cthreads -> pthreads
Get rid of cthreads; switch to pthreads. Most of the issues raised on this page has been resolved, a few remain.
IRC, freenode, #hurd, 2012-04-26
<pinotree> youpi: just to be sure: even if libpthread is compiled inside
glibc (with proper symbols forwarding etc), it doesn't change that you
cannot use both cthreads and pthreads in the same app, right?
<youpi> it's the same libpthread
<youpi> symbol forwarding does not magically resolve that libpthread lacks
some libthread features :)
<pinotree> i know, i was referring about the clash between actively using
both
<youpi> there'll still be the issue that only one will be initialized
<youpi> and one that provides libc thread safety functions, etc.
<pinotree> that's what i wanted to knew, thanks :)
IRC, freenode, #hurd, 2012-07-23
<bddebian> So I am not sure what to do with the hurd_condition_wait stuff
<braunr> i would also like to know what's the real issue with cancellation
here
<braunr> because my understanding is that libpthread already implements it
<braunr> does it look ok to you to make hurd_condition_timedwait return an
errno code (like ETIMEDOUT and ECANCELED) ?
<youpi> braunr: that's what pthread_* function usually do, yes
<braunr> i thought they used their own code
<youpi> no
<braunr> thanks
<braunr> well, first, do you understand what hurd_condition_wait is ?
<braunr> it's similar to condition_wait or pthread_cond_wait with a subtle
difference
<braunr> it differs from the original cthreads version by handling
cancellation
<braunr> but it also differs from the second by how it handles cancellation
<braunr> instead of calling registered cleanup routines and leaving, it
returns an error code
<braunr> (well simply !0 in this case)
<braunr> so there are two ways
<braunr> first, change the call to pthread_cond_wait
<bddebian> Are you saying we could fix stuff to use pthread_cond_wait()
properly?
<braunr> it's possible but not easy
<braunr> because you'd have to rewrite the cancellation code
<braunr> probably writing cleanup routines
<braunr> this can be hard and error prone
<braunr> and is useless if the code already exists
<braunr> so it seems reasonable to keep this hurd extension
<braunr> but now, as it *is* a hurd extension noone else uses
<antrik> braunr: BTW, when trying to figure out a tricky problem with the
auth server, cfhammer digged into the RPC cancellation code quite a bit,
and it's really a horrible complex monstrosity... plus the whole concept
is actually broken in some regards I think -- though I don't remember the
details
<braunr> antrik: i had the same kind of thoughts
<braunr> antrik: the hurd or pthreads ones ?
<antrik> not sure what you mean. I mean the RPC cancellation code -- which
is involves thread management too
<braunr> ok
<antrik> I don't know how it is related to hurd_condition_wait though
<braunr> well i found two main entry points there
<braunr> hurd_thread_cancel and hurd_condition_wait
<braunr> and it didn't look that bad
<braunr> whereas in the pthreads code, there are many corner cases
<braunr> and even the standard itself looks insane
<antrik> well, perhaps the threading part is not that bad...
<antrik> it's not where we saw the problems at any rate :-)
<braunr> rpc interruption maybe ?
<antrik> oh, right... interruption is probably the right term
<braunr> yes that thing looks scary
<braunr> :))
<braunr> the migration thread paper mentions some things about the problems
concerning threads controllability
<antrik> I believe it's a very strong example for why building around
standard Mach features is a bad idea, instead of adapting the primitives
to our actual needs...
<braunr> i wouldn't be surprised if the "monstrosities" are work arounds
<braunr> right
IRC, freenode, #hurd, 2012-07-26
<bddebian> Uhm, where does /usr/include/hurd/signal.h come from?
<pinotree> head -n4 /usr/include/hurd/signal.
<pinotree> h
<bddebian> Ohh glibc?
<bddebian> That makes things a little more difficult :(
<braunr> why ?
<bddebian> Hurd includes it which brings in cthreads
<braunr> ?
<braunr> the hurd already brings in cthreads
<braunr> i don't see what you mean
<bddebian> Not anymore :)
<braunr> the system cthreads header ?
<braunr> well it's not that difficult to trick the compiler not to include
them
<bddebian> signal.h includes cthreads.h I need to stop that
<braunr> just define the _CTHREADS_ macro before including anything
<braunr> remember that header files are normally enclosed in such macros to
avoid multiple inclusions
<braunr> this isn't specific to cthreads
<pinotree> converting hurd from cthreads to pthreads will make hurd and
glibc break source and binary compatibility
<bddebian> Of course
<braunr> reminds me of the similar issues of the late 90s
<bddebian> Ugh, why is he using _pthread_self()?
<pinotree> maybe because it accesses to the internals
<braunr> "he" ?
<bddebian> Thomas in his modified cancel-cond.c
<braunr> well, you need the internals to implement it
<braunr> hurd_condition_wait is similar to pthread_condition_wait, except
that instead of stopping the thread and calling cleanup routines, it
returns 1 if cancelled
<pinotree> not that i looked at it, but there's really no way to implement
it using public api?
<bddebian> Even if I am using glibc pthreads?
<braunr> unlikely
<bddebian> God I had all of this worked out before I dropped off for a
couple years.. :(
<braunr> this will come back :p
<pinotree> that makes you the perfect guy to work on it ;)
<bddebian> I can't find a pt-internal.h anywhere.. :(
<pinotree> clone the hurd/libpthread.git repo from savannah
<bddebian> Of course when I was doing this libpthread was still in hurd
sources...
<bddebian> So if I am using glibc pthread, why can't I use pthread_self()
instead?
<pinotree> that won't give you access to the internals
<bddebian> OK, dumb question time. What internals?
<pinotree> the libpthread ones
<braunr> that's where you will find if your thread has been cancelled or
not
<bddebian> pinotree: But isn't that assuming that I am using hurd's
libpthread?
<pinotree> if you aren't inside libpthread, no
<braunr> pthread_self is normally not portable
<braunr> you can only use it with pthread_equal
<braunr> so unless you *know* the internals, you can't use it
<braunr> and you won't be able to do much
<braunr> so, as it was done with cthreads, hurd_condition_wait should be
close to the libpthread implementation
<braunr> inside, normally
<braunr> now, if it's too long for you (i assume you don't want to build
glibc)
<braunr> you can just implement it outside, grabbing the internal headers
for now
<pinotree> another "not that i looked at it" question: isn't there no way
to rewrite the code using that custom condwait stuff to use the standard
libpthread one?
<braunr> and once it works, it'll get integrated
<braunr> pinotree: it looks very hard
<bddebian> braunr: But the internal headers are assuming hurd libpthread
which isn't in the source anymore
<braunr> from what i could see while working on select, servers very often
call hurd_condition_wait
<braunr> and they return EINTR if canceleld
<braunr> so if you use the standard pthread_cond_wait function, your thread
won't be able to return anything, unless you push the reply in a
completely separate callback
<braunr> i'm not sure how well mig can cope with that
<braunr> i'd say it can't :)
<braunr> no really it looks ugly
<braunr> it's far better to have this hurd specific function and keep the
existing user code as it is
<braunr> bddebian: you don't need the implementation, only the headers
<braunr> the thread, cond, mutex structures mostly
<bddebian> I should turn <pt-internal.h> to "pt-internal.h" and just put it
in libshouldbelibc, no?
<pinotree> no, that header is not installed
<bddebian> Obviously not the "best" way
<bddebian> pinotree: ??
<braunr> pinotree: what does it change ?
<pinotree> braunr: it == ?
<braunr> bddebian: you could even copy it entirely in your new
cancel-cond.C and mention where it was copied from
<braunr> pinotree: it == pt-internal.H not being installed
<pinotree> that he cannot include it in libshouldbelibc sources?
<pinotree> ah, he wants to copy it?
<braunr> yes
<braunr> i want him to copy it actually :p
<braunr> it may be hard if there are a lot of macro options
<pinotree> the __pthread struct changes size and content depending on other
internal sysdeps headers
<braunr> well he needs to copy those too :p
<bddebian> Well even if this works we are going to have to do something
more "correct" about hurd_condition_wait. Maybe even putting it in
glibc?
<braunr> sure
<braunr> but again, don't waste time on this for now
<braunr> make it *work*, then it'll get integrated
<bddebian> Like it has already? This "patch" is only about 5 years old
now... ;-P
<braunr> but is it complete ?
<bddebian> Probably not :)
<bddebian> Hmm, I wonder how many undefined references I am going to get
though.. :(
<bddebian> Shit, 5
<bddebian> One of which is ___pthread_self.. :(
<bddebian> Does that mean I am actually going to have to build hurds
libpthreads in libshouldbeinlibc?
<bddebian> Seriously, do I really need ___pthread_self, __pthread_self,
_pthread_self and pthread_self???
<bddebian> I'm still unclear what to do with cancel-cond.c. It seems to me
that if I leave it the way it is currently I am going to have to either
re-add libpthreads or still all of the libpthreads code under
libshouldbeinlibc.
<braunr> then add it in libc
<braunr> glib
<braunr> glibc
<braunr> maybe under the name __hurd_condition_wait
<bddebian> Shouldn't I be able to interrupt cancel-cond stuff to use glibc
pthreads?
<braunr> interrupt ?
<bddebian> Meaning interject like they are doing. I may be missing the
point but they are just obfuscating libpthreads thread with some other
"namespace"? (I know my terminology is wrong, sorry).
<braunr> they ?
<bddebian> Well Thomas in this case but even in the old cthreads code,
whoever wrote cancel-cond.c
<braunr> but they use internal thread structures ..
<bddebian> Understood but at some level they are still just getting to a
libpthread thread, no?
<braunr> absolutely not ..
<braunr> there is *no* pthread stuff in the hurd
<braunr> that's the problem :p
<bddebian> Bah damnit...
<braunr> cthreads are directly implement on top of mach threads
<braunr> implemeneted*
<braunr> implemented*
<bddebian> Sure but hurd_condition_wait wasn't
<braunr> of course it is
<braunr> it's almost the same as condition_wait
<braunr> but returns 1 if a cancelation request was made
<bddebian> Grr, maybe I am just confusing myself because I am looking at
the modified (pthreads) version instead of the original cthreads version
of cancel-cond.c
<braunr> well if the modified version is fine, why not directly use that ?
<braunr> normally, hurd_condition_wait should sit next to other pthread
internal stuff
<braunr> it could be renamed __hurd_condition_wait, i'm not sure
<braunr> that's irrelevant for your work anyway
<bddebian> I am using it but it relies on libpthread and I am trying to use
glibc pthreads
<braunr> hum
<braunr> what's the difference between libpthread and "glibc pthreads" ?
<braunr> aren't glibc pthreads the merged libpthread ?
<bddebian> quite possibly but then I am missing something obvious. I'm
getting ___pthread_self in libshouldbeinlibc but it is *UND*
<braunr> bddebian: with unmodified binaries ?
<bddebian> braunr: No I added cancel-cond.c to libshouldbeinlibc
<bddebian> And some of the pt-xxx.h headers
<braunr> well it's normal then
<braunr> i suppose
<bddebian> braunr: So how do I get those defined without including
pthreads.c from libpthreads? :)
<antrik> pinotree: hm... I think we should try to make sure glibc works
both whith cthreads hurd and pthreads hurd. I hope that shoudn't be so
hard.
<antrik> breaking binary compatibility for the Hurd libs is not too
terrible I'd say -- as much as I'd like that, we do not exactly have a
lot of external stuff depending on them :-)
<braunr> bddebian: *sigh*
<braunr> bddebian: just add cancel-cond to glibc, near the pthread code :p
<bddebian> braunr: Wouldn't I still have the same issue?
<braunr> bddebian: what issue ?
<antrik> is hurd_condition_wait() the name of the original cthreads-based
function?
<braunr> antrik: the original is condition_wait
<antrik> I'm confused
<antrik> is condition_wait() a standard cthreads function, or a
Hurd-specific extension?
<braunr> antrik: as standard as you can get for something like cthreads
<bddebian> braunr: Where hurd_condition_wait is looking for "internals" as
you call them. I.E. there is no __pthread_self() in glibc pthreads :)
<braunr> hurd_condition_wait is the hurd-specific addition for cancelation
<braunr> bddebian: who cares ?
<braunr> bddebian: there is a pthread structure, and conditions, and
mutexes
<braunr> you need those definitions
<braunr> so you either import them in the hurd
<antrik> braunr: so hurd_condition_wait() *is* also used in the original
cthread-based implementation?
<braunr> or you write your code directly where they're available
<braunr> antrik: what do you call "original" ?
<antrik> not transitioned to pthreads
<braunr> ok, let's simply call that cthreads
<braunr> yes, it's used by every hurd servers
<braunr> virtually
<braunr> if not really everyone of them
<bddebian> braunr: That is where you are losing me. If I can just use
glibc pthreads structures, why can't I just use them in the new pthreads
version of cancel-cond.c which is what I was originally asking.. :)
<braunr> you *have* to do that
<braunr> but then, you have to build the whole glibc
* bddebian shoots himself
<braunr> and i was under the impression you wanted to avoid that
<antrik> do any standard pthread functions use identical names to any
standard cthread functions?
<braunr> what you *can't* do is use the standard pthreads interface
<braunr> no, not identical
<braunr> but very close
<braunr> bddebian: there is a difference between using pthreads, which
means using the standard posix interface, and using the glibc pthreads
structure, which means toying with the internale implementation
<braunr> you *cannot* implement hurd_condition_wait with the standard posix
interface, you need to use the internal structures
<braunr> hurd_condition_wait is actually a shurd specific addition to the
threading library
<braunr> hurd*
<antrik> well, in that case, the new pthread-based variant of
hurd_condition_wait() should also use a different name from the
cthread-based one
<braunr> so it's normal to put it in that threading library, like it was
done for cthreads
<braunr> 21:35 < braunr> it could be renamed __hurd_condition_wait, i'm not
sure
<bddebian> Except that I am trying to avoid using that threading library
<braunr> what ?
<bddebian> If I am understanding you correctly it is an extention to the
hurd specific libpthreads?
<braunr> to the threading library, whichever it is
<braunr> antrik: although, why not keeping the same name ?
<antrik> braunr: I don't think having hurd_condition_wait() for the cthread
variant and __hurd_condition_wait() would exactly help clarity...
<antrik> I was talking about a really new name. something like
pthread_hurd_condition_wait() or so
<antrik> braunr: to avoid confusion. to avoid accidentally pulling in the
wrong one at build and/or runtime.
<antrik> to avoid possible namespace conflicts
<braunr> ok
<braunr> well yes, makes sense
<bddebian> braunr: Let me state this as plainly as I hope I can. If I want
to use glibc's pthreads, I have no choice but to add it to glibc?
<braunr> and pthread_hurd_condition_wait is a fine name
<braunr> bddebian: no
<braunr> bddebian: you either add it there
<braunr> bddebian: or you copy the headers defining the internal structures
somewhere else and implement it there
<braunr> but adding it to glibc is better
<braunr> it's just longer in the beginning, and now i'm working on it, i'm
really not sure
<braunr> add it to glibc directly :p
<bddebian> That's what I am trying to do but the headers use pthread
specific stuff would should be coming from glibc's pthreads
<braunr> yes
<braunr> well it's not the headers you need
<braunr> you need the internal structure definitions
<braunr> sometimes they're in c files for opacity
<bddebian> So ___pthread_self() should eventually be an obfuscation of
glibcs pthread_self(), no?
<braunr> i don't know what it is
<braunr> read the cthreads variant of hurd_condition_wait, understand it,
do the same for pthreads
<braunr> it's easy :p
<bddebian> For you bastards that have a clue!! ;-P
<antrik> I definitely vote for adding it to the hurd pthreads
implementation in glibc right away. trying to do it externally only adds
unnecessary complications
<antrik> and we seem to agree that this new pthread function should be
named pthread_hurd_condition_wait(), not just hurd_condition_wait() :-)
IRC, freenode, #hurd, 2012-07-27
<bddebian> OK this hurd_condition_wait stuff is getting ridiculous the way
I am trying to tackle it. :( I think I need a new tactic.
<braunr> bddebian: what do you mean ?
<bddebian> braunr: I know I am thick headed but I still don't get why I
cannot implement it in libshouldbeinlibc for now but still use glibc
pthreads internals
<bddebian> I thought I was getting close last night by bringing in all of
the hurd pthread headers and .c files but it just keeps getting uglier
and uglier
<bddebian> youpi: Just to verify. The /usr/lib/i386-gnu/libpthread.so that
ships with Debian now is from glibc, NOT libpthreads from Hurd right?
Everything I need should be available in glibc's libpthreads? (Except for
hurd_condition_wait obviously).
<braunr> 22:35 < antrik> I definitely vote for adding it to the hurd
pthreads implementation in glibc right away. trying to do it externally
only adds unnecessary complications
<youpi> bddebian: yes
<youpi> same as antrik
<bddebian> fuck
<youpi> libpthread *already* provides some odd symbols (cthread
compatibility), it can provide others
<braunr> bddebian: don't curse :p it will be easier in the long run
* bddebian breaks out glibc :(
<braunr> but you should tell thomas that too
<bddebian> braunr: I know it just adds a level of complexity that I may not
be able to deal with
<braunr> we wouldn't want him to waste too much time on the external
libpthread
<braunr> which one ?
<bddebian> glibc for one. hurd_condition_wait() for another which I don't
have a great grasp on. Remember my knowledge/skillsets are limited
currently.
<braunr> bddebian: tschwinge has good instructions to build glibc
<braunr> keep your tree around and it shouldn't be long to hack on it
<braunr> for hurd_condition_wait, i can help
<bddebian> Oh I was thinking about using Debian glibc for now. You think I
should do it from git?
<braunr> no
<braunr> debian rules are even more reliable
<braunr> (just don't build all the variants)
<pinotree> `debian/rules build_libc` builds the plain i386 variant only
<bddebian> So put pthread_hurd_cond_wait in it's own .c file or just put it
in pt-cond-wait.c ?
<braunr> i'd put it in pt-cond-wait.C
<bddebian> youpi or braunr: OK, another dumb question. What (if anything)
should I do about hurd/hurd/signal.h. Should I stop it from including
cthreads?
<youpi> it's not a dumb question. it should probably stop, yes, but there
might be uncovered issues, which we'll have to take care of
<bddebian> Well I know antrik suggested trying to keep compatibility but I
don't see how you would do that
<braunr> compability between what ?
<braunr> and source and/or binary ?
<youpi> hurd/signal.h implicitly including cthreads.h
<braunr> ah
<braunr> well yes, it has to change obviously
<bddebian> Which will break all the cthreads stuff of course
<bddebian> So are we agreeing on pthread_hurd_cond_wait()?
<braunr> that's fine
<bddebian> Ugh, shit there is stuff in glibc using cthreads??
<braunr> like what ?
<bddebian> hurdsig, hurdsock, setauth, dtable, ...
<youpi> it's just using the compatibility stuff, that pthread does provide
<bddebian> but it includes cthreads.h implicitly
<bddebian> s/it/they in many cases
<youpi> not a problem, we provide the functions
<bddebian> Hmm, then what do I do about signal.h? It includes chtreads.h
because it uses extern struct mutex ...
<youpi> ah, then keep the include
<youpi> the pthread mutexes are compatible with that
<youpi> we'll clean that afterwards
<bddebian> arf, OK
<youpi> that's what I meant by "uncover issues"
IRC, freenode, #hurd, 2012-07-28
<bddebian> Well crap, glibc built but I have no symbol for
pthread_hurd_cond_wait in libpthread.so :(
<bddebian> Hmm, I wonder if I have to add pthread_hurd_cond_wait to
forward.c and Versions? (Versions obviously eventually)
<pinotree> bddebian: most probably not about forward.c, but definitely you
have to export public stuff using Versions
IRC, freenode, #hurd, 2012-07-29
<bddebian> braunr: http://paste.debian.net/181078/
<braunr> ugh, inline functions :/
<braunr> "Tell hurd_thread_cancel how to unblock us"
<braunr> i think you need that one too :p
<bddebian> ??
<braunr> well, they work in pair
<braunr> one cancels, the other notices it
<braunr> hurd_thread_cancel is in the hurd though, iirc
<braunr> or uh wait
<braunr> no it's in glibc, hurd/thread-cancel.c
<braunr> otherwise it looks like a correct reuse of the original code, but
i need to understand the pthreads internals better to really say anything
IRC, freenode, #hurd, 2012-08-03
<braunr> pinotree: what do you think of
condition_implies/condition_unimplies ?
<braunr> the work on pthread will have to replace those
IRC, freenode, #hurd, 2012-08-06
<braunr> bddebian: so, where is the work being done ?
<bddebian> braunr: Right now I would just like to testing getting my glibc
with pthread_hurd_cond_wait installed on the clubber subhurd. It is in
/home/bdefreese/glibc-debian2
<braunr> we need a git branch
<bddebian> braunr: Then I want to rebuild hurd with Thomas's pthread
patches against that new libc
<bddebian> Aye
<braunr> i don't remember, did thomas set a git repository somewhere for
that ?
<bddebian> He has one but I didn't have much luck with it since he is using
an external libpthreads
<braunr> i can manage the branches
<bddebian> I was actually patching debian/hurd then adding his patches on
top of that. It is in /home/bdefreese/debian-hurd but he has updateds
some stuff since then
<bddebian> Well we need to agree on a strategy. libpthreads only exists in
debian/glibc
<braunr> it would be better to have something upstream than to work on a
debian specific branch :/
<braunr> tschwinge: do you think it can be done
<braunr> ?
IRC, freenode, #hurd, 2012-08-07
<tschwinge> braunr: You mean to create on Savannah branches for the
libpthread conversion? Sure -- that's what I have been suggesting to
Barry and Thomas D. all the time.
<bddebian> braunr: OK, so I installed my glibc with
pthread_hurd_condition_wait in the subhurd and now I have built Debian
Hurd with Thomas D's pthread patches.
<braunr> bddebian: i'm not sure we're ready for tests yet :p
<bddebian> braunr: Why not? :)
<braunr> bddebian: a few important bits are missing
<bddebian> braunr: Like?
<braunr> like condition_implies
<braunr> i'm not sure they have been handled everywhere
<braunr> it's still interesting to try, but i bet your system won't finish
booting
<bddebian> Well I haven't "installed" the built hurd yet
<bddebian> I was trying to think of a way to test a little bit first, like
maybe ext2fs.static or something
<bddebian> Ohh, it actually mounted the partition
<bddebian> How would I actually "test" it?
<braunr> git clone :p
<braunr> building a debian package inside
<braunr> removing the whole content after
<braunr> that sort of things
<bddebian> Hmm, I think I killed clubber :(
<bddebian> Yep.. Crap! :(
<braunr> ?
<braunr> how did you do that ?
<bddebian> Mounted a new partition with the pthreads ext2fs.static then did
an apt-get source hurd to it..
<braunr> what partition, and what mount point ?
<bddebian> I added a new 2Gb partition on /dev/hd0s6 and set the translator
on /home/bdefreese/part6
<braunr> shouldn't kill your hurd
<bddebian> Well it might still be up but killed my ssh session at the very
least :)
<braunr> ouch
<bddebian> braunr: Do you have debugging enabled in that custom kernel you
installed? Apparently it is sitting at the debug prompt.
IRC, freenode, #hurd, 2012-08-12
<braunr> hmm, it seems the hurd notion of cancellation is actually not the
pthread one at all
<braunr> pthread_cancel merely marks a thread as being cancelled, while
hurd_thread_cancel interrupts it
<braunr> ok, i have a pthread_hurd_cond_wait_np function in glibc
IRC, freenode, #hurd, 2012-08-13
<braunr> nice, i got ext2fs work with pthreads
<braunr> there are issues with the stack size strongly limiting the number
of concurrent threads, but that's easy to fix
<braunr> one problem with the hurd side is the condition implications
<braunr> i think it should be deal separately, and before doing anything
with pthreads
<braunr> but that's minor, the most complex part is, again, the term server
<braunr> other than that, it was pretty easy to do
<braunr> but, i shouldn't speak too soon, who knows what tricky bootstrap
issue i'm gonna face ;p
<braunr> tschwinge: i'd like to know how i should proceed if i want a
symbol in a library overriden by that of a main executable
<braunr> e.g. have libpthread define a default stack size, and let
executables define their own if they want to change it
<braunr> tschwinge: i suppose i should create a weak alias in the library
and a normal variable in the executable, right ?
<braunr> hm i'm making this too complicated
<braunr> don't mind that stupid question
<tschwinge> braunr: A simple variable definition would do, too, I think?
<tschwinge> braunr: Anyway, I'd first like to know why we can'T reduce the
size of libpthread threads from 2 MiB to 64 KiB as libthreads had. Is
that a requirement of the pthread specification?
<braunr> tschwinge: it's a requirement yes
<braunr> the main reason i see is that hurd threadvars (which are still
present) rely on common stack sizes and alignment to work
<tschwinge> Mhm, I see.
<braunr> so for now, i'm using this approach as a hack only
<tschwinge> I'm working on phasing out threadvars, but we're not there yet.
<tschwinge> Yes, that's fine for the moment.
<braunr> tschwinge: a simple definition wouldn't work
<braunr> tschwinge: i resorted to a weak symbol, and see how it goes
<braunr> tschwinge: i supposed i need to export my symbol as a global one,
otherwise making it weak makes no sense, right ?
<braunr> suppose*
<braunr> tschwinge: also, i'm not actually sure what you meant is a
requirement about the stack size, i shouldn't have answered right away
<braunr> no there is actually no requirement
<braunr> i misunderstood your question
<braunr> hm when adding this weak variable, starting a program segfaults :(
<braunr> apparently on ___pthread_self, a tls variable
<braunr> fighting black magic begins
<braunr> arg, i can't manage to use that weak symbol to reduce stack sizes
:(
<braunr> ah yes, finally
<braunr> git clone /path/to/glibc.git on a pthread-powered ext2fs server :>
<braunr> tschwinge: seems i have problems using __thread in hurd code
<braunr> tschwinge: they produce undefined symbols
<braunr> tschwinge: forget that, another mistake on my part
<braunr> so, current state: i just need to create another patch, for the
code that is included in the debian hurd package but not in the upstream
hurd repository (e.g. procfs, netdde), and i should be able to create
hurd packages taht completely use pthreads
IRC, freenode, #hurd, 2012-08-14
<braunr> tschwinge: i have weird bootstrap issues, as expected
<braunr> tschwinge: can you point me to important files involved during
bootstrap ?
<braunr> my ext2fs.static server refuses to start as a rootfs, whereas it
seems to work fine otherwise
<braunr> hm, it looks like it's related to global signal dispositions
IRC, freenode, #hurd, 2012-08-15
<braunr> ahah, a subhurd running pthreads-powered hurd servers only
<LarstiQ> braunr: \o/
<braunr> i can even long on ssh
<braunr> log
<braunr> pinotree: for reference, i uploaded my debian-specific changes
there :
<braunr> http://git.sceen.net/rbraun/debian_hurd.git/
<braunr> darnassus is now running a pthreads-enabled hurd system :)
IRC, freenode, #hurd, 2012-08-16
<braunr> my pthreads-enabled hurd systems can quickly die under load
<braunr> youpi: with hurd servers using pthreads, i occasionally see thread
storms apparently due to a deadlock
<braunr> youpi: it makes me think of the problem you sometimes have (and
had often with the page cache patch)
<braunr> in cthreads, mutex and condition operations are macros, and they
check the mutex/condition queue without holding the internal
mutex/condition lock
<braunr> i'm not sure where this can lead to, but it doesn't seem right
<pinotree> isn't that a bit dangerous?
<braunr> i believe it is
<braunr> i mean
<braunr> it looks dangerous
<braunr> but it may be perfectly safe
<pinotree> could it be?
<braunr> aiui, it's an optimization, e.g. "dont take the internal lock if
there are no thread to wake"
<braunr> but if there is a thread enqueuing itself at the same time, it
might not be waken
<pinotree> yeah
<braunr> pthreads don't have this issue
<braunr> and what i see looks like a deadlock
<pinotree> anything can happen between the unlocked checking and the
following instruction
<braunr> so i'm not sure how a situation working around a faulty
implementation would result in a deadlock with a correct one
<braunr> on the other hand, the error youpi reported
(http://lists.gnu.org/archive/html/bug-hurd/2012-07/msg00051.html) seems
to indicate something is deeply wrong with libports
<pinotree> it could also be the current code does not really "works around"
that, but simply implicitly relies on the so-generated behaviour
<braunr> luckily not often
<braunr> maybe
<braunr> i think we have to find and fix these issues before moving to
pthreads entirely
<braunr> (ofc, using pthreads to trigger those bugs is a good procedure)
<pinotree> indeed
<braunr> i wonder if tweaking the error checking mode of pthreads to abort
on EDEADLK is a good approach to detecting this problem
<braunr> let's try !
<braunr> youpi: eh, i think i've spotted the libports ref mistake
<youpi> ooo!
<youpi> .oOo.!!
<gnu_srs> Same problem but different patches
<braunr> look at libports/bucket-iterate.c
<braunr> in the HURD_IHASH_ITERATE loop, pi->refcnt is incremented without
a lock
<youpi> Mmm, the incrementation itself would probably be compiled into an
INC, which is safe in UP
<youpi> it's an add currently actually
<youpi> 0x00004343 <+163>: addl $0x1,0x4(%edi)
<braunr> 40c4: 83 47 04 01 addl $0x1,0x4(%edi)
<youpi> that makes it SMP unsafe, but not UP unsafe
<braunr> right
<braunr> too bad
<youpi> that still deserves fixing :)
<braunr> the good side is my mind is already wired for smp
<youpi> well, it's actually not UP either
<youpi> in general
<youpi> when the processor is not able to do the add in one instruction
<braunr> sure
<braunr> youpi: looks like i'm wrong, refcnt is protected by the global
libports lock
<youpi> braunr: but aren't there pieces of code which manipulate the refcnt
while taking another lock than the global libports lock
<youpi> it'd not be scalable to use the global libports lock to protect
refcnt
<braunr> youpi: imo, the scalability issues are present because global
locks are taken all the time, indeed
<youpi> urgl
<braunr> yes ..
<braunr> when enabling mutex checks in libpthread, pfinet dies :/
<braunr> grmbl, when trying to start "ls" using my deadlock-detection
libpthread, the terminal gets unresponsive, and i can't even use ps .. :(
<pinotree> braunr: one could say your deadlock detection works too
good... :P
<braunr> pinotree: no, i made a mistake :p
<braunr> it works now :)
<braunr> well, works is a bit fast
<braunr> i can't attach gdb now :(
<braunr> *sigh*
<braunr> i guess i'd better revert to a cthreads hurd and debug from there
<braunr> eh, with my deadlock-detection changes, recursive mutexes are now
failing on _pthread_self(), which for some obscure reason generates this
<braunr> => 0x0107223b <+283>: jmp 0x107223b
<__pthread_mutex_timedlock_internal+283>
<braunr> *sigh*
IRC, freenode, #hurd, 2012-08-17
<braunr> aw, the thread storm i see isn't a deadlock
<braunr> seems to be mere contention ....
<braunr> youpi: what do you think of the way
ports_manage_port_operations_multithread determines it needs to spawn a
new thread ?
<braunr> it grabs a lock protecting the number of threads to determine if
it needs a new thread
<braunr> then releases it, to retake it right after if a new thread must be
created
<braunr> aiui, it could lead to a situation where many threads could
determine they need to create threads
<youpi> braunr: there's no reason to release the spinlock before re-taking
it
<youpi> that can indeed lead to too much thread creations
<braunr> youpi: a harder question
<braunr> youpi: what if thread creation fails ? :/
<braunr> if i'm right, hurd servers simply never expect thread creation to
fail
<youpi> indeed
<braunr> and as some patterns have threads blocking until another produce
an event
<braunr> i'm not sure there is any point handling the failure at all :/
<youpi> well, at least produce some output
<braunr> i added a perror
<youpi> so we know that happened
<braunr> async messaging is quite evil actually
<braunr> the bug i sometimes have with pfinet is usually triggered by
fakeroot
<braunr> it seems to use select a lot
<braunr> and select often destroys ports when it has something to return to
the caller
<braunr> which creates dead name notifications
<braunr> and if done often enough, a lot of them
<youpi> uh
<braunr> and as pfinet is creating threads to service new messages, already
existing threads are starved and can't continue
<braunr> which leads to pfinet exhausting its address space with thread
stacks (at about 30k threads)
<braunr> i initially thought it was a deadlock, but my modified libpthread
didn't detect one, and indeed, after i killed fakeroot (the whole
dpkg-buildpackage process hierarchy), pfinet just "cooled down"
<braunr> with almost all 30k threads simply waiting for requests to
service, and the few expected select calls blocking (a few ssh sessions,
exim probably, possibly others)
<braunr> i wonder why this doesn't happen with cthreads
<youpi> there's a 4k guard between stacks, otherwise I don't see anything
obvious
<braunr> i'll test my pthreads package with the fixed
ports_manage_port_operations_multithread
<braunr> but even if this "fix" should reduce thread creation, it doesn't
prevent the starvation i observed
<braunr> evil concurrency :p
<braunr> youpi: hm i've just spotted an important difference actually
<braunr> youpi: glibc sched_yield is __swtch(), cthreads is
thread_switch(MACH_PORT_NULL, SWITCH_OPTION_DEPRESS, 10)
<braunr> i'll change the glibc implementation, see how it affects the whole
system
<braunr> youpi: do you think bootsting the priority or cancellation
requests is an acceptable workaround ?
<braunr> boosting
<braunr> of*
<youpi> workaround for what?
<braunr> youpi: the starvation i described earlier
<youpi> well, I guess I'm not into the thing enough to understand
<youpi> you meant the dead port notifications, right?
<braunr> yes
<braunr> they are the cancellation triggers
<youpi> cancelling whaT?
<braunr> a blocking select for example
<braunr> ports_do_mach_notify_dead_name -> ports_dead_name ->
ports_interrupt_notified_rpcs -> hurd_thread_cancel
<braunr> so it's important they are processed quickly, to allow blocking
threads to unblock, reply, and be recycled
<youpi> you mean the threads in pfinet?
<braunr> the issue applies to all servers, but yes
<youpi> k
<youpi> well, it can not not be useful :)
<braunr> whatever the choice, it seems to be there will be a security issue
(a denial of service of some kind)
<youpi> well, it's not only in that case
<youpi> you can always queue a lot of requests to a server
<braunr> sure, i'm just focusing on this particular problem
<braunr> hm
<braunr> max POLICY_TIMESHARE or min POLICY_FIXEDPRI ?
<braunr> i'd say POLICY_TIMESHARE just in case
<braunr> (and i'm not sure mach handles fixed priority threads first
actually :/)
<braunr> hm my current hack which consists of calling swtch_pri(0) from a
freshly created thread seems to do the job eh
<braunr> (it may be what cthreads unintentionally does by acquiring a spin
lock from the entry function)
<braunr> not a single issue any more with this hack
<bddebian> Nice
<braunr> bddebian: well it's a hack :p
<braunr> and the problem is that, in order to boost a thread's priority,
one would need to implement that in libpthread
<bddebian> there isn't thread priority in libpthread?
<braunr> it's not implemented
<bddebian> Interesting
<braunr> if you want to do it, be my guest :p
<braunr> mach should provide the basic stuff for a partial implementation
<braunr> but for now, i'll fall back on the hack, because that's what
cthreads "does", and it's "reliable enough"
<antrik> braunr: I don't think the locking approach in
ports_manage_port_operations_multithread() could cause issues. the worst
that can happen is that some other thread becomes idle between the check
and creating a new thread -- and I can't think of a situation where this
could have any impact...
<braunr> antrik: hm ?
<braunr> the worst case is that many threads will evalute spawn to 1 and
create threads, whereas only one of them should have
<antrik> braunr: I'm not sure perror() is a good way to handle the
situation where thread creation failed. this would usually happen because
of resource shortage, right? in that case, it should work in non-debug
builds too
<braunr> perror isn't specific to debug builds
<braunr> i'm building glibc packages with a pthreads-enabled hurd :>
<braunr> (which at one point run the test allocating and filling 2 GiB of
memory, which passed)
<braunr> (with a kernel using a 3/1 split of course, swap usage reached
something like 1.6 GiB)
<antrik> braunr: BTW, I think the observation that thread storms tend to
happen on destroying stuff more than on creating stuff has been made
before...
<braunr> ok
<antrik> braunr: you are right about perror() of course. brain fart -- was
thinking about assert_perror()
<antrik> (which is misused in some places in existing Hurd code...)
<antrik> braunr: I still don't see the issue with the "spawn"
locking... the only situation where this code can be executed
concurrently is when multiple threads are idle and handling incoming
request -- but in that case spawning does *not* happen anyways...
<antrik> unless you are talking about something else than what I'm thinking
of...
<braunr> well imagine you have idle threads, yes
<braunr> let's say a lot like a thousand
<braunr> and the server gets a thousand requests
<braunr> a one more :p
<braunr> normally only one thread should be created to handle it
<braunr> but here, the worst case is that all threads run internal_demuxer
roughly at the same time
<braunr> and they all determine they need to spawn a thread
<braunr> leading to another thousand
<braunr> (that's extreme and very unlikely in practice of course)
<antrik> oh, I see... you mean all the idle threads decide that no spawning
is necessary; but before they proceed, finally one comes in and decides
that it needs to spawn; and when the other ones are scheduled again they
all spawn unnecessarily?
<braunr> no, spawn is a local variable
<braunr> it's rather, all idle threads become busy, and right before
servicing their request, they all decide they must spawn a thread
<antrik> I don't think that's how it works. changing the status to busy (by
decrementing the idle counter) and checking that there are no idle
threads is atomic, isn't it?
<braunr> no
<antrik> oh
<antrik> I guess I should actually look at that code (again) before
commenting ;-)
<braunr> let me check
<braunr> no sorry you're right
<braunr> so right, you can't lead to that situation
<braunr> i don't even understand how i can't see that :/
<braunr> let's say it's the heat :p
<braunr> 22:08 < braunr> so right, you can't lead to that situation
<braunr> it can't lead to that situation
IRC, freenode, #hurd, 2012-08-18
<braunr> one more attempt at fixing netdde, hope i get it right this time
<braunr> some parts assume a ddekit thread is a cthread, because they share
the same address
<braunr> it's not as easy when using pthread_self :/
<braunr> good, i got netdde work with pthreads
<braunr> youpi: for reference, there are now glibc, hurd and netdde
packages on my repository
<braunr> youpi: the debian specific patches can be found at my git
repository (http://git.sceen.net/rbraun/debian_hurd.git/ and
http://git.sceen.net/rbraun/debian_netdde.git/)
<braunr> except a freeze during boot (between exec and init) which happens
rarely, and the starvation which still exists to some extent (fakeroot
can cause many threads to be created in pfinet and pflocal), the
glibc/hurd packages have been working fine for a few days now
<braunr> the threading issue in pfinet/pflocal is directly related to
select, which the io_select_timeout patches should fix once merged
<braunr> well, considerably reduce at least
<braunr> and maybe fix completely, i'm not sure
IRC, freenode, #hurd, 2012-08-27
<pinotree> braunr: wrt a78a95d in your pthread branch of hurd.git,
shouldn't that job theorically been done using pthread api (of course
after implementing it)?
<braunr> pinotree: sure, it could be done through pthreads
<braunr> pinotree: i simply restricted myself to moving the hurd to
pthreads, not augment libpthread
<braunr> (you need to remember that i work on hurd with pthreads because it
became a dependency of my work on fixing select :p)
<braunr> and even if it wasn't the reason, it is best to do these tasks
(replace cthreads and implement pthread scheduling api) separately
<pinotree> braunr: hm ok
<pinotree> implementing the pthread priority bits could be done
independently though
<braunr> youpi: there are more than 9000 threads for /hurd/streamio kmsg on
ironforge oO
<youpi> kmsg ?!
<youpi> it's only /dev/klog right?
<braunr> not sure but it seems so
<pinotree> which syslog daemon is running?
<youpi> inetutils
<youpi> I've restarted the klog translator, to see whether when it grows
again
<braunr> 6 hours and 21 minutes to build glibc on darnassus
<braunr> pfinet still runs only 24 threads
<braunr> the ext2 instance used for the build runs 2k threads, but that's
because of the pageouts
<braunr> so indeed, the priority patch helps a lot
<braunr> (pfinet used to have several hundreds, sometimes more than a
thousand threads after a glibc build, and potentially increasing with
each use of fakeroot)
<braunr> exec weights 164M eww, we definitely have to fix that leak
<braunr> the leaks are probably due to wrong mmap/munmap usage
IRC, freenode, #hurd, 2012-08-29
<braunr> youpi: btw, after my glibc build, there were as little as between
20 and 30 threads for pflocal and pfinet
<braunr> with the priority patch
<braunr> ext2fs still had around 2k because of pageouts, but that's
expected
<youpi> ok
<braunr> overall the results seem very good and allow the switch to
pthreads
<youpi> yep, so it seems
<braunr> youpi: i think my first integration branch will include only a few
changes, such as this priority tuning, and the replacement of
condition_implies
<youpi> sure
<braunr> so we can push the move to pthreads after all its small
dependencies
<youpi> yep, that's the most readable way
IRC, freenode, #hurd, 2012-09-03
<gnu_srs> braunr: Compiling yodl-3.00.0-7:
<gnu_srs> pthreads: real 13m42.460s, user 0m0.000s, sys 0m0.030s
<gnu_srs> cthreads: real 9m 6.950s, user 0m0.000s, sys 0m0.020s
<braunr> thanks
<braunr> i'm not exactly certain about what causes the problem though
<braunr> it could be due to libpthread using doubly-linked lists, but i
don't think the overhead would be so heavier because of that alone
<braunr> there is so much contention sometimes that it could
<braunr> the hurd would have been better off with single threaded servers
:/
<braunr> we should probably replace spin locks with mutexes everywhere
<braunr> on the other hand, i don't have any more starvation problem with
the current code
IRC, freenode, #hurd, 2012-09-06
<gnu_srs> braunr: Yes you are right, the new pthread-based Hurd is _much_
slower.
<gnu_srs> One annoying example is when compiling, the standard output is
written in bursts with _long_ periods of no output in between:-(
<braunr> that's more probably because of the priority boost, not the
overhead
<braunr> that's one of the big issues with our mach-based model
<braunr> we either give high priorities to our servers, or we can suffer
from message floods
<braunr> that's in fact more a hurd problem than a mach one
<gnu_srs> braunr: any immediate ideas how to speed up responsiveness the
pthread-hurd. It is annoyingly slow (slow-witted)
<braunr> gnu_srs: i already answered that
<braunr> it doesn't look that slower on my machines though
<gnu_srs> you said you had some ideas, not which. except for mcsims work.
<braunr> i have ideas about what makes it slower
<braunr> it doesn't mean i have solutions for that
<braunr> if i had, don't you think i'd have applied them ? :)
<gnu_srs> ok, how to make it more responsive on the console? and printing
stdout more regularly, now several pages are stored and then flushed.
<braunr> give more details please
<gnu_srs> it behaves like a loaded linux desktop, with little memory
left...
<braunr> details about what you're doing
<gnu_srs> apt-get source any big package and: fakeroot debian/rules binary
2>&1 | tee ../binary.logg
<braunr> isee
<braunr> well no, we can't improve responsiveness
<braunr> without reintroducing the starvation problem
<braunr> they are linked
<braunr> and what you're doing involes a few buffers, so the laggy feel is
expected
<braunr> if we can fix that simply, we'll do so after it is merged upstream
IRC, freenode, #hurd, 2012-09-07
<braunr> gnu_srs: i really don't feel the sluggishness you described with
hurd+pthreads on my machines
<braunr> gnu_srs: what's your hardware ?
<braunr> and your VM configuration ?
<gnu_srs> Intel(R) Core(TM)2 Quad CPU Q6600 @ 2.40GHz
<gnu_srs> kvm -m 1024 -net nic,model=rtl8139 -net
user,hostfwd=tcp::5562-:22 -drive
cache=writeback,index=0,media=disk,file=hurd-experimental.img -vnc :6
-cdrom isos/netinst_2012-07-15.iso
<braunr> what is the file system type where your disk image is stored ?
<gnu_srs> ext3
<braunr> and how much physical memory on the host ?
<braunr> (paste meminfo somewhere please)
<gnu_srs> 4G, and it's on the limit, 2 kvm instances+gnome,etc
<gnu_srs> 80% in use by programs, 14% in cache.
<braunr> ok, that's probably the reason then
<braunr> the writeback option doesn't help a lot if you don't have much
cache
<gnu_srs> well the other instance is cthreads based, and not so sluggish.
<braunr> we know hurd+pthreads is slower
<braunr> i just wondered why i didn't feel it that much
<gnu_srs> try to fire up more kvm instances, and do a heavy compile...
<braunr> i don't do that :)
<braunr> that's why i never had the problem
<braunr> most of the time i have like 2-3 GiB of cache
<braunr> and of course more on shattrath
<braunr> (the host of the sceen.net hurdboxes, which has 16 GiB of ram)
IRC, freenode, #hurd, 2012-09-11
<gnu_srs> Monitoring the cthreads and the pthreads load under Linux shows:
<gnu_srs> cthread version: load can jump very high, less cpu usage than
pthread version
<gnu_srs> pthread version: less memory usage, background cpu usage higher
than for cthread version
<braunr> that's the expected behaviour
<braunr> gnu_srs: are you using the lifothreads gnumach kernel ?
<gnu_srs> for experimental, yes.
<gnu_srs> i.e. pthreads
<braunr> i mean, you're measuring on it right now, right ?
<gnu_srs> yes, one instance running cthreads, and one pthreads (with lifo
gnumach)
<braunr> ok
<gnu_srs> no swap used in either instance, will try a heavy compile later
on.
<braunr> what for ?
<gnu_srs> E.g. for memory when linking. I have swap available, but no swap
is used currently.
<braunr> yes but, what do you intend to measure ?
<gnu_srs> don't know, just to see if swap is used at all. it seems to be
used not very much.
<braunr> depends
<braunr> be warned that using the swap means there is pageout, which is one
of the triggers for global system freeze :p
<braunr> anonymous memory pageout
<gnu_srs> for linux swap is used constructively, why not on hurd?
<braunr> because of hard to squash bugs
<gnu_srs> aha, so it is bugs hindering swap usage:-/
<braunr> yup :/
<gnu_srs> Let's find them thenO:-), piece of cake
<braunr> remember my page cache branch in gnumach ? :)
<gnu_srs> not much
<braunr> i started it before fixing non blocking select
<braunr> anyway, as a side effect, it should solve this stability issue
too, but it'll probably take time
<gnu_srs> is that branch integrated? I only remember slab and the lifo
stuff.
<gnu_srs> and mcsims work
<braunr> no it's not
<braunr> it's unfinished
<gnu_srs> k!
<braunr> it correctly extends the page cache to all available physical
memory, but since the hurd doesn't scale well, it slows the system down
IRC, freenode, #hurd, 2012-09-14
<braunr> arg
<braunr> darnassus seems to eat 100% cpu and make top freeze after some
time
<braunr> seems like there is an important leak in the pthreads version
<braunr> could be the lifothreads patch :/
<cjbirk> there's a memory leak?
<cjbirk> in pthreads?
<braunr> i don't think so, and it's not a memory leak
<braunr> it's a port leak
<braunr> probably in the kernel
IRC, freenode, #hurd, 2012-09-17
<braunr> nice, the port leak is actually caused by the exim4 loop bug
IRC, freenode, #hurd, 2012-09-23
<braunr> the port leak i observed a few days ago is because of exim4 (the
infamous loop eating the cpu we've been seeing regularly)
<youpi> oh
<braunr> next time it happens, and if i have the occasion, i'll examine the
problem
<braunr> tip: when you can't use top or ps -e, you can use ps -e -o
pid=,args=
<youpi> or -M ?
<braunr> haven't tested
IRC, freenode, #hurd, 2013-01-26
<braunr> ah great, one of the recent fixes (probably select-eintr or
setitimer) fixed exim4 :)
IRC, freenode, #hurd, 2012-09-23
<braunr> tschwinge: i committed the last hurd pthread change,
https://git.savannah.gnu.org/cgit/hurd/hurd.git/log/?h=master-pthreads
<braunr> tschwinge: please tell me if you consider it ok for merging
IRC, freenode, #hurd, 2012-11-27
<youpi> braunr: btw, I forgot to forward here, with the glibc patch it does
boot fine, I'll push all that and build some almost-official packages for
people to try out what will come when eglibc gets the change in unstable
<braunr> youpi: great :)
<youpi> thanks for managing the final bits of this
<youpi> (and thanks for everybody involved)
<braunr> sorry again for the non obvious parts
<braunr> if you need the debian specific parts refined (e.g. nice commits
for procfs & others), i can do that
<youpi> I'll do that, no pb
<braunr> ok
<braunr> after that (well, during also), we should focus more on bug
hunting
IRC, freenode, #hurd, 2012-10-26
<mcsim1> hello. What does following error message means? "unable to adjust
libports thread priority: Operation not permitted" It appears when I set
translators.
<mcsim1> Seems has some attitude to libpthread. Also following appeared
when I tried to remove translator: "pthread_create: Resource temporarily
unavailable"
<mcsim1> Oh, first message appears very often, when I use translator I set.
<braunr> mcsim1: it's related to a recent patch i sent
<braunr> mcsim1: hurd servers attempt to increase their priority on startup
(when a thread is created actually)
<braunr> to reduce message floods and thread storms (such sweet names :))
<braunr> but if you start them as an unprivileged user, it fails, which is
ok, it's just a warning
<braunr> the second way is weird
<braunr> it normally happens when you're out of available virtual space,
not when shutting a translator donw
<mcsim1> braunr: you mean this patch: libports: reduce thread starvation on
message floods?
<braunr> yes
<braunr> remember you're running on darnassus
<braunr> with a heavily modified hurd/glibc
<braunr> you can go back to the cthreads version if you wish
<mcsim1> it's better to check translators privileges, before attempting to
increase their priority, I think.
<braunr> no
<mcsim1> it's just a bit annoying
<braunr> privileges can be changed during execution
<braunr> well remove it
<mcsim1> But warning should not appear.
<braunr> what could be done is to limit the warning to one occurrence
<braunr> mcsim1: i prefer that it appears
<mcsim1> ok
<braunr> it's always better to be explicit and verbose
<braunr> well not always, but very often
<braunr> one of the reasons the hurd is so difficult to debug is the lack
of a "message server" à la dmesg
IRC, freenode, #hurd, 2012-12-10
<youpi> braunr: unable to adjust libports thread priority: (ipc/send)
invalid destination port
<youpi> I'll see what package brought that
<youpi> (that was on a buildd)
<braunr> wow
<youpi> mkvtoolnix_5.9.0-1:
<pinotree> shouldn't that code be done in pthreads and then using such
pthread api? :p
<braunr> pinotree: you've already asked that question :p
<pinotree> i know :p
<braunr> the semantics of pthreads are larger than what we need, so that
will be done "later"
<braunr> but this error shouldn't happen
<braunr> it looks more like a random mach bug
<braunr> youpi: anything else on the console ?
<youpi> nope
<braunr> i'll add traces to know which step causes the error
IRC, freenode, #hurd, 2012-12-11
<youpi> braunr: mktoolnix seems like a reproducer for the libports thread
priority issue
<youpi> (3 times)
<braunr> youpi: thanks
<braunr> youpi: where is that tool packaged ?
<pinotree> he probably means the mkvtoolnix source
<braunr> seems so
<braunr> i don't find anything else
<youpi> that's it, yes
IRC, freenode, #hurd, 2013-03-01
<youpi> braunr: btw, "unable to adjust libports thread priority: (ipc/send)
invalid destination port" is actually not a sign of fatality
<youpi> bach recovered from it
<braunr> youpi: well, it never was a sign of fatality
<braunr> but it means that, for some reason, a process looses a right for a
very obscure reason :/
<braunr> weird sentence, agreed :p
IRC, freenode, #hurd, 2013-06-14
<gnu_srs> Hi, when running check for gccgo the following occurs (multiple
times) locking up the console
<gnu_srs> unable to adjust libports thread priority: (ipc/send) invalid
destination port
<gnu_srs> (not locking up the console, it was just completely filled with
messages))
<braunr> gnu_srs: are you running your translator as root ?
<braunr> or, do you have a translator running as an unprivileged user ?
<braunr> hm, invalid dest port
<braunr> that's a bug :p
<braunr> but i don't know why
<braunr> i'll have to take some time to track it down
<braunr> it might be a user ref overflow or something similarly tricky
<braunr> gnu_srs: does it happen everytime you run gccgo checks or only
after the system has been alive for some time ?
<braunr> (some time being at least a few hours, more probably days)
IRC, freenode, #hurd, 2013-07-05
<braunr> ok, found the bug about invalid ports when adjusting priorities
<braunr> thhe hurd must be plagued with wrong deallocations :(
<braunr> i have so many problems when trying to cleanly destroy threads
IRC, freenode, #hurd, 2013-11-25
<braunr> youpi: btw, my last commit on the hurd repo fixes the urefs
overflow we've sometimes seen in the past in the priority adjusting code
of libports
IRC, freenode, #hurd, 2013-11-29
See also fix have kernel resources.
<braunr> there still are some leak ports making servers spawn threads with
non-elevated priorities :/
<braunr> leaks*
<teythoon> issues with your thread destruction work ?
<teythoon> err, wait
<teythoon> why does a port leak cause that ?
<braunr> because it causes urefs overflows
<braunr> and the priority adjustment code does check errors :p
<teythoon> ^^
<teythoon> ah yes, urefs...
<braunr> apparently it only affects the root file system
<teythoon> hm
<braunr> i'll spend an hour looking for it, and whatever i find, i'll
install the upstream debian packages so you can build glibc without too
much trouble
<teythoon> we need a clean build chroot on darnassus for this situation
<braunr> ah yes
<braunr> i should have time to set things up this week end
<braunr> 1: send (refs: 65534)
<braunr> i wonder what the first right is in the root file system
<teythoon> hm
<braunr> search doesn't help so i'm pretty sure it's a kernel object
<braunr> perhaps the host priv port
<teythoon> could be the thread port or something ?
<braunr> no, not the thread port
<teythoon> why would it have so many refs ?
<braunr> the task port maybe but it's fine if it overflows
<teythoon> also, some urefs are clamped at max, so maybe this is fine ?
<braunr> it may be fine yes
<braunr> err = get_privileged_ports (&host_priv, NULL);
<braunr> iirc, this function should pass copies of the name, not increment
the urefs counter
<braunr> it may behave differently if built statically
<teythoon> o_O y would it ?
<braunr> no idea
<braunr> something doesn't behave as it should :)
<braunr> i'm not asking why, i'm asking where :)
<braunr> the proc server is also affected
<braunr> so it does look like it has something to do with bootstrap
<teythoon> I'm not surprised :/
IRC, freenode, #hurd, 2013-11-30
<braunr> so yes, the host_priv port gets a reference when calling
get_privileged_ports
<braunr> but only in the rootfs and proc servers, probably because others
use the code path to fetch it from proc
<teythoon> ah
<teythoon> well, it shouldn't behave differently
<braunr> ?
<teythoon> get_privileged_ports
<braunr> get_privileged_ports is explictely described to cache references
<teythoon> i don't get it
<teythoon> you said it behaved differently for proc and the rootfs
<teythoon> that's undesireable, isn't it ?
<braunr> yes
<teythoon> ok
<braunr> so it should behave differently than it does
<teythoon> yes
<teythoon> right
<braunr> teythoon: during your work this summer, have you come across the
bootstrap port of a task ?
<braunr> i wonder what the bootstrap port of the root file system is
<braunr> maybe i got the description wrong since references on host or
master are deallocated where get_privileged_ports is used ..
<teythoon> no, I do not believe i did anything bootstrap port related
<braunr> ok
<braunr> i don't need that any more fortunately
<braunr> i just wonder how someone could write a description so error-prone
..
<braunr> and apparently, this problem should affect all servers, but for
some reason i didn't see it
<braunr> there, problem fixed
<teythoon> ?
<braunr> last leak eliminated
<teythoon> cool :)
<teythoon> how ?
<braunr> i simply deallocate host_priv in addition to the others when
adjusting thread priority
<braunr> as simple as that ..
<teythoon> uh
<teythoon> sure ?
<braunr> so many system calls just for reference counting
<braunr> yes
<teythoon> i did that, and broke the rootfs
<braunr> well i'm using one right now
<teythoon> ok
<braunr> maybe i should let it run a bit :)
<teythoon> no, for me it failed on the first write
<braunr> teythoon: looks weird
<teythoon> so i figured it was wrong to deallocate that port
<braunr> i'll reboot it and see if there may be a race
<teythoon> thought i didn't get a reference after all or something
<teythoon> I believe there is a race in ext2fs
<braunr> teythoon: that's not good news for me
<teythoon> when doing fsysopts --update / (which remounts /)
<teythoon> sometimes, the system hangs
<braunr> :/
<teythoon> might be a deadlock, or the rootfs dies and noone notices
<teythoon> with my protected payload stuff, the system would reboot instead
of just hanging
<braunr> oh
<teythoon> which might point to a segfault in ext2fs
<teythoon> maybe the exception message carries a bad payload
<braunr> makes sense
<braunr> exception handling in ext2fs is messy ..
<teythoon> braunr: and, doing sleep 0.1 before remounting / makes the
problem less likely to appear
<braunr> ugh
<teythoon> and system load on my host system seems to affect this
<teythoon> but it is hard to tell
<teythoon> sometimes, this doesn't show up at all
<teythoon> sometimes several times in a row
<braunr> the system load might simply indicate very short lived processes
<braunr> (or threads)
<teythoon> system load on my host
<braunr> ah
<teythoon> this makes me believe that it is a race somewhere
<teythoon> all of this
<braunr> well, i can't get anything wrong with my patched rootfs
<teythoon> braunr: ok, maybe I messed up
<braunr> or maybe you were very unlucky
<braunr> and there is a rare race
<braunr> but i'll commit anyway
<teythoon> no, i never got it to work, always hung at the first write
<braunr> it won't be the first or last rare problem we'll have to live with
<braunr> hm
<braunr> then you probably did something wrong, yes
<braunr> that's reassuring
IRC, freenode, #hurd, 2013-03-11
<braunr> youpi: oh btw, i noticed a problem with the priority adjustement
code
<braunr> a thread created by a privileged server (e.g. an ext2fs
translator) can then spawn a server from a node owned by an unprivileged
user
<braunr> which inherits the priority
<braunr> easy to fix but worth saying to keep in mind
<youpi> uh
<youpi> indeed
IRC, freenode, #hurd, 2013-07-01
<youpi> braunr: it seems as if pfinet is not prioritized enough
<youpi> I'm getting network connectivity issues when the system is quite
loaded
<braunr> loaded with what ?
<braunr> it could be ext2fs having a lot more threads than other servers
<youpi> building packages
<youpi> I'm talking about the buildds
<braunr> ok
<braunr> ironforge or others ?
<youpi> they're having troubles uploading packages while building stuff
<youpi> ironforge and others
<youpi> that happened already in the past sometimes
<youpi> but at the moment it's really pronounced
<braunr> i don't think it's a priority issue
<braunr> i think it's swapping
<youpi> ah, that's not impossible indeed
<youpi> but why would it swap?
<youpi> there's a lot of available memory
<braunr> a big file is enough
<braunr> it pushes anonymous memory out
<youpi> to fill 900MiB memory ?
<braunr> i see 535M of swap on if
<braunr> yes
<youpi> ironforge is just building libc
<braunr> and for some reason, swapping is orders of magnitude slower than
anything else
<youpi> not linking it yet
<braunr> i also see 1G of free memory on it
<youpi> that's what I meant with 900MiB
<braunr> so at some point, it needed a lot of memory, caused swapping
<braunr> and from time to time it's probably swapping back
<youpi> well, pfinet had all the time to swap back already
<youpi> I don't see why it should be still suffering from it
<braunr> swapping is a kernel activity
<youpi> ok, but once you're back, you're back
<youpi> unless something else pushes you out
<braunr> if the kernel is busy waiting for the default pager, nothing makes
progress
<braunr> (eccept the default pager hopefully)
<youpi> sure but pfinet should be back already, since it does work
<youpi> so I don't see why it should wait for something
<braunr> the kernel is waiting
<braunr> and the kernel isn't preemptibl
<braunr> e
<braunr> although i'm not sure preemption is the problem here
<youpi> well
<youpi> what I don't understand is what we have changed that could have so
much impact
<youpi> the only culprit I can see is the priorities we have changed
recently
<braunr> do you mean it happens a lot more frequently than before ?
<youpi> yes
<youpi> way
<braunr> ok
<youpi> ironforge is almost unusable while building glibc
<youpi> I've never seen that
<braunr> that's weird, i don't have these problems on darnassus
<braunr> but i think i reboot it more often
<braunr> could be a scalability issue then
<braunr> combined with the increased priorities
<braunr> if is indeed running full time on the host, whereas swapping
issues show the cpu being almost idle
<braunr> loadavg is high too so i guess there are many threads
<braunr> 0 971 3 -20 -20 1553 305358625 866485906 523M 63M * S<o
? 13hrs /hurd/ext2fs.static -A /dev/hd0s2
<braunr> 0 972 3 -20 -20 1434 125237556 719443981 483M 5.85M * S<o
? 13hrs /hurd/ext2fs.static -A /dev/hd0s3
<braunr> around 1k5 each
<youpi> that's quite usual
<braunr> could be the priorities then
<braunr> but i'm afraid that if we lower them, the number of threads will
grow out of control
<braunr> (good thing is i'm currently working on a way to make libpthread
actually remove kernel resources)
<youpi> but the priorities should be the same in ext2fs and pfinet,
shouldn't they?
<braunr> yes but ext2 has a lot more threads than pfinet
<braunr> the scheduler only sees threads, we don't have a grouping feature
<youpi> right
<braunr> we also should remove priority depressing in glibc
<braunr> (in sched_yield)
<braunr> it affects spin locks
<braunr> youpi: is it normal to see priorities of 26 ?
<youpi> braunr: we have changed the nice factor
<braunr> ah, factor
<youpi> Mm, I'm however realizing the gnumach kernel running these systems
hasn't been upgraded in a while
<youpi> it may not even have the needed priority levels
<braunr> ar euare you using top right now on if ?
<braunr> hm no i don't see it any more
<braunr> well yes, could be the priorities ..
<youpi> I've rebooted with an upgraded kernel
<youpi> no issue so far
<youpi> package uploads will tell me on the long run
<braunr> i bet it's also a scalability issue
<youpi> but why would it appear now only?
<braunr> until the cache and other data containers start to get filled,
processing is fast enough that we don't see it hapenning
<youpi> sure, but I haven't seen that in the past
<braunr> oh it's combined with the increased priorities
<youpi> even after a week building packages
<braunr> what i mean is, increased priorities don't affect much if threads
porcess things fast
<braunr> things get longer with more data, and then increased prioritis
give more time to these threads
<braunr> and that's when the problem appears
<youpi> but increased priorities give more time to the pfinet threads too,
don't they?
<braunr> yes
<youpi> so what is different ?
<braunr> but again, there are a lot more threads elsewhere
<braunr> with a lot more data to process
<youpi> sure, but that has alwasy been so
<braunr> hm
<youpi> really, 1k5 threads does not surprise me at all :)
<youpi> 10k would
<braunr> there aren't all active either
<youpi> yes
<braunr> but right, i don't know why pfinet would be given less time than
other threads ..
<braunr> compared to before
<youpi> particularly on xen-based buildds
<braunr> libpthread is slower than cthreads
<youpi> where it doesn't even have to wait for netdde
<braunr> threads need more quanta to achieve the same ting
<braunr> perhaps processing could usually be completed in one go before,
and not any more
<braunr> we had a discussion about this with antrik
<braunr> youpi: concerning the buildd issue, i don't think pfinet is
affected actually
<braunr> but the applications using the network may be
<youpi> why using the network would be a difference ?
<braunr> normal applications have a lower priority
<braunr> what i mean is, pfinet simply has nothing to do, because normal
applications don't have enough cpu time
<braunr> (what you said earlier seemed to imply pfinet had issues, i don't
think it has)
<braunr> it should be easy to test by pinging the machine while under load
<braunr> we should also check the priority of the special thread used to
handle packets, both in pfinet and netdde
<braunr> this one isn't spawned by libports and is likely to have a lower
priority as well
<braunr> youpi: you're right, something very recent slowed things down a
lot
<braunr> perhaps the new priority factor
<braunr> well not the factor but i suppose the priority range has been
increased
nice vs mach thread priorities.
<youpi> braunr: haven't had any upload issue so far
<youpi> over 20 uploads
<youpi> while it was usually 1 every 2 before...
<youpi> so it was very probably the kernel missing the priorities levels
<braunr> ok
<braunr> i think i've had the same problem on another virtual machine
<braunr> with a custom kernel i built a few weeks ago
<braunr> same kind of issue i guess
<braunr> it's fine now, and always was on darnassus
IRC, freenode, #hurd, 2012-12-05
<braunr> tschwinge: i'm currently working on a few easy bugs and i have
planned improvements for libpthreads soon
<pinotree> wotwot, which ones?
<braunr> pinotree: first, fixing pthread_cond_timedwait (and everything
timedsomething actually)
<braunr> pinotree: then, fixing cancellation
<braunr> pinotree: and last but not least, optimizing thread wakeup
<braunr> i also want to try replacing spin locks and see if it does what i
expect
<pinotree> which fixes do you plan applying to cond_timedwait?
<braunr> see sysdeps/generic/pt-cond-timedwait.c
<braunr> the FIXME comment
<pinotree> ah that
<braunr> well that's important :)
<braunr> did you have something else in mind ?
<pinotree> hm, __pthread_timedblock... do you plan fixing directly there? i
remember having seem something related to that (but not on conditions),
but wasn't able to see further
<braunr> it has the same issue
<braunr> i don't remember the details, but i wrote a cthreads version that
does it right
<braunr> in the io_select_timeout branch
<braunr> see
https://git.savannah.gnu.org/cgit/hurd/hurd.git/tree/libthreads/cancel-cond.c?h=rbraun/select_timeout
for example
* pinotree looks
<braunr> what matters is the msg_delivered member used to synchronize
sleeper and waker
<braunr> the waker code is in
https://git.savannah.gnu.org/cgit/hurd/hurd.git/tree/libthreads/cprocs.c?h=rbraun/select_timeout
<pinotree> never seen cthreads' code before :)
<braunr> soon you shouldn't have any more reason to :p
<pinotree> ah, so basically the cthread version of the pthread cleanup
stack + cancelation (ie the cancel hook) broadcasts the condition
<braunr> yes
<pinotree> so a similar fix would be needed in all the places using
__pthread_timedblock, that is conditions and mutexes
<braunr> and that's what's missing in glibc that prevents deploying a
pthreads based hurd currently
<braunr> no that's unrelated
<pinotree> ok
<braunr> the problem is how __pthread_block/__pthread_timedblock is
synchronized with __pthread_wakeup
<braunr> libpthreads does exactly the same thing as cthreads for that,
i.e. use messages
<braunr> but the message alone isn't enough, since, as explained in the
FIXME comment, it can arrive too late
<braunr> it's not a problem for __pthread_block because this function can
only resume after receiving a message
<braunr> but it's a problem for __pthread_timedblock which can resume
because of a timeout
<braunr> my solution is to add a flag that says whether a message was
actually sent, and lock around sending the message, so that the thread
resume can accurately tell in which state it is
<braunr> and drain the message queue if needed
<pinotree> i see, race between the "i stop blocking because of timeout" and
"i stop because i got a message" with the actual check for the real cause
<braunr> locking around mach_msg may seem overkill but it's not in
practice, since there can only be one message at most in the message
queue
<braunr> and i checked that in practice by limiting the message queue size
and check for such errors
<braunr> but again, it would be far better with mutexes only, and no spin
locks
<braunr> i wondered for a long time why the load average was so high on the
hurd under even "light" loads
<braunr> now i know :)
IRC, freenode, #hurd, 2012-12-27
<youpi> btw, good news: the installer works with libpthread
<youpi> (well, at least boots, I haven't tested the installation)
<braunr> i can do that if the image is available publically
<braunr> youpi: the one thing i suspect won't work right is the hurd
console :/
<braunr> so we might need to not enable it by default
<youpi> braunr: you mean the mode setting?
<braunr> youpi: i don't know what's wrong with the hurd console, but it
seems to deadlock with pthreads
<youpi> ah?
<youpi> I don't have such issue
<braunr> ah ? i need to retest that then
Same issue as term blocking perhaps?
IRC, freenode, #hurd, 2013-01-06
<youpi> it seems fakeroot has become slow as hell
<braunr> fakeroot is the main source of dead name notifications
<braunr> well, a very heavy one
<braunr> with pthreads hurd servers, their priority is raised, precisely to
give them time to handle those dead name notifications
<braunr> which slows everything else down, but strongly reduces the rate at
which additional threads are created to handle dn notifications
<braunr> so this is expected
<youpi> ok :/
<braunr> which is why i mentioned a rewrite of io_select into a completely
synchronous io_poll
<braunr> so that the client themselves remove their requests, instead of
the servers doing it asynchronously when notified
<youpi> by "slows everything else down", you mean, if the servers do take
cpu time?
<braunr> but considering the amount of messaging it requires, it will be
slow on moderate to large fd sets with frequent calls (non blocking or
low timeout)
<braunr> yes
<youpi> well here the problem is not really it gets slowed down
<youpi> but that e.g. for gtk+2.0 build, it took 5h cpu time
<youpi> (and counting)
<braunr> ah, the hurd with pthreads is noticeably slower too
<braunr> i'm not sure why, but i suspect the amount of internal function
calls could account for some of the overhead
<youpi> I mean the fakeroot process
<youpi> not the server process
<braunr> hum
<braunr> that's not normal :)
<youpi> that's what I meant
<braunr> well, i should try to build gtk+20 some day
<braunr> i've been building glibc today and it's going fine for now
<youpi> it's the install stage which poses problem
<youpi> I've noticed it with the hurd package too
<braunr> the hurd is easier to build
<braunr> that's a good test case
<braunr> there are many times when fakeroot just doesn't use cpu, and it
doesn't look like a select timeout issue (it still behaved that way with
my fixed branch)
<youpi> in general, pfinet is taking really a lot of cpu time
<youpi> that's surprising
<braunr> why ?
<braunr> fakeroot uses it a lot
<youpi> I know
<youpi> but still
<youpi> 40% cpu time is not normal
<youpi> I don't see why it would need so much cpu time
<braunr> 17:57 < braunr> but considering the amount of messaging it
requires, it will be slow on moderate to large fd sets with frequent
calls (non blocking or low timeout)
<youpi> by "it", what did you mean?
<youpi> I thought you meant the synchronous select implementation
<braunr> something irrelevant here
<braunr> yes
<braunr> what matters here is the second part of my sentence, which is what
i think happens now
<youpi> you mean it's the IPC overhead which is taking so much time?
<braunr> i mean, it doesn't matter if io_select synchronously removes
requests, or does it by destroying ports and relying on notifications,
there are lots of messages in this case anyway
<braunr> yes
<youpi> why "a lot" ?
<youpi> more than one per select call?
<braunr> yes
<youpi> why ?
<braunr> one per fd
<braunr> then one to wait
<youpi> there are two in faked
<braunr> hum :)
<braunr> i remember the timeout is low
<braunr> but i don't remember its value
<youpi> the timeout is NULL in faked
<braunr> the client then
<youpi> the client doesn't use select
<braunr> i must be confused
<braunr> i thought it did through the fakeroot library
<braunr> but yes, i see the same behaviour, 30 times more cpu for pfinet
than faked-tcp
<braunr> or let's say between 10 to 30
<braunr> and during my tests, these were the moments the kernel would
create lots of threads in servers and fail because of lack of memory,
either kernel memory, or virtual in the client space (filled with thread
stacks)
<braunr> it could be due to threads spinning too much
<braunr> (inside pfinet)
<youpi> attaching a gdb shows it mostly inside __pthread_block
<youpi> uh, how awful pfinet's select is
<youpi> a big global lock
<youpi> whenever something happens all threads get woken up
<pinotree> BKL!
* pinotree runs
<braunr> we have many big hurd locks :p
<youpi> it's rather a big translator lock
<braunr> more than a global lock it seems, a global condvar too, isn't it ?
<youpi> sure
<braunr> we have a similar problem with the hurd-specific cancellation
code, it's in my todo list with io_select
<youpi> ah, no, the condvar is not global
IRC, freenode, #hurd, 2013-01-14
<braunr> *sigh* thread cancellable is totally broken :(
<braunr> cancellation*
<braunr> it looks like playing with thread cancellability can make some
functions completely restart
<braunr> (e.g. one call to printf to write twice its output)
git duplicated content, git-core-2.
* braunr is cooking a patch to fix pthread cancellation in
pthread_cond_{,timed}wait, smells good
<braunr> youpi: ever heard of something that would make libc functions
"restart" ?
<youpi> you mean as a feature, or as a bug ?
<braunr> when changing the pthread cancellation state of a thread, i
sometimes see printf print its output twice
<youpi> or perhaps after a signal dispatch?
<braunr> i'll post my test code
<youpi> that could be a duplicate write
<youpi> due to restarting after signal
<braunr> http://www.sceen.net/~rbraun/pthreads_test_cancel.c
#include <stdio.h>
#include <stdarg.h>
#include <stdlib.h>
#include <pthread.h>
#include <unistd.h>
static pthread_cond_t cond = PTHREAD_COND_INITIALIZER;
static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
static int predicate;
static int ready;
static int cancelled;
static void
uncancellable_printf(const char *format, ...)
{
int oldstate;
va_list ap;
va_start(ap, format);
pthread_setcancelstate(PTHREAD_CANCEL_DISABLE, &oldstate);
vprintf(format, ap);
pthread_setcancelstate(oldstate, &oldstate);
va_end(ap);
}
static void *
run(void *arg)
{
uncancellable_printf("thread: setting ready\n");
ready = 1;
uncancellable_printf("thread: spin until cancellation is sent\n");
while (!cancelled)
sched_yield();
uncancellable_printf("thread: locking mutex\n");
pthread_mutex_lock(&mutex);
uncancellable_printf("thread: waiting for predicate\n");
while (!predicate)
pthread_cond_wait(&cond, &mutex);
uncancellable_printf("thread: unlocking mutex\n");
pthread_mutex_unlock(&mutex);
uncancellable_printf("thread: exit\n");
return NULL;
}
int
main(int argc, char *argv[])
{
pthread_t thread;
uncancellable_printf("main: create thread\n");
pthread_create(&thread, NULL, run, NULL);
uncancellable_printf("main: spin until thread is ready\n");
while (!ready)
sched_yield();
uncancellable_printf("main: sending cancellation\n");
pthread_cancel(thread);
uncancellable_printf("main: setting cancelled\n");
cancelled = 1;
uncancellable_printf("main: joining thread\n");
pthread_join(thread, NULL);
uncancellable_printf("main: exit\n");
return EXIT_SUCCESS;
}
<braunr> youpi: i'd see two calls to write, the second because of a signal,
as normal, as long as the second call resumes, but not restarts after
finishing :/
<braunr> or restarts because nothing was done (or everything was entirely
rolled back)
<youpi> well, with an RPC you may not be sure whether it's finished or not
<braunr> ah
<youpi> we don't really have rollback
<braunr> i don't really see the difference with a syscall there
<youpi> the kernel controls the interruption in the case of the syscall
<braunr> except that write is normally atomic if i'm right
<youpi> it can't happen on the way back to userland
<braunr> but that could be exactly the same with RPCs
<youpi> while perhaps it can happen on the mach_msg back to userland
<braunr> back to userland ok, back to the application, no
<braunr> anyway, that's a side issue
<braunr> i'm fixing a few bugs in libpthread
<braunr> and noticed that
<braunr> (i should soon have patches to fix - at least partially - thread
cancellation and timed blocking)
<braunr> i was just wondering how cancellation how handled in glibc wrt
libpthread
<youpi> I don't know
<braunr> (because the non standard hurd cancellation has nothing to do with
pthread cancellation)à
<braunr> ok
<braunr> s/how h/is h/
IRC, freenode, #hurd, 2013-01-15
<tschwinge> braunr: Re »one call to printf to write twice its output«:
sounds familiar:
http://www.gnu.org/software/hurd/open_issues/git_duplicated_content.html
and http://www.gnu.org/software/hurd/open_issues/git-core-2.html
<braunr> tschwinge: what i find strange with the duplicated operations i've
seen is that i merely use pthreads and printf, nothing else
<braunr> no setitimer, no alarm, no select
<braunr> so i wonder how cancellation/syscall restart is actually handled
in our glibc
<braunr> but i agree with you on the analysis
IRC, freenode, #hurd, 2013-01-16
<braunr> neal: do you (by any chance) remember if there could possibly be
spurious wakeups in your libpthread implementation ?
<neal> braunr: There probably are.
<neal> but I don't recall
<braunr> i think the duplicated content issue is due to the libmach/glibc
mach_msg wrapper
<braunr> which restarts a message send if interrupted
<tschwinge> Hrm, depending on which point it has been interrupted you mean?
<braunr> yes
<braunr> not sure yet and i could be wrong
<braunr> but i suspect that if interrupted after send and during receive,
the restart might be wrongfully done
<braunr> i'm currently reworking the timed* pthreads functions, doing the
same kind of changes i did last summer when working on select (since
implement the timeout at the server side requires pthread_cond_timedwait)
<braunr> and i limit the message queue size of the port used to wake up
threads to 1
<braunr> and it seems i have the same kind of problems, i.e. blocking
because of a second, unexpected send
<braunr> i'll try using __mach_msg_trap directly and see how it goes
<tschwinge> Hrm, mach/msg.c:__mach_msg does look correct to me, but yeah,
won't hurd to confirm this by looking what direct usage of
__mach_msg_trap is doing.
<braunr> tschwinge: can i ask if you still have a cthreads based hurd
around ?
<braunr> tschwinge: and if so, to send me libthreads.so.0.3 ... :)
<tschwinge> braunr: darnassus:~tschwinge/libthreads.so.0.3
<braunr> call 19c0 <mach_msg@plt>
<braunr> so, cthreads were also using the glibc wrapper
<braunr> and i never had a single MACH_SEND_INTERRUPTED
<braunr> or a busy queue :/
<braunr> (IOW, no duplicated messages, and the wrapper indeed looks
correct, so it's something else)
<tschwinge> (Assuming Mach is doing the correct thing re interruptions, of
course...)
<braunr> mach doesn't implement it
<braunr> it's explicitely meant to be done in userspace
<braunr> mach merely reports the error
<braunr> i checked the osfmach code of libmach, it's almost exactly the
same as ours
<tschwinge> Yeah, I meant Mach returns the interurption code but anyway
completed the RPC.
<braunr> ok
<braunr> i don't expect mach wouldn't do it right
<braunr> the only difference in osf libmach is that, when retrying,
MACH_SEND_INTERRUPT|MACH_RCV_INTERRUPT are both masked (for both the
send/send+receive and receive cases)
<tschwinge> Hrm.
<braunr> but they say it's for performance, i.e. mach won't take the slow
path because of unexpected bits in the options
<braunr> we probably should do the same anyway
IRC, freenode, #hurd, 2013-01-17
<braunr> tschwinge: i think our duplicated RPCs come from
hurd/intr-msg.c:148 (err == MACH_SEND_INTERRUPTED but !(option &
MACH_SEND_MSG))
<braunr> a thread is interrupted by a signal meant for a different thread
<braunr> hum no, still not that ..
<braunr> or maybe .. :)
<tschwinge> Hrm. Why would it matter for for the current thread for which
reason (different thread) mach_msg_trap returns *_INTERRUPTED?
<braunr> mach_msg wouldn't return it, as explained in the comment
<braunr> the signal thread would, to indicate the send was completed but
the receive must be retried
<braunr> however, when retrying, the original user_options are used again,
which contain MACH_SEND_MSG
<braunr> i'll test with a modified version that masks it
<braunr> tschwinge: hm no, doesn't fix anything :(
IRC, freenode, #hurd, 2013-01-18
<braunr> the duplicated rpc calls is one i find very very frustrating :/
<youpi> you mean the dup writes we've seen lately?
<braunr> yes
<youpi> k
IRC, freenode, #hurd, 2013-01-19
<braunr> all right, i think the duplicated message sends are due to thread
creation
<braunr> the duplicated message seems to be sent by the newly created
thread
<braunr> arg no, misread
IRC, freenode, #hurd, 2013-01-20
<braunr> tschwinge: youpi: about the diplucated messages issue, it seems to
be caused by two threads (with pthreads) doing an rpc concurrently
<braunr> duplicated*
IRC, freenode, #hurd, 2013-01-21
<braunr> ah, found something interesting
<braunr> tschwinge: there seems to be a race on our file descriptors
<braunr> the content written by one thread seems to be retained somewhere
and another thread writing data to the file descriptor will resend what
the first already did
<braunr> it could be a FILE race instead of fd one though
<braunr> yes, it's not at the fd level, it's above
<braunr> so good news, seems like the low level message/signalling code
isn't faulty here
<braunr> all right, simple explanation: our IO_lockfile functions are
no-ops
<pinotree> braunr: i found that out days ago, and samuel said they were
okay
glibc, flockfile
/ftrylockfile
/funlockfile
.
IRC, freenode, #hurd, 2013-01-15
<braunr> hmm, looks like subhurds have been broken by the pthreads patch :/
<braunr> arg, we really do have broken subhurds :((
<braunr> time for an immersion in the early hurd bootstrapping stuff
<tschwinge> Hrm. Narrowed down to cthreads -> pthread you say.
<braunr> i think so
<braunr> but i think the problem is only exposed
<braunr> it was already present before
<braunr> even for the main hurd, i sometimes have systems blocking on exec
<braunr> there must be a race there that showed far less frequently with
cthreads
<braunr> youpi: we broke subhurds :/
<youpi> ?
<braunr> i can't start one
<braunr> exec seems to die and prevent the root file system from
progressing
<braunr> there must be a race, exposed by the switch to pthreads
<braunr> arg, looks like exec doesn't even reach main :(
<braunr> now, i'm wondering if it could be the tls support that stops exec
<braunr> although i wonder why exec would start correctly on a main hurd,
and not on a subhurd :(
<braunr> i even wonder how much progress ld.so.1 is able to make, and don't
have much idea on how to debug that
IRC, freenode, #hurd, 2013-01-22
<braunr> hm, subhurds seem to be broken because of select
<braunr> damn select !
<braunr> hm i see, we can't boot a subhurd that still uses libthreads from
a main hurd that doesn't
<braunr> the linker can't find it and doesn't start exec
<braunr> pinotree: do you understand what the fmh function does in
sysdeps/mach/hurd/dl-sysdep.c ?
<braunr> i think we broke subhurds by fixing vm_map with size 0
<pinotree> braunr: no idea, but i remember thomas talking about this code
<braunr> it checks for KERN_INVALID_ADDRESS and KERN_NO_SPACE
<braunr> and calls assert_perror(err); to make sure it's one of them
<braunr> but now, KERN_INVALID_ARGUMENT can be returned
<braunr> ok i understand what it does
<braunr> and youpi has changed the code, so he does too
<braunr> (now i'm wondering why he didn't think of it when we fixed vm_map
size with 0 but his head must already be filled with other things so ..)
<braunr> anyway, once this is dealt with, we get subhurds back :)
<braunr> yes, with a slight change, my subhurd starts again \o/
<braunr> youpi: i found the bug that prevents subhurds from booting
<braunr> it's caused by our fixing of vm_map with size 0
<braunr> when ld.so.1 starts exec, the code in
sysdeps/mach/hurd/dl-sysdep.c fails because it doesn't expect the new
error code we introduced
<braunr> (the fmh functions)
<youpi> ah :)
<youpi> good :)
<braunr> adding KERN_INVALID_ARGUMENT to the list should do the job, but if
i understand the code correctly, checking if fmhs isn't 0 before calling
vm_map should do the work too
<braunr> s/do the work/work/
<braunr> i'm not sure which is the preferred way
<youpi> otherwise I believe fmh could be just fixed to avoid calling vm_map
in the !fmhs case
<braunr> yes that's what i currently do
<braunr> at the start of the loop, just after computing it
<braunr> seems to work so far
IRC, freenode, #hurd, 2013-01-22
<braunr> i have almost completed fixing both cancellation and timeout
handling, but there are still a few bugs remaining
<braunr> fyi, the related discussion was
https://lists.gnu.org/archive/html/bug-hurd/2012-08/msg00057.html
IRC, freenode, #hurd, 2014-01-01
<youpi> braunr: I have an issue with tls_thread_leak
<youpi> int main(void) {
<youpi> pthread_create(&t, NULL, foo, NULL);
<youpi> pthread_exit(0);
<youpi> }
<youpi> this fails at least with the libpthread without your libpthread
thread termination patch
<youpi> because for the main thread, tcb->self doesn't contain thread_self
<youpi> where is tcb->self supposed to be initialized for the main thread?
<youpi> there's also the case of fork()ing from main(), then calling
pthread_exit()
<youpi> (calling pthread_exit() from the child)
<youpi> the child would inherit the tcb->self value from the parent, and
thus pthread_exit() would try to kill the father
<youpi> can't we still do tcb->self = self, even if we don't keep a
reference over the name?
<youpi> (the pthread_exit() issue above should be fixed by your thread
termination patch actually)
<youpi> Mmm, it seems the thread_t port that the child inherits actually
properly references the thread of the child, and not the thread of the
father?
<youpi> “For the name we use for our own thread port, we will insert the
thread port for the child main user thread after we create it.” Oh, good
:)
<youpi> and, “Skip the name we use for any of our own thread ports.”, good
too :)
<braunr> youpi: reading
<braunr> youpi: if we do tcb->self = self, we have to keep the reference
<braunr> this is strange though, i had tests that did exactlt what you're
talking about, and they didn't fail
<youpi> why?
<braunr> if you don't keep the reference, it means you deallocate self
<youpi> with the thread termination patch, tcb->self is not used for
destruction
<braunr> hum
<braunr> no it isn't
<braunr> but it must be deallocated at some point if it's not temporary
<braunr> normally, libpthread should set it for the main thread too, i
don't understand
<youpi> I don't see which code is supposed to do it
<youpi> sure it needs to be deallocated at some point
<youpi> but does tcb->self has to wear the reference?
<braunr> init_routine should do it
<braunr> it calls __pthread_create_internal
<braunr> which allocates the tcb
<braunr> i think at some point, __pthread_setup should be called for it too
<youpi> but what makes pthread->kernel_thread contain the port for the
thread?
<braunr> but i have to check that
<braunr> __pthread_thread_alloc does that
<braunr> so normally it should work
<braunr> is your libpthread up to date as well ?
<youpi> no, as I said it doesn't contain the thread destruction patch
<braunr> ah
<braunr> that may explain
<youpi> but the tcb->self uninitialized issue happens on darnassus too
<youpi> it just doesn't happen to crash because it's not used
<braunr> that's weird :/
<youpi> see ~youpi/test.c there for instance
<braunr> humpf
<braunr> i don't see why :/
<braunr> i'll debug that later
<braunr> youpi: did you find the problem ?
<youpi> no
<youpi> I'm working on fixing the libpthread hell in the glibc debian
package :)
<youpi> i.e. replace a dozen patches with a git snapshot
<braunr> ah you reverted commit
<braunr> +a
<braunr> i imagine it's hairy :)
<youpi> not too much actually
<braunr> wow :)
<youpi> with the latest commits, things have converged
<youpi> it's now about small build details
<youpi> I just take time to make sure I'm getting the same source code in
the end :)
<braunr> :)
<braunr> i hope i can determine what's going wrong tonight
<braunr> youpi: avec mach_print, je vois bien self setté par la libpthread
..
<youpi> mais à autre chose que 0 ?
<braunr> oui
<braunr> bizarrement, l'autre thread n'as pas la même valeur
<braunr> tu es bien sûr que c'est self que tu affiches avec l'assembleur ?
<braunr> oops, english
<youpi> see test2
<youpi> so I'm positive
<braunr> well, there obviously is a bug
<braunr> but are you certain your assembly code displays the thread port
name ?
<youpi> I'm certain it displays tcb->self
<braunr> oh wait, hexadecimal, ok
<youpi> and the value happens to be what mach_thread_self returns
<braunr> ah right
<youpi> ah, right, names are usually decimals :)
<braunr> hm
<braunr> what's the problem with test2 ?
<youpi> none
<braunr> ok
<youpi> I was just checking what happens on fork from another thread
<braunr> ok i do have 0x68 now
<braunr> so the self field gets erased somehow
<braunr> 15:34 < youpi> this fails at least with the libpthread without
your libpthread thread termination patch
<braunr> how does it fail ?
<youpi> ../libpthread/sysdeps/mach/pt-thread-halt.c:44:
__pthread_thread_halt: Unexpected error: (ipc/send) invalid destination
port.
<braunr> hm
<braunr> i don't have that problem on darnassus
<youpi> with the new libc?
<braunr> the pthread destruction patch actually doesn't use the tcb->self
name if i'm right
<braunr> yes
<braunr> what is tcb->self used for ?
<youpi> it used to be used by pt-thread-halt
<youpi> but is darnassus using your thread destruction patch?
<youpi> as I said, since your thread destruction pathc doesn't use
tcb->self, it doesn't have the issue
<braunr> the patched libpthread merely uses the sysdeps kernel_thread
member
<braunr> ok
<youpi> it's the old libpthread against the new libc which has issues
<braunr> yes it is
<braunr> so for me, the only thing to do is make sure tcb->self remains
valid
<braunr> we could simply add a third user ref but i don't like the idea
<youpi> well, as you said the issue is rather that tcb->self gets
overwritten
<youpi> there is no reason why it should
<braunr> the value is still valid when init_routine exits, so it must be in
libc
<youpi> or perhaps for some reason tls gets initialized twice
<braunr> maybe
<youpi> and thus what libpthread's init writes to is not what's used later
<braunr> i've add a print in pthread_create, to see if self actually got
overwritten
<braunr> and it doesn't
<braunr> there is a disrepancy between the tcb member in libpthread and
what libc uses for tls
<braunr> added*
<braunr> (the print is at the very start of pthread_create, and displays
the thread name of the caller only)
<youpi> well, yes, for the main thread libpthread shouldn't be allocating a
new tcb
<youpi> and just use the existing one
<braunr> ?
<youpi> the main thread's tcb is initialized before the threading library
iirc
<braunr> hmm
<braunr> it would make sense if we actually had non-threaded programs :)
<youpi> at any rate, the address of the tcb allocated by libpthread is not
put into registers
<braunr> how does it get there for the other threads ?
<youpi> __pthread_setup does it
<braunr> so
<braunr> looks like dl_main is called after init_routine
<braunr> and it then calls init_tls
<braunr> init_tls returns the tcb for the main thread, and that's what
overrides the libpthread one
<youpi> yes, _hurd_tls_init is called very early, before init_routine
<youpi> __pthread_create_internal could fetch the tcb pointer from gs:0
when it's the main thread
<braunr> so there is something i didn't get right
<braunr> i thought _hurd_tls_init was called as part of dl_main
<youpi> well, it's not a bug of yours, it has always been bug :)
<braunr> which is called *after* init_routine
<braunr> and that explains why the libpthread tcb isn't the one installed
in the thread register
<braunr> i can actually check that quite easily
<youpi> where do you see dl_main called after init_routine?
<braunr> well no i got that wrong somehow
<braunr> or i'm unable to find it again
<braunr> let's see
<braunr> init_routine is called by init which is called by _dl_init_first
<braunr> which i can only find in the macro RTLD_START_SPECIAL_INIT
<braunr> with print traces, i see dl_main called before init_routine
<braunr> so yes, libpthread should reuse it
<braunr> the tcb isn't overriden, it's just never installed
<braunr> i'm not sure how to achieve that cleanly
<youpi> well, it is installed, by _hurd_tls_init
<youpi> it's the linker which creates the main thread's tcb
<youpi> and calls _hurd_tls_init to install it
<youpi> before the thread library enters into action
<braunr> agreed
IRC, freenode, #hurd, 2014-01-14
<braunr> btw, are you planning to do something with regard to the main
thread tcb initialization issue ?
<youpi> well, I thought you were working on it
<braunr> ok
<braunr> i wasn't sure
IRC, freenode, #hurd, 2014-01-19
<braunr> i have some fixup code for the main thread tcb
<braunr> but it sometimes crashes on tcb deallocation
<braunr> is there anything particular that you would know about the tcb of
the main thread ?
<braunr> (that could help explaining this)
<youpi> Mmmm, I don't think there is anything particular
<braunr> doesn't look like the first tcb can be reused safely
<braunr> i think we should instead update the thread register to point to
the pthread tcb
<youpi> what do you mean by "the first tcb" exactly?
IRC, freenode, #hurd, 2014-01-03
<gg0> braunr: hurd from your repo can't boot. restored debian one
<braunr> gg0: it does boot
<braunr> gg0: but you need everything (gnumach and glibc) in order to make
it work
<braunr> i think youpi did take care of compatibility with older kernels
<teythoon> braunr: so do we need a rebuilt libc for the latest hurd from
git ?
<braunr> teythoon: no, the hurd isn't the problem
<teythoon> ok
<teythoon> good
<braunr> the problem is the libports_stability patch
<teythoon> what about it ?
<braunr> the hurd can't work correctly without it since the switch to
pthreads
<braunr> because of subtle bugs concerning resource recycling
<teythoon> ok
<braunr> these have been fixed recently by youpi and me (youpi fixed them
exactly as i did, which made my life very easy when merging :))
<braunr> there is also the problem of the stack sizes, which means the hurd
servers could use 2M stacks with an older glibc
<braunr> or perhaps it chokes on an error when attempting to set the stack
size because it was unsupported
<braunr> i don't know
<braunr> that may be what gg0 suffered from
<gg0> yes, both gnumach and eglibc were from debian. seems i didn't
manually upgrade eglibc from yours
<gg0> i'll reinstall them now. let's screw it up once again
<braunr> :)
<braunr> bbl
<gg0> ok it boots
<gg0> # apt-get install
{hurd,hurd-dev,hurd-libs0.3}=1:0.5.git20131101-1+rbraun.7
{libc0.3,libc0.3-dev,libc0.3-dbg,libc-dev-bin}=2.17-97+hurd.0+rbraun.1+threadterm.1
<gg0> there must a simpler way
<gg0> besides apt-pinning
<gg0> making it a real "experimental" release might help with -t option for
instance
<gg0> btw locales still segfaults
<gg0> rpctrace from teythoon gets stuck at
http://paste.debian.net/plain/74072/
<gg0> ("rpctrace locale-gen", last 300 lines)