-0 MB memory
Anonymous
26 Feb 2005 15:28:46 UTC
Topic 12733
(moderation:
> I installed Boing client 2.19 on two Linux-boxes. Installation was easy and
> OK. Now, after launching "boinc_4.19_i686-pc-linux-gnu" I get a message: No
> work sent ( there was work but your computer doesn't have enough memory).
> Searching in my account, I find the memory of both boxes given with -0 MB !
> However, the first box has at that moment a physical memory availaible of near
> 140 MB, the second one has had some 112 MB. What is going wrong ? And a new
> tryial is only possible after one day ! ( Einstein-Home)
> Could you advise me ? Thank you in advance
I hope you meant the BOINC client 4.19. Since it appears to be incorrectly reporting the physical memory, I would guess that something sent wrong when BOINC was installed and recorded the memory.
Here are two things to try. First, check that the user who is running BOINC can read meminfo.
Then try uninstalling and re-installing BOINC (eg, remove all the .xml files from the BOINC working directory). Finally, if both of these fail, I suggest you try the latest BOINC experimental client for Linux (probably 4.24) which you can get by going to the download page and clicking the link at the bottom that says something like 'also show development versions'. If you do this be sure that the boinc 4.19 client is stopped before you start the new boinc client (and client GUI).
Bruce
-0 MB memory
)
> I installed Boing client 2.19 on two Linux-boxes. Installation was easy and
> OK. Now, after launching "boinc_4.19_i686-pc-linux-gnu" I get a message: No
> work sent ( there was work but your computer doesn't have enough memory).
> Searching in my account, I find the memory of both boxes given with -0 MB !
> However, the first box has at that moment a physical memory availaible of near
> 140 MB, the second one has had some 112 MB. What is going wrong ? And a new
> tryial is only possible after one day ! ( Einstein-Home)
> Could you advise me ? Thank you in advance
I hope you meant the BOINC client 4.19. Since it appears to be incorrectly reporting the physical memory, I would guess that something sent wrong when BOINC was installed and recorded the memory.
Try uninstalling and re-installing BOINC (eg, remove all the .xml files from the BOINC working directory). If this fails to show more memory, I suggest you try the latest BOINC experimental client for Linux (probably 4.24) which you can get by going to the download page and clicking the link at the bottom that says something like 'also show development versions'. If you do this be sure that the boinc 4.19 client is stopped before you start the new boinc client (and client GUI).
Bruce
> Hi Bruce, > > also with
)
> Hi Bruce,
>
> also with new installation of experimental version 4.24 the amount of memory
> reported by "computers on this account" continues beeing -0 MB.
> /proc/meminfo has the permissions -r--r--r--, which should be OK. How does
> boinc transmit the memory data ? If I give "/proc/meminfo" I get "permission
> denied", either as root or as user. Only "cat /proc/meminfo" gives results.
> I am behind an Allnet router, but I don't think this could be of influence, as
> all other data concerning my PC's has been cleanly transmitted.
Thank you for following up. Here's where BOINC is supposed to get this.
http://boinc.berkeley.edu/cgi-bin/cvsweb.cgi/boinc/client/hostinfo_unix.C?rev=1.94&content-type=text/x-cvsweb-markup
code is:
#elif defined(_SC_PHYS_PAGES)
m_nbytes = (double)sysconf(_SC_PAGESIZE) * (double)sysconf(_SC_PHYS_PAGES); /* Linux */
Do you have a way of determining what this returns on your system?
Cheers,
Bruce
> I could not find sysconf
)
> I could not find sysconf defines in my unistd.h file. Unfortunately, I am not
> an C programmer, although my system is a linux one I compiled from the
> scratch. Would it be possible for you to give me an extract of the programm
> which I could compile and run and that prints the returned memory information,
> so I could provide you with the info my system is returning ?
I'm adding this to the FAQ on the front page (currently the very last item). The forums strip out too many characters for me to put the code here.
Bruce
> Hi Bruce, > > thank you
)
> Hi Bruce,
>
> thank you very much for the smalll program. Compilation is Ok, with a little
> warning:memory.c:27:2: warning: no newline at end of file.
> The results of running the programm are:
> quote
> -------
> bash-2.05b$ ./memory
> size = 4096 number = -1 bytes=-4096
> bash-2.05b$
>> And I am dealing with the box ID 48907.
Edgar,
No need for header files, thanks!
[1] what Linux distribution are you using?
[2] Please do:
man sysconf
and see if it documents _SC_PHYS_PAGES
If it does, then you need to file a bug report with the distribution, saying that memory is being reported incorrectly and that sysconf doesn't work as advertised. You can send them memory.c as an example.
Bruce
> Hi Bruce, > > > [1] what
)
> Hi Bruce,
>
> > [1] what Linux distribution are you using?
>
> Linux from the Scratch, Version 5.1.1 with Kernel 2.4.26. It is not a
> distribution but a self compiled system "from the scratch", taylormade. Please
> have a look in http://www.tldp.org/LDP/lfs/html/ or in
> http://www.linuxfromscratch.org/lfs/news.html
OK. This is almost certainly a problem with how the distribution is building glibc.
> > [2] Please do:
> > man sysconf
> > and see if it documents _SC_PHYS_PAGES
> >
> quote ( extract )
> -----------
> These values also exist, but may not be standard.
>
> - _SC_PHYS_PAGES
> The number of pages of physical memory. Note that it is
> possi-
> ble for the product of this value and the value of
> _SC_PAGE_SIZE
> to overflow.
> ------------
> unquote
Good enough. What's happening is that your glibc library is complaining that _SC_PHYS_PAGES is not defined or available to sysconf.
> > If it does, then you need to file a bug report with the distribution,
> saying
> > that memory is being reported incorrectly and that sysconf doesn't work
> as
> > advertised. You can send them memory.c as an example.
>
> Mhmm, the components from Linux from the scratch are the original ones
> downloaded eg. from Sourceforge or so. So this kind of bug should also be
> found in distributions.
Not in (for example) Fedora Core 3, RH9, or Debian (I have tested on all three).
> I will try to find out, if a bug was reported on
> sysconf or a patch given. Will we remain in contact ?
I suggest you contact the linuxfromstcratch maintainers and complain that memory.c does not compile and run as expected, and as it does under other distributions. We can use this thread to stay in touch about it.
I'll try and add something to the scheduler in the next few days so that if the memory is reported as
> Hi Bruce, > > > I suggest
)
> Hi Bruce,
>
> > I suggest you contact the linuxfromstcratch maintainers and complain
> that
> > memory.c does not compile and run as expected, and as it does under
> other
> > distributions. We can use this thread to stay in touch about it.
> >
> Here five answers to my posting in the "Beyond Linux From Scratch"-List:
> Quote
> ----------
> Does this have to be a c program?
Yes. Linuxfromscratch needs to fix sysconf() so that it behaves as documented.
Bruce
> Hallo Bruce, > > > Yes.
)
> Hallo Bruce,
>
> > Yes. Linuxfromscratch needs to fix sysconf() so that it behaves as
> > documented.
>
> The problem is _my_ Scratch 5.1.1 installation on this machine. I booted some
> hours ago with the old version 3 I still have installed on this box, and
> sysconf was reporting the correct amount on memory. Also my friend's box with
> 5.1.1 is working correct. It is my problem, and I have to try to find out what
> is happening. I will come back.
Something is probably wrong with your glibc installation or with your header file installation.
Bruce