The attached kernel patch (applied on top of 4.18.5) that I've tried, almost completely eliminates the disk thrashing(the constant reading of executable(and .so) files on every context switch) associated with freezing the OS and so, with this patch, the OOM-killer is triggered within a maxium of 1 second when it is needed, rather than, without this patch, freeze the OS for minutes(or just a long time, it may even auto reboot depending on your kernel .config options set to panic(reboot) on hang after xx seconds) with constant disk reading well before OOM-killer gets triggered.
More info as to why OS freezes with constant disk reading when running out of RAM, here: https://unix.stackexchange.com/a/463469/306023 The patch is from inside this question: https://stackoverflow.com/q/52067753/10239615 Note: I don't recommend using this patch in production, without actual programmers saying that it's ok first. I'm not a programmer. Maybe some programmer could improve it? I am using the patch currently inside a Qubes OS R4.0 Fedora 28 AppVM where I can easily reproduce freezing the VM's OS (+ the constant disk thrashing, seen from `sudo iotop` in dom0) without the patch, while attempting to compile firefox with 4000MB max RAM setting for the VM, and with the patch basically no freezing and little to no disk thrashing during the 1 second it takes for OOM-killer to kill the offending process(es). ** Patch added: "le9b.patch" https://bugs.launchpad.net/ubuntu/+source/linux/+bug/159356/+attachment/5183022/+files/le9b.patch -- You received this bug notification because you are a member of Kernel Packages, which is subscribed to linux in Ubuntu. https://bugs.launchpad.net/bugs/159356 Title: System freeze on high memory usage when DMA is disabled Status in linux package in Ubuntu: Incomplete Bug description: I run a batch matlab job server here at my lab, running Dapper 6.06 (for the LTS). One of the users has submitted a very memory-consuming job, which successfully crashes the server. Upon closer inspection, the crash happens like this: 1. I run matlab with the given file (as an ordinary, unpriveleged user) 2. RAM usage quickly fills up 3. Once the RAM meter hits 100%, the system freezes: All SSH connections freeze up, and while switching VTs directly on the machine works, no new processes run - so one can't log in, or do anything if he is logged in. (Sometimes typing doesn't work at all) Note that the swap - while 7 gigs of it are available - is never used. (The machine has 7 gigs of RAM as well) I've tried the same on my Gutsy 32-bit box, and there was no system freezeup - matlab simply notified that the system was out of memory. However, it did this once memory was 100% in use - and still, swap didn't get used at all! (Though it is mounted correctly and shows up in "top" and "free"). So first thing's first - I'd like to eliminate the crash issue. I suppose I could switch the server to 32-bit, but I think that would be a performance loss, considering that it does a lot of heavy computation. There is no reason, however, that this should happen on a 64-bit machine anyway. Why does it? WORKAROUND: Enabling DMA in the BIOS To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/159356/+subscriptions -- Mailing list: https://launchpad.net/~kernel-packages Post to : kernel-packages@lists.launchpad.net Unsubscribe : https://launchpad.net/~kernel-packages More help : https://help.launchpad.net/ListHelp
