<rsalvaterra>
Had to revert it, otherwise I had no connectivity (network unreachable).
<rsalvaterra>
No crashes, no.
<Ansuel>
segfault anything?
<Ansuel>
will check tomorrow thanks for the report
<rsalvaterra>
Sorry, didn't dig too deeply. I didn't see any crashes, but I haven't invoked fw3 manually.
schwicht has joined #openwrt-devel
<slh>
Mangix: it's been a few years since I last (tried to-) built for 4/32 devices, but that's going to bite you big time and TP9343 support afaik isn't quite great either
cmonroe has quit [Read error: Connection timed out]
danitool has quit [Remote host closed the connection]
cmonroe has joined #openwrt-devel
<\x>
anyone who has "APM82181" ? I kinda want to know how it scores on coremark
<stintel>
rsalvaterra: firewall3? what year are you in? :D
<dwfreed>
(hint: the return isn't always malloc'ed)
<dwfreed>
given fw3 is not long-running, it's probably better to just let it leak
tSYS has quit [Quit: *squeak*]
tSYS has joined #openwrt-devel
schwicht has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
dangole has quit [Ping timeout: 480 seconds]
rua has quit [Quit: Leaving.]
valku has quit [Quit: valku]
goliath has joined #openwrt-devel
Tapper has joined #openwrt-devel
<mrkiko>
slh: any chance you may ask for someone in the forum to post high-res pictures of the eMMC part of the nbg7815 ? To diagnose problems with i/o errors
<Ansuel>
rsalvaterra can i use you as a tester for a patch?
<Ansuel>
dwfreed just notice and thanks for the link... what a shit implementation of having a function that can return both malloced or static pointer -.-
Tapper has quit [Ping timeout: 480 seconds]
danitool has joined #openwrt-devel
robimarko has joined #openwrt-devel
<robimarko>
Ansuel: They CC-ed somebody from Micron regarding the eMMC
<Ansuel>
oh! also I notice the commented on my nvmem patch... ahhhhh so much stress
<robimarko>
Ansuel: It was Dmitry, he is totally missing the purpose of that driver
<Ansuel>
let me try to give them reason
<robimarko>
I am planning on replying today, basically its not fun once the vendor replaces the SoC
<robimarko>
Or QCA bins them differently
<robimarko>
BTW, I even have a IPQ6018 that per datasheet runs at 1.8GHz but then this older version is capped at 1.5GHz
<robimarko>
Hence why the need for driver
<robimarko>
IPQ8064 is even messier with the PVS and speedbin mess
<robimarko>
mrkiko: Somebody just replied with a high res photo of eMMC on NBG
<robimarko>
Well, they dont even mention TRIM in the datasheet for it
<robimarko>
Well, they do in the newer datasheet
<robimarko>
And that datasheet is for revision 0x3 aka X03U model
<Ansuel>
anyway guess we will have to change it to enum to make dmitry happy
<robimarko>
That is basically just a cosmetic change, nothing else.
<Ansuel>
yep and considering it's really 2 define... IMHO it would also look ugly as they are bits
<\x>
robimarko: your MR7350 also stuck at 1.5GHz on the cpu?
<robimarko>
\x: Yes
<robimarko>
And rigthly so as speedbin fuse also indicates its limited to 1.5GHz
<\x>
its been like that whatever firmware i install
<\x>
i thought it was a bug like a power thing wont let it do it
<robimarko>
\x: Yes, its either older SKU or they intentionally bought the lower binned ones
<Ansuel>
fuse don't lie
<\x>
so i guess theres no more bugs per se
<Ansuel>
probably it's a defective chip and using 1.8 would result in the thing explode
<\x>
atleast usability wise
<Ansuel>
pretty common in modern cpu (disabling stuff due to silicon defects)
<\x>
wifi on/off works with 2.4 and 2.5 firmware on kirdesde's github
<robimarko>
Ansuel: Nope, even stock FW is exposing all OPP-s to 1.8GHz, its just that even forcing 1.8GHz wont work
<\x>
usb works
<robimarko>
It will just run at 1.5GHZ
<robimarko>
\x: Well, remoteproc craps out for me
<\x>
what firmware?
<robimarko>
As soon as it loads BDF and caldata it just resets the board
<robimarko>
2.4 to 2.7
<\x>
I tried 2.4, 2.5, 2.6 and 2.7
<\x>
2.4 and 2.5 works nice
<\x>
the bdf I use is the FCC one
<\x>
from latest firmware
<\x>
from latest vendor* firmware
<robimarko>
However you put it, its not supposed to happen
<robimarko>
99% its the shit remoteproc implementation we currently have
<robimarko>
I am tempted to nuke the current implementation with the secure PIL
<robimarko>
And just port the downstream QCA nonsecure one
<Ansuel>
is that a lot of code?
<robimarko>
QCA "pinky" promised to add IPQ8074 and IPQ6018 to their new remoteproc driver after IPQ9574 and IPQ5018 are merged
<Ansuel>
i bet it's the secure PIl that makes the soc crash
<robimarko>
Ansuel: Nope, its just manually writing registers rather than holding the breath after an SCM call
<robimarko>
Cause, IPQ6018 TZ has bugs
<Ansuel>
it really sounds like illegal memory access from secure zone and TZ just killing the soc
<Ansuel>
read some practice like that on some phone soc
<robimarko>
Like, I have a hack to not advertise OSI mode in PSCI cause stupid TZ is advertising it but until a certain version it will just make your board hang
<robimarko>
BTW, they are really upstreaming IPQ9574 so far
<\x>
if only I didnot get 300/300 from isp i would be dropping my ipq60xx in favor of my ipq40xx + a wifi6 card on its mini pcie, but then ipq40xx cant cake 300/300 nicely
<robimarko>
So, if anybody knows any IPQ9574 HW that is available for a reasonable price I am all ears
<robimarko>
That is my on my TODO after ipq60xx
<\x>
isnt there only two?
<\x>
xiaomi 10000 and 7000?
<Ansuel>
ipq50xx no ?
<\x>
on wifi 7 news though
<robimarko>
Well, lets say that I would prefer to avoid XIaomi
<linusw>
right now for a newcomer it's a bit ... OK what syntax is this now?
rua has joined #openwrt-devel
<PaulFertser>
linusw: so where exactly in there did you plan to put this reference?
<linusw>
PaulFertser: somewhere on the top? I might be odd but I would expect it to be one of the first things.
<PaulFertser>
linusw: I'd put it on the bottom, near the sidebar note. As it would look inconsistent with the current document structure otherwise I'm afraid.
<mrkiko>
robimarko: great to know you got the hi-res pictures
dangole has joined #openwrt-devel
rua has quit [Ping timeout: 480 seconds]
<linusw>
PaulFertser: OK I trust your judgement :)
<mrkiko>
robimarko: what are the informations you needed that are present in the photos but not exported by the kernel? I guess they're not exportable in any way
<mrkiko>
linusw: great work with the bcm northstar SOC
<mrkiko>
linusw: thanks
<linusw>
mrkiko: thanks! yeah it was a fun thing to fix up, I don't know if the patches will go in soon I have no idea who actually merges it, maybe Christian or Rafal.
<\x>
ah yeah, 512M memory profile still not working for 60xx robimarko
<Ansuel>
\x ... what....
<\x>
yeah
<Ansuel>
so what is the profile to use 1g ?
<\x>
we run 1GB profile on 512MB devices
<Ansuel>
on a low spec soc
<Ansuel>
very nice... but are we sure it's not a missing piece in ath11k ?
<\x>
its fine
<\x>
but yeah its like 384 vs 408MB usable
<Ansuel>
afaik we have a reduced patch for the memory profile so i wonder if with 512 profile applied they also now require the reduced ring setup
<Ansuel>
we will shortly start to upload ipq60xx file to the bdf repository
<Ansuel>
and update the regdb in it
<robimarko>
Ansuel: Its crashing even with FW mode 0 aka the default
<robimarko>
So its lower level than that
<Ansuel>
mhhh worth to check what magic that guy is doing?
<Ansuel>
probably we are missing a patch?
<robimarko>
I did
<robimarko>
He backported the QCA non-secure changes for remoteproc as well
<Ansuel>
well we have our answer...
<Ansuel>
as always secure zone on ipq is implemented like shit
<\x>
Ansuel: I have the dump of MR7350's vendor firmware, theres a lot of bdwlan inside, i tested all of them and FCC is the best for me, do you want the dumps?
<Ansuel>
we should not use the bed one inside but the one tweaked by the OEM
<PaulFertser>
linusw: ok, I made "wiki syntax" a hyper link https://openwrt.org/wiki/wikirules#structure and also I noticed earlier "Wiki pages" entry references a page that describes everything about the wiki, including the syntax.
floof58 has quit [Ping timeout: 480 seconds]
floof58 has joined #openwrt-devel
_mfjlo54ys has joined #openwrt-devel
<_mfjlo54ys>
hi, i got this error. i think a file did not download right because of a internet glitch, How can i patch right, quick andeasy before deleting the whole folder and re-compiling from scratch? patching file drivers/usb/chipidea/ci.h es). patching file drivers/usb/chipidea/otg.c patching file include/linux/usb/chipidea.h Patch failed! Please fix /home/**/target/linux/ar71xx/patches-4.14/930-chipidea-pullup.patch
<_mfjlo54ys>
nvm
_mfjlo54ys has quit [Quit: Page closed]
<Ansuel>
aparcar ok to merge the ccache pr?
<schmars[m]>
blocktrron: hey hey, we briefly chatted about USW-Flex and 60 Ghz devices at WCW. You mentioned the AF60's wifi card, i think it's a Peraso Falcon yeah?
<blocktrron>
schmars[m]: i thought it was a wilocity
<schmars[m]>
ah yeah thanks for the name wilocity! my AF60-LR strongly looks like Peraso judging from dmesg and available commands. it's very possible they changed chips between AF60 iterations and models
lucenera has joined #openwrt-devel
<schmars[m]>
ooh and WIlocity being acquired by QCA in 2014, it makes sense they would try to strengthen alternative vendors
<aparcar[m]>
Ansuel: works for me
<aparcar[m]>
can you add the secrets?
<Ansuel>
yep
<schmars[m]>
the downside seems to be that Peraso Falcon looks like a hard no at the moment regarding open linux drivers
<Ansuel>
already done and limited to openwrt repository usage
<aparcar[m]>
let's see how it goes
<aparcar[m]>
as I said before, I can also ask a second s3 provider
<robimarko>
dangole: Does BPI-R3 BL2 work if inside the GPT?
<robimarko>
Cause, I put BL2 at 17k but it doesnt boot
<Mangix>
slh: I mean... I assume 18.06 works
<Mangix>
4.9 is a higher version than dd-wrt's 3.18
<dangole>
robimarko: ... and you are on an SD card which works with the OpenWrt sdcard image for the R3?
<dangole>
robimarko: because some cards don't work with the bootrom at all for some unknown reason, others have problems with the minimal MMC driver in U-Boot, all of them work fine in Linux...
<robimarko>
dangole: Yeah, it boots OpenWrt
<robimarko>
Here, I just get the bootrom print and a halt
<dangole>
robimarko: that's odd then. are you using 'mkimage -T mtk_image' or mediatek's proprietary 'bromimage' tool to create bl2 image?
<dangole>
asking because on the R3 I've never tried bromimage, just went with U-Boot's mkimage from the beginning and that worked for me
<dangole>
but mtk's tf-a sources use bromimage by default afaik
<robimarko>
dangole: Well, I tried mimicing OpenWrt and am just passing USE_MKIMAGE=1
<dangole>
ok, that should be the same then. I guess the next thing I'd do would be compare the hexdump of both sdcard images and look for significant differences
<robimarko>
Il try using U-boot BL2 and packaging that
<robimarko>
*OpenWRt
floof58 has quit [Ping timeout: 480 seconds]
<robimarko>
Yeah, its my BL2 builds fault
floof58 has joined #openwrt-devel
schwicht has joined #openwrt-devel
<robimarko>
Ansuel: Oh, you merged the ccache
<robimarko>
Lets see how that works out
<Ansuel>
Yep now long wait for everything to build and refresh cache
<robimarko>
BTW, how could I trigger a refresh for all target patches?
<robimarko>
I want to update 6.1 to .31
rua has joined #openwrt-devel
<mrkiko>
robimarko: make target/linux/refresh
<Ansuel>
think he was referring to something for CI
<mrkiko>
robimarko: I don't know if it works for a specific target only as well
<mrkiko>
Ansuel: ah, ok; sorry then
<tmn505>
robimarko: in maintainer-tools.git there's a update_kernel.sh script. Personaly never used it as my machines are slow to bump all targets (no space on HDD also)
<Ansuel>
why perl is so difficult to understand...
<robimarko>
Perfect, I need something to refresh all of the targets running 6.1
<robimarko>
That was my question, I have used make target/linux/refresh plenty of times
<Ansuel>
it would be at least useful to know what targets are with 6.1
<Ansuel>
but dump-targets-info doesn't support that
<robimarko>
ipq806x, ipq807x, malta and bmips for sure
<robimarko>
Just searching for the test version is fine
<mrkiko>
apm821xx
<mrkiko>
uml armvirt rockchip
<Ansuel>
robimarko my idea is to extend the label CI so we can test testing thing on demand
<robimarko>
Ansuel: Yeah, that would be neat
<robimarko>
Well, guess who used bl2.bin instead of bl2.img
<robimarko>
Now bl2 works, but it doesnt find the U-boot FIP
<mrkiko>
robimarko: did you add my eMMC to quirks list in the end?
<robimarko>
mrkiko: No, I am waiting to see what upstream has to say about this one
<robimarko>
Maybe there is a better solution than simply disabling TRIM
<robimarko>
If they merge it, I will add that one as well and make a PR
<mrkiko>
thanks. Was curious about what you needed from the photos you couldn't see from the kernel
<robimarko>
Kernel doesnt know the exact eMMC model
<robimarko>
It just has the 8 character string
rua has quit [Quit: Leaving.]
<mrkiko>
robimarko: in any case, feel free to ping me and/or pm in case, so I can test things out in the device
<robimarko>
Will do if I need something
danitool has quit [Quit: Cubum autem in duos cubos, aut quadratoquadratum in duos quadratoquadratos]
<hurricos>
@blocktrron: Are you willing to share either a compiled u-boot image, or even the compilation source for the MSM560? I'm willing to accept "No", I don't enjoy u-boot replacements either.
<Ansuel>
thanks for linking it i was searching it from a bit..
<mrkiko>
:)
<Ansuel>
i will leave some message but in the current state is a NACK for me...
<Ansuel>
it's a big hack so the only way to make it less of an hack is to use compatible instead of a binding
fakuivan has quit [Remote host closed the connection]
<mrkiko>
Ansuel: well, on the DTS part it may be agreeable; for the rest, based on what I got from the ML, it looked like there was no other solution rather than change the PCI id
fakuivan has joined #openwrt-devel
<Ansuel>
i don't really like adding all the stuff to the struct
<Ansuel>
it's an hack anyway... lets bind everything to a compatible
<Ansuel>
there is an idea to make that in a upstreamable state while that can't happen like ever.
finger has joined #openwrt-devel
minimal has joined #openwrt-devel
finger has left #openwrt-devel [#openwrt-devel]
finger has joined #openwrt-devel
<robimarko>
That is a complete hack and will have to be carried forever
<Ansuel>
yep a from a porting point it's better to have simple if condition than putting additional function in struct
<mrkiko>
barhom: try to ask in #openwrt, you might get some answers there
shibboleth has joined #openwrt-devel
_mfjlo54ys has joined #openwrt-devel
<_mfjlo54ys>
hey this error message is common...what's the command to fix this compile error again? make[1]: *** [package/Makefile:107: staging_dir/target-mips_24kc_musl/stamp/.package_compile] Error 2 make[1]: Leaving directory ''
<_mfjlo54ys>
Is it: ./scripts/feeds install -a after a failed compile?
noltari has quit [Quit: Bye ~ Happy Hacking!]
<_mfjlo54ys>
some mssing library pr else.. i was outside, got back and thought iu would have a compile.. gonna fix an dget out
<_mfjlo54ys>
cant be inside now
<_mfjlo54ys>
so a quick fixing command would be cool thanks
<_mfjlo54ys>
gonna try that in the meanwhile..
<f00b4r0>
just stumbled upon #3818: it can be closed, the device support has already been merged through a different PR
<wigyori>
hauke: can't see any objections to it - if it's still okay to add new target into it
<wigyori>
hauke: btw, if you want to take a look at it, i've prepped a common target for the jh7100/jh7110 socs - https://git.openwrt.org/?p=openwrt/staging/wigyori.git;a=shortlog;h=refs/heads/riscv-vf-202305
<wigyori>
needs some polishing for usb on both boards, but it's a good start on 6.1 with patches backported from upstream kernels