Bug 198733
Summary: | nvme AMD-Vi IO_PAGE_FAULT | ||
---|---|---|---|
Product: | Platform Specific/Hardware | Reporter: | Andreas (andreas.thalhammer) |
Component: | x86-64 | Assignee: | platform_x86_64 (platform_x86_64) |
Status: | RESOLVED DUPLICATE | ||
Severity: | normal | CC: | aladjev.andrew, chenhan.hsiao.tw, nutodafozo |
Priority: | P1 | ||
Hardware: | IA-32 | ||
OS: | Linux | ||
Kernel Version: | 4.15 | Subsystem: | |
Regression: | No | Bisected commit-id: |
Description
Andreas
2018-02-08 21:48:54 UTC
Hello. Today I've reproduced similar issue #202665. I don't know whether it is connected with your issue. Please try to disable discard option in fstab, remove iommu=soft boot param, reboot and run "fstrim -v /". If "IO_PAGE_FAULT" will appear - we will have the same issue. We need here a developer who works with iommu. He can send us something like debug patch so we can provide better log. I think that there is a side effect between nvme and iommu implementations. Fstrim can be just a way to reproduce it. Yes, I must confirm this behaviour, thus it seems to be the same issue. I've removed all "discard" mount options I had for the filesystems on NVMe in my /etc/fstab, and removed the "iommu=soft" mount option I had used before. No errors appeared. (With discard I still have those errors on a regular basis.) The fstrim command then triggered these messages. nvme 0000:01:00.0: Event logged [IO_PAGE_FAULT domain=0x0000 address=0xXXXXXXXX flags=0x0000] AMD-Vi: Event logged [IO_PAGE_FAULT device=01:00.0 domain=0x0000 address=0xXXXXXXXX flags=0x0000] Only the addresses varied, domain is always 0x0000 as well as flags. I can confirm this issue on Asus X470 PRO motherboard (latest bios from today 4406) and HP EX920 ssd (same sm2265 controller) - running fstrim spams dmesg with IO_PAGE_FAULT. See https://bugzilla.kernel.org/show_bug.cgi?id=202665#c7 *** This bug has been marked as a duplicate of bug 202665 *** |