Hello Kim,
There is probably some kind of issue or a missed update. I definitely see the required-opps properties in the latest kernel in that branch.
See https://git.codelinaro.org/linaro/qcomlt/kernel/-/blob/release/db820c/qcomlt...
On 15/02/2023 21:13, Kim Steiner wrote:
** comments
-----Original Message----- From: Dmitry Baryshkov dmitry.baryshkov@linaro.org Sent: Tuesday, February 14, 2023 9:06 AM To: sightlineapplications@lists.linaro.org Cc: Kim Steiner kim.steiner@sightlineapplications.com Subject: Project updates
Hello,
A set of updates from our side.
- We have started the process of migrating our repositories to new service. Please consider migrating the repository url on your side from git.linaro.org to https://git.codelinaro.org/linaro/qcomlt/kernel/ .
The branch name states intact (release/db820c/qcomlt-5.15) The old repository location will mirror the new location, but it might take some time for the changes to propagate to it.
** I noticed in the new location you mention that under the gpu_opp_table there is no required-opps = on any of the frequencies.
Could you please doublecheck, which branch were you looking at? There are required-opps in the 'release/db820c/qcomlt-5.15' branch. Please do get confused with the bare 'release/qcomlt-5.15' branch, it doesn't contain patches developed for you for the apq8096/msm8996 support.
In the db820c/qcomlt-5.15 that there is such a 'required-opps' field in each of the frequencies: opp-133000000 { opp-hz = /bits/ 64 <133000000>; opp-supported-hw = <0xFF>; required-opps = <&cpr3_gfx_opp1 &rpmpd_opp4>; };
- I have pushed a set of patches fixing two issues I narrowed down during my tests of the GPU. I will continue working on the driver.
I will pull the latest and try. 2) Regarding fixing the GPU clocks. Please consider applying the attached patch, which should fix the minimum clock rate to 510 MHz. ** the patch causes error messages when we run our software. Plus the patch is missing any mention of 'required-opps' field so it isn't a direct patch:
drmModeGetResources returned NULL *** ERROR -1 creating dumb buffer *** *** ERROR -1 mapping dumb buffer *** Could not map buffer exported through PRIME : Bad file descriptor (9) Buffer : 0xffffffffffffffff *** ERROR -1 creating dumb buffer *** *** ERROR -1 mapping dumb buffer *** Could not map buffer exported through PRIME : Bad file descriptor (9) If I interpret your last test results correctly, the issue really is in the GPU rate handling, so this patch should provide a workaround.
-- With best wishes Dmitry