Michal Hocko wrote: > On Sat 12-08-17 00:46:18, Tetsuo Handa wrote: > > Michal Hocko wrote: > > > On Fri 11-08-17 16:54:36, Tetsuo Handa wrote: > > > > Michal Hocko wrote: > > > > > On Fri 11-08-17 11:28:52, Tetsuo Handa wrote: > > > > > > Will you explain the mechanism why random values are written instead of zeros > > > > > > so that this patch can actually fix the race problem? > > > > > > > > > > I am not sure what you mean here. Were you able to see a write with an > > > > > unexpected content? > > > > > > > > Yes. See http://lkml.kernel.org/r/201708072228.FAJ09347.tOOVOFFQJSHMFL@xxxxxxxxxxxxxxxxxxx . > > > > > > Ahh, I've missed that random part of your output. That is really strange > > > because AFAICS the oom reaper shouldn't really interact here. We are > > > only unmapping anonymous memory and even if a refault slips through we > > > should always get zeros. > > > > > > Your test case doesn't mmap MAP_PRIVATE of a file so we shouldn't even > > > get any uninitialized data from a file by missing CoWed content. The > > > only possible explanations would be that a page fault returned a > > > non-zero data which would be a bug on its own or that a file write > > > extend the file without actually writing to it which smells like a fs > > > bug to me. > > > > As I wrote at http://lkml.kernel.org/r/201708112053.FIG52141.tHJSOQFLOFMFOV@xxxxxxxxxxxxxxxxxxx , > > I don't think it is a fs bug. > > Were you able to reproduce with other filesystems? Yes, I can reproduce this problem using both xfs and ext4 on 4.11.11-200.fc25.x86_64 on Oracle VM VirtualBox on Windows. I believe that this is not old data from disk, for I can reproduce this problem using newly attached /dev/sdb which has never written any data (other than data written by mkfs.xfs and mkfs.ext4). /dev/sdb /tmp ext4 rw,seclabel,relatime,data=ordered 0 0 The garbage pattern (the last 4096 bytes) is identical for both xfs and ext4. > I wonder what is > different in my testing because I cannot reproduce this at all. Well, I > had to reduce the number of competing writer threads to 128 because I > quickly hit the trashing behavior with more of them (and 4 CPUs). I will > try on a larger machine. I don't think a larger machine is necessary. I can reproduce this problem with 8 competing writer threads on 4 CPUs. I don't have native Linux environment. Maybe that is the difference. Can you try VMware Workstation Player or Oracle VM VirtualBox environment? -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>
