This is a significant simplification, it eliminates all the remaining
'hmm' stuff in mm_struct, eliminates krefing along the critical notifier
paths, and takes away all the ugly locking and abuse of page_table_lock.
mmu_notifier_get() provides the single struct hmm per struct mm which
eliminates mm->hmm.
It also directly guarantees that no mmu_notifier op callback is callable
while concurrent free is possible, this eliminates all the krefs inside
the mmu_notifier callbacks.
The remaining krefs in the range code were overly cautious, drivers are
already not permitted to free the mirror while a range exists.
Link: https://lore.kernel.org/r/20190806231548.25242-6-jgg@ziepe.ca
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Ralph Campbell <rcampbell@nvidia.com>
Tested-by: Ralph Campbell <rcampbell@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Systems in lockdown mode should block the kexec of untrusted kernels.
For x86 and ARM we can ensure that a kernel is trustworthy by validating
a PE signature, but this isn't possible on other architectures. On those
platforms we can use IMA digital signatures instead. Add a function to
determine whether IMA has or will verify signatures for a given event type,
and if so permit kexec_file() even if the kernel is otherwise locked down.
This is restricted to cases where CONFIG_INTEGRITY_TRUSTED_KEYRING is set
in order to prevent an attacker from loading additional keys at runtime.
Signed-off-by: Matthew Garrett <mjg59@google.com>
Acked-by: Mimi Zohar <zohar@linux.ibm.com>
Cc: Dmitry Kasatkin <dmitry.kasatkin@gmail.com>
Cc: linux-integrity@vger.kernel.org
Signed-off-by: James Morris <jmorris@namei.org>
When KEXEC_SIG is not enabled, kernel should not load images through
kexec_file systemcall if the kernel is locked down.
[Modified by David Howells to fit with modifications to the previous patch
and to return -EPERM if the kernel is locked down for consistency with
other lockdowns. Modified by Matthew Garrett to remove the IMA
integration, which will be replaced by integrating with the IMA
architecture policy patches.]
Signed-off-by: Jiri Bohac <jbohac@suse.cz>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Matthew Garrett <mjg59@google.com>
cc: kexec@lists.infradead.org
Signed-off-by: James Morris <jmorris@namei.org>
This is a preparatory patch for kexec_file_load() lockdown. A locked down
kernel needs to prevent unsigned kernel images from being loaded with
kexec_file_load(). Currently, the only way to force the signature
verification is compiling with KEXEC_VERIFY_SIG. This prevents loading
usigned images even when the kernel is not locked down at runtime.
This patch splits KEXEC_VERIFY_SIG into KEXEC_SIG and KEXEC_SIG_FORCE.
Analogous to the MODULE_SIG and MODULE_SIG_FORCE for modules, KEXEC_SIG
turns on the signature verification but allows unsigned images to be
loaded. KEXEC_SIG_FORCE disallows images without a valid signature.
Signed-off-by: Jiri Bohac <jbohac@suse.cz>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Matthew Garrett <mjg59@google.com>
cc: kexec@lists.infradead.org
Signed-off-by: James Morris <jmorris@namei.org>
The kexec_load() syscall permits the loading and execution of arbitrary
code in ring 0, which is something that lock-down is meant to prevent. It
makes sense to disable kexec_load() in this situation.
This does not affect kexec_file_load() syscall which can check for a
signature on the image to be booted.
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Matthew Garrett <mjg59@google.com>
Acked-by: Dave Young <dyoung@redhat.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
cc: kexec@lists.infradead.org
Signed-off-by: James Morris <jmorris@namei.org>
If the kernel is locked down, require that all modules have valid
signatures that we can verify.
I have adjusted the errors generated:
(1) If there's no signature (ENODATA) or we can't check it (ENOPKG,
ENOKEY), then:
(a) If signatures are enforced then EKEYREJECTED is returned.
(b) If there's no signature or we can't check it, but the kernel is
locked down then EPERM is returned (this is then consistent with
other lockdown cases).
(2) If the signature is unparseable (EBADMSG, EINVAL), the signature fails
the check (EKEYREJECTED) or a system error occurs (eg. ENOMEM), we
return the error we got.
Note that the X.509 code doesn't check for key expiry as the RTC might not
be valid or might not have been transferred to the kernel's clock yet.
[Modified by Matthew Garrett to remove the IMA integration. This will
be replaced with integration with the IMA architecture policy
patchset.]
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Matthew Garrett <matthewgarrett@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Cc: Jessica Yu <jeyu@kernel.org>
Signed-off-by: James Morris <jmorris@namei.org>
Pull kernel thread signal handling fix from Eric Biederman:
"I overlooked the fact that kernel threads are created with all signals
set to SIG_IGN, and accidentally caused a regression in cifs and drbd
when replacing force_sig with send_sig.
This is my fix for that regression. I add a new function
allow_kernel_signal which allows kernel threads to receive signals
sent from the kernel, but continues to ignore all signals sent from
userspace. This ensures the user space interface for cifs and drbd
remain the same.
These kernel threads depend on blocking networking calls which block
until something is received or a signal is pending. Making receiving
of signals somewhat necessary for these kernel threads.
Perhaps someday we can cleanup those interfaces and remove
allow_kernel_signal. If not allow_kernel_signal is pretty trivial and
clearly documents what is going on so I don't think we will mind
carrying it"
* 'siginfo-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace:
signal: Allow cifs and drbd to receive their terminating signals
If alloc_descs() fails before irq_sysfs_init() has run, free_desc() in the
cleanup path will call kobject_del() even though the kobject has not been
added with kobject_add().
Fix this by making the call to kobject_del() conditional on whether
irq_sysfs_init() has run.
This problem surfaced because commit aa30f47cf6 ("kobject: Add support
for default attribute groups to kobj_type") makes kobject_del() stricter
about pairing with kobject_add(). If the pairing is incorrrect, a WARNING
and backtrace occur in sysfs_remove_group() because there is no parent.
[ tglx: Add a comment to the code and make it work with CONFIG_SYSFS=n ]
Fixes: ecb3f394c5 ("genirq: Expose interrupt information through sysfs")
Signed-off-by: Michael Kelley <mikelley@microsoft.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/1564703564-4116-1-git-send-email-mikelley@microsoft.com
My recent to change to only use force_sig for a synchronous events
wound up breaking signal reception cifs and drbd. I had overlooked
the fact that by default kthreads start out with all signals set to
SIG_IGN. So a change I thought was safe turned out to have made it
impossible for those kernel thread to catch their signals.
Reverting the work on force_sig is a bad idea because what the code
was doing was very much a misuse of force_sig. As the way force_sig
ultimately allowed the signal to happen was to change the signal
handler to SIG_DFL. Which after the first signal will allow userspace
to send signals to these kernel threads. At least for
wake_ack_receiver in drbd that does not appear actively wrong.
So correct this problem by adding allow_kernel_signal that will allow
signals whose siginfo reports they were sent by the kernel through,
but will not allow userspace generated signals, and update cifs and
drbd to call allow_kernel_signal in an appropriate place so that their
thread can receive this signal.
Fixing things this way ensures that userspace won't be able to send
signals and cause problems, that it is clear which signals the
threads are expecting to receive, and it guarantees that nothing
else in the system will be affected.
This change was partly inspired by similar cifs and drbd patches that
added allow_signal.
Reported-by: ronnie sahlberg <ronniesahlberg@gmail.com>
Reported-by: Christoph Böhmwalder <christoph.boehmwalder@linbit.com>
Tested-by: Christoph Böhmwalder <christoph.boehmwalder@linbit.com>
Cc: Steve French <smfrench@gmail.com>
Cc: Philipp Reisner <philipp.reisner@linbit.com>
Cc: David Laight <David.Laight@ACULAB.COM>
Fixes: 247bc9470b ("cifs: fix rmmod regression in cifs.ko caused by force_sig changes")
Fixes: 72abe3bcf0 ("signal/cifs: Fix cifs_put_tcp_session to call send_sig instead of force_sig")
Fixes: fee109901f ("signal/drbd: Use send_sig not force_sig")
Fixes: 3cf5d076fb ("signal: Remove task parameter from force_sig")
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
klp_module_coming() is called for every module appearing in the system.
It sets obj->mod to a patched module for klp_object obj. Unfortunately
it leaves it set even if an error happens later in the function and the
patched module is not allowed to be loaded.
klp_is_object_loaded() uses obj->mod variable and could currently give a
wrong return value. The bug is probably harmless as of now.
Signed-off-by: Miroslav Benes <mbenes@suse.cz>
Reviewed-by: Petr Mladek <pmladek@suse.com>
Acked-by: Josh Poimboeuf <jpoimboe@redhat.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
lockdep reports the following deadlock scenario:
WARNING: possible circular locking dependency detected
kworker/1:1/48 is trying to acquire lock:
000000008d7a62b2 (text_mutex){+.+.}, at: kprobe_optimizer+0x163/0x290
but task is already holding lock:
00000000850b5e2d (module_mutex){+.+.}, at: kprobe_optimizer+0x31/0x290
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #1 (module_mutex){+.+.}:
__mutex_lock+0xac/0x9f0
mutex_lock_nested+0x1b/0x20
set_all_modules_text_rw+0x22/0x90
ftrace_arch_code_modify_prepare+0x1c/0x20
ftrace_run_update_code+0xe/0x30
ftrace_startup_enable+0x2e/0x50
ftrace_startup+0xa7/0x100
register_ftrace_function+0x27/0x70
arm_kprobe+0xb3/0x130
enable_kprobe+0x83/0xa0
enable_trace_kprobe.part.0+0x2e/0x80
kprobe_register+0x6f/0xc0
perf_trace_event_init+0x16b/0x270
perf_kprobe_init+0xa7/0xe0
perf_kprobe_event_init+0x3e/0x70
perf_try_init_event+0x4a/0x140
perf_event_alloc+0x93a/0xde0
__do_sys_perf_event_open+0x19f/0xf30
__x64_sys_perf_event_open+0x20/0x30
do_syscall_64+0x65/0x1d0
entry_SYSCALL_64_after_hwframe+0x49/0xbe
-> #0 (text_mutex){+.+.}:
__lock_acquire+0xfcb/0x1b60
lock_acquire+0xca/0x1d0
__mutex_lock+0xac/0x9f0
mutex_lock_nested+0x1b/0x20
kprobe_optimizer+0x163/0x290
process_one_work+0x22b/0x560
worker_thread+0x50/0x3c0
kthread+0x112/0x150
ret_from_fork+0x3a/0x50
other info that might help us debug this:
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(module_mutex);
lock(text_mutex);
lock(module_mutex);
lock(text_mutex);
*** DEADLOCK ***
As a reproducer I've been using bcc's funccount.py
(https://github.com/iovisor/bcc/blob/master/tools/funccount.py),
for example:
# ./funccount.py '*interrupt*'
That immediately triggers the lockdep splat.
Fix by acquiring text_mutex before module_mutex in kprobe_optimizer().
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Acked-by: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Naveen N. Rao <naveen.n.rao@linux.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: d5b844a2cf ("ftrace/x86: Remove possible deadlock between register_kprobe() and ftrace_run_update_code()")
Link: http://lkml.kernel.org/r/20190812184302.GA7010@xps-13
Signed-off-by: Ingo Molnar <mingo@kernel.org>
If a task is PI-blocked (blocking on sleeping spinlock) then we don't want to
schedule a new kworker if we schedule out due to lock contention because !RT
does not do that as well. A spinning spinlock disables preemption and a worker
does not schedule out on lock contention (but spin).
On RT the RW-semaphore implementation uses an rtmutex so
tsk_is_pi_blocked() will return true if a task blocks on it. In this case we
will now start a new worker which may deadlock if one worker is waiting on
progress from another worker. Since a RW-semaphore starts a new worker on !RT,
we should do the same on RT.
XFS is able to trigger this deadlock.
Allow to schedule new worker if the current worker is PI-blocked.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20190816160626.12742-1-bigeasy@linutronix.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Pull SPDX fixes from Greg KH:
"Here are four small SPDX fixes for 5.3-rc5.
A few style fixes for some SPDX comments, added an SPDX tag for one
file, and fix up some GPL boilerplate for another file.
All of these have been in linux-next for a few weeks with no reported
issues (they are comment changes only, so that's to be expected...)"
* tag 'spdx-5.3-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/spdx:
i2c: stm32: Use the correct style for SPDX License Identifier
intel_th: Use the correct style for SPDX License Identifier
coccinelle: api/atomic_as_refcounter: add SPDX License Identifier
kernel/configs: Replace GPL boilerplate code with SPDX identifier
The XSKMAP did not honor the BPF_EXIST/BPF_NOEXIST flags when updating
an entry. This patch addresses that.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
When an AF_XDP socket is released/closed the XSKMAP still holds a
reference to the socket in a "released" state. The socket will still
use the netdev queue resource, and block newly created sockets from
attaching to that queue, but no user application can access the
fill/complete/rx/tx queues. This results in that all applications need
to explicitly clear the map entry from the old "zombie state"
socket. This should be done automatically.
In this patch, the sockets tracks, and have a reference to, which maps
it resides in. When the socket is released, it will remove itself from
all maps.
Suggested-by: Bruce Richardson <bruce.richardson@intel.com>
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Rename existing bpf_map_inc_not_zero to __bpf_map_inc_not_zero to
indicate that it's caller's responsibility to do proper locking.
Create and export bpf_map_inc_not_zero wrapper that properly
locks map_idr_lock. Will be used in the next commit to
hold a map while cloning a socket.
Cc: Martin KaFai Lau <kafai@fb.com>
Cc: Yonghong Song <yhs@fb.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Stanislav Fomichev <sdf@google.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Pull power management fixes from Rafael Wysocki:
"These add a check to avoid recent suspend-to-idle power regression on
systems with NVMe drives where the PCIe ASPM policy is "performance"
(or when the kernel is built without ASPM support), fix an issue
related to frequency limits in the schedutil cpufreq governor and fix
a mistake related to the PM QoS usage in the cpufreq core introduced
recently.
Specifics:
- Disable NVMe power optimization related to suspend-to-idle added
recently on systems where PCIe ASPM is not able to put PCIe links
into low-power states to prevent excess power from being drawn by
the system while suspended (Rafael Wysocki).
- Make the schedutil governor handle frequency limits changes
properly in all cases (Viresh Kumar).
- Prevent the cpufreq core from treating positive values returned by
dev_pm_qos_update_request() as errors (Viresh Kumar)"
* tag 'pm-5.3-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
nvme-pci: Allow PCI bus-level PM to be used if ASPM is disabled
PCI/ASPM: Add pcie_aspm_enabled()
cpufreq: schedutil: Don't skip freq update when limits change
cpufreq: dev_pm_qos_update_request() can return 1 on success
Use str_has_prefix() instead of strncmp() which is less
straightforward in decode_state().
Signed-off-by: Chuhong Yuan <hslester96@gmail.com>
[ rjw: Subject & changelog ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
In case of error, the function kobject_create_and_add() returns NULL
pointer not ERR_PTR(). The IS_ERR() test in the return value check
should be replaced with NULL test.
Fixes: 341dfcf8d7 ("btf: expose BTF info through sysfs")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Pull dma-mapping fixes from Christoph Hellwig:
- fix the handling of the bus_dma_mask in dma_get_required_mask, which
caused a regression in this merge window (Lucas Stach)
- fix a regression in the handling of DMA_ATTR_NO_KERNEL_MAPPING (me)
- fix dma_mmap_coherent to not cause page attribute mismatches on
coherent architectures like x86 (me)
* tag 'dma-mapping-5.3-4' of git://git.infradead.org/users/hch/dma-mapping:
dma-mapping: fix page attributes for dma_mmap_*
dma-direct: don't truncate dma_required_mask to bus addressing capabilities
dma-direct: fix DMA_ATTR_NO_KERNEL_MAPPING
Daniel Borkmann says:
====================
The following pull-request contains BPF updates for your *net-next* tree.
There is a small merge conflict in libbpf (Cc Andrii so he's in the loop
as well):
for (i = 1; i <= btf__get_nr_types(btf); i++) {
t = (struct btf_type *)btf__type_by_id(btf, i);
if (!has_datasec && btf_is_var(t)) {
/* replace VAR with INT */
t->info = BTF_INFO_ENC(BTF_KIND_INT, 0, 0);
<<<<<<< HEAD
/*
* using size = 1 is the safest choice, 4 will be too
* big and cause kernel BTF validation failure if
* original variable took less than 4 bytes
*/
t->size = 1;
*(int *)(t+1) = BTF_INT_ENC(0, 0, 8);
} else if (!has_datasec && kind == BTF_KIND_DATASEC) {
=======
t->size = sizeof(int);
*(int *)(t + 1) = BTF_INT_ENC(0, 0, 32);
} else if (!has_datasec && btf_is_datasec(t)) {
>>>>>>> 72ef80b5ee
/* replace DATASEC with STRUCT */
Conflict is between the two commits 1d4126c4e1 ("libbpf: sanitize VAR to
conservative 1-byte INT") and b03bc6853c ("libbpf: convert libbpf code to
use new btf helpers"), so we need to pick the sanitation fixup as well as
use the new btf_is_datasec() helper and the whitespace cleanup. Looks like
the following:
[...]
if (!has_datasec && btf_is_var(t)) {
/* replace VAR with INT */
t->info = BTF_INFO_ENC(BTF_KIND_INT, 0, 0);
/*
* using size = 1 is the safest choice, 4 will be too
* big and cause kernel BTF validation failure if
* original variable took less than 4 bytes
*/
t->size = 1;
*(int *)(t + 1) = BTF_INT_ENC(0, 0, 8);
} else if (!has_datasec && btf_is_datasec(t)) {
/* replace DATASEC with STRUCT */
[...]
The main changes are:
1) Addition of core parts of compile once - run everywhere (co-re) effort,
that is, relocation of fields offsets in libbpf as well as exposure of
kernel's own BTF via sysfs and loading through libbpf, from Andrii.
More info on co-re: http://vger.kernel.org/bpfconf2019.html#session-2
and http://vger.kernel.org/lpc-bpf2018.html#session-2
2) Enable passing input flags to the BPF flow dissector to customize parsing
and allowing it to stop early similar to the C based one, from Stanislav.
3) Add a BPF helper function that allows generating SYN cookies from XDP and
tc BPF, from Petar.
4) Add devmap hash-based map type for more flexibility in device lookup for
redirects, from Toke.
5) Improvements to XDP forwarding sample code now utilizing recently enabled
devmap lookups, from Jesper.
6) Add support for reporting the effective cgroup progs in bpftool, from Jakub
and Takshak.
7) Fix reading kernel config from bpftool via /proc/config.gz, from Peter.
8) Fix AF_XDP umem pages mapping for 32 bit architectures, from Ivan.
9) Follow-up to add two more BPF loop tests for the selftest suite, from Alexei.
10) Add perf event output helper also for other skb-based program types, from Allan.
11) Fix a co-re related compilation error in selftests, from Yonghong.
====================
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Bimodal behavior of rcu_do_batch() is not really suited to Google
applications like gfe servers.
When a process with millions of sockets exits, closing all files
queues two rcu callbacks per socket.
This eventually reaches the point where RCU enters an emergency
mode, where rcu_do_batch() do not return until whole queue is flushed.
Each rcu callback lasts at least 70 nsec, so with millions of
elements, we easily spend more than 100 msec without rescheduling.
Goal of this patch is to avoid the infamous message like following
"need_resched set for > 51999388 ns (52 ticks) without schedule"
We dynamically adjust the number of elements we process, instead
of 10 / INFINITE choices, we use a floor of ~1 % of current entries.
If the number is above 1000, we switch to a time based limit of 3 msec
per batch, adjustable with /sys/module/rcutree/parameters/rcu_resched_ns
Signed-off-by: Eric Dumazet <edumazet@google.com>
[ paulmck: Forward-port and remove debug statements. ]
Signed-off-by: Paul E. McKenney <paulmck@linux.ibm.com>
When under overload conditions, __call_rcu_nocb_wake() will wake the
no-CBs GP kthread any time the no-CBs CB kthread is asleep or there
are no ready-to-invoke callbacks, but only after a timer delay. If the
no-CBs GP kthread has a ->nocb_bypass_timer pending, the deferred wakeup
from __call_rcu_nocb_wake() is redundant. This commit therefore makes
__call_rcu_nocb_wake() avoid posting the redundant deferred wakeup if
->nocb_bypass_timer is pending. This requires adding a bit of ordering
of timer actions.
Signed-off-by: Paul E. McKenney <paulmck@linux.ibm.com>
Currently, __call_rcu_nocb_wake() advances callbacks each time that it
detects excessive numbers of callbacks, though only if it succeeds in
conditionally acquiring its leaf rcu_node structure's ->lock. Despite
the conditional acquisition of ->lock, this does increase contention.
This commit therefore avoids advancing callbacks unless there are
callbacks in ->cblist whose grace period has completed and advancing
has not yet been done during this jiffy.
Note that this decision does not take the presence of new callbacks
into account. That is because on this code path, there will always be
at least one new callback, namely the one we just enqueued.
Signed-off-by: Paul E. McKenney <paulmck@linux.ibm.com>
Currently, nocb_cb_wait() advances callbacks on each pass through its
loop, though only if it succeeds in conditionally acquiring its leaf
rcu_node structure's ->lock. Despite the conditional acquisition of
->lock, this does increase contention. This commit therefore avoids
advancing callbacks unless there are callbacks in ->cblist whose grace
period has completed.
Note that nocb_cb_wait() doesn't worry about callbacks that have not
yet been assigned a grace period. The idea is that the only reason for
nocb_cb_wait() to advance callbacks is to allow it to continue invoking
callbacks. Time will tell whether this is the correct choice.
Signed-off-by: Paul E. McKenney <paulmck@linux.ibm.com>
The rcutree_migrate_callbacks() invokes rcu_advance_cbs() on both the
offlined CPU's ->cblist and that of the surviving CPU, then merges
them. However, after the merge, and of the offlined CPU's callbacks
that were not ready to be invoked will no longer be associated with a
grace-period number. This commit therefore invokes rcu_advance_cbs()
one more time on the merged ->cblist in order to assign a grace-period
number to these callbacks.
Signed-off-by: Paul E. McKenney <paulmck@linux.ibm.com>
When callbacks are in full flow, the common case is waiting for a
grace period, and this grace period will normally take a few jiffies to
complete. It therefore isn't all that helpful for __call_rcu_nocb_wake()
to do a synchronous wakeup in this case. This commit therefore turns this
into a timer-based deferred wakeup of the no-CBs grace-period kthread.
Signed-off-by: Paul E. McKenney <paulmck@linux.ibm.com>
This commit causes locking, sleeping, and callback state to be printed
for no-CBs CPUs when the rcutorture writer is delayed sufficiently for
rcutorture to complain.
Signed-off-by: Paul E. McKenney <paulmck@linux.ibm.com>
Use of the rcu_data structure's segmented ->cblist for no-CBs CPUs
takes advantage of unrelated grace periods, thus reducing the memory
footprint in the face of floods of call_rcu() invocations. However,
the ->cblist field is a more-complex rcu_segcblist structure which must
be protected via locking. Even though there are only three entities
which can acquire this lock (the CPU invoking call_rcu(), the no-CBs
grace-period kthread, and the no-CBs callbacks kthread), the contention
on this lock is excessive under heavy stress.
This commit therefore greatly reduces contention by provisioning
an rcu_cblist structure field named ->nocb_bypass within the
rcu_data structure. Each no-CBs CPU is permitted only a limited
number of enqueues onto the ->cblist per jiffy, controlled by a new
nocb_nobypass_lim_per_jiffy kernel boot parameter that defaults to
about 16 enqueues per millisecond (16 * 1000 / HZ). When that limit is
exceeded, the CPU instead enqueues onto the new ->nocb_bypass.
The ->nocb_bypass is flushed into the ->cblist every jiffy or when
the number of callbacks on ->nocb_bypass exceeds qhimark, whichever
happens first. During call_rcu() floods, this flushing is carried out
by the CPU during the course of its call_rcu() invocations. However,
a CPU could simply stop invoking call_rcu() at any time. The no-CBs
grace-period kthread therefore carries out less-aggressive flushing
(every few jiffies or when the number of callbacks on ->nocb_bypass
exceeds (2 * qhimark), whichever comes first). This means that the
no-CBs grace-period kthread cannot be permitted to do unbounded waits
while there are callbacks on ->nocb_bypass. A ->nocb_bypass_timer is
used to provide the needed wakeups.
[ paulmck: Apply Coverity feedback reported by Colin Ian King. ]
Signed-off-by: Paul E. McKenney <paulmck@linux.ibm.com>
Upcoming ->nocb_lock contention-reduction work requires that the
rcu_segcblist structure's ->len field be concurrently manipulated,
but only if there are no-CBs CPUs in the kernel. This commit
therefore makes this ->len field be an atomic_long_t, but only
in CONFIG_RCU_NOCB_CPU=y kernels.
Signed-off-by: Paul E. McKenney <paulmck@linux.ibm.com>
When there are excessive numbers of callbacks, and when either the
corresponding no-CBs callback kthread is asleep or there is no more
ready-to-invoke callbacks, and when least one callback is pending,
__call_rcu_nocb_wake() will advance the callbacks, but refrain from
awakening the corresponding no-CBs grace-period kthread. However,
because rcu_advance_cbs_nowake() is used, it is possible (if a bit
unlikely) that the needed advancement could not happen due to a grace
period not being in progress. Plus there will always be at least one
pending callback due to one having just now been enqueued.
This commit therefore attempts to advance callbacks and awakens the
no-CBs grace-period kthread when there are excessive numbers of callbacks
posted and when the no-CBs callback kthread is not in a position to do
anything helpful.
Signed-off-by: Paul E. McKenney <paulmck@linux.ibm.com>
The sleep/wakeup of the no-CBs grace-period kthreads is synchronized
using the ->nocb_lock of the first CPU corresponding to that kthread.
This commit provides a separate ->nocb_gp_lock for this purpose, thus
reducing contention on ->nocb_lock.
Signed-off-by: Paul E. McKenney <paulmck@linux.ibm.com>
Currently, nocb_cb_wait() unconditionally acquires the leaf rcu_node
->lock to advance callbacks when done invoking the previous batch.
It does this while holding ->nocb_lock, which means that contention on
the leaf rcu_node ->lock visits itself on the ->nocb_lock. This commit
therefore makes this lock acquisition conditional, forgoing callback
advancement when the leaf rcu_node ->lock is not immediately available.
(In this case, the no-CBs grace-period kthread will eventually do any
needed callback advancement.)
Signed-off-by: Paul E. McKenney <paulmck@linux.ibm.com>
Currently, __call_rcu_nocb_wake() conditionally acquires the leaf rcu_node
structure's ->lock, and only afterwards does rcu_advance_cbs_nowake()
check to see if it is possible to advance callbacks without potentially
needing to awaken the grace-period kthread. Given that the no-awaken
check can be done locklessly, this commit reverses the order, so that
rcu_advance_cbs_nowake() is invoked without holding the leaf rcu_node
structure's ->lock and rcu_advance_cbs_nowake() checks the grace-period
state before conditionally acquiring that lock, thus reducing the number
of needless acquistions of the leaf rcu_node structure's ->lock.
Signed-off-by: Paul E. McKenney <paulmck@linux.ibm.com>
Currently, when the square root of the number of CPUs is rounded down
by int_sqrt(), this round-down is applied to the number of callback
kthreads per grace-period kthreads. This makes almost no difference
for large systems, but results in oddities such as three no-CBs
grace-period kthreads for a five-CPU system, which is a bit excessive.
This commit therefore causes the round-down to apply to the number of
no-CBs grace-period kthreads, so that systems with from four to eight
CPUs have only two no-CBs grace period kthreads.
Signed-off-by: Paul E. McKenney <paulmck@linux.ibm.com>
A given rcu_data structure's ->nocb_lock can be acquired very frequently
by the corresponding CPU and occasionally by the corresponding no-CBs
grace-period and callbacks kthreads. In particular, these two kthreads
will have frequent gaps between ->nocb_lock acquisitions that are roughly
a grace period in duration. This means that any excessive ->nocb_lock
contention will be due to the CPU's acquisitions, and this in turn
enables a very naive contention-avoidance strategy to be quite effective.
This commit therefore modifies rcu_nocb_lock() to first
attempt a raw_spin_trylock(), and to atomically increment a
separate ->nocb_lock_contended across a raw_spin_lock(). This new
->nocb_lock_contended field is checked in __call_rcu_nocb_wake() when
interrupts are enabled, with a spin-wait for contending acquisitions
to complete, thus allowing the kthreads a chance to acquire the lock.
Signed-off-by: Paul E. McKenney <paulmck@linux.ibm.com>
Currently, the code provides an extra wakeup for the no-CBs grace-period
kthread if one of its CPUs is generating excessive numbers of callbacks.
But satisfying though it is to wake something up when things are going
south, unless the thing being awakened can actually help solve the
problem, that extra wakeup does nothing but consume additional CPU time,
which is exactly what you don't want during a call_rcu() flood.
This commit therefore avoids doing anything if the corresponding
no-CBs callback kthread is going full tilt. Otherwise, if advancing
callbacks immediately might help and if the leaf rcu_node structure's
lock is immediately available, this commit invokes a new variant of
rcu_advance_cbs() that advances callbacks only if doing so won't require
awakening the grace-period kthread (not to be confused with any of the
no-CBs grace-period kthreads).
Signed-off-by: Paul E. McKenney <paulmck@linux.ibm.com>
It might be hard to imagine having more than two billion callbacks
queued on a single CPU's ->cblist, but someone will do it sometime.
This commit therefore makes __call_rcu_nocb_wake() handle this situation
by upgrading local variable "len" from "int" to "long".
Signed-off-by: Paul E. McKenney <paulmck@linux.ibm.com>
