level page table, and to that extent we need
a significant amount more memory simply for the page tables. To address
this, the allocations will be split up in finer amounts.

v2: Replace size_t with uint64_t (Chris, Imre)

v3: Fix size in gen8_ppgtt_init (Ben)
Fix Size in i915_gem_suspend_gtt_mappings/restore (Imre)

Reviewed-by: Imre Deak <imre.deak@intel.com> (v2)
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Like cleanup in an earlier patch, the code becomes much more readable,
and easier to extend if we extract out helper functions for the various
stages of init.

Note that with this patch it becomes really simple, and tempting to begin
using the 'goto out' idiom with explicit free/fini semantics. I've
kept the error path as similar as possible to the cleanup() function to
make sure cleanup is as robust as possible

v2: Remove comment "NB:From here on, ppgtt->base.cleanup() should
function properly"
Update commit message to reflect above

v3: Rebased on top of bugfixes found in the previous patch by Imre
Moved number of pd pages assertion to the proper place (Imre)

v4:
Allocate dma address space for num_pd_pages, not num_pd_entries (Ben)
Don't use gen8_pt_dma_addr after free on error path (Imre)
With new fix from v4 of the previous patch.
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Reviewed-by: NImre Deak <imre.deak@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Create 3 clear stages in PPGTT init. This will help with upcoming
changes be more readable. The 3 stages are, allocation, dma mapping, and
writing the P[DT]Es

One nice benefit to the patches is that it makes 2 very clear error
points, allocation, and mapping, and avoids having to do any handling
after writing PTEs (something which was likely buggy before). This
simplified error handling I suspect will be helpful when we move to
deferred/dynamic page table allocation and mapping.

The patches also attempts to break up some of the steps into more
logical reviewable chunks, particularly when we free.

v2: Don't call cleanup on the error path since that takes down the
drm_mm and list entry, which aren't setup at this point.

v3: Fixes addressing Imre's comments from:
<1392821989.19792.13.camel@intelbox>

Don't do dynamic allocation for the page table DMA addresses. I can't
remember why I did it in the first place. This addresses one of Imre's
other issues.

Fix error path leak of page tables.

v4: Fix the fix of the error path leak. Original fix still leaked page
tables. (Imre)
Reviewed-by: NImre Deak <imre.deak@intel.com>
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

GEN8 never freed the PPGTT struct. As GEN8 doesn't use full PPGTT, the
leak is small and only found on a module reload. ie. I don't think this
needs to go to stable.

v2: The very naive, kfree in gen8 ppgtt cleanup, is subject to a double
free on PPGTT initialization failure. (Spotted by Imre). Instead this
patch pulls the ppgtt struct freeing out of the cleanup and leaves it to
the allocators/callers or the one doing the last kref_put as in standard
convention
Reported-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Reviewed-by: NImre Deak <imre.deak@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

There's no need not to, really.

Split out from Chris vma-bind rework.

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ben Widawsky <benjamin.widawsky@intel.com>
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Anything more than just one bool parameter is just a pain to read,
symbolic constants are much better.

Split out from Chris' vma-binding rework patch.

v2: Undo the behaviour change in object_pin that Chris spotted.

v3: Split out misplaced hunk to handle set_cache_level errors,
spotted by Jani.

v4: Keep the current over-zealous binding logic in the execbuffer code
working with a quick hack while the overall binding code gets shuffled
around.

v5: Reorder the PIN_ flags for more natural patch splitup.

v6: Pull out the PIN_GLOBAL split-up again.

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ben Widawsky <benjamin.widawsky@intel.com>
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This will make the code more readable, and extensible which is needed
for upcoming feature work. Eventually, we'll do the same for init.
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

With 20+ module parameters, I think referring to them via a struct
improves clarity over just having a bunch of globals. While at it, move
the parameter initialization and definitions into a new file
i915_params.c to reduce clutter in i915_drv.c.

Apart from the ill-named i915_enable_rc6, i915_enable_fbc and
i915_enable_ppgtt parameters, for which we lose the "i915_" prefix
internally, the module parameters now look the same both on the kernel
command line and in code. For example, "i915.modeset".

The downsides of the change are losing static on a couple of variables
and not having the initialization and module_param_named() right next to
each other. On the other hand, all module parameters are now defined in
one place at i915_params.c. Plus you can do this to find all module
parameter references:

$ git grep "i915\." -- drivers/gpu/drm/i915

v2:
- move the definitions into a new file
- s/i915_params/i915/
- make i915_try_reset i915.reset, for consistency
Signed-off-by: NJani Nikula <jani.nikula@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

A regression in the topic/ppgtt branch introduce in

commit 87d60b63
Author: Ben Widawsky <ben@bwidawsk.net>
Date:   Fri Dec 6 14:11:29 2013 -0800

    drm/i915: Add PPGTT dumper

The issue is that we're missing the definitions for the seq_file
functions and hence compilation fails.

v2: Just include the right header instead of splattering #ifdefs all
over the place (Chris).

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reported-by: Nkbuild test robot <fengguang.wu@intel.com>
Reported-by: NAntti Koskipaa <antti.koskipaa@linux.intel.com>
Cc: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

We're dealing with CPU physical addresses here, which may be different from
bus addresses, so rename gtt_bus_addr to gtt_phys_addr to avoid confusion.

No functional change.
Signed-off-by: NBjorn Helgaas <bhelgaas@google.com>
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The bug from gen6_ppgtt_insert_entries() was replicated into
gen8_ppgtt_insert_entries(). This applies the fix for the OOPS from the
previous patch to the gen8 routine.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Ben Widawsky <benjamin.widawsky@intel.com>
Reviewed-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

[   89.237347] BUG: unable to handle kernel paging request at ffff880096326000
[   89.237369] IP: [<ffffffff81347227>] gen6_ppgtt_insert_entries+0x117/0x170
[   89.237382] PGD 2272067 PUD 25df0e067 PMD 25de5c067 PTE 8000000096326060
[   89.237394] Oops: 0000 [#1] SMP DEBUG_PAGEALLOC
[   89.237404] CPU: 1 PID: 1981 Comm: gem_concurrent_ Not tainted 3.13.0-rc4+ #639
[   89.237411] Hardware name: Intel Corporation 2012 Client Platform/Emerald Lake 2, BIOS ACRVMBY1.86C.0078.P00.1201161002 01/16/2012
[   89.237420] task: ffff88024c038030 ti: ffff88024b130000 task.ti: ffff88024b130000
[   89.237425] RIP: 0010:[<ffffffff81347227>]  [<ffffffff81347227>] gen6_ppgtt_insert_entries+0x117/0x170
[   89.237435] RSP: 0018:ffff88024b131ae0  EFLAGS: 00010286
[   89.237440] RAX: ffff880096325000 RBX: 0000000000000400 RCX: 0000000000001000
[   89.237445] RDX: 0000000000000200 RSI: 0000000000000001 RDI: 0000000000000010
[   89.237451] RBP: ffff88024b131b30 R08: ffff88024cc3aef0 R09: 0000000000000000
[   89.237456] R10: 0000000000000000 R11: 0000000000000000 R12: ffff88024cc3ae00
[   89.237462] R13: ffff88024a578000 R14: 0000000000000001 R15: ffff88024a578ffc
[   89.237469] FS:  00007ff5475d8900(0000) GS:ffff88025d020000(0000) knlGS:0000000000000000
[   89.237475] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   89.237480] CR2: ffff880096326000 CR3: 000000024d531000 CR4: 00000000001407e0
[   89.237485] Stack:
[   89.237488]  ffff880000000000 0000020000000000 ffff88024b23f2c0 0000000100000000
[   89.237499]  0000000000000001 000000000007ffff ffff8801e7bf5ac0 ffff8801e7bf5ac0
[   89.237510]  ffff88024cc3ae00 ffff880248a2ee40 ffff88024b131b58 ffffffff813455ed
[   89.237521] Call Trace:
[   89.237528]  [<ffffffff813455ed>] ppgtt_bind_vma+0x3d/0x60
[   89.237534]  [<ffffffff8133d8dc>] i915_gem_object_pin+0x55c/0x6a0
[   89.237541]  [<ffffffff8134275b>] i915_gem_execbuffer_reserve_vma.isra.14+0x5b/0x110
[   89.237548]  [<ffffffff81342a88>] i915_gem_execbuffer_reserve+0x278/0x2c0
[   89.237555]  [<ffffffff81343d29>] i915_gem_do_execbuffer.isra.22+0x699/0x1250
[   89.237562]  [<ffffffff81344d91>] ? i915_gem_execbuffer2+0x51/0x290
[   89.237569]  [<ffffffff81344de6>] i915_gem_execbuffer2+0xa6/0x290
[   89.237575]  [<ffffffff813014f2>] drm_ioctl+0x4d2/0x610
[   89.237582]  [<ffffffff81080bf1>] ? cpuacct_account_field+0xa1/0xc0
[   89.237588]  [<ffffffff81080b55>] ? cpuacct_account_field+0x5/0xc0
[   89.237597]  [<ffffffff811371c0>] do_vfs_ioctl+0x300/0x520
[   89.237603]  [<ffffffff810757a1>] ? vtime_account_user+0x91/0xa0
[   89.237610]  [<ffffffff810e40eb>] ?  context_tracking_user_exit+0x9b/0xe0
[   89.237617]  [<ffffffff81083d7d>] ? trace_hardirqs_on+0xd/0x10
[   89.237623]  [<ffffffff81137425>] SyS_ioctl+0x45/0x80
[   89.237630]  [<ffffffff815afffa>] tracesys+0xd4/0xd9
[   89.237634] Code: 5b 41 5c 41 5d 41 5e 41 5f 5d c3 66 0f 1f 84 00 00 00 00 00 83 45 bc 01 49 8b 84 24 78 01 00 00 65 ff 0c 25 e0 b8 00 00 8b 55 bc <4c> 8b 2c d0 65 ff 04 25 e0 b8 00 00 49 8b 45 00 48 c1 e8 2d 48
[   89.237741] RIP  [<ffffffff81347227>] gen6_ppgtt_insert_entries+0x117/0x170
[   89.237749]  RSP <ffff88024b131ae0>
[   89.237753] CR2: ffff880096326000
[   89.237758] ---[ end trace 27416ba8b18d496c ]---

This bug dates back to the original introduction of the
gen6_ppgtt_insert_entries()
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Ben Widawsky <benjamin.widawsky@intel.com>
Reviewed-by: NBen Widawsky <ben@bwidawsk.net>
[danvet: Dropped cc: stable since without full ppgtt there's no way
we'll access the last page directory with this function since that
range is occupied (only in the allocator) with the ppgtt pdes. Without
aliasing we can start to use that range and blow up.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The iommu and gfx on Ironlake do not like each other and require a
big hammer to prevent hard machine hangs. In

commit 5c042287
Author: Ben Widawsky <ben@bwidawsk.net>
Date:   Mon Oct 17 15:51:55 2011 -0700

    drm/i915: ILK + VT-d workaround

we added the workaround, but never emitted any debug message that it was
active. Doing so should help identify known performance regressions.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

gem_gtt_cpu_tlb seems to indicate that it is needed.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=72869Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Dump the aliasing PPGTT with it. The aliasing PPGTT should actually
always be empty.

TODO: Broadwell. Since we don't yet use full PPGTT on Broadwell, not
having the dumper is okay.
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Originally this commit message said:
Now that do_switch does the mm switch, and we always enable the aliasing
PPGTT, and contexts at the same time, there is no need to continue doing
this during PPGTT enabling.

Since originally writing the patch however, I introduced the concept of
synchronous mm switching (using MMIO). Since this is generally not
recommended in the spec (for reasons unknown), I've isolated its usage
as much as possible. As such the "extraneous" switch only ever will
occur when we have full PPGTT.
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

As with processes which run on the CPU, the goal of multiple VMs is to
provide process isolation. Specific to GEN, there is also the ability to
map more objects per process (2GB each instead of 2Gb-2k total).

For the most part, all the pipes have been laid, and all we need to do
is remove asserts and actually start changing address spaces with the
context switch. Since prior to this we've converted the setting of the
page tables to a streamed version, this is quite easy.

One important thing to point out (since it'd been hotly contested) is
that with this patch, every context created will have it's own address
space (provided the HW can do it).

v2: Disable BDW on rebase

NOTE: I tried to make this commit as small as possible. I needed one
place where I could "turn everything on" and that is here. It could be
split into finer commits, but I didn't really see much point.

Cc: Eric Anholt <eric@anholt.net>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

We have a default context which suits the aliasing PPGTT well. Tie them
together so it looks like any other context/PPGTT pair. This makes the
code cleaner as it won't have to special case aliasing as often.

The patch has one slightly tricky part in the default context creation
function. In the future (and on aliased setup) we create a new VM for a
context (potentially). However, if we have aliasing PPGTT, which occurs
at this point in time for all platforms GEN6+, we can simply manage the
refcounting to allow things to behave as normal. Now is a good time to
recall that the aliasing_ppgtt doesn't have a real VM, it uses the GGTT
drm_mm.
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

In following with the old restore code, we must now restore ever PPGTT's
PDEs, since they aren't proper GEM ojbects.

v2: Rebased on BDW. Only do restore pdes for gen6 & 7
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

We won't be calling enable() for all PPGTTs. We do need to write PDEs
for all PPGTTs however. By moving the writing to init (which is called
for all PPGTTs) we should accomplish this.

ADD NOTE ABOUT PDE restore

TODO: Eventually, we should allocate the page tables on demand.

v2: Rebased on BDW. Only do PDEs for pre-gen8
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Pretty straightforward so far except for the bit about the refcounting.
The PPGTT will potentially be shared amongst multiple contexts. Because
contexts themselves have a refcounted lifecycle, the easiest way to
manage this will be to refcount the PPGTT. To acheive this, we piggy
back off of the existing context refcount, and will increment and
decrement the PPGTT refcount with context creation, and destruction.

To put it more clearly, if context A, and context B both use PPGTT 0, we
can't free the PPGTT until both A, and B are destroyed.

Note that because the PPGTT is permanently pinned (for now), it really
just matters for the PPGTT destruction, as opposed to making space under
memory pressure.
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This patch consolidates the way in which we handle the various supported
PPGTT by module parameter in addition to what the hardware supports. It
strives to make doing the right thing in the code as simple as possible,
with the USES_ macros.

I've opted to add the full PPGTT argument simply so one can see how I
intend to use this function. It will not/cannot be used until later.
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Rearrange the initialization code to try to special case the aliasing
PPGTT less, and provide usable interfaces for the general case later.
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

I've found this by accident. The docs don't really come out and say you
need to do this. What the docs do tell you is you need to flush the TLBs
before you set the PP_DIR_BASE, and that the RCS will invalidate its
TLBs upon setting the new PP_DIR_BASE. It makes no such comment about
any of the other rings.

Empirically, this indeed fixes a really obvious bug whereby the batches
being sent to the blitter were not executing (we were executing the
HSWP somehow instead).

NOTE: This should make no difference with the current code. It only
applies when we start using multiple VMs.

NOTE2: HSW appears to be immune to this.
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The docs seem to suggest this is the appropriate method (though it
doesn't say so outright). In other words, we probably should have done
this before. We certainly must do this for switching VMs on the fly,
since synchronizing the rings to MMIO updates isn't acceptable.

v2:
Make the reset code actually work for all rings. Note that this was
fixed in subsequent commits, but was indeed broken for this commit.

Add a posting read to the reset case. It probably should have existed
before hand, but since we have no failures; there is no reason to make
it a separate commit.

Make IS_GEN6 not use the ring because I am seeing crashes when using it.
It is a bit of a hack in this patch, it will get fixed up in a couple of
patches.
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

In order to do the full context switch with address space, it's
convenient to have a way to switch the address space. We already have
this in our code - just pull it out to be called by the context switch
code later.

v2: Rebased on BDW support. Required adding BDW.
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The patch before this changed the way in which we allocate space for the
PPGTT PDEs. It began carving out the PPGTT PDEs (which live in the
Global GTT) from the GGTT's drm_mm. Prior to that patch, the PDEs were
hidden from the drm_mm, and therefore could never fail to be allocated.

In unfortunate cases, the drm_mm may be full when we want to allocate
the space. This can technically occur whenever we try to allocate, which
happens in two places currently. Practically, it can only really ever
happen at GPU reset.

Later, when we allocate more PDEs for multiple PPGTTs this will
potentially even more useful.
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

When PPGTT support was originally enabled, it was only designed to
support 1 PPGTT. It therefore made sense to simply hide the GGTT space
required to enable this from the drm_mm allocator.

Since we intend to support full PPGTT, which means more than 1, and they
can be created and destroyed ad hoc it will be required to use the
proper allocation techniques we already have.

The first step here is to make the existing single PPGTT use the
allocator.

The astute observer will notice that we are reserving space in the GGTT
for the PDEs for the lifetime of the address space, and would be right
to question whether or not this is a good idea. It does not make a
difference with this current patch only the aliasing PPGTT (indeed the
PDEs should still be hidden from the shrinker). For the future, we are
allocating from top to bottom to avoid using the precious "gtt
space" The GGTT space at that point should only be used for scanout, HW
contexts, ringbuffers, HWSP, PDEs, and a couple of other small buffers
(potentially) used by the kernel. Everything else should be mapped into
a PPGTT. To put the consumption in more tangible terms, it takes
approximately 4 sets of PDEs to equal one 19x10 framebuffer (with no
fancy stride or alignment constraints). 3/4 of the total [average] GGTT
can be used for PDEs, and hopefully never touch the 1/4 that the
framebuffer needs.

The astute, and persistent observer might ask about the page tables
which are also pinned for the address space. This waste is unfortunate.
We use 2MB of memory per address space. We leave wrapping the PDEs as a
real GEM object as a TODO.

v2: Align PDEs to 64b in GTT
Allocate the node dynamically so we can use drm_mm_put_block
Now tested on IGT
Allocate node at the top to avoid fragmentation (Chris)

v3: Use Chris' top down allocator

v4: Embed drm_mm_node into ppgtt struct (Jesse)
Remove hunks which didn't belong (Jesse)

v5: Don't subtract guard page since we now killed the guard page prior
to this patch. (Ben)

v6: Rebased and removed guard page stuff.
Added a chunk to the commit message
Allow adding a context to mappable region

v7: Undo v3, so we can make the drm patch last in the series

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org> (v4)
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>

squash: drm/i915: allow PPGTT to use mappable
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

To sum up what goes on here, we abstract the vma binding, similarly to
the previous object binding. This helps for distinguishing legacy
binding, versus modern binding. To keep the code churn as minimal as
possible, I am leaving in insert_entries(). It serves as the per
platform pte writing basically. bind_vma and insert_entries do share a
lot of similarities, and I did have designs to combine the two, but as
mentioned already... too much churn in an already massive patchset.

What follows are the 3 commits which existed discretely in the original
submissions. Upon rebasing on Broadwell support, it became clear that
separation was not good, and only made for more error prone code. Below
are the 3 commit messages with all their history.

drm/i915: Add bind/unbind object functions to VMA
drm/i915: Use the new vm [un]bind functions
drm/i915: reduce vm->insert_entries() usage

drm/i915: Add bind/unbind object functions to VMA

As we plumb the code with more VM information, it has become more
obvious that the easiest way to deal with bind and unbind is to simply
put the function pointers in the vm, and let those choose the correct
way to handle the page table updates. This change allows many places in
the code to simply be vm->bind, and not have to worry about
distinguishing PPGTT vs GGTT.

Notice that this patch has no impact on functionality. I've decided to
save the actual change until the next patch because I think it's easier
to review that way. I'm happy to squash the two, or let Daniel do it on
merge.

v2:
Make ggtt handle the quirky aliasing ppgtt
Add flags to bind object to support above
Don't ever call bind/unbind directly for PPGTT until we have real, full
PPGTT (use NULLs to assert this)
Make sure we rebind the ggtt if there already is a ggtt binding.  This
happens on set cache levels.
Use VMA for bind/unbind (Daniel, Ben)

v3: Reorganize ggtt_vma_bind to be more concise and easier to read
(Ville). Change logic in unbind to only unbind ggtt when there is a
global mapping, and to remove a redundant check if the aliasing ppgtt
exists.

v4: Make the bind function a bit smarter about the cache levels to avoid
unnecessary multiple remaps. "I accept it is a wart, I think unifying
the pin_vma / bind_vma could be unified later" (Chris)
Removed the git notes, and put version info here. (Daniel)

v5: Update the comment to not suck (Chris)

v6:
Move bind/unbind to the VMA. It makes more sense in the VMA structure
(always has, but I was previously lazy). With this change, it will allow
us to keep a distinct insert_entries.
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>

drm/i915: Use the new vm [un]bind functions

Building on the last patch which created the new function pointers in
the VM for bind/unbind, here we actually put those new function pointers
to use.

Split out as a separate patch to aid in review. I'm fine with squashing
into the previous patch if people request it.

v2: Updated to address the smart ggtt which can do aliasing as needed
Make sure we bind to global gtt when mappable and fenceable. I thought
we could get away without this initialy, but we cannot.

v3: Make the global GTT binding explicitly use the ggtt VM for
bind_vma(). While at it, use the new ggtt_vma helper (Chris)

At this point the original mailing list thread diverges. ie.

v4^:
use target_obj instead of obj for gen6 relocate_entry
vma->bind_vma() can be called safely during pin. So simply do that
instead of the complicated conditionals.
Don't restore PPGTT bound objects on resume path
Bug fix in resume path for globally bound Bos
Properly handle secure dispatch
Rebased on vma bind/unbind conversion
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>

drm/i915: reduce vm->insert_entries() usage

FKA: drm/i915: eliminate vm->insert_entries()

With bind/unbind function pointers in place, we no longer need
insert_entries. We could, and want, to remove clear_range, however it's
not totally easy at this point. Since it's used in a couple of place
still that don't only deal in objects: setup, ppgtt init, and restore
gtt mappings.

v2: Don't actually remove insert_entries, just limit its usage. It will
be useful when we introduce gen8. It will always be called from the vma
bind/unbind.

Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> (v1)
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The initial implementation of this function used MMIO to write the PDPs.
Upon review it was determined (correctly) that the docs say to use LRI.
The issue is there are times where we want to do a synchronous write
(GPU reset).

I've tested this, and it works. I've verified with as many people as
possible that it should work.

This should fix the failing reset problems.
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Our VM code already has a cleanup function, and this is a nice place to
put the drm_mm_takedown. This should have no functional impact, it just
leaves the unload function a bit cleaer, and is more logical IMO
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>