splat.ko is packaged

Bug #1566074 reported by Seth Arnold on 2016-04-05
This bug affects 2 people
Affects Status Importance Assigned to Milestone
linux (Ubuntu)
Colin Ian King

Bug Description

I just noticed that splat.ko from the SPL upstream is packaged in our linux-image-* packages:


$ zgrep splat.ko buildlog_ubuntu-xenial-amd64.linux_4.4.0-16.32_BUILDING.txt.gz
  CC [M] /«PKGBUILDDIR»/debian/build/build-generic/spl/module/splat/splat-kobj.o
  LD [M] /«PKGBUILDDIR»/debian/build/build-generic/spl/module/splat/splat.ko
  CC [M] /«PKGBUILDDIR»/debian/build/build-lowlatency/spl/module/splat/splat-kobj.o
  LD [M] /«PKGBUILDDIR»/debian/build/build-lowlatency/spl/module/splat/splat.ko
  INSTALL /«PKGBUILDDIR»/debian/build/build-generic/spl/module/splat/splat.ko
  INSTALL /«PKGBUILDDIR»/debian/build/build-lowlatency/spl/module/splat/splat.ko
-rw-r--r-- root/root 344662 2016-03-25 01:59 ./lib/modules/4.4.0-16-generic/kernel/zfs/splat/splat.ko
-rw-r--r-- root/root 325110 2016-03-25 03:12 ./lib/modules/4.4.0-16-lowlatency/kernel/zfs/splat/splat.ko

splat.ko is used only with the /usr/sbin/splat testing tool from the spl package. I suspect we shouldn't include this module in packages that we expect users to install.


Richard Laager (rlaager) on 2016-04-05
Changed in linux (Ubuntu):
status: New → Confirmed
Colin Ian King (colin-king) wrote :

This module does have use, namely for regression testing the spl, so I'd prefer to keep it in rather than rip it just because users may not use it.

Changed in linux (Ubuntu):
assignee: nobody → Colin Ian King (colin-king)
importance: Undecided → Wishlist
status: Confirmed → Won't Fix
Seth Arnold (seth-arnold) wrote :

I can understand that but I'd really rather see it someplace that's hard for users to get to. Every now and then someone finds a bug that allows loading any kernel module on demand and this module seems likely to have functionality that should not be exposed to untrusted users.

(I haven't reviewed it, and it's my goal to not review it. :)


Colin Ian King (colin-king) wrote :

My view is that removing a key testing module reduces our ability to sanity check for regressions. I'd like to be able to ensure we can test the layer between ZFS and the kernel before we find out it eats data.

To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Other bug subscribers