Vince reported that when we do IOC_ENABLE/IOC_DISABLE while the task
is SIGSTOP'ed state the timestamps go wobbly.
It turns out we indeed fail to correctly account time while in 'OFF'
state and doing IOC_ENABLE without getting scheduled in exposes the
problem.
Further thinking about this problem, it occurred to me that we can
suffer a similar fate when we migrate an uncore event between CPUs.
The perf_event_install() on the 'new' CPU will do add_event_to_ctx()
which will reset all the time stamp, resulting in a subsequent
update_event_times() to overwrite the total_time_* fields with smaller
values.
Reported-by: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Enable the newly added jump opcodes, main parts are in two
different areas, namely direct packet access and dynamic map
value access. For the direct packet access, we now allow for
the following two new patterns to match in order to trigger
markings with find_good_pkt_pointers():
Variant 1 (access ok when taking the branch):
0: (61) r2 = *(u32 *)(r1 +76)
1: (61) r3 = *(u32 *)(r1 +80)
2: (bf) r0 = r2
3: (07) r0 += 8
4: (ad) if r0 < r3 goto pc+2
R0=pkt(id=0,off=8,r=0) R1=ctx R2=pkt(id=0,off=0,r=0)
R3=pkt_end R10=fp
5: (b7) r0 = 0
6: (95) exit
from 4 to 7: R0=pkt(id=0,off=8,r=8) R1=ctx
R2=pkt(id=0,off=0,r=8) R3=pkt_end R10=fp
7: (71) r0 = *(u8 *)(r2 +0)
8: (05) goto pc-4
5: (b7) r0 = 0
6: (95) exit
processed 11 insns, stack depth 0
Variant 2 (access ok on fall-through):
0: (61) r2 = *(u32 *)(r1 +76)
1: (61) r3 = *(u32 *)(r1 +80)
2: (bf) r0 = r2
3: (07) r0 += 8
4: (bd) if r3 <= r0 goto pc+1
R0=pkt(id=0,off=8,r=8) R1=ctx R2=pkt(id=0,off=0,r=8)
R3=pkt_end R10=fp
5: (71) r0 = *(u8 *)(r2 +0)
6: (b7) r0 = 1
7: (95) exit
from 4 to 6: R0=pkt(id=0,off=8,r=0) R1=ctx
R2=pkt(id=0,off=0,r=0) R3=pkt_end R10=fp
6: (b7) r0 = 1
7: (95) exit
processed 10 insns, stack depth 0
The above two basically just swap the branches where we need
to handle an exception and allow packet access compared to the
two already existing variants for find_good_pkt_pointers().
For the dynamic map value access, we add the new instructions
to reg_set_min_max() and reg_set_min_max_inv() in order to
learn bounds. Verifier test cases for both are added in a
follow-up patch.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, eBPF only understands BPF_JGT (>), BPF_JGE (>=),
BPF_JSGT (s>), BPF_JSGE (s>=) instructions, this means that
particularly *JLT/*JLE counterparts involving immediates need
to be rewritten from e.g. X < [IMM] by swapping arguments into
[IMM] > X, meaning the immediate first is required to be loaded
into a register Y := [IMM], such that then we can compare with
Y > X. Note that the destination operand is always required to
be a register.
This has the downside of having unnecessarily increased register
pressure, meaning complex program would need to spill other
registers temporarily to stack in order to obtain an unused
register for the [IMM]. Loading to registers will thus also
affect state pruning since we need to account for that register
use and potentially those registers that had to be spilled/filled
again. As a consequence slightly more stack space might have
been used due to spilling, and BPF programs are a bit longer
due to extra code involving the register load and potentially
required spill/fills.
Thus, add BPF_JLT (<), BPF_JLE (<=), BPF_JSLT (s<), BPF_JSLE (s<=)
counterparts to the eBPF instruction set. Modifying LLVM to
remove the NegateCC() workaround in a PoC patch at [1] and
allowing it to also emit the new instructions resulted in
cilium's BPF programs that are injected into the fast-path to
have a reduced program length in the range of 2-3% (e.g.
accumulated main and tail call sections from one of the object
file reduced from 4864 to 4729 insns), reduced complexity in
the range of 10-30% (e.g. accumulated sections reduced in one
of the cases from 116432 to 88428 insns), and reduced stack
usage in the range of 1-5% (e.g. accumulated sections from one
of the object files reduced from 824 to 784b).
The modification for LLVM will be incorporated in a backwards
compatible way. Plan is for LLVM to have i) a target specific
option to offer a possibility to explicitly enable the extension
by the user (as we have with -m target specific extensions today
for various CPU insns), and ii) have the kernel checked for
presence of the extensions and enable them transparently when
the user is selecting more aggressive options such as -march=native
in a bpf target context. (Other frontends generating BPF byte
code, e.g. ply can probe the kernel directly for its code
generation.)
[1] https://github.com/borkmann/llvm/tree/bpf-insns
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The UDP offload conflict is dealt with by simply taking what is
in net-next where we have removed all of the UFO handling code
entirely.
The TCP conflict was a case of local variables in a function
being removed from both net and net-next.
In netvsc we had an assignment right next to where a missing
set of u64 stats sync object inits were added.
Signed-off-by: David S. Miller <davem@davemloft.net>
CPUFREQ_ENTRY_INVALID is a special symbol which is used to specify that
an entry in the cpufreq table is invalid. But using it outside of the
scope of the cpufreq table looks a bit incorrect.
We can represent an invalid frequency by writing it as 0 instead if we
need. Note that it is already done that way for the return value of the
->get() callback.
Lets do the same for ->fast_switch() and not use CPUFREQ_ENTRY_INVALID
outside of the scope of cpufreq table.
Also update the comment over cpufreq_driver_fast_switch() to clearly
mention what this returns.
None of the drivers return CPUFREQ_ENTRY_INVALID as of now from
->fast_switch() callback and so we don't need to update any of those.
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Commit 65d8fc777f ("futex: Remove requirement for lock_page() in
get_futex_key()") removed an unnecessary lock_page() with the
side-effect that page->mapping needed to be treated very carefully.
Two defensive warnings were added in case any assumption was missed and
the first warning assumed a correct application would not alter a
mapping backing a futex key. Since merging, it has not triggered for
any unexpected case but Mark Rutland reported the following bug
triggering due to the first warning.
kernel BUG at kernel/futex.c:679!
Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
Modules linked in:
CPU: 0 PID: 3695 Comm: syz-executor1 Not tainted 4.13.0-rc3-00020-g307fec773ba3 #3
Hardware name: linux,dummy-virt (DT)
task: ffff80001e271780 task.stack: ffff000010908000
PC is at get_futex_key+0x6a4/0xcf0 kernel/futex.c:679
LR is at get_futex_key+0x6a4/0xcf0 kernel/futex.c:679
pc : [<ffff00000821ac14>] lr : [<ffff00000821ac14>] pstate: 80000145
The fact that it's a bug instead of a warning was due to an unrelated
arm64 problem, but the warning itself triggered because the underlying
mapping changed.
This is an application issue but from a kernel perspective it's a
recoverable situation and the warning is unnecessary so this patch
removes the warning. The warning may potentially be triggered with the
following test program from Mark although it may be necessary to adjust
NR_FUTEX_THREADS to be a value smaller than the number of CPUs in the
system.
#include <linux/futex.h>
#include <pthread.h>
#include <stdio.h>
#include <stdlib.h>
#include <sys/mman.h>
#include <sys/syscall.h>
#include <sys/time.h>
#include <unistd.h>
#define NR_FUTEX_THREADS 16
pthread_t threads[NR_FUTEX_THREADS];
void *mem;
#define MEM_PROT (PROT_READ | PROT_WRITE)
#define MEM_SIZE 65536
static int futex_wrapper(int *uaddr, int op, int val,
const struct timespec *timeout,
int *uaddr2, int val3)
{
syscall(SYS_futex, uaddr, op, val, timeout, uaddr2, val3);
}
void *poll_futex(void *unused)
{
for (;;) {
futex_wrapper(mem, FUTEX_CMP_REQUEUE_PI, 1, NULL, mem + 4, 1);
}
}
int main(int argc, char *argv[])
{
int i;
mem = mmap(NULL, MEM_SIZE, MEM_PROT,
MAP_SHARED | MAP_ANONYMOUS, -1, 0);
printf("Mapping @ %p\n", mem);
printf("Creating futex threads...\n");
for (i = 0; i < NR_FUTEX_THREADS; i++)
pthread_create(&threads[i], NULL, poll_futex, NULL);
printf("Flipping mapping...\n");
for (;;) {
mmap(mem, MEM_SIZE, MEM_PROT,
MAP_FIXED | MAP_SHARED | MAP_ANONYMOUS, -1, 0);
}
return 0;
}
Reported-and-tested-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: stable@vger.kernel.org # 4.7+
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The more detailed value tracking can reduce the effectiveness of pruning
for some programs. So, to avoid rejecting previously valid programs, up
the limit to 128kinsns. Hopefully we will be able to bring this back
down later by improving pruning performance.
Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Allows us to, sometimes, combine information from a signed check of one
bound and an unsigned check of the other.
We now track the full range of possible values, rather than restricting
ourselves to [0, 1<<30) and considering anything beyond that as
unknown. While this is probably not necessary, it makes the code more
straightforward and symmetrical between signed and unsigned bounds.
Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Unifies adjusted and unadjusted register value types (e.g. FRAME_POINTER is
now just a PTR_TO_STACK with zero offset).
Tracks value alignment by means of tracking known & unknown bits. This
also replaces the 'reg->imm' (leading zero bits) calculations for (what
were) UNKNOWN_VALUEs.
If pointer leaks are allowed, and adjust_ptr_min_max_vals returns -EACCES,
treat the pointer as an unknown scalar and try again, because we might be
able to conclude something about the result (e.g. pointer & 0x40 is either
0 or 0x40).
Verifier hooks in the netronome/nfp driver were changed to match the new
data structures.
Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Originally we used a mutex to protect concurrent devmap update
and delete operations from racing with netdev unregister notifier
callbacks.
The notifier hook is needed because we increment the netdev ref
count when a dev is added to the devmap. This ensures the netdev
reference is valid in the datapath. However, we don't want to block
unregister events, hence the initial mutex and notifier handler.
The concern was in the notifier hook we search the map for dev
entries that hold a refcnt on the net device being torn down. But,
in order to do this we require two steps,
(i) dereference the netdev: dev = rcu_dereference(map[i])
(ii) test ifindex: dev->ifindex == removing_ifindex
and then finally we can swap in the NULL dev in the map via an
xchg operation,
xchg(map[i], NULL)
The danger here is a concurrent update could run a different
xchg op concurrently leading us to replace the new dev with a
NULL dev incorrectly.
CPU 1 CPU 2
notifier hook bpf devmap update
dev = rcu_dereference(map[i])
dev = rcu_dereference(map[i])
xchg(map[i]), new_dev);
rcu_call(dev,...)
xchg(map[i], NULL)
The above flow would create the incorrect state with the dev
reference in the update path being lost. To resolve this the
original code used a mutex around the above block. However,
updates, deletes, and lookups occur inside rcu critical sections
so we can't use a mutex in this context safely.
Fortunately, by writing slightly better code we can avoid the
mutex altogether. If CPU 1 in the above example uses a cmpxchg
and _only_ replaces the dev reference in the map when it is in
fact the expected dev the race is removed completely. The two
cases being illustrated here, first the race condition,
CPU 1 CPU 2
notifier hook bpf devmap update
dev = rcu_dereference(map[i])
dev = rcu_dereference(map[i])
xchg(map[i]), new_dev);
rcu_call(dev,...)
odev = cmpxchg(map[i], dev, NULL)
Now we can test the cmpxchg return value, detect odev != dev and
abort. Or in the good case,
CPU 1 CPU 2
notifier hook bpf devmap update
dev = rcu_dereference(map[i])
odev = cmpxchg(map[i], dev, NULL)
[...]
Now 'odev == dev' and we can do proper cleanup.
And viola the original race we tried to solve with a mutex is
corrected and the trace noted by Sasha below is resolved due
to removal of the mutex.
Note: When walking the devmap and removing dev references as needed
we depend on the core to fail any calls to dev_get_by_index() using
the ifindex of the device being removed. This way we do not race with
the user while searching the devmap.
Additionally, the mutex was also protecting list add/del/read on
the list of maps in-use. This patch converts this to an RCU list
and spinlock implementation. This protects the list from concurrent
alloc/free operations. The notifier hook walks this list so it uses
RCU read semantics.
BUG: sleeping function called from invalid context at kernel/locking/mutex.c:747
in_atomic(): 1, irqs_disabled(): 0, pid: 16315, name: syz-executor1
1 lock held by syz-executor1/16315:
#0: (rcu_read_lock){......}, at: [<ffffffff8c363bc2>] map_delete_elem kernel/bpf/syscall.c:577 [inline]
#0: (rcu_read_lock){......}, at: [<ffffffff8c363bc2>] SYSC_bpf kernel/bpf/syscall.c:1427 [inline]
#0: (rcu_read_lock){......}, at: [<ffffffff8c363bc2>] SyS_bpf+0x1d32/0x4ba0 kernel/bpf/syscall.c:1388
Fixes: 2ddf71e23c ("net: add notifier hooks for devmap bpf map")
Reported-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, bpf programs cannot be attached to sys_enter_* and sys_exit_*
style tracepoints. The iovisor/bcc issue #748
(https://github.com/iovisor/bcc/issues/748) documents this issue.
For example, if you try to attach a bpf program to tracepoints
syscalls/sys_enter_newfstat, you will get the following error:
# ./tools/trace.py t:syscalls:sys_enter_newfstat
Ioctl(PERF_EVENT_IOC_SET_BPF): Invalid argument
Failed to attach BPF to tracepoint
The main reason is that syscalls/sys_enter_* and syscalls/sys_exit_*
tracepoints are treated differently from other tracepoints and there
is no bpf hook to it.
This patch adds bpf support for these syscalls tracepoints by
. permitting bpf attachment in ioctl PERF_EVENT_IOC_SET_BPF
. calling bpf programs in perf_syscall_enter and perf_syscall_exit
The legality of bpf program ctx access is also checked.
Function trace_event_get_offsets returns correct max offset for each
specific syscall tracepoint, which is compared against the maximum offset
access in bpf program.
Signed-off-by: Yonghong Song <yhs@fb.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Saeed Mahameed says:
====================
mlx5-shared-2017-08-07
This series includes some mlx5 updates for both net-next and rdma trees.
From Saeed,
Core driver updates to allow selectively building the driver with
or without some large driver components, such as
- E-Switch (Ethernet SRIOV support).
- Multi-Physical Function Switch (MPFs) support.
For that we split E-Switch and MPFs functionalities into separate files.
From Erez,
Delay mlx5_core events when mlx5 interfaces, namely mlx5_ib, registration
is taking place and until it completes.
From Rabie,
Increase the maximum supported flow counters.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
The latest change of compat_sys_sigpending in commit 8f13621abc
("sigpending(): move compat to native") has broken it in two ways.
First, it tries to write 4 bytes more than userspace expects:
sizeof(old_sigset_t) == sizeof(long) == 8 instead of
sizeof(compat_old_sigset_t) == sizeof(u32) == 4.
Second, on big endian architectures these bytes are being written in the
wrong order.
This bug was found by strace test suite.
Reported-by: Anatoly Pugachev <matorola@gmail.com>
Inspired-by: Eugene Syromyatnikov <evgsyr@gmail.com>
Fixes: 8f13621abc ("sigpending(): move compat to native")
Signed-off-by: Dmitry V. Levin <ldv@altlinux.org>
Acked-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Modify the ACPI system sleep support setup code to select
suspend-to-idle as the default system sleep state if
(1) the ACPI_FADT_LOW_POWER_S0 flag is set in the FADT and
(2) the Low Power Idle S0 _DSM interface has been discovered and
(3) the default sleep state was not selected from the kernel command
line.
The main motivation for this change is that systems where the (1) and
(2) conditions are met typically ship with OSes that don't exercise
the S3 path in the platform firmware which remains untested and turns
out to be non-functional at least in some cases.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Tested-by: Mario Limonciello <mario.limonciello@dell.com>
Pull timer fix from Thomas Gleixner:
"A single fix for a multiplication overflow in the timer code on 32bit
systems"
* 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
timers: Fix overflow in get_next_timer_interrupt
In codepaths that use the begin/retry interface for reading
mems_allowed_seq with irqs disabled, there exists a race condition that
stalls the patch process after only modifying a subset of the
static_branch call sites.
This problem manifested itself as a deadlock in the slub allocator,
inside get_any_partial. The loop reads mems_allowed_seq value (via
read_mems_allowed_begin), performs the defrag operation, and then
verifies the consistency of mem_allowed via the read_mems_allowed_retry
and the cookie returned by xxx_begin.
The issue here is that both begin and retry first check if cpusets are
enabled via cpusets_enabled() static branch. This branch can be
rewritted dynamically (via cpuset_inc) if a new cpuset is created. The
x86 jump label code fully synchronizes across all CPUs for every entry
it rewrites. If it rewrites only one of the callsites (specifically the
one in read_mems_allowed_retry) and then waits for the
smp_call_function(do_sync_core) to complete while a CPU is inside the
begin/retry section with IRQs off and the mems_allowed value is changed,
we can hang.
This is because begin() will always return 0 (since it wasn't patched
yet) while retry() will test the 0 against the actual value of the seq
counter.
The fix is to use two different static keys: one for begin
(pre_enable_key) and one for retry (enable_key). In cpuset_inc(), we
first bump the pre_enable key to ensure that cpuset_mems_allowed_begin()
always return a valid seqcount if are enabling cpusets. Similarly, when
disabling cpusets via cpuset_dec(), we first ensure that callers of
cpuset_mems_allowed_retry() will start ignoring the seqcount value
before we let cpuset_mems_allowed_begin() return 0.
The relevant stack traces of the two stuck threads:
CPU: 1 PID: 1415 Comm: mkdir Tainted: G L 4.9.36-00104-g540c51286237 #4
Hardware name: Default string Default string/Hardware, BIOS 4.29.1-20170526215256 05/26/2017
task: ffff8817f9c28000 task.stack: ffffc9000ffa4000
RIP: smp_call_function_many+0x1f9/0x260
Call Trace:
smp_call_function+0x3b/0x70
on_each_cpu+0x2f/0x90
text_poke_bp+0x87/0xd0
arch_jump_label_transform+0x93/0x100
__jump_label_update+0x77/0x90
jump_label_update+0xaa/0xc0
static_key_slow_inc+0x9e/0xb0
cpuset_css_online+0x70/0x2e0
online_css+0x2c/0xa0
cgroup_apply_control_enable+0x27f/0x3d0
cgroup_mkdir+0x2b7/0x420
kernfs_iop_mkdir+0x5a/0x80
vfs_mkdir+0xf6/0x1a0
SyS_mkdir+0xb7/0xe0
entry_SYSCALL_64_fastpath+0x18/0xad
...
CPU: 2 PID: 1 Comm: init Tainted: G L 4.9.36-00104-g540c51286237 #4
Hardware name: Default string Default string/Hardware, BIOS 4.29.1-20170526215256 05/26/2017
task: ffff8818087c0000 task.stack: ffffc90000030000
RIP: int3+0x39/0x70
Call Trace:
<#DB> ? ___slab_alloc+0x28b/0x5a0
<EOE> ? copy_process.part.40+0xf7/0x1de0
__slab_alloc.isra.80+0x54/0x90
copy_process.part.40+0xf7/0x1de0
copy_process.part.40+0xf7/0x1de0
kmem_cache_alloc_node+0x8a/0x280
copy_process.part.40+0xf7/0x1de0
_do_fork+0xe7/0x6c0
_raw_spin_unlock_irq+0x2d/0x60
trace_hardirqs_on_caller+0x136/0x1d0
entry_SYSCALL_64_fastpath+0x5/0xad
do_syscall_64+0x27/0x350
SyS_clone+0x19/0x20
do_syscall_64+0x60/0x350
entry_SYSCALL64_slow_path+0x25/0x25
Link: http://lkml.kernel.org/r/20170731040113.14197-1-dmitriyz@waymo.com
Fixes: 46e700abc4 ("mm, page_alloc: remove unnecessary taking of a seqlock when cpusets are disabled")
Signed-off-by: Dima Zavin <dmitriyz@waymo.com>
Reported-by: Cliff Spradlin <cspradlin@waymo.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Christopher Lameter <cl@linux.com>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Each css_set directly points to the default cgroup it belongs to, so
there's no reason to walk the cgrp_links list on the default
hierarchy.
Signed-off-by: Tejun Heo <tj@kernel.org>
A cgroup can consume resources even after being deleted by a user.
For example, writing back dirty pages should be accounted and
limited, despite the corresponding cgroup might contain no processes
and being deleted by a user.
In the current implementation a cgroup can remain in such "dying" state
for an undefined amount of time. For instance, if a memory cgroup
contains a pge, mlocked by a process belonging to an other cgroup.
Although the lifecycle of a dying cgroup is out of user's control,
it's important to have some insight of what's going on under the hood.
In particular, it's handy to have a counter which will allow
to detect css leaks.
To solve this problem, add a cgroup.stat interface to
the base cgroup control files with the following metrics:
nr_descendants total number of visible descendant cgroups
nr_dying_descendants total number of dying descendant cgroups
Signed-off-by: Roman Gushchin <guro@fb.com>
Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Zefan Li <lizefan@huawei.com>
Cc: Waiman Long <longman@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: kernel-team@fb.com
Cc: cgroups@vger.kernel.org
Cc: linux-doc@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Creating cgroup hierearchies of unreasonable size can affect
overall system performance. A user might want to limit the
size of cgroup hierarchy. This is especially important if a user
is delegating some cgroup sub-tree.
To address this issue, introduce an ability to control
the size of cgroup hierarchy.
The cgroup.max.descendants control file allows to set the maximum
allowed number of descendant cgroups.
The cgroup.max.depth file controls the maximum depth of the cgroup
tree. Both are single value r/w files, with "max" default value.
The control files exist on each hierarchy level (including root).
When a new cgroup is created, we check the total descendants
and depth limits on each level, and if none of them are exceeded,
a new cgroup is created.
Only alive cgroups are counted, removed (dying) cgroups are
ignored.
Signed-off-by: Roman Gushchin <guro@fb.com>
Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Zefan Li <lizefan@huawei.com>
Cc: Waiman Long <longman@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: kernel-team@fb.com
Cc: cgroups@vger.kernel.org
Cc: linux-doc@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Chunyu Hu reported:
"per_cpu trace directories and files are created for all possible cpus,
but only the cpus which have ever been on-lined have their own per cpu
ring buffer (allocated by cpuhp threads). While trace_buffers_open, the
open handler for trace file 'trace_pipe_raw' is always trying to access
field of ring_buffer_per_cpu, and would panic with the NULL pointer.
Align the behavior of trace_pipe_raw with trace_pipe, that returns -NODEV
when openning it if that cpu does not have trace ring buffer.
Reproduce:
cat /sys/kernel/debug/tracing/per_cpu/cpu31/trace_pipe_raw
(cpu31 is never on-lined, this is a 16 cores x86_64 box)
Tested with:
1) boot with maxcpus=14, read trace_pipe_raw of cpu15.
Got -NODEV.
2) oneline cpu15, read trace_pipe_raw of cpu15.
Get the raw trace data.
Call trace:
[ 5760.950995] RIP: 0010:ring_buffer_alloc_read_page+0x32/0xe0
[ 5760.961678] tracing_buffers_read+0x1f6/0x230
[ 5760.962695] __vfs_read+0x37/0x160
[ 5760.963498] ? __vfs_read+0x5/0x160
[ 5760.964339] ? security_file_permission+0x9d/0xc0
[ 5760.965451] ? __vfs_read+0x5/0x160
[ 5760.966280] vfs_read+0x8c/0x130
[ 5760.967070] SyS_read+0x55/0xc0
[ 5760.967779] do_syscall_64+0x67/0x150
[ 5760.968687] entry_SYSCALL64_slow_path+0x25/0x25"
This was introduced by the addition of the feature to reuse reader pages
instead of re-allocating them. The problem is that the allocation of a
reader page (which is per cpu) does not check if the cpu is online and set
up for the ring buffer.
Link: http://lkml.kernel.org/r/1500880866-1177-1-git-send-email-chuhu@redhat.com
Cc: stable@vger.kernel.org
Fixes: 73a757e631 ("ring-buffer: Return reader page back into existing ring buffer")
Reported-by: Chunyu Hu <chuhu@redhat.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
The clear_boot_tracer function is used to reset the default_bootup_tracer
string to prevent it from being accessed after boot, as it originally points
to init data. But since clear_boot_tracer() is called via the
init_lateinit() call, it races with the initcall for registering the hwlat
tracer. If someone adds "ftrace=hwlat" to the kernel command line, depending
on how the linker sets up the text, the saved command line may be cleared,
and the hwlat tracer never is initialized.
Simply have the clear_boot_tracer() be called by initcall_lateinit_sync() as
that's for tasks to be called after lateinit.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=196551
Cc: stable@vger.kernel.org
Fixes: e7c15cd8a ("tracing: Added hardware latency tracer")
Reported-by: Zamir SUN <sztsian@gmail.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
'perf cqm' never worked due to the incompatibility between perf
infrastructure and cqm hardware support. The hardware uses RMIDs to
track the llc occupancy of tasks and these RMIDs are per package. This
makes monitoring a hierarchy like cgroup along with monitoring of tasks
separately difficult and several patches sent to lkml to fix them were
NACKed. Further more, the following issues in the current perf cqm make
it almost unusable:
1. No support to monitor the same group of tasks for which we do
allocation using resctrl.
2. It gives random and inaccurate data (mostly 0s) once we run out
of RMIDs due to issues in Recycling.
3. Recycling results in inaccuracy of data because we cannot
guarantee that the RMID was stolen from a task when it was not
pulling data into cache or even when it pulled the least data. Also
for monitoring llc_occupancy, if we stop using an RMID_x and then
start using an RMID_y after we reclaim an RMID from an other event,
we miss accounting all the occupancy that was tagged to RMID_x at a
later perf_count.
2. Recycling code makes the monitoring code complex including
scheduling because the event can lose RMID any time. Since MBM
counters count bandwidth for a period of time by taking snap shot of
total bytes at two different times, recycling complicates the way we
count MBM in a hierarchy. Also we need a spin lock while we do the
processing to account for MBM counter overflow. We also currently
use a spin lock in scheduling to prevent the RMID from being taken
away.
4. Lack of support when we run different kind of event like task,
system-wide and cgroup events together. Data mostly prints 0s. This
is also because we can have only one RMID tied to a cpu as defined
by the cqm hardware but a perf can at the same time tie multiple
events during one sched_in.
5. No support of monitoring a group of tasks. There is partial support
for cgroup but it does not work once there is a hierarchy of cgroups
or if we want to monitor a task in a cgroup and the cgroup itself.
6. No support for monitoring tasks for the lifetime without perf
overhead.
7. It reported the aggregate cache occupancy or memory bandwidth over
all sockets. But most cloud and VMM based use cases want to know the
individual per-socket usage.
Signed-off-by: Vikas Shivappa <vikas.shivappa@linux.intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: ravi.v.shankar@intel.com
Cc: tony.luck@intel.com
Cc: fenghua.yu@intel.com
Cc: peterz@infradead.org
Cc: eranian@google.com
Cc: vikas.shivappa@intel.com
Cc: ak@linux.intel.com
Cc: davidcc@google.com
Cc: reinette.chatre@intel.com
Link: http://lkml.kernel.org/r/1501017287-28083-2-git-send-email-vikas.shivappa@linux.intel.com
Two minor conflicts in virtio_net driver (bug fix overlapping addition
of a helper) and MAINTAINERS (new driver edit overlapping revamp of
PHY entry).
Signed-off-by: David S. Miller <davem@davemloft.net>
With Android UI and benchmarks the latency of cpufreq response to
certain scheduling events can become very critical. Currently, callbacks
into cpufreq governors are only made from the scheduler if the target
CPU of the event is the same as the current CPU. This means there are
certain situations where a target CPU may not run the cpufreq governor
for some time.
One testcase to show this behavior is where a task starts running on
CPU0, then a new task is also spawned on CPU0 by a task on CPU1. If the
system is configured such that the new tasks should receive maximum
demand initially, this should result in CPU0 increasing frequency
immediately. But because of the above mentioned limitation though, this
does not occur.
This patch updates the scheduler core to call the cpufreq callbacks for
remote CPUs as well.
The schedutil, ondemand and conservative governors are updated to
process cpufreq utilization update hooks called for remote CPUs where
the remote CPU is managed by the cpufreq policy of the local CPU.
The intel_pstate driver is updated to always reject remote callbacks.
This is tested with couple of usecases (Android: hackbench, recentfling,
galleryfling, vellamo, Ubuntu: hackbench) on ARM hikey board (64 bit
octa-core, single policy). Only galleryfling showed minor improvements,
while others didn't had much deviation.
The reason being that this patch only targets a corner case, where
following are required to be true to improve performance and that
doesn't happen too often with these tests:
- Task is migrated to another CPU.
- The task has high demand, and should take the target CPU to higher
OPPs.
- And the target CPU doesn't call into the cpufreq governor until the
next tick.
Based on initial work from Steve Muckle.
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Acked-by: Saravana Kannan <skannan@codeaurora.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Pull networking fixes from David Miller:
1) Handle notifier registry failures properly in tun/tap driver, from
Tonghao Zhang.
2) Fix bpf verifier handling of subtraction bounds and add a testcase
for this, from Edward Cree.
3) Increase reset timeout in ftgmac100 driver, from Ben Herrenschmidt.
4) Fix use after free in prd_retire_rx_blk_timer_exired() in AF_PACKET,
from Cong Wang.
5) Fix SElinux regression due to recent UDP optimizations, from Paolo
Abeni.
6) We accidently increment IPSTATS_MIB_FRAGFAILS in the ipv6 code
paths, fix from Stefano Brivio.
7) Fix some mem leaks in dccp, from Xin Long.
8) Adjust MDIO_BUS kconfig deps to avoid build errors, from Arnd
Bergmann.
9) Mac address length check and buffer size fixes from Cong Wang.
10) Don't leak sockets in ipv6 udp early demux, from Paolo Abeni.
11) Fix return value when copy_from_user() fails in
bpf_prog_get_info_by_fd(), from Daniel Borkmann.
12) Handle PHY_HALTED properly in phy library state machine, from
Florian Fainelli.
13) Fix OOPS in fib_sync_down_dev(), from Ido Schimmel.
14) Fix truesize calculation in virtio_net which led to performance
regressions, from Michael S Tsirkin.
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (76 commits)
samples/bpf: fix bpf tunnel cleanup
udp6: fix jumbogram reception
ppp: Fix a scheduling-while-atomic bug in del_chan
Revert "net: bcmgenet: Remove init parameter from bcmgenet_mii_config"
virtio_net: fix truesize for mergeable buffers
mv643xx_eth: fix of_irq_to_resource() error check
MAINTAINERS: Add more files to the PHY LIBRARY section
ipv4: fib: Fix NULL pointer deref during fib_sync_down_dev()
net: phy: Correctly process PHY_HALTED in phy_stop_machine()
sunhme: fix up GREG_STAT and GREG_IMASK register offsets
bpf: fix bpf_prog_get_info_by_fd to dump correct xlated_prog_len
tcp: avoid bogus gcc-7 array-bounds warning
net: tc35815: fix spelling mistake: "Intterrupt" -> "Interrupt"
bpf: don't indicate success when copy_from_user fails
udp6: fix socket leak on early demux
net: thunderx: Fix BGX transmit stall due to underflow
Revert "vhost: cache used event for better performance"
team: use a larger struct for mac address
net: check dev->addr_len for dev_set_mac_address()
phy: bcm-ns-usb3: fix MDIO_BUS dependency
...
Pull cgroup fixes from Tejun Heo:
"Several cgroup bug fixes.
- cgroup core was calling a migration callback on empty migrations,
which could make cpuset crash.
- There was a very subtle bug where the controller interface files
aren't created directly when cgroup2 is mounted. Because later
operations create them, this bug didn't get noticed earlier.
- Failed writes to cgroup.subtree_control were incorrectly returning
zero"
* 'for-4.13-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup:
cgroup: fix error return value from cgroup_subtree_control()
cgroup: create dfl_root files on subsys registration
cgroup: don't call migration methods if there are no tasks to migrate
Pull workqueue fixes from Tejun Heo:
"Two notable fixes.
- While adding NUMA affinity support to unbound workqueues, the
assumption that an unbound workqueue with max_active == 1 is
ordered was broken.
The plan was to use explicit alloc_ordered_workqueue() for those
cases. Unfortunately, I forgot to update the documentation properly
and we grew a handful of use cases which depend on that assumption.
While we want to convert them to alloc_ordered_workqueue(), we
don't really lose anything by enforcing ordered execution on
unbound max_active == 1 workqueues and it doesn't make sense to
risk subtle bugs. Restore the assumption.
- Workqueue assumes that CPU <-> NUMA node mapping remains static.
This is a general assumption - we don't have any synchronization
mechanism around CPU <-> node mapping. Unfortunately, powerpc may
change the mapping dynamically leading to crashes. Michael added a
workaround so that we at least don't crash while powerpc hotplug
code gets updated"
* 'for-4.13-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq:
workqueue: Work around edge cases for calc of pool's cpumask
workqueue: implicit ordered attribute should be overridable
workqueue: restore WQ_UNBOUND/max_active==1 to be ordered
This patch replaces an rwlock and raw notifier by an atomic notifier
protected by a spin_lock and RCU.
The main reason for this change is due to a 'scheduling while atomic'
bug with RT kernels on ARM/ARM64. On ARM/ARM64, the rwlock
cpu_pm_notifier_lock in cpu_pm_enter/exit() causes a potential
schedule after IRQ disable in the idle call chain:
cpu_startup_entry
cpu_idle_loop
local_irq_disable()
cpuidle_idle_call
call_cpuidle
cpuidle_enter
cpuidle_enter_state
->enter :arm_enter_idle_state
cpu_pm_enter/exit
CPU_PM_CPU_IDLE_ENTER
read_lock(&cpu_pm_notifier_lock); <-- sleep in idle
__rt_spin_lock();
schedule();
The kernel panic is here:
[ 4.609601] BUG: scheduling while atomic: swapper/1/0/0x00000002
[ 4.609608] [<ffff0000086fae70>] arm_enter_idle_state+0x18/0x70
[ 4.609614] Modules linked in:
[ 4.609615] [<ffff0000086f9298>] cpuidle_enter_state+0xf0/0x218
[ 4.609620] [<ffff0000086f93f8>] cpuidle_enter+0x18/0x20
[ 4.609626] Preemption disabled at:
[ 4.609627] [<ffff0000080fa234>] call_cpuidle+0x24/0x40
[ 4.609635] [<ffff000008882fa4>] schedule_preempt_disabled+0x1c/0x28
[ 4.609639] [<ffff0000080fa49c>] cpu_startup_entry+0x154/0x1f8
[ 4.609645] [<ffff00000808e004>] secondary_start_kernel+0x15c/0x1a0
Daniel Lezcano said this notification is needed on arm/arm64 platforms.
Sebastian suggested using atomic_notifier instead of rwlock, which is not
only removing the sleeping in idle, but also improving latency.
Tony Lindgren found a miss use that rcu_read_lock used after rcu_idle_enter
Paul McKenney suggested trying RCU_NONIDLE.
Signed-off-by: Alex Shi <alex.shi@linaro.org>
Tested-by: Tony Lindgren <tony@atomide.com>
Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
[ rjw: Subject & changelog ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Pull scheduler fixes from Thomas Gleixner:
"Two patches addressing build warnings caused by inconsistent kernel
doc comments"
* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
sched/wait: Clean up some documentation warnings
sched/core: Fix some documentation build warnings
bpf_prog_size(prog->len) is not the correct length we want to dump
back to user space. The code in bpf_prog_get_info_by_fd() uses this
to copy prog->insnsi to user space, but bpf_prog_size(prog->len) also
includes the size of struct bpf_prog itself plus program instructions
and is usually used either in context of accounting or for bpf_prog_alloc()
et al, thus we copy out of bounds in bpf_prog_get_info_by_fd()
potentially. Use the correct bpf_prog_insn_size() instead.
Fixes: 1e27097690 ("bpf: Add BPF_OBJ_GET_INFO_BY_FD")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
err in bpf_prog_get_info_by_fd() still holds 0 at that time from prior
check_uarg_tail_zero() check. Explicitly return -EFAULT instead, so
user space can be notified of buggy behavior.
Fixes: 1e27097690 ("bpf: Add BPF_OBJ_GET_INFO_BY_FD")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently cfq/bfq/blk-throttle output cgroup info in trace in their own
way. Now we have standard blktrace API for this, so convert them to use
it.
Note, this changes the behavior a little bit. cgroup info isn't output
by default, we only do this with 'blk_cgroup' option enabled. cgroup
info isn't output as a string by default too, we only do this with
'blk_cgname' option enabled. Also cgroup info is output in different
position of the note string. I think these behavior changes aren't a big
issue (actually we make trace data shorter which is good), since the
blktrace note is solely for debugging.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
By default we output cgroup id in blktrace. This adds an option to
display cgroup path. Since get cgroup path is a relativly heavy
operation, we don't enable it by default.
with the option enabled, blktrace will output something like this:
dd-1353 [007] d..2 293.015252: 8,0 /test/level D R 24 + 8 [dd]
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Currently blktrace isn't cgroup aware. blktrace prints out task name of
current context, but the task of current context isn't always in the
cgroup where the BIO comes from. We can't use task name to find out IO
cgroup. For example, Writeback BIOs always comes from flusher thread but
the BIOs are for different blk cgroups. Request could be requeued and
dispatched from completely different tasks. MD/DM are another examples.
This patch tries to fix the gap. We print out cgroup fhandle info in
blktrace. Userspace can use open_by_handle_at() syscall to find the
cgroup by fhandle. Or userspace can use name_to_handle_at() syscall to
find fhandle for a cgroup and use a BPF program to filter out blktrace
for a specific cgroup.
We add a new 'blk_cgroup' trace option for blk tracer. It's default off.
Application which doesn't know the new option isn't affected. When it's
on, we output fhandle info right after blk_io_trace with an extra bit
set in event action. So from application point of view, blktrace with
the option will output new actions.
I didn't change blk trace event yet, since I'm not sure if changing the
trace event output is an ABI issue. If not, I'll do it later.
Acked-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Now we have the facilities to implement exportfs operations. The idea is
cgroup can export the fhandle info to userspace, then userspace uses
fhandle to find the cgroup name. Another example is userspace can get
fhandle for a cgroup and BPF uses the fhandle to filter info for the
cgroup.
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Per-cpu workqueues have been tripping CPU affinity sanity checks while
a CPU is being offlined. A per-cpu kworker ends up running on a CPU
which isn't its target CPU while the CPU is online but inactive.
While the scheduler allows kthreads to wake up on an online but
inactive CPU, it doesn't allow a running kthread to be migrated to
such a CPU, which leads to an odd situation where setting affinity on
a sleeping and running kthread leads to different results.
Each mem-reclaim workqueue has one rescuer which guarantees forward
progress and the rescuer needs to bind itself to the CPU which needs
help in making forward progress; however, due to the above issue,
while set_cpus_allowed_ptr() succeeds, the rescuer doesn't end up on
the correct CPU if the CPU is in the process of going offline,
tripping the sanity check and executing the work item on the wrong
CPU.
This patch updates __migrate_task() so that kthreads can be migrated
into an inactive but online CPU.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Reported-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
There is an underlying assumption/trade-off in many layers of the Linux
system that CPU <-> node mapping is static. This is despite the presence
of features like NUMA and 'hotplug' that support the dynamic addition/
removal of fundamental system resources like CPUs and memory. PowerPC
systems, however, do provide extensive features for the dynamic change
of resources available to a system.
Currently, there is little or no synchronization protection around the
updating of the CPU <-> node mapping, and the export/update of this
information for other layers / modules. In systems which can change
this mapping during 'hotplug', like PowerPC, the information is changing
underneath all layers that might reference it.
This patch attempts to ensure that a valid, usable cpumask attribute
is used by the workqueue infrastructure when setting up new resource
pools. It prevents a crash that has been observed when an 'empty'
cpumask is passed along to the worker/task scheduling code. It is
intended as a temporary workaround until a more fundamental review and
correction of the issue can be done.
[With additions to the patch provided by Tejun Hao <tj@kernel.org>]
Signed-off-by: Michael Bringmann <mwb@linux.vnet.ibm.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Tree RCU guarantees that every online CPU has a memory barrier between
any given grace period and any of that CPU's RCU read-side sections that
must be ordered against that grace period. Since RCU doesn't always
know where read-side critical sections are, the actual implementation
guarantees order against prior and subsequent non-idle non-offline code,
whether in an RCU read-side critical section or not. As a result, there
does not need to be a memory barrier at the end of synchronize_rcu()
and friends because the ordering internal to the grace period has
ordered every CPU's post-grace-period execution against each CPU's
pre-grace-period execution, again for all non-idle online CPUs.
In contrast, SRCU can have non-idle online CPUs that are completely
uninvolved in a given SRCU grace period, for example, a CPU that
never runs any SRCU read-side critical sections and took no part in
the grace-period processing. It is in theory possible for a given
synchronize_srcu()'s wakeup to be delivered to a CPU that was completely
uninvolved in the prior SRCU grace period, which could mean that the
code following that synchronize_srcu() would end up being unordered with
respect to both the grace period and any pre-existing SRCU read-side
critical sections.
This commit therefore adds an smp_mb() to the end of __synchronize_srcu(),
which prevents this scenario from occurring.
Reported-by: Lance Roy <ldr709@gmail.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Acked-by: Lance Roy <ldr709@gmail.com>
Cc: <stable@vger.kernel.org> # 4.12.x
