On Tue, Jul 22, 2014 at 11:42:32PM +0900, Jungseok Lee wrote: > On Jul 21, 2014, at 9:45 PM, Richard W.M. Jones wrote: > > [ ... ] > > > I have spent a bit of time inserting printks around memory allocations > > in arch/arm/kvm/mmu.c (since that file seems to be responsible for > > guest page tables and guest memory allocation) but I haven't come up > > with anything conclusive. > > Hi, Richard > > Have you figure out this issue? No. > IMHO, it is needed to dive into three functions: mmu_topup_memory_cache, > mmu_free_memory_cache and mum_memory_cache_alloc. They manage both the > number of allocated (cached) pages and the number of usable pages with > a single parameter, nobjs. However, they can have different values and > mmu_free_memory_cache only refers to the number of usable pages, not the > total number of allocated pages. > > When VM launches, mmu_topup_memory_cache is called with min=2 and max=40. > nobjs is set to 40 at this time frame. As mmu_memory_cache_alloc is called, > nobjs is decreased. Let's try to destroy VM now. mmu_free_memory_cache tries > to free smaller pages instead of 40 pages. It reaches to memory leak. Yes I tried adding printk's into these functions. There is a possible leak. Unfortunately I'm not clear on how / if `put_page' frees a page. I kind of assume it's using some kind of reference counting, but I'm not sure. If `put_page' is releasing the page, then maybe there isn't a leak in these functions. I also looked at the code in those functions, and I couldn't see any error. It would be really useful if there was one place in the kernel where I could see all page allocations made by the KVM code ... It's possibly something I could do with systemtap, but I've not got around to that yet. Rich. -- Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones Read my programming and virtualization blog: http://rwmj.wordpress.com Fedora Windows cross-compiler. Compile Windows programs, test, and build Windows installers. Over 100 libraries supported. http://fedoraproject.org/wiki/MinGW _______________________________________________ kvmarm mailing list kvmarm@xxxxxxxxxxxxxxxxxxxxx https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
