LKML News v5.5-rc3

PROBLEM: Large (256MB) CMA allocation fails from time to time when using Intel VT-d

https://lkml.kernel.org/r/5e75d900cb50457784ee1d2050b511ec@MAILER13.tech.ceia

The author reports a CMA allocation failure that occurs time to time if the system is using Intel VT-d.

[PATCHv2 0/3] Allow ZRAM to use any zpool-compatible backend

https://lkml.kernel.org/r/20191219151928.ad4ccf732b64b7f8a26116db@gmail.com

ZRAM is an in-memory block device that compresses the content. Currently, ZRAM is used only with zsmalloc. However, the zsmalloc is not an optimal selection in some environment. The author gives embedded environment which fast response is important as an example. In such environment, though compress ratio is a little bit less, faster response (e..g, z3fold) is more valuable. Thus, this patchset changes ZRAM to be able to use multiple backend.

[PATCH RFC v4 00/42] Add KernelMemorySanitizer infrastructure

https://lkml.kernel.org/r/20191220184955.223741-1-glider@google.com

KMSAN detects errors related to uses of uninitialized memory. With a help from syzkaller, KMSAN has already reported more than 200 bugs in the past two years. This patch aims to be merged in the mainline.

Regression seen when using MADV_FREE vs MADV_DONTNEED

https://lkml.kernel.org/r/CAKgT0UeSs_qy8g5QoXpgcw2685+JtJPY+rehggraoGSYPrP1Aw@mail.gmail.com

The author introduced a patch using MADV_FREE instead of MADV_DONTNEED in the QEMU. On next-20191120 kernel, it worked as the author expected. On next-20191120, however, the author was able to encounter an issue. The MADV_FREE has been significantly slower than MADV_DONTNEED.

Linux 5.5-rc3

https://lkml.kernel.org/r/CAHk-=wiGKfWeUv294FD6bs4cw0m=nO_LdZkLz-vkp-2zwBzLcA@mail.gmail.com

This release has delayed a couple of hors due to Christmas preparation, but seems the Christmas has made no big effect to the development. Rather than that, this release is not only bigger than rc2 (rc3 is almost always bigger than rc2), but also bigger than recent few other rc3 releases.

Torvalds assumes that this is just a normal fluctuation, as the changes are spread around various subsystems. Also, the merge window timing could made some effect. As Santa is coming to the town, Torvalds also believe that next release will be more calm than now. That said, he also asks some tests.