Don't instrument 3 more files that contain debugging facilities and
produce large amounts of uninteresting coverage for every syscall.
The following snippets are sprinkled all over the place in kcov traces
in a debugging kernel. We already try to disable instrumentation of
stack unwinding code and of most debug facilities. I guess we did not
use fault-inject.c at the time, and stacktrace.c was somehow missed (or
something has changed in kernel/configs). This change both speeds up
kcov (kernel doesn't need to store these PCs, user-space doesn't need to
process them) and frees trace buffer capacity for more useful coverage.
should_fail
lib/fault-inject.c:149
fail_dump
lib/fault-inject.c:45
stack_trace_save
kernel/stacktrace.c:124
stack_trace_consume_entry
kernel/stacktrace.c:86
stack_trace_consume_entry
kernel/stacktrace.c:89
... a hundred frames skipped ...
stack_trace_consume_entry
kernel/stacktrace.c:93
stack_trace_consume_entry
kernel/stacktrace.c:86
Link: http://lkml.kernel.org/r/20200116111449.217744-1-dvyukov@gmail.com
Signed-off-by: Dmitry Vyukov <dvyukov@google.com>
Reviewed-by: Andrey Konovalov <andreyknvl@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The patch 'tracing: Fix histogram code when expression has same var as
value' added code to return an existing variable reference when
creating a new variable reference, which resulted in var_ref_vals
slots being reused instead of being duplicated.
The implementation of the trace action assumes that the end of the
var_ref_vals array starting at action_data.var_ref_idx corresponds to
the values that will be assigned to the trace params. The patch
mentioned above invalidates that assumption, which means that each
param needs to explicitly specify its index into var_ref_vals.
This fix changes action_data.var_ref_idx to an array of var ref
indexes to account for that.
Link: https://lore.kernel.org/r/1580335695.6220.8.camel@kernel.org
Fixes: 8bcebc77e8 ("tracing: Fix histogram code when expression has same var as value")
Signed-off-by: Tom Zanussi <zanussi@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
If you switch to a sleeping task with the "pid" command and then type
"rd", kdb tells you this:
No current kdb registers. You may need to select another task
diag: -17: Invalid register name
The first message makes sense, but not the second. Fix it by just
returning 0 after commands accessing the current registers finish if
we've already printed the "No current kdb registers" error.
While fixing kdb_rd(), change the function to use "if" rather than
"ifdef". It cleans the function up a bit and any modern compiler will
have no trouble handling still producing good code.
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Link: https://lore.kernel.org/r/20191109111624.5.I121f4c6f0c19266200bf6ef003de78841e5bfc3d@changeid
Signed-off-by: Daniel Thompson <daniel.thompson@linaro.org>
Some (but not all?) of the kdb backtrace paths would cause the
kdb_current_task and kdb_current_regs to remain changed. As discussed
in a review of a previous patch [1], this doesn't seem intuitive, so
let's fix that.
...but, it turns out that there's actually no longer any reason to set
the current task / current regs while backtracing anymore anyway. As
of commit 2277b49258 ("kdb: Fix stack crawling on 'running' CPUs
that aren't the master") if we're backtracing on a task running on a
CPU we ask that CPU to do the backtrace itself. Linux can do that
without anything fancy. If we're doing backtrace on a sleeping task
we can also do that fine without updating globals. So this patch
mostly just turns into deleting a bunch of code.
[1] https://lore.kernel.org/r/20191010150735.dhrj3pbjgmjrdpwr@holly.lan
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Link: https://lore.kernel.org/r/20191109111624.4.Ibc3d982bbeb9e46872d43973ba808cd4c79537c7@changeid
Signed-off-by: Daniel Thompson <daniel.thompson@linaro.org>
5153faac18 ("cgroup: remove cgroup_enable_task_cg_lists()
optimization") removed lazy initialization of css_sets so that new
tasks are always lniked to its css_set. In the process, it incorrectly
ended up adding init_tasks to root css_set. They show up as PID 0's in
root's cgroup.procs triggering warnings in systemd and generally
confusing people.
Fix it by skip css_set linking for init_tasks.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: https://github.com/joanbm
Link: https://github.com/systemd/systemd/issues/14682
Fixes: 5153faac18 ("cgroup: remove cgroup_enable_task_cg_lists() optimization")
Cc: stable@vger.kernel.org # v5.5+
Move all the tracing selftest configs to the bottom of the tracing menu.
There's no reason for them to be interspersed throughout.
Also, move the bootconfig menu to the top.
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Move the config that enables the mmiotracer with the other tracers such that
all the tracers are together.
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
The MMIO test module was by itself, move it to the other test modules. Also,
add the text "Test module" to PREEMPTIRQ_DELAY_TEST as that create a test
module as well.
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
The features that depend on the function tracer were spread out through the
tracing menu, pull them together as it is easier to manage.
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Add functions used to generate kprobe event commands, built on top of
the dynevent_cmd interface.
kprobe_event_gen_cmd_start() is used to create a kprobe event command
using a variable arg list, and kretprobe_event_gen_cmd_start() does
the same for kretprobe event commands. kprobe_event_add_fields() can
be used to add single fields one by one or as a group. Once all
desired fields are added, kprobe_event_gen_cmd_end() or
kretprobe_event_gen_cmd_end() respectively are used to actually
execute the command and create the event.
Link: http://lkml.kernel.org/r/95cc4696502bb6017f9126f306a45ad19b4cc14f.1580323897.git.zanussi@kernel.org
Acked-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Tom Zanussi <zanussi@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Add an exported function named synth_event_trace(), allowing modules
or other kernel code to trace synthetic events.
Also added are several functions that allow the same functionality to
be broken out in a piecewise fashion, which are useful in situations
where tracing an event from a full array of values would be
cumbersome. Those functions are synth_event_trace_start/end() and
synth_event_add_(next)_val().
Link: http://lkml.kernel.org/r/7a84de5f1854acf4144b57efe835ca645afa764f.1580323897.git.zanussi@kernel.org
Acked-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Tom Zanussi <zanussi@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Add functions used to generate synthetic event commands, built on top
of the dynevent_cmd interface.
synth_event_gen_cmd_start() is used to create a synthetic event
command using a variable arg list and
synth_event_gen_cmd_array_start() does the same thing but using an
array of field descriptors. synth_event_add_field(),
synth_event_add_field_str() and synth_event_add_fields() can be used
to add single fields one by one or as a group. Once all desired
fields are added, synth_event_gen_cmd_end() is used to actually
execute the command and create the event.
synth_event_create() does everything, including creating the event, in
a single call.
Link: http://lkml.kernel.org/r/38fef702fad5ef208009f459552f34a94befd860.1580323897.git.zanussi@kernel.org
Acked-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Tom Zanussi <zanussi@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Add an interface used to build up dynamic event creation commands,
such as synthetic and kprobe events. Interfaces specific to those
particular types of events and others can be built on top of this
interface.
Command creation is started by first using the dynevent_cmd_init()
function to initialize the dynevent_cmd object. Following that, args
are appended and optionally checked by the dynevent_arg_add() and
dynevent_arg_pair_add() functions, which use objects representing
arguments and pairs of arguments, initialized respectively by
dynevent_arg_init() and dynevent_arg_pair_init(). Finally, once all
args have been successfully added, the command is finalized and
actually created using dynevent_create().
The code here for actually printing into the dyn_event->cmd buffer
using snprintf() etc was adapted from v4 of Masami's 'tracing/boot:
Add synthetic event support' patch.
Link: http://lkml.kernel.org/r/1f65fa44390b6f238f6036777c3784ced1dcc6a0.1580323897.git.zanussi@kernel.org
Signed-off-by: Tom Zanussi <zanussi@kernel.org>
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Add a function to get an event file and prevent it from going away on
module or instance removal.
trace_get_event_file() will find an event file in a given instance (if
instance is NULL, it assumes the top trace array) and return it,
pinning the instance's trace array as well as the event's module, if
applicable, so they won't go away while in use.
trace_put_event_file() does the matching release.
Link: http://lkml.kernel.org/r/bb31ac4bdda168d5ed3c4b5f5a4c8f633e8d9118.1580323897.git.zanussi@kernel.org
Acked-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Tom Zanussi <zanussi@kernel.org>
[ Moved trace_array_put() to end of trace_put_event_file() ]
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
if seq_file .next fuction does not change position index,
read after some lseek can generate unexpected output.
Without patch:
# dd bs=30 skip=1 if=/sys/kernel/tracing/events/sched/sched_switch/trigger
dd: /sys/kernel/tracing/events/sched/sched_switch/trigger: cannot skip to specified offset
n traceoff snapshot stacktrace enable_event disable_event enable_hist disable_hist hist
# Available triggers:
# traceon traceoff snapshot stacktrace enable_event disable_event enable_hist disable_hist hist
6+1 records in
6+1 records out
206 bytes copied, 0.00027916 s, 738 kB/s
Notice the printing of "# Available triggers:..." after the line.
With the patch:
# dd bs=30 skip=1 if=/sys/kernel/tracing/events/sched/sched_switch/trigger
dd: /sys/kernel/tracing/events/sched/sched_switch/trigger: cannot skip to specified offset
n traceoff snapshot stacktrace enable_event disable_event enable_hist disable_hist hist
2+1 records in
2+1 records out
88 bytes copied, 0.000526867 s, 167 kB/s
It only prints the end of the file, and does not restart.
Link: http://lkml.kernel.org/r/3c35ee24-dd3a-8119-9c19-552ed253388a@virtuozzo.comhttps://bugzilla.kernel.org/show_bug.cgi?id=206283
Signed-off-by: Vasily Averin <vvs@virtuozzo.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
if seq_file .next fuction does not change position index,
read after some lseek can generate unexpected output.
Link: http://lkml.kernel.org/r/7ad85b22-1866-977c-db17-88ac438bc764@virtuozzo.com
Signed-off-by: Vasily Averin <vvs@virtuozzo.com>
[ This is not a bug fix, it just makes it "technically correct"
which is why I applied it. NULL is only returned on an anomaly
which triggers a WARN_ON ]
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
if seq_file .next fuction does not change position index,
read after some lseek can generate unexpected output.
Without patch:
# dd bs=4 skip=1 if=/sys/kernel/tracing/set_ftrace_pid
dd: /sys/kernel/tracing/set_ftrace_pid: cannot skip to specified offset
id
no pid
2+1 records in
2+1 records out
10 bytes copied, 0.000213285 s, 46.9 kB/s
Notice the "id" followed by "no pid".
With the patch:
# dd bs=4 skip=1 if=/sys/kernel/tracing/set_ftrace_pid
dd: /sys/kernel/tracing/set_ftrace_pid: cannot skip to specified offset
id
0+1 records in
0+1 records out
3 bytes copied, 0.000202112 s, 14.8 kB/s
Notice that it only prints "id" and not the "no pid" afterward.
Link: http://lkml.kernel.org/r/4f87c6ad-f114-30bb-8506-c32274ce2992@virtuozzo.comhttps://bugzilla.kernel.org/show_bug.cgi?id=206283
Signed-off-by: Vasily Averin <vvs@virtuozzo.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Although the device tree might contain a reserved-memory DT node
dedicated as the default CMA pool, users might want to change CMA's
parameters using the kernel command line for debugging purposes and
whatnot. Honor this by bypassing the reserved memory CMA setup, which
will ultimately end up freeing the memblock and allow the command line
CMA configuration routine to run.
Signed-off-by: Nicolas Saenz Julienne <nsaenzjulienne@suse.de>
Reviewed-by: Phil Elwell <phil@raspberrypi.org>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Pull mmu_notifier updates from Jason Gunthorpe:
"This small series revises the names in mmu_notifier to make the code
clearer and more readable"
* tag 'for-linus-hmm' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma:
mm/mmu_notifiers: Use 'interval_sub' as the variable for mmu_interval_notifier
mm/mmu_notifiers: Use 'subscription' as the variable name for mmu_notifier
mm/mmu_notifier: Rename struct mmu_notifier_mm to mmu_notifier_subscriptions
Pull thread management updates from Christian Brauner:
"Sargun Dhillon over the last cycle has worked on the pidfd_getfd()
syscall.
This syscall allows for the retrieval of file descriptors of a process
based on its pidfd. A task needs to have ptrace_may_access()
permissions with PTRACE_MODE_ATTACH_REALCREDS (suggested by Oleg and
Andy) on the target.
One of the main use-cases is in combination with seccomp's user
notification feature. As a reminder, seccomp's user notification
feature was made available in v5.0. It allows a task to retrieve a
file descriptor for its seccomp filter. The file descriptor is usually
handed of to a more privileged supervising process. The supervisor can
then listen for syscall events caught by the seccomp filter of the
supervisee and perform actions in lieu of the supervisee, usually
emulating syscalls. pidfd_getfd() is needed to expand its uses.
There are currently two major users that wait on pidfd_getfd() and one
future user:
- Netflix, Sargun said, is working on a service mesh where users
should be able to connect to a dns-based VIP. When a user connects
to e.g. 1.2.3.4:80 that runs e.g. service "foo" they will be
redirected to an envoy process. This service mesh uses seccomp user
notifications and pidfd to intercept all connect calls and instead
of connecting them to 1.2.3.4:80 connects them to e.g.
127.0.0.1:8080.
- LXD uses the seccomp notifier heavily to intercept and emulate
mknod() and mount() syscalls for unprivileged containers/processes.
With pidfd_getfd() more uses-cases e.g. bridging socket connections
will be possible.
- The patchset has also seen some interest from the browser corner.
Right now, Firefox is using a SECCOMP_RET_TRAP sandbox managed by a
broker process. In the future glibc will start blocking all signals
during dlopen() rendering this type of sandbox impossible. Hence,
in the future Firefox will switch to a seccomp-user-nofication
based sandbox which also makes use of file descriptor retrieval.
The thread for this can be found at
https://sourceware.org/ml/libc-alpha/2019-12/msg00079.html
With pidfd_getfd() it is e.g. possible to bridge socket connections
for the supervisee (binding to a privileged port) and taking actions
on file descriptors on behalf of the supervisee in general.
Sargun's first version was using an ioctl on pidfds but various people
pushed for it to be a proper syscall which he duely implemented as
well over various review cycles. Selftests are of course included.
I've also added instructions how to deal with merge conflicts below.
There's also a small fix coming from the kernel mentee project to
correctly annotate struct sighand_struct with __rcu to fix various
sparse warnings. We've received a few more such fixes and even though
they are mostly trivial I've decided to postpone them until after -rc1
since they came in rather late and I don't want to risk introducing
build warnings.
Finally, there's a new prctl() command PR_{G,S}ET_IO_FLUSHER which is
needed to avoid allocation recursions triggerable by storage drivers
that have userspace parts that run in the IO path (e.g. dm-multipath,
iscsi, etc). These allocation recursions deadlock the device.
The new prctl() allows such privileged userspace components to avoid
allocation recursions by setting the PF_MEMALLOC_NOIO and
PF_LESS_THROTTLE flags. The patch carries the necessary acks from the
relevant maintainers and is routed here as part of prctl()
thread-management."
* tag 'threads-v5.6' of git://git.kernel.org/pub/scm/linux/kernel/git/brauner/linux:
prctl: PR_{G,S}ET_IO_FLUSHER to support controlling memory reclaim
sched.h: Annotate sighand_struct with __rcu
test: Add test for pidfd getfd
arch: wire up pidfd_getfd syscall
pid: Implement pidfd_getfd syscall
vfs, fdtable: Add fget_task helper
Pull Kselftest kunit updates from Shuah Khan:
"This kunit update consists of:
- Support for building kunit as a module from Alan Maguire
- AppArmor KUnit tests for policy unpack from Mike Salvatore"
* tag 'linux-kselftest-5.6-rc1-kunit' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest:
kunit: building kunit as a module breaks allmodconfig
kunit: update documentation to describe module-based build
kunit: allow kunit to be loaded as a module
kunit: remove timeout dependence on sysctl_hung_task_timeout_seconds
kunit: allow kunit tests to be loaded as a module
kunit: hide unexported try-catch interface in try-catch-impl.h
kunit: move string-stream.h to lib/kunit
apparmor: add AppArmor KUnit tests for policy unpack
Pull y2038 updates from Arnd Bergmann:
"Core, driver and file system changes
These are updates to device drivers and file systems that for some
reason or another were not included in the kernel in the previous
y2038 series.
I've gone through all users of time_t again to make sure the kernel is
in a long-term maintainable state, replacing all remaining references
to time_t with safe alternatives.
Some related parts of the series were picked up into the nfsd, xfs,
alsa and v4l2 trees. A final set of patches in linux-mm removes the
now unused time_t/timeval/timespec types and helper functions after
all five branches are merged for linux-5.6, ensuring that no new users
get merged.
As a result, linux-5.6, or my backport of the patches to 5.4 [1],
should be the first release that can serve as a base for a 32-bit
system designed to run beyond year 2038, with a few remaining caveats:
- All user space must be compiled with a 64-bit time_t, which will be
supported in the coming musl-1.2 and glibc-2.32 releases, along
with installed kernel headers from linux-5.6 or higher.
- Applications that use the system call interfaces directly need to
be ported to use the time64 syscalls added in linux-5.1 in place of
the existing system calls. This impacts most users of futex() and
seccomp() as well as programming languages that have their own
runtime environment not based on libc.
- Applications that use a private copy of kernel uapi header files or
their contents may need to update to the linux-5.6 version, in
particular for sound/asound.h, xfs/xfs_fs.h, linux/input.h,
linux/elfcore.h, linux/sockios.h, linux/timex.h and
linux/can/bcm.h.
- A few remaining interfaces cannot be changed to pass a 64-bit
time_t in a compatible way, so they must be configured to use
CLOCK_MONOTONIC times or (with a y2106 problem) unsigned 32-bit
timestamps. Most importantly this impacts all users of 'struct
input_event'.
- All y2038 problems that are present on 64-bit machines also apply
to 32-bit machines. In particular this affects file systems with
on-disk timestamps using signed 32-bit seconds: ext4 with
ext3-style small inodes, ext2, xfs (to be fixed soon) and ufs"
[1] https://git.kernel.org/pub/scm/linux/kernel/git/arnd/playground.git/log/?h=y2038-endgame
* tag 'y2038-drivers-for-v5.6-signed' of git://git.kernel.org:/pub/scm/linux/kernel/git/arnd/playground: (21 commits)
Revert "drm/etnaviv: reject timeouts with tv_nsec >= NSEC_PER_SEC"
y2038: sh: remove timeval/timespec usage from headers
y2038: sparc: remove use of struct timex
y2038: rename itimerval to __kernel_old_itimerval
y2038: remove obsolete jiffies conversion functions
nfs: fscache: use timespec64 in inode auxdata
nfs: fix timstamp debug prints
nfs: use time64_t internally
sunrpc: convert to time64_t for expiry
drm/etnaviv: avoid deprecated timespec
drm/etnaviv: reject timeouts with tv_nsec >= NSEC_PER_SEC
drm/msm: avoid using 'timespec'
hfs/hfsplus: use 64-bit inode timestamps
hostfs: pass 64-bit timestamps to/from user space
packet: clarify timestamp overflow
tsacct: add 64-bit btime field
acct: stop using get_seconds()
um: ubd: use 64-bit time_t where possible
xtensa: ISS: avoid struct timeval
dlm: use SO_SNDTIMEO_NEW instead of SO_SNDTIMEO_OLD
...
Pull printk update from Petr Mladek:
"Prevent replaying log on all consoles"
* tag 'printk-for-5.6' of git://git.kernel.org/pub/scm/linux/kernel/git/pmladek/printk:
printk: fix exclusive_console replaying
Pull openat2 support from Al Viro:
"This is the openat2() series from Aleksa Sarai.
I'm afraid that the rest of namei stuff will have to wait - it got
zero review the last time I'd posted #work.namei, and there had been a
leak in the posted series I'd caught only last weekend. I was going to
repost it on Monday, but the window opened and the odds of getting any
review during that... Oh, well.
Anyway, openat2 part should be ready; that _did_ get sane amount of
review and public testing, so here it comes"
From Aleksa's description of the series:
"For a very long time, extending openat(2) with new features has been
incredibly frustrating. This stems from the fact that openat(2) is
possibly the most famous counter-example to the mantra "don't silently
accept garbage from userspace" -- it doesn't check whether unknown
flags are present[1].
This means that (generally) the addition of new flags to openat(2) has
been fraught with backwards-compatibility issues (O_TMPFILE has to be
defined as __O_TMPFILE|O_DIRECTORY|[O_RDWR or O_WRONLY] to ensure old
kernels gave errors, since it's insecure to silently ignore the
flag[2]). All new security-related flags therefore have a tough road
to being added to openat(2).
Furthermore, the need for some sort of control over VFS's path
resolution (to avoid malicious paths resulting in inadvertent
breakouts) has been a very long-standing desire of many userspace
applications.
This patchset is a revival of Al Viro's old AT_NO_JUMPS[3] patchset
(which was a variant of David Drysdale's O_BENEATH patchset[4] which
was a spin-off of the Capsicum project[5]) with a few additions and
changes made based on the previous discussion within [6] as well as
others I felt were useful.
In line with the conclusions of the original discussion of
AT_NO_JUMPS, the flag has been split up into separate flags. However,
instead of being an openat(2) flag it is provided through a new
syscall openat2(2) which provides several other improvements to the
openat(2) interface (see the patch description for more details). The
following new LOOKUP_* flags are added:
LOOKUP_NO_XDEV:
Blocks all mountpoint crossings (upwards, downwards, or through
absolute links). Absolute pathnames alone in openat(2) do not
trigger this. Magic-link traversal which implies a vfsmount jump is
also blocked (though magic-link jumps on the same vfsmount are
permitted).
LOOKUP_NO_MAGICLINKS:
Blocks resolution through /proc/$pid/fd-style links. This is done
by blocking the usage of nd_jump_link() during resolution in a
filesystem. The term "magic-links" is used to match with the only
reference to these links in Documentation/, but I'm happy to change
the name.
It should be noted that this is different to the scope of
~LOOKUP_FOLLOW in that it applies to all path components. However,
you can do openat2(NO_FOLLOW|NO_MAGICLINKS) on a magic-link and it
will *not* fail (assuming that no parent component was a
magic-link), and you will have an fd for the magic-link.
In order to correctly detect magic-links, the introduction of a new
LOOKUP_MAGICLINK_JUMPED state flag was required.
LOOKUP_BENEATH:
Disallows escapes to outside the starting dirfd's
tree, using techniques such as ".." or absolute links. Absolute
paths in openat(2) are also disallowed.
Conceptually this flag is to ensure you "stay below" a certain
point in the filesystem tree -- but this requires some additional
to protect against various races that would allow escape using
"..".
Currently LOOKUP_BENEATH implies LOOKUP_NO_MAGICLINKS, because it
can trivially beam you around the filesystem (breaking the
protection). In future, there might be similar safety checks done
as in LOOKUP_IN_ROOT, but that requires more discussion.
In addition, two new flags are added that expand on the above ideas:
LOOKUP_NO_SYMLINKS:
Does what it says on the tin. No symlink resolution is allowed at
all, including magic-links. Just as with LOOKUP_NO_MAGICLINKS this
can still be used with NOFOLLOW to open an fd for the symlink as
long as no parent path had a symlink component.
LOOKUP_IN_ROOT:
This is an extension of LOOKUP_BENEATH that, rather than blocking
attempts to move past the root, forces all such movements to be
scoped to the starting point. This provides chroot(2)-like
protection but without the cost of a chroot(2) for each filesystem
operation, as well as being safe against race attacks that
chroot(2) is not.
If a race is detected (as with LOOKUP_BENEATH) then an error is
generated, and similar to LOOKUP_BENEATH it is not permitted to
cross magic-links with LOOKUP_IN_ROOT.
The primary need for this is from container runtimes, which
currently need to do symlink scoping in userspace[7] when opening
paths in a potentially malicious container.
There is a long list of CVEs that could have bene mitigated by
having RESOLVE_THIS_ROOT (such as CVE-2017-1002101,
CVE-2017-1002102, CVE-2018-15664, and CVE-2019-5736, just to name a
few).
In order to make all of the above more usable, I'm working on
libpathrs[8] which is a C-friendly library for safe path resolution.
It features a userspace-emulated backend if the kernel doesn't support
openat2(2). Hopefully we can get userspace to switch to using it, and
thus get openat2(2) support for free once it's ready.
Future work would include implementing things like
RESOLVE_NO_AUTOMOUNT and possibly a RESOLVE_NO_REMOTE (to allow
programs to be sure they don't hit DoSes though stale NFS handles)"
* 'work.openat2' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
Documentation: path-lookup: include new LOOKUP flags
selftests: add openat2(2) selftests
open: introduce openat2(2) syscall
namei: LOOKUP_{IN_ROOT,BENEATH}: permit limited ".." resolution
namei: LOOKUP_IN_ROOT: chroot-like scoped resolution
namei: LOOKUP_BENEATH: O_BENEATH-like scoped resolution
namei: LOOKUP_NO_XDEV: block mountpoint crossing
namei: LOOKUP_NO_MAGICLINKS: block magic-link resolution
namei: LOOKUP_NO_SYMLINKS: block symlink resolution
namei: allow set_root() to produce errors
namei: allow nd_jump_link() to produce errors
nsfs: clean-up ns_get_path() signature to return int
namei: only return -ECHILD from follow_dotdot_rcu()
Pull RCU warning removal from Paul McKenney:
"A single commit that fixes an embarrassing bug discussed here:
https://lore.kernel.org/lkml/20200125131425.GB16136@zn.tnic/
which apparently also affects smaller systems"
[ This was sent to Ingo, but since I see the issue on the laptop I use for
testing during the merge window, I'm doing the pull directly - Linus ]
* 'urgent-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-rcu:
rcu: Forgive slow expedited grace periods at boot time
Instead of using a locally defined "struct bpf_verifier_log log = {}",
btf_struct_ops_init() should reuse the "log" from its calling
function "btf_parse_vmlinux()". It should also resolve the
frame-size too large compiler warning in some ARCH.
Fixes: 27ae7997a6 ("bpf: Introduce BPF_PROG_TYPE_STRUCT_OPS")
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20200127175145.1154438-1-kafai@fb.com
In the process of adding better error messages for sorting, I realized
that strsep was being used incorrectly and some of the error paths I
was expecting to be hit weren't and just fell through to the common
invalid key error case.
It also became obvious that for keyword assignments, it wasn't
necessary to save the full assignment and reparse it later, and having
a common empty-assignment check would also make more sense in terms of
error processing.
Change the code to fix these problems and simplify it for new error
message changes in a subsequent patch.
Link: http://lkml.kernel.org/r/1c3ef0b6655deaf345f6faee2584a0298ac2d743.1561743018.git.zanussi@kernel.org
Fixes: e62347d245 ("tracing: Add hist trigger support for user-defined sorting ('sort=' param)")
Fixes: 7ef224d1d0 ("tracing: Add 'hist' event trigger command")
Fixes: a4072fe85b ("tracing: Add a clock attribute for hist triggers")
Reported-by: Masami Hiramatsu <mhiramat@kernel.org>
Reviewed-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Tom Zanussi <zanussi@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Pull UML updates from Anton Ivanov:
"I am sending this on behalf of Richard who is traveling.
This contains the following changes for UML:
- Fix for time travel mode
- Disable CONFIG_CONSTRUCTORS again
- A new command line option to have an non-raw serial line
- Preparations to remove obsolete UML network drivers"
* tag 'for-linus-5.6-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rw/uml:
um: Fix time-travel=inf-cpu with xor/raid6
Revert "um: Enable CONFIG_CONSTRUCTORS"
um: Mark non-vector net transports as obsolete
um: Add an option to make serial driver non-raw
Pull tracing fix from Steven Rostedt:
"Kprobe events added 'ustring' to distinguish reading strings from
kernel space or user space.
But the creating of the event format file only checks for 'string' to
display string formats. 'ustring' must also be handled"
* tag 'trace-v5.5-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
tracing/kprobes: Have uname use __get_str() in print_fmt
Pull networking updates from David Miller:
1) Add WireGuard
2) Add HE and TWT support to ath11k driver, from John Crispin.
3) Add ESP in TCP encapsulation support, from Sabrina Dubroca.
4) Add variable window congestion control to TIPC, from Jon Maloy.
5) Add BCM84881 PHY driver, from Russell King.
6) Start adding netlink support for ethtool operations, from Michal
Kubecek.
7) Add XDP drop and TX action support to ena driver, from Sameeh
Jubran.
8) Add new ipv4 route notifications so that mlxsw driver does not have
to handle identical routes itself. From Ido Schimmel.
9) Add BPF dynamic program extensions, from Alexei Starovoitov.
10) Support RX and TX timestamping in igc, from Vinicius Costa Gomes.
11) Add support for macsec HW offloading, from Antoine Tenart.
12) Add initial support for MPTCP protocol, from Christoph Paasch,
Matthieu Baerts, Florian Westphal, Peter Krystad, and many others.
13) Add Octeontx2 PF support, from Sunil Goutham, Geetha sowjanya, Linu
Cherian, and others.
* git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next: (1469 commits)
net: phy: add default ARCH_BCM_IPROC for MDIO_BCM_IPROC
udp: segment looped gso packets correctly
netem: change mailing list
qed: FW 8.42.2.0 debug features
qed: rt init valid initialization changed
qed: Debug feature: ilt and mdump
qed: FW 8.42.2.0 Add fw overlay feature
qed: FW 8.42.2.0 HSI changes
qed: FW 8.42.2.0 iscsi/fcoe changes
qed: Add abstraction for different hsi values per chip
qed: FW 8.42.2.0 Additional ll2 type
qed: Use dmae to write to widebus registers in fw_funcs
qed: FW 8.42.2.0 Parser offsets modified
qed: FW 8.42.2.0 Queue Manager changes
qed: FW 8.42.2.0 Expose new registers and change windows
qed: FW 8.42.2.0 Internal ram offsets modifications
MAINTAINERS: Add entry for Marvell OcteonTX2 Physical Function driver
Documentation: net: octeontx2: Add RVU HW and drivers overview
octeontx2-pf: ethtool RSS config support
octeontx2-pf: Add basic ethtool support
...
Pull crypto updates from Herbert Xu:
"API:
- Removed CRYPTO_TFM_RES flags
- Extended spawn grabbing to all algorithm types
- Moved hash descsize verification into API code
Algorithms:
- Fixed recursive pcrypt dead-lock
- Added new 32 and 64-bit generic versions of poly1305
- Added cryptogams implementation of x86/poly1305
Drivers:
- Added support for i.MX8M Mini in caam
- Added support for i.MX8M Nano in caam
- Added support for i.MX8M Plus in caam
- Added support for A33 variant of SS in sun4i-ss
- Added TEE support for Raven Ridge in ccp
- Added in-kernel API to submit TEE commands in ccp
- Added AMD-TEE driver
- Added support for BCM2711 in iproc-rng200
- Added support for AES256-GCM based ciphers for chtls
- Added aead support on SEC2 in hisilicon"
* 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6: (244 commits)
crypto: arm/chacha - fix build failured when kernel mode NEON is disabled
crypto: caam - add support for i.MX8M Plus
crypto: x86/poly1305 - emit does base conversion itself
crypto: hisilicon - fix spelling mistake "disgest" -> "digest"
crypto: chacha20poly1305 - add back missing test vectors and test chunking
crypto: x86/poly1305 - fix .gitignore typo
tee: fix memory allocation failure checks on drv_data and amdtee
crypto: ccree - erase unneeded inline funcs
crypto: ccree - make cc_pm_put_suspend() void
crypto: ccree - split overloaded usage of irq field
crypto: ccree - fix PM race condition
crypto: ccree - fix FDE descriptor sequence
crypto: ccree - cc_do_send_request() is void func
crypto: ccree - fix pm wrongful error reporting
crypto: ccree - turn errors to debug msgs
crypto: ccree - fix AEAD decrypt auth fail
crypto: ccree - fix typo in comment
crypto: ccree - fix typos in error msgs
crypto: atmel-{aes,sha,tdes} - Retire crypto_platform_data
crypto: x86/sha - Eliminate casts on asm implementations
...
Group RT scheduler contains protection against setting zero runtime for
cgroup with RT tasks. Right now function tg_set_rt_bandwidth() iterates
over all CPU cgroups and calls tg_has_rt_tasks() for any cgroup which
runtime is zero (not only for changed one). Default RT runtime is zero,
thus tg_has_rt_tasks() will is called for almost at CPU cgroups.
This protection already is slightly racy: runtime limit could be changed
between cpu_cgroup_can_attach() and cpu_cgroup_attach() because changing
cgroup attribute does not lock cgroup_mutex while attach does not lock
rt_constraints_mutex. Changing task scheduler class also races with
changing rt runtime: check in __sched_setscheduler() isn't protected.
Function tg_has_rt_tasks() iterates over all threads in the system.
This gives NR_CGROUPS * NR_TASKS operations under single tasklist_lock
locked for read tg_set_rt_bandwidth(). Any concurrent attempt of locking
tasklist_lock for write (for example fork) will stuck with disabled irqs.
This patch makes two optimizations:
1) Remove locking tasklist_lock and iterate only tasks in cgroup
2) Call tg_has_rt_tasks() iff rt runtime changes from non-zero to zero
All changed code is under CONFIG_RT_GROUP_SCHED.
Testcase:
# mkdir /sys/fs/cgroup/cpu/test{1..10000}
# echo 0 | tee /sys/fs/cgroup/cpu/test*/cpu.rt_runtime_us
At the same time without patch fork time will be >100ms:
# perf trace -e clone --duration 100 stress-ng --fork 1
Also remote ping will show timings >100ms caused by irq latency.
Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lkml.kernel.org/r/157996383820.4651.11292439232549211693.stgit@buzz
Currently we loop through all threads of a core to evaluate if the core is
idle or not. This is unnecessary. If a thread of a core is not idle, skip
evaluating other threads of a core. Also while clearing the cpumask, bits
of all CPUs of a core can be cleared in one-shot.
Collecting ticks on a Power 9 SMT 8 system around select_idle_core
while running schbench shows us
(units are in ticks, hence lesser is better)
Without patch
N Min Max Median Avg Stddev
x 130 151 1083 284 322.72308 144.41494
With patch
N Min Max Median Avg Stddev Improvement
x 164 88 610 201 225.79268 106.78943 30.03%
Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Valentin Schneider <valentin.schneider@arm.com>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Link: https://lkml.kernel.org/r/20191206172422.6578-1-srikar@linux.vnet.ibm.com
Implement arch_scale_freq_capacity() for 'modern' x86. This function
is used by the scheduler to correctly account usage in the face of
DVFS.
The present patch addresses Intel processors specifically and has positive
performance and performance-per-watt implications for the schedutil cpufreq
governor, bringing it closer to, if not on-par with, the powersave governor
from the intel_pstate driver/framework.
Large performance gains are obtained when the machine is lightly loaded and
no regression are observed at saturation. The benchmarks with the largest
gains are kernel compilation, tbench (the networking version of dbench) and
shell-intensive workloads.
1. FREQUENCY INVARIANCE: MOTIVATION
* Without it, a task looks larger if the CPU runs slower
2. PECULIARITIES OF X86
* freq invariance accounting requires knowing the ratio freq_curr/freq_max
2.1 CURRENT FREQUENCY
* Use delta_APERF / delta_MPERF * freq_base (a.k.a "BusyMHz")
2.2 MAX FREQUENCY
* It varies with time (turbo). As an approximation, we set it to a
constant, i.e. 4-cores turbo frequency.
3. EFFECTS ON THE SCHEDUTIL FREQUENCY GOVERNOR
* The invariant schedutil's formula has no feedback loop and reacts faster
to utilization changes
4. KNOWN LIMITATIONS
* In some cases tasks can't reach max util despite how hard they try
5. PERFORMANCE TESTING
5.1 MACHINES
* Skylake, Broadwell, Haswell
5.2 SETUP
* baseline Linux v5.2 w/ non-invariant schedutil. Tested freq_max = 1-2-3-4-8-12
active cores turbo w/ invariant schedutil, and intel_pstate/powersave
5.3 BENCHMARK RESULTS
5.3.1 NEUTRAL BENCHMARKS
* NAS Parallel Benchmark (HPC), hackbench
5.3.2 NON-NEUTRAL BENCHMARKS
* tbench (10-30% better), kernbench (10-15% better),
shell-intensive-scripts (30-50% better)
* no regressions
5.3.3 SELECTION OF DETAILED RESULTS
5.3.4 POWER CONSUMPTION, PERFORMANCE-PER-WATT
* dbench (5% worse on one machine), kernbench (3% worse),
tbench (5-10% better), shell-intensive-scripts (10-40% better)
6. MICROARCH'ES ADDRESSED HERE
* Xeon Core before Scalable Performance processors line (Xeon Gold/Platinum
etc have different MSRs semantic for querying turbo levels)
7. REFERENCES
* MMTests performance testing framework, github.com/gormanm/mmtests
+-------------------------------------------------------------------------+
| 1. FREQUENCY INVARIANCE: MOTIVATION
+-------------------------------------------------------------------------+
For example; suppose a CPU has two frequencies: 500 and 1000 Mhz. When
running a task that would consume 1/3rd of a CPU at 1000 MHz, it would
appear to consume 2/3rd (or 66.6%) when running at 500 MHz, giving the
false impression this CPU is almost at capacity, even though it can go
faster [*]. In a nutshell, without frequency scale-invariance tasks look
larger just because the CPU is running slower.
[*] (footnote: this assumes a linear frequency/performance relation; which
everybody knows to be false, but given realities its the best approximation
we can make.)
+-------------------------------------------------------------------------+
| 2. PECULIARITIES OF X86
+-------------------------------------------------------------------------+
Accounting for frequency changes in PELT signals requires the computation of
the ratio freq_curr / freq_max. On x86 neither of those terms is readily
available.
2.1 CURRENT FREQUENCY
====================
Since modern x86 has hardware control over the actual frequency we run
at (because amongst other things, Turbo-Mode), we cannot simply use
the frequency as requested through cpufreq.
Instead we use the APERF/MPERF MSRs to compute the effective frequency
over the recent past. Also, because reading MSRs is expensive, don't
do so every time we need the value, but amortize the cost by doing it
every tick.
2.2 MAX FREQUENCY
=================
Obtaining freq_max is also non-trivial because at any time the hardware can
provide a frequency boost to a selected subset of cores if the package has
enough power to spare (eg: Turbo Boost). This means that the maximum frequency
available to a given core changes with time.
The approach taken in this change is to arbitrarily set freq_max to a constant
value at boot. The value chosen is the "4-cores (4C) turbo frequency" on most
microarchitectures, after evaluating the following candidates:
* 1-core (1C) turbo frequency (the fastest turbo state available)
* around base frequency (a.k.a. max P-state)
* something in between, such as 4C turbo
To interpret these options, consider that this is the denominator in
freq_curr/freq_max, and that ratio will be used to scale PELT signals such as
util_avg and load_avg. A large denominator will undershoot (util_avg looks a
bit smaller than it really is), viceversa with a smaller denominator PELT
signals will tend to overshoot. Given that PELT drives frequency selection
in the schedutil governor, we will have:
freq_max set to | effect on DVFS
--------------------+------------------
1C turbo | power efficiency (lower freq choices)
base freq | performance (higher util_avg, higher freq requests)
4C turbo | a bit of both
4C turbo proves to be a good compromise in a number of benchmarks (see below).
+-------------------------------------------------------------------------+
| 3. EFFECTS ON THE SCHEDUTIL FREQUENCY GOVERNOR
+-------------------------------------------------------------------------+
Once an architecture implements a frequency scale-invariant utilization (the
PELT signal util_avg), schedutil switches its frequency selection formula from
freq_next = 1.25 * freq_curr * util [non-invariant util signal]
to
freq_next = 1.25 * freq_max * util [invariant util signal]
where, in the second formula, freq_max is set to the 1C turbo frequency (max
turbo). The advantage of the second formula, whose usage we unlock with this
patch, is that freq_next doesn't depend on the current frequency in an
iterative fashion, but can jump to any frequency in a single update. This
absence of feedback in the formula makes it quicker to react to utilization
changes and more robust against pathological instabilities.
Compare it to the update formula of intel_pstate/powersave:
freq_next = 1.25 * freq_max * Busy%
where again freq_max is 1C turbo and Busy% is the percentage of time not spent
idling (calculated with delta_MPERF / delta_TSC); essentially the same as
invariant schedutil, and largely responsible for intel_pstate/powersave good
reputation. The non-invariant schedutil formula is derived from the invariant
one by approximating util_inv with util_raw * freq_curr / freq_max, but this
has limitations.
Testing shows improved performances due to better frequency selections when
the machine is lightly loaded, and essentially no change in behaviour at
saturation / overutilization.
+-------------------------------------------------------------------------+
| 4. KNOWN LIMITATIONS
+-------------------------------------------------------------------------+
It's been shown that it is possible to create pathological scenarios where a
CPU-bound task cannot reach max utilization, if the normalizing factor
freq_max is fixed to a constant value (see [Lelli-2018]).
If freq_max is set to 4C turbo as we do here, one needs to peg at least 5
cores in a package doing some busywork, and observe that none of those task
will ever reach max util (1024) because they're all running at less than the
4C turbo frequency.
While this concern still applies, we believe the performance benefit of
frequency scale-invariant PELT signals outweights the cost of this limitation.
[Lelli-2018]
https://lore.kernel.org/lkml/20180517150418.GF22493@localhost.localdomain/
+-------------------------------------------------------------------------+
| 5. PERFORMANCE TESTING
+-------------------------------------------------------------------------+
5.1 MACHINES
============
We tested the patch on three machines, with Skylake, Broadwell and Haswell
CPUs. The details are below, together with the available turbo ratios as
reported by the appropriate MSRs.
* 8x-SKYLAKE-UMA:
Single socket E3-1240 v5, Skylake 4 cores/8 threads
Max EFFiciency, BASE frequency and available turbo levels (MHz):
EFFIC 800 |********
BASE 3500 |***********************************
4C 3700 |*************************************
3C 3800 |**************************************
2C 3900 |***************************************
1C 3900 |***************************************
* 80x-BROADWELL-NUMA:
Two sockets E5-2698 v4, 2x Broadwell 20 cores/40 threads
Max EFFiciency, BASE frequency and available turbo levels (MHz):
EFFIC 1200 |************
BASE 2200 |**********************
8C 2900 |*****************************
7C 3000 |******************************
6C 3100 |*******************************
5C 3200 |********************************
4C 3300 |*********************************
3C 3400 |**********************************
2C 3600 |************************************
1C 3600 |************************************
* 48x-HASWELL-NUMA
Two sockets E5-2670 v3, 2x Haswell 12 cores/24 threads
Max EFFiciency, BASE frequency and available turbo levels (MHz):
EFFIC 1200 |************
BASE 2300 |***********************
12C 2600 |**************************
11C 2600 |**************************
10C 2600 |**************************
9C 2600 |**************************
8C 2600 |**************************
7C 2600 |**************************
6C 2600 |**************************
5C 2700 |***************************
4C 2800 |****************************
3C 2900 |*****************************
2C 3100 |*******************************
1C 3100 |*******************************
5.2 SETUP
=========
* The baseline is Linux v5.2 with schedutil (non-invariant) and the intel_pstate
driver in passive mode.
* The rationale for choosing the various freq_max values to test have been to
try all the 1-2-3-4C turbo levels (note that 1C and 2C turbo are identical
on all machines), plus one more value closer to base_freq but still in the
turbo range (8C turbo for both 80x-BROADWELL-NUMA and 48x-HASWELL-NUMA).
* In addition we've run all tests with intel_pstate/powersave for comparison.
* The filesystem is always XFS, the userspace is openSUSE Leap 15.1.
* 8x-SKYLAKE-UMA is capable of HWP (Hardware-Managed P-States), so the runs
with active intel_pstate on this machine use that.
This gives, in terms of combinations tested on each machine:
* 8x-SKYLAKE-UMA
* Baseline: Linux v5.2, non-invariant schedutil, intel_pstate passive
* intel_pstate active + powersave + HWP
* invariant schedutil, freq_max = 1C turbo
* invariant schedutil, freq_max = 3C turbo
* invariant schedutil, freq_max = 4C turbo
* both 80x-BROADWELL-NUMA and 48x-HASWELL-NUMA
* [same as 8x-SKYLAKE-UMA, but no HWP capable]
* invariant schedutil, freq_max = 8C turbo
(which on 48x-HASWELL-NUMA is the same as 12C turbo, or "all cores turbo")
5.3 BENCHMARK RESULTS
=====================
5.3.1 NEUTRAL BENCHMARKS
------------------------
Tests that didn't show any measurable difference in performance on any of the
test machines between non-invariant schedutil and our patch are:
* NAS Parallel Benchmarks (NPB) using either MPI or openMP for IPC, any
computational kernel
* flexible I/O (FIO)
* hackbench (using threads or processes, and using pipes or sockets)
5.3.2 NON-NEUTRAL BENCHMARKS
----------------------------
What follow are summary tables where each benchmark result is given a score.
* A tilde (~) means a neutral result, i.e. no difference from baseline.
* Scores are computed with the ratio result_new / result_baseline, so a tilde
means a score of 1.00.
* The results in the score ratio are the geometric means of results running
the benchmark with different parameters (eg: for kernbench: using 1, 2, 4,
... number of processes; for pgbench: varying the number of clients, and so
on).
* The first three tables show higher-is-better kind of tests (i.e. measured in
operations/second), the subsequent three show lower-is-better kind of tests
(i.e. the workload is fixed and we measure elapsed time, think kernbench).
* "gitsource" is a name we made up for the test consisting in running the
entire unit tests suite of the Git SCM and measuring how long it takes. We
take it as a typical example of shell-intensive serialized workload.
* In the "I_PSTATE" column we have the results for intel_pstate/powersave. Other
columns show invariant schedutil for different values of freq_max. 4C turbo
is circled as it's the value we've chosen for the final implementation.
80x-BROADWELL-NUMA (comparison ratio; higher is better)
+------+
I_PSTATE 1C 3C | 4C | 8C
pgbench-ro 1.14 ~ ~ | 1.11 | 1.14
pgbench-rw ~ ~ ~ | ~ | ~
netperf-udp 1.06 ~ 1.06 | 1.05 | 1.07
netperf-tcp ~ 1.03 ~ | 1.01 | 1.02
tbench4 1.57 1.18 1.22 | 1.30 | 1.56
+------+
8x-SKYLAKE-UMA (comparison ratio; higher is better)
+------+
I_PSTATE/HWP 1C 3C | 4C |
pgbench-ro ~ ~ ~ | ~ |
pgbench-rw ~ ~ ~ | ~ |
netperf-udp ~ ~ ~ | ~ |
netperf-tcp ~ ~ ~ | ~ |
tbench4 1.30 1.14 1.14 | 1.16 |
+------+
48x-HASWELL-NUMA (comparison ratio; higher is better)
+------+
I_PSTATE 1C 3C | 4C | 12C
pgbench-ro 1.15 ~ ~ | 1.06 | 1.16
pgbench-rw ~ ~ ~ | ~ | ~
netperf-udp 1.05 0.97 1.04 | 1.04 | 1.02
netperf-tcp 0.96 1.01 1.01 | 1.01 | 1.01
tbench4 1.50 1.05 1.13 | 1.13 | 1.25
+------+
In the table above we see that active intel_pstate is slightly better than our
4C-turbo patch (both in reference to the baseline non-invariant schedutil) on
read-only pgbench and much better on tbench. Both cases are notable in which
it shows that lowering our freq_max (to 8C-turbo and 12C-turbo on
80x-BROADWELL-NUMA and 48x-HASWELL-NUMA respectively) helps invariant
schedutil to get closer.
If we ignore active intel_pstate and focus on the comparison with baseline
alone, there are several instances of double-digit performance improvement.
80x-BROADWELL-NUMA (comparison ratio; lower is better)
+------+
I_PSTATE 1C 3C | 4C | 8C
dbench4 1.23 0.95 0.95 | 0.95 | 0.95
kernbench 0.93 0.83 0.83 | 0.83 | 0.82
gitsource 0.98 0.49 0.49 | 0.49 | 0.48
+------+
8x-SKYLAKE-UMA (comparison ratio; lower is better)
+------+
I_PSTATE/HWP 1C 3C | 4C |
dbench4 ~ ~ ~ | ~ |
kernbench ~ ~ ~ | ~ |
gitsource 0.92 0.55 0.55 | 0.55 |
+------+
48x-HASWELL-NUMA (comparison ratio; lower is better)
+------+
I_PSTATE 1C 3C | 4C | 8C
dbench4 ~ ~ ~ | ~ | ~
kernbench 0.94 0.90 0.89 | 0.90 | 0.90
gitsource 0.97 0.69 0.69 | 0.69 | 0.69
+------+
dbench is not very remarkable here, unless we notice how poorly active
intel_pstate is performing on 80x-BROADWELL-NUMA: 23% regression versus
non-invariant schedutil. We repeated that run getting consistent results. Out
of scope for the patch at hand, but deserving future investigation. Other than
that, we previously ran this campaign with Linux v5.0 and saw the patch doing
better on dbench a the time. We haven't checked closely and can only speculate
at this point.
On the NUMA boxes kernbench gets 10-15% improvements on average; we'll see in
the detailed tables that the gains concentrate on low process counts (lightly
loaded machines).
The test we call "gitsource" (running the git unit test suite, a long-running
single-threaded shell script) appears rather spectacular in this table (gains
of 30-50% depending on the machine). It is to be noted, however, that
gitsource has no adjustable parameters (such as the number of jobs in
kernbench, which we average over in order to get a single-number summary
score) and is exactly the kind of low-parallelism workload that benefits the
most from this patch. When looking at the detailed tables of kernbench or
tbench4, at low process or client counts one can see similar numbers.
5.3.3 SELECTION OF DETAILED RESULTS
-----------------------------------
Machine : 48x-HASWELL-NUMA
Benchmark : tbench4 (i.e. dbench4 over the network, actually loopback)
Varying parameter : number of clients
Unit : MB/sec (higher is better)
5.2.0 vanilla (BASELINE) 5.2.0 intel_pstate 5.2.0 1C-turbo
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Hmean 1 126.73 +- 0.31% ( ) 315.91 +- 0.66% ( 149.28%) 125.03 +- 0.76% ( -1.34%)
Hmean 2 258.04 +- 0.62% ( ) 614.16 +- 0.51% ( 138.01%) 269.58 +- 1.45% ( 4.47%)
Hmean 4 514.30 +- 0.67% ( ) 1146.58 +- 0.54% ( 122.94%) 533.84 +- 1.99% ( 3.80%)
Hmean 8 1111.38 +- 2.52% ( ) 2159.78 +- 0.38% ( 94.33%) 1359.92 +- 1.56% ( 22.36%)
Hmean 16 2286.47 +- 1.36% ( ) 3338.29 +- 0.21% ( 46.00%) 2720.20 +- 0.52% ( 18.97%)
Hmean 32 4704.84 +- 0.35% ( ) 4759.03 +- 0.43% ( 1.15%) 4774.48 +- 0.30% ( 1.48%)
Hmean 64 7578.04 +- 0.27% ( ) 7533.70 +- 0.43% ( -0.59%) 7462.17 +- 0.65% ( -1.53%)
Hmean 128 6998.52 +- 0.16% ( ) 6987.59 +- 0.12% ( -0.16%) 6909.17 +- 0.14% ( -1.28%)
Hmean 192 6901.35 +- 0.25% ( ) 6913.16 +- 0.10% ( 0.17%) 6855.47 +- 0.21% ( -0.66%)
5.2.0 3C-turbo 5.2.0 4C-turbo 5.2.0 12C-turbo
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Hmean 1 128.43 +- 0.28% ( 1.34%) 130.64 +- 3.81% ( 3.09%) 153.71 +- 5.89% ( 21.30%)
Hmean 2 311.70 +- 6.15% ( 20.79%) 281.66 +- 3.40% ( 9.15%) 305.08 +- 5.70% ( 18.23%)
Hmean 4 641.98 +- 2.32% ( 24.83%) 623.88 +- 5.28% ( 21.31%) 906.84 +- 4.65% ( 76.32%)
Hmean 8 1633.31 +- 1.56% ( 46.96%) 1714.16 +- 0.93% ( 54.24%) 2095.74 +- 0.47% ( 88.57%)
Hmean 16 3047.24 +- 0.42% ( 33.27%) 3155.02 +- 0.30% ( 37.99%) 3634.58 +- 0.15% ( 58.96%)
Hmean 32 4734.31 +- 0.60% ( 0.63%) 4804.38 +- 0.23% ( 2.12%) 4674.62 +- 0.27% ( -0.64%)
Hmean 64 7699.74 +- 0.35% ( 1.61%) 7499.72 +- 0.34% ( -1.03%) 7659.03 +- 0.25% ( 1.07%)
Hmean 128 6935.18 +- 0.15% ( -0.91%) 6942.54 +- 0.10% ( -0.80%) 7004.85 +- 0.12% ( 0.09%)
Hmean 192 6901.62 +- 0.12% ( 0.00%) 6856.93 +- 0.10% ( -0.64%) 6978.74 +- 0.10% ( 1.12%)
This is one of the cases where the patch still can't surpass active
intel_pstate, not even when freq_max is as low as 12C-turbo. Otherwise, gains are
visible up to 16 clients and the saturated scenario is the same as baseline.
The scores in the summary table from the previous sections are ratios of
geometric means of the results over different clients, as seen in this table.
Machine : 80x-BROADWELL-NUMA
Benchmark : kernbench (kernel compilation)
Varying parameter : number of jobs
Unit : seconds (lower is better)
5.2.0 vanilla (BASELINE) 5.2.0 intel_pstate 5.2.0 1C-turbo
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Amean 2 379.68 +- 0.06% ( ) 330.20 +- 0.43% ( 13.03%) 285.93 +- 0.07% ( 24.69%)
Amean 4 200.15 +- 0.24% ( ) 175.89 +- 0.22% ( 12.12%) 153.78 +- 0.25% ( 23.17%)
Amean 8 106.20 +- 0.31% ( ) 95.54 +- 0.23% ( 10.03%) 86.74 +- 0.10% ( 18.32%)
Amean 16 56.96 +- 1.31% ( ) 53.25 +- 1.22% ( 6.50%) 48.34 +- 1.73% ( 15.13%)
Amean 32 34.80 +- 2.46% ( ) 33.81 +- 0.77% ( 2.83%) 30.28 +- 1.59% ( 12.99%)
Amean 64 26.11 +- 1.63% ( ) 25.04 +- 1.07% ( 4.10%) 22.41 +- 2.37% ( 14.16%)
Amean 128 24.80 +- 1.36% ( ) 23.57 +- 1.23% ( 4.93%) 21.44 +- 1.37% ( 13.55%)
Amean 160 24.85 +- 0.56% ( ) 23.85 +- 1.17% ( 4.06%) 21.25 +- 1.12% ( 14.49%)
5.2.0 3C-turbo 5.2.0 4C-turbo 5.2.0 8C-turbo
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Amean 2 284.08 +- 0.13% ( 25.18%) 283.96 +- 0.51% ( 25.21%) 285.05 +- 0.21% ( 24.92%)
Amean 4 153.18 +- 0.22% ( 23.47%) 154.70 +- 1.64% ( 22.71%) 153.64 +- 0.30% ( 23.24%)
Amean 8 87.06 +- 0.28% ( 18.02%) 86.77 +- 0.46% ( 18.29%) 86.78 +- 0.22% ( 18.28%)
Amean 16 48.03 +- 0.93% ( 15.68%) 47.75 +- 1.99% ( 16.17%) 47.52 +- 1.61% ( 16.57%)
Amean 32 30.23 +- 1.20% ( 13.14%) 30.08 +- 1.67% ( 13.57%) 30.07 +- 1.67% ( 13.60%)
Amean 64 22.59 +- 2.02% ( 13.50%) 22.63 +- 0.81% ( 13.32%) 22.42 +- 0.76% ( 14.12%)
Amean 128 21.37 +- 0.67% ( 13.82%) 21.31 +- 1.15% ( 14.07%) 21.17 +- 1.93% ( 14.63%)
Amean 160 21.68 +- 0.57% ( 12.76%) 21.18 +- 1.74% ( 14.77%) 21.22 +- 1.00% ( 14.61%)
The patch outperform active intel_pstate (and baseline) by a considerable
margin; the summary table from the previous section says 4C turbo and active
intel_pstate are 0.83 and 0.93 against baseline respectively, so 4C turbo is
0.83/0.93=0.89 against intel_pstate (~10% better on average). There is no
noticeable difference with regard to the value of freq_max.
Machine : 8x-SKYLAKE-UMA
Benchmark : gitsource (time to run the git unit test suite)
Varying parameter : none
Unit : seconds (lower is better)
5.2.0 vanilla 5.2.0 intel_pstate/hwp 5.2.0 1C-turbo
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Amean 858.85 +- 1.16% ( ) 791.94 +- 0.21% ( 7.79%) 474.95 ( 44.70%)
5.2.0 3C-turbo 5.2.0 4C-turbo
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Amean 475.26 +- 0.20% ( 44.66%) 474.34 +- 0.13% ( 44.77%)
In this test, which is of interest as representing shell-intensive
(i.e. fork-intensive) serialized workloads, invariant schedutil outperforms
intel_pstate/powersave by a whopping 40% margin.
5.3.4 POWER CONSUMPTION, PERFORMANCE-PER-WATT
---------------------------------------------
The following table shows average power consumption in watt for each
benchmark. Data comes from turbostat (package average), which in turn is read
from the RAPL interface on CPUs. We know the patch affects CPU frequencies so
it's reasonable to ignore other power consumers (such as memory or I/O). Also,
we don't have a power meter available in the lab so RAPL is the best we have.
turbostat sampled average power every 10 seconds for the entire duration of
each benchmark. We took all those values and averaged them (i.e. with don't
have detail on a per-parameter granularity, only on whole benchmarks).
80x-BROADWELL-NUMA (power consumption, watts)
+--------+
BASELINE I_PSTATE 1C 3C | 4C | 8C
pgbench-ro 130.01 142.77 131.11 132.45 | 134.65 | 136.84
pgbench-rw 68.30 60.83 71.45 71.70 | 71.65 | 72.54
dbench4 90.25 59.06 101.43 99.89 | 101.10 | 102.94
netperf-udp 65.70 69.81 66.02 68.03 | 68.27 | 68.95
netperf-tcp 88.08 87.96 88.97 88.89 | 88.85 | 88.20
tbench4 142.32 176.73 153.02 163.91 | 165.58 | 176.07
kernbench 92.94 101.95 114.91 115.47 | 115.52 | 115.10
gitsource 40.92 41.87 75.14 75.20 | 75.40 | 75.70
+--------+
8x-SKYLAKE-UMA (power consumption, watts)
+--------+
BASELINE I_PSTATE/HWP 1C 3C | 4C |
pgbench-ro 46.49 46.68 46.56 46.59 | 46.52 |
pgbench-rw 29.34 31.38 30.98 31.00 | 31.00 |
dbench4 27.28 27.37 27.49 27.41 | 27.38 |
netperf-udp 22.33 22.41 22.36 22.35 | 22.36 |
netperf-tcp 27.29 27.29 27.30 27.31 | 27.33 |
tbench4 41.13 45.61 43.10 43.33 | 43.56 |
kernbench 42.56 42.63 43.01 43.01 | 43.01 |
gitsource 13.32 13.69 17.33 17.30 | 17.35 |
+--------+
48x-HASWELL-NUMA (power consumption, watts)
+--------+
BASELINE I_PSTATE 1C 3C | 4C | 12C
pgbench-ro 128.84 136.04 129.87 132.43 | 132.30 | 134.86
pgbench-rw 37.68 37.92 37.17 37.74 | 37.73 | 37.31
dbench4 28.56 28.73 28.60 28.73 | 28.70 | 28.79
netperf-udp 56.70 60.44 56.79 57.42 | 57.54 | 57.52
netperf-tcp 75.49 75.27 75.87 76.02 | 76.01 | 75.95
tbench4 115.44 139.51 119.53 123.07 | 123.97 | 130.22
kernbench 83.23 91.55 95.58 95.69 | 95.72 | 96.04
gitsource 36.79 36.99 39.99 40.34 | 40.35 | 40.23
+--------+
A lower power consumption isn't necessarily better, it depends on what is done
with that energy. Here are tables with the ratio of performance-per-watt on
each machine and benchmark. Higher is always better; a tilde (~) means a
neutral ratio (i.e. 1.00).
80x-BROADWELL-NUMA (performance-per-watt ratios; higher is better)
+------+
I_PSTATE 1C 3C | 4C | 8C
pgbench-ro 1.04 1.06 0.94 | 1.07 | 1.08
pgbench-rw 1.10 0.97 0.96 | 0.96 | 0.97
dbench4 1.24 0.94 0.95 | 0.94 | 0.92
netperf-udp ~ 1.02 1.02 | ~ | 1.02
netperf-tcp ~ 1.02 ~ | ~ | 1.02
tbench4 1.26 1.10 1.06 | 1.12 | 1.26
kernbench 0.98 0.97 0.97 | 0.97 | 0.98
gitsource ~ 1.11 1.11 | 1.11 | 1.13
+------+
8x-SKYLAKE-UMA (performance-per-watt ratios; higher is better)
+------+
I_PSTATE/HWP 1C 3C | 4C |
pgbench-ro ~ ~ ~ | ~ |
pgbench-rw 0.95 0.97 0.96 | 0.96 |
dbench4 ~ ~ ~ | ~ |
netperf-udp ~ ~ ~ | ~ |
netperf-tcp ~ ~ ~ | ~ |
tbench4 1.17 1.09 1.08 | 1.10 |
kernbench ~ ~ ~ | ~ |
gitsource 1.06 1.40 1.40 | 1.40 |
+------+
48x-HASWELL-NUMA (performance-per-watt ratios; higher is better)
+------+
I_PSTATE 1C 3C | 4C | 12C
pgbench-ro 1.09 ~ 1.09 | 1.03 | 1.11
pgbench-rw ~ 0.86 ~ | ~ | 0.86
dbench4 ~ 1.02 1.02 | 1.02 | ~
netperf-udp ~ 0.97 1.03 | 1.02 | ~
netperf-tcp 0.96 ~ ~ | ~ | ~
tbench4 1.24 ~ 1.06 | 1.05 | 1.11
kernbench 0.97 0.97 0.98 | 0.97 | 0.96
gitsource 1.03 1.33 1.32 | 1.32 | 1.33
+------+
These results are overall pleasing: in plenty of cases we observe
performance-per-watt improvements. The few regressions (read/write pgbench and
dbench on the Broadwell machine) are of small magnitude. kernbench loses a few
percentage points (it has a 10-15% performance improvement, but apparently the
increase in power consumption is larger than that). tbench4 and gitsource, which
benefit the most from the patch, keep a positive score in this table which is
a welcome surprise; that suggests that in those particular workloads the
non-invariant schedutil (and active intel_pstate, too) makes some rather
suboptimal frequency selections.
+-------------------------------------------------------------------------+
| 6. MICROARCH'ES ADDRESSED HERE
+-------------------------------------------------------------------------+
The patch addresses Xeon Core processors that use MSR_PLATFORM_INFO and
MSR_TURBO_RATIO_LIMIT to advertise their base frequency and turbo frequencies
respectively. This excludes the recent Xeon Scalable Performance processors
line (Xeon Gold, Platinum etc) whose MSRs have to be parsed differently.
Subsequent patches will address:
* Xeon Scalable Performance processors and Atom Goldmont/Goldmont Plus
* Xeon Phi (Knights Landing, Knights Mill)
* Atom Silvermont
+-------------------------------------------------------------------------+
| 7. REFERENCES
+-------------------------------------------------------------------------+
Tests have been run with the help of the MMTests performance testing
framework, see github.com/gormanm/mmtests. The configuration file names for
the benchmark used are:
db-pgbench-timed-ro-small-xfs
db-pgbench-timed-rw-small-xfs
io-dbench4-async-xfs
network-netperf-unbound
network-tbench
scheduler-unbound
workload-kerndevel-xfs
workload-shellscripts-xfs
hpc-nas-c-class-mpi-full-xfs
hpc-nas-c-class-omp-full
All those benchmarks are generally available on the web:
pgbench: https://www.postgresql.org/docs/10/pgbench.html
netperf: https://hewlettpackard.github.io/netperf/
dbench/tbench: https://dbench.samba.org/
gitsource: git unit test suite, github.com/git/git
NAS Parallel Benchmarks: https://www.nas.nasa.gov/publications/npb.html
hackbench: https://people.redhat.com/mingo/cfs-scheduler/tools/hackbench.c
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Giovanni Gherdovich <ggherdovich@suse.cz>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Doug Smythies <dsmythies@telus.net>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Link: https://lkml.kernel.org/r/20200122151617.531-2-ggherdovich@suse.cz
When a running task is moved on a throttled task group and there is no
other task enqueued on the CPU, the task can keep running using 100% CPU
whatever the allocated bandwidth for the group and although its cfs rq is
throttled. Furthermore, the group entity of the cfs_rq and its parents are
not enqueued but only set as curr on their respective cfs_rqs.
We have the following sequence:
sched_move_task
-dequeue_task: dequeue task and group_entities.
-put_prev_task: put task and group entities.
-sched_change_group: move task to new group.
-enqueue_task: enqueue only task but not group entities because cfs_rq is
throttled.
-set_next_task : set task and group_entities as current sched_entity of
their cfs_rq.
Another impact is that the root cfs_rq runnable_load_avg at root rq stays
null because the group_entities are not enqueued. This situation will stay
the same until an "external" event triggers a reschedule. Let trigger it
immediately instead.
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Ben Segall <bsegall@google.com>
Link: https://lkml.kernel.org/r/1579011236-31256-1-git-send-email-vincent.guittot@linaro.org
