Testing Raspberry Pi OS Bookworm in the Pi 4 was definitely worthwhile. That combined with the x86 report on this threadFirst of all I have just run some tests that I should have run a few days back. As I have said I tested file transfer from a Pi4 running Bullseye without any problems. As the page file size is 4K I have rerun the tests to ensure I was trying with enough file transfers. Today I have transferred over 120GB without issue.I also can't see a memory leak go undetected by those who test for such in the kernel. Could it be a protocol incompatibility with the Synology NAS? The fact that the Pi 4 seems fine is mysterious as is the fact that both Pi 5's lock up.
The telling test was when I moved the Bookworm SD to the Pi 4 (with the 4k page size). Each time I managed to move a single 10GB file but the system hung every time I tried to move a second file.
From this we can conclude there is not, as I had originally supposed a NVME problem, as it also occurs when only a SD card is involved. As the problem can also be seen on a Pi 4 it is not a Pi5 problem. This leaves us with Bookworm and the NAS. Suggesting there is a change in the handling of CIFS data transfers, in Bookworm, which triggers a problem. I must admit my thought is to agree with the possibility of a memory leak but obviously an obscure one. Otherwise it would have been discovered by others.
makes me wonder whether something actually went wrong with Bookworm upstream.not exactly a RPi item, but very consistent with the subject.
At this point the lockup seems more serious than it initially appeared. Maybe posting the output of
# testparm -s
as suggested when
is the next step.I have been following this and one thing I haven't seen is the output of:From the rpiCode:
testparm -s
Another thought is to avoid the Synology and set up one of the Pi computers as a samba server. I'm not sure if it would be more or less interesting if Pi to Pi transfers worked or not.
Statistics: Posted by ejolson — Mon Nov 18, 2024 3:17 am