Re: [GIT PULL] RCU changes for v5.14

From: Paul E. McKenney
Date: Sun Jul 04 2021 - 16:32:04 EST


On Sun, Jul 04, 2021 at 01:01:33PM -0700, Linus Torvalds wrote:
> On Sun, Jul 4, 2021 at 10:24 AM Paul E. McKenney <paulmck@xxxxxxxxxx> wrote:
> >
> > An example merge resolution may be found here:
> >
> > 4e2b64e124c7 ("Merge remote-tracking branch 'linus/master' into HEAD")
>
> There sommit ID's are completely useless, because I have no idea where
> they come from. They aren't in the linux-next tree as far as I can
> tell, for example.
>
> So they are just random noise.
>
> Now, none of the conflicts looked in the least complicated, so it's
> not like I _need_ the examples, but this "send random shortened SHA1s
> to Linus" is simply not useful.
>
> At a guess, it's actually from your merge-example branch in your own tree.
>
> The point being, that a SHA1 may be globally unique, but without
> telling me where that SHA1 can be _found_, it is entirely useless.
>
> If you have example merges - which I do like seeing, and I will
> compare against just to double-check even when I have no reason to
> doubt my own merge - you need to point to it the same way you point to
> the actual real branch.
>
> IOW, say something like
>
> "I've done an example merge, and you can find it in
>
> git://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-rcu.git
> rcu-example-merge"
>
> or similar.
>
> Because actual SHA1 names are only useful WHEN THEY ARE IN MY TREE. So
> you can point to history that I have (or that was in your actual pull
> request), and I can see _those_ just fine.
>
> So when you say
>
> "The second is a trivial whitespace conflict between these two commits:
>
> 76c8eaafe4f0 ("rcu: Create an unrcu_pointer() to remove __rcu
> from a pointer")
> b9964ce74544 ("rcu: Create an unrcu_pointer() to remove __rcu
> from a pointer")"
>
> then that makes sense, because those are two commits that I actually
> have as part of the merge conflict).
>
> But that example merge? I don't have it, unless you actually tell me
> where it is.
>
> Then I can just do
>
> git fetch <paul-told-me-where-to-fetch>
>
> and can do
>
> git show FETCH_HEAD
>
> or (more commonly) just compare my merge result with yours:
>
> git diff FETCH_HEAD kernel/rcu/tree_stall.h
>
> and it's all golden. But if you send me a random SHA1 of somethign
> that only exists in your trees, I just go "oh, ok, not useful".

Once again, please accept my apologies, and thank you for the explanation.
I should have sent something like the following, correct? Or is there
still something I am missing?

Thanx, Paul

------------------------------------------------------------------------

Hello, Linus,

In a deviation from long-standing -tip tradition, please pull the latest
RCU git tree from:

git://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-rcu.git core-rcu-2021.07.04
# HEAD: 641faf1b9064c270a476a424e60063bb05df3ee9: Merge branches 'bitmaprange.2021.05.10c', 'doc.2021.05.10c', 'fixes.2021.05.13a', 'kvfree_rcu.2021.05.10c', 'mmdumpobj.2021.05.10c', 'nocb.2021.05.12a', 'srcu.2021.05.12a', 'tasks.2021.05.18a' and 'torture.2021.05.10c' into HEAD (2021-05-18 10:56:19 -0700)

RCU changes for this cycle were:

o Bitmap support for "all" as an alias for all bits.

o Documentation updates.

o Miscellaneous fixes, including some that overlap into mm and lockdep.

o kvfree_rcu() updates.

o mem_dump_obj() updates, with acks from one of the slab-allocator
maintainers.

o RCU NOCB CPU updates, including limited deoffloading.

o SRCU updates.

o Tasks-RCU updates.

o Torture-test updates.

These changes have two merge conflicts with mainline. The first is
a semantic conflict detected by -next between these two commits:

2f064a59a11f ("sched: Change task_struct::state")
e44111ed20d8 ("rcu: Add ->rt_priority and ->gp_start to show_rcu_gp_kthreads() output")

I have done an example merge here:

git://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-rcu.git merge-example-rcu

The second is a trivial whitespace conflict between these two commits:

76c8eaafe4f0 ("rcu: Create an unrcu_pointer() to remove __rcu from a pointer")
b9964ce74544 ("rcu: Create an unrcu_pointer() to remove __rcu from a pointer")

I have done an example merge on top of the merge commit for the first
conflict:

git://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-rcu.git merge-example-rcu2

And yes, this second conflict did in fact happen because I handed out
a patch containing this whitespace error, and failed to follow up with
the corrected patch. :-/

Thanx, Paul

----------------------------------------------------------------
Frederic Weisbecker (17):
doc: Fix diagram references in memory-ordering document
rcu/nocb: Use the rcuog CPU's ->nocb_timer
timer: Revert "timer: Add timer_curr_running()"
srcu: Remove superfluous sdp->srcu_lock_count zero filling
srcu: Remove superfluous ssp initialization for early callbacks
srcu: Unconditionally embed struct lockdep_map
srcu: Initialize SRCU after timers
srcu: Fix broken node geometry after early ssp init
torture: Correctly fetch number of CPUs for non-English languages
rcu/nocb: Directly call __wake_nocb_gp() from bypass timer
rcu/nocb: Allow de-offloading rdp leader
rcu/nocb: Cancel nocb_timer upon nocb_gp wakeup
rcu/nocb: Delete bypass_timer upon nocb_gp wakeup
rcu/nocb: Only cancel nocb timer if not polling
rcu/nocb: Prepare for fine-grained deferred wakeup
rcu/nocb: Unify timers
srcu: Early test SRCU polling start

Ingo Molnar (1):
rcu: Fix various typos in comments

Jules Irenge (1):
rcu: Add missing __releases() annotation

Maninder Singh (2):
mm/slub: Fix backtrace of objects to handle redzone adjustment
mm/slub: Add Support for free path information of an object

Paul E. McKenney (47):
doc: Fix statement of RCU's memory-ordering requirements
tools/rcu: Add drgn script to dump number of RCU callbacks
rcu-tasks: Add block comment laying out RCU Tasks design
rcu-tasks: Add block comment laying out RCU Rude design
torture: Fix remaining erroneous torture.sh instance of $*
torture: Add "scenarios" option to kvm.sh --dryrun parameter
torture: Make kvm-again.sh use "scenarios" rather than "batches" file
refscale: Allow CPU hotplug to be enabled
rcuscale: Allow CPU hotplug to be enabled
torture: Add kvm-remote.sh script for distributed rcutorture test runs
refscale: Add acqrel, lock, and lock-irq
rcutorture: Abstract read-lock-held checks
torture: Fix grace-period rate output
torture: Abstract end-of-run summary
torture: Make kvm.sh use abstracted kvm-end-run-stats.sh
torture: Make the build machine control N in "make -jN"
torture: Make kvm-find-errors.sh account for kvm-remote.sh
rcutorture: Judge RCU priority boosting on grace periods, not callbacks
torture: Set kvm.sh language to English
rcutorture: Delay-based false positives for RCU priority boosting tests
rcutorture: Consolidate rcu_torture_boost() timing and statistics
rcutorture: Make rcu_torture_boost_failed() check for GP end
rcutorture: Add BUSTED-BOOST to test RCU priority boosting tests
rcutorture: Forgive RCU boost failures when CPUs don't pass through QS
rcutorture: Don't count CPU-stalled time against priority boosting
torture: Make kvm-remote.sh account for network failure in pathname checks
torture: Don't cap remote runs by build-system number of CPUs
rcutorture: Move mem_dump_obj() tests into separate function
rcu: Remove the unused rcu_irq_exit_preempt() function
rcu: Invoke rcu_spawn_core_kthreads() from rcu_spawn_gp_kthread()
rcu: Add ->rt_priority and ->gp_start to show_rcu_gp_kthreads() output
rcu: Add ->gp_max to show_rcu_gp_kthreads() output
lockdep: Explicitly flag likely false-positive report
rcu: Reject RCU_LOCKDEP_WARN() false positives
rcu: Add quiescent states and boost states to show_rcu_gp_kthreads() output
rcu: Make RCU priority boosting work on single-CPU rcu_node structures
rcu: Make show_rcu_gp_kthreads() dump rcu_node structures blocking GP
rcu: Restrict RCU_STRICT_GRACE_PERIOD to at most four CPUs
rcu: Make rcu_gp_cleanup() be noinline for tracing
rcu: Point to documentation of ordering guarantees
rcu: Don't penalize priority boosting when there is nothing to boost
rcu: Create an unrcu_pointer() to remove __rcu from a pointer
rcu: Improve comments describing RCU read-side critical sections
rcu: Remove obsolete rcu_read_unlock() deadlock commentary
rcu-tasks: Make ksoftirqd provide RCU Tasks quiescent states
tasks-rcu: Make show_rcu_tasks_gp_kthreads() be static inline
Merge branches 'bitmaprange.2021.05.10c', 'doc.2021.05.10c', 'fixes.2021.05.13a', 'kvfree_rcu.2021.05.10c', 'mmdumpobj.2021.05.10c', 'nocb.2021.05.12a', 'srcu.2021.05.12a', 'tasks.2021.05.18a' and 'torture.2021.05.10c' into HEAD

Rolf Eike Beer (1):
rcu: Fix typo in comment: kthead -> kthread

Uladzislau Rezki (Sony) (6):
kvfree_rcu: Use [READ/WRITE]_ONCE() macros to access to nr_bkv_objs
kvfree_rcu: Add a bulk-list check when a scheduler is run
kvfree_rcu: Update "monitor_todo" once a batch is started
kvfree_rcu: Use kfree_rcu_monitor() instead of open-coded variant
kvfree_rcu: Fix comments according to current code
kvfree_rcu: Refactor kfree_rcu_monitor()

Yury Norov (2):
bitmap_parse: Support 'all' semantics
rcu/tree_plugin: Don't handle the case of 'all' CPU range

Zhang Qiang (1):
kvfree_rcu: Release a page cache under memory pressure

Zhouyi Zhou (1):
rcu: Improve tree.c comments and add code cleanups

.../Memory-Ordering/Tree-RCU-Memory-Ordering.rst | 6 +-
Documentation/admin-guide/kernel-parameters.rst | 5 +
Documentation/admin-guide/kernel-parameters.txt | 5 +
include/linux/rcupdate.h | 84 +++---
include/linux/rcutiny.h | 1 -
include/linux/rcutree.h | 1 -
include/linux/srcu.h | 6 +
include/linux/srcutree.h | 2 -
include/linux/timer.h | 2 -
include/trace/events/rcu.h | 1 +
init/main.c | 2 +
kernel/locking/lockdep.c | 6 +-
kernel/rcu/Kconfig.debug | 2 +-
kernel/rcu/rcu.h | 14 +-
kernel/rcu/rcutorture.c | 315 +++++++++++----------
kernel/rcu/refscale.c | 109 ++++++-
kernel/rcu/srcutree.c | 28 +-
kernel/rcu/sync.c | 4 +-
kernel/rcu/tasks.h | 58 +++-
kernel/rcu/tiny.c | 1 -
kernel/rcu/tree.c | 313 +++++++++++---------
kernel/rcu/tree.h | 14 +-
kernel/rcu/tree_plugin.h | 239 ++++++++--------
kernel/rcu/tree_stall.h | 84 +++++-
kernel/rcu/update.c | 8 +-
kernel/time/timer.c | 14 -
lib/bitmap.c | 9 +
lib/test_bitmap.c | 7 +
mm/oom_kill.c | 2 +-
mm/slab.h | 1 +
mm/slab_common.c | 12 +-
mm/slub.c | 8 +
mm/util.c | 2 +-
tools/rcu/rcu-cbs.py | 46 +++
.../testing/selftests/rcutorture/bin/kvm-again.sh | 33 +--
.../testing/selftests/rcutorture/bin/kvm-build.sh | 6 +-
.../selftests/rcutorture/bin/kvm-end-run-stats.sh | 40 +++
.../selftests/rcutorture/bin/kvm-find-errors.sh | 2 +-
.../selftests/rcutorture/bin/kvm-recheck-rcu.sh | 2 +-
.../testing/selftests/rcutorture/bin/kvm-remote.sh | 249 ++++++++++++++++
tools/testing/selftests/rcutorture/bin/kvm.sh | 61 ++--
tools/testing/selftests/rcutorture/bin/torture.sh | 2 +-
.../selftests/rcutorture/configs/rcu/BUSTED-BOOST | 17 ++
.../rcutorture/configs/rcu/BUSTED-BOOST.boot | 8 +
.../selftests/rcutorture/configs/rcuscale/TREE | 2 +-
.../selftests/rcutorture/configs/rcuscale/TREE54 | 2 +-
.../rcutorture/configs/refscale/NOPREEMPT | 2 +-
.../selftests/rcutorture/configs/refscale/PREEMPT | 2 +-
.../rcutorture/formal/srcu-cbmc/src/locks.h | 2 +-
49 files changed, 1265 insertions(+), 576 deletions(-)
create mode 100644 tools/rcu/rcu-cbs.py
create mode 100755 tools/testing/selftests/rcutorture/bin/kvm-end-run-stats.sh
create mode 100755 tools/testing/selftests/rcutorture/bin/kvm-remote.sh
create mode 100644 tools/testing/selftests/rcutorture/configs/rcu/BUSTED-BOOST
create mode 100644 tools/testing/selftests/rcutorture/configs/rcu/BUSTED-BOOST.boot