linux-stable.git/kernel, branch v5.4.133

tracing: Resize tgid_map to pid_max, not PID_MAX_DEFAULT

2021-07-19T06:53:17+00:00

commit 4030a6e6a6a4a42ff8c18414c9e0c93e24cc70b8 upstream.

Currently tgid_map is sized at PID_MAX_DEFAULT entries, which means that
on systems where pid_max is configured higher than PID_MAX_DEFAULT the
ftrace record-tgid option doesn't work so well. Any tasks with PIDs
higher than PID_MAX_DEFAULT are simply not recorded in tgid_map, and
don't show up in the saved_tgids file.

In particular since systemd v243 & above configure pid_max to its
highest possible 1<<22 value by default on 64 bit systems this renders
the record-tgids option of little use.

Increase the size of tgid_map to the configured pid_max instead,
allowing it to cover the full range of PIDs up to the maximum value of
PID_MAX_LIMIT if the system is configured that way.

On 64 bit systems with pid_max == PID_MAX_LIMIT this will increase the
size of tgid_map from 256KiB to 16MiB. Whilst this 64x increase in
memory overhead sounds significant 64 bit systems are presumably best
placed to accommodate it, and since tgid_map is only allocated when the
record-tgid option is actually used presumably the user would rather it
spends sufficient memory to actually record the tgids they expect.

The size of tgid_map could also increase for CONFIG_BASE_SMALL=y
configurations, but these seem unlikely to be systems upon which people
are both configuring a large pid_max and running ftrace with record-tgid
anyway.

Of note is that we only allocate tgid_map once, the first time that the
record-tgid option is enabled. Therefore its size is only set once, to
the value of pid_max at the time the record-tgid option is first
enabled. If a user increases pid_max after that point, the saved_tgids
file will not contain entries for any tasks with pids beyond the earlier
value of pid_max.

Link: https://lkml.kernel.org/r/20210701172407.889626-2-paulburton@google.com

Fixes: d914ba37d714 ("tracing: Add support for recording tgid of tasks")
Cc: Ingo Molnar 
Cc: Joel Fernandes 
Cc: 
Signed-off-by: Paul Burton 
[ Fixed comment coding style ]
Signed-off-by: Steven Rostedt (VMware) 
Signed-off-by: Greg Kroah-Hartman

tracing: Simplify & fix saved_tgids logic

2021-07-19T06:53:17+00:00

commit b81b3e959adb107cd5b36c7dc5ba1364bbd31eb2 upstream.

The tgid_map array records a mapping from pid to tgid, where the index
of an entry within the array is the pid & the value stored at that index
is the tgid.

The saved_tgids_next() function iterates over pointers into the tgid_map
array & dereferences the pointers which results in the tgid, but then it
passes that dereferenced value to trace_find_tgid() which treats it as a
pid & does a further lookup within the tgid_map array. It seems likely
that the intent here was to skip over entries in tgid_map for which the
recorded tgid is zero, but instead we end up skipping over entries for
which the thread group leader hasn't yet had its own tgid recorded in
tgid_map.

A minimal fix would be to remove the call to trace_find_tgid, turning:

  if (trace_find_tgid(*ptr))

into:

  if (*ptr)

..but it seems like this logic can be much simpler if we simply let
seq_read() iterate over the whole tgid_map array & filter out empty
entries by returning SEQ_SKIP from saved_tgids_show(). Here we take that
approach, removing the incorrect logic here entirely.

Link: https://lkml.kernel.org/r/20210630003406.4013668-1-paulburton@google.com

Fixes: d914ba37d714 ("tracing: Add support for recording tgid of tasks")
Cc: Ingo Molnar 
Cc: Joel Fernandes 
Cc: 
Signed-off-by: Paul Burton 
Signed-off-by: Steven Rostedt (VMware) 
Signed-off-by: Greg Kroah-Hartman

rq-qos: fix missed wake-ups in rq_qos_throttle try two

2021-07-19T06:53:16+00:00

commit 11c7aa0ddea8611007768d3e6b58d45dc60a19e1 upstream.

Commit 545fbd0775ba ("rq-qos: fix missed wake-ups in rq_qos_throttle")
tried to fix a problem that a process could be sleeping in rq_qos_wait()
without anyone to wake it up. However the fix is not complete and the
following can still happen:

CPU1 (waiter1)		CPU2 (waiter2)		CPU3 (waker)
rq_qos_wait()		rq_qos_wait()
  acquire_inflight_cb() -> fails
			  acquire_inflight_cb() -> fails

						completes IOs, inflight
						  decreased
  prepare_to_wait_exclusive()
			  prepare_to_wait_exclusive()
  has_sleeper = !wq_has_single_sleeper() -> true as there are two sleepers
			  has_sleeper = !wq_has_single_sleeper() -> true
  io_schedule()		  io_schedule()

Deadlock as now there's nobody to wakeup the two waiters. The logic
automatically blocking when there are already sleepers is really subtle
and the only way to make it work reliably is that we check whether there
are some waiters in the queue when adding ourselves there. That way, we
are guaranteed that at least the first process to enter the wait queue
will recheck the waiting condition before going to sleep and thus
guarantee forward progress.

Fixes: 545fbd0775ba ("rq-qos: fix missed wake-ups in rq_qos_throttle")
CC: stable@vger.kernel.org
Signed-off-by: Jan Kara 
Link: https://lore.kernel.org/r/20210607112613.25344-1-jack@suse.cz
Signed-off-by: Jens Axboe 
Signed-off-by: Greg Kroah-Hartman

cpu/hotplug: Cure the cpusets trainwreck

2021-07-19T06:53:15+00:00

commit b22afcdf04c96ca58327784e280e10288cfd3303 upstream.

Alexey and Joshua tried to solve a cpusets related hotplug problem which is
user space visible and results in unexpected behaviour for some time after
a CPU has been plugged in and the corresponding uevent was delivered.

cpusets delegate the hotplug work (rebuilding cpumasks etc.) to a
workqueue. This is done because the cpusets code has already a lock
nesting of cgroups_mutex -> cpu_hotplug_lock. A synchronous callback or
waiting for the work to finish with cpu_hotplug_lock held can and will
deadlock because that results in the reverse lock order.

As a consequence the uevent can be delivered before cpusets have consistent
state which means that a user space invocation of sched_setaffinity() to
move a task to the plugged CPU fails up to the point where the scheduled
work has been processed.

The same is true for CPU unplug, but that does not create user observable
failure (yet).

It's still inconsistent to claim that an operation is finished before it
actually is and that's the real issue at hand. uevents just make it
reliably observable.

Obviously the problem should be fixed in cpusets/cgroups, but untangling
that is pretty much impossible because according to the changelog of the
commit which introduced this 8 years ago:

 3a5a6d0c2b03("cpuset: don't nest cgroup_mutex inside get_online_cpus()")

the lock order cgroups_mutex -> cpu_hotplug_lock is a design decision and
the whole code is built around that.

So bite the bullet and invoke the relevant cpuset function, which waits for
the work to finish, in _cpu_up/down() after dropping cpu_hotplug_lock and
only when tasks are not frozen by suspend/hibernate because that would
obviously wait forever.

Waiting there with cpu_add_remove_lock, which is protecting the present
and possible CPU maps, held is not a problem at all because neither work
queues nor cpusets/cgroups have any lockchains related to that lock.

Waiting in the hotplug machinery is not problematic either because there
are already state callbacks which wait for hardware queues to drain. It
makes the operations slightly slower, but hotplug is slow anyway.

This ensures that state is consistent before returning from a hotplug
up/down operation. It's still inconsistent during the operation, but that's
a different story.

Add a large comment which explains why this is done and why this is not a
dump ground for the hack of the day to work around half thought out locking
schemes. Document also the implications vs. hotplug operations and
serialization or the lack of it.

Thanks to Alexy and Joshua for analyzing why this temporary
sched_setaffinity() failure happened.

Fixes: 3a5a6d0c2b03("cpuset: don't nest cgroup_mutex inside get_online_cpus()")
Reported-by: Alexey Klimov 
Reported-by: Joshua Baker 
Signed-off-by: Thomas Gleixner 
Tested-by: Alexey Klimov 
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/87tuowcnv3.ffs@nanos.tec.linutronix.de
Signed-off-by: Greg Kroah-Hartman

bpf: Fix up register-based shifts in interpreter to silence KUBSAN

2021-07-19T06:53:11+00:00

[ Upstream commit 28131e9d933339a92f78e7ab6429f4aaaa07061c ]

syzbot reported a shift-out-of-bounds that KUBSAN observed in the
interpreter:

  [...]
  UBSAN: shift-out-of-bounds in kernel/bpf/core.c:1420:2
  shift exponent 255 is too large for 64-bit type 'long long unsigned int'
  CPU: 1 PID: 11097 Comm: syz-executor.4 Not tainted 5.12.0-rc2-syzkaller #0
  Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
  Call Trace:
   __dump_stack lib/dump_stack.c:79 [inline]
   dump_stack+0x141/0x1d7 lib/dump_stack.c:120
   ubsan_epilogue+0xb/0x5a lib/ubsan.c:148
   __ubsan_handle_shift_out_of_bounds.cold+0xb1/0x181 lib/ubsan.c:327
   ___bpf_prog_run.cold+0x19/0x56c kernel/bpf/core.c:1420
   __bpf_prog_run32+0x8f/0xd0 kernel/bpf/core.c:1735
   bpf_dispatcher_nop_func include/linux/bpf.h:644 [inline]
   bpf_prog_run_pin_on_cpu include/linux/filter.h:624 [inline]
   bpf_prog_run_clear_cb include/linux/filter.h:755 [inline]
   run_filter+0x1a1/0x470 net/packet/af_packet.c:2031
   packet_rcv+0x313/0x13e0 net/packet/af_packet.c:2104
   dev_queue_xmit_nit+0x7c2/0xa90 net/core/dev.c:2387
   xmit_one net/core/dev.c:3588 [inline]
   dev_hard_start_xmit+0xad/0x920 net/core/dev.c:3609
   __dev_queue_xmit+0x2121/0x2e00 net/core/dev.c:4182
   __bpf_tx_skb net/core/filter.c:2116 [inline]
   __bpf_redirect_no_mac net/core/filter.c:2141 [inline]
   __bpf_redirect+0x548/0xc80 net/core/filter.c:2164
   ____bpf_clone_redirect net/core/filter.c:2448 [inline]
   bpf_clone_redirect+0x2ae/0x420 net/core/filter.c:2420
   ___bpf_prog_run+0x34e1/0x77d0 kernel/bpf/core.c:1523
   __bpf_prog_run512+0x99/0xe0 kernel/bpf/core.c:1737
   bpf_dispatcher_nop_func include/linux/bpf.h:644 [inline]
   bpf_test_run+0x3ed/0xc50 net/bpf/test_run.c:50
   bpf_prog_test_run_skb+0xabc/0x1c50 net/bpf/test_run.c:582
   bpf_prog_test_run kernel/bpf/syscall.c:3127 [inline]
   __do_sys_bpf+0x1ea9/0x4f00 kernel/bpf/syscall.c:4406
   do_syscall_64+0x2d/0x70 arch/x86/entry/common.c:46
   entry_SYSCALL_64_after_hwframe+0x44/0xae
  [...]

Generally speaking, KUBSAN reports from the kernel should be fixed.
However, in case of BPF, this particular report caused concerns since
the large shift is not wrong from BPF point of view, just undefined.
In the verifier, K-based shifts that are >= {64,32} (depending on the
bitwidth of the instruction) are already rejected. The register-based
cases were not given their content might not be known at verification
time. Ideas such as verifier instruction rewrite with an additional
AND instruction for the source register were brought up, but regularly
rejected due to the additional runtime overhead they incur.

As Edward Cree rightly put it:

  Shifts by more than insn bitness are legal in the BPF ISA; they are
  implementation-defined behaviour [of the underlying architecture],
  rather than UB, and have been made legal for performance reasons.
  Each of the JIT backends compiles the BPF shift operations to machine
  instructions which produce implementation-defined results in such a
  case; the resulting contents of the register may be arbitrary but
  program behaviour as a whole remains defined.

  Guard checks in the fast path (i.e. affecting JITted code) will thus
  not be accepted.

  The case of division by zero is not truly analogous here, as division
  instructions on many of the JIT-targeted architectures will raise a
  machine exception / fault on division by zero, whereas (to the best
  of my knowledge) none will do so on an out-of-bounds shift.

Given the KUBSAN report only affects the BPF interpreter, but not JITs,
one solution is to add the ANDs with 63 or 31 into ___bpf_prog_run().
That would make the shifts defined, and thus shuts up KUBSAN, and the
compiler would optimize out the AND on any CPU that interprets the shift
amounts modulo the width anyway (e.g., confirmed from disassembly that
on x86-64 and arm64 the generated interpreter code is the same before
and after this fix).

The BPF interpreter is slow path, and most likely compiled out anyway
as distros select BPF_JIT_ALWAYS_ON to avoid speculative execution of
BPF instructions by the interpreter. Given the main argument was to
avoid sacrificing performance, the fact that the AND is optimized away
from compiler for mainstream archs helps as well as a solution moving
forward. Also add a comment on LSH/RSH/ARSH translation for JIT authors
to provide guidance when they see the ___bpf_prog_run() interpreter
code and use it as a model for a new JIT backend.

Reported-by: syzbot+bed360704c521841c85d@syzkaller.appspotmail.com
Reported-by: Kurt Manucredo 
Signed-off-by: Eric Biggers 
Co-developed-by: Eric Biggers 
Signed-off-by: Daniel Borkmann 
Acked-by: Alexei Starovoitov 
Acked-by: Andrii Nakryiko 
Tested-by: syzbot+bed360704c521841c85d@syzkaller.appspotmail.com
Cc: Edward Cree 
Link: https://lore.kernel.org/bpf/0000000000008f912605bd30d5d7@google.com
Link: https://lore.kernel.org/bpf/bac16d8d-c174-bdc4-91bd-bfa62b410190@gmail.com
Signed-off-by: Sasha Levin

rcu: Invoke rcu_spawn_core_kthreads() from rcu_spawn_gp_kthread()

2021-07-14T14:53:36+00:00

[ Upstream commit 8e4b1d2bc198e34b48fc7cc3a3c5a2fcb269e271 ]

Currently, rcu_spawn_core_kthreads() is invoked via an early_initcall(),
which works, except that rcu_spawn_gp_kthread() is also invoked via an
early_initcall() and rcu_spawn_core_kthreads() relies on adjustments to
kthread_prio that are carried out by rcu_spawn_gp_kthread().  There is
no guaranttee of ordering among early_initcall() handlers, and thus no
guarantee that kthread_prio will be properly checked and range-limited
at the time that rcu_spawn_core_kthreads() needs it.

In most cases, this bug is harmless.  After all, the only reason that
rcu_spawn_gp_kthread() adjusts the value of kthread_prio is if the user
specified a nonsensical value for this boot parameter, which experience
indicates is rare.

Nevertheless, a bug is a bug.  This commit therefore causes the
rcu_spawn_core_kthreads() function to be invoked directly from
rcu_spawn_gp_kthread() after any needed adjustments to kthread_prio have
been carried out.

Fixes: 48d07c04b4cc ("rcu: Enable elimination of Tree-RCU softirq processing")
Signed-off-by: Paul E. McKenney 
Signed-off-by: Sasha Levin

sched/uclamp: Fix uclamp_tg_restrict()

2021-07-14T14:53:24+00:00

[ Upstream commit 0213b7083e81f4acd69db32cb72eb4e5f220329a ]

Now cpu.uclamp.min acts as a protection, we need to make sure that the
uclamp request of the task is within the allowed range of the cgroup,
that is it is clamp()'ed correctly by tg->uclamp[UCLAMP_MIN] and
tg->uclamp[UCLAMP_MAX].

As reported by Xuewen [1] we can have some corner cases where there's
inversion between uclamp requested by task (p) and the uclamp values of
the taskgroup it's attached to (tg). Following table demonstrates
2 corner cases:

	           |  p  |  tg  |  effective
	-----------+-----+------+-----------
	CASE 1
	-----------+-----+------+-----------
	uclamp_min | 60% | 0%   |  60%
	-----------+-----+------+-----------
	uclamp_max | 80% | 50%  |  50%
	-----------+-----+------+-----------
	CASE 2
	-----------+-----+------+-----------
	uclamp_min | 0%  | 30%  |  30%
	-----------+-----+------+-----------
	uclamp_max | 20% | 50%  |  20%
	-----------+-----+------+-----------

With this fix we get:

	           |  p  |  tg  |  effective
	-----------+-----+------+-----------
	CASE 1
	-----------+-----+------+-----------
	uclamp_min | 60% | 0%   |  50%
	-----------+-----+------+-----------
	uclamp_max | 80% | 50%  |  50%
	-----------+-----+------+-----------
	CASE 2
	-----------+-----+------+-----------
	uclamp_min | 0%  | 30%  |  30%
	-----------+-----+------+-----------
	uclamp_max | 20% | 50%  |  30%
	-----------+-----+------+-----------

Additionally uclamp_update_active_tasks() must now unconditionally
update both UCLAMP_MIN/MAX because changing the tg's UCLAMP_MAX for
instance could have an impact on the effective UCLAMP_MIN of the tasks.

	           |  p  |  tg  |  effective
	-----------+-----+------+-----------
	old
	-----------+-----+------+-----------
	uclamp_min | 60% | 0%   |  50%
	-----------+-----+------+-----------
	uclamp_max | 80% | 50%  |  50%
	-----------+-----+------+-----------
	*new*
	-----------+-----+------+-----------
	uclamp_min | 60% | 0%   | *60%*
	-----------+-----+------+-----------
	uclamp_max | 80% |*70%* | *70%*
	-----------+-----+------+-----------

[1] https://lore.kernel.org/lkml/CAB8ipk_a6VFNjiEnHRHkUMBKbA+qzPQvhtNjJ_YNzQhqV_o8Zw@mail.gmail.com/

Fixes: 0c18f2ecfcc2 ("sched/uclamp: Fix wrong implementation of cpu.uclamp.min")
Reported-by: Xuewen Yan 
Signed-off-by: Qais Yousef 
Signed-off-by: Peter Zijlstra (Intel) 
Link: https://lkml.kernel.org/r/20210617165155.3774110-1-qais.yousef@arm.com
Signed-off-by: Sasha Levin

sched/rt: Fix Deadline utilization tracking during policy change

2021-07-14T14:53:24+00:00

[ Upstream commit d7d607096ae6d378b4e92d49946d22739c047d4c ]

DL keeps track of the utilization on a per-rq basis with the structure
avg_dl. This utilization is updated during task_tick_dl(),
put_prev_task_dl() and set_next_task_dl(). However, when the current
running task changes its policy, set_next_task_dl() which would usually
take care of updating the utilization when the rq starts running DL
tasks, will not see a such change, leaving the avg_dl structure outdated.
When that very same task will be dequeued later, put_prev_task_dl() will
then update the utilization, based on a wrong last_update_time, leading to
a huge spike in the DL utilization signal.

The signal would eventually recover from this issue after few ms. Even
if no DL tasks are run, avg_dl is also updated in
__update_blocked_others(). But as the CPU capacity depends partly on the
avg_dl, this issue has nonetheless a significant impact on the scheduler.

Fix this issue by ensuring a load update when a running task changes
its policy to DL.

Fixes: 3727e0e ("sched/dl: Add dl_rq utilization tracking")
Signed-off-by: Vincent Donnefort 
Signed-off-by: Peter Zijlstra (Intel) 
Reviewed-by: Vincent Guittot 
Link: https://lore.kernel.org/r/1624271872-211872-3-git-send-email-vincent.donnefort@arm.com
Signed-off-by: Sasha Levin

sched/rt: Fix RT utilization tracking during policy change

2021-07-14T14:53:24+00:00

[ Upstream commit fecfcbc288e9f4923f40fd23ca78a6acdc7fdf6c ]

RT keeps track of the utilization on a per-rq basis with the structure
avg_rt. This utilization is updated during task_tick_rt(),
put_prev_task_rt() and set_next_task_rt(). However, when the current
running task changes its policy, set_next_task_rt() which would usually
take care of updating the utilization when the rq starts running RT tasks,
will not see a such change, leaving the avg_rt structure outdated. When
that very same task will be dequeued later, put_prev_task_rt() will then
update the utilization, based on a wrong last_update_time, leading to a
huge spike in the RT utilization signal.

The signal would eventually recover from this issue after few ms. Even if
no RT tasks are run, avg_rt is also updated in __update_blocked_others().
But as the CPU capacity depends partly on the avg_rt, this issue has
nonetheless a significant impact on the scheduler.

Fix this issue by ensuring a load update when a running task changes
its policy to RT.

Fixes: 371bf427 ("sched/rt: Add rt_rq utilization tracking")
Signed-off-by: Vincent Donnefort 
Signed-off-by: Peter Zijlstra (Intel) 
Reviewed-by: Vincent Guittot 
Link: https://lore.kernel.org/r/1624271872-211872-2-git-send-email-vincent.donnefort@arm.com
Signed-off-by: Sasha Levin

sched/uclamp: Fix locking around cpu_util_update_eff()

2021-07-14T14:53:20+00:00

[ Upstream commit 93b73858701fd01de26a4a874eb95f9b7156fd4b ]

cpu_cgroup_css_online() calls cpu_util_update_eff() without holding the
uclamp_mutex or rcu_read_lock() like other call sites, which is
a mistake.

The uclamp_mutex is required to protect against concurrent reads and
writes that could update the cgroup hierarchy.

The rcu_read_lock() is required to traverse the cgroup data structures
in cpu_util_update_eff().

Surround the caller with the required locks and add some asserts to
better document the dependency in cpu_util_update_eff().

Fixes: 7226017ad37a ("sched/uclamp: Fix a bug in propagating uclamp value in new cgroups")
Reported-by: Quentin Perret 
Signed-off-by: Qais Yousef 
Signed-off-by: Peter Zijlstra (Intel) 
Link: https://lkml.kernel.org/r/20210510145032.1934078-3-qais.yousef@arm.com
Signed-off-by: Sasha Levin