<html> <body> I'm running the current WS 3 kernel (although older kernels exhibit the same problem) on an Athlon 3200+ in a Soyo KT880 motherboard with 1GB, using it as a Samba file server for my little home network. Nothing runs on it except a software raid 5 with four 200GB drives, which is shared with a couple of Wintel boxes and Macs in my house using Samba. If I boot it up and just let it sit there for a day or two, then log in to a gnome desktop, the system monitor shows 990MB of memory used plus a couple of hundred MB of swap used too. This is with no users ever having logged in, except root just to check the system monitor, one time only. Basically nothing has happened since boot, not even any network activity against the raid to speak of. Haven't even played solitaire or run a browser. The memory usage accounted for on the ps -e display is a couple hundred MB, mostly basic GUI infrastructure stuff, nothing you'd even call an application. There's something like 700 or 800MB unaccounted for, and it would probably be even more if I knew how to check for the memory leak without going into a GUI. The only thing different about this system from a totally stock binary kernel is the installation of 3 driver modules: libsata and sata-via (to handle two of the IDE drives in the raid) and the GBE network chip driver module, 'velocityget', which I downloaded and compiled from Via to handle their onboard 6122 network controller. To all outward appearances these devices work just fine. The sata drivers are supplied by Red Hat but unsupported. The net driver is supplied by Via but unsupported. Not being a kernel guru (or any kind of guru), I don't know how to even start finding out if this egregious leak is in one of these drivers or someplace else. Any advice would be appreciated. --Ken S. </body> </html>