Hi,
Am 02.10.2018 um 17:29 schrieb Michael Mrozek EvilDragon@openpandora.org:
Am Tue, 2 Oct 2018 14:39:28 +0300 hat Grazvydas Ignotas notasas@gmail.com geschrieben:
Hi,
I'm back and I've done some tests, no useful conclusions though. I have 2 4GB boards now, I'll call them old one and new one.
- the old board also hangs, although it seems to take much longer (3
hours at least)
- the new one hangs in minutes, but occasionally it can run for half
an hour or so
- u-boot tester doesn't fail on new one, although I haven't tried it
for too long (less than 10 minutes)
I've run them for half an hour without any issues, so yes, it doesn't seem to fail.
- memtester (on Linux) doesn't fail on new one, it also seems to allow
the new board to survive much longer. I think it survived overnight but hung later when I was at work.
THAT is interesting...
ideas:
- perhaps the problem has something to do with idle/power saving. Need
to check if there are RAM registers related to that. Does anyone know if it currently tries to do RAM self-refresh while idle?
Hm, but would idle / power saving trigger when the system is doing something?
Well, I think it will trigger if the system is NOT doing something :)
Because when doing my test runs, I always run a script that spits out the voltages, temperatures and remaining temperature left every 10 seconds. So Linux doesn't really idle around, but freezes anyways after some minutes. And sometimes it even freezes while booting - and it's anything but idle there.
Or do you mean something else?
I remember that there is a patch which does control some CPU idling and is reported to have an effect in the range of minutes:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?h...
But:
a) it is also in omap5evm mainline kernels b) I have tried reverting this patch some time ago but don't remember to have an effect wrt. freezes (I didn't measure power demand)
- try without any TILER patches. I can vaguely remember it was said
the TILER can't be used at all because of some errata, but Matt proved otherwise. Maybe with 4GB/LPAE the problem actually triggers.
I remember that (that's why it was so slow at the beginning), haven't found much about TILER in the errata though. Only one which doesn't seem to be too critical.
I tried that a while ago (perhaps several times), but it would be good to repeat and cross-check. My findings may have been disturbed by somthing else.
But it's a good idea trying that, yes. If someone compiles me a kernel, I'll test anything on various boards whatever ideas you have :)
I can't code much, but I can test.
Here's what I planned to do:
- See if a non-LPAE-Kernel has a different effect
- Boot the kernel directly into a shell (without loading any other
drivers except for the ones directly included in the kernel). Test whether this runs more stable and then modprobe the remaining drivers one after the other. After all, Tony was running the memtest similar as well, and that didn't fail.
Any other ideas or comments are very welcome :)
BTW: what I found is that Freezes seem to have the characteristics that the Reset button has no function. I.e. the OMAP5 isn't responding to external NRESWARM any more. This seems to be different from the omap5evm freezes I have. There I can always reboot with RESET if I remember correctly.
And I had a bad setup on one SD card some days ago for the LC15-eval board. This repeatably triggered a Freeze ca. 3 seconds into boot time (last messages were that LDO.8 and LDO.5 were initialized). The RESET button was responsive up to ca. 0.1 seconds before that, but not after.
Currently, I am trying to understand the details of the setup that makes this happen. Most likely it was a wrong U-Boot version for the board or a wrong DTB.
BR, Nikolaus