minimal has quit [Quit: Leaving]
rz has joined #openwrt-devel
skynet2 has joined #openwrt-devel
skynet2 has quit [Ping timeout: 480 seconds]
torv has quit [Remote host closed the connection]
torv has joined #openwrt-devel
parazyd has joined #openwrt-devel
tSYS has quit [Quit: *squeak*]
tSYS has joined #openwrt-devel
gch981213 has joined #openwrt-devel
<owrt-images-builds> Build [#164](https://buildbot.openwrt.org/images/#/builders/205/builds/164) of `master_mpc85xx/p1020` completed successfully.
<dwfreed> ah, yes, let's compare risc-v to arm, great comparison there
<\x> it has both which is ahaha nice
<\x> so you can boot aarch64, a single 1GHz A53 then with a flick of a switch you can boot dual core riscv its like 1+0.7GHz
<KGB-2> https://tests.reproducible-builds.org/openwrt/openwrt_mpc85xx.html has been updated. (100.0% images and 100.0% packages reproducible in our current test framework.)
goliath has joined #openwrt-devel
mcbridematt has quit [Quit: Leaving]
mcbridematt has joined #openwrt-devel
robimarko has joined #openwrt-devel
fakuivan has quit [Remote host closed the connection]
<Ansuel> nbd any chance for the rewrite of the thing?
<nbd> haven't had time yet
<aparcar> Ansuel: what thing?
<aparcar> nbd: please let me know if you have time to debug unetd hole punching with me 🙂 it kind of never works
<Ansuel> also did you made some experiment with the bucket?
<aparcar> Ansuel: not yet, working on mbedtls support for apk
<aparcar> in case you want to help with that, please reach out 🙂
<aparcar> I wanted to have mbedtls support before compiling the images
<Ansuel> did we make some space difference between opkg and apk ?
<aparcar> no since it's compiled with openssl right now
<aparcar> 675 130.0K -rwxr-xr-x 1 root root 129.8K Apr 10 16:42 apk
<aparcar> 744 194.0K -rwxr-xr-x 1 root root 193.9K Apr 10 16:42 libapk.so.2.99.0
<aparcar> 37 195.5K -rwxr-xr-x 1 root root 195.5K Mar 22 23:09 opkg
<aparcar> I think we can later on strip some stuff from APK, right now it got an audit feature and a container runtime somewhat included
<Ansuel> and what are the progress with mbedtls ?
<aparcar> mostly done'ish
<aparcar> we need mbedtls 3.60
<Ansuel> mhhh what is missing?
<Ansuel> the fixme?
<Ansuel> oh nice that single check is the only part ?
<aparcar> I think one needs to decode the public key and read it
<Ansuel> are we sure mbedtls_pk_parse_key doesn't already do what openssl i2d_PublicKey does in 2 step ?
<aparcar> Nope
<Ansuel> aparcar any idea how to test the thing?
<aparcar> first make it compile 🙂
<Ansuel> ouch
<Ansuel> HAHAHA
<Ansuel> but maybe i'm on to something
robimarko has quit [Remote host closed the connection]
<aparcar> Ansuel: please see the DMs 🙂
robimarko has joined #openwrt-devel
<KanjiMonster> Mangix: ah, wasn't aware we have a local backport of it
skynet2 has joined #openwrt-devel
rsalvaterra has joined #openwrt-devel
minimal has joined #openwrt-devel
<Ansuel> robimarko did you notice nss-dp makes the kernel panic on rmmod ?
<mrnuke> Ansuel: It's written in the ancient scriptures that rmmod shall crash. You haven;t been keeping up on your QCA mythology, have you?
goliath has quit [Quit: SIGSEGV]
<Ansuel> well this is not true...
<Ansuel> * The cleanup can happen from data plane override
<Ansuel> * or from module_exit, we want to cleanup only once
<Ansuel> */
<Ansuel> they have some logic to support that
<mrnuke> QCA and logic in the same context ...?
<mrnuke> I have noticed that too. On ipq8074, on ipq6018, and lately on ipq95xx. It crashes equally well
<Ansuel> think i got why the thing crash... flawed logic i guess? They call the cleanup once but at the first call instead of at the last call (aka when there is no usr of the context)
<Ansuel> also this is all loverly caused by them allocating a shared struct at the start of the driver instead of refcount and putting the priv struct in the nss_dp_data_plane_ctx
<mrnuke> I noticed this trend in kmod-qca-... to allocate a metric tonne of global data structures
<mrnuke> Ansuel: Do you think it's worth trying to put in the time ti fix the rmmod ?
<Ansuel> if i'm right about this it's a quick fix
<Ansuel> but i recently wiped my buildroot with all my staging stuff so I'm waiting for everything to rebuild
<Ansuel> mrnuke tempted to drop the global struct and add a priv entry in the nss_dp_data_plane_ctx struct :D
<mrnuke> Have you seen QCA's attempt at an upstream driver? ( https://patchwork.kernel.org/project/linux-arm-msm/list/?series=815719&state=* )
<Ansuel> well aware
<mrnuke> It seems to be missing the part that calls netdev_register(), so pretty useless
<robimarko> Ansuel: Well, I would be shocked if it did not crash
<Ansuel> also offered jie some support to it but their respons was "sure we will review the changes internally"
<mrnuke> But I was wondering if we could adopt that as an interim solution, and just add the missing bits
<Ansuel> (like it wasn't was I was saying but ok...)
<mrnuke> robimarko: Well said!
<Ansuel> robimarko they assume everything is single core maybe?
<robimarko> That shouldnt matter at all, they probably dont care about rmmod as its never going to be removed
<robimarko> And something is not freeing properly
<Ansuel> mrnuke also that ppe is for ipq95xx
<Ansuel> robimarko problem is that things are freed while other port still needs to use the global struct AHAHAH
<Ansuel> indeed the kernel crash is on the second core that try to drop the second dp port... the first one correctly goes down
<robimarko> Hehe, that about sums up QCA QC
<robimarko> BTW, I think that our biggest issue is the remoteproc driver
<Ansuel> ?
<robimarko> Its working but not fully
<robimarko> Remoteproc reboot has been broken from the start and its rather a mess
<Ansuel> oh you mean the recovery part
<robimarko> Recovery is just an side effect of reboot being broken
<Ansuel> yes but that would be the main usage
<robimarko> And since remoteproc reset/reboot is broken there is no FW recovery after crash
<Ansuel> thing is that i don't think it's fixable by us
<robimarko> Sadly I think you are right
<robimarko> Something is fundamentaly broken
<robimarko> I am not sure if we chased all of the broken clocks as well
<Ansuel> the golden thest is to check if the thing works on original driver
<Ansuel> maybe qsdk ?
<robimarko> Well, the thing is that we cannot 1:1 reproduce things
<robimarko> And there remoteproc reboot works
<robimarko> But the whole remoteproc subsystem is so patched in their 5.4 kernel that it literaly can be anything
<Ansuel> well still remember a single reg to be set that caused wifi offload not working on ne qsdk version
<Ansuel> considering the tune you can do on thing i bet it's one single option not set
<Ansuel> robimarko do you have link of the patched one
<Ansuel> ?
<robimarko> Its just their 5.4 kernel, everything regarding to remoteproc has a ton of commits on top
<Ansuel> or should i do THE_DIFF TM again (i made an entire .diff of linux_msm kernel version 3.18 compared to upstream version)
<Ansuel> 21mb worth of changes
<robimarko> And then its going to be some unrelated change most likely
<Ansuel> on the big patch i notice some good changes even in the asm code... (spectre patch on krait cpu)
<Ansuel> btw i bet it's some clock or mux that are set by bootloader or even sbl and us resetting it makes it die?
<robimarko> Well, if only our QCA friends actually wrote commit descriptions in a normal way
<robimarko> So we would understand what they are doing
<Ansuel> sir it's Changed-id: 9y7823gfi92t7g10 or nothing.
danitool has joined #openwrt-devel
bbezak has quit [Quit: The Lounge - https://thelounge.chat]
bbezak has joined #openwrt-devel
robimarko has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
fakuivan has joined #openwrt-devel
fakuivan has quit [Remote host closed the connection]
fakuivan has joined #openwrt-devel
Lechu has joined #openwrt-devel
parazyd has quit [Ping timeout: 480 seconds]