On Fri, 05 Oct 2007 00:39:16 +0200 Miklos Szeredi <[EMAIL PROTECTED]> wrote:
> > throttle_vm_writeout() should be a per-zone thing, I guess. Perhaps fixing > > that would fix your deadlock. That's doubtful, but I don't know anything > > about your deadlock so I cannot say. > > No, doing the throttling per-zone won't in itself fix the deadlock. > > Here's a deadlock example: > > Total memory = 32M > /proc/sys/vm/dirty_ratio = 10 > dirty_threshold = 3M > ratelimit_pages = 1M > > Some program dirties 4M (dirty_threshold + ratelimit_pages) of mmap on > a fuse fs. Page balancing is called which turns all these into > writeback pages. > > Then userspace filesystem gets a write request, and tries to allocate > memory needed to complete the writeout. > > That will possibly trigger direct reclaim, and throttle_vm_writeout() > will be called. That will block until nr_writeback goes below 3.3M > (dirty_threshold + 10%). But since all 4M of writeback is from the > fuse fs, that will never happen. > > Does that explain it better? > yup, thanks. This is a somewhat general problem: a userspace process is in the IO path. Userspace block drivers, for example - pretty much anything which involves kernel->userspace upcalls for storage applications. I solved it once in the past by marking the userspace process as PF_MEMALLOC and I beleive that others have implemented the same hack. I suspect that what we need is a general solution, and that the solution will involve explicitly telling the kernel that this process is one which actually cleans memory and needs special treatment. Because I bet there will be other corner-cases where such a process needs kernel help, and there might be optimisation opportunities as well. Problem is, any such mark-me-as-special syscall would need to be privileged, and FUSE servers presently don't require special perms (do they?) - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
