The current map creation API does not allow to provide the numa-node
preference. The memory usually comes from where the map-creation-process
is running. The performance is not ideal if the bpf_prog is known to
always run in a numa node different from the map-creation-process.
One of the use case is sharding on CPU to different LRU maps (i.e.
an array of LRU maps). Here is the test result of map_perf_test on
the INNER_LRU_HASH_PREALLOC test if we force the lru map used by
CPU0 to be allocated from a remote numa node:
[ The machine has 20 cores. CPU0-9 at node 0. CPU10-19 at node 1 ]
># taskset -c 10 ./map_perf_test 512 8 1260000 8000000
5:inner_lru_hash_map_perf pre-alloc 1628380 events per sec
4:inner_lru_hash_map_perf pre-alloc 1626396 events per sec
3:inner_lru_hash_map_perf pre-alloc 1626144 events per sec
6:inner_lru_hash_map_perf pre-alloc 1621657 events per sec
2:inner_lru_hash_map_perf pre-alloc 1621534 events per sec
1:inner_lru_hash_map_perf pre-alloc 1620292 events per sec
7:inner_lru_hash_map_perf pre-alloc 1613305 events per sec
0:inner_lru_hash_map_perf pre-alloc 1239150 events per sec #<<<
After specifying numa node:
># taskset -c 10 ./map_perf_test 512 8 1260000 8000000
5:inner_lru_hash_map_perf pre-alloc 1629627 events per sec
3:inner_lru_hash_map_perf pre-alloc 1628057 events per sec
1:inner_lru_hash_map_perf pre-alloc 1623054 events per sec
6:inner_lru_hash_map_perf pre-alloc 1616033 events per sec
2:inner_lru_hash_map_perf pre-alloc 1614630 events per sec
4:inner_lru_hash_map_perf pre-alloc 1612651 events per sec
7:inner_lru_hash_map_perf pre-alloc 1609337 events per sec
0:inner_lru_hash_map_perf pre-alloc 1619340 events per sec #<<<
This patch adds one field, numa_node, to the bpf_attr. Since numa node 0
is a valid node, a new flag BPF_F_NUMA_NODE is also added. The numa_node
field is honored if and only if the BPF_F_NUMA_NODE flag is set.
Numa node selection is not supported for percpu map.
This patch does not change all the kmalloc. F.e.
'htab = kzalloc()' is not changed since the object
is small enough to stay in the cache.
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
to be used in combination with tcp option set support to mimic
iptables TCPMSS --clamp-mss-to-pmtu.
v2: Eric Dumazet points out dst must be initialized.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
This allows setting 2 and 4 byte quantities in the tcp option space.
Main purpose is to allow native replacement for xt_TCPMSS to
work around pmtu blackholes.
Writes to kind and len are now allowed at the moment, it does not seem
useful to do this as it causes corruption of the tcp option space.
We can always lift this restriction later if a use-case appears.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
The show and store functions don't need/use the attribute.
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
These functions are wrapper arount class_create_file which can take a
const attribute.
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There are currently macros to set and test an ETHTOOL_LINK_MODE_ setting,
but not to clear one. Add a macro to clear an ETHTOOL_LINK_MODE_ setting.
Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The return error code need to be included in the tracepoint
xdp:xdp_redirect, else its not possible to distinguish successful or
failed XDP_REDIRECT transmits.
XDP have no queuing mechanism. Thus, it is fairly easily to overrun a
NIC transmit queue. The eBPF program invoking helpers (bpf_redirect
or bpf_redirect_map) to redirect a packet doesn't get any feedback
whether the packet was actually transmitted.
Info on failed transmits in the tracepoint xdp:xdp_redirect, is
interesting as this opens for providing a feedback-loop to the
receiving XDP program.
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is useful for directly looking up a task based on class id rather than
having to scan through all open file descriptors.
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Wenwei Tao has noticed that our current assumption that the oom victim
is dying and never doing any visible changes after it dies, and so the
oom_reaper can tear it down, is not entirely true.
__task_will_free_mem consider a task dying when SIGNAL_GROUP_EXIT is set
but do_group_exit sends SIGKILL to all threads _after_ the flag is set.
So there is a race window when some threads won't have
fatal_signal_pending while the oom_reaper could start unmapping the
address space. Moreover some paths might not check for fatal signals
before each PF/g-u-p/copy_from_user.
We already have a protection for oom_reaper vs. PF races by checking
MMF_UNSTABLE. This has been, however, checked only for kernel threads
(use_mm users) which can outlive the oom victim. A simple fix would be
to extend the current check in handle_mm_fault for all tasks but that
wouldn't be sufficient because the current check assumes that a kernel
thread would bail out after EFAULT from get_user*/copy_from_user and
never re-read the same address which would succeed because the PF path
has established page tables already. This seems to be the case for the
only existing use_mm user currently (virtio driver) but it is rather
fragile in general.
This is even more fragile in general for more complex paths such as
generic_perform_write which can re-read the same address more times
(e.g. iov_iter_copy_from_user_atomic to fail and then
iov_iter_fault_in_readable on retry).
Therefore we have to implement MMF_UNSTABLE protection in a robust way
and never make a potentially corrupted content visible. That requires
to hook deeper into the PF path and check for the flag _every time_
before a pte for anonymous memory is established (that means all
!VM_SHARED mappings).
The corruption can be triggered artificially
(http://lkml.kernel.org/r/201708040646.v746kkhC024636@www262.sakura.ne.jp)
but there doesn't seem to be any real life bug report. The race window
should be quite tight to trigger most of the time.
Link: http://lkml.kernel.org/r/20170807113839.16695-3-mhocko@kernel.org
Fixes: aac4536355 ("mm, oom: introduce oom reaper")
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reported-by: Wenwei Tao <wenwei.tww@alibaba-inc.com>
Tested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Cc: Andrea Argangeli <andrea@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
There is existing use after free bug when deferred struct pages are
enabled:
The memblock_add() allocates memory for the memory array if more than
128 entries are needed. See comment in e820__memblock_setup():
* The bootstrap memblock region count maximum is 128 entries
* (INIT_MEMBLOCK_REGIONS), but EFI might pass us more E820 entries
* than that - so allow memblock resizing.
This memblock memory is freed here:
free_low_memory_core_early()
We access the freed memblock.memory later in boot when deferred pages
are initialized in this path:
deferred_init_memmap()
for_each_mem_pfn_range()
__next_mem_pfn_range()
type = &memblock.memory;
One possible explanation for why this use-after-free hasn't been hit
before is that the limit of INIT_MEMBLOCK_REGIONS has never been
exceeded at least on systems where deferred struct pages were enabled.
Tested by reducing INIT_MEMBLOCK_REGIONS down to 4 from the current 128,
and verifying in qemu that this code is getting excuted and that the
freed pages are sane.
Link: http://lkml.kernel.org/r/1502485554-318703-2-git-send-email-pasha.tatashin@oracle.com
Fixes: 7e18adb4f8 ("mm: meminit: initialise remaining struct pages in parallel with kswapd")
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.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>
These are the few pending fixes I have queued up for v4.13-final. One
is a a generic regression fix for recursive loops on kmod and the other
one is a trivial print out correction.
During the v4.13 development we assumed that recursive kmod loops were
no longer possible. Clearly that is not true. The regression fix makes
use of a new killable wait. We use a killable wait to be paranoid in
how signals might be sent to modprobe and only accept a proper SIGKILL.
The signal will only be available to userspace to issue *iff* a thread
has already entered a wait state, and that happens only if we've already
throttled after 50 kmod threads have been hit.
Note that although it may seem excessive to trigger a failure afer 5
seconds if all kmod thread remain busy, prior to the series of changes
that went into v4.13 we would actually *always* fatally fail any request
which came in if the limit was already reached. The new waiting
implemented in v4.13 actually gives us *more* breathing room -- the wait
for 5 seconds is a wait for *any* kmod thread to finish. We give up and
fail *iff* no kmod thread has finished and they're *all* running
straight for 5 consecutive seconds. If 50 kmod threads are running
consecutively for 5 seconds something else must be really bad.
Recursive loops with kmod are bad but they're also hard to implement
properly as a selftest without currently fooling current userspace tools
like kmod [1]. For instance kmod will complain when you run depmod if
it finds a recursive loop with symbol dependency between modules as such
this type of recursive loop cannot go upstream as the modules_install
target will fail after running depmod.
These tests already exist on userspace kmod upstream though (refer to
the testsuite/module-playground/mod-loop-*.c files). The same is not
true if request_module() is used though, or worst if aliases are used.
Likewise the issue with 64-bit kernels booting 32-bit userspace without
a binfmt handler built-in is also currently not detected and proactively
avoided by userspace kmod tools, or kconfig for all architectures.
Although we could complain in the kernel when some of these individual
recursive issues creep up, proactively avoiding these situations in
userspace at build time is what we should keep striving for.
Lastly, since recursive loops could happen with kmod it may mean
recursive loops may also be possible with other kernel usermode helpers,
this should be investigated and long term if we can come up with a more
sensible generic solution even better!
[0] https://git.kernel.org/pub/scm/linux/kernel/git/mcgrof/linux.git/log/?h=20170809-kmod-for-v4.13-final
[1] https://git.kernel.org/pub/scm/utils/kernel/kmod/kmod.git
This patch (of 3):
This wait is similar to wait_event_interruptible_timeout() but only
accepts SIGKILL interrupt signal. Other signals are ignored.
Link: http://lkml.kernel.org/r/20170809234635.13443-2-mcgrof@kernel.org
Signed-off-by: Luis R. Rodriguez <mcgrof@kernel.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Cc: Jessica Yu <jeyu@redhat.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Michal Marek <mmarek@suse.com>
Cc: Petr Mladek <pmladek@suse.com>
Cc: Miroslav Benes <mbenes@suse.cz>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Matt Redfearn <matt.redfearn@imgtec.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Colin Ian King <colin.king@canonical.com>
Cc: Daniel Mentz <danielmentz@google.com>
Cc: David Binderman <dcb314@hotmail.com>
Cc: Matt Redfearn <matt.redfearn@imgetc.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Jaegeuk and Brad report a NULL pointer crash when writeback ending tries
to update the memcg stats:
BUG: unable to handle kernel NULL pointer dereference at 00000000000003b0
IP: test_clear_page_writeback+0x12e/0x2c0
[...]
RIP: 0010:test_clear_page_writeback+0x12e/0x2c0
Call Trace:
<IRQ>
end_page_writeback+0x47/0x70
f2fs_write_end_io+0x76/0x180 [f2fs]
bio_endio+0x9f/0x120
blk_update_request+0xa8/0x2f0
scsi_end_request+0x39/0x1d0
scsi_io_completion+0x211/0x690
scsi_finish_command+0xd9/0x120
scsi_softirq_done+0x127/0x150
__blk_mq_complete_request_remote+0x13/0x20
flush_smp_call_function_queue+0x56/0x110
generic_smp_call_function_single_interrupt+0x13/0x30
smp_call_function_single_interrupt+0x27/0x40
call_function_single_interrupt+0x89/0x90
RIP: 0010:native_safe_halt+0x6/0x10
(gdb) l *(test_clear_page_writeback+0x12e)
0xffffffff811bae3e is in test_clear_page_writeback (./include/linux/memcontrol.h:619).
614 mod_node_page_state(page_pgdat(page), idx, val);
615 if (mem_cgroup_disabled() || !page->mem_cgroup)
616 return;
617 mod_memcg_state(page->mem_cgroup, idx, val);
618 pn = page->mem_cgroup->nodeinfo[page_to_nid(page)];
619 this_cpu_add(pn->lruvec_stat->count[idx], val);
620 }
621
622 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
623 gfp_t gfp_mask,
The issue is that writeback doesn't hold a page reference and the page
might get freed after PG_writeback is cleared (and the mapping is
unlocked) in test_clear_page_writeback(). The stat functions looking up
the page's node or zone are safe, as those attributes are static across
allocation and free cycles. But page->mem_cgroup is not, and it will
get cleared if we race with truncation or migration.
It appears this race window has been around for a while, but less likely
to trigger when the memcg stats were updated first thing after
PG_writeback is cleared. Recent changes reshuffled this code to update
the global node stats before the memcg ones, though, stretching the race
window out to an extent where people can reproduce the problem.
Update test_clear_page_writeback() to look up and pin page->mem_cgroup
before clearing PG_writeback, then not use that pointer afterward. It
is a partial revert of 62cccb8c8e ("mm: simplify lock_page_memcg()")
but leaves the pageref-holding callsites that aren't affected alone.
Link: http://lkml.kernel.org/r/20170809183825.GA26387@cmpxchg.org
Fixes: 62cccb8c8e ("mm: simplify lock_page_memcg()")
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reported-by: Jaegeuk Kim <jaegeuk@kernel.org>
Tested-by: Jaegeuk Kim <jaegeuk@kernel.org>
Reported-by: Bradley Bolen <bradleybolen@gmail.com>
Tested-by: Brad Bolen <bradleybolen@gmail.com>
Cc: Vladimir Davydov <vdavydov@virtuozzo.com>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: <stable@vger.kernel.org> [4.6+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
refcount_t type and corresponding API should be
used instead of atomic_t when the variable is used as
a reference counter. This allows to avoid accidental
refcounter overflows that might lead to use-after-free
situations.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Due to commit e6afc8ace6 ("udp: remove
headers from UDP packets before queueing"), when udp packets are being
peeked the requested extra offset is always 0 as there is no need to skip
the udp header. However, when the offset is 0 and the next skb is
of length 0, it is only returned once. The behaviour can be seen with
the following python script:
from socket import *;
f=socket(AF_INET6, SOCK_DGRAM | SOCK_NONBLOCK, 0);
g=socket(AF_INET6, SOCK_DGRAM | SOCK_NONBLOCK, 0);
f.bind(('::', 0));
addr=('::1', f.getsockname()[1]);
g.sendto(b'', addr)
g.sendto(b'b', addr)
print(f.recvfrom(10, MSG_PEEK));
print(f.recvfrom(10, MSG_PEEK));
Where the expected output should be the empty string twice.
Instead, make sk_peek_offset return negative values, and pass those values
to __skb_try_recv_datagram/__skb_try_recv_from_queue. If the passed offset
to __skb_try_recv_from_queue is negative, the checked skb is never skipped.
__skb_try_recv_from_queue will then ensure the offset is reset back to 0
if a peek is requested without an offset, unless no packets are found.
Also simplify the if condition in __skb_try_recv_from_queue. If _off is
greater then 0, and off is greater then or equal to skb->len, then
(_off || skb->len) must always be true assuming skb->len >= 0 is always
true.
Also remove a redundant check around a call to sk_peek_offset in af_unix.c,
as it double checked if MSG_PEEK was set in the flags.
V2:
- Moved the negative fixup into __skb_try_recv_from_queue, and remove now
redundant checks
- Fix peeking in udp{,v6}_recvmsg to report the right value when the
offset is 0
V3:
- Marked new branch in __skb_try_recv_from_queue as unlikely.
Signed-off-by: Matthew Dawson <matthew@mjdsystems.ca>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull "Reset controller changes for v4.14" from Philipp Zabel:
- constify zx2967 reset_ops
- add a convenience API to manage an array of resets
- let deassert report success and let assert report success for shared resets
if the reset controller driver does not implement (de)assert.
- add HSDKv1 reset driver
- remove Gemini reset controller, the driver is made obsolete
by a combined clock/reset driver in drivers/clk
- fix the total number of reset lines in the sunxi driver
- various uniphier updates and fixes:
- remove sLD3 SoC support
- simplify system reset register and bit definitions
- add audio systems, video input subsystem, and analog amplifiers reset
controls
* tag 'reset-for-4.14' of git://git.pengutronix.de/git/pza/linux:
reset: uniphier: add analog amplifiers reset control
reset: uniphier: add video input subsystem reset control
reset: uniphier: add audio systems reset control
reset: sunxi: fix number of reset lines
reset: uniphier: do not use per-SoC macro for system reset block
reset: uniphier: remove sLD3 SoC support
Revert "reset: Add a Gemini reset controller"
ARC: reset: introduce HSDKv1 reset driver
reset: make (de)assert report success for self-deasserting reset drivers
reset: Add APIs to manage array of resets
reset: zx2967: constify zx2967_reset_ops.
Pull "soc/tegra: Changes for v4.14-rc1" from Thierry Reding:
Contains a fix for unbalanced reference counting of device tree nodes in
the PMC-based generic power domains code.
A second change moves the SoC device registration code from its old
location in arch/arm/mach-tegra to drivers/soc/tegra so that it can be
shared between 32-bit and 64-bit ARM Tegra SoCs.
* tag 'tegra-for-4.14-soc' of git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux:
soc/tegra: Register SoC device
soc/tegra: Fix bad of_node_put() in powergate init
This patch series primarily increases sizes of variables that hold
lid values from 16 to 32 bits. Additionally, it adds a check in
the IB mad stack to verify a properly formatted MAD when OPA
extended LIDs are used.
Signed-off-by: Don Hiatt <don.hiatt@intel.com>
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Conflicts:
drivers/infiniband/core/iwcm.c - The rdma_netlink patches in
HEAD and the iwarp cm workqueue fix (don't use WQ_MEM_RECLAIM,
we aren't safe for that context) touched the same code.
Signed-off-by: Doug Ledford <dledford@redhat.com>
A sockaddr_in structure on the stack getting passed into rdma_ip2gid
triggers this warning, since we memcpy into a larger sockaddr_in6
structure:
In function 'memcpy',
inlined from 'rdma_ip2gid' at include/rdma/ib_addr.h:175:3,
inlined from 'addr_event.isra.4.constprop' at drivers/infiniband/core/roce_gid_mgmt.c:693:2,
inlined from 'inetaddr_event' at drivers/infiniband/core/roce_gid_mgmt.c:716:9:
include/linux/string.h:305:4: error: call to '__read_overflow2' declared with attribute error: detected read beyond size of object passed as 2nd parameter
The warning seems appropriate here, but the code is also clearly
correct, so we really just want to shut up this instance of the
output.
The best way I found so far is to avoid the memcpy() call and instead
replace it with a struct assignment.
Fixes: 6974f0c455 ("include/linux/string.h: add the option of fortified string.h functions")
Cc: Daniel Micay <danielmicay@gmail.com>
Cc: Kees Cook <keescook@chromium.org>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Some drivers (specifically the nes IB driver), want to create a lot of
sysfs driver attributes. Instead of open-coding the creation and
removal of these files (and getting it wrong btw), it's a better idea to
let the driver core handle all of this logic for us.
So add a new field to the pci driver structure, **groups, that allows
pci drivers to specify an attribute group list it wishes to have created
when it is registered with the driver core.
Big bonus is now the driver doesn't race with userspace when the sysfs
files are created vs. when the kobject is announced, so any script/tool
that actually wanted to use these files will not have to poll waiting
for them to show up.
Cc: Faisal Latif <faisal.latif@intel.com>
Cc: Doug Ledford <dledford@redhat.com>
Cc: Sean Hefty <sean.hefty@intel.com>
Cc: Hal Rosenstock <hal.rosenstock@gmail.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Add an API to get "pci_epf_device_id" matching the EPF name. This can be
used by the EPF driver to get the driver data corresponding to the EPF
device name.
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
[bhelgaas: folded in "while" loop termination fix from Colin Ian King
<colin.king@canonical.com>]
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
A new mount option "cpuset_v2_mode" is added to the v1 cgroupfs
filesystem to enable cpuset controller to use v2 behavior in a v1
cgroup. This mount option applies only to cpuset controller and have
no effect on other controllers.
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Archit requested this backmerge to facilitate merging some patches
depending on changes between -rc2 & -rc5
Signed-off-by: Sean Paul <seanpaul@chromium.org>
Since blk_mq_ops.reinit_request is only called from inside
blk_mq_reinit_tagset(), make this function pointer an argument of
blk_mq_reinit_tagset() instead of a member of struct blk_mq_ops.
This patch does not change any functionality but makes
blk_mq_reinit_tagset() calls easier to read and to analyze.
Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Sagi Grimberg <sagi@grimberg.me>
Cc: James Smart <james.smart@broadcom.com>
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The hardlockup detector on x86 uses a performance counter based on unhalted
CPU cycles and a periodic hrtimer. The hrtimer period is about 2/5 of the
performance counter period, so the hrtimer should fire 2-3 times before the
performance counter NMI fires. The NMI code checks whether the hrtimer
fired since the last invocation. If not, it assumess a hard lockup.
The calculation of those periods is based on the nominal CPU
frequency. Turbo modes increase the CPU clock frequency and therefore
shorten the period of the perf/NMI watchdog. With extreme Turbo-modes (3x
nominal frequency) the perf/NMI period is shorter than the hrtimer period
which leads to false positives.
A simple fix would be to shorten the hrtimer period, but that comes with
the side effect of more frequent hrtimer and softlockup thread wakeups,
which is not desired.
Implement a low pass filter, which checks the perf/NMI period against
kernel time. If the perf/NMI fires before 4/5 of the watchdog period has
elapsed then the event is ignored and postponed to the next perf/NMI.
That solves the problem and avoids the overhead of shorter hrtimer periods
and more frequent softlockup thread wakeups.
Fixes: 58687acba5 ("lockup_detector: Combine nmi_watchdog and softlockup detector")
Reported-and-tested-by: Kan Liang <Kan.liang@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: dzickus@redhat.com
Cc: prarit@redhat.com
Cc: ak@linux.intel.com
Cc: babu.moger@oracle.com
Cc: peterz@infradead.org
Cc: eranian@google.com
Cc: acme@redhat.com
Cc: stable@vger.kernel.org
Cc: atomlin@redhat.com
Cc: akpm@linux-foundation.org
Cc: torvalds@linux-foundation.org
Link: http://lkml.kernel.org/r/alpine.DEB.2.20.1708150931310.1886@nanos
For an already existing irqdomain hierarchy, as might be obtained via
a call to pci_enable_msix_range(), a PCI driver wishing to add an
additional irqdomain to the hierarchy needs to be able to insert the
irqdomain to that already initialized hierarchy. Calling
irq_domain_create_hierarchy() allows the new irqdomain to be created,
but no existing code allows for initializing the associated irq_data.
Add a couple of helper functions (irq_domain_push_irq() and
irq_domain_pop_irq()) to initialize the irq_data for the new
irqdomain added to an existing hierarchy.
Signed-off-by: David Daney <david.daney@cavium.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Marc Zyngier <marc.zyngier@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Alexandre Courbot <gnurou@gmail.com>
Cc: Linus Walleij <linus.walleij@linaro.org>
Cc: linux-gpio@vger.kernel.org
Link: http://lkml.kernel.org/r/1503017616-3252-6-git-send-email-david.daney@cavium.com
For SoC to achieve its lowest power platform idle state a set of hardware
preconditions must be met. These preconditions or constraints can be
obtained by issuing a device specific method (_DSM) with function "1".
Refer to the document provided in the link below.
Here during initialization (from attach() callback of LPS0 device), invoke
function 1 to get the device constraints. Each enabled constraint is
stored in a table.
The devices in this table are used to check whether they were in required
minimum state, while entering suspend. This check is done from platform
freeze wake() callback, only when /sys/power/pm_debug_messages attribute
is non zero.
If any constraint is not met and device is ACPI power managed then it
prints the device information to kernel logs.
Also if debug is enabled in acpi/sleep.c, the constraint table and state
of each device on wake is dumped in kernel logs.
Since pm_debug_messages_on setting is used as condition to check
constraints outside kernel/power/main.c, pm_debug_messages_on is changed
to a global variable.
Link: http://www.uefi.org/sites/default/files/resources/Intel_ACPI_Low_Power_S0_Idle.pdf
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
This reverts commit 68c4a4f8ab, with
various conflict clean-ups.
The capability check required too much privilege compared to simple DAC
controls. A system builder was forced to have crash handler processes
run with CAP_SYSLOG which would give it the ability to read (and wipe)
the _current_ dmesg, which is much more access than being given access
only to the historical log stored in pstorefs.
With the prior commit to make the root directory 0750, the files are
protected by default but a system builder can now opt to give access
to a specific group (via chgrp on the pstorefs root directory) without
being forced to also give away CAP_SYSLOG.
Suggested-by: Nick Kralevich <nnk@google.com>
Signed-off-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Petr Mladek <pmladek@suse.cz>
Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
dq_data_lock is currently used to protect all modifications of quota
accounting information, consistency of quota accounting on the inode,
and dquot pointers from inode. As a result contention on the lock can be
pretty heavy.
Reduce the contention on the lock by protecting quota accounting
information by a new dquot->dq_dqb_lock and consistency of quota
accounting with inode usage by inode->i_lock.
This change reduces time to create 500000 files on ext4 on ramdisk by 50
different processes in separate directories by 6% when user quota is
turned on. When those 50 processes belong to 50 different users, the
improvement is about 9%.
Signed-off-by: Jan Kara <jack@suse.cz>
Provide helper __inode_get_bytes() which assumes i_lock is already
acquired. Quota code will need this to be able to use i_lock to protect
consistency of quota accounting information and inode usage.
Signed-off-by: Jan Kara <jack@suse.cz>
inode_add_rsv_space() and inode_sub_rsv_space() had only one callsite.
Inline them there directly. inode_claim_rsv_space() and
inode_reclaim_rsv_space() had two callsites so inline them there as
well. This will simplify further locking changes.
Signed-off-by: Jan Kara <jack@suse.cz>
Filesystems that are journalling quotas generally don't need tracking of
dirty dquots in a list since forcing a transaction commit flushes all
quotas anyway. Allow filesystem to say it doesn't want dquots to be
tracked as it reduces contention on the dq_list_lock.
Signed-off-by: Jan Kara <jack@suse.cz>
Currently every dquot carries a wait_queue_head_t used only when we are
turning quotas off to wait for last users to drop dquot references.
Since such rare case is not performance sensitive in any means, just use
a global waitqueue for this and save space in struct dquot. Also convert
the logic to use wait_event() instead of open-coding it.
Signed-off-by: Jan Kara <jack@suse.cz>
omapdrm changes for v4.14
* HDMI hot plug IRQ support (instead of polling)
* Big driver cleanup from Laurent (no functional changes)
* OMAP5 DSI support (only the pinmuxing was missing)
* tag 'omapdrm-4.14' of git://git.kernel.org/pub/scm/linux/kernel/git/tomba/linux: (60 commits)
drm/omap: Potential NULL deref in omap_crtc_duplicate_state()
drm/omap: remove no-op cleanup code
drm/omap: rename omapdrm device back
drm: omapdrm: Remove omapdrm platform data
ARM: OMAP2+: Don't register omapdss device for omapdrm
ARM: OMAP2+: Remove unused omapdrm platform device
drm: omapdrm: Remove the omapdss driver
drm: omapdrm: Register omapdrm platform device in omapdss driver
drm: omapdrm: hdmi: Don't allocate PHY features dynamically
drm: omapdrm: hdmi: Configure the PHY from the HDMI core version
drm: omapdrm: hdmi: Configure the PLL from the HDMI core version
drm: omapdrm: hdmi: Pass HDMI core version as integer to HDMI audio
drm: omapdrm: hdmi: Replace OMAP SoC model check with HDMI xmit version
drm: omapdrm: hdmi: Rename functions and structures to use hdmi_ prefix
drm/omap: add OMAP5 DSIPHY lane-enable support
drm/omap: use regmap_update_bit() when muxing DSI pads
drm: omapdrm: Remove dss_features.h
drm: omapdrm: Move supported outputs feature to dss driver
drm: omapdrm: Move DSS_FCK feature to dss driver
drm: omapdrm: Move PCD, LINEWIDTH and DOWNSCALE features to dispc driver
...
Runtime GPEs have corresponding _Lxx/_Exx methods and are enabled
automatically during the initialization of the ACPI subsystem through
acpi_update_all_gpes() with the assumption that acpi_setup_gpe_for_wake()
will be called in advance for all of the GPEs pointed to by _PRW
objects in the namespace that may be affected by acpi_update_all_gpes().
That is, acpi_ev_initialize_gpe_block() can only be called for a GPE
block after acpi_setup_gpe_for_wake() has been called for all of the
_PRW (wakeup) GPEs in it.
The platform firmware on some systems, however, expects GPEs to be
enabled before the enumeration of devices which is when
acpi_setup_gpe_for_wake() is called and that goes against the above
assumption.
For this reason, introduce a new flag to be set by
acpi_ev_initialize_gpe_block() when automatically enabling a GPE
to indicate to acpi_setup_gpe_for_wake() that it needs to drop the
reference to the GPE coming from acpi_ev_initialize_gpe_block()
and modify acpi_setup_gpe_for_wake() accordingly. These changes
allow acpi_setup_gpe_for_wake() and acpi_ev_initialize_gpe_block()
to be invoked in any order.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Tested-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Christian Brauner reported that if you use the TIOCGPTPEER ioctl() to
get a slave pty file descriptor, the resulting file descriptor doesn't
look right in /proc/<pid>/fd/<fd>. In particular, he wanted to use
readlink() on /proc/self/fd/<fd> to get the pathname of the slave pty
(basically implementing "ptsname{_r}()").
The reason for that was that we had generated the wrong 'struct path'
when we create the pty in ptmx_open().
In particular, the dentry was correct, but the vfsmount pointed to the
mount of the ptmx node. That _can_ be correct - in case you use
"/dev/pts/ptmx" to open the master - but usually is not. The normal
case is to use /dev/ptmx, which then looks up the pts/ directory, and
then the vfsmount of the ptmx node is obviously the /dev directory, not
the /dev/pts/ directory.
We actually did have the right vfsmount available, but in the wrong
place (it gets looked up in 'devpts_acquire()' when we get a reference
to the pts filesystem), and so ptmx_open() used the wrong mnt pointer.
The end result of this confusion was that the pty worked fine, but when
if you did TIOCGPTPEER to get the slave side of the pty, end end result
would also work, but have that dodgy 'struct path'.
And then when doing "d_path()" on to get the pathname, the vfsmount
would not match the root of the pts directory, and d_path() would return
an empty pathname thinking that the entry had escaped a bind mount into
another mount.
This fixes the problem by making devpts_acquire() return the vfsmount
for the pts filesystem, allowing ptmx_open() to trivially just use the
right mount for the pts dentry, and create the proper 'struct path'.
Reported-by: Christian Brauner <christian.brauner@ubuntu.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Acked-by: Eric Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Move this code from arch/arm/mach-tegra and make it common among 32-bit
and 64-bit Tegra SoCs. This is slightly complicated by the fact that on
32-bit Tegra, the SoC device is used as the parent for all devices that
are instantiated from device tree.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Implement MEMBARRIER_CMD_PRIVATE_EXPEDITED with IPIs using cpumask built
from all runqueues for which current thread's mm is the same as the
thread calling sys_membarrier. It executes faster than the non-expedited
variant (no blocking). It also works on NOHZ_FULL configurations.
Scheduler-wise, it requires a memory barrier before and after context
switching between processes (which have different mm). The memory
barrier before context switch is already present. For the barrier after
context switch:
* Our TSO archs can do RELEASE without being a full barrier. Look at
x86 spin_unlock() being a regular STORE for example. But for those
archs, all atomics imply smp_mb and all of them have atomic ops in
switch_mm() for mm_cpumask(), and on x86 the CR3 load acts as a full
barrier.
* From all weakly ordered machines, only ARM64 and PPC can do RELEASE,
the rest does indeed do smp_mb(), so there the spin_unlock() is a full
barrier and we're good.
* ARM64 has a very heavy barrier in switch_to(), which suffices.
* PPC just removed its barrier from switch_to(), but appears to be
talking about adding something to switch_mm(). So add a
smp_mb__after_unlock_lock() for now, until this is settled on the PPC
side.
Changes since v3:
- Properly document the memory barriers provided by each architecture.
Changes since v2:
- Address comments from Peter Zijlstra,
- Add smp_mb__after_unlock_lock() after finish_lock_switch() in
finish_task_switch() to add the memory barrier we need after storing
to rq->curr. This is much simpler than the previous approach relying
on atomic_dec_and_test() in mmdrop(), which actually added a memory
barrier in the common case of switching between userspace processes.
- Return -EINVAL when MEMBARRIER_CMD_SHARED is used on a nohz_full
kernel, rather than having the whole membarrier system call returning
-ENOSYS. Indeed, CMD_PRIVATE_EXPEDITED is compatible with nohz_full.
Adapt the CMD_QUERY mask accordingly.
Changes since v1:
- move membarrier code under kernel/sched/ because it uses the
scheduler runqueue,
- only add the barrier when we switch from a kernel thread. The case
where we switch from a user-space thread is already handled by
the atomic_dec_and_test() in mmdrop().
- add a comment to mmdrop() documenting the requirement on the implicit
memory barrier.
CC: Peter Zijlstra <peterz@infradead.org>
CC: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
CC: Boqun Feng <boqun.feng@gmail.com>
CC: Andrew Hunter <ahh@google.com>
CC: Maged Michael <maged.michael@gmail.com>
CC: gromer@google.com
CC: Avi Kivity <avi@scylladb.com>
CC: Benjamin Herrenschmidt <benh@kernel.crashing.org>
CC: Paul Mackerras <paulus@samba.org>
CC: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Tested-by: Dave Watson <davejwatson@fb.com>
