01:36 <agentzh> fche: heh

18:23 <irker921> systemtap: alizhang systemtap.git:master * release-4.4-4-g25afbb60a / main.cxx runtime/softfloat.c runtime/softfloat.h runtime/softfloat/internals.h runtime/softfloat/primitives.h runtime/softfloat/specialize.h tapset/floatingpoint.stp testsuite/buildok/floatingpoint.stp: PR13838: Add float32 support and corresponding test cases

18:43 <kerneltoast> fche hiya

18:44 <kerneltoast> i've got something else for you to review: https://gist.github.com/kerneltoast/ad75180b735bf3e733e64a746168fc3a

19:59 <agentzh> fche: this is a deadlock caught by lockdep in our own kernel fuzzer for stap.

20:05 <fche> yo

20:21 <kerneltoast> yo yo

20:23 <kerneltoast> btw fche, just to clarify, that bug wasn't introduced by me :)

20:23 <kerneltoast> it's an old bug

20:23 <kerneltoast> i'm innocent

20:24 <fche> aha yeah, an irqsave/restore 'upgrade' is fine, ship it

20:24 <kerneltoast> noice

20:24 <fche> noice

20:24 <kerneltoast> i've got some print questions for ya too

20:24 <kerneltoast> since `git blame` is useless in this part of stap

20:25 <kerneltoast> i'm not sure why Stp_lbuf exists

20:25 <fche> git blame -C -C -C -C -C -C -M -M -M -M -R -R -R -R -R ftw

20:26 <kerneltoast> it's a dedicated print buffer for _stp_vlog

20:26 <kerneltoast> but then

20:27 <kerneltoast> _stp_vlog dumps its buffer into the main print buffer, Stp_pbuf, via _stp_print()

20:27 <kerneltoast> so why the custom buffer?

20:28 <kerneltoast> additionally, the size of Stp_lbuf can be customized via STP_LOG_BUF_LEN

20:28 <kerneltoast> i don't wanna trash someone's usecase, but at the same time, this doesn't really make sense

20:28 <fche> I'm looking over these old functions

20:28 <kerneltoast> remember back to 2012

20:29 <kerneltoast> when we thought the world was going to end

20:29 <fche> but seem to recall wanting to have a pre-reserved pool of message buffers for use by emergency sorts of messages

20:29 <fche> p.s. the world did end

20:29 <kerneltoast> that's all fine and dandy, but then the emergency buffer depends on the main buffer

20:30 <kerneltoast> so the purpose is defeated :/

20:30 <kerneltoast> the best theory i could come up with was that a big buffer was needed to construct the vlog message and stack space was limited

20:31 <fche> yea not sure, still checking

20:32 <kerneltoast> p.s. making the print stuff robust while supporting v2.6.9 is painful :))))

20:32 <fche> hm no ... I think this part was for something else

20:33 <fche> and probably quite obsolete

20:33 <fche> 2.6.9 is jettisonable, don't worry about it

20:33 <fche> 2.6.32 (rhel6) is the oldest worth considering for us

20:33 <kerneltoast> so you're fine if my patch makes 2.6.9 not compile?

20:33 <fche> yes

20:34 <kerneltoast> wowza

20:34 <fche> (rhel4 era IIRC)

20:34 <kerneltoast> there is still stuff for <2.6.32

20:34 <kerneltoast> but cool

20:34 <kerneltoast> this makes life easier

21:48 <fche> kerneltoast, so I did some more digging w.r.t. Stp-lbuf and suspect that it worked around some old misbehaviour

21:48 <fche> probably before we had the stp_reserve capability, maybe to keep intact any printfs formatted piecemeal

21:59 <kerneltoast> ah ok

21:59 <kerneltoast> so i can nuke it away

21:59 <fche> I think so

22:03 <agentzh> fche: will you please add kerneltoast's ssh key to the official repo? so that i don't need to be the middleman any more ;)

22:03 <fche> sure

22:03 <agentzh> thanks

22:04 <fche> https://sourceware.org/cgi-bin/pdw/ps_form.cgi

22:04 <kerneltoast> https://imgur.com/vScHSRz

22:05 <fche> errr....

22:05 <kerneltoast> JUST KIDDING

22:07 <kerneltoast> submitted

22:08 <fche> ok done

22:09 <kerneltoast> cool

22:12 <kerneltoast> fche, what should i set as my git url?

22:12 <fche> pushurl ssh://sourceware.org/git/systemtap.git

22:13 <irker921> systemtap: sultan systemtap.git:master * release-4.4-5-g374d37118 / runtime/stp_utrace.c: stp_utrace: disable IRQs when holding the bucket spin lock

22:13 <agentzh> fche: i like the "never give up" saying, btw ;)

22:14 <fche> so do Thermians

22:14 <agentzh> heh

22:27 <kerneltoast> no amount of googling is helping me understand what fche said about Thermians

22:29 <fche> https://www.youtube.com/watch?v=SJ2hJezvd2I

22:30 <kerneltoast> that clip felt very familiar

22:31 <kerneltoast> i may have seen that movie

22:31 <fche> highly recommendd

22:31 <kerneltoast> the upside to having a terrible memory is that i get to experience good films for the first time more than once

22:32 <kerneltoast> just gotta wait 5 years and bam, it's all fresh again