From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.7.5) Gecko/20041111 Firefox/1.0 Description of problem: After installing ndiswrapper 1.0 and setting up my wireless network the other day, I left my laptop on with the wireless running. After coming back a few hours later, the computer was very slow to respond. I checked the memory usage and my RAM was close to full when not including buffers. The swap file usage was also unusually high. slabtop revealed that size-64 was over 250 MB in Cache Size. With ndiswrapper loaded and the wireless going, this value climbs at a steady rate. Without ndiswrapper, it appears to be fairly stable. Others have also noticed this problem with other distros, so it seems to be a general kernel problem. A message to the ndiswrapper list says this is definitely a problem with kernels 2.6.10 through 11-rc3 and the status of 11-rc4 is unknown. Version-Release number of selected component (if applicable): kernel-2.6.10-1.766_FC3.x86_64 How reproducible: Always Steps to Reproduce: 1. Get ndiswrapper 1.0 and build it 2. Run slabtop and notice the Cache Size column for size-64 entry, also note amount of free RAM reported by free 3. modprobe ndiswrapper and setup wireless for use 4. Run slabtop again and watch Cache Size column for size-64 entry climb up at about 4 kB per update 5. Eventually, RAM will be used and the swap file will begin to be used heavily Actual Results: After leaving my laptop running (idle) for several hours with ndiswrapper loaded and wireless setup, the indicated Cache Size of the size-64 entry in slabtop was over 250 MB. Right after boot, this value is a few hundred kB. My RAM was almost completely full (not including buffers) and the used swap file was a couple hundered MB (it is normally fairly small). Expected Results: The kernel should not leak memory. Additional info: I've marked this as high because it mentions a memory leak. I don't know if I'd call it a severe leak, but it is substantial. I have 512 MB of RAM, but I suspect that with less RAM, this could be a major problem much more quickly. A rough estimate is about 4 kB every couple seconds or so.
Well, it seems I may have spoken too soon. It seems that ndiswrapper misinterpreted something in the NDIS protocol. This lead to a situation where the Windows driver was making numerous calls to a function that the wrapper allocated memory in, but corresponding free calls were not made. I was told this behavior was only seen in certain kernel versions, thus making it appear to be a kernel thing. It now seems to be an ndiswrapper thing, so I'm marking it NOTABUG (as far as the kernel goes). For those interested, it should be fixed in ndiswrapper 1.1rc1 as of 2005-02-21. There is also a patch for version 1.0 if you can only use that as I have only been able to with 64-bit Broadcom drivers. Sorry for the false alarm :).