I recently debugged a DMA mapping oops where a driver was trying to map
a buffer returned from request_firmware() with dma_map_single(). Memory
returned from request_firmware() is mapped into the vmalloc region and
this isn't a valid region to map with dma_map_single() per the DMA
documentation's "What memory is DMA'able?" section.
Unfortunately, we don't really check that in the DMA debugging code, so
enabling DMA debugging doesn't help catch this problem. Let's add a new
DMA debug function to check for a vmalloc address or an invalid virtual
address and print a warning if this happens. This makes it a little
easier to debug these sorts of problems, instead of seeing odd behavior
or crashes when drivers attempt to map the vmalloc space for DMA.
Cc: Marek Szyprowski <m.szyprowski@samsung.com>
Reviewed-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Andrei Vagin <avagin@gmail.com> reported:
> Accoding to the man page, the user should not set si_signo, it has to be set
> by kernel.
>
> $ man 2 rt_sigqueueinfo
>
> The uinfo argument specifies the data to accompany the signal. This
> argument is a pointer to a structure of type siginfo_t, described in
> sigaction(2) (and defined by including <sigaction.h>). The caller
> should set the following fields in this structure:
>
> si_code
> This must be one of the SI_* codes in the Linux kernel source
> file include/asm-generic/siginfo.h, with the restriction that
> the code must be negative (i.e., cannot be SI_USER, which is
> used by the kernel to indicate a signal sent by kill(2)) and
> cannot (since Linux 2.6.39) be SI_TKILL (which is used by the
> kernel to indicate a signal sent using tgkill(2)).
>
> si_pid This should be set to a process ID, typically the process ID of
> the sender.
>
> si_uid This should be set to a user ID, typically the real user ID of
> the sender.
>
> si_value
> This field contains the user data to accompany the signal. For
> more information, see the description of the last (union sigval)
> argument of sigqueue(3).
>
> Internally, the kernel sets the si_signo field to the value specified
> in sig, so that the receiver of the signal can also obtain the signal
> number via that field.
>
> On Tue, Sep 25, 2018 at 07:19:02PM +0200, Eric W. Biederman wrote:
>>
>> If there is some application that calls sigqueueinfo directly that has
>> a problem with this added sanity check we can revisit this when we see
>> what kind of crazy that application is doing.
>
>
> I already know two "applications" ;)
>
> https://github.com/torvalds/linux/blob/master/tools/testing/selftests/ptrace/peeksiginfo.c
> https://github.com/checkpoint-restore/criu/blob/master/test/zdtm/static/sigpending.c
>
> Disclaimer: I'm the author of both of them.
Looking at the kernel code the historical behavior has alwasy been to prefer
the signal number passed in by the kernel.
So sigh. Implmenet __copy_siginfo_from_user and __copy_siginfo_from_user32 to
take that signal number and prefer it. The user of ptrace will still
use copy_siginfo_from_user and copy_siginfo_from_user32 as they do not and
never have had a signal number there.
Luckily this change has never made it farther than linux-next.
Fixes: e75dc036c4 ("signal: Fail sigqueueinfo if si_signo != sig")
Reported-by: Andrei Vagin <avagin@gmail.com>
Tested-by: Andrei Vagin <avagin@gmail.com>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Dave writes:
"Networking fixes:
1) Fix truncation of 32-bit right shift in bpf, from Jann Horn.
2) Fix memory leak in wireless wext compat, from Stefan Seyfried.
3) Use after free in cfg80211's reg_process_hint(), from Yu Zhao.
4) Need to cancel pending work when unbinding in smsc75xx otherwise
we oops, also from Yu Zhao.
5) Don't allow enslaving a team device to itself, from Ido Schimmel.
6) Fix backwards compat with older userspace for rtnetlink FDB dumps.
From Mauricio Faria.
7) Add validation of tc policy netlink attributes, from David Ahern.
8) Fix RCU locking in rawv6_send_hdrinc(), from Wei Wang."
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (26 commits)
net: mvpp2: Extract the correct ethtype from the skb for tx csum offload
ipv6: take rcu lock in rawv6_send_hdrinc()
net: sched: Add policy validation for tc attributes
rtnetlink: fix rtnl_fdb_dump() for ndmsg header
yam: fix a missing-check bug
net: bpfilter: Fix type cast and pointer warnings
net: cxgb3_main: fix a missing-check bug
bpf: 32-bit RSH verification must truncate input before the ALU op
net: phy: phylink: fix SFP interface autodetection
be2net: don't flip hw_features when VXLANs are added/deleted
net/packet: fix packet drop as of virtio gso
net: dsa: b53: Keep CPU port as tagged in all VLANs
openvswitch: load NAT helper
bnxt_en: get the reduced max_irqs by the ones used by RDMA
bnxt_en: free hwrm resources, if driver probe fails.
bnxt_en: Fix enables field in HWRM_QUEUE_COS2BW_CFG request
bnxt_en: Fix VNIC reservations on the PF.
team: Forbid enslaving team device to itself
net/usb: cancel pending work when unbinding smsc75xx
mlxsw: spectrum: Delete RIF when VLAN device is removed
...
Ingo writes:
"perf fixes:
- fix a CPU#0 hot unplug bug and a PCI enumeration bug in the x86 Intel uncore PMU driver
- fix a CPU event enumeration bug in the x86 AMD PMU driver
- fix a perf ring-buffer corruption bug when using tracepoints
- fix a PMU unregister locking bug"
* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
perf/x86/amd/uncore: Set ThreadMask and SliceMask for L3 Cache perf events
perf/x86/intel/uncore: Fix PCI BDF address of M3UPI on SKX
perf/ring_buffer: Prevent concurent ring buffer access
perf/x86/intel/uncore: Use boot_cpu_data.phys_proc_id instead of hardcorded physical package ID 0
perf/core: Fix perf_pmu_unregister() locking
Ingo writes:
"scheduler fixes:
These fixes address a rather involved performance regression between
v4.17->v4.19 in the sched/numa auto-balancing code. Since distros
really need this fix we accelerated it to sched/urgent for a faster
upstream merge.
NUMA scheduling and balancing performance is now largely back to
v4.17 levels, without reintroducing the NUMA placement bugs that
v4.18 and v4.19 fixed.
Many thanks to Srikar Dronamraju, Mel Gorman and Jirka Hladky, for
reporting, testing, re-testing and solving this rather complex set of
bugs."
* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
sched/numa: Migrate pages to local nodes quicker early in the lifetime of a task
mm, sched/numa: Remove rate-limiting of automatic NUMA balancing migration
sched/numa: Avoid task migration for small NUMA improvement
mm/migrate: Use spin_trylock() while resetting rate limit
sched/numa: Limit the conditions where scan period is reset
sched/numa: Reset scan rate whenever task moves across nodes
sched/numa: Pass destination CPU as a parameter to migrate_task_rq
sched/numa: Stop multiple tasks from moving to the CPU at the same time
Daniel Borkmann says:
====================
pull-request: bpf 2018-10-05
The following pull-request contains BPF updates for your *net* tree.
The main changes are:
1) Fix to truncate input on ALU operations in 32 bit mode, from Jann.
2) Fixes for cgroup local storage to reject reserved flags on element
update and rejection of map allocation with zero-sized value, from Roman.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
When I wrote commit 468f6eafa6 ("bpf: fix 32-bit ALU op verification"), I
assumed that, in order to emulate 64-bit arithmetic with 32-bit logic, it
is sufficient to just truncate the output to 32 bits; and so I just moved
the register size coercion that used to be at the start of the function to
the end of the function.
That assumption is true for almost every op, but not for 32-bit right
shifts, because those can propagate information towards the least
significant bit. Fix it by always truncating inputs for 32-bit ops to 32
bits.
Also get rid of the coerce_reg_to_size() after the ALU op, since that has
no effect.
Fixes: 468f6eafa6 ("bpf: fix 32-bit ALU op verification")
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Jann Horn <jannh@google.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
log_buf_len_setup does not check input argument before passing it to
simple_strtoull. The argument would be a NULL pointer if "log_buf_len",
without its value, is set in command line and thus causes the following
panic.
PANIC: early exception 0xe3 IP 10:ffffffffaaeacd0d error 0 cr2 0x0
[ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 4.19.0-rc4-yocto-standard+ #1
[ 0.000000] RIP: 0010:_parse_integer_fixup_radix+0xd/0x70
...
[ 0.000000] Call Trace:
[ 0.000000] simple_strtoull+0x29/0x70
[ 0.000000] memparse+0x26/0x90
[ 0.000000] log_buf_len_setup+0x17/0x22
[ 0.000000] do_early_param+0x57/0x8e
[ 0.000000] parse_args+0x208/0x320
[ 0.000000] ? rdinit_setup+0x30/0x30
[ 0.000000] parse_early_options+0x29/0x2d
[ 0.000000] ? rdinit_setup+0x30/0x30
[ 0.000000] parse_early_param+0x36/0x4d
[ 0.000000] setup_arch+0x336/0x99e
[ 0.000000] start_kernel+0x6f/0x4ee
[ 0.000000] x86_64_start_reservations+0x24/0x26
[ 0.000000] x86_64_start_kernel+0x6f/0x72
[ 0.000000] secondary_startup_64+0xa4/0xb0
This patch adds a check to prevent the panic.
Link: http://lkml.kernel.org/r/1538239553-81805-1-git-send-email-zhe.he@windriver.com
Cc: stable@vger.kernel.org
Cc: rostedt@goodmis.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: He Zhe <zhe.he@windriver.com>
Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
It appears that in commit 9d7a224b46 ("dma-direct: always allow dma mask
<= physiscal memory size") the logic of the test was changed from a "<" to
a ">=" however I don't see any reason for that change. I am assuming that
there was some additional change planned, specifically I suspect the logic
was intended to be reversed and possibly used for a return. Since that is
the case I have gone ahead and done that.
This addresses issues I had on my system that prevented me from booting
with the above mentioned commit applied on an x86_64 system w/ Intel IOMMU.
Fixes: 9d7a224b46 ("dma-direct: always allow dma mask <= physiscal memory size")
Signed-off-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Acked-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
A cgroup which is already a threaded domain may be converted into a
threaded cgroup if the prerequisite conditions are met. When this
happens, all threaded descendant should also have their ->dom_cgrp
updated to the new threaded domain cgroup. Unfortunately, this
propagation was missing leading to the following failure.
# cd /sys/fs/cgroup/unified
# cat cgroup.subtree_control # show that no controllers are enabled
# mkdir -p mycgrp/a/b/c
# echo threaded > mycgrp/a/b/cgroup.type
At this point, the hierarchy looks as follows:
mycgrp [d]
a [dt]
b [t]
c [inv]
Now let's make node "a" threaded (and thus "mycgrp" s made "domain threaded"):
# echo threaded > mycgrp/a/cgroup.type
By this point, we now have a hierarchy that looks as follows:
mycgrp [dt]
a [t]
b [t]
c [inv]
But, when we try to convert the node "c" from "domain invalid" to
"threaded", we get ENOTSUP on the write():
# echo threaded > mycgrp/a/b/c/cgroup.type
sh: echo: write error: Operation not supported
This patch fixes the problem by
* Moving the opencoded ->dom_cgrp save and restoration in
cgroup_enable_threaded() into cgroup_{save|restore}_control() so
that mulitple cgroups can be handled.
* Updating all threaded descendants' ->dom_cgrp to point to the new
dom_cgrp when enabling threaded mode.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-and-tested-by: "Michael Kerrisk (man-pages)" <mtk.manpages@gmail.com>
Reported-by: Amin Jamali <ajamali@pivotal.io>
Reported-by: Joao De Almeida Pereira <jpereira@pivotal.io>
Link: https://lore.kernel.org/r/CAKgNAkhHYCMn74TCNiMJ=ccLd7DcmXSbvw3CbZ1YREeG7iJM5g@mail.gmail.com
Fixes: 454000adaa ("cgroup: introduce cgroup->dom_cgrp and threaded css_set handling")
Cc: stable@vger.kernel.org # v4.14+
Minor conflict in net/core/rtnetlink.c, David Ahern's bug fix in 'net'
overlapped the renaming of a netlink attribute in net-next.
Signed-off-by: David S. Miller <davem@davemloft.net>
We reserve 128 bytes for struct siginfo but only use about 48 bytes on
64bit and 32 bytes on 32bit. Someday we might use more but it is unlikely
to be anytime soon.
Userspace seems content with just enough bytes of siginfo to implement
sigqueue. Or in the case of checkpoint/restart reinjecting signals
the kernel has sent.
Reducing the stack footprint and the work to copy siginfo around from
2 cachelines to 1 cachelines seems worth doing even if I don't have
benchmarks to show a performance difference.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Linus recently observed that if we did not worry about the padding
member in struct siginfo it is only about 48 bytes, and 48 bytes is
much nicer than 128 bytes for allocating on the stack and copying
around in the kernel.
The obvious thing of only adding the padding when userspace is
including siginfo.h won't work as there are sigframe definitions in
the kernel that embed struct siginfo.
So split siginfo in two; kernel_siginfo and siginfo. Keeping the
traditional name for the userspace definition. While the version that
is used internally to the kernel and ultimately will not be padded to
128 bytes is called kernel_siginfo.
The definition of struct kernel_siginfo I have put in include/signal_types.h
A set of buildtime checks has been added to verify the two structures have
the same field offsets.
To make it easy to verify the change kernel_siginfo retains the same
size as siginfo. The reduction in size comes in a following change.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
In preparation for using a smaller version of siginfo in the kernel
introduce copy_siginfo_from_user and use it when siginfo is copied from
userspace.
Make the pattern for using copy_siginfo_from_user and
copy_siginfo_from_user32 to capture the return value and return that
value on error.
This is a necessary prerequisite for using a smaller siginfo
in the kernel than the kernel exports to userspace.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Rework the defintion of struct siginfo so that the array padding
struct siginfo to SI_MAX_SIZE can be placed in a union along side of
the rest of the struct siginfo members. The result is that we no
longer need the __ARCH_SI_PREAMBLE_SIZE or SI_PAD_SIZE definitions.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
The kernel needs to validate that the contents of struct siginfo make
sense as siginfo is copied into the kernel, so that the proper union
members can be put in the appropriate locations. The field si_signo
is a fundamental part of that validation. As such changing the
contents of si_signo after the validation make no sense and can result
in nonsense values in the kernel.
As such simply fail if someone is silly enough to set si_signo out of
sync with the signal number passed to sigqueueinfo.
I don't expect a problem as glibc's sigqueue implementation sets
"si_signo = sig" and CRIU just returns to the kernel what the kernel
gave to it.
If there is some application that calls sigqueueinfo directly that has
a problem with this added sanity check we can revisit this when we see
what kind of crazy that application is doing.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
When moving all of the architectures specific si_codes into
siginfo.h, I apparently overlooked EMT_TAGOVF. Move it now.
Remove the now redundant test in siginfo_layout for SIGEMT
as now NSIGEMT is always defined.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
If CONFIG_WW_MUTEX_SELFTEST=y is enabled, booting an image
in an arm64 virtual machine results in the following
traceback if 8 CPUs are enabled:
DEBUG_LOCKS_WARN_ON(__owner_task(owner) != current)
WARNING: CPU: 2 PID: 537 at kernel/locking/mutex.c:1033 __mutex_unlock_slowpath+0x1a8/0x2e0
...
Call trace:
__mutex_unlock_slowpath()
ww_mutex_unlock()
test_cycle_work()
process_one_work()
worker_thread()
kthread()
ret_from_fork()
If requesting b_mutex fails with -EDEADLK, the error variable
is reassigned to the return value from calling ww_mutex_lock
on a_mutex again. If this call fails, a_mutex is not locked.
It is, however, unconditionally unlocked subsequently, causing
the reported warning. Fix the problem by using two error variables.
With this change, the selftest still fails as follows:
cyclic deadlock not resolved, ret[7/8] = -35
However, the traceback is gone.
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will.deacon@arm.com>
Fixes: d1b42b800e ("locking/ww_mutex: Add kselftests for resolving ww_mutex cyclic deadlocks")
Link: http://lkml.kernel.org/r/1538516929-9734-1-git-send-email-linux@roeck-us.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
This patch adds new BPF helper functions, bpf_sk_lookup_tcp() and
bpf_sk_lookup_udp() which allows BPF programs to find out if there is a
socket listening on this host, and returns a socket pointer which the
BPF program can then access to determine, for instance, whether to
forward or drop traffic. bpf_sk_lookup_xxx() may take a reference on the
socket, so when a BPF program makes use of this function, it must
subsequently pass the returned pointer into the newly added sk_release()
to return the reference.
By way of example, the following pseudocode would filter inbound
connections at XDP if there is no corresponding service listening for
the traffic:
struct bpf_sock_tuple tuple;
struct bpf_sock_ops *sk;
populate_tuple(ctx, &tuple); // Extract the 5tuple from the packet
sk = bpf_sk_lookup_tcp(ctx, &tuple, sizeof tuple, netns, 0);
if (!sk) {
// Couldn't find a socket listening for this traffic. Drop.
return TC_ACT_SHOT;
}
bpf_sk_release(sk, 0);
return TC_ACT_OK;
Signed-off-by: Joe Stringer <joe@wand.net.nz>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Allow helper functions to acquire a reference and return it into a
register. Specific pointer types such as the PTR_TO_SOCKET will
implicitly represent such a reference. The verifier must ensure that
these references are released exactly once in each path through the
program.
To achieve this, this commit assigns an id to the pointer and tracks it
in the 'bpf_func_state', then when the function or program exits,
verifies that all of the acquired references have been freed. When the
pointer is passed to a function that frees the reference, it is removed
from the 'bpf_func_state` and all existing copies of the pointer in
registers are marked invalid.
Signed-off-by: Joe Stringer <joe@wand.net.nz>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
An upcoming commit will need very similar copy/realloc boilerplate, so
refactor the existing stack copy/realloc functions into macros to
simplify it.
Signed-off-by: Joe Stringer <joe@wand.net.nz>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Teach the verifier a little bit about a new type of pointer, a
PTR_TO_SOCKET. This pointer type is accessed from BPF through the
'struct bpf_sock' structure.
Signed-off-by: Joe Stringer <joe@wand.net.nz>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This check will be reused by an upcoming commit for conditional jump
checks for sockets. Refactor it a bit to simplify the later commit.
Signed-off-by: Joe Stringer <joe@wand.net.nz>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
The array "reg_type_str" provides canonical formatting of register
types, however a couple of places would previously check whether a
register represented the context and write the name "context" directly.
An upcoming commit will add another pointer type to these statements, so
to provide more accurate error messages in the verifier, update these
error messages to use "reg_type_str" instead.
Signed-off-by: Joe Stringer <joe@wand.net.nz>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
An upcoming commit will add another two pointer types that need very
similar behaviour, so generalise this function now.
Signed-off-by: Joe Stringer <joe@wand.net.nz>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Add this iterator for spilled registers, it concentrates the details of
how to get the current frame's spilled registers into a single macro
while clarifying the intention of the code which is calling the macro.
Signed-off-by: Joe Stringer <joe@wand.net.nz>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
CON_PRINTBUFFER console registration requires us to do several
preparation steps:
- Rollback console_seq to replay logbuf messages which were already
seen on other consoles;
- Set exclusive_console flag so console_unlock() will ->write() logbuf
messages only to the exclusive_console driver.
The way we do it, however, is a bit racy
logbuf_lock_irqsave(flags);
console_seq = syslog_seq;
console_idx = syslog_idx;
logbuf_unlock_irqrestore(flags);
<< preemption enabled
<< irqs enabled
exclusive_console = newcon;
console_unlock();
We rollback console_seq under logbuf_lock with IRQs disabled, but
we set exclusive_console with local IRQs enabled and logbuf unlocked.
If the system oops-es or panic-s before we set exclusive_console - and
given that we have IRQs and preemption enabled there is such a
possibility - we will re-play all logbuf messages to every registered
console, which may be a bit annoying and time consuming.
Move exclusive_console assignment to the same IRQs-disabled and
logbuf_lock-protected section where we rollback console_seq.
Link: http://lkml.kernel.org/r/20180928095304.9972-1-sergey.senozhatsky@gmail.com
To: Steven Rostedt <rostedt@goodmis.org>
Cc: Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Automatic NUMA Balancing uses a multi-stage pass to decide whether a page
should migrate to a local node. This filter avoids excessive ping-ponging
if a page is shared or used by threads that migrate cross-node frequently.
Threads inherit both page tables and the preferred node ID from the
parent. This means that threads can trigger hinting faults earlier than
a new task which delays scanning for a number of seconds. As it can be
load balanced very early in its lifetime there can be an unnecessary delay
before it starts migrating thread-local data. This patch migrates private
pages faster early in the lifetime of a thread using the sequence counter
as an identifier of new tasks.
With this patch applied, STREAM performance is the same as 4.17 even though
processes are not spread cross-node prematurely. Other workloads showed
a mix of minor gains and losses. This is somewhat expected most workloads
are not very sensitive to the starting conditions of a process.
4.19.0-rc5 4.19.0-rc5 4.17.0
numab-v1r1 fastmigrate-v1r1 vanilla
MB/sec copy 43298.52 ( 0.00%) 47335.46 ( 9.32%) 47219.24 ( 9.06%)
MB/sec scale 30115.06 ( 0.00%) 32568.12 ( 8.15%) 32527.56 ( 8.01%)
MB/sec add 32825.12 ( 0.00%) 36078.94 ( 9.91%) 35928.02 ( 9.45%)
MB/sec triad 32549.52 ( 0.00%) 35935.94 ( 10.40%) 35969.88 ( 10.51%)
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Reviewed-by: Rik van Riel <riel@surriel.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Jirka Hladky <jhladky@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Linux-MM <linux-mm@kvack.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20181001100525.29789-3-mgorman@techsingularity.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Pull v4.20 RCU changes from Paul E. McKenney:
- Documentation updates, including some good-eye catches from
Joel Fernandes.
- SRCU updates, most notably changes enabling call_srcu() to be
invoked very early in the boot sequence.
- Torture-test updates, including some preliminary work towards
making rcutorture better able to find problems that result in
insufficient grace-period forward progress.
- Consolidate the RCU-bh, RCU-preempt, and RCU-sched flavors into
a single flavor similar to RCU-sched in !PREEMPT kernels and
into a single flavor similar to RCU-preempt (but also waiting
on preempt-disabled sequences of code) in PREEMPT kernels. This
branch also includes a refactoring of rcu_{nmi,irq}_{enter,exit}()
from Byungchul Park.
- Now that there is only one RCU flavor in any given running kernel,
the many "rsp" pointers are no longer required, and this cleanup
series removes them.
- This branch carries out additional cleanups made possible by
the RCU flavor consolidation, including inlining how-trivial
functions, updating comments and definitions, and removing
now-unneeded rcutorture scenarios.
- Initial changes to RCU to better promote forward progress of
grace periods, including fixing a bug found by Marius Hillenbrand
and David Woodhouse, with the fix suggested by Peter Zijlstra.
- Now that there is only one flavor of RCU in any running kernel,
there is also only on rcu_data structure per CPU. This means
that the rcu_dynticks structure can be merged into the rcu_data
structure, a task taken on by this branch. This branch also
contains a -rt-related fix from Mike Galbraith.
Signed-off-by: Ingo Molnar <mingo@kernel.org>
A CFS (SCHED_OTHER, SCHED_BATCH or SCHED_IDLE policy) task's
se->runnable_weight must always be in sync with its se->load.weight.
se->runnable_weight is set to se->load.weight when the task is
forked (init_entity_runnable_average()) or reniced (reweight_entity()).
There are two cases in set_load_weight() which since they currently only
set se->load.weight could lead to a situation in which se->load.weight
is different to se->runnable_weight for a CFS task:
(1) A task switches to SCHED_IDLE.
(2) A SCHED_FIFO, SCHED_RR or SCHED_DEADLINE task which has been reniced
(during which only its static priority gets set) switches to
SCHED_OTHER or SCHED_BATCH.
Set se->runnable_weight to se->load.weight in these two cases to prevent
this. This eliminates the need to explicitly set it to se->load.weight
during PELT updates in the CFS scheduler fastpath.
Signed-off-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Joel Fernandes <joelaf@google.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Morten Rasmussen <morten.rasmussen@arm.com>
Cc: Patrick Bellasi <patrick.bellasi@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Quentin Perret <quentin.perret@arm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vincent Guittot <vincent.guittot@linaro.org>
Link: http://lkml.kernel.org/r/20180803140538.1178-1-dietmar.eggemann@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
LB_BIAS allows the adjustment on how conservative load should be
balanced.
The rq->cpu_load[idx] array is used for this functionality. It contains
weighted CPU load decayed average values over different intervals
(idx = 1..4). Idx = 0 is the weighted CPU load itself.
The values are updated during scheduler_tick, before idle balance and at
nohz exit.
There are 5 different types of idx's per sched domain (sd). Each of them
is used to index into the rq->cpu_load[idx] array in a specific scenario
(busy, idle and newidle for load balancing, forkexec for wake-up
slow-path load balancing and wake for affine wakeup based on weight).
Only the sd idx's for busy and idle load balancing are set to 2,3 or 1,2
respectively. All the other sd idx's are set to 0.
Conservative load balancing is achieved for sd idx's >= 1 by using the
min/max (source_load()/target_load()) value between the current weighted
CPU load and the rq->cpu_load[sd idx -1] for the busiest(idlest)/local
CPU load in load balancing or vice versa in the wake-up slow-path load
balancing.
There is no conservative balancing for sd idx = 0 since only current
weighted CPU load is used in this case.
It is very likely that LB_BIAS' influence on load balancing can be
neglected (see test results below). This is further supported by:
(1) Weighted CPU load today is by itself a decayed average value (PELT)
(cfs_rq->avg->runnable_load_avg) and not the instantaneous load
(rq->load.weight) it was when LB_BIAS was introduced.
(2) Sd imbalance_pct is used for CPU_NEWLY_IDLE and CPU_NOT_IDLE (relate
to sd's newidle and busy idx) in find_busiest_group() when comparing
busiest and local avg load to make load balancing even more
conservative.
(3) The sd forkexec and newidle idx are always set to 0 so there is no
adjustment on how conservatively load balancing is done here.
(4) Affine wakeup based on weight (wake_affine_weight()) will not be
impacted since the sd wake idx is always set to 0.
Let's disable LB_BIAS by default for a few kernel releases to make sure
that no workload and no scheduler topology is affected. The benefit of
being able to remove the LB_BIAS dependency from source_load() and
target_load() is that the entire rq->cpu_load[idx] code could be removed
in this case.
It is really hard to say if there is no regression w/o testing this with
a lot of different workloads on a lot of different platforms, especially
NUMA machines.
The following 104 LKP (Linux Kernel Performance) tests were run by the
0-Day guys mostly on multi-socket hosts with a larger number of logical
cpus (88, 192).
The base for the test was commit b3dae109fa ("sched/swait: Rename to
exclusive") (tip/sched/core v4.18-rc1).
Only 2 out of the 104 tests had a significant change in one of the
metrics (fsmark/1x-1t-1HDD-btrfs-nfsv4-4M-60G-NoSync-performance +7%
files_per_sec, unixbench/300s-100%-syscall-performance -11% score).
Tests which showed a change in one of the metrics are marked with a '*'
and this change is listed as well.
(a) lkp-bdw-ep3:
88 threads Intel(R) Xeon(R) CPU E5-2699 v4 @ 2.20GHz 64G
dd-write/10m-1HDD-cfq-btrfs-100dd-performance
fsmark/1x-1t-1HDD-xfs-nfsv4-4M-60G-NoSync-performance
* fsmark/1x-1t-1HDD-btrfs-nfsv4-4M-60G-NoSync-performance
7.50 7% 8.00 ± 6% fsmark.files_per_sec
fsmark/1x-1t-1HDD-btrfs-nfsv4-4M-60G-fsyncBeforeClose-performance
fsmark/1x-1t-1HDD-btrfs-4M-60G-NoSync-performance
fsmark/1x-1t-1HDD-btrfs-4M-60G-fsyncBeforeClose-performance
kbuild/300s-50%-vmlinux_prereq-performance
kbuild/300s-200%-vmlinux_prereq-performance
kbuild/300s-50%-vmlinux_prereq-performance-1HDD-ext4
kbuild/300s-200%-vmlinux_prereq-performance-1HDD-ext4
(b) lkp-skl-4sp1:
192 threads Intel(R) Xeon(R) Platinum 8160 768G
dbench/100%-performance
ebizzy/200%-100x-10s-performance
hackbench/1600%-process-pipe-performance
iperf/300s-cs-localhost-tcp-performance
iperf/300s-cs-localhost-udp-performance
perf-bench-numa-mem/2t-300M-performance
perf-bench-sched-pipe/10000000ops-process-performance
perf-bench-sched-pipe/10000000ops-threads-performance
schbench/2-16-300-30000-30000-performance
tbench/100%-cs-localhost-performance
(c) lkp-bdw-ep6:
88 threads Intel(R) Xeon(R) CPU E5-2699 v4 @ 2.20GHz 128G
stress-ng/100%-60s-pipe-performance
unixbench/300s-1-whetstone-double-performance
unixbench/300s-1-shell1-performance
unixbench/300s-1-shell8-performance
unixbench/300s-1-pipe-performance
* unixbench/300s-1-context1-performance
312 315 unixbench.score
unixbench/300s-1-spawn-performance
unixbench/300s-1-syscall-performance
unixbench/300s-1-dhry2reg-performance
unixbench/300s-1-fstime-performance
unixbench/300s-1-fsbuffer-performance
unixbench/300s-1-fsdisk-performance
unixbench/300s-100%-whetstone-double-performance
unixbench/300s-100%-shell1-performance
unixbench/300s-100%-shell8-performance
unixbench/300s-100%-pipe-performance
unixbench/300s-100%-context1-performance
unixbench/300s-100%-spawn-performance
* unixbench/300s-100%-syscall-performance
3571 ± 3% -11% 3183 ± 4% unixbench.score
unixbench/300s-100%-dhry2reg-performance
unixbench/300s-100%-fstime-performance
unixbench/300s-100%-fsbuffer-performance
unixbench/300s-100%-fsdisk-performance
unixbench/300s-1-execl-performance
unixbench/300s-100%-execl-performance
* will-it-scale/brk1-performance
365004 360387 will-it-scale.per_thread_ops
* will-it-scale/dup1-performance
432401 437596 will-it-scale.per_thread_ops
will-it-scale/eventfd1-performance
will-it-scale/futex1-performance
will-it-scale/futex2-performance
will-it-scale/futex3-performance
will-it-scale/futex4-performance
will-it-scale/getppid1-performance
will-it-scale/lock1-performance
will-it-scale/lseek1-performance
will-it-scale/lseek2-performance
* will-it-scale/malloc1-performance
47025 45817 will-it-scale.per_thread_ops
77499 76529 will-it-scale.per_process_ops
will-it-scale/malloc2-performance
* will-it-scale/mmap1-performance
123399 120815 will-it-scale.per_thread_ops
152219 149833 will-it-scale.per_process_ops
* will-it-scale/mmap2-performance
107327 104714 will-it-scale.per_thread_ops
136405 133765 will-it-scale.per_process_ops
will-it-scale/open1-performance
* will-it-scale/open2-performance
171570 168805 will-it-scale.per_thread_ops
532644 526202 will-it-scale.per_process_ops
will-it-scale/page_fault1-performance
will-it-scale/page_fault2-performance
will-it-scale/page_fault3-performance
will-it-scale/pipe1-performance
will-it-scale/poll1-performance
* will-it-scale/poll2-performance
176134 172848 will-it-scale.per_thread_ops
281361 275053 will-it-scale.per_process_ops
will-it-scale/posix_semaphore1-performance
will-it-scale/pread1-performance
will-it-scale/pread2-performance
will-it-scale/pread3-performance
will-it-scale/pthread_mutex1-performance
will-it-scale/pthread_mutex2-performance
will-it-scale/pwrite1-performance
will-it-scale/pwrite2-performance
will-it-scale/pwrite3-performance
* will-it-scale/read1-performance
1190563 1174833 will-it-scale.per_thread_ops
* will-it-scale/read2-performance
1105369 1080427 will-it-scale.per_thread_ops
will-it-scale/readseek1-performance
* will-it-scale/readseek2-performance
261818 259040 will-it-scale.per_thread_ops
will-it-scale/readseek3-performance
* will-it-scale/sched_yield-performance
2408059 2382034 will-it-scale.per_thread_ops
will-it-scale/signal1-performance
will-it-scale/unix1-performance
will-it-scale/unlink1-performance
will-it-scale/unlink2-performance
* will-it-scale/write1-performance
976701 961588 will-it-scale.per_thread_ops
* will-it-scale/writeseek1-performance
831898 822448 will-it-scale.per_thread_ops
* will-it-scale/writeseek2-performance
228248 225065 will-it-scale.per_thread_ops
* will-it-scale/writeseek3-performance
226670 224058 will-it-scale.per_thread_ops
will-it-scale/context_switch1-performance
aim7/performance-fork_test-2000
* aim7/performance-brk_test-3000
74869 76676 aim7.jobs-per-min
aim7/performance-disk_cp-3000
aim7/performance-disk_rd-3000
aim7/performance-sieve-3000
aim7/performance-page_test-3000
aim7/performance-creat-clo-3000
aim7/performance-mem_rtns_1-8000
aim7/performance-disk_wrt-8000
aim7/performance-pipe_cpy-8000
aim7/performance-ram_copy-8000
(d) lkp-avoton3:
8 threads Intel(R) Atom(TM) CPU C2750 @ 2.40GHz 16G
netperf/ipv4-900s-200%-cs-localhost-TCP_STREAM-performance
Signed-off-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Li Zhijian <zhijianx.li@intel.com>
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/20180809135753.21077-1-dietmar.eggemann@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
