<f00b4r0>
enyc: I've seen hostapd trigger slab unreclaimable leaks, haven't checked status in 24.10.1. nbd is aware
<f00b4r0>
bbiab
gromero has quit [Ping timeout: 480 seconds]
rua has joined #openwrt-devel
MAbeeTT3 has joined #openwrt-devel
<enyc>
f00b4r0: thankyou; is that relatively recent known issue? worth my while trying wpad-basic-openssl instead of wpad-basic-mbedtls now?
rua has quit [Quit: Leaving.]
<f00b4r0>
enyc: it's a 24.10 issue, hadn't seen that in 23.05. That's with stock hostapd (wpad-basic?) and it seems to predominently affect some ramips hardware (at least in the cases witnessed so far)
<enyc>
f00b4r0: is there a bug report? follow what is happening when any 24.10 patch worth testing?
<f00b4r0>
the most visible symptom is a growing SUnreclaim. It may be slow growing or occasionally pretty fast (several 10s of MB within a couple hours), until the machine OOMs and hostapd is eventually killed, which restores the situation to "normal".
fakuivan has quit [Remote host closed the connection]
<f00b4r0>
issuing "wifi up" also clears that (and disconnects clients of course)
<enyc>
f00b4r0: so, who/hat is nbd? how will this be sorted/investigated further?
<f00b4r0>
¯\_(ツ)_/¯
<enyc>
would it be helpful to confirm the openssl version over the mbedtls version?
<f00b4r0>
as I said this occurs on non-tls versions
<enyc>
So, no? hrrm
<enyc>
f00b4r0: nbd some developer? not just network block device?
<enyc>
nbd: AHA ;O
<enyc>
f00b4r0: fwiw the "stock" hostapd seems to *be* wpad-basic-mbedtls
<f00b4r0>
oh indeed
<f00b4r0>
i was confused
<enyc>
f00b4r0: SUnreclaim: 16876 kB
<f00b4r0>
looks normal
<enyc>
wifi up and goes down to 15xxx kB at first but then SUnreclaim: 16948 kB
<enyc>
slowly climbing
<enyc>
dropped
<f00b4r0>
i have a script that kicks in at 50MB (on 256MB devices)