On 6/9/23 12:06, Linus Torvalds wrote:
On Fri, Jun 9, 2023 at 11:42 AM Guenter Roeck linux@roeck-us.net wrote:
I managed to revise my bisect script sufficiently enough to get reliable results. It looks like the culprit is commit 503e554782c9 (" debugobject: Ensure pool refill (again)"); see bisect log below. Bisect on four different systems all have the same result. After reverting this patch, I do not see the problem anymore (again, confirmed on four different systems).
Does this happen on mainline too? It's commit 0af462f19e63 in the upstream tree.
It was in 6.4-rc1, and I see a clean result from you at least for -rc2, so for some reason it sounds like upstream is ok. But I don't really see why that would be the case...
I see the problem only in v5.15.y, to the point where it is almost impossible to get a clean test of all arm-v7 systems. Affected are npcm (Nuvoton) boards (kudo-bmc, quanta-gsj, npcm750-evb) as well as orangepi-pc. I don't see it in any other branch or with any other platform/architecture.
Mainline is fine; I have not seen any problems since -rc2.
I have no idea what is going on either, only that I can reliably reproduce the problem (and of course it disappears if CONFIG_DEBUG_OBJECTS is disabled).
Guenter