Re: PowerVR, 3.3 kernel and omap_drm
That's the one I'm using. And it seems to be working fine. The problem I encountered was with the libdri2 package, where I had overlooked one of the patches which obviously was needed. - Martin On 05/04/12 18:21, Boudet, Xavier wrote: Hi Can you let me know which version of libdrm-omap1 package you are using? The libdrm - 2.4.32-1ubuntu1+ti2.0 available into the trunk PPA shall have all the latest fixes. Regards, /*Xavier Boudet -*/ Texas Instruments France Linux SW Integration - Platform Enablement Organization Office: +33 4 93 22 */26 83/* Texas Instruments France SA, 821 Avenue Jack Kilby, 06270 Villeneuve Loubet. 036 420 040 R.C.S Antibes. Capital de EUR 753.920 On Fri, May 4, 2012 at 5:25 PM, Dechesne, Nicolas n-deche...@ti.com mailto:n-deche...@ti.com wrote: On Thu, May 3, 2012 at 5:34 PM, Christian Robottom Reis k...@linaro.org mailto:k...@linaro.org wrote: any chance you can use current upstream libdrm (http://cgit.freedesktop.org/mesa/drm)? there was a fix for this issue.. I guess what you are using has an earlier version of libdrm_omap patches compared to what is upstream.. Wouldn't it make sense to update the version in ~tiomap-dev, then? yes, it does... we just don't update the PPA for every single commit ;-) ___ linaro-dev mailing list linaro-dev@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-dev ___ linaro-dev mailing list linaro-dev@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-dev
PowerVR, 3.3 kernel and omap_drm
Hi. I have gotten a lot of help from you guys getting the PowerVR drivers up and running with the 3.3 kernel on the Pandaboard ES. Problem now is that all I tested then, was that X was running. After some more work, we tried out EGL, and found out that this is not working. The 3.3 kernel I'm using is revision f8e851d03e884b60b5207f59a342da9cafdb415f from tilt-3.3 from the ti-landing team. I'm using the PowerVR drivers in http://ppa.launchpad.net/tiomap-dev/omap-trunk/ubuntu/pool/main/, that is pvr-omap4_1.7.15.0.1.57 and pvr-omap4-dkms_1.7.15.0.1.54. I'm also using the libdrm, libdri2 and xorg server from the same place, using the versions from: https://launchpad.net/~tiomap-dev/+archive/omap-trunk?field.series_filter=precise Problem is, when trying to run something using EGL, I get the output from: http://pastebin.com/e5deVmiD Running it through gdb on the device, I see that it segfaults in omap_bo_handle in /usr/lib/libdrm_omap.so.1 This file is compiled, using the patches listed in debian/patches/series in the latest commit from git://gitorious.org/ubuntu-omap/libdrm.git, and with configure arguments: --enable-omap-experimental-api --enable-udev --enable-libkms --disable-radeon Any ideas on what might be wrong or what I can try? I have tried copying the libdrm_omap binary from the ppa, as it looks like that might be at fault, but no luck. Best Regards Martin Ertsas ___ linaro-dev mailing list linaro-dev@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-dev
Possible issues with going from OMAP4 to OMAP5 with Linux
Hi. We are currently working towards deploying on a OMAP4460 SoC, and are using the kernel from the TI landing team in tilt-3.3, but would be willing to go to 4.4 if necessary. We are currently discussing using the OMAP5432 instead of the OMAP4460, and are wondering which issues we might encounter with doing this? Do you expect every module on the 5432 to be working well when the chip is released, or in about a year from now? I got some documentation, which basically told us to design for 4460 now, and when 5432 comes out, most of the work was already done. This was for Android though, and I'm wondering if you expect this to be the case also on Linux? Best regards Martin Ertssas ___ linaro-dev mailing list linaro-dev@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-dev
Re: 3.3 kernel from TI landing teams and powervr problems.
Thanks to this, I got x running, problem is that I don't seem to have direct rendering. First it complained that glx could not be loaded. When adding the glx extensions from the xserver, I get some problems with AIGLX. It sais that /usr/lib/dri/omap_dri.so could not be found. Is this something I should have with these three packages, and I have not compiled everything I should, or is it a fourth package which I haven't installed? Searched through the xf86-video-omap, and doesn't seem to build anything but omap_drv.so, but it do contain a file omap_dri2.c. Should I let the glx warning appear, and just ignore it maybe? Do the driver provide dri2 by itself? Thank you so much for your help so far. - Martin On 04/17/12 01:17, Clark, Rob wrote: On Mon, Apr 16, 2012 at 10:07 AM, Martin Ertsås marti...@gmail.com wrote: True, I renamed that file, since I got the message that it couldn't find module omap. First I renamed it to omap_drv.so, which gave me the same error, then I tried pvr_drm.so which is the attatched logfile. Both of them are the omap_pvr_drv.so file. yeah, you probably shouldn't really rename the xorg modules.. the xorg module loader will look for symbols within the driver matching the name. I expect omap_drv.so is in a different package (maybe missing dependency?).. it is opensrc, anyways, you can rebuild it from the git tree at freedesktop.org http://cgit.freedesktop.org/xorg/driver/xf86-video-omap/ git://anongit.freedesktop.org/xorg/driver/xf86-video-omap BR, -R Should there be a omap_drv.so file in the same directory as the omap_pvr_drv.so file? In that case there isn't, the omap_pvr_drv.so file is the only one in that directory when extracting that tarball. Actually there is no file at all named omap_drv.so: $ find . -name omap_drv.so return nothing, while just omap gives $ find . -name omap* ./usr/lib/debug/usr/lib/xorg/modules/drivers/omap_pvr_drv.so ./usr/lib/xorg/modules/drivers/omap_pvr_drv.so which is the same as *drv* gives. - Martin On 04/16/12 16:42, Clark, Rob wrote: There is an omap_drv.so which should be the xorg driver to load.. it will attempt to load (if present) omap_pvr_drv.so as a submodule for EXA accel. I'm not entirely sure why it is named pvr_drv.so in your filesystem, but from list of symbols it seems to match what I would expect to be omap_pvr_drv.so BR, -R On Mon, Apr 16, 2012 at 8:45 AM, Martin Ertsåsmarti...@gmail.com wrote: Hi. I am not 100% sure this is the correct place to ask, but since I know at least Ricardo Salveti are working on these drivers for omap4, and I haven't heard anything from the ppa's mailing list, I try to ask here. If this was completely wrong I'm sorry, but would still appreciate a hint as to where to ask and who to contact. We are compiling what is now HEAD in the tilt-3.3 branch from the TI landing team. To compile the module needed for the powervr drivers, we had to get the pvr-omap4-dkms-1.7.15 and pvr-omap4-1.7.15 from http://ppa.launchpad.net/tiomap-dev/omap-trunk/ubuntu/pool/main/p/ We downloaded the tar.gz from those, extracted them and compiled them. Problem starts when we boot the board and try to run startx. The xorg config which comes with the later package is loaded, but when trying to start X I get the error: (EE) Failed to load /usr/lib/xorg/modules/drivers/pvr_drv.so: /usr/lib/xorg/modules/drivers/pvr_drv.so: undefined symbol: OMAPFinishAccess None of the libraries in pvr-omap4 contains this symbol, so I'm wondering if I'm doing something wrong, or if there is something wrong with the binary? When comparing to the library from rsalveti, from people/rsalveti/pvr-omap4.git, I see that the dynsym table contains 232 entries, but there are only 64 entries in the same table in the other library. Is this correct, or should there be more entries in the 1.7.15 library? Both the Xorg log and the readelf output from both libraries are attached, if you have questions/need more info, please do not hesitate to ask. Thank you in advance for the help, and thank you for the great work you put in for making embedded Linux better. Best regards Martin Ertsaas ___ linaro-dev mailing list linaro-dev@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-dev [1792017.897] X.Org X Server 1.11.2 Release Date: 2011-11-04 [1792017.897] X Protocol Version 11, Revision 0 [1792017.898] Build Operating System: Linux 3.2.12-gentoo x86_64 [1792017.898] Current Operating System: Linux nimbus-panda 3.3.0-wl-gf852b48-dirty #1 SMP PREEMPT Mon Apr 16 11:26:47 CEST 2012 armv7l [1792017.898] Kernel command line: console=ttyO2,115200n8 vram=16M root=/dev/mmcblk0p2 rw rootfstype=ext3 rootwait [1792017.898] Build Date: 16 April 2012 11:51:58AM [1792017.898] [1792017.898] Current version of pixman: 0.25.2 [1792017.898] Before reporting problems, check http://wiki.x.org to make sure that you have
Re: 3.3 kernel from TI landing teams and powervr problems.
On 04/17/12 10:50, Dechesne, Nicolas wrote: On Tue, Apr 17, 2012 at 9:52 AM, Martin Ertsås marti...@gmail.com mailto:marti...@gmail.com wrote: Searched through the xf86-video-omap, and doesn't seem to build anything but omap_drv.so, but it do contain a file omap_dri2.c. Should I let the glx warning appear, and just ignore it maybe? Do the driver provide dri2 by itself? Thank you so much for your help so far. Martin, if you look at the TI OMAP PPA on Launchpad, we have 2 main PPAs: release and trunk. 'trunk' PPA for precise has *all* the packages needed for X with direct rendering, see https://launchpad.net/~tiomap-dev/+archive/omap-trunk?field.series_filter=precise https://launchpad.net/%7Etiomap-dev/+archive/omap-trunk?field.series_filter=precise and that includes the kernel. to make sure you have a working system, I would suggest that you start with 12.04 (desktop or core), add the 'trunk' PPA , install our kernel from PPA, reboot and install ubuntu-omap4-extras-graphics. at that point you should have X with acceleration. Thanks. I'll use this then. ___ linaro-dev mailing list linaro-dev@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-dev
Re: 3.3 kernel from TI landing teams and powervr problems.
True, I renamed that file, since I got the message that it couldn't find module omap. First I renamed it to omap_drv.so, which gave me the same error, then I tried pvr_drm.so which is the attatched logfile. Both of them are the omap_pvr_drv.so file. Should there be a omap_drv.so file in the same directory as the omap_pvr_drv.so file? In that case there isn't, the omap_pvr_drv.so file is the only one in that directory when extracting that tarball. Actually there is no file at all named omap_drv.so: $ find . -name omap_drv.so return nothing, while just omap gives $ find . -name omap* ./usr/lib/debug/usr/lib/xorg/modules/drivers/omap_pvr_drv.so ./usr/lib/xorg/modules/drivers/omap_pvr_drv.so which is the same as *drv* gives. - Martin On 04/16/12 16:42, Clark, Rob wrote: There is an omap_drv.so which should be the xorg driver to load.. it will attempt to load (if present) omap_pvr_drv.so as a submodule for EXA accel. I'm not entirely sure why it is named pvr_drv.so in your filesystem, but from list of symbols it seems to match what I would expect to be omap_pvr_drv.so BR, -R On Mon, Apr 16, 2012 at 8:45 AM, Martin Ertsåsmarti...@gmail.com wrote: Hi. I am not 100% sure this is the correct place to ask, but since I know at least Ricardo Salveti are working on these drivers for omap4, and I haven't heard anything from the ppa's mailing list, I try to ask here. If this was completely wrong I'm sorry, but would still appreciate a hint as to where to ask and who to contact. We are compiling what is now HEAD in the tilt-3.3 branch from the TI landing team. To compile the module needed for the powervr drivers, we had to get the pvr-omap4-dkms-1.7.15 and pvr-omap4-1.7.15 from http://ppa.launchpad.net/tiomap-dev/omap-trunk/ubuntu/pool/main/p/ We downloaded the tar.gz from those, extracted them and compiled them. Problem starts when we boot the board and try to run startx. The xorg config which comes with the later package is loaded, but when trying to start X I get the error: (EE) Failed to load /usr/lib/xorg/modules/drivers/pvr_drv.so: /usr/lib/xorg/modules/drivers/pvr_drv.so: undefined symbol: OMAPFinishAccess None of the libraries in pvr-omap4 contains this symbol, so I'm wondering if I'm doing something wrong, or if there is something wrong with the binary? When comparing to the library from rsalveti, from people/rsalveti/pvr-omap4.git, I see that the dynsym table contains 232 entries, but there are only 64 entries in the same table in the other library. Is this correct, or should there be more entries in the 1.7.15 library? Both the Xorg log and the readelf output from both libraries are attached, if you have questions/need more info, please do not hesitate to ask. Thank you in advance for the help, and thank you for the great work you put in for making embedded Linux better. Best regards Martin Ertsaas ___ linaro-dev mailing list linaro-dev@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-dev ___ linaro-dev mailing list linaro-dev@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-dev
Re: 3.3 kernel from TI landing teams and powervr problems.
On 04/17/12 01:17, Clark, Rob wrote: On Mon, Apr 16, 2012 at 10:07 AM, Martin Ertsåsmarti...@gmail.com wrote: True, I renamed that file, since I got the message that it couldn't find module omap. First I renamed it to omap_drv.so, which gave me the same error, then I tried pvr_drm.so which is the attatched logfile. Both of them are the omap_pvr_drv.so file. yeah, you probably shouldn't really rename the xorg modules.. the xorg module loader will look for symbols within the driver matching the name. I expect omap_drv.so is in a different package (maybe missing dependency?).. it is opensrc, anyways, you can rebuild it from the git tree at freedesktop.org http://cgit.freedesktop.org/xorg/driver/xf86-video-omap/ git://anongit.freedesktop.org/xorg/driver/xf86-video-omap BR, -R I know, it was kind of a last attempt to get stuff working. I'll try compiling the xf86-video-omap then. Thanks for the help. - Martin Should there be a omap_drv.so file in the same directory as the omap_pvr_drv.so file? In that case there isn't, the omap_pvr_drv.so file is the only one in that directory when extracting that tarball. Actually there is no file at all named omap_drv.so: $ find . -name omap_drv.so return nothing, while just omap gives $ find . -name omap* ./usr/lib/debug/usr/lib/xorg/modules/drivers/omap_pvr_drv.so ./usr/lib/xorg/modules/drivers/omap_pvr_drv.so which is the same as *drv* gives. - Martin On 04/16/12 16:42, Clark, Rob wrote: There is an omap_drv.so which should be the xorg driver to load.. it will attempt to load (if present) omap_pvr_drv.so as a submodule for EXA accel. I'm not entirely sure why it is named pvr_drv.so in your filesystem, but from list of symbols it seems to match what I would expect to be omap_pvr_drv.so BR, -R On Mon, Apr 16, 2012 at 8:45 AM, Martin Ertsåsmarti...@gmail.com wrote: Hi. I am not 100% sure this is the correct place to ask, but since I know at least Ricardo Salveti are working on these drivers for omap4, and I haven't heard anything from the ppa's mailing list, I try to ask here. If this was completely wrong I'm sorry, but would still appreciate a hint as to where to ask and who to contact. We are compiling what is now HEAD in the tilt-3.3 branch from the TI landing team. To compile the module needed for the powervr drivers, we had to get the pvr-omap4-dkms-1.7.15 and pvr-omap4-1.7.15 from http://ppa.launchpad.net/tiomap-dev/omap-trunk/ubuntu/pool/main/p/ We downloaded the tar.gz from those, extracted them and compiled them. Problem starts when we boot the board and try to run startx. The xorg config which comes with the later package is loaded, but when trying to start X I get the error: (EE) Failed to load /usr/lib/xorg/modules/drivers/pvr_drv.so: /usr/lib/xorg/modules/drivers/pvr_drv.so: undefined symbol: OMAPFinishAccess None of the libraries in pvr-omap4 contains this symbol, so I'm wondering if I'm doing something wrong, or if there is something wrong with the binary? When comparing to the library from rsalveti, from people/rsalveti/pvr-omap4.git, I see that the dynsym table contains 232 entries, but there are only 64 entries in the same table in the other library. Is this correct, or should there be more entries in the 1.7.15 library? Both the Xorg log and the readelf output from both libraries are attached, if you have questions/need more info, please do not hesitate to ask. Thank you in advance for the help, and thank you for the great work you put in for making embedded Linux better. Best regards Martin Ertsaas ___ linaro-dev mailing list linaro-dev@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-dev ___ linaro-dev mailing list linaro-dev@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-dev
Re: Linux kernel with rpmsg and possibility for compiling powervr drivers
On 03/17/12 01:13, Andy Green wrote: On 03/17/2012 04:14 AM, Somebody in the thread at some point said: Mertsas, On Fri, Mar 16, 2012 at 10:58 AM, Martin Ertsas (mertsas) mert...@cisco.com wrote: Sent from Samsung Mobile Andy Green andy.gr...@linaro.org wrote: On 03/16/2012 10:46 PM, Somebody in the thread at some point said: If you look at tilt-android-tracking, there is a complete 1.8 SGX (usable on ICS) on fairly recent basis which includes its own rpmsg stack as part of the SGX port. The matching userlands are available via google AOSP. http://git.linaro.org/gitweb?p=landing-teams/working/ti/kernel.git;a=shortlog;h=refs/heads/tilt-android-tracking We also have what's currently only working on I was a bit unclear. Sorry. First, we are thinking of using Linux for this task instead of Android. We want rpmsg support to get access to the m3 devices for h264 encoding and decoding. with pvr I basically meant a dss which can be used to compile rsalvetis pvr omap 4 kernel module. Bit unclear is an understatement in this case ^^ rpmsg is a complete red herring then. 1.7 SGX as currently used in Ubuntu does not use rpmsg. All the currently available mm decode solutions for Ubuntu don't use it yet either, they use its predecessor syslink. If you check out tilt-3.1 branch from our repo that has working syslink + tiler pieces needed by the mm decode pieces in Ubuntu and will build against SGX dkms. I seemed to have missed some conversation here. Just wanted to check if you still have trouble getting a kernel with rpmsg and pvr? I am not aware of the pvr part of things but for the rpmsg stuff, like Andy mentioned the one tilt has is a different from the up-streamed one. You will get a lot more features in the tilt version of it. The flip side being, it is dependent on ion. Can you elaborate what you are trying to achieve with rpmsg? You definitely need tiler driver support too to get h264 decoder working. Personally, I think , you are better off moving to rpmsg, as you can expect more support in future on this. I might not be aware of all the minute details but could help you out on the rpmsg part. Agreed rpmsg has deprecated syslink going on, and medium or long term planning will need to be around that. However for Ubuntu, there's not quite yet anything rpmsg-based that's workable to give him mm decode (it seems to be that he's interested in) that will provide userland pieces he can have as a normal mortal as well. Rob Clark is working on something that should bear fruit soon, Nice guys will provide matching Ubuntu pieces and TILT will integrate it into our tracking. Until then tilt-3.1 + Ubuntu packages already there seems the best choice to getting something known to work end-end quickly. Since it integrates to gstreamer already when he upgrades to rpmsg-based solution (userland too) it should hopefully be transparent. -Andy The reason we want rpmsg is to get h264 endcoding and decoding using gstreamer elements. I'm not sure if gst-ducati works with syslink as well? Andy is right that it is mm decode/encode I am interested in. So if I have understood you right: For the time being, I should use tilt-3.1 with syslink and gst-ducati (or some other gst element?) to get mm h264 encoding/decoding working on the omap. And when rpmsg is in a stable upstream/tilt kernel, I can switch to that and still use gst-ducati? - Martin ___ linaro-dev mailing list linaro-dev@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-dev
Re: Linux kernel with rpmsg and possibility for compiling powervr drivers
On 03/19/12 11:22, Dechesne, Nicolas wrote: Martin, On Mon, Mar 19, 2012 at 8:14 AM, Martin Ertsås mert...@cisco.com mailto:mert...@cisco.com wrote: The reason we want rpmsg is to get h264 endcoding and decoding using gstreamer elements. I'm not sure if gst-ducati works with syslink as well? Andy is right that it is mm decode/encode I am interested in. So if I have understood you right: For the time being, I should use tilt-3.1 with syslink and gst-ducati (or some other gst element?) to get mm h264 encoding/decoding working on the omap. And when rpmsg is in a stable upstream/tilt kernel, I can switch to that and still use gst-ducati? at the current time, as andy said if you want a workable GST with decoders and encoders using gst-ducat, the only solution is 3.1 kernel with syslink2.0. I am not entirely sure about the exact status of Linaro releases with this regards (rsavelti can comment on that), but for sure if you use a plain 'ubuntu 11.10' with the TI OMAP 'release' PPA packages installed (see http://groups.google.com/group/pandaboard/browse_thread/thread/6aa2cbe99c36919d) then you will get such a system. PVR drivers are working for UI, and PVR is used as well for video rendering (through GST pvrvideosink element). With all these packages installed, you can use gst-launch or any gst based application, and video codecs will be h/w accelerated through gst-ducati. we are *currently* working on the next gen GST stack which will be based on 3.3+ kernel with rpmsg (aka syslink3) and a new graphics architecture based on our new omapdrm GEM based driver. This is work-in-progress at the moment, and we are publishing our 'experimental' work in our OMAP 'trunk' PPA (~tiomap-dev/omap-trunk in LP). At this point of time, we have updated libDCE , kernel, and firmware and we have some basic video decode working with this new stack, but 1) we haven't started the GST updates, 2) it's not recommended if you want something that 'just works'. Support for encoders, and support at GST level (gst-ducati) will be available in the coming weeks. Ok. Thank you so much for all your help, it have been really helpful. Will work with linux-3.1 and syslink 2.0 then. Really appreciate the response. - Martin ___ linaro-dev mailing list linaro-dev@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-dev
Linux kernel with rpmsg and possibility for compiling powervr drivers
Hi. I'm currently fumbling around in the dark, trying to find a kernel which I can use to compile the powervr drivers, as well as with rpmsg support. Ideally the rpmsg should already be in the kernel, but I can see that this might be hard, and cherry picking is an ok solution. I'm wondering if one of the ti landing team kernels might be suitable for this, and if one of these is considered a stable kernel? Last kernel I looked at was the LEB 3.1.5+ kernel, release-linux-2011-12, which worked fine for compiling the pvr drivers from rsalveti, but didn't contain rpmsg drivers. When I tried the release-linux-2012-01 kernel, 3.2.0+, it could compile the powervr driver, but didn't contain the rpmsg drivers and didn't actually run, as I got the message 'coherent pool not initialized' from the kernel when trying to allocate for the ehci-omap. The kernel will be run on a device with an omap4460 if this is of interest. I would really appreciate your help on this, as I feel quite lost in all the different branches and tags which reside in the different linaro kernels, so a repo with a stable commit would really help. As most rpmsg implementations I have seen follows the 3.2 kernel, that would really be preferable if I need to cherry pick, but if not I don't really care if it's 3.1 or 3.2. If you have any questions, please do not hesitate to ask them. Best regard Martin Ertsas ___ linaro-dev mailing list linaro-dev@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-dev