The gpu coredump typically takes place in a dma_fence signalling
critical path, and hence can't use GFP_KERNEL allocations, as that
means we might hit deadlocks under memory pressure. However
changing to __GFP_KSWAPD_RECLAIM which will be done in an upcoming
patch will instead mean a lower chance of the allocation succeeding.
In particular large contigous allocations like the coredump page
vector.
Remove the page vector in favor of a linked list of single pages.
Use the page lru list head as the list link, as the page owner is
allowed to do that.
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Ramalingam C <ramalingam.c@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211108174547.979714-2-thomas.hellstrom@linux.intel.com
I doubt anyone has used the display error state since CS flips
went the way of the dodo. Just nuke it.
It might be semi interesting to have something like this for
FIFO underruns and the like, but as it stands this wouldn't
provide a sufficient amount of information. So would need
an extensive rewrite anyway.
The lockless power well handling is also racy, so this could
just be contributing noise to test results if we end up
accessing something with the relevant power well already
disabled.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210505191140.14215-1-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Between events which trigger engine and GPU resets and capturing the error
state we lose information on which engine triggered the reset. Improve
this by passing in the hung engine mask down to error capture.
Result is that the list of engines in user visible "GPU HANG: ecode
<gen>:<engines>:<ecode>, <process>" is now a list of hanging and not just
active engines. Most importantly the displayed process is now the one
which was actually hung.
v2:
* Stub prototype. (Chris)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20201104134743.916027-1-tvrtko.ursulin@linux.intel.com
Since the engines belong to the GT, move the runtime-updated list of
available engines to the intel_gt struct. The original mask has been
renamed to indicate it contains the maximum engine list that can be
found on a matching device.
In preparation for other info being moved to the gt in follow up patches
(sseu), introduce an intel_gt_info structure to group all gt-related
runtime info.
v2: s/max_engine_mask/platform_engine_mask (tvrtko), fix selftest
Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Andi Shyti <andi.shyti@intel.com>
Cc: Venkata Sandeep Dhanalakota <venkata.s.dhanalakota@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> #v1
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20200708003952.21831-5-daniele.ceraolospurio@intel.com
The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:
struct foo {
int stuff;
struct boo array[];
};
By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.
Also, notice that, dynamic memory allocations won't be affected by
this change:
"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]
sizeof(flexible-array-member) triggers a warning because flexible array
members have incomplete type[1]. There are some instances of code in
which the sizeof operator is being incorrectly/erroneously applied to
zero-length arrays and the result is zero. Such instances may be hiding
some bugs. So, this work (flexible-array member conversions) will also
help to get completely rid of those sorts of issues.
This issue was found with the help of Coccinelle.
[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 7649773293 ("cxgb3/l2t: Fix undefined behaviour")
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20200507185408.GA14561@embeddedor
We only hold the active spinlock while dumping the error state, and this
does not prevent another thread from retiring the request -- as it is
quite possible that despite us capturing the current state, the GPU has
completed the request. As such, it is dangerous to dereference state
below the request as it may already be freed, and the simplest way to
avoid the danger is not include it in the error state.
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/1788
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Andi Shyti <andi.shyti@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200424191410.27570-1-chris@chris-wilson.co.uk
GPU saves accumulated context runtime (in CS timestamp units) in PPHWSP
which will be useful for us in cases when we are not able to track context
busyness ourselves (like with GuC). Keep a copy of this in struct
intel_context from where it can be easily read even if the context is not
pinned.
v2:
(Chris)
* Do not store pphwsp address in intel_context.
* Log CS wrap-around.
* Simplify calculation by relying on integer wraparound.
v3:
* Include total/avg in traces and error state for debugging
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20200216133620.394962-1-chris@chris-wilson.co.uk
Now that we have offline error capture and can reset an engine from
inside an atomic context while also preserving the GPU state for
post-mortem analysis, it is time to handle error interrupts thrown by
the command parser.
This provides a much, much faster mechanism for us to detect known
problems than using heartbeats/hangchecks, and also provides a mechanism
for when those are disabled. However, it is limited to problems the HW
can detect in the CS and so not a complete solution for detecting lockups.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200128204318.4182039-2-chris@chris-wilson.co.uk
i915_gpu_coreddump_put is currently only defined if
CONFIG_DRM_I915_CAPTURE_ERROR is enabled, provide a stub otherwise.
Reported-by: Mike Lothian <mike@fireburn.co.uk>
Fixes: 742379c0c4 ("drm/i915: Start chopping up the GPU error capture")
Fixes: 748317386a ("drm/i915/execlists: Offline error capture")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mike Lothian <mike@fireburn.co.uk>
Cc: Andi Shyti <andi.shyti@intel.com>
Reviewed-by: Andi Shyti <andi.shyti@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200124192255.541355-1-chris@chris-wilson.co.uk
Since commit 742379c0c4 ("drm/i915: Start chopping up the GPU error
capture"), function 'i915_error_state_store' was defined and used with
only one parameter.
But if no 'CONFIG_DRM_I915_CAPTURE_ERROR', this function was defined
with two parameter.
This may lead compile error. This patch fix it.
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Zhang Xiaoxu <zhangxiaoxu5@huawei.com>
Reviewed-by: Andi Shyti <andi.shyti@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20200117073436.6507-1-zhangxiaoxu5@huawei.com
If 'CONFIG_DRM_I915_CAPTURE_ERROR' not configured, there is an error
when compile the kernel:
drivers/gpu/drm/i915/gt/intel_reset.c:
In function intel_gt_handle_error:
drivers/gpu/drm/i915/gt/intel_reset.c:1233:3:
error: too few arguments to function i915_capture_error_state
i915_capture_error_state(gt->i915);
^~~~~~~~~~~~~~~~~~~~~~~~
In file included
from ./drivers/gpu/drm/i915/i915_drv.h:97:0,
from ./drivers/gpu/drm/i915/display/intel_display_types.h:46,
from drivers/gpu/drm/i915/gt/intel_reset.c:10:
./drivers/gpu/drm/i915/i915_gpu_error.h:267:20: note: declared here
static inline void i915_capture_error_state(struct drm_i915_private *dev_priv,
Fixes: 742379c0c4 ("drm/i915: Start chopping up the GPU error capture")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Zhang Xiaoxu <zhangxiaoxu5@huawei.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20200113081942.15982-1-zhangxiaoxu5@huawei.com
If 'CONFIG_DRM_I915_CAPTURE_ERROR' not configured, there are some
errors like:
drivers/gpu/drm/i915/i915_irq.o:
In function `i915_vma_capture_finish':
./drivers/gpu/drm/i915/i915_gpu_error.h:312:
multiple definition of `i915_vma_capture_finish'
drivers/gpu/drm/i915/i915_drv.o:
./drivers/gpu/drm/i915/i915_gpu_error.h:312: first defined here
So, add 'static inline' on the defineation of the 'i915_vma_capture_finish'
Fixes: d713e3ab93fdc("drm/i915: Correct typo in i915_vma_compress_finish stub")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Zhang Xiaoxu <zhangxiaoxu5@huawei.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20200113104009.13274-1-zhangxiaoxu5@huawei.com
The list of requests from after the hang tells little about the hang
itself, only how busy userspace was after the fact. As it pertains
nothing to the HW state, drop it from the error state.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Andi Shyti <andi.shyti@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200110123059.1348712-4-chris@chris-wilson.co.uk
In the near future, we will want to start a GPU error capture from a new
context, from inside the softirq region of a forced preemption. To do
so requires us to break up the monolithic error capture to provide new
entry points with finer control; in particular focusing on one
engine/gt, and being able to compose an error state from little pieces
of HW capture.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Andi Shyti <andi.shyti@intel.com>
Acked-by: Andi Shyti <andi.shyti@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200110123059.1348712-1-chris@chris-wilson.co.uk
TGL introduced a feature in which we map the main surface to the
auxiliary surface. If we screw up the page tables, the HW has a
register to tell us which engine encounters a fault in the page table
walk.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
[ickle: Be brave and apply to gen12]
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20191025121718.18806-1-lionel.g.landwerlin@intel.com
Replace sampling the engine state every so often with a periodic
heartbeat request to measure the health of an engine. This is coupled
with the forced-preemption to allow long running requests to survive so
long as they do not block other users.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Jon Bloomfield <jon.bloomfield@intel.com>
Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20191023133108.21401-5-chris@chris-wilson.co.uk
With the introduction of ctx->engines[] we allow multiple logical
contexts to be used on the same engine (e.g. with virtual engines).
According to bspec, aach logical context requires a unique tag in order
for context-switching to occur correctly between them. [Simple
experiments show that it is not so easy to trick the HW into performing
a lite-restore with matching logical IDs, though my memory from early
Broadwell experiments do suggest that it should be generating
lite-restores.]
We only need to keep a unique tag for the active lifetime of the
context, and for as long as we need to identify that context. The HW
uses the tag to determine if it should use a lite-restore (why not the
LRCA?) and passes the tag back for various status identifies. The only
status we need to track is for OA, so when using perf, we assign the
specific context a unique tag.
v2: Calculate required number of tags to fill ELSP.
Fixes: 976b55f0e1 ("drm/i915: Allow a context to define its set of engines")
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=111895
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20191004134015.13204-14-chris@chris-wilson.co.uk
It might prove useful in the future to know if the vma is utilising
huge-GTT-pages. Related to this is the GTT cache, where there is some HW
"quirkiness" where it must be disabled if using 2M pages, so include
that for good measure.
Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190909171646.22090-1-matthew.auld@intel.com
Skip printing out idle engines that did not contribute to the GPU hang.
As the number of engines gets ever larger, we have increasing noise in
the error state where typically there is only one guilty request on one
engine that we need to inspect.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190808144511.32269-1-chris@chris-wilson.co.uk
Trust that we now have adequate protection over the low level structures
via the engine->active.lock to allow ourselves to capture the GPU error
state without the heavy hammer of stop_machine(). Sadly this does mean
that we have to forgo some of the lesser used information (not derived
from the active state) that is not controlled by the active locks. This
includes the list of buffers in the ppGTT and pinned globally in the
GGTT. Originally this was used to manually verify relocations, but
hasn't been required for sometime and modern mesa now has the habit of
ensuring that all interesting buffers within a batch are captured in their
entirety (that are the auxiliary state buffers, but not the textures).
A useful side-effect is that this allows us to restore error capturing
for Braswell and Broxton.
v2: Use pagevec for a typical arbitrary number of preallocated pages
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190722222847.24178-1-chris@chris-wilson.co.uk
Both microcontrollers are part of the GT HW and are closely related to
GT operations. To keep all the files cleanly together, they've been
placed in their own subdir inside the gt/ folder
Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190713100016.8026-6-chris@chris-wilson.co.uk
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Having taken the first step in encapsulating the functionality by moving
the related files under gt/, the next step is to start encapsulating by
passing around the relevant structs rather than the global
drm_i915_private. In this step, we pass intel_gt to intel_reset.c
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Reviewed-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190712192953.9187-1-chris@chris-wilson.co.uk
In the current scheme, on submitting a request we take a single global
GEM wakeref, which trickles down to wake up all GT power domains. This
is undesirable as we would like to be able to localise our power
management to the available power domains and to remove the global GEM
operations from the heart of the driver. (The intent there is to push
global GEM decisions to the boundary as used by the GEM user interface.)
Now during request construction, each request is responsible via its
logical context to acquire a wakeref on each power domain it intends to
utilize. Currently, each request takes a wakeref on the engine(s) and
the engines themselves take a chipset wakeref. This gives us a
transition on each engine which we can extend if we want to insert more
powermangement control (such as soft rc6). The global GEM operations
that currently require a struct_mutex are reduced to listening to pm
events from the chipset GT wakeref. As we reduce the struct_mutex
requirement, these listeners should evaporate.
Perhaps the biggest immediate change is that this removes the
struct_mutex requirement around GT power management, allowing us greater
flexibility in request construction. Another important knock-on effect,
is that by tracking engine usage, we can insert a switch back to the
kernel context on that engine immediately, avoiding any extra delay or
inserting global synchronisation barriers. This makes tracking when an
engine and its associated contexts are idle much easier -- important for
when we forgo our assumed execution ordering and need idle barriers to
unpin used contexts. In the process, it means we remove a large chunk of
code whose only purpose was to switch back to the kernel context.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190424200717.1686-5-chris@chris-wilson.co.uk
Start partitioning off the code that talks to the hardware (GT) from the
uapi layers and move the device facing code under gt/
One casualty is s/intel_ringbuffer.h/intel_engine.h/ with the plan to
subdivide that header and body further (and split out the submission
code from the ringbuffer and logical context handling). This patch aims
to be simple motion so git can fixup inflight patches with little mess.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Acked-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190424174839.7141-1-chris@chris-wilson.co.uk
We want to use intel_engine_mask_t inside i915_request.h, which means
extracting it from the general header file mess and placing it inside a
types.h. A knock on effect is that the compiler wants to warn about
type-contraction of ALL_ENGINES into intel_engine_maskt_t, so prepare
for the worst.
v2: Use intel_engine_mask_t consistently
v3: Move I915_NUM_ENGINES to its natural home at the end of the enum
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: John Harrison <John.C.Harrison@Intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190401162641.10963-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
We no longer use the semaphore sync registers on gen6/7, so including
them in the GPU error state is mere noise.
References: 6faf5916e6 ("drm/i915: Remove HW semaphores for gen7 inter-engine synchronisation")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190305150914.11340-2-chris@chris-wilson.co.uk
Having weaned the interrupt handling off using a single global execution
queue, we no longer need to emit a global_seqno. Note that we still have
a few assumptions about execution order along engine timelines, but this
removes the most obvious artefact!
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190226094922.31617-3-chris@chris-wilson.co.uk
Currently, we accumulate each time a context hangs the GPU, offset
against the number of requests it submits, and if that score exceeds a
certain threshold, we ban that context from submitting any more requests
(cancelling any work in flight). In contrast, we use a simple timer on
the file, that if we see more than a 9 hangs faster than 60s apart in
total across all of its contexts, we will ban the client from creating
any more contexts. This leads to a confusing situation where the file
may be banned before the context, so lets use a simple timer scheme for
each.
If the context submits 3 hanging requests within a 120s period, declare
it forbidden to ever send more requests.
This has the advantage of not being easy to repair by simply sending
empty requests, but has the disadvantage that if the context is idle
then it is forgiven. However, if the context is idle, it is not
disrupting the system, but a hog can evade the request counting and
cause much more severe disruption to the system.
Updating ban_score from request retirement is dubious as the retirement
is purposely not in sync with request submission (i.e. we try and batch
retirement to reduce overhead and avoid latency on submission), which
leads to surprising situations where we can forgive a hang immediately
due to a backlog of requests from before the hang being retired
afterwards.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190219122215.8941-2-chris@chris-wilson.co.uk
Previously, we were able to rely on the recursive properties of
struct_mutex to allow us to serialise revoking mmaps and reacquiring the
FENCE registers with them being clobbered over a global device reset.
I then proceeded to throw out the baby with the bath water in order to
pursue a struct_mutex-less reset.
Perusing LWN for alternative strategies, the dilemma on how to serialise
access to a global resource on one side was answered by
https://lwn.net/Articles/202847/ -- Sleepable RCU:
1 int readside(void) {
2 int idx;
3 rcu_read_lock();
4 if (nomoresrcu) {
5 rcu_read_unlock();
6 return -EINVAL;
7 }
8 idx = srcu_read_lock(&ss);
9 rcu_read_unlock();
10 /* SRCU read-side critical section. */
11 srcu_read_unlock(&ss, idx);
12 return 0;
13 }
14
15 void cleanup(void)
16 {
17 nomoresrcu = 1;
18 synchronize_rcu();
19 synchronize_srcu(&ss);
20 cleanup_srcu_struct(&ss);
21 }
No more worrying about stop_machine, just an uber-complex mutex,
optimised for reads, with the overhead pushed to the rare reset path.
However, we do run the risk of a deadlock as we allocate underneath the
SRCU read lock, and the allocation may require a GPU reset, causing a
dependency cycle via the in-flight requests. We resolve that by declaring
the driver wedged and cancelling all in-flight rendering.
v2: Use expedited rcu barriers to match our earlier timing
characteristics.
v3: Try to annotate locking contexts for sparse
v4: Reduce selftest lock duration to avoid a reset deadlock with fences
v5: s/srcu/reset_backoff_srcu/
v6: Remove more stale comments
Testcase: igt/gem_mmap_gtt/hang
Fixes: eb8d0f5af4 ("drm/i915: Remove GPU reset dependence on struct_mutex")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190208153708.20023-2-chris@chris-wilson.co.uk
Missed breadcrumb detection is defunct due to the tight coupling with
dma_fence signaling and the myriad ways we may signal fences from
everywhere but from an interrupt, i.e. we frequently signal a fence
before we even see its interrupt. This means that even if we miss an
interrupt for a fence, it still is signaled before our breadcrumb
hangcheck fires, so simplify the breadcrumb hangchecking by moving it
into the GPU hangcheck and forgo fake interrupts.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190129205230.19056-3-chris@chris-wilson.co.uk
A few years ago, see commit 688e6c7258 ("drm/i915: Slaughter the
thundering i915_wait_request herd"), the issue of handling multiple
clients waiting in parallel was brought to our attention. The
requirement was that every client should be woken immediately upon its
request being signaled, without incurring any cpu overhead.
To handle certain fragility of our hw meant that we could not do a
simple check inside the irq handler (some generations required almost
unbounded delays before we could be sure of seqno coherency) and so
request completion checking required delegation.
Before commit 688e6c7258, the solution was simple. Every client
waiting on a request would be woken on every interrupt and each would do
a heavyweight check to see if their request was complete. Commit
688e6c7258 introduced an rbtree so that only the earliest waiter on
the global timeline would woken, and would wake the next and so on.
(Along with various complications to handle requests being reordered
along the global timeline, and also a requirement for kthread to provide
a delegate for fence signaling that had no process context.)
The global rbtree depends on knowing the execution timeline (and global
seqno). Without knowing that order, we must instead check all contexts
queued to the HW to see which may have advanced. We trim that list by
only checking queued contexts that are being waited on, but still we
keep a list of all active contexts and their active signalers that we
inspect from inside the irq handler. By moving the waiters onto the fence
signal list, we can combine the client wakeup with the dma_fence
signaling (a dramatic reduction in complexity, but does require the HW
being coherent, the seqno must be visible from the cpu before the
interrupt is raised - we keep a timer backup just in case).
Having previously fixed all the issues with irq-seqno serialisation (by
inserting delays onto the GPU after each request instead of random delays
on the CPU after each interrupt), we can rely on the seqno state to
perfom direct wakeups from the interrupt handler. This allows us to
preserve our single context switch behaviour of the current routine,
with the only downside that we lose the RT priority sorting of wakeups.
In general, direct wakeup latency of multiple clients is about the same
(about 10% better in most cases) with a reduction in total CPU time spent
in the waiter (about 20-50% depending on gen). Average herd behaviour is
improved, but at the cost of not delegating wakeups on task_prio.
v2: Capture fence signaling state for error state and add comments to
warm even the most cold of hearts.
v3: Check if the request is still active before busywaiting
v4: Reduce the amount of pointer misdirection with list_for_each_safe
and using a local i915_request variable inside the loops
v5: Add a missing pluralisation to a purely informative selftest message.
References: 688e6c7258 ("drm/i915: Slaughter the thundering i915_wait_request herd")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190129205230.19056-2-chris@chris-wilson.co.uk
Now that the submission backends are controlled via their own spinlocks,
with a wave of a magic wand we can lift the struct_mutex requirement
around GPU reset. That is we allow the submission frontend (userspace)
to keep on submitting while we process the GPU reset as we can suspend
the backend independently.
The major change is around the backoff/handoff strategy for performing
the reset. With no mutex deadlock, we no longer have to coordinate with
any waiter, and just perform the reset immediately.
Testcase: igt/gem_mmap_gtt/hang # regresses
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190125132230.22221-3-chris@chris-wilson.co.uk
Make i915_gem_set_wedged() and i915_gem_unset_wedged() behaviour more
consistent if called concurrently, and only do the wedging and reporting
once, curtailing any possible race where we start unwedging in the middle
of a wedge.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190114210408.4561-2-chris@chris-wilson.co.uk
Currently, we convert the error state into a string every time we read
from sysfs (and sysfs reads in page size (4KiB) chunks). We do try to
window the string and only capture the portion that is being read, but
that means that we must always convert up to the window to find the
start. For a very large error state bordering on EXEC_OBJECT_CAPTURE
abuse, this is noticeable as it degrades to O(N^2)!
As we do not have a convenient hook for sysfs open(), and we would like
to keep the lazy conversion into a string, do the conversion of the
whole string on the first read and keep the string until the error state
is freed.
v2: Don't double advance simple_read_from_buffer
v3: Due to extreme pain of lack of vrealloc, use a scatterlist
v4: Keep the forward iterator loosely cached
v5: Stylistic improvements to reduce patch size
Reported-by: Jason Ekstrand <jason@jlekstrand.net>
Testcase: igt/gem_exec_capture/many*
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181123132325.26541-1-chris@chris-wilson.co.uk
Since capturing the error state requires fiddling around with the GGTT
to read arbitrary buffers and is itself run under stop_machine(), it
deadlocks the machine (effectively a hard hang) when run in conjunction
with Broxton's VTd workaround to serialize GGTT access.
v2: Store the ERR_PTR in first_error so that the error can be reported
to the user via sysfs.
v3: Mention the quirk in dmesg (using info as per usual)
Fixes: 0ef34ad622 ("drm/i915: Serialize GTT/Aperture accesses on BXT")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Jon Bloomfield <jon.bloomfield@intel.com>
Cc: John Harrison <john.C.Harrison@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181102161232.17742-5-chris@chris-wilson.co.uk
The final call to zlib_deflate(Z_FINISH) may require more output
space to be allocated and so needs to re-invoked. Failure to do so in
the current code leads to incomplete zlib streams (albeit intact due to
the use of Z_SYNC_FLUSH) resulting in the occasional short object
capture.
v2: Check against overrunning our pre-allocated page array
v3: Drop Z_SYNC_FLUSH entirely
Testcase: igt/i915-error-capture.js
Fixes: 0a97015d45 ("drm/i915: Compress GPU objects in error state")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: <stable@vger.kernel.org> # v4.10+
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181003082422.23214-1-chris@chris-wilson.co.uk
In the next patch, we will want to be able to use more flexible request
timelines that can hop between engines. From the vma pov, we can then
not rely on the binding of this request to an engine and so can not
ensure that different requests are ordered through a per-engine
timeline, and so we must track activity of all timelines. (We track
activity on the vma itself to prevent unbinding from HW before the HW
has finished accessing it.)
v2: Switch to a rbtree for 32b safety (since using u64 as a radixtree
index is fraught with aliasing of unsigned longs).
v3: s/lookup_active/active_instance/ because we can never agree on names
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180706103947.15919-5-chris@chris-wilson.co.uk
Stop reading some now deprecated interrupt registers in both
debugfs and error state. Instead, read the new equivalents in the
Gen11 interrupt repartitioning scheme.
Note that the equivalent to the PM ISR & IIR cannot be read without
affecting the current state of the system, so I've opted for leaving
them out. See gen11_reset_one_iir() for more info.
v2: else if !!! (Paulo)
v3: another else if (Vinay)
v4:
- Rebased
- Renamed patch
- Improved the ordering of GENs
- Improved the printing of per-GEN info
v5: Avoid maybe-unitialized & add comment explaining the lack
of PM ISR & IIR
Suggested-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Cc: Vinay Belgaumkar <vinay.belgaumkar@intel.com>
Reviewed-by: Vinay Belgaumkar <vinay.belgaumkar@intel.com>
[Paulo: fix commit message and coding style.]
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1525989595-18220-1-git-send-email-oscar.mateo@intel.com
Since the advent of execlists, the HW no longer executes from a single
statically assigned ring, but instead switches to a different ring for
each context (logical ringbuffer contexts as it is called). So a good way
to tally the executing context against what we have queued is by
comparing the RING_START register against our requests. Make it so.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180502104150.29874-1-chris@chris-wilson.co.uk
Using plain jiffies in error state output makes the output
time differences relative to the current system time. This
is wrong as it makes output time differences dependent
of when the error state is printed rather than when it is
captured.
Store capture jiffies into error state and use it
when outputting the state to fix time differences output.
v2: use engine timestamp as epoch, output formatting (Chris)
v3: pass epoch to print_engine/request (Chris)
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20180430075259.4476-1-mika.kuoppala@linux.intel.com
