hello... i'm trying to put project reliable and i'm testing now the possibility of using the network for long periods of time with FLIPS. but when i do a considerable amount of sendto's it gives me the following error:
[C.A]dde_linux/lib/src/mm.c:117: __more_kcore(): Error: out of memory
i've seen that this error ocurrs when something in DDE is out of memory. i presume the driver's buffers. What can i do to increase this memory? is this possible? or is it something not flushing some buffers in a correct way?
thanks
Tiago
Hello,
On Sa, 2005-06-04 at 17:02 +0100, Tiago Jorge wrote:
hello... i'm trying to put project reliable and i'm testing now the possibility of using the network for long periods of time with FLIPS. but when i do a considerable amount of sendto's it gives me the following error:
[C.A]dde_linux/lib/src/mm.c:117: __more_kcore(): Error: out of memory
i've seen that this error ocurrs when something in DDE is out of memory. i presume the driver's buffers. What can i do to increase this memory? is this possible? or is it something not flushing some buffers in a correct way?
IMO FLIPS suffers from some memory leaks I never had time to fix or even analyse thoroughly. You may increase the size of the DDE memory pool by editing flips/server/src/main_lxdrv.c:22:
if (liblinux_init(640*1024, 1024*1024))
The first value describes the vmem pool (vmalloc() and friends) - the second describes the kmem pool (kmalloc()).
Good luck
following this issue i'm having maybe a combination of both problem that i raised in this list with flips... memory and parelelism. My server has several iterations. For each iteration, a node (because it is distributed) interacts with the network, doing sendto's and recvfrom's (in average 15/20 calls for each iteration). In earlier discussions, we concluded that Flips wasn't parallel and has some memory leaks. Since the beginning i've had a strange problem... sometimes the sendto hangs. When i didn't knew that flips wasn't paralel this was more notorius, becauses he hanged lacking the support to the concurrent and blocking calls, but now he has no reason to do that. I don't block calls for an infinite period of time and when i do a blocking call, i've surrounded all the other possible calls with locks, so that each blocking call becomes atomic and sequential (sendto, recvfrom and select). He calls the sendto of the client library, but nothing happens on the flips side... so i presume that the call hangs somewhere. This happens after some iterations... around 200 iterations, so nearly 2000 sendto's. It's not the same problem that gave me when i tried to do 10000 sendto's. Then it gave me the "no memory available...". Now only hangs. Any ideia for what might be causing this issue?
Thanks in advance
Tiago
Hello,
sorry, but I've no idea what blocks FLIPS after the specified amount of sendto's and I currently have no resources to investigate this further.
Good luck
l4-hackers@os.inf.tu-dresden.de