Closed b-rad-NDi closed 7 years ago
Something happened between 06/20 and 07/01, awaiting hardware to debug and get back in sync with Linuxtv.org media tree.
I have some patches for this (using all 22). How would you like them? They aren't perfect for the quadhd 885 (166200/1) 2157A30 (fw 3.0.5 internal) 2168B40 (fw4.0.25 from antti) but they don't crash anymore and I can record on some inputs.
Note I'm running mythtv with 4.12.6 (debian testing) regards Mark
Thanks Mark. I have been extremely busy all summer and hadn't gotten a chance to dedicate time to this. I will review your patch set.
Hi Mark. The i2c lifetime bug is fairly savage. Since it is a driver wide issue I'm going to apply it before I make my changes, then adapt my changes to fit the new allocator.
For the T2 auto fix, i'm going to work that into my T2 auto patch and get our euro testers to give it a shot.
I'll be updating the linux-oops branch with the latest work as I'm testing this out. Thanks again, fixing this had just made its way to the front of my queue. If you want to make separate issues for the other pieces you say are missing or broken still I can look into whatever you're not currently sorting.
@mspieth : Could you create a new issue and post logs indicating the failure you are seeing with our T/T2 scanning patch? This patch you've provided just tries T or T2 if the other mode fails. We would like further info.
@mspieth : I have applied your cx23885 i2c alloc patch. I added some missing kfrees. It is 0012-cx23885-i2c-platform-data-cannot-be-stack-allocated.patch in zesty extra.
It is in the linuxtv-oops-with-extra-applied branch.
You missed a couple.
Patch attached.
The si2157 bit is part of the corresponding patch, as a sanity check. I will fix the other three parts tomorrow.
I fixed those 3 issues. I also added the same i2c platform data fix to cx231xx driver, as it has the same problem.
Unfortunately the i2c lifetime issue is not the sole source of the oops in this issue. I am continuing to sort out why I get a very similar oops with our usb devices.
I reversed this patch: https://patchwork.linuxtv.org/patch/41960 and now all the devices which were previously oopsing are perfectly fine. Once I noticed this was affecting a lot of devices I had, each using independent unrelated drivers, I identified that it had to be a core issue. This patch was coincidentally applied briefly after the last working linuxtv patch I made, so it was reviewed quickly and immediately made me suspicious.
I am closing this issue. The reverse of the kvmalloc patch has fixed this particular oops. Everything appears all good with the various devices tested. I will continue to investigate your other issues.
All changes are applied and a fresh linuxtv.org patch was generated in the linuxtv-oops-with-extra-applied directory.