RCU git Tree


The following branches are of interest:

  1. dev: Bleeding-edge code, both RCU and the Linux-kernel memory model. Please do any new RCU development against this branch.
  2. rcu/next: Outside of the merge window, RCU commits that have passed a reasonable amount of testing, and have not yet been proven to be unready for the merge window. During the merge window, a commit merging the topic branches which have been accepted in -tip or which will be submitted directly to mainline during that merge window. If you want to look at or test newish RCU code, but nevertheless want something reasonably stable, this is your branch.
  3. kcsan: Kernel concurrency sanitizer (KCSAN) updates intended or the next merge window. Prior to being added to this branch, KCSAN commits often spend some time in the dev branch.
  4. lkmm: Linux-kernel memory model (LKMM) updates intended for the next merge window. Note that LKMM patches require at least one Acked-by: (or Reviewed-by:) from someone other than the author, and that Paul E. McKenney's Signed-off-by: does not count.
  5. lkmm-dev: Linux-kernel memory model (LKMM) updates not deemed ready for the next merge window. Prior to being added to this branch, LKMM commits often spend some time in the dev branch.
  6. urgent: Fixes for regressions in mainline.
  7. main: The branch master will be a synonym for main until such time as it can be reasonably certain that users' git installations know about main. The presence of these two branches mean that someone cloning the -rcu tree will get a recent mainline RCU snapshot, perhaps excluding post-merge-window bug-fix patches. It is hoped that these branches will reduce the number of questions from beginners about ca. 2011 versions of RCU. (Unless they really do want to know about old versions of RCU, but experience indicates that such desires are not the common case.)

All of the above branches are subject to rebase. However, the old commits are kept around for at least six months by date-stamped branches, for example, “dev.2020.11.05a”.

This tree may be accessed as follows:

git clone git://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-rcu.git
cd linux-rcu
git checkout origin/dev

Once created, you can make your local copy incorporate changes as follows:

git remote update
git checkout origin/dev

If your development takes some time, please either rebase onto current origin/dev before sending your patches.

Regardless of how long your development takes, please test your code appropriately. The rcutorture test suite is easy to run on systems supporting KVM and qemu, so please give that a try, especially if your change is at all non-trivial. Of course, if you submit too many patches that break things, I am likely to insist that you run rcutorture on subsequent patches.

RCU Workflow

RCU Current Tasks

  1. Make RCU more friendly towards large deployments. The kvm-remote.sh script is a large step in this direction.
  2. Add an automated “acceptance test”. The torture.sh script is a large step in this direction.
  3. Add more forward-progress testing to rcutorture and fix any more bugs located thereby.
  4. Come up with some way to warn about RCU callback overload conditions.

Additional RCU Information

  1. Read-Copy Update (RCU) Papers
  2. A Tour Through rcutorture
  3. Stupid RCU Tricks
  4. Is Parallel Programming Hard, And If So, What Can You Do About It?