NVMe EFI booting fix, and also large-swap tests completed

[Matthew Dillon]

master now has a bug fix for NVMe EFI booting.  There was a bug in devfs
that could cause writes into the UFS /boot partition on the NVMe device to
corrupt the filesystem.   This bug is now fixed.   It looks like many
recent BIOSes support booting directly from NVMe using an EFI boot.
Booting via NVMe is not typically supported with legacy boot.

If your BIOS supports EFI booting from NVMe devices, DragonFlyBSD can now
be booted this way.  Several bug fixes have just gone in to master,
particularly one related to UFS that gets tickled by a UFS partition on a
NVMe device which corrupts the UFS /boot.  This only happens with
UFS-on-NVMe.

If you have a NVMe-capable EFI BIOS you can try it out.  Be sure that you
do so running a kernel with the fix, and also be sure that the kernel you
install in the partition has the fix.  Instructions can be found here:

http://apollo.backplane.com/DFlyMisc/uefi.txt

--

In other news, recent improvements to swap management and the RSS resource
limit handling in master should now be stable.  Fairly extensive testing is
now complete.  Swap was tested paging out and paging in 14TB of actual
data, and I used synth with a RSS resource limit applied to seriously test
the RSS resource limit under very heavy VM loads.

-Matt
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.dragonflybsd.org/pipermail/users/attachments/20170101/68e71b3a/attachment.html>