Hello Hamish, > I'll figure it out eventually
perf(1)'s ‘kmem’ lets one record slab allocations, amongst over things. It should be possible to wait until your thirty-second loop shows the Sunreclaim steadily growing and the record the allocation across the whole system, say for ten seconds, and then format that to show the callers within the kernel. This may give a clue as to the type of activity or driver which is responsible. If perf doesn't help then BPF can be used to collect data from inside the kernel, e.g. slabratetop. https://github.com/iovisor/bcc#readme -- Cheers, Ralph. -- Next meeting: Online, Jitsi, Tuesday, 2020-10-06 20:00 Check to whom you are replying Meetings, mailing list, IRC, ... http://dorset.lug.org.uk New thread, don't hijack: mailto:dorset@mailman.lug.org.uk
