- 11 Nov, 2019 8 commits
-
-
Chris Wilson authored
We report "frequencies" (actual-frequency, requested-frequency) as the number of accumulated cycles so that the average frequency over that period may be determined by the user. This means the units we report to the user are Mcycles (or just M), not MHz. Signed-off-by:
Chris Wilson <chris@chris-wilson.co.uk> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: stable@vger.kernel.org Reviewed-by:
Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191109105356.5273-1-chris@chris-wilson.co.uk
-
Chris Wilson authored
If we detect a hang in a closed context, just flush all of its requests and cancel any remaining execution along the context. Note that after closing the context, the last reference to the context may be dropped, leaving it only valid under RCU. Signed-off-by:
Mika Kuoppala <mika.kuoppala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191111114323.5833-5-chris@chris-wilson.co.uk
-
Chris Wilson authored
We mention that we are resetting the GPU, and dump the device state for post mortem debugging. However, while that dump contains the active processes and the one flagged as causing the error, we do not always include that information in dmesg. Include the name of the guilty process in dmesg for reference. Signed-off-by:
-
Chris Wilson authored
Update the context.name on closing so that the persistent requests are clear in debug prints. Signed-off-by:
-
Chris Wilson authored
Use a small char buffer inside the i915_gem_context to store the user friendly name so that ctx->name has the same lifetime as the RCU protected GEM context. That is, e.g. when using print_request() that prints the timeline name (ctx->name), the name will not be prematurely freed upon the context being closed and the last reference dropped. Signed-off-by:
-
Chris Wilson authored
Inside print_request(), we query the context/timeline name. Nothing immediately protects the context from being freed if the request is complete -- we rely on serialisation by the caller to keep the name valid until they finish using it. Inside intel_engine_dump(), we generally only print the requests in the execution queue protected by the engine->active.lock, but we also show the pending execlists ports which are not protected and so require a rcu_read_lock to keep the pointer valid. [ 1695.700883] BUG: KASAN: use-after-free in i915_fence_get_timeline_name+0x53/0x90 [i915] [ 1695.700981] Read of size 8 at addr ffff8887344f4d50 by task gem_ctx_persist/2968 [ 1695.701068] [ 1695.701156] CPU: 1 PID: 2968 Comm: gem_ctx_persist Tainted: G U 5.4.0-rc6+ #331 [ 1695.701246] Hardware name: Intel Corporation NUC7i5BNK/NUC7i5BNB, BIOS BNKBL357.86A.0052.2017.0918.1346 09/18/2017 [ 1695.701334] Call Trace: [ 1695.701424] dump_stack+0x5b/0x90 [ 1695.701870] ? i915_fence_get_timeline_name+0x53/0x90 [i915] [ 1695.701964] print_address_description.constprop.7+0x36/0x50 [ 1695.702408] ? i915_fence_get_timeline_name+0x53/0x90 [i915] [ 1695.702856] ? i915_fence_get_timeline_name+0x53/0x90 [i915] [ 1695.702947] __kasan_report.cold.10+0x1a/0x3a [ 1695.703390] ? i915_fence_get_timeline_name+0x53/0x90 [i915] [ 1695.703836] i915_fence_get_timeline_name+0x53/0x90 [i915] [ 1695.704241] print_request+0x82/0x2e0 [i915] [ 1695.704638] ? fwtable_read32+0x133/0x360 [i915] [ 1695.705042] ? write_timestamp+0x110/0x110 [i915] [ 1695.705133] ? _raw_spin_lock_irqsave+0x79/0xc0 [ 1695.705221] ? refcount_inc_not_zero_checked+0x91/0x110 [ 1695.705306] ? refcount_dec_and_mutex_lock+0x50/0x50 [ 1695.705709] ? intel_engine_find_active_request+0x202/0x230 [i915] [ 1695.706115] intel_engine_dump+0x2c9/0x900 [i915] Fixes: c36eebd9 ("drm/i915/gt: execlists->active is serialised by the tasklet") Signed-off-by:
-
Chris Wilson authored
After doing some measuring, Icelake behaves on a par with Broadwell, and without having to compromise for low power cores with long latencies, we can reduce the powergating hysteresis so that the powersaving is enabled faster. No impact observed on client side throughput measures (so negligible increase in extra switching), and inspection from high frequency polling using igt/gem_exec_nop/sequential, provided an estimate for the upper bound before we can measure a substantial impact on latency. Signed-off-by:
-
Lionel Landwerlin authored
The ordering of the checks in the existing code can lead to holding preemption not being considered as privileged op. Signed-off-by:
Lionel Landwerlin <lionel.g.landwerlin@intel.com> Fixes: 9cd20ef7 ("drm/i915/perf: allow holding preemption on filtered ctx") Reviewed-by:
-
- 08 Nov, 2019 7 commits
-
-
Ville Syrjälä authored
Make sure we have a crtc before probing its primary plane's max stride. Initially I thought we can't get this far without crtcs, but looks like we can via the dumb_create ioctl. Not sure if we shouldn't disable dumb buffer support entirely when we have no crtcs, but that would require some amount of work as the only thing currently being checked is dev->driver->dumb_create which we'd have to convert to some device specific dynamic thing. Cc: stable@vger.kernel.org Reported-by:
Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191106172349.11987-1-ville.syrjala@linux.intel.comReviewed-by:
-
Chris Wilson authored
On gen7, we have to avoid concurrent access to the same mmio cacheline, and so coordinate all mmio access with the uncore->lock. However, for pmu, we want to avoid perturbing the system and disabling interrupts unnecessarily, so restrict the w/a to gen7 where it is requied to prevent machine hangs. Signed-off-by:
-
Chris Wilson authored
We want to avoid taking forcewake when querying the performance stats, as we wish to avoid perturbing the system under observation. (And with the forcewake being kept alive for 1ms after use, sampling the frequency from a 200Hz timer keeps forcewake 40% active.) Signed-off-by:
-
Chris Wilson authored
In the selftests, where we are accessing a private ctx from within the confines of a single test, we know that the ctx->vm pointer is static and bounded by the lifetime of the test. We can use a simple helper to provide the RCU annotations to keep sparse happy. Signed-off-by:
-
Chris Wilson authored
Since drm provided us with a real struct file we can use for our anonymous internal clients (mock_file), complete our transition to using that as the primary interface (and not the mocked up struct drm_file we previous were using). Signed-off-by:
Matthew Auld <matthew.auld@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191107213929.23286-1-chris@chris-wilson.co.uk
-
Masahiro Yamada authored
The headers in the gem/selftests/, gt/selftests, gvt/, selftests/ directories have never been compile-tested, but it would be possible to make them self-contained. This commit only addresses missing <linux/types.h> and forward struct declarations. Signed-off-by:
Masahiro Yamada <yamada.masahiro@socionext.com> Reviewed-by:
-
Masahiro Yamada authored
Since this function is defined in a header file, it should be 'static inline' instead of 'static'. Signed-off-by:
-
- 07 Nov, 2019 18 commits
-
-
Matt Roper authored
Rather than just specifying the bullet numbers from the bspec (e.g., "4.b") actually include the description of what the bspec wants us to do. Steps can be renumbered or moved so including the description will help us match the code up to the spec. Plus if we add support for new platforms, some of the steps may be added/removed so more descriptive comments will be useful for ensuring all of the bspec requirements are met. Cc: José Roberto de Souza <jose.souza@intel.com> Cc: Lucas De Marchi <lucas.demarchi@intel.com> Signed-off-by:
Matt Roper <matthew.d.roper@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191107174527.11165-1-matthew.d.roper@intel.comReviewed-by:
Lucas De Marchi <lucas.demarchi@intel.com> Reviewed-by:
José Roberto de Souza <jose.souza@intel.com>
-
Chris Wilson authored
Whenever, we unbind (or change fence registers) on an object, we must revoke any and all mmap_gtt using the previous bindings. Those user PTEs point at the GGTT which know points into a new object, the wrong object. Ergo, those PTEs must be cleared so that any user access provokes a new page fault. Signed-off-by:
Abdiel Janulgue <abdiel.janulgue@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191107180601.30815-5-chris@chris-wilson.co.uk
-
Chris Wilson authored
Provide a utility function to create a vma corresponding to an mmap() of our device. And use it to exercise the equivalent of userspace performing a GTT mmap of our objects. Signed-off-by:
-
Chris Wilson authored
As drm now exports a method to create an anonymous struct file around a drm_device for internal use, make use of it to avoid our horrible hacks. Danial suggested that the mock_file_put() wrapper was suitable for drm-core, along with the mock_drm_getfile() [and that the vestigal mock_drm_file() in this patch should perhaps be the drm interface itself]. However, the eventual goal is to remove the mock_drm_file() and use the struct file and fput() directly, in this patch we take a simple transition in that direction. Signed-off-by:
Daniel Vetter <daniel.vetter@ffwll.ch> Link: https://patchwork.freedesktop.org/patch/msgid/20191107180601.30815-3-chris@chris-wilson.co.uk
-
Chris Wilson authored
Sometimes we need to create a struct file to wrap a drm_device, as it the user were to have opened /dev/dri/card0 but to do so anonymously (i.e. for internal use). Provide a utility method to create a struct file with the drm_device->driver.fops, that wrap the drm_device. v2: Restrict usage to selftests Signed-off-by:
-
Chris Wilson authored
Currently, we only export symbols for drm-selftests which are either compiled as modules or into the main drm builtin. However, if we want to export symbols from drm.ko for the drivers' selftests, we require a means of controlling that export separately. So we add a new Kconfig to determine whether or not the EXPORT_SYMBOL_FOR_TESTS_ONLY() takes effect. Signed-off-by:
-
Chris Wilson authored
As we read the ctx->vm unlocked before cloning/exporting, we should validate our reference is correct before returning it. We already do for clone_vm() but were not so strict around get_ppgtt(). Signed-off-by:
Niranjana Vishwanathapura <niranjana.vishwanathapura@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191106091312.12921-1-chris@chris-wilson.co.uk
-
Chris Wilson authored
Only add the engine to the available set of uabi engines once it has been fully initialised and we know we want it in the public set. Signed-off-by:
Andi Shyti <andi.shyti@intel.com> Reviewed-by:
-
Ville Syrjälä authored
The LUTs are single buffered so in order to program them without tearing we'd have to do it during vblank (actually to be 100% effective it has to happen between start of vblank and frame start). We have no proper mechanism for that at the moment so we just defer loading them after the vblank waits have happened. That is not quite sufficient (especially when committing multiple pipes whose vblanks don't line up) so the LUT load will often leak into the following frame causing tearing. However in case the hardware wasn't previously using the LUT we can preload it before setting the enable bit (which is double buffered so won't tear). Let's determine if we can do such preloading and make it happen. Slight variation between the hardware requires some platforms specifics in the checks. Hans is seeing ugly colored flash on VLV/CHV macchines (GPD win and Asus T100HA) when the gamma LUT gets loaded for the first time as the BIOS has left some junk in the LUT memory. v2: Deal with uapi vs. hw crtc state split s/GCM/CGM/ typo fix Cc: Hans de Goede <hdegoede@redhat.com> Fixes: 051a6d8d ("drm/i915: Move LUT programming to happen after vblank waits") Signed-off-by:Hans de Goede <hdegoede@redhat.com> Reviewed-by:
-
Ramalingam C authored
If Local memory is supported by hardware, we want framebuffer backing gem objects from local memory. if the backing obj is not from LMEM, pin_to_display is failed. v2: memory regions are correctly assigned to obj->memory_regions [tvrtko] migration failure is reported as debug log [Tvrtko] v3: Migration is dropped. only error is reported [Daniel] mem region check is move to pin_to_display [Chris] v4: s/dev_priv/i915 [chris] v5: i915_gem_object_is_lmem is used for detecting the obj mem type. [Matt] cc: Matthew Auld <matthew.auld@intel.com> Signed-off-by:
Ramalingam C <ramalingam.c@intel.com> Reviewed-by:
-
Chris Wilson authored
The hidden aliasing-ppgtt's size is never revealed, as we only inspect the front GTT when engaged. However, we were "fixing" the hidden ppgtt to match, with the net result that we ended up leaking the unused portion on Braswell were we preallocated the entire set of top level PDP, see gen8_preallocate_top_level_pdp(). [ 26.025364] DMA-API: pci 0000:00:02.0: device driver has pending DMA allocations while released from device [count=2] [ 26.025364] One of leaked entries details: [device address=0x0000000230778000] [size=4096 bytes] [mapped with DMA_BIDIRECTIONAL] [mapped as single] [ 26.025683] WARNING: CPU: 0 PID: 415 at kernel/dma/debug.c:894 dma_debug_device_change+0x1a4/0x1f0 [ 26.025905] Modules linked in: i915(E-) intel_powerclamp(E) nls_ascii(E) nls_cp437(E) crct10dif_pclmul(E) crc32_pclmul(E) vfat(E) crc32c_intel(E) fat(E) ghash_clmulni_intel(E) prime_numbers(E) intel_gtt(E) i2c_algo_bit(E) efi_pstore(E) drm_kms_helper(E) syscopyarea(E) sysfillrect(E) sysimgblt(E) fb_sys_fops(E) evdev(E) drm(E) aesni_intel(E) glue_helper(E) crypto_simd(E) cryptd(E) intel_cstate(E) sg(E) efivars(E) pcspkr(E) video(E) button(E) efivarfs(E) ip_tables(E) x_tables(E) autofs4(E) sd_mod(E) lpc_ich(E) ahci(E) mfd_core(E) i2c_i801(E) libahci(E) i2c_designware_pci(E) i2c_designware_core(E) [ 26.026613] CPU: 0 PID: 415 Comm: rmmod Tainted: G E 5.4.0-rc6+ #25 [ 26.026837] Hardware name: /, BIOS PYBSWCEL.86A.0027.2015.0507.1758 05/07/2015 [ 26.027080] RIP: 0010:dma_debug_device_change+0x1a4/0x1f0 [ 26.027319] Code: 89 54 24 08 e8 ad 60 62 00 48 8b 54 24 08 48 89 c6 41 57 4d 89 e9 49 89 d8 44 89 f1 41 54 48 c7 c7 e0 61 06 82 e8 c1 aa f5 ff <0f> 0b 5a 59 48 83 3c 24 00 0f 85 97 26 00 00 8b 05 77 47 92 01 85 [ 26.027600] RSP: 0018:ffff888228d2fcc8 EFLAGS: 00010282 [ 26.027831] RAX: 0000000000000000 RBX: 0000000230778000 RCX: 0000000000000000 [ 26.028053] RDX: 0000000000000001 RSI: 0000000000000008 RDI: ffffed10451a5f8f [ 26.028279] RBP: ffff88823480c0b0 R08: 0000000000000001 R09: ffffed1046e83eb1 [ 26.028500] R10: ffffed1046e83eb0 R11: ffff88823741f587 R12: ffffffff82067340 [ 26.028725] R13: 0000000000001000 R14: 0000000000000002 R15: ffffffff82067480 [ 26.028952] FS: 00007fdf3ed174c0(0000) GS:ffff888237400000(0000) knlGS:0000000000000000 [ 26.029185] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 26.029405] CR2: 000055e211109030 CR3: 0000000230139000 CR4: 00000000001006f0 [ 26.029622] Call Trace: [ 26.029846] notifier_call_chain+0x67/0xa0 [ 26.030076] blocking_notifier_call_chain+0x5a/0x80 [ 26.030305] device_release_driver_internal+0x20d/0x260 [ 26.030535] driver_detach+0x7b/0xe1 [ 26.030761] bus_remove_driver+0x8c/0x153 [ 26.030993] pci_unregister_driver+0x2d/0xf0 [ 26.032603] i915_exit+0x16/0x1c [i915] Reported-by:
-
Jani Nikula authored
The intel_dp_link_training.h include has no need or place in intel_display.h. Include it in intel_display.c instead. Cc: Manasi Navare <manasi.d.navare@intel.com> Fixes: eadf6f91 ("drm/i915/display/icl: Enable master-slaves in trans port sync") Reviewed-by:
Manasi Navare <manasi.d.navare@intel.com> Signed-off-by:
Jani Nikula <jani.nikula@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191029103947.7535-1-jani.nikula@intel.com
-
Daniel Vetter authored
So strictly speaking the existing annotation is also ok, because we have a chain of obj->mm.lock#I915_MM_GET_PAGES -> fs_reclaim -> obj->mm.lock (the shrinker cannot get at an object while we're in get_pages, hence this is safe). But it's confusing, so try to take the right subclass of the lock. This does a bit reduce our lockdep based checking, but then it's also less fragile, in case we ever change the nesting around. Signed-off-by:
Daniel Vetter <daniel.vetter@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Will Deacon <will@kernel.org> Cc: linux-kernel@vger.kernel.org Signed-off-by:
Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191104173720.2696-3-daniel.vetter@ffwll.ch
-
Daniel Vetter authored
Necessary to annotate functions where we might acquire a mutex_lock_nested() or similar. Needed by i915. Acked-by:
Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by:
-
Daniel Vetter authored
The trouble with having a plain nesting flag for locks which do not naturally nest (unlike block devices and their partitions, which is the original motivation for nesting levels) is that lockdep will never spot a true deadlock if you screw up. This patch is an attempt at trying better, by highlighting a bit more of the actual nature of the nesting that's going on. Essentially we have two kinds of objects: - objects without pages allocated, which cannot be on any lru and are hence inaccessible to the shrinker. - objects which have pages allocated, which are on an lru, and which the shrinker can decide to throw out. For the former type of object, memory allocations while holding obj->mm.lock are permissible. For the latter they are not. And get/put_pages transitions between the two types of objects. This is still not entirely fool-proof since the rules might change. But as long as we run such a code ever at runtime lockdep should be able to observe the inconsistency and complain (like with any other lockdep class that we've split up in multiple classes). But there are a few clear benefits: - We can drop the nesting flag parameter from __i915_gem_object_put_pages, because that function by definition is never going allocate memory, and calling it on an object which doesn't have its pages allocated would be a bug. - We strictly catch more bugs, since there's not only one place in the entire tree which is annotated with the special class. All the other places that had explicit lockdep nesting annotations we're now going to leave up to lockdep again. - Specifically this catches stuff like calling get_pages from put_pages (which isn't really a good idea, if we can call get_pages so could the shrinker). I've seen patches do exactly that. Of course I fully expect CI will show me for the fool I am with this one here :-) v2: There can only be one (lockdep only has a cache for the first subclass, not for deeper ones, and we don't want to make these locks even slower). Still separate enums for better documentation. Real fix: don't forget about phys objs and pin_map(), and fix the shrinker to have the right annotations ... silly me. v3: Forgot usertptr too ... v4: Improve comment for pages_pin_count, drop the IMPORTANT comment and instead prime lockdep (Chris). v5: Appease checkpatch, no double empty lines (Chris) v6: More rebasing over selftest changes. Also somehow I forgot to push this patch :-/ Also format comments consistently while at it. v7: Fix typo in commit message (Joonas) Also drop the priming, with the lmem merge we now have allocations while holding the lmem lock, which wreaks the generic priming I've done in earlier patches. Should probably be resurrected when lmem is fixed. See commit 232a6eba Author: Matthew Auld <matthew.auld@intel.com> Date: Tue Oct 8 17:01:14 2019 +0100 drm/i915: introduce intel_memory_region I'm keeping the priming patch locally so it wont get lost. Cc: Matthew Auld <matthew.auld@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: "Tang, CQ" <cq.tang@intel.com> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> (v5) Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> (v6) Signed-off-by:
-
Chris Wilson authored
Before we grab the engine wakeref, tidy up the previous heartbeat request. If we then abort because the engine powerwell is off, we ensure the request is freed as we know we will not have freed it when cancelling the work (as the work is running!). Fixes: 841e8672 ("drm/i915/gt: Only drop heartbeat.systole if the sole owner") References: 058179e7 ("drm/i915/gt: Replace hangcheck by heartbeats") Signed-off-by:
-
Lucas De Marchi authored
Prefer using intel_encoder and pass the base where needed rather than keeping both encoder and intel_encoder variables around. v2: actually add all changes to the patch Signed-off-by:
-
James Ausmus authored
Another TGP ID has shown up, so let's add it to avoid South Display breakage on systems that have this ID. Cc: Lucas De Marchi <lucas.demarchi@intel.com> Cc: José Roberto de Souza <jose.souza@intel.com> Signed-off-by:
James Ausmus <james.ausmus@intel.com> Reviewed-by:
-
- 06 Nov, 2019 7 commits
-
-
Niranjana Vishwanathapura authored
We don't need rcu read side critical section to call pid_nr(), remove it. Signed-off-by:
-
Chris Wilson authored
When inside the lock, remember to unlock even if you want to leave early. Reported-by:
Dan Carpenter <dan.carpenter@oracle.com> Fixes: a4e7ccda ("drm/i915: Move context management under GEM") Signed-off-by:
-
Chris Wilson authored
Mika spotted that only using cancel_delayed_work() could mean that we attempted to clear the heartbeat.systole while the worker was still running. Rectify the situation by only touching the systole from outside the worker if we suceeded in cancelling the worker before it could run. The worker is expected to clean up by itself upon idling. Reported-by:
-
Matthew Auld authored
We should not be unconditionally calling release_fake_lmem_bar. Signed-off-by:
-
Ville Syrjälä authored
The uapi vs. hw state split introduced a bug in intel_crtc_disable_noatomic() where it's now frobbing an already freed temp crtc state instead of adjusting the crtc state we are really left with. Fix that by making a cleaner separation beteen the two. This causes explosions on any machine that boots up with pipes already running but not hooked up to any encoder (typical behaviour for gen2-4 VBIOS). Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Fixes: 58d124ea ("drm/i915: Complete crtc hw/uapi split, v6.") Signed-off-by:
-
Chris Wilson authored
If the device does not have an aperture through which we can indirectly access and detile the buffers, simply reject the ioctl. Later we can extend the ioctl to support different modes, but as an extension the user must opt in and explicitly control the mmap type (viz MMAP_OFFSET_IOCTL). Signed-off-by:
Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191105145305.14314-1-chris@chris-wilson.co.uk
-
Tvrtko Ursulin authored
It sounds like the hardware only needs the DSB object to be in global GTT and not in the mappable region. Signed-off-by:
Animesh Manna <animesh.manna@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191017155810.21654-1-tvrtko.ursulin@linux.intel.com
-
