- 29 Apr, 2016 13 commits
-
-
Chris Wilson authored
At the start of request emission, we flush some space for the request, estimating the typical size for the request body. The common tail is now much larger than the typical body, so we can shrink the flush substantially. Signed-off-by:
Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by:
Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1461917226-9132-3-git-send-email-chris@chris-wilson.co.uk
-
Chris Wilson authored
At the start of request emission, we flush some space for the request, estimating the typical size for the request body. The tail is now much larger than the typical body, so we can shrink the flush slightly. Signed-off-by:
-
Chris Wilson authored
With 5 rings and a flush, we need 192 bytes of space to emit the breadcrumb and semaphores. However, we need some spare room the size of the single largest packet (36 dwords, 144 bytes) to accommodate wraparound giving a grand total of 336 bytes Signed-off-by:
-
Ander Conselvan de Oliveira authored
This moves the last phy specific code from the encoders to the phy specific file. Signed-off-by:
Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com> Reviewed-by:
Jim Bride <jim.bride@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1461761065-21195-11-git-send-email-ander.conselvan.de.oliveira@intel.com
-
Ander Conselvan de Oliveira authored
The phy code in vlv_pre_enable_dp() and vlv_hdmi_pre_enable() is exectly the same, so extract it to intel_dpio_phy.c. Signed-off-by:
-
Ander Conselvan de Oliveira authored
The code used by the DP and HDMI paths was very similar, so make them share it. Note that this removes the write to signal level registers from the HDMI pre pll enable path, but that's OK since those are set in vlv_hdmi_pre_enable() function. Signed-off-by:
-
Ander Conselvan de Oliveira authored
The logic for setting signal levels is used for both HDMI and DP with small variations. But it is similar enough to put behind a function called from the encoders. v2: Remove unrelated MST changes due to rebase fumble. (Jim Bride) Fix typo in the commit message. (Jim Bride) v3: Really fix the typo. (Jim) Cc: Jim Bride <jim.bride@linux.intel.com> Signed-off-by: -
Ander Conselvan de Oliveira authored
The exact same code was used by HDMI and DP encoders, so move it to intel_dpio_phy.c. v2: Fix typo in the commit message. (Jim Bride) v3: Call the new function chv_phy_post_pll_disable() instead of chv_phy_post_disable(), as it should be called after the pll is disabled. (Ville) Cc: Jim Bride <jim.bride@linux.intel.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by: -
Ander Conselvan de Oliveira authored
The only difference between the DP and HDMI versions was the lane count. Since lane_count is now set appropriately for HDMI too, get rid of the duplication and move this to intel_dpio_phy.c v2: Don't move comments about 2nd common lane staying alive. (Ville) Signed-off-by:
-
Ander Conselvan de Oliveira authored
The same logic is used for DP and HDMI so move it to intel_dpio_phy.c. v2: Rebase Signed-off-by:
-
Ander Conselvan de Oliveira authored
The function chv_data_lane_soft_reset() was duplicated in DP and HDMI code. Move it to intel_dpio_phy.c. Signed-off-by:
-
Ander Conselvan de Oliveira authored
The code for programming voltage swing and emphasis was duplicated between DP and HDMI code. Move that to a new file, intel_dpio_phy.c. v2: Keep the "Use 800mV-0dB" comment in the HDMI code. (Ville) Signed-off-by:
-
Ander Conselvan de Oliveira authored
Set the lane count for HDMI to 4. This will make it easier to unduplicate CHV phy code. This also fixes the the soft reset programming for HDMI with CHV. After commit a8f327fb ("drm/i915: Clean up CHV lane soft reset programming"), it wouldn't set the right bits for PCS23 since it relied on a lane count that was never set. v2: Set lane_count in *_get_config() to please state checker. (0day) v3: Set lane_count for DDI in DVI mode too. (CI) v4: Add note about CHV soft lane reset. (Ander) Fixes: a8f327fb ("drm/i915: Clean up CHV lane soft reset programming") Signed-off-by:
-
- 28 Apr, 2016 27 commits
-
-
Ville Syrjälä authored
The comment about GMBUSFREQ is confused. The spec actually explains the 4MHz thing perfectly by noting that the 4MHz divider values is actually just bits [9:2] not [9:0], hence the divide by 1000 correct. Replace the confused note with a quote from the spec, and eliminate the duplicated comment that snuck in. Signed-off-by:
Ville Syrjälä <ville.syrjala@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1461689194-6079-4-git-send-email-ville.syrjala@linux.intel.comReviewed-by:
Mika Kahola <mika.kahola@intel.com>
-
Ville Syrjälä authored
No point in reading the cdclk out from the hardware every single time since we have it cached already. Just return the cached value to the audio driver. Signed-off-by:
-
Ville Syrjälä authored
Update CDCLK_FREQ on BDW after changing the cdclk frequency. Not sure if this is a late addition to the spec, or if I simply overlooked this step when writing the original code. This is what Bspec has to say about CDCLK_FREQ: "Program this field to the CD clock frequency minus one. This is used to generate a divided down clock for miscellaneous timers in display." And the "Broadwell Sequences for Changing CD Clock Frequency" section clarifies this further: "For CD clock 337.5 MHz, program 337 decimal. For CD clock 450 MHz, program 449 decimal. For CD clock 540 MHz, program 539 decimal. For CD clock 675 MHz, program 674 decimal." Cc: stable@vger.kernel.org Cc: Mika Kahola <mika.kahola@intel.com> Fixes: b432e5cf ("drm/i915: BDW clock change support") Signed-off-by:
-
Ramalingam C authored
In BXT DSI there is no regs programmed with few horizontal timings in Pixels but txbyteclkhs.. So retrieval process adds some ROUND_UP ERRORS in the process of PIXELS<==>txbyteclkhs. Actually here for the given adjusted_mode, we are calculating the value programmed to the port and then back to the horizontal timing param in pixels. This is the expected value at the end of get_config, including roundup errors. And if that is same as retrieved value from port, then retrieved (HW state) adjusted_mode's horizontal timings are corrected to match with SW state to nullify the errors. Signed-off-by:
Ramalingam C <ramalingam.c@intel.com> Acked-by:
Daniel Vetter <daniel.vetter@ffwll.ch> Acked-by:
Jani Nikula <jani.nikula@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1461053894-5058-2-git-send-email-ramalingam.c@intel.com
-
Ramalingam C authored
Retriving the horizontal timings from the port registers as part of get_config(). This fixes a division by zero: [ 56.916557] divide error: 0000 [#1] PREEMPT SMP [ 56.921741] Modules linked in: i915(+) drm_kms_helper syscopyarea sysfillrect sysimgblt fb_sys_fops drm intel_gtt agpgart cf g80211 rfkill binfmt_misc ax88179_178a kvm_intel kvm irqbypass crc32c_intel efivars tpm_tis tpm fuse [ 56.944106] CPU: 3 PID: 1097 Comm: modprobe Not tainted 4.6.0-rc4+ #433 [ 56.951501] Hardware name: Intel Corp. Broxton M/RVP, BIOS BXT1RVPA.X64.0131.B30.1604142217 04/14/2016 [ 56.961908] task: ffff88007a854d00 ti: ffff88007aea0000 task.ti: ffff88007aea0000 [ 56.970273] RIP: 0010:[<ffffffffa01235b2>] [<ffffffffa01235b2>] drm_mode_hsync+0x22/0x40 [drm] [ 56.980043] RSP: 0018:ffff88007aea3788 EFLAGS: 00010206 [ 56.985982] RAX: 000000000788b600 RBX: ffff880073c22108 RCX: 0000000000000000 [ 56.993957] RDX: 0000000000000000 RSI: ffff88007ab06800 RDI: ffff880073c22108 [ 57.001935] RBP: ffff88007aea3788 R08: 0000000000000001 R09: ffff880073c221e8 [ 57.009903] R10: ffff880073c22108 R11: 0000000000000001 R12: ffff88007a300000 [ 57.017872] R13: ffff880073c22000 R14: ffff880175f78000 R15: ffff880175f78798 [ 57.025849] FS: 00007f105d3e6700(0000) GS:ffff88017fd80000(0000) knlGS:0000000000000000 [ 57.034894] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 57.041317] CR2: 00007f4d485101d0 CR3: 000000007a820000 CR4: 00000000003406e0 [ 57.049292] Stack: [ 57.051539] ffff88007aea37a0 ffffffffa043b632 ffff880175f787c8 ffff88007aea3810 [ 57.059825] ffffffffa043d59e ffff880175f787b0 ffff88007ab68c00 ffff88007aea37f0 [ 57.068128] ffff880073c221e8 ffff880073c22108 ffff880175f78780 ffff880100000000 [ 57.076430] Call Trace: [ 57.079254] [<ffffffffa043b632>] intel_mode_from_pipe_config+0x82/0xb0 [i915] [ 57.087405] [<ffffffffa043d59e>] intel_modeset_setup_hw_state+0x55e/0xd60 [i915] [ 57.095847] [<ffffffffa043ff94>] intel_modeset_init+0x8e4/0x1630 [i915] [ 57.103415] [<ffffffffa047bcf0>] i915_driver_load+0xbe0/0x1980 [i915] [ 57.110745] [<ffffffffa0116c19>] drm_dev_register+0xa9/0xc0 [drm] [ 57.117681] [<ffffffffa011921d>] drm_get_pci_dev+0x8d/0x1e0 [drm] [ 57.124600] [<ffffffff8195f942>] ? _raw_spin_unlock_irqrestore+0x42/0x70 [ 57.132253] [<ffffffffa03b0384>] i915_pci_probe+0x34/0x50 [i915] [ 57.139070] [<ffffffff8149c375>] local_pci_probe+0x45/0xa0 [ 57.145303] [<ffffffff8149d300>] ? pci_match_device+0xe0/0x110 [ 57.151924] [<ffffffff8149d6cb>] pci_device_probe+0xdb/0x130 [ 57.158355] [<ffffffff81579b93>] driver_probe_device+0x223/0x440 [ 57.165169] [<ffffffff81579e85>] __driver_attach+0xd5/0x100 [ 57.171500] [<ffffffff81579db0>] ? driver_probe_device+0x440/0x440 [ 57.178510] [<ffffffff81577736>] bus_for_each_dev+0x66/0xa0 [ 57.184841] [<ffffffff815793de>] driver_attach+0x1e/0x20 [ 57.190881] [<ffffffff81578d6e>] bus_add_driver+0x1ee/0x280 [ 57.197212] [<ffffffff8157abc0>] driver_register+0x60/0xe0 [ 57.203447] [<ffffffff8149bc50>] __pci_register_driver+0x60/0x70 [ 57.210285] [<ffffffffa0119450>] drm_pci_init+0xe0/0x110 [drm] [ 57.216911] [<ffffffff810dcd8d>] ? trace_hardirqs_on+0xd/0x10 [ 57.223434] [<ffffffffa023a000>] ? 0xffffffffa023a000 [ 57.229237] [<ffffffffa023a092>] i915_init+0x92/0x99 [i915] [ 57.235570] [<ffffffff810003db>] do_one_initcall+0xab/0x1d0 [ 57.241900] [<ffffffff810f9eef>] ? rcu_read_lock_sched_held+0x7f/0x90 [ 57.249205] [<ffffffff81204f18>] ? kmem_cache_alloc_trace+0x248/0x2b0 [ 57.256509] [<ffffffff811a5eee>] ? do_init_module+0x27/0x1d9 [ 57.262934] [<ffffffff811a5f26>] do_init_module+0x5f/0x1d9 [ 57.269167] [<ffffffff8112392f>] load_module+0x20ef/0x27b0 [ 57.275401] [<ffffffff8111f8e0>] ? store_uevent+0x40/0x40 [ 57.281541] [<ffffffff81124243>] SYSC_finit_module+0xc3/0xf0 [ 57.287969] [<ffffffff8112428e>] SyS_finit_module+0xe/0x10 [ 57.294203] [<ffffffff81960069>] entry_SYSCALL_64_fastpath+0x1c/0xac [ 57.301406] Code: ff 5d c3 66 0f 1f 44 00 00 0f 1f 44 00 00 8b 87 d8 00 00 00 55 48 89 e5 85 c0 75 22 8b 4f 68 85 c9 78 1b 69 47 58 e8 03 00 00 99 <f7> f9 b9 d3 4d 62 10 05 f4 01 00 00 f7 e1 89 d0 c1 e8 06 5d c3 [ 57.322964] RIP [<ffffffffa01235b2>] drm_mode_hsync+0x22/0x40 [drm] [ 57.330103] RSP <ffff88007aea3788> [ 57.334276] ---[ end trace d414224cb2e2a4cf ]--- [ 57.339861] modprobe (1097) used greatest stack depth: 12048 bytes left Fixes: 6f0e7535 ("drm/i915/BXT: Get pipe conf from the port registers") Signed-off-by:
Imre Deak <imre.deak@intel.com> Signed-off-by:
-
Chris Wilson authored
Both execlists and legacy need to reset the context (and mode) of the GPU before we lose control of the system. By resetting the GPU, we revert back to default settings. This simplifies the life of any subsequent driver (in particular for virtualized setups) as it does not then have to try and recover from an unknown condition. As both paths need to reset for the same reason, move the reset to a common point. This unifies the resets added in a647828a (drm/i915: Also perform gpu reset under execlist mode) and 8e96d9c4 (drm/i915: reset the GPU on context fini). v2: Restrict the reset to "modern" gen (where we enable HW contexts) to try and avoid leaving the machine in an unusable state with a risky reset on older GPU. This should keep the status quo as to who performs resets (i.e. currently only GPUs with HW contexts perform a reset on shutdown). Signed-off-by:
Mika Kuoppala <mika.kuoppala@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1461833819-3991-25-git-send-email-chris@chris-wilson.co.uk
-
Tvrtko Ursulin authored
With the previous patch having extended the pinned lifetime of contexts by referencing the previous context from the current request until the latter is retired (completed by the GPU), we can now remove usage of execlist retired queue entirely. This is because the above now guarantees that all execlist object access requirements are satisfied by this new tracking, and we can stop taking additional references and stop keeping request on the execlists retired queue. The latter was a source of significant scalability issues in the driver causing performance hits on some tests. Most dramatical of which was igt/gem_close_race which had run time in tens of minutes which is now reduced to tens of seconds. Signed-off-by:
Tvrtko Ursulin <tvrtko@ursulin.net> Cc: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by:
-
Tvrtko Ursulin authored
This way in the following patch we can disconnect requests from contexts. Signed-off-by:
Tvrtko Ursulin <tvrtko.ursulin@intel.com> Reviewed-by:
-
Chris Wilson authored
As the contexts are accessed by the hardware until the switch is completed to a new context, the hardware may still be writing to the context object after the breadcrumb is visible. We must not unpin/unbind/prune that object whilst still active and so we keep the previous context pinned until the following request. We can generalise the tracking we already do via the engine->last_context and move it to the request so that it works equally for execlists and GuC. v2: Drop the execlists double pin as that exposes a race inside the lrc irq handler as it tries to access the context after it may be retired. Signed-off-by:
-
Chris Wilson authored
If we move the release of the GEM request (i.e. decoupling it from the various lists used for client and context tracking) after it is complete (either by the GPU retiring the request, or by the caller cancelling the request), we can remove the requirement that the final unreference of the GEM request need to be under the struct_mutex. The careful reader may notice that one or two impossible NULL pointer tests are dropped for readability. These pointers cannot be NULL since they are assigned during request construction and never unset. v2,v3: Rebalance execlists by moving the context unpinning. v4: Rebase onto -nightly v5: Avoid trying to rebalance execlist/GuC context pinning, leave that to the next step Signed-off-by:
-
Chris Wilson authored
We can hide more details of execlists from higher level code by removing the explicit call to create an execlist context from execbuffer and into its first use by execlists. Signed-off-by:
-
Chris Wilson authored
Refactor pinning and unpinning of contexts, such that the default context for an engine is pinned during initialisation and unpinned during teardown (pinning of the context handles the reference counting). Thus we can eliminate the special case handling of the default context that was required to mask that it was not being pinned normally. v2: Rebalance context_queue after rebasing. v3: Rebase to -nightly (not 40 patches in) v4: Rebase onto request_alloc unwinding Signed-off-by:
Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> Reviewed-by:
-
Chris Wilson authored
Rather than reuse the current location of the context in the global GTT for its hardware identifier, use the context's unique ID assigned to it for its whole lifetime. Signed-off-by:
-
Chris Wilson authored
The hardware tracks contexts and expects all live contexts (those active on the hardware) to have a unique identifier. This is used by the hardware to assign pagefaults and the like to a particular context. v2: Reorder to make sure ctx->link is not left dangling if the assignment of a hw_id fails (Mika). v3: We have 21bits of context space, not 20. Signed-off-by:
-
Chris Wilson authored
The comments describing the Context Descriptor Format are off by a bit for the size of the context ID. Signed-off-by:
-
Chris Wilson authored
Rather than being interrupted when we run out of space halfway through the request, and having to restart from the beginning (and returning to userspace), flush a little more free space when we prepare the request. Signed-off-by:
-
Chris Wilson authored
In the next patches, we want to move the work out of freeing the request and into its retirement (so that we can free the request without requiring the struct_mutex). This means that we cannot rely on unreferencing the request to completely teardown the request any more and so we need to manually unwind the failed allocation. In doing so, we reorder the allocation in order to make the unwind simple (and ensure that we don't try to unwind a partial request that may have modified global state) and so we end up pushing the initial preallocation down into the engine request initialisation functions where we have the requisite control over the state of the request. Moving the initial preallocation into the engine is less than ideal: it moves logic to handle a specific problem with request handling out of the common code. On the other hand, it does allow those backends significantly more flexibility in performing its allocations. Signed-off-by:
-
Chris Wilson authored
Now that we share intel_ring_begin(), reserving space for the tail of the request is identical between legacy/execlists and so the tautology can be removed. In the process, we move the reserved space tracking from the ringbuffer on to the request. This is to enable us to reorder the reserved space allocation in the next patch. Signed-off-by:
-
Chris Wilson authored
Combine the near identical implementations of intel_logical_ring_begin() and intel_ring_begin() - the only difference is that the logical wait has to check for a matching ring (which is assumed by legacy). In the process some debug messages are culled as there were following a WARN if we hit an actual error. v2: Updated commentary Signed-off-by:
-
Chris Wilson authored
The code to switch_mm() is already handled by i915_switch_context(), the only difference required to setup the aliasing ppgtt is that we need to emit te switch_mm() on the first context, i.e. when transitioning from engine->last_context == NULL. This allows us to defer the initialisation of the GPU from early device initialisation to first use, which should marginally speed up both. The caveat is that we then defer the context initialisation until first use - i.e. we cannot assume that the GPU engines are initialised. For example, this means that power contexts for rc6 (Ironlake) need to explicitly loaded, as they are. Signed-off-by:
-
Chris Wilson authored
Since we do the l3-remap on context switch, we can remove the redundant early call to set the mapping prior to performing the first context switch. Signed-off-by:
-
Chris Wilson authored
We can use a single MI_LOAD_REGISTER_IMM command packet to write all the L3 remapping registers, shrinking the number of bytes required to emit the context switch. Signed-off-by:
-
Chris Wilson authored
Move the i915_gem_l3_remap function such that it next to the context switching, which is where we perform the L3 remap. Signed-off-by:
-
Chris Wilson authored
In order to force a reload of the context image upon resume, we first need to mark its absence on suspend. Currently we are failing to restore the golden context state and any context w/a to the default context after resume. One oversight corrected, is that we had forgotten to reapply the L3 remapping when restoring the lost default context. v2: Remove deprecated WARN. Signed-off-by:
Mika Kuoppala <mika.kuoppala@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1461833819-3991-7-git-send-email-chris@chris-wilson.co.uk
-
Chris Wilson authored
Similarly to i915_gem_object_pin_map on LLC platforms, we can use the new VMA based io mapping on !LLC to amoritize the cost of ringbuffer pinning and unpinning. Signed-off-by:
-
Chris Wilson authored
By tracking the iomapping on the VMA itself, we can share that area between multiple users. Also by only revoking the iomapping upon unbinding from the mappable portion of the GGTT, we can keep that iomap across multiple invocations (e.g. execlists context pinning). Note that by moving the iounnmap tracking to the VMA, we actually end up fixing a leak of the iomapping in intel_fbdev. v1.5: Rebase prompted by Tvrtko v2: Drop dev_priv parameter, we can recover the i915_ggtt from the vma. v3: Move handling of ioremap space exhaustion to vmap_purge and also allow vmallocs to recover old iomaps. Add Tvrtko's kerneldoc. v4: Fix a use-after-free in shrinker and rearrange i915_vma_iomap v5: Back to i915_vm_to_ggtt v6: Use i915_vma_pin_iomap and i915_vma_unpin_iomap to mark critical sections and ensure the VMA cannot be reaped whilst mapped. v7: Move i915_vma_iounmap so that consumers of the API are not tempted, and add iomem annotations Signed-off-by:
-
Chris Wilson authored
In a couple of places, we have an i915_address_space that we know is really an i915_ggtt that we want to use. Create an inline helper to convert from the i915_address_space subclass into its container. Signed-off-by:
-
