Thread: Re: [Quickfix-developers] Build problems on Solaris
Brought to you by:
orenmnero
Menu
▾
▴
quickfix-developers
|
From: Joerg T. <Joe...@ma...> - 2003-04-03 11:35:00
|
Hi Barry,
this is an easy one...
> The only problem is that I don't seem able to execute pbind properly against
> any process that I own. Insufficient rights I am guessing. For example:
No, invalid argument as pbind tells you.
> bash-2.03$ /usr/sbin/psrinfo
> 1 on-line since 07/19/02 18:35:00
> 2 on-line since 07/19/02 18:35:02
From this output, you see that you have processors 1 and 2 (see first column).
> bash-2.03$ /usr/sbin/pbind -b 0 3761
> /usr/sbin/pbind: cannot bind pid 3761: Invalid argument
But you try to bind to processor 0 (-b 0) which simply does not exist --
therefore the invalid argument.
I am sorry for the rather sketchy explanation in my last mail. I thought that it was
obvious to use psrinfo to look the available processor number (at my Sun: 0, but at
your Sun: 1 and 2) and I also thought that there is always a processor 0.
Nevertheless, feel free to ask further questions.
Actually, we used this approach to hotfix a threading problem under heavy load on a
multiprocessor machine while development took place on a single processor one.
But later we discovered, that we had used C library functions which are not thread save.
After we fixed this, we could drop the pbind.
For all non-thread save functions, there are thread-safe (re-entrant) variants suffixed
by "_r". In our case, gethostbyname() was one of the culprits. We replaced it by
gethostbyname_r(), which lead to a somewhat more complex call syntax.
Oren, perhaps QF uses some non-thread safe functions (e.g. gethostbyname()). In my
experience, this can work for a long time, but if you switch to a real fast MP machine
(for production), the program starts to dump core.
On my machine, I looked in the following way for all "_r" suffixed functions:
joerg@polaris:/usr/lib $ nm -A -g -n lib*.so | egrep '_r$' | grep -v UNDEF
libXm.so: [6411] | 1542716| 252|FUNC |GLOB |0 |12 |XmFontListCreate_r
libXm.so: [7899] | 1542040| 256|FUNC |GLOB |0 |12 |XmFontListEntryCreate_r
libXm.so: [8495] | 1542980| 12|FUNC |GLOB |0 |12 |XmStringCreateFontList_r
libXm.so: [6958] | 483476| 96|FUNC |GLOB |0 |12 |_XmReCacheLabG_r
libbsm.so: [865] | 30892| 408|FUNC |GLOB |0 |9 |getauclassent_r
libbsm.so: [586] | 31364| 76|FUNC |GLOB |0 |9 |getauclassnam_r
libbsm.so: [587] | 36924| 424|FUNC |GLOB |0 |9 |getauevent_r
libbsm.so: [727] | 37412| 92|FUNC |GLOB |0 |9 |getauevnam_r
libbsm.so: [814] | 37504| 88|FUNC |GLOB |0 |9 |getauevnum_r
libbsm.so: [826] | 54888| 424|FUNC |GLOB |0 |9 |getauuserent_r
libbsm.so: [583] | 55368| 168|FUNC |GLOB |0 |9 |getauusernam_r
libc.so: [2699] | 227336| 332|FUNC |GLOB |0 |9 |__posix_asctime_r
libc.so: [3019] | 227852| 68|FUNC |GLOB |0 |9 |__posix_ctime_r
libc.so: [3752] | 608620| 100|FUNC |GLOB |0 |9 |__posix_getgrgid_r
libc.so: [4553] | 608856| 92|FUNC |GLOB |0 |9 |__posix_getgrnam_r
libc.so: [4510] | 240376| 120|FUNC |GLOB |0 |9 |__posix_getlogin_r
libc.so: [3910] | 611572| 100|FUNC |GLOB |0 |9 |__posix_getpwnam_r
libc.so: [3028] | 611336| 100|FUNC |GLOB |0 |9 |__posix_getpwuid_r
libc.so: [3346] | 615168| 200|FUNC |GLOB |0 |9 |__posix_readdir_r
libc.so: [3199] | 349788| 148|FUNC |GLOB |0 |9 |__posix_ttyname_r
libc.so: [4378] | 227668| 48|FUNC |GLOB |0 |9 |_asctime_r
libc.so: [3478] | 615368| 48|FUNC |GLOB |0 |9 |_ctermid_r
libc.so: [4082] | 227780| 72|FUNC |GLOB |0 |9 |_ctime_r
libc.so: [3184] | 609252| 124|FUNC |GLOB |0 |9 |_fgetgrent_r
libc.so: [3038] | 611976| 124|FUNC |GLOB |0 |9 |_fgetpwent_r
libc.so: [3294] | 613520| 124|FUNC |GLOB |0 |9 |_fgetspent_r
libc.so: [3835] | 609044| 208|FUNC |GLOB |0 |9 |_getgrent_r
libc.so: [3797] | 607960| 256|FUNC |GLOB |0 |9 |_getgrgid_r
libc.so: [4619] | 607628| 332|FUNC |GLOB |0 |9 |_getgrnam_r
libc.so: [3589] | 240080| 296|FUNC |GLOB |0 |9 |_getlogin_r
libc.so: [4636] | 611768| 208|FUNC |GLOB |0 |9 |_getpwent_r
libc.so: [3228] | 610772| 244|FUNC |GLOB |0 |9 |_getpwnam_r
libc.so: [4694] | 611016| 184|FUNC |GLOB |0 |9 |_getpwuid_r
libc.so: [3016] | 613312| 208|FUNC |GLOB |0 |9 |_getspent_r
libc.so: [3807] | 613080| 136|FUNC |GLOB |0 |9 |_getspnam_r
libc.so: [3116] | 336028| 72|FUNC |GLOB |0 |9 |_gmtime_r
libc.so: [3720] | 335940| 68|FUNC |GLOB |0 |9 |_localtime_r
libc.so: [3929] | 614524| 92|FUNC |GLOB |0 |9 |_rand_r
libc.so: [3110] | 614616| 348|FUNC |GLOB |0 |9 |_readdir64_r
libc.so: [3092] | 614964| 204|FUNC |GLOB |0 |9 |_readdir_r
libc.so: [4684] | 324656| 128|FUNC |GLOB |0 |9 |_strtok_r
libc.so: [3166] | 616336| 176|FUNC |GLOB |0 |9 |_tmpnam_r
libc.so: [4325] | 349056| 732|FUNC |GLOB |0 |9 |_ttyname_r
libc.so: [3870] | 608484| 136|FUNC |GLOB |0 |9 |_uncached_getgrgid_r
libc.so: [4153] | 608720| 136|FUNC |GLOB |0 |9 |_uncached_getgrnam_r
libc.so: [4681] | 611436| 136|FUNC |GLOB |0 |9 |_uncached_getpwnam_r
libc.so: [4479] | 611200| 136|FUNC |GLOB |0 |9 |_uncached_getpwuid_r
libc.so: [3587] | 227668| 48|FUNC |WEAK |0 |9 |asctime_r
libc.so: [3559] | 615368| 48|FUNC |WEAK |0 |9 |ctermid_r
libc.so: [2899] | 227780| 72|FUNC |WEAK |0 |9 |ctime_r
libc.so: [4579] | 609252| 124|FUNC |WEAK |0 |9 |fgetgrent_r
libc.so: [3187] | 611976| 124|FUNC |WEAK |0 |9 |fgetpwent_r
libc.so: [3755] | 613520| 124|FUNC |WEAK |0 |9 |fgetspent_r
libc.so: [3185] | 609044| 208|FUNC |WEAK |0 |9 |getgrent_r
libc.so: [2760] | 607960| 256|FUNC |WEAK |0 |9 |getgrgid_r
libc.so: [3782] | 607628| 332|FUNC |WEAK |0 |9 |getgrnam_r
libc.so: [4251] | 240080| 296|FUNC |WEAK |0 |9 |getlogin_r
libc.so: [3271] | 243412| 184|FUNC |GLOB |0 |9 |getnetgrent_r
libc.so: [3520] | 611768| 208|FUNC |WEAK |0 |9 |getpwent_r
libc.so: [4097] | 610772| 244|FUNC |WEAK |0 |9 |getpwnam_r
libc.so: [3320] | 611016| 184|FUNC |WEAK |0 |9 |getpwuid_r
libc.so: [3851] | 613312| 208|FUNC |WEAK |0 |9 |getspent_r
libc.so: [4436] | 613080| 136|FUNC |WEAK |0 |9 |getspnam_r
libc.so: [3362] | 336028| 72|FUNC |WEAK |0 |9 |gmtime_r
libc.so: [3898] | 335940| 68|FUNC |WEAK |0 |9 |localtime_r
libc.so: [3182] | 614524| 92|FUNC |WEAK |0 |9 |rand_r
libc.so: [3279] | 614616| 348|FUNC |WEAK |0 |9 |readdir64_r
libc.so: [3418] | 614964| 204|FUNC |WEAK |0 |9 |readdir_r
libc.so: [3288] | 324656| 128|FUNC |WEAK |0 |9 |strtok_r
libc.so: [4516] | 616336| 176|FUNC |WEAK |0 |9 |tmpnam_r
libc.so: [3538] | 349056| 732|FUNC |WEAK |0 |9 |ttyname_r
libm.so: [594] | 18020| 16|FUNC |GLOB |0 |9 |__gamma_r
libm.so: [612] | 28796| 16|FUNC |GLOB |0 |9 |__lgamma_r
libm.so: [658] | 18020| 16|FUNC |WEAK |0 |9 |gamma_r
libm.so: [632] | 28796| 16|FUNC |WEAK |0 |9 |lgamma_r
libnsl.so: [3766] | 435368| 52|FUNC |GLOB |0 |9 |__nis_map_group_r
libnsl.so: [3773] | 234608| 160|FUNC |GLOB |0 |9 |_switch_gethostbyaddr_r
libnsl.so: [3501] | 119800| 148|FUNC |GLOB |0 |9 |_switch_gethostbyname_r
libnsl.so: [3929] | 234768| 160|FUNC |GLOB |0 |9 |_switch_getipnodebyaddr_r
libnsl.so: [3543] | 234460| 148|FUNC |GLOB |0 |9 |_switch_getipnodebyname_r
libnsl.so: [3646] | 216360| 48|FUNC |GLOB |0 |9 |_uncached_gethostbyaddr_r
libnsl.so: [4061] | 216348| 12|FUNC |GLOB |0 |9 |_uncached_gethostbyname_r
libnsl.so: [4253] | 216612| 200|FUNC |GLOB |0 |9 |gethostbyaddr_r
libnsl.so: [3714] | 216408| 204|FUNC |GLOB |0 |9 |gethostbyname_r
libnsl.so: [4141] | 218508| 144|FUNC |GLOB |0 |9 |gethostent_r
libnsl.so: [4225] | 228912| 148|FUNC |GLOB |0 |9 |getrpcbyname_r
libnsl.so: [3561] | 229060| 148|FUNC |GLOB |0 |9 |getrpcbynumber_r
libnsl.so: [4424] | 229328| 132|FUNC |GLOB |0 |9 |getrpcent_r
libnsl.so: [3817] | 114092| 76|FUNC |GLOB |0 |9 |inet_ntoa_r
libnsl.so: [3526] | 415004| 96|FUNC |GLOB |0 |9 |nis_leaf_of_r
libnsl.so: [4178] | 430712| 176|FUNC |GLOB |0 |9 |nis_sperror_r
libsocket.so: [273] | 20944| 152|FUNC |GLOB |0 |9 |getnetbyaddr_r
libsocket.so: [261] | 20796| 148|FUNC |GLOB |0 |9 |getnetbyname_r
libsocket.so: [343] | 21220| 132|FUNC |GLOB |0 |9 |getnetent_r
libsocket.so: [307] | 22304| 148|FUNC |GLOB |0 |9 |getprotobyname_r
libsocket.so: [234] | 22452| 148|FUNC |GLOB |0 |9 |getprotobynumber_r
libsocket.so: [463] | 22716| 132|FUNC |GLOB |0 |9 |getprotoent_r
libsocket.so: [458] | 23488| 156|FUNC |GLOB |0 |9 |getservbyname_r
libsocket.so: [464] | 23644| 160|FUNC |GLOB |0 |9 |getservbyport_r
libsocket.so: [378] | 24268| 136|FUNC |GLOB |0 |9 |getservent_r
Cheers, Jörg
--
Joerg Thoennes
http://macd.com
Tel.: +49 (0)241 44597-24 Macdonald Associates GmbH
Fax : +49 (0)241 44597-10 Lothringer Str. 52, D-52070 Aachen
|
|
From: Joerg T. <Joe...@ma...> - 2003-04-03 11:47:10
|
Hi Barry,
> No, invalid argument as pbind tells you.
>
>> bash-2.03$ /usr/sbin/psrinfo
>> 1 on-line since 07/19/02 18:35:00
>> 2 on-line since 07/19/02 18:35:02
>
>
> From this output, you see that you have processors 1 and 2 (see first
> column).
>
>> bash-2.03$ /usr/sbin/pbind -b 0 3761
>> /usr/sbin/pbind: cannot bind pid 3761: Invalid argument
>
>
> But you try to bind to processor 0 (-b 0) which simply does not exist --
> therefore the invalid argument.
Please, can you confirm whether this works now?
Does it fix your problem?
> Actually, we used this approach to hotfix a threading problem under
> heavy load on a
> multiprocessor machine while development took place on a single
> processor one.
> But later we discovered, that we had used C library functions which are
> not thread save.
> After we fixed this, we could drop the pbind.
>
> For all non-thread save functions, there are thread-safe (re-entrant)
> variants suffixed
> by "_r". In our case, gethostbyname() was one of the culprits. We
> replaced it by
> gethostbyname_r(), which lead to a somewhat more complex call syntax.
Perhaps you have been experiencing exactly this kind of situation. It would be
good to have a stack trace of core dump:
$ pstack core
In the stack trace we could see at which place the error occured. For more such
nice utilities see "man proc", "man truss", "man gcore", "man coreadm".
In addition, beginning with Solaris 2.8 there are two thread libraries:
1. Old T1 lib in /lib/libthread.so implements complex N:M mappings of N user threads
to M<<N kernel threads callend LWPs( Light Weight Processes ). Due to the complexity
there are many issues in the bug database.
2. New T2 lib in /lib/lwp/libthread.so. This simply maps every user thread to an LWP.
This is much simpler and new kernels can handle lots of LWPs. Therefore, this is the
default in Solaris 9.
Prepend /usr/lib/lwp to your LD_LIBRARY_PATH to activate it:
export LD_LIBRARY_PATH=/usr/lib/lwp:$LD_LIBRARY_PATH
or
setenv LD_LIBRARY_PATH /usr/lib/lwp:$LD_LIBRARY_PATH
In our experience, if you mix Java application and JNI/native libs which create threads
on their own, this is always a good choice to avoid threading problem by buggy libs.
Cheers, Jörg
--
Joerg Thoennes
http://macd.com
Tel.: +49 (0)241 44597-24 Macdonald Associates GmbH
Fax : +49 (0)241 44597-10 Lothringer Str. 52, D-52070 Aachen
|
×
Want the latest updates on software, tech news, and AI?
Get latest updates about software, tech news, and AI from SourceForge directly in your inbox once a month.
Thanks for helping keep SourceForge clean.
X