[PATCH RFC 000/111] Etnaviv DRM driver
Hi Lucas, Thanks for the patch series ! It sounds great, even if it probably needs to be squashed down to some patches, as most of the patches are fixes. 2015-04-02 22:01 GMT+02:00 Robert Nelson : > On Thu, Apr 2, 2015 at 10:59 AM, Lucas Stach > wrote: >> Am Donnerstag, den 02.04.2015, 16:43 +0100 schrieb Russell King - ARM >> Linux: >>> On Thu, Apr 02, 2015 at 05:29:02PM +0200, Lucas Stach wrote: >>> > Hey all, >>> > >>> > this is the Etnaviv DRM driver for Vivante embedded GPUs. It is heavily >>> > influenced by the MSM driver, as can be clearly seen with the first >>> > commits. >>> >>> You should be copying Greg KH for staging too. >>> >> I didn't do that on purpose. As stated below in the cover letter I'm not >> really happy to push things into staging. Especially after the >> experience with imx-drm, where it took us a considerable amount of work >> to even get people to look at the code after it had landed in staging. >> >> If possible I would like to collect feedback now and only if someone >> feels genuinely unhappy about this code living under drivers/gpu/drm >> then keep it in staging. Otherwise I would like to move it when removing >> the RFC from this patchset. > > Looks good so far Lucas on my wand quad.. > > Where's your libdrm/mesa tree located that you've been working on? > > debian at arm:~$ uname -r > 4.0.0-rc6-armv7-devel-r24 > debian at arm:~$ dmesg | grep etnaviv > [3.711866] etnaviv gpu-subsystem: bound 134000.gpu (ops gpu_ops) > [3.718015] etnaviv gpu-subsystem: bound 13.gpu (ops gpu_ops) > [3.724133] etnaviv gpu-subsystem: bound 2204000.gpu (ops gpu_ops) > [3.730351] etnaviv-gpu 134000.gpu: model: GC320, revision: 5007 > [3.771045] etnaviv-gpu 13.gpu: model: GC2000, revision: 5108 > [3.802887] etnaviv-gpu 2204000.gpu: model: GC355, revision: 1215 > [3.848287] [drm] Initialized etnaviv 1.0.0 20150302 on minor 1 Exactly the same question, I would like to give it a try on my board, how can I test it ? Thanks, JM
[PATCH RFC 000/111] Etnaviv DRM driver
On Thu, Apr 02, 2015 at 05:59:36PM +0200, Lucas Stach wrote: > Am Donnerstag, den 02.04.2015, 16:43 +0100 schrieb Russell King - ARM > Linux: > > On Thu, Apr 02, 2015 at 05:29:02PM +0200, Lucas Stach wrote: > > > Hey all, > > > > > > this is the Etnaviv DRM driver for Vivante embedded GPUs. It is heavily > > > influenced by the MSM driver, as can be clearly seen with the first > > > commits. > > > > You should be copying Greg KH for staging too. > > > I didn't do that on purpose. As stated below in the cover letter I'm not > really happy to push things into staging. Especially after the > experience with imx-drm, where it took us a considerable amount of work > to even get people to look at the code after it had landed in staging. > > If possible I would like to collect feedback now and only if someone > feels genuinely unhappy about this code living under drivers/gpu/drm > then keep it in staging. Otherwise I would like to move it when removing > the RFC from this patchset. If you want feedback can you please squash down the patch series so that it's clean? I don't like reading drivers and spotting problems just to realize that a few patches down an issue is fixed. -Daniel -- Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch
[PATCH RFC 000/111] Etnaviv DRM driver
2015-04-02 22:01 GMT+02:00 Robert Nelson : > On Thu, Apr 2, 2015 at 10:59 AM, Lucas Stach > wrote: >> Am Donnerstag, den 02.04.2015, 16:43 +0100 schrieb Russell King - ARM >> Linux: >>> On Thu, Apr 02, 2015 at 05:29:02PM +0200, Lucas Stach wrote: >>> > Hey all, >>> > >>> > this is the Etnaviv DRM driver for Vivante embedded GPUs. It is heavily >>> > influenced by the MSM driver, as can be clearly seen with the first >>> > commits. >>> >>> You should be copying Greg KH for staging too. >>> >> I didn't do that on purpose. As stated below in the cover letter I'm not >> really happy to push things into staging. Especially after the >> experience with imx-drm, where it took us a considerable amount of work >> to even get people to look at the code after it had landed in staging. >> >> If possible I would like to collect feedback now and only if someone >> feels genuinely unhappy about this code living under drivers/gpu/drm >> then keep it in staging. Otherwise I would like to move it when removing >> the RFC from this patchset. > > Looks good so far Lucas on my wand quad.. > > Where's your libdrm/mesa tree located that you've been working on? > > debian at arm:~$ uname -r > 4.0.0-rc6-armv7-devel-r24 > debian at arm:~$ dmesg | grep etnaviv > [3.711866] etnaviv gpu-subsystem: bound 134000.gpu (ops gpu_ops) > [3.718015] etnaviv gpu-subsystem: bound 13.gpu (ops gpu_ops) > [3.724133] etnaviv gpu-subsystem: bound 2204000.gpu (ops gpu_ops) > [3.730351] etnaviv-gpu 134000.gpu: model: GC320, revision: 5007 > [3.771045] etnaviv-gpu 13.gpu: model: GC2000, revision: 5108 > [3.802887] etnaviv-gpu 2204000.gpu: model: GC355, revision: 1215 > [3.848287] [drm] Initialized etnaviv 1.0.0 20150302 on minor 1 > > Regards, > > -- Where can we find the userspace (libdrm, mesa, ..)? -- Christian Gmeiner, MSc https://soundcloud.com/christian-gmeiner
[PATCH RFC 000/111] Etnaviv DRM driver
Am Donnerstag, den 02.04.2015, 16:43 +0100 schrieb Russell King - ARM Linux: > On Thu, Apr 02, 2015 at 05:29:02PM +0200, Lucas Stach wrote: > > Hey all, > > > > this is the Etnaviv DRM driver for Vivante embedded GPUs. It is heavily > > influenced by the MSM driver, as can be clearly seen with the first commits. > > You should be copying Greg KH for staging too. > I didn't do that on purpose. As stated below in the cover letter I'm not really happy to push things into staging. Especially after the experience with imx-drm, where it took us a considerable amount of work to even get people to look at the code after it had landed in staging. If possible I would like to collect feedback now and only if someone feels genuinely unhappy about this code living under drivers/gpu/drm then keep it in staging. Otherwise I would like to move it when removing the RFC from this patchset. Regards, Lucas -- Pengutronix e.K. | Lucas Stach | Industrial Linux Solutions | http://www.pengutronix.de/ |
[PATCH RFC 000/111] Etnaviv DRM driver
Hey all, this is the Etnaviv DRM driver for Vivante embedded GPUs. It is heavily influenced by the MSM driver, as can be clearly seen with the first commits. The userspace interface does a look a lot like the MSM one, with some small differences: - Each GPU core is a pipe as with MSM, but Vivante doesn't have a strict separation of tasks between the pipes. On some SoCs like the i.MX6 each pipe feeds one rendering backend (2D, 3D, VG), but there are also SoCs out there where on core (pipe) houses more than one backend. So pipes on Etnaviv represent one core, that may be switched between multiple execution state through the command stream. To allow for proper separation between processes each process may specify the expected execution state on submit. - OR-ing and shifting of BO reloc addresses has been removed, as there is need for this on Vivante GPUs. The register interface is designed in a way that one always fills in complete 32bit addresses without any additional informations. - Presumption of BO addresses is not used right now, as the GPU MMU v1 can not quarantee full protection. There is a 2GB window into physical memory without any MMU translation in between, so we always have to process all relocs to guard against malicious userspace. I've left it in the interface though as MMU v2 seems to be able to give full protection and it might become useful at that point. Unfinished stuff: - GPU PM and context switching. This already works for the GPU 2D where there isn't much state to retain on the GPU itself. For full context switching and power down support on the GPU 3D the userspace needs to aid the kernel with a context restore buffer. This part isn't done yet. It's a rather long series. I already tried to squash some commits together, but wanted to retain the authorship of the individual people that worked on this driver for now. Maybe if everyone involved is okay with it we could squash some of the fixup commits a bit more. I've kept things in staging for now, as that's the place where Christian started this driver, but would really like to move it to DRM proper _before_ merging. So please review stuff with that in mind. Russell King has some experimental support in the xf86-video-armada driver to get some X accel running atop of this. I have a working libdrm/MESA stack that basically works for some simple applications, but needs a good deal more work to clean up. If you would like to look at this stuff as a git tree feel free to fetch: git://git.pengutronix.de/git/lst/linux.git etnaviv-for-upstream Regards, Lucas Christian Gmeiner (2): staging: etnaviv: add drm driver staging: etnaviv: quiten down kernel log output Lucas Stach (28): staging: etnaviv: add devicetree bindings staging: etnaviv: import new headers staging: etnaviv: remove IOMMUv2 stubs staging: etnaviv: allow to draw up to 256 rectangles in one draw call staging: etnaviv: align command stream size to 64 bit staging: etnaviv: correct instruction count for GC2000 and GC880 staging: etnaviv: reconfigure bus mapping on GC2000 staging: etnaviv: fix cache cleaning for uncached SHM buffers staging: etnaviv: properly flush all TLBs on MMUv1 staging: etnaviv: convert to_etnaviv_bo() to real function staging: etnaviv: take gpu instead of pipe as input to fence wait function staging: etnaviv: plug in fence waiting in cpu_prepare staging: etnaviv: allow to map buffer object into multiple address spaces staging: etnaviv: don't pretend to have a single MMU staging: etnaviv: use GPU device to construct MMU staging: etnaviv: flush MMU when switching context staging: etnaviv: add flag to force buffer through MMU staging: etnaviv: use more natural devicetree abstraction staging: etnaviv: don't override platform provided IRQ flags staging: etnaviv: separate GPU pipes from execution state staging: etnaviv: make sure to unlock DRM mutex when component bind fails staging: etnaviv: clean up public API staging: etnaviv: prune dumb buffer support staging: etnaviv: properly prefix all prime functions to etnaviv staging: etnaviv: rename last remaining bits from msm to etnaviv staging: etnaviv: add proper license header to all files staging: etnaviv: some final trivial changes to the module ARM: imx6: add Vivante GPU nodes Philipp Zabel (1): of: Add vendor prefix for Vivante Corporation Russell King (80): staging: etnaviv: fix oops on unbind staging: etnaviv: fix oops in timer subsystem caused by hangcheck timer staging: etnaviv: fix etnaviv_add_components() staging: etnaviv: fix etnaviv_hw_reset() staging: etnaviv: fix etnaviv gpu debugfs output staging: etnaviv: fix fence implementation staging: etnaviv: fix buffer dumping code staging: etnaviv: fix ring buffer overflow check staging: etnaviv: fix cleanup of imported dmabufs staging: etnaviv: fix printk formats staging: etnaviv: validation: ensure space for
[PATCH RFC 000/111] Etnaviv DRM driver
On Thu, Apr 02, 2015 at 05:29:02PM +0200, Lucas Stach wrote: > Hey all, > > this is the Etnaviv DRM driver for Vivante embedded GPUs. It is heavily > influenced by the MSM driver, as can be clearly seen with the first commits. You should be copying Greg KH for staging too. -- FTTC broadband for 0.8mile line: currently at 10.5Mbps down 400kbps up according to speedtest.net.
[PATCH RFC 000/111] Etnaviv DRM driver
On Thu, Apr 2, 2015 at 10:59 AM, Lucas Stach wrote: > Am Donnerstag, den 02.04.2015, 16:43 +0100 schrieb Russell King - ARM > Linux: >> On Thu, Apr 02, 2015 at 05:29:02PM +0200, Lucas Stach wrote: >> > Hey all, >> > >> > this is the Etnaviv DRM driver for Vivante embedded GPUs. It is heavily >> > influenced by the MSM driver, as can be clearly seen with the first >> > commits. >> >> You should be copying Greg KH for staging too. >> > I didn't do that on purpose. As stated below in the cover letter I'm not > really happy to push things into staging. Especially after the > experience with imx-drm, where it took us a considerable amount of work > to even get people to look at the code after it had landed in staging. > > If possible I would like to collect feedback now and only if someone > feels genuinely unhappy about this code living under drivers/gpu/drm > then keep it in staging. Otherwise I would like to move it when removing > the RFC from this patchset. Looks good so far Lucas on my wand quad.. Where's your libdrm/mesa tree located that you've been working on? debian at arm:~$ uname -r 4.0.0-rc6-armv7-devel-r24 debian at arm:~$ dmesg | grep etnaviv [3.711866] etnaviv gpu-subsystem: bound 134000.gpu (ops gpu_ops) [3.718015] etnaviv gpu-subsystem: bound 13.gpu (ops gpu_ops) [3.724133] etnaviv gpu-subsystem: bound 2204000.gpu (ops gpu_ops) [3.730351] etnaviv-gpu 134000.gpu: model: GC320, revision: 5007 [3.771045] etnaviv-gpu 13.gpu: model: GC2000, revision: 5108 [3.802887] etnaviv-gpu 2204000.gpu: model: GC355, revision: 1215 [3.848287] [drm] Initialized etnaviv 1.0.0 20150302 on minor 1 Regards, -- Robert Nelson https://rcn-ee.com/
