Adding linaro-multimedia who also cannot see video on Panda.
On 3 May 2011 03:48, Jim Huang jim.huang@linaro.org wrote:
On 3 May 2011 16:19, Alexander Sack asac@linaro.org wrote:
On Mon, May 2, 2011 at 10:44 PM, Zach Pfeffer zach.pfeffer@linaro.org wrote:
E/FramebufferNativeWindow( 2057): couldn't open framebuffer HAL (No such device)
yes, this is the problem. Jim says this works for him. He should have exact info what is needed to make this happen. The kernel Jim gave me didnt work for me either, so I think that Jim must have more changes locally that are not in our clean leb-panda build.
Alternatively, we can roll back to Froyo based GFX, which worked for me, too.
I'm pulling from omap kernel again and rebuilding from scratch. Will confirm you guys soon.
Thanks, -jserv
On Mon, May 2, 2011 at 10:44 PM, Zach Pfeffer zach.pfeffer@linaro.org wrote:
E/FramebufferNativeWindow( 2057): couldn't open framebuffer HAL (No such device)
btw, usually a failure here could be a mismatch between kernel and userspace pvr stuff.. version mismatch or configuration mismatch could cause this.. (unfortunately non debug builds have all the traces stripped out making identifying *what* is misaligned a bit tricky to identify)
BR, -R
Yeah...just fought through similar issues with the DM8168.
Jim, do you have Code Composer and the JTAG or are you debugging the SGX without it?
-Zach
On 3 May 2011 08:34, Clark, Rob rob@ti.com wrote:
On Mon, May 2, 2011 at 10:44 PM, Zach Pfeffer zach.pfeffer@linaro.org wrote:
E/FramebufferNativeWindow( 2057): couldn't open framebuffer HAL (No such device)
btw, usually a failure here could be a mismatch between kernel and userspace pvr stuff.. version mismatch or configuration mismatch could cause this.. (unfortunately non debug builds have all the traces stripped out making identifying *what* is misaligned a bit tricky to identify)
BR, -R
On 3 May 2011 23:52, Zach Pfeffer zach.pfeffer@linaro.org wrote:
Yeah...just fought through similar issues with the DM8168.
Jim, do you have Code Composer and the JTAG or are you debugging the SGX without it?
hi Zach and Rob,
I have no tool to debug SGX, and I have no background knowledge to resolve SGX issues.
How can I do?
Thanks, -jserv
On 3 May 2011 08:34, Clark, Rob rob@ti.com wrote:
On Mon, May 2, 2011 at 10:44 PM, Zach Pfeffer zach.pfeffer@linaro.org wrote:
E/FramebufferNativeWindow( 2057): couldn't open framebuffer HAL (No such device)
btw, usually a failure here could be a mismatch between kernel and userspace pvr stuff.. version mismatch or configuration mismatch could cause this.. (unfortunately non debug builds have all the traces stripped out making identifying *what* is misaligned a bit tricky to identify)
On Tue, May 3, 2011 at 9:40 PM, Jim Huang jim.huang@linaro.org wrote:
On 3 May 2011 23:52, Zach Pfeffer zach.pfeffer@linaro.org wrote:
Yeah...just fought through similar issues with the DM8168.
Jim, do you have Code Composer and the JTAG or are you debugging the SGX without it?
hi Zach and Rob,
I have no tool to debug SGX, and I have no background knowledge to resolve SGX issues.
How can I do?
Not sure if there is any way to get a build w/ debug enabled for both kernel and userspace side (because again, debug is one of those options that if you enable on kernel side you have to enable on user side and visa versa)
How are you getting the userspace binaries? Are those prebuilt by someone? I'm not quite sure how it works for android, but I guess if we can track down who can provide debug binaries for userspace (BUILD=debug), then for kernel build in menuconfig go to Device Drivers -> Graphics Support -> PowerVR build type, and select 'SGX540 DEBUG'
I'm not sure JTAG would help much.. getting all the traces that are disabled at build time by default would, because this would tell you which bits in the features mask mismatch, etc.
BR, -R
Thanks, -jserv
On 3 May 2011 08:34, Clark, Rob rob@ti.com wrote:
On Mon, May 2, 2011 at 10:44 PM, Zach Pfeffer zach.pfeffer@linaro.org wrote: > E/FramebufferNativeWindow( 2057): couldn't open framebuffer HAL (No such device)
btw, usually a failure here could be a mismatch between kernel and userspace pvr stuff.. version mismatch or configuration mismatch could cause this.. (unfortunately non debug builds have all the traces stripped out making identifying *what* is misaligned a bit tricky to identify)
jim, have you double checked with TI folks in linaro-android channel on the exact kernel/binary version combination to use? when i talked to them they were quiet helpful. only problem was that the had not pushed the defconfig yet and that there were build failures for kernel. anyway, first step is really to make our leb-panda build use a) their proposed kernel tree/version and b) include their exact binaries. then when we have a build we can daily show we can probably get them to fix things or at least get ideas from them on what might be wrong.
at least we have then done our part of the homework. zach, have you tried with the omapzoom.org kernel? maybe try 35 and 38 and 39 branches and see what happens. fact is that we cannot fix and debug this on our own right now, but should report our findings as bugs properly.
thanks!
- Alexander Am 04.05.2011 03:40 schrieb "Jim Huang" jim.huang@linaro.org:
On 3 May 2011 23:52, Zach Pfeffer zach.pfeffer@linaro.org wrote:
Yeah...just fought through similar issues with the DM8168.
Jim, do you have Code Composer and the JTAG or are you debugging the SGX without it?
hi Zach and Rob,
I have no tool to debug SGX, and I have no background knowledge to resolve SGX issues.
How can I do?
Thanks, -jserv
On 3 May 2011 08:34, Clark, Rob rob@ti.com wrote:
On Mon, May 2, 2011 at 10:44 PM, Zach Pfeffer <
zach.pfeffer@linaro.org> wrote:
> E/FramebufferNativeWindow( 2057): couldn't open framebuffer HAL (No
such device)
btw, usually a failure here could be a mismatch between kernel and userspace pvr stuff.. version mismatch or configuration mismatch could cause this.. (unfortunately non debug builds have all the traces stripped out making identifying *what* is misaligned a bit tricky to identify)
On 4 May 2011 15:34, Alexander Sack asac@linaro.org wrote:
jim, have you double checked with TI folks in linaro-android channel on the exact kernel/binary version combination to use? when i talked to them they were quiet helpful. only problem was that the had not pushed the defconfig yet and that there were build failures for kernel. anyway, first step is really to make our leb-panda build use a) their proposed kernel tree/version and b) include their exact binaries. then when we have a build we can daily show we can probably get them to fix things or at least get ideas from them on what might be wrong.
hi Alexander,
The perfect one I knew is the package from Froyo_L27.8.2_Graphics-1.0-Linux-x86-Install.
Our OESF friend, sola, indicated before: http://blog.sola-dolphin-1.net/archives/3173852.html
I was using this kernel tree: https://github.com/sola-dolphin1/OHA-Android-2.3_r1.0/tree/master/kernel
However, after switching to p-android-omap-2.6.38 (omap-kernel), I encountered the same problem as you.
According to git log:
commit 2cf3c35decc02cc198d835596e0393490ef3414e Author: seonkon.choi seonkon.choi@ti.com Date: Wed Apr 13 10:20:20 2011 +0900
OMAP3: HWMOD: Fixed bugs of GPU hwmod.
The GPU(SGX) has a type2 SYSCONFIG at 0x50000000 + 0xfe10.
Needed for SGX-KM 1.7.17.4403.
Change-Id: Id07801673d2a57858cb22a0fbb9e109cd7d2f2d7 Signed-off-by: seonkon.choi seonkon.choi@ti.com
Also, the kernel configuration:
$ grep SGX .config CONFIG_SGX540=y # CONFIG_SGX_REV110 is not set CONFIG_SGX_REV120=y CONFIG_SGX_540_BUILD_RELEASE=y
I am checking the kernel messages triggered by CONFIG_SGX_540_BUILD_DEBUG.
Thanks, -jserv
prob with froyo is that we dont have a license. ti folks sais that 35 kernel should work with binaries they have in git. maybe we use binaries from wrong sgx brannch? please double check with them which versions exactly are supposed to work.
- Alexander Am 04.05.2011 08:43 schrieb "Jim Huang" jim.huang@linaro.org:
On 4 May 2011 15:34, Alexander Sack asac@linaro.org wrote:
jim, have you double checked with TI folks in linaro-android channel on
the
exact kernel/binary version combination to use? when i talked to them
they
were quiet helpful. only problem was that the had not pushed the
defconfig
yet and that there were build failures for kernel. anyway, first step is really to make our leb-panda build use a) their proposed kernel
tree/version
and b) include their exact binaries. then when we have a build we can
daily
show we can probably get them to fix things or at least get ideas from
them
on what might be wrong.
hi Alexander,
The perfect one I knew is the package from Froyo_L27.8.2_Graphics-1.0-Linux-x86-Install.
Our OESF friend, sola, indicated before: http://blog.sola-dolphin-1.net/archives/3173852.html
I was using this kernel tree: https://github.com/sola-dolphin1/OHA-Android-2.3_r1.0/tree/master/kernel
However, after switching to p-android-omap-2.6.38 (omap-kernel), I encountered the same problem as you.
According to git log:
commit 2cf3c35decc02cc198d835596e0393490ef3414e Author: seonkon.choi seonkon.choi@ti.com Date: Wed Apr 13 10:20:20 2011 +0900
OMAP3: HWMOD: Fixed bugs of GPU hwmod.
The GPU(SGX) has a type2 SYSCONFIG at 0x50000000 + 0xfe10.
Needed for SGX-KM 1.7.17.4403.
Change-Id: Id07801673d2a57858cb22a0fbb9e109cd7d2f2d7 Signed-off-by: seonkon.choi seonkon.choi@ti.com
Also, the kernel configuration:
$ grep SGX .config CONFIG_SGX540=y # CONFIG_SGX_REV110 is not set CONFIG_SGX_REV120=y CONFIG_SGX_540_BUILD_RELEASE=y
I am checking the kernel messages triggered by CONFIG_SGX_540_BUILD_DEBUG.
Thanks, -jserv
linaro-multimedia@lists.linaro.org
-
Alexander Sack -
Andy Green -
Clark, Rob -
Jim Huang -
Zach Pfeffer