ChanServ changed the topic of #armlinux to: ARM kernel talk [Upstream kernel, find your vendor forums for questions about their kernels] | https://libera.irclog.whitequark.org/armlinux
<ukleinek>
broonie: ah, nevermind, this is PTR_ERR_OR_ZERO, not IS_ERR_OR_NULL.
tre has joined #armlinux
shoragan has quit [Excess Flood]
shoragan has joined #armlinux
sszy has joined #armlinux
mcoquelin has joined #armlinux
scosu has joined #armlinux
Pali has joined #armlinux
frieder has quit [Remote host closed the connection]
frieder has joined #armlinux
frieder has quit [Remote host closed the connection]
<geertu>
arnd: Is swap over NFS actually working? I tried that a while ago on the poor RZ/A1 with only 32 MiB of RAM, and it didn't work (hence I'm used to just rebooting when the background apt-update goes OOM)
frieder has joined #armlinux
alexels has joined #armlinux
<arnd>
geertu: I've never done it myself, but I'm fairly sure I've seen references to it in the code
<arnd>
there is a fundamental problem with memory allocations that are done as part of reclaim, so if the kernel is trying to swap out data in order to serve a memory allocation, this must complete without any need to do another allocation that triggers reclaim. We have __GFP_FS and __GFP_IO to mark allocations that are done as part of this
<arnd>
but if this causes a network driver to allocate memory in turn, it must behave the same way
<arnd>
I think there is a hack in the kernel to trigger that behavior in order to allow swap over nfs, but I'm not sure
* ukleinek
thought there was a GPT_NO_NETWORK or something for kmalloc?
<geertu>
_jannau__: thx, not yet applied, apparently?
nsaenz has joined #armlinux
<arnd>
swap-over-nbd support was merged 10 years ago https://lwn.net/Articles/504261/, this has a number of the same issues, but it's not quite as bad
<_jannau__>
geertu: no idea, I just remembered seeing it
<arnd>
or maybe this only deals with a total lockup in the allocator, not with the OOM issue
<arnd>
sk_set_memalloc() is set by sunrpc (nfs), nbd, wireguard nvme-over-tcp, cifs, and iscsi, which all have the same fundamental issue
matthias_bgg has joined #armlinux
headless has joined #armlinux
<geertu>
_jannau__: Thx, v3 seems to work on arm (running apt-update, 11 MiB swap in use)
<arnd>
geertu: are you running a regular Debian install on 32MB, or did you have to spend a lot of time disabling services?
<arnd>
apparently both openwrt (on mips32 and armv5) and tinycorelinux (on x86) require at least 64MB even for a minimal boot now, and I would have expected them to need less than Debian
apritzel has joined #armlinux
<shenki>
we have played with booting linux from the cache, but that's using buildroot. real userspace would oom
<shenki>
(8mb)
<arnd>
shenki: right, you can get fairly low with a lot of customizations, the bottom so far was Vitaly Wool booting an stm32 from ROM with under 256KB RAM a few years ago. My question was just about general-purpose distros
headless has quit [Quit: Konversation terminated!]
atorgue has joined #armlinux
<tmlind>
arnd: what's the status of arm-soc fixes for the -rc cycle? i though ojn was going to send the fixes from the arm-soc later branch..
<arnd>
tmlind: I thought so too, but apparently he didn't get around to it. I'll take over and apply the pending fixes today, will forward them this week
<tmlind>
arnd: ok thanks, i have "[GIT PULL] Fixes for omaps
<tmlind>
pull request i sent, should be still valid. from jan 10th
<tmlind>
that got pulled into arm-soc/for-next but then the merge window ended :)
<tmlind>
using that as a base for some clock clean up, that's why i'm asking
jlinton has quit [Quit: Client closed]
<geertu>
arnd: Debian 9 nfsroot.
<arnd>
geertu: ah, using an older release probably helps as well. I guess this would be without systemd?
<geertu>
Started with debootstrap many years ago, and cloned when new boards arrived. The initial install was not done on the 2 MiB machine, but probably on R-Car M2-W.
<geertu>
arnd: my risc-v boards are running bookworm. landik (SH) runs bullseye
<geertu>
most of my arm/arm64 boards are still at stretch, but some run buster
<geertu>
the rbtx4927 is stuck at jessie
<geertu>
swap over nfs is slow, though
<arnd>
geertu: you mean you wouldn't do that over a zorro8390?
prabhakarlad has joined #armlinux
<geertu>
arnd: :-)
<geertu>
arnd: the rskrza1 has 100Mbps Ethernet
tre has quit [Remote host closed the connection]
torez has joined #armlinux
iivanov__ has joined #armlinux
iivanov has quit [Ping timeout: 256 seconds]
matthias_bgg has quit [Read error: Connection reset by peer]
headless has joined #armlinux
Tokamak has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<robher>
mwalle: You are correct. The exception is if only one device is enabled at a time.
<mwalle>
robher: mh, so its ok for the dtsi to include both nodes with 'status = "disabled";'?
<robher>
mwalle: yes.
<mwalle>
robher: ok thanks
Tokamak has joined #armlinux
torez has quit [Ping timeout: 240 seconds]
matthias_bgg has joined #armlinux
Tokamak has quit [Read error: Connection reset by peer]
Tokamak has joined #armlinux
prabhakarlad has quit [Quit: Client closed]
kos_tom has quit [Remote host closed the connection]
kos_tom has joined #armlinux
prabhakarlad has joined #armlinux
torez has joined #armlinux
headless has quit [Quit: Konversation terminated!]
iivanov__ has quit [Ping timeout: 256 seconds]
iivanov__ has joined #armlinux
monstr has joined #armlinux
mraynal has joined #armlinux
torez has quit [Ping timeout: 240 seconds]
<mraynal>
geertu: Hello o/ I was about to apply "[PATCH v2] mtd: Replace the expert mode symbols with a single helper" in my tree but I figured out that you didn't officially acked it, I'm not sure if this is something you plan to review or if you marked it done in your inbox already?
Tokamak has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
torez has joined #armlinux
<geertu>
mraynal: Rb provide
<geertu>
d
<mraynal>
geertu: perfect, thanks!
Tokamak has joined #armlinux
Tokamak has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<arnd>
geertu: that sounds like a sam9x60, the SoC is only two years old. It's interesting that they managed to get wifi6e on it though, as there is no PCI
Tokamak has joined #armlinux
<jn>
via SDIO perhaps
<arnd>
probably actually faster than a MIPS 24Kc MT7620, which is the more common choice for this type of device
matthias_bgg has quit [Quit: Leaving]
guillaume_g has quit [Ping timeout: 250 seconds]
headless_ has joined #armlinux
headless has quit [Ping timeout: 240 seconds]
frieder has quit [Remote host closed the connection]
alexels has quit [Quit: WeeChat 3.4]
monstr has quit [Remote host closed the connection]