linux-stable.git/kernel/rcu/srcutree.c, branch v5.13.4

srcu: Fix broken node geometry after early ssp init

2021-07-20T14:00:12+00:00

[ Upstream commit b5befe842e6612cf894cf4a199924ee872d8b7d8 ]

An srcu_struct structure that is initialized before rcu_init_geometry()
will have its srcu_node hierarchy based on CONFIG_NR_CPUS.  Once
rcu_init_geometry() is called, this hierarchy is compressed as needed
for the actual maximum number of CPUs for this system.

Later on, that srcu_struct structure is confused, sometimes referring
to its initial CONFIG_NR_CPUS-based hierarchy, and sometimes instead
to the new num_possible_cpus() hierarchy.  For example, each of its
->mynode fields continues to reference the original leaf rcu_node
structures, some of which might no longer exist.  On the other hand,
srcu_for_each_node_breadth_first() traverses to the new node hierarchy.

There are at least two bad possible outcomes to this:

1) a) A callback enqueued early on an srcu_data structure (call it
      *sdp) is recorded pending on sdp->mynode->srcu_data_have_cbs in
      srcu_funnel_gp_start() with sdp->mynode pointing to a deep leaf
      (say 3 levels).

   b) The grace period ends after rcu_init_geometry() shrinks the
      nodes level to a single one.  srcu_gp_end() walks through the new
      srcu_node hierarchy without ever reaching the old leaves so the
      callback is never executed.

   This is easily reproduced on an 8 CPUs machine with CONFIG_NR_CPUS >= 32
   and "rcupdate.rcu_self_test=1". The srcu_barrier() after early tests
   verification never completes and the boot hangs:

	[ 5413.141029] INFO: task swapper/0:1 blocked for more than 4915 seconds.
	[ 5413.147564]       Not tainted 5.12.0-rc4+ #28
	[ 5413.151927] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
	[ 5413.159753] task:swapper/0       state:D stack:    0 pid:    1 ppid:     0 flags:0x00004000
	[ 5413.168099] Call Trace:
	[ 5413.170555]  __schedule+0x36c/0x930
	[ 5413.174057]  ? wait_for_completion+0x88/0x110
	[ 5413.178423]  schedule+0x46/0xf0
	[ 5413.181575]  schedule_timeout+0x284/0x380
	[ 5413.185591]  ? wait_for_completion+0x88/0x110
	[ 5413.189957]  ? mark_held_locks+0x61/0x80
	[ 5413.193882]  ? mark_held_locks+0x61/0x80
	[ 5413.197809]  ? _raw_spin_unlock_irq+0x24/0x50
	[ 5413.202173]  ? wait_for_completion+0x88/0x110
	[ 5413.206535]  wait_for_completion+0xb4/0x110
	[ 5413.210724]  ? srcu_torture_stats_print+0x110/0x110
	[ 5413.215610]  srcu_barrier+0x187/0x200
	[ 5413.219277]  ? rcu_tasks_verify_self_tests+0x50/0x50
	[ 5413.224244]  ? rdinit_setup+0x2b/0x2b
	[ 5413.227907]  rcu_verify_early_boot_tests+0x2d/0x40
	[ 5413.232700]  do_one_initcall+0x63/0x310
	[ 5413.236541]  ? rdinit_setup+0x2b/0x2b
	[ 5413.240207]  ? rcu_read_lock_sched_held+0x52/0x80
	[ 5413.244912]  kernel_init_freeable+0x253/0x28f
	[ 5413.249273]  ? rest_init+0x250/0x250
	[ 5413.252846]  kernel_init+0xa/0x110
	[ 5413.256257]  ret_from_fork+0x22/0x30

2) An srcu_struct structure that is initialized before rcu_init_geometry()
   and used afterward will always have stale rdp->mynode references,
   resulting in callbacks to be missed in srcu_gp_end(), just like in
   the previous scenario.

This commit therefore causes init_srcu_struct_nodes to initialize the
geometry, if needed.  This ensures that the srcu_node hierarchy is
properly built and distributed from the get-go.

Suggested-by: Paul E. McKenney 
Signed-off-by: Frederic Weisbecker 
Cc: Boqun Feng 
Cc: Lai Jiangshan 
Cc: Neeraj Upadhyay 
Cc: Josh Triplett 
Cc: Joel Fernandes 
Cc: Uladzislau Rezki 
Signed-off-by: Paul E. McKenney 
Signed-off-by: Sasha Levin

Merge branches 'doc.2021.01.06a', 'fixes.2021.01.04b', 'kfree_rcu.2021.01.04a', 'mmdumpobj.2021.01.22a', 'nocb.2021.01.06a', 'rt.2021.01.04a', 'stall.2021.01.06a', 'torture.2021.01.12a' and 'tortureall.2021.01.06a' into HEAD

2021-01-22T23:26:44+00:00

doc.2021.01.06a: Documentation updates.
fixes.2021.01.04b: Miscellaneous fixes.
kfree_rcu.2021.01.04a: kfree_rcu() updates.
mmdumpobj.2021.01.22a: Dump allocation point for memory blocks.
nocb.2021.01.06a: RCU callback offload updates and cblist segment lengths.
rt.2021.01.04a: Real-time updates.
stall.2021.01.06a: RCU CPU stall warning updates.
torture.2021.01.12a: Torture-test updates and polling SRCU grace-period API.
tortureall.2021.01.06a: Torture-test script updates.

rcu/segcblist: Add counters to segcblist datastructure

2021-01-07T00:24:19+00:00

Add counting of segment lengths of segmented callback list.

This will be useful for a number of things such as knowing how big the
ready-to-execute segment have gotten. The immediate benefit is ability
to trace how the callbacks in the segmented callback list change.

Also this patch remove hacks related to using donecbs's ->len field as a
temporary variable to save the segmented callback list's length. This cannot be
done anymore and is not needed.

Also fix SRCU:
The negative counting of the unsegmented list cannot be used to adjust
the segmented one. To fix this, sample the unsegmented length in
advance, and use it after CB execution to adjust the segmented list's
length.

Reviewed-by: Frederic Weisbecker 
Signed-off-by: Joel Fernandes (Google) 
Signed-off-by: Paul E. McKenney

srcu: Add comment explaining cookie overflow/wrap

2021-01-04T21:53:39+00:00

This commit adds to the poll_state_synchronize_srcu() header comment
describing the issues surrounding SRCU cookie overflow/wrap for the
different kernel configurations.

Link: https://lore.kernel.org/rcu/20201112201547.GF3365678@moria.home.lan/
Reported-by: Kent Overstreet 
Signed-off-by: Paul E. McKenney

srcu: Provide polling interfaces for Tree SRCU grace periods

2021-01-04T21:53:38+00:00

There is a need for a polling interface for SRCU grace
periods, so this commit supplies get_state_synchronize_srcu(),
start_poll_synchronize_srcu(), and poll_state_synchronize_srcu() for this
purpose.  The first can be used if future grace periods are inevitable
(perhaps due to a later call_srcu() invocation), the second if future
grace periods might not otherwise happen, and the third to check if a
grace period has elapsed since the corresponding call to either of the
first two.

As with get_state_synchronize_rcu() and cond_synchronize_rcu(),
the return value from either get_state_synchronize_srcu() or
start_poll_synchronize_srcu() must be passed in to a later call to
poll_state_synchronize_srcu().

Link: https://lore.kernel.org/rcu/20201112201547.GF3365678@moria.home.lan/
Reported-by: Kent Overstreet 
[ paulmck: Add EXPORT_SYMBOL_GPL() per kernel test robot feedback. ]
[ paulmck: Apply feedback from Neeraj Upadhyay. ]
Link: https://lore.kernel.org/lkml/20201117004017.GA7444@paulmck-ThinkPad-P72/
Reviewed-by: Neeraj Upadhyay 
Signed-off-by: Paul E. McKenney

srcu: Provide internal interface to start a Tree SRCU grace period

2021-01-04T21:53:37+00:00

There is a need for a polling interface for SRCU grace periods.
This polling needs to initiate an SRCU grace period without having
to queue (and manage) a callback.  This commit therefore splits the
Tree SRCU __call_srcu() function into callback-initialization and
queuing/start-grace-period portions, with the latter in a new function
named srcu_gp_start_if_needed().  This function may be passed a NULL
callback pointer, in which case it will refrain from queuing anything.

Why have the new function mess with queuing?  Locking considerations,
of course!

Link: https://lore.kernel.org/rcu/20201112201547.GF3365678@moria.home.lan/
Reported-by: Kent Overstreet 
Reviewed-by: Neeraj Upadhyay 
Signed-off-by: Paul E. McKenney

Merge branches 'cpuinfo.2020.11.06a', 'doc.2020.11.06a', 'fixes.2020.11.19b', 'lockdep.2020.11.02a', 'tasks.2020.11.06a' and 'torture.2020.11.06a' into HEAD

2020-11-20T03:37:47+00:00

cpuinfo.2020.11.06a: Speedups for /proc/cpuinfo.
doc.2020.11.06a: Documentation updates.
fixes.2020.11.19b: Miscellaneous fixes.
lockdep.2020.11.02a: Lockdep-RCU updates to avoid "unused variable".
tasks.2020.11.06a: Tasks-RCU updates.
torture.2020.11.06a': Torture-test updates.

srcu: Take early exit on memory-allocation failure

2020-11-20T03:37:17+00:00

It turns out that init_srcu_struct() can be invoked from usermode tasks,
and that fatal signals received by these tasks can cause memory-allocation
failures.  These failures are not handled well by init_srcu_struct(),
so much so that NULL pointer dereferences can result.  This commit
therefore causes init_srcu_struct() to take an early exit upon detection
of memory-allocation failure.

Link: https://lore.kernel.org/lkml/20200908144306.33355-1-aik@ozlabs.ru/
Reported-by: Alexey Kardashevskiy 
Tested-by: Alexey Kardashevskiy 
Signed-off-by: Paul E. McKenney

srcu: Use a more appropriate lockdep helper

2020-11-03T01:10:00+00:00

The lockdep_is_held() macro is defined as:

 #define lockdep_is_held(lock)		lock_is_held(&(lock)->dep_map)

This hides away the dereference, so that builds with !LOCKDEP don't break.
This works in current kernels because the RCU_LOCKDEP_WARN() eliminates
its condition at preprocessor time in !LOCKDEP kernels.  However, later
patches in this series will cause the compiler to see this condition even
in !LOCKDEP kernels.  This commit prepares for this upcoming change by
switching from lock_is_held() to lockdep_is_held().

Signed-off-by: Jakub Kicinski 
--
CC: jiangshanlai@gmail.com
CC: paulmck@kernel.org
CC: josh@joshtriplett.org
CC: rostedt@goodmis.org
CC: mathieu.desnoyers@efficios.com
Signed-off-by: Paul E. McKenney

srcu: Remove KCSAN stubs

2020-08-25T01:36:03+00:00

KCSAN is now in mainline, so this commit removes the stubs for the
data_race(), ASSERT_EXCLUSIVE_WRITER(), and ASSERT_EXCLUSIVE_ACCESS()
macros.

Signed-off-by: Paul E. McKenney