Hi! Luckily I've now got a PC wit 2GB RAM, so that allocating 900MB memory works quite well. So I found out that xosview does come up after allocating the 900MB. There realy seems to be some kind of race condition while xosview reads the information for software-raid. Disabling watching software-RAID: xosview*RAID: False in .Xresources workarounds the problem. xosview only allocates only some kB. On Samstag 09 April 2005 01:16, David Huecking wrote:
Hello!
I run xosview on some machines to be informed about their status. Now with SuSE Linux 9.2 x86 I've go the prob that from time to time xosview does eat all memory available on the machine running the X-server to display the xosview-windows (my workstation). At the same time it may happen that from remote-machine A the xosview works correct, from remote-machine B it does not and from local machine it also does not work. Some time later when I switched my workstation off and on again all xosviews are running smoothly... Though it happens only from time to time it seems like some kind of race condition?! I did a strace of a error xosview session: [...] mmap2(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x4001b000 read(4, "Personalities : [raid1] \nmd1 : a"..., 8192) = 195 read(4, "", 7168) = 0 close(4) = 0 munmap(0x4001b000, 4096) = 0 mmap2(NULL, 943079424, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x40357000 mmap2(NULL, 943079424, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x786bb000
The last mmap2-call seems to be the bad one. As I understand the mmap2-call it does allocate 943079424 bytes of memory (would be around 900MB). On a correctly running xosview session the program goes into the loop to grab the informations from /proc after the munmap(): [...] mmap2(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x4001b000 read(4, " CPU0 \n 0: 234"..., 8192) = 490 read(4, "", 7168) = 0 close(4) = 0 munmap(0x4001b000, 4096) = 0 open("/proc/mdstat", O_RDONLY) = 4 fstat64(4, {st_mode=S_IFREG|0444, st_size=0, ...}) = 0 mmap2(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x4001b000 read(4, "Personalities : [raid1] \nmd1 : a"..., 8192) = 196 read(4, "", 7168) = 0 close(4) = 0 munmap(0x4001b000, 4096) = 0 write(3, "\20\1\t\0\33\0`\5XDCCC_LINEAR_RGB_CORRECT"..., 36) = 36 read(3, 0xbfffea70, 32) = -1 EAGAIN (Resource temporarily unavailable) [...]
I do remember this behaviour from a SuSE 9.0 based system where I switched to a self-compiled Kernel 2.6 (9.0 was with 2.4...). But there was a patch for xosview available because the structure of /proc changed from 2.4 to 2.6.
To customize xosview I added the following to my .Xresources ! =========================================================================== ! XOsview ! =========================================================================== xosview*netBandwidth: 10000000 ! for 100Mbit/s Ethernet (could be 12500000) !xosview*netBandwidth: 5000000 xosview*netGraph: True xosview*cpuGraph: False !True xosview*loadGraph: True xosview*disk: True xosview*diskBandwidth: 20000000 xosview*RAID: True
But even if I comment out the last line (RAID /proc/mdstat) or even every line, so that the resource are standard it does not work...
-- Eat, sleep and go running, David Hücking. Encrypted eMail welcome! GnuPG/ PGP-Key: 0x57809216. Fingerprint: 3DF2 CBE0 DFAA 4164 02C2 4E2A E005 8DF7 5780 9216