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
djrscally has quit [Ping timeout: 256 seconds]
Pali has joined #armlinux
Tokamak has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
apritzel_ has quit [Ping timeout: 256 seconds]
Pali has quit [Read error: Connection reset by peer]
Pali has joined #armlinux
Pali has quit [Ping timeout: 256 seconds]
darkapex has quit [Quit: ZNC 1.7.2+deb3 - https://znc.in]
<javierm>
broonie: even if isn't finished, the changes for each driver has merits on its own. That is, adding an OF and SPI device ID tables to drivers that are missing is always a good thing
<broonie>
javierm: If someone wants to mentor it I guess.
<javierm>
broonie: I'm happy to apply as a mentor. Will figure out what's the process to do that
<broonie>
I've got patches for adding in the ID tables that are missing already, it's mostly just a question of getting people to actually respond to patches sent to them.
<javierm>
broonie: cool, so that stewardship could be done by a mentee to make sure that things don't fall into the cracks
<broonie>
I left it for a while since the one person really bothered about the non-OF aliases was being a bit demotivating.
<broonie>
Who knows, by the time outreachy gets round it it it might be done?
<javierm>
broonie: yeah, that's true too
<broonie>
and it might be a bit small a project as well? Dunno.
<javierm>
not really familiar of the size and scope of the outreachy projects tbh. Your drm patches just reminded me of this issue
<javierm>
broonie: btw, I'll push those to the drm-misc tree in a week or so. Just in case someone else wants to comment on them
Amit_T has joined #armlinux
snalty_ has quit [Remote host closed the connection]
snalty has joined #armlinux
snalty has quit [Client Quit]
snalty has joined #armlinux
<arnd>
mrutland: I'm trying out a way to optionally move linux/atomic.h into lib/atomic.c with only extern declarations left in the header, as linux/atomic.h is now the single largest source of indirect includes in mingo's fast-header tree
<arnd>
for the most part this is really easy thanks to your scripts to generate the headers, I just add a couple more scripts along the same lines
<arnd>
but I think I just came across a bug in the 32-bit arm version which does not have a custom arch_atomic64_read_acquire() and arch_atomic64_set_release() implementation
<arnd>
include/linux/atomic/atomic-arch-fallback.h falls back to smp_load_acquire() and smp_store_release() instead, but those are not allowed on 64-bit pointers
<mrutland>
arnd: oh no!
<arnd>
I can make it build with this two-liner, but I don't know if that is safe:
<mrutland>
arnd: that'll need barriers, since the arm implementation of arch_atomic64_read and arch_atomic64_set are weakly ordered
<mrutland>
... lemme just dig a bit, because we should be able to build that in the wrappers with the usual barriers there if we don't use smp_load_acquire() and smp_store_release() as the fallbacks
<arnd>
at least I'm fairly sure this has not caused an actual bug with atomicity, as any attempt to use these functions would result in "include/linux/compiler_types.h:342:45: error: call to '__compiletime_assert_8' declared with attribute error: Need native word sized stores/loads for atomicity."
<mrutland>
That's good to hear!
<mrutland>
I reckon the right thing to do is to alter the fallbacks to only use the smp_*() helpers when the tpye is __native_word(), and do use the regulare acquire/release fallbacks otherwise ; lemme just play about with that
<mrutland>
I suspect that just means a little duplication for now
apritzel_ has joined #armlinux
<arnd>
mrutland: would it be easier to provide a correct arch_atomic64_read_acquire()/arch_atomic64_set_release() implementation for each 32-bit architecture that implements atomic64_t? I think there are only three implementations aside from asm-generic: arm, i386 and arcv2
<mrutland>
arnd: maybe, but it looks like the scripting change might be trivial -- e.g. half-written https://paste.debian.net/1229465/
<mrutland>
(assuming that if-else doesn't need ot be CPP #if #else #endif)
apritzel_ has quit [Ping timeout: 256 seconds]
<mrutland>
Right, that seems to work for me, having hacked a test into arch/arm/kernel/setup.o
<mrutland>
I'll write up a commit message and send that out shortly, with you Cc'd and given a Reported-by
<mrutland>
Thansk for the report, and sorry for the bug! :)
<ardb>
mrutland: must be missing something but why isn't that code using arch_atomic64_set() et al ?
jlinton has joined #armlinux
suzukikp has joined #armlinux
suzukikp1 has joined #armlinux
<mrutland>
ardb: you mean why does it use smp_store_release() today?
<ardb>
mrutland: no why does arch_atomic64_read_acquire() fallback arch_atomic_read() rather than arch_atomic64_read()?
<ardb>
fallback *to*
<mrutland>
beczuse I made a typo :/
<mrutland>
that should be using arch_${atomic}_read and arch_${atomic}_set
<mrutland>
Do you want to reply so on list, or should I just self-reply that you pointed that out here?
<ardb>
i'm not cc'ed to linux-kernel so please reply yourself
<ardb>
not subscribed
<ardb>
jeez
* ukleinek
hands ardb the manpage of lei
<ardb>
ta
<mrutland>
ardb: will do, thanks!
<mrutland>
I'm confused how it generated ldrd/strd now
<mrutland>
Ah, because the 32-bit arch_atomic_read() and arch_atomic_set are macros wrapping READ_ONCE()/WRITE_ONCE()
<ardb>
right
<ardb>
mrutland: that's fine for LPAE
Tokamak has joined #armlinux
<mrutland>
ardb: yup -- I just tested the fixed version with multi_v7_defconfig (!LPAE) and keystone_defconfig (LPAE), and it uses exclusives for the former, LDRD/STRD for the latter
<ardb>
excellent
<ardb>
mrutland: strange though that this wasn't caught by any type checks
<mrutland>
ardb: the problem is that arm's arch_atomic_{read,set}() are macros, as are READ_ONCE/WRITE_ONCE(), so there are no types to check
<mrutland>
Likewise for arm64
<ardb>
right
torez has quit [Ping timeout: 240 seconds]
System_Error has quit [Ping timeout: 276 seconds]
torez has joined #armlinux
System_Error has joined #armlinux
System_Error has quit [Remote host closed the connection]
System_Error has joined #armlinux
frieder has quit [Remote host closed the connection]
headless has joined #armlinux
alexels has quit [Quit: WeeChat 3.4]
apritzel has quit [Ping timeout: 256 seconds]
russ has quit [Ping timeout: 250 seconds]
russ has joined #armlinux
suzpou01__ has joined #armlinux
prabhakarlad has joined #armlinux
suzukikp1 has quit [Ping timeout: 256 seconds]
suzukikp has quit [Ping timeout: 256 seconds]
suzukikp has joined #armlinux
<sudeepholla>
ardb, sorry checking irc now
<sudeepholla>
ardb, for cpuidle to work on ACPI, you need _LPI methods in CPU device
<sudeepholla>
ardb, I think upstream edk2-platforms has it for Juno(I vaguely remember adding one)
<sudeepholla>
arnd, hope that helps, let me know if you need any more info
<sudeepholla>
arnd, sorry not for you :(
jlinton has quit [Ping timeout: 256 seconds]
Amit_T has quit [Remote host closed the connection]
<ardb>
sudeepholla: thanks!
Tokamak has quit [Ping timeout: 256 seconds]
Tokamak has joined #armlinux
prabhakarlad has quit [Ping timeout: 256 seconds]
monstr has quit [Remote host closed the connection]
guillaume_g has quit [Quit: Konversation terminated!]