Formerly: "drm/i915: Create VMAs (part 1)"
In a previous patch, the notion of a VM was introduced. A VMA describes
an area of part of the VM address space. A VMA is similar to the concept
in the linux mm. However, instead of representing regular memory, a VMA
is backed by a GEM BO. There may be many VMAs for a given object, one
for each VM the object is to be used in. This may occur through flink,
dma-buf, or a number of other transient states.
Currently the code depends on only 1 VMA per object, for the global GTT
(and aliasing PPGTT). The following patches will address this and make
the rest of the infrastructure more suited
v2: s/i915_obj/i915_gem_obj (Chris)
v3: Only move an object to the now global unbound list if there are no
more VMAs for the object which are bound into a VM (ie. the list is
empty).
v4: killed obj->gtt_space
some reworks due to rebase
v5: Free vma on error path (Imre)
v6: Another missed vma free in i915_gem_object_bind_to_gtt error path
(Imre)
Fixed vma freeing in stolen preallocation (Imre)
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Imre Deak <imre.deak@intel.com>
[danvet: Squash in fixup from Ben to not deref a non-existing vma in
set_cache_level, reported by Chris.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Shamelessly manipulated out of Daniel :-)
"When moving the lists around explain that the active/inactive stuff is
used by eviction when we run out of address space, so needs to be
per-vma and per-address space. Bound/unbound otoh is used by the
shrinker which only cares about the amount of memory used and not one
bit about in which address space this memory is all used in. Of course
to actual kick out an object we need to unbind it from every address
space, but for that we have the per-object list of vmas."
v2: Leave the bound list as a global one. (Chris, indirectly)
v3: Rebased with no i915_gtt_vm. In most places I added a new *vm local,
since it will eventually be replaces by a vm argument.
Put comment back inline, since it no longer makes sense to do otherwise.
v4: Rebased on hangcheck/error state movement
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
After we plumb our code to support multiple address spaces (VMs), there
are a few situations where we want to be able to traverse the list of
all address spaces in the system. Cases like eviction, or error state
collection are obvious example.
v2: Delete the global link instead of the list head. While this in and
of itself shouldn't be really be a problem, doing this allows us to WARN
on an non-empty list, which is a problem. (Daniel)
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Every address space should support object allocation. It therefore makes
sense to have the allocator be part of the "superclass" which GGTT and
PPGTT will derive.
Since our maximum address space size is only 2GB we're not yet able to
avoid doing allocation/eviction; but we'd hope one day this becomes
almost irrelvant.
v2: Rebased
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The GTT and PPGTT can be thought of more generally as GPU address
spaces. Many of their actions (insert entries), state (LRU lists), and
many of their characteristics (size) can be shared. Do that.
The change itself doesn't actually impact most of the VMA/VM rework
coming up, it just fits in with the grand scheme of abstracting the GPU
VM operations. GGTT will usually be a special case where we either know
an object must be in the GGTT (dislay engine, workarounds, etc.).
The scratch page is left as part of the VM (even though it's currently
shared with the ppgtt code) because in the future when we have Full
PPGTT, I intend to create a separate scratch page for each.
v2: Drop usage of i915_gtt_vm (Daniel)
Make cleanup also part of the parent class (Ben)
Modified commit msg
Rebased
v3: Properly share scratch page (Imre)
Finish commit message (Daniel, Imre)
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
intel_enable_rc6() is used to check if we can compute the RC6 residency
in the sysfs code. Disable this for platforms older than Ironlake.
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
At the moment we have the following interrupt enabling sequence:
1. irq preinstall hook
2. enabling the interrupt handler and calling irq postinstall hook
3. enable rps interrupts from the async work
And the folliwing disable sequence:
1. disabling the interrupt handler and calling the uninstall hook
2. disabling the rps interrupt
Since the postinstall hook now always sets up PMIIR, PMIER and PMIMR
to known-good states there no way for an interrupt to sneak in in the
enable sequence, so we can reinstate the WARN lost in
commit eda63ffb90
Author: Ben Widawsky <ben@bwidawsk.net>
Date: Tue May 28 19:22:26 2013 -0700
drm/i915: Add PM regs to pre/post install
Note that there's some room for future cleanups since most of the
interrupt register clearing in the disable function is rather
redundant. But that's better done in follow-up patches, if at all.
Cc: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The VECS enabling required some changes to how rps interrupts are
enabled/disabled since VECS interrupts are handling with the PM
interrupt registers.
But now that the pre/postinstall sequences is identical for all
platforms with rps support (snb, ivb, hsw, vlv) we can also use the
exact same sequence to actually enable the rps interrupts. Strictly
speaking using spinlocks is overkill on snb/ivb & vlv since they have
no VECS ring, but imo that's more than made up by the common code.
Hence this just unifies the vlv code with the snb-hsw code which
matched exactly before the VECS enabling. See
commit eda63ffb90
Author: Ben Widawsky <ben@bwidawsk.net>
Date: Tue May 28 19:22:26 2013 -0700
drm/i915: Add PM regs to pre/post install
and
commit 4848405cce
Author: Ben Widawsky <ben@bwidawsk.net>
Date: Tue May 28 19:22:27 2013 -0700
drm/i915: make PM interrupt writes non-destructive
for why the gen6 code (shared between snb, ivb and hsw) needed to be
changed originally.
v3: Improve the commit message to more clearly spell out why we want
to unify the code and what exactly changes.
Cc: Paulo Zanoni <przanoni@gmail.com>
Cc: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Again extract a common helper. For the postinstall hook things are a
bit more complicated since we have more cases on ilk-hsw/vlv here.
But since vlv was clearly broken by failing to initialize
dev_priv->gt_irq_mask correctly the shared code is clearly justified.
Also kill the PMIER setting in the async rps enable work. I should
have been save, but also clearly looked rather fragile. PMIER setup is
now all down in the irq pre/postinstall hooks.
With this we now have the usual interrupt register sequence for GT/PM
irq registers:
- IER is setup once with all the interrupts we ever need in the
postinstall hook and never touched again. Exceptions are SDEIER,
which is touched in the preinstall hook (when the irq handler isn't
enabled) and then only from the irq handler. And DEIER/VLV_IER with
is used in the irq handler but also written to once in the
postinstall hook. But since that write is essentially what enables
the interrupt and we should always have MSI interrupts we should be
save. In case we ever have non-MSI interrupts we'd be screwed.
- IIR is cleared in the postinstall hook before we enable/unmask the
respective interrupt sources. Hence we can't steal an interrupt
event an accidentally trigger the spurious interrupt logic in the
core kernel. Note that after some discussion with Ben Widawsky we
think that we actually should clear the IIR registers in the
preinstall hook. But doing that is a much larger patch series.
- IMR regs are (usually) all masked off. Those are the only regs
changed at runtime, which is all protected by dev_priv->irq_lock.
This unification also kills the cargo-culted read-modify-write PM
register setup for VECS. Interrupt setup is done without userspace
being able to interfere, so we better know what values we want to put
into those registers. RMW cycles otoh are really good at papering over
races, until stuff magically blows up and no one has a clue why.
v2: Touch the gen6+ PM interrupt registers only on gen6+.
v3: Improve the commit message to more clearly spell out why we want
to unify the code and what exactly changes.
Cc: Ben Widawsky <ben@bwidawsk.net>
Cc: Paulo Zanoni <przanoni@gmail.com>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
[danvet: Add a comment to explain why the l3 parity interrupt is
special.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Since the addition of VECS we have a slightly different enable
sequence for PM interrupts on ivb/hsw vs snb and vlv. Usually that
will end up in hard to track down surprises.
Hence unifiy things and since we have copies of this code in 3 places
now, extract it into its own little helper.
Note that this changes the irq preinstall sequence a bit for snb and
vlv: We now also clear the PM registers in the preinstall hook, in
addition to the PM register clearing/setup already done when actually
enabling rps. So this doesn't fix a bug but simply unifies the code
across all platforms. After the postinstall hook is similarly unified
we can rip out the then redundant PM interrupt setup from the rps
code.
v3: Rebase on top of the retained double-GTIIR clearing. Also
resurrect the masking/disabling of the gen6+ PM interrupts as spotted
by Ben Widaswky.
v4: Move the DE interrupt reset code out of gen5_gt_irq_preinstall
back to ironlake_irq_preinstall where it really belongs. Spotted by
Paulo.
v3: Improve the commit message to more clearly spell out why we want
to unify the code and what exactly changes.
Cc: Paulo Zanoni <przanoni@gmail.com>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
[danvet: s/GT/PM/ to fix up a comment which Ben spotted while
reviewing.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
To make users life a little easier figuring out what they have on their
system.
Ideally, I'd really like to report LLC size, but it turned out to be a
bit of a pain. Maybe I'll revisit it in the future.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
DRI clients really should be using MOCS to get fine grained streaming
cache controls. With that note, I *hope* that this patch doesn't improve
performance overwhelmingly, because if it does - it means there is a
problem elsewhere.
In any case, the kernel, and old userspace should get some benefit from
this, so let's do it. eLLC is always a good default, and really not
using it is the special case for MOCS.
References: http://www.intel.com/newsroom/kits/restricted/ha$well!/pdfs/4th_Gen_Intel_Core_PressBriefing_5-29.pdf (page 57)
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The eLLC cannot be determined by PCIID because as far as we know, even
machines supporting eLLC may not have it enabled, or fused off or
whatever. It's possible this isn't actually true, and at that point we
can switch to a DEV_INFO flag instead.
I've defined everything where the docs are clear, and left the rest as
magic.
But we need it before we set the pte_encode function pointers, which
happens really early, in gtt_init.
The problem with just doing the normal sequence earlier is we don't have
the ability to use forcewake until after the pte functions have been set
up.
Since all solutions are somewhat ugly (barring rewriting all the init
ordering), I've opted to do the detection really early, and the enabling
later - since the register to detect doesn't require forcewake.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
[danvet: Move dev_priv->ellc_size away from the dri1 dungeon to a nice
place right next to the l3 parity stuff. Also squash in the follow-up
commit to read out the eLLC size a bit earlier.]
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The EDRAM present register isn't really defined in the docs. It just
says check to see if it's set to 1. So I haven't defined the 1 value not
knowing what it actually means.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The cacheability controls have changed, and the bits have been
rearranged in general.
Note that age 0 is the oldest (most likely to get evicted) and age 3
is the youngest (most likely to stick around for a bit). We've picked
0 for no reason, but atm it shouldn't matter anyway (since we don't
yet try to differentiate between different objects).
v2: Remove comments for snb/ivb cache leves, that's a separate change.
v3: Resolve conflicts due to patch series reordering.
v4: Rebased on top of Kenneth Graunke's ->pte_encode refactoring.
v5: Removed eLLC bits for separate patch.
In the internal repository this was:
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
[danvet: Add comment about cache ages as requested by Ben provoked due
to a question from Damien.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Toghether with the hw state readout this should catch cases where we
don't properly updated the pll state (either in sw or hw). At least
for the shared dpll code the equivalent tricke helped a lot in
catching bugs.
Also rename the function prefix, it's not a generic piece of
infrastructure.
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Bspec for the "DPLL HDMI multiplier" field says:
"Restriction : The DPLL must be enabled and stable before setting these bits.
These bits must be programmed after DPLL_SEL is programmed."
There is apparently no restriction on programming the DPLL_SEL
register wrt the DPLL. So let's just move that up before we enable the
pch dpll.
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Ben Widawsky
Mika Kuoppala
Chris Wilson
Damien Lespiau
Daniel Vetter
Ben Widawsky