- 14 Feb, 2014 15 commits
-
-
Daniel Vetter authored
One side-effect of the introduction of ppgtt was that we needed to rebind the object into the appropriate vm (and global gtt in some peculiar cases). For simplicity this was done twice for every object on every call to execbuffer. However, that adds a tremendous amount of CPU overhead (rewriting all the PTE for all objects into WC memory) per draw. The fix is to push all the decision about which vm to bind into and when down into the low-level bind routines through hints rather than performing the bind unconditionally in the execbuffer routine. Note that this is a regression introduced in the full ppgtt feature branch, before this we've only done re-bound objects when the relevant has_(aliasing_ppgtt|global_gtt)_mapping flag was clear. But since that's per-object and not per-vma that optimization broke. v2: Split out prep work and unrelated changes. v3: Bring back functional change around PIN_GLOBAL that I've accidentally split out. v4: Remove the temporary hack for the old binding logic to avoid bisection issues. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=72906 Tested-by: jianx.zhou@intel.com Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> (v1) Cc: Ben Widawsky <benjamin.widawsky@intel.com> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Acked-by:
Ben Widawsky <ben@bwidawsk.net> Signed-off-by:
Daniel Vetter <daniel.vetter@ffwll.ch>
-
Daniel Vetter authored
This is prep work for reworking the object_pin logic. Atm it still does a (now redundant) lookup of the vma. The next patch will fix this. Split out from Chris vma-bind rework. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Ben Widawsky <benjamin.widawsky@intel.com> Reviewed-by:
Jani Nikula <jani.nikula@intel.com> Reviewed-by:
Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by:
-
Daniel Vetter authored
Split out from Chris vma-bind rework. Jani wondered why this is save, and the reason is that i915_vma_unbind does all these checks, too. So they're redundant. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Ben Widawsky <benjamin.widawsky@intel.com> Cc: Jani Nikula <jani.nikula@intel.com> Reviewed-by:
-
Daniel Vetter authored
There's no need not to, really. Split out from Chris vma-bind rework. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Ben Widawsky <benjamin.widawsky@intel.com> Reviewed-by:
-
Daniel Vetter authored
Only the hardware really access them, so no need to have cpu gtt access available. Split out from Chris vma-bind rework. Note that this is only possible due to the split-up of the mappable pin flag into PIN_GLOBAL and PIN_MAPPABLE. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Ben Widawsky <benjamin.widawsky@intel.com> Reviewed-by:
-
Daniel Vetter authored
Split out from Chris vma-bind rework. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Ben Widawsky <benjamin.widawsky@intel.com> Reviewed-by:
-
Daniel Vetter authored
We access it through the cpu window. No functional difference expected atm since we default to a bottom-up allocation scheme. But that might eventually change so that we prefer the unmappable range for buffers that don't need cpu gtt access. Split out from Chris vma-bind rework. Note that this is only possible due to the split-up of the mappable pin flag into PIN_GLOBAL and PIN_MAPPABLE. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Ben Widawsky <benjamin.widawsky@intel.com> Reviewed-by:
-
Daniel Vetter authored
Tighter code since legacy gem has only mappable anyway. Split out from Chris vma-bind rework. Note that this is only possible due to the split-up of the mappable pin flag into PIN_GLOBAL and PIN_MAPPABLE. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Ben Widawsky <benjamin.widawsky@intel.com> Reviewed-by:
-
Daniel Vetter authored
Split out from Chris vma-bind rework. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Ben Widawsky <benjamin.widawsky@intel.com> Cc: Jani Nikula <jani.nikula@linux.intel.com> Reviewed-by:
-
Daniel Vetter authored
With abitrary pin flags it makes sense to split out a "please bind this into global gtt" from the "please allocate in the mappable range". Use this unconditionally in our global gtt pin helper since this is what its callers want. Later patches will drop PIN_MAPPABLE where it's not strictly needed. Reviewed-by:
-
Daniel Vetter authored
Anything more than just one bool parameter is just a pain to read, symbolic constants are much better. Split out from Chris' vma-binding rework patch. v2: Undo the behaviour change in object_pin that Chris spotted. v3: Split out misplaced hunk to handle set_cache_level errors, spotted by Jani. v4: Keep the current over-zealous binding logic in the execbuffer code working with a quick hack while the overall binding code gets shuffled around. v5: Reorder the PIN_ flags for more natural patch splitup. v6: Pull out the PIN_GLOBAL split-up again. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Ben Widawsky <benjamin.widawsky@intel.com> Reviewed-by:
-
Imre Deak authored
This is the same what we do for DP connectors, so make things more consistent. Signed-off-by:
Imre Deak <imre.deak@intel.com> Reviewed-by:
Antti Koskipää <antti.koskipaa@linux.intel.com> Signed-off-by:
-
Imre Deak authored
Signed-off-by:
-
Imre Deak authored
Atm we set the parent of the dp i2c device to be the correspondig connector device. During driver cleanup we first remove the connector device through intel_modeset_cleanup()->drm_sysfs_connector_remove() and only after that the i2c device through the encoder's destroy callback. This order is not supported by the device core and we'll get a warning, see the below bugzilla ticket. The proper order is to remove first any child device and only then the parent device. The first part of the fix changes the i2c device's parent to be the drm device. Its logical owner is not the connector anyway, but the encoder. Since the encoder doesn't have a device object, the next best choice is the drm device. This is the same what we do in the case of the sdvo i2c device and what the nouveau driver does. The second part creates a symlink in the connector's sysfs directory pointing to the i2c device. This is so, that we keep the current ABI, which also makes sense in case someone wants to look up the i2c device belonging to a specific connector. Reference: http://lists.freedesktop.org/archives/intel-gfx/2014-January/038782.html Reference: http://lists.freedesktop.org/archives/intel-gfx/2014-February/039427.html Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=70523Signed-off-by:
-
Imre Deak authored
Since commit d9255d57 Author: Paulo Zanoni <paulo.r.zanoni@intel.com> Date: Thu Sep 26 20:05:59 2013 -0300 it became clear that we need to separate the unload sequence into two parts: 1. remove all interfaces through which new operations on some object (crtc, encoder, connector) can be started and make sure all pending operations are completed 2. do the actual tear down of the internal representation of the above objects The above commit achieved this separation for connectors by splitting out the sysfs removal part from the connector's destroy callback and doing this removal before calling drm_mode_config_cleanup() which does the actual tear-down of all the drm objects. Since we'll have to customize the interface removal part for different types of connectors in the upcoming patches, add a new unregister callback and move the interface removal part to it. No functional change. Signed-off-by:
-
- 13 Feb, 2014 5 commits
-
-
Paulo Zanoni authored
Reproducible by runtime suspending a Haswell machine with eDP + HDMI outputs connected. [ 209.600086] [drm:i915_runtime_suspend], Suspending device [ 209.688435] BUG: unable to handle kernel NULL pointer dereference at 0000000000000060 [ 209.688500] IP: [<ffffffffa0109d4e>] i915_sink_crc+0x6e/0xf0 [i915] [ 209.688577] PGD 36aba067 PUD 35d7f067 PMD 0 [ 209.688613] Oops: 0000 [#1] SMP [ 209.688641] Modules linked in: fuse ip6table_filter ip6_tables ebtable_nat ebtables iTCO_wdt iTCO_vendor_support x86_pkg_temp_thermal coretemp microcode serio_raw e1000e pcspkr i2c_i801 ptp mei_me mei lpc_ich mfd_core pps_core dm_crypt i915 i2c_algo_bit crc32_pclmul drm_kms_helper crc32c_intel drm ghash_clmulni_intel video [ 209.688893] CPU: 1 PID: 1797 Comm: pm_pc8 Not tainted 3.13.0+ #118 [ 209.688937] Hardware name: Intel Corporation Shark Bay Client platform/WhiteTip Mountain 1, BIOS HSWLPTU1.86C.0133.R00.1309172123 09/17/2013 [ 209.689023] task: ffff88007fb4b690 ti: ffff88007d9d2000 task.ti: ffff88007d9d2000 [ 209.689074] RIP: 0010:[<ffffffffa0109d4e>] [<ffffffffa0109d4e>] i915_sink_crc+0x6e/0xf0 [i915] [ 209.689169] RSP: 0018:ffff88007d9d3e68 EFLAGS: 00010246 [ 209.689205] RAX: 0000000000000000 RBX: ffff880036a03478 RCX: ffff8800366c9770 [ 209.689252] RDX: ffff88014325cf38 RSI: ffff88007fb4bd08 RDI: ffff88007fb4b690 [ 209.689299] RBP: ffff88007d9d3e98 R08: 0000000000000000 R09: 0000000000000000 [ 209.689346] R10: 0000000000000001 R11: 0000000000000000 R12: ffff8800366c9148 [ 209.689393] R13: 00000000ffffffed R14: ffff88007d9d3f50 R15: ffff880036a03478 [ 209.689441] FS: 00007f5a74bc29c0(0000) GS:ffff88014f240000(0000) knlGS:0000000000000000 [ 209.689494] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 209.689533] CR2: 0000000000000060 CR3: 0000000079d7e000 CR4: 00000000001407e0 [ 209.689580] Stack: [ 209.689594] 0000000000001000 ffff880146083980 ffff880146083980 0000000000000000 [ 209.689649] ffff880146083980 0000000000000001 ffff88007d9d3f00 ffffffff811d0744 [ 209.689702] 0000000000000046 00007fff7949fe20 ffff880036a034b8 0000000000000080 [ 209.689756] Call Trace: [ 209.689778] [<ffffffff811d0744>] seq_read+0x164/0x3e0 [ 209.689816] [<ffffffff811ab165>] vfs_read+0x95/0x160 [ 209.689851] [<ffffffff811abc79>] SyS_read+0x49/0xa0 [ 209.689888] [<ffffffff810ef64c>] ? __audit_syscall_entry+0x9c/0xf0 [ 209.689933] [<ffffffff81659412>] system_call_fastpath+0x16/0x1b Testcase: igt/pm_pc8 (do a full run, it will fail at the debugfs-read subtest) Signed-off-by:
Paulo Zanoni <paulo.r.zanoni@intel.com> [danvet: Flip around NULL check for robustness.] Signed-off-by:
-
Damien Lespiau authored
Coverity points out that, if we end up in the 'failed' label, that's precisely because we couldn't retrieve a fixed mode (ie fixed_mode is NULL) and then "if (fixed_mode)" is always false. Remove that dead code. Signed-off-by:
Damien Lespiau <damien.lespiau@intel.com> Reviewed-by:
-
Jesse Barnes authored
It can be corrected later and may be what was actually desired, but generally isn't, so if we find nothing is enabled, let the core DRM fb helper figure something out. v2: free the array too (Jesse) Note that this also undoes any changes in case we bail out due to hw cloning. Signed-off-by:
Jesse Barnes <jbarnes@virtuousgeek.org> Signed-off-by:
-
Ben Widawsky authored
This will make the code more readable, and extensible which is needed for upcoming feature work. Eventually, we'll do the same for init. Signed-off-by:
-
Ville Syrjälä authored
We have a couple of switch cases to compute the port value for the VIDEO_DIP_CTL register. Replace them with a simple macro. We do lose a few BUG() calls, but many people may consider that an improvement. Signed-off-by:
Ville Syrjälä <ville.syrjala@linux.intel.com> Reviewed-by:
-
- 12 Feb, 2014 20 commits
-
-
Daniel Vetter authored
... past the check for DRIVER_MODESET. Avoids races with userspace opening a master and our sarea setup. Cc: Signed-off-by: Stéphane Marchesin <marcheu@chromium.org> Signed-off-by: -
Daniel Vetter authored
Avoids surprises when userspace races open/closes against this. Cc: Stéphane Marchesin <marcheu@chromium.org> Signed-off-by: -
Daniel Vetter authored
We assign the sarea_priv pointer only in the dma ioctl, which is disallowed when kernel modesetting is enabled. So this is dead code. Cc: Stéphane Marchesin <marcheu@chromium.org> Signed-off-by: -
Jesse Barnes authored
The BIOS or boot loader will generally create an initial display configuration for us that includes some set of active pipes and displays. This routine tries to figure out which pipes and connectors are active and stuffs them into the crtcs and modes array given to us by the drm_fb_helper code. The overall sequence is: intel_fbdev_init - from driver load intel_fbdev_init_bios - initialize the intel_fbdev using BIOS data drm_fb_helper_init - build fb helper structs drm_fb_helper_single_add_all_connectors - more fb helper structs intel_fbdev_initial_config - apply the config drm_fb_helper_initial_config - call ->probe then register_framebuffer() drm_setup_crtcs - build crtc config for fbdev intel_fb_initial_config - find active connectors etc drm_fb_helper_single_fb_probe - set up fbdev intelfb_create - re-use or alloc fb, build out fbdev structs v2: use BIOS connector config unconditionally if possible (Daniel) check for crtc cloning and reject (Daniel) fix up comments (Daniel) v3: use command line args and preferred modes first (Ville) Signed-off-by: -
Jesse Barnes authored
This allows drivers to use them in custom initial_config functions. Signed-off-by:
Dave Airlie <airlied@gmail.com> Signed-off-by:
-
Daniel Vetter authored
Just a bit of polish which I hope will help me with massaging some internal patches to use Imre's reworked pipestat handling: - Don't check for underrun reporting or enable pipestat interrupts twice. - Frob the comments a bit. - Do the iir PIPE_EVENT to pipe mapping explicitly with a switch. We only have one place which does this, so better to make it explicit. v2: Ville noticed that I've broken the logic a bit with trying to avoid checking whether we're interested in a given pipe twice. push the PIPESTAT read down after we've computed the mask of interesting bits first to avoid that duplication properly. v3: Squash in fixups from Imre on irc. Cc: Imre Deak <imre.deak@intel.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Reviewed-by:
-
Daniel Vetter authored
We want to reuse this in the fbdev initial config code independently from any fastboot hacks. So allow a bit more flexibility. v2: Forgot to git add ... v3: make non-static (Jesse) Signed-off-by: -
Jesse Barnes authored
We want to do this early on before we try to fetch the plane config, which depends on some of the pipe config state. Note that the important part is that we do this before we initialize gem, since otherwise we can't properly pre-reserve the stolen memory for framebuffers inherited from the bios. v2: split back out from get_plane_config change (Daniel) update for recent locking & reset changes (Jesse) Signed-off-by: -
Imre Deak authored
So far during driver unload we called drm_framebuffer_cleanup() for the fbdev fb, which only removes the fb from the drm fb list regardless of its reference count, but leaves the fb bound on an active crtc. Since the fb's backing storage was freed this could mean we scan some random memory content out afterwards. It's not a big issue since the fb is allocated from stolen memory and afaik there is no other user for that than i915. It's still cleaner to properly unbind the fb and disable the crtc, which is what drm_framebuffer_remove() does. Note that after commit 88891eb1e9eca0ba619518bed31580f91e9cf84d Author: Daniel Vetter <daniel.vetter@ffwll.ch> Date: Mon Feb 10 18:00:38 2014 +0100 we call drm_framebuffer_cleanup() only after dropping the last reference on the fb, but that won't happen since we don't unbind the fb. This results in a drm core warn about a leaked fb. Signed-off-by:
-
Ben Widawsky authored
Signed-off-by:
Deepak S <deepak.s@intel.com> Reviewed-by:
Rodrigo Vivi <rodrigo.vivi@gmail.com> Signed-off-by:
-
Ben Widawsky authored
Everything can be overridden by module parameters, so don't confuse the users that are using them. We have RC6 turned on for all platforms which support it, but Ironlake, so the need to explain the situation is no longer pressing. Signed-off-by:
-
Ben Widawsky authored
It wasn't ever used by the caller anyway with the exception of what we show in sysfs. Signed-off-by:
-
Ben Widawsky authored
At one time, we though all future platforms would have the deeper RC6 states. As it turned out, they killed it after Ivybridge, and began using other means to achieve the power savings (the stuff we need to get to PC7+). The enable function was left in a weird state of odd corner cases as a result. Since the future is now, and we also have some insight into what's currently the future, we have an opportunity to simplify, and future proof the function. NOTE: VLV will be addressed in a subsequent patch. This patch was trying not to change functionality. NOTE2: All callers sanitize the return value anyway, so this patch is simply to have the code make a bit more sense. Signed-off-by:
-
Paulo Zanoni authored
... and QUIRK_PIPEA_FORCE is not present. I initially thought that case was impossible and just added a WARN on it, but then I was told this case is possible due to QUIRK_PIPEA_FORCE. So let's add a WARN that serves two purposes: - tell us in case we have done something wrong; - document the only case where we expect this. Signed-off-by:
-
Paulo Zanoni authored
Add a nice comment explaining why we shouldn't wait for a vblank on all cases, wait based on the HW gen, and add a comment saying we should probably skip that wait on some of the previous HW gens. Signed-off-by:
-
Paulo Zanoni authored
Now that we pass struct intel_crtc as an argument, we can check for DSI inside the function, removing one more of those confusing boolean arguments. Signed-off-by:
-
Paulo Zanoni authored
Now that we pass struct intel_crtc as an argument, there's no need for it. Signed-off-by:
-
Paulo Zanoni authored
We want to remove those 3 boolean arguments. This is the first step. The "pipe" passed as the argument is always intel_crtc->pipe. Also adjust the function documentation. Signed-off-by:
-
Paulo Zanoni authored
When I forked haswell_crtc_enable I copied all the code from ironlake_crtc_enable. The last piece of the function contains a big comment with a call to intel_wait_for_vblank. After this fork, we rearranged the Haswell code so that it enables the planes as the very last step of the modeset sequence, so we're sure that we call intel_enable_primary_plane after the pipe is really running, so the vblank waiting functions work as expected. I really believe this is what fixes the problem described by the big comment, so let's give it a try and get rid of that intel_wait_for_vblank, saving around 16ms per modeset (and init/resume). We can always revert if needed :) Signed-off-by:
-
Paulo Zanoni authored
Because on Haswell, the pipe is never running at this point, so we hit the 50ms timeout waiting for nothing. We already have two other places where we wait for vblanks on haswell_crtc_enable, so we're safe. This gets us rid of one instance of "vblank wait timed out" for each mode set, which means driver init and resume are also 50ms faster. Signed-off-by:
-
