mt7986 only supports 8/12/16/24/32/48/96/192 kHz
Signed-off-by: Maso Huang <maso.huang@mediatek.com>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Link: https://lore.kernel.org/r/20231024035019.11732-4-maso.huang@mediatek.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Remove the mt7986_wm8960_priv structure.
Signed-off-by: Maso Huang <maso.huang@mediatek.com>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Link: https://lore.kernel.org/r/20231024035019.11732-3-maso.huang@mediatek.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Drop the remove callback of mt7986_wm8960.
Signed-off-by: Maso Huang <maso.huang@mediatek.com>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Link: https://lore.kernel.org/r/20231024035019.11732-2-maso.huang@mediatek.com
Signed-off-by: Mark Brown <broonie@kernel.org>
The bpf_user_ringbuf_drain() BPF_CALL function uses an atomic_set()
immediately preceded by smp_mb__before_atomic() so as to order storing
of ring-buffer consumer and producer positions prior to the atomic_set()
call's clearing of the ->busy flag, as follows:
smp_mb__before_atomic();
atomic_set(&rb->busy, 0);
Although this works given current architectures and implementations, and
given that this only needs to order prior writes against a later write.
However, it does so by accident because the smp_mb__before_atomic()
is only guaranteed to work with read-modify-write atomic operations, and
not at all with things like atomic_set() and atomic_read().
Note especially that smp_mb__before_atomic() will not, repeat *not*,
order the prior write to "a" before the subsequent non-read-modify-write
atomic read from "b", even on strongly ordered systems such as x86:
WRITE_ONCE(a, 1);
smp_mb__before_atomic();
r1 = atomic_read(&b);
Therefore, replace the smp_mb__before_atomic() and atomic_set() with
atomic_set_release() as follows:
atomic_set_release(&rb->busy, 0);
This is no slower (and sometimes is faster) than the original, and also
provides a formal guarantee of ordering that the original lacks.
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: David Vernet <void@manifault.com>
Link: https://lore.kernel.org/bpf/ec86d38e-cfb4-44aa-8fdb-6c925922d93c@paulmck-laptop
htab_lock_bucket uses the following logic to avoid recursion:
1. preempt_disable();
2. check percpu counter htab->map_locked[hash] for recursion;
2.1. if map_lock[hash] is already taken, return -BUSY;
3. raw_spin_lock_irqsave();
However, if an IRQ hits between 2 and 3, BPF programs attached to the IRQ
logic will not able to access the same hash of the hashtab and get -EBUSY.
This -EBUSY is not really necessary. Fix it by disabling IRQ before
checking map_locked:
1. preempt_disable();
2. local_irq_save();
3. check percpu counter htab->map_locked[hash] for recursion;
3.1. if map_lock[hash] is already taken, return -BUSY;
4. raw_spin_lock().
Similarly, use raw_spin_unlock() and local_irq_restore() in
htab_unlock_bucket().
Fixes: 20b6cc34ea ("bpf: Avoid hashtab deadlock with map_locked")
Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Song Liu <song@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/7a9576222aa40b1c84ad3a9ba3e64011d1a04d41.camel@linux.ibm.com
Link: https://lore.kernel.org/bpf/20231012055741.3375999-1-song@kernel.org
acpi_scan_device_not_present() is called when a device in the
hierarchy is not available for enumeration. Historically enumeration
was only based on whether the device was present.
To add support for only enumerating devices that are both present
and enabled, this helper should be renamed. It was only ever about
enumeration, rename it acpi_scan_device_not_enumerated().
No change in behaviour is intended.
Signed-off-by: James Morse <james.morse@arm.com>
Reviewed-by: Gavin Shan <gshan@redhat.com>
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Miguel Luis <miguel.luis@oracle.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Return struct nft_elem_priv instead of struct nft_set_ext for
consistency with ("netfilter: nf_tables: expose opaque set element as
struct nft_elem_priv") and to prepare the introduction of element
timeout updates from control path.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Instead of copying struct nft_set_elem into struct nft_trans_elem, store
the pointer to the opaque set element object in the transaction. Adapt
set backend API (and set backend implementations) to take the pointer to
opaque set element representation whenever required.
This patch deconstifies .remove() and .activate() set backend API since
these modify the set element opaque object. And it also constify
nft_set_elem_ext() this provides access to the nft_set_ext struct
without updating the object.
According to pahole on x86_64, this patch shrinks struct nft_trans_elem
size from 216 to 24 bytes.
This patch also reduces stack memory consumption by removing the
template struct nft_set_elem object, using the opaque set element object
instead such as from the set iterator API, catchall elements and the get
element command.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Check whether the event type matches the PMU type firstly in
pmu::event_init() before touching the event. Otherwise we'll
change the events of others and lead to incorrect results.
Since in perf_init_event() we may call every pmu's event_init()
in a certain case, we should not modify the event if it's not
ours.
Fixes: 8404b0fbc7 ("drivers/perf: hisi: Add driver for HiSilicon PCIe PMU")
Signed-off-by: Yicong Yang <yangyicong@hisilicon.com>
Link: https://lore.kernel.org/r/20231024092954.42297-2-yangyicong@huawei.com
Signed-off-by: Will Deacon <will@kernel.org>
Add placeholder structure and place it at the beginning of each struct
nft_*_elem for each existing set backend, instead of exposing elements
as void type to the frontend which defeats compiler type checks. Use
this pointer to this new type to replace void *.
This patch updates the following set backend API to use this new struct
nft_elem_priv placeholder structure:
- update
- deactivate
- flush
- get
as well as the following helper functions:
- nft_set_elem_ext()
- nft_set_elem_init()
- nft_set_elem_destroy()
- nf_tables_set_elem_destroy()
This patch adds nft_elem_priv_cast() to cast struct nft_elem_priv to
native element representation from the corresponding set backend.
BUILD_BUG_ON() makes sure this .priv placeholder is always at the top
of the opaque set element representation.
Suggested-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
.flush is always successful since this results from iterating over the
set elements to toggle mark the element as inactive in the next
generation.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Relieve the dump callback from having to inspect nlmsg_type upon each
call, just do it once at start of the dump.
Signed-off-by: Phil Sutter <phil@nwl.cc>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
No need to allocate it if one may just use struct netlink_callback's
scratch area for it.
Signed-off-by: Phil Sutter <phil@nwl.cc>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Prep work for moving the context into struct netlink_callback scratch
area.
Signed-off-by: Phil Sutter <phil@nwl.cc>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Name it for what it is supposed to become, a real nft_obj_dump_ctx. No
functional change intended.
Signed-off-by: Phil Sutter <phil@nwl.cc>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Prep work for moving the filter into struct netlink_callback's scratch
area.
Signed-off-by: Phil Sutter <phil@nwl.cc>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
The code does not make use of cb->args fields past the first one, no
need to zero them.
Signed-off-by: Phil Sutter <phil@nwl.cc>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
The spinlock is back from the day when connabels did not have
a fixed size and reallocation had to be supported.
Remove it. This change also allows to call the helpers from
softirq or timers without deadlocks.
Also add WARN()s to catch refcounting imbalances.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
br_netfilter registers two forward hooks, one for ip and one for arp.
Just use a common function for both and then call the arp/ip helper
as needed.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Rule reset is not concurrency-safe per-se, so multiple CPUs may reset
the same rule at the same time. At least counter and quota expressions
will suffer from value underruns in this case.
Prevent this by introducing dedicated locking callbacks for nfnetlink
and the asynchronous dump handling to serialize access.
Signed-off-by: Phil Sutter <phil@nwl.cc>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Outsource the reply skb preparation for non-dump getrule requests into a
distinct function. Prep work for rule reset locking.
Signed-off-by: Phil Sutter <phil@nwl.cc>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
The table lookup will be dropped from that function, so remove that
dependency from audit logging code. Using whatever is in
nla[NFTA_RULE_TABLE] is sufficient as long as the previous rule info
filling succeded.
Signed-off-by: Phil Sutter <phil@nwl.cc>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
There is no need for asynchronous garbage collection, rbtree inserts
can only happen from the netlink control plane.
We already perform on-demand gc on insertion, in the area of the
tree where the insertion takes place, but we don't do a full tree
walk there for performance reasons.
Do a full gc walk at the end of the transaction instead and
remove the async worker.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Next patch adds a cllaer that doesn't hold the priv->write lock and
will need a similar function.
Rename the existing function to make it clear that it can only
be used for opportunistic gc during insertion.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
A helper function for printing non-work-conserving alarms is added in
commit b00355db3f ("pkt_sched: sch_hfsc: sch_htb: Add non-work-conserving
warning handler."). In this commit, use qdisc_warn_nonwc() instead of
WARN_ONCE() to handle the non-work-conserving warning in qfq Qdisc.
Signed-off-by: Liu Jian <liujian56@huawei.com>
Link: https://lore.kernel.org/r/20231023064729.370649-1-liujian56@huawei.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
QM215 is typically paired with a PM8916 PMIC and uses its SMPA1 and
LDOA2 regulators in voltage level mode for VDDCX and VDDMX, respectively.
Signed-off-by: Otto Pflüger <otto.pflueger@abscue.de>
Reviewed-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Link: https://lore.kernel.org/r/20231014133823.14088-4-otto.pflueger@abscue.de
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
MSM8917 uses the SMPA2 and LDOA3 regulators provided by the PM8937 PMIC
for the VDDCX and VDDMX power domains in voltage level mode,
respectively. These definitions should also work on MSM8937.
Signed-off-by: Otto Pflüger <otto.pflueger@abscue.de>
Reviewed-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Link: https://lore.kernel.org/r/20231014133823.14088-3-otto.pflueger@abscue.de
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Merge the immutable branch genpd_dt into next, to allow the DT bindings to
be tested together with new pmdomain changes that are targeted for v6.7.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
The MSM8917, MSM8937 and QM215 SoCs have VDDCX and VDDMX power domains
controlled in voltage level mode. Define the MSM8937 and QM215 power
domains as aliases because these SoCs are similar to MSM8917 and may
share some parts of the device tree.
Also add the compatibles for these SoCs to the documentation, with
qcom,msm8937-rpmpd using qcom,msm8917-rpmpd as a fallback compatible
because there are no known differences. QM215 is not compatible with
these because it uses different regulators.
Signed-off-by: Otto Pflüger <otto.pflueger@abscue.de>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Link: https://lore.kernel.org/r/20231014133823.14088-2-otto.pflueger@abscue.de
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
The commit c494a447c1 ("soc: bcm: bcm2835-power: Refactor ASB control")
refactored the ASB control by using a general function to handle both
the enable and disable. But this patch introduced a subtle regression:
we need to check if !!(readl(base + reg) & ASB_ACK) == enable, not just
check if (readl(base + reg) & ASB_ACK) == true.
Currently, this is causing an invalid register state in V3D when
unloading and loading the driver, because `bcm2835_asb_disable()` will
return -ETIMEDOUT and `bcm2835_asb_power_off()` will fail to disable the
ASB slave for V3D.
Fixes: c494a447c1 ("soc: bcm: bcm2835-power: Refactor ASB control")
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Stefan Wahren <stefan.wahren@i2se.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20231024101251.6357-2-mcanal@igalia.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Pablo Neira Ayuso says:
====================
GTP tunnel driver fixes
The following patchset contains two fixes for the GTP tunnel driver:
1) Incorrect GTPA_MAX definition in UAPI headers. This is updating an
existing UAPI definition but for a good reason, this is certainly
broken. Similar fixes for incorrect _MAX definition in netlink
headers were applied in the past too.
2) Fix GTP driver PMTU with GRO packets, add missing call to
skb_gso_validate_network_len() to handle GRO packets.
====================
Link: https://lore.kernel.org/r/20231022202519.659526-1-pablo@netfilter.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Call skb_gso_validate_network_len() to check if packet is over PMTU.
Fixes: 459aa660eb ("gtp: add initial driver for datapath of GPRS Tunneling Protocol (GTP-U)")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Subtract one to __GTPA_MAX, otherwise GTPA_MAX is off by 2.
Fixes: 459aa660eb ("gtp: add initial driver for datapath of GPRS Tunneling Protocol (GTP-U)")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
In the previous implementation, when multiple xsk sockets were
associated with a single xsk_buff_pool, a situation could arise
where the xsk_tx_list maintained data at the front for one xsk
socket while starving the xsk sockets at the back of the list.
This could result in issues such as the inability to transmit packets,
increased latency, and jitter. To address this problem, we introduce
a new variable called tx_budget_spent, which limits each xsk to transmit
a maximum of MAX_PER_SOCKET_BUDGET tx descriptors. This allocation ensures
equitable opportunities for subsequent xsk sockets to send tx descriptors.
The value of MAX_PER_SOCKET_BUDGET is set to 32.
Signed-off-by: Albert Huang <huangjie.albert@bytedance.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Magnus Karlsson <magnus.karlsson@intel.com>
Link: https://lore.kernel.org/bpf/20231023125732.82261-1-huangjie.albert@bytedance.com
This reverts commit eb3f1d9e42.
The patches for the features that these bindings described are still
under active development, so odds are these bindings will also have to
be changed in the future. As no in-tree code requires these bindings at
the moment, revert them.
Reported-by: Johan Hovold <johan@kernel.org>
Link: https://lore.kernel.org/r/ZTeObdjSSok0tttg@hovoldconsulting.com
Cc: Bjorn Andersson <quic_bjorande@quicinc.com>
Cc: Krishna Kurapati <quic_kriskura@quicinc.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This reverts commit ca58c4ae75.
The patches for the features that these bindings described are still
under active development, so odds are these bindings will also have to
be changed in the future. As no in-tree code requires these bindings at
the moment, revert them.
Reported-by: Johan Hovold <johan@kernel.org>
Link: https://lore.kernel.org/r/ZTeObdjSSok0tttg@hovoldconsulting.com
Cc: Krishna Kurapati <quic_kriskura@quicinc.com>
Cc: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Richard reported that a serial port may end up sometimes with tx data
pending in the buffer for long periods of time.
Turns out we bail out early on any errors from pm_runtime_get(),
including -EINPROGRESS. To fix the issue, we need to ignore -EINPROGRESS
as we only care about the runtime PM usage count at this point. We check
for an active runtime PM state later on for tx.
Fixes: 84a9582fd2 ("serial: core: Start managing serial controllers to enable runtime PM")
Cc: stable <stable@kernel.org>
Reported-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Cc: Bruce Ashfield <bruce.ashfield@gmail.com>
Cc: Mikko Rapeli <mikko.rapeli@linaro.org>
Cc: Paul Gortmaker <paul.gortmaker@windriver.com>
Cc: Randy MacLeod <randy.macleod@windriver.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
Tested-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Link: https://lore.kernel.org/r/20231023074856.61896-1-tony@atomide.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently there is a device tree entry called "local-mac-address"
which can be filled by the bootloader or manually set.This is
useful when the user does not want to use the MAC address
programmed into the SoC.
Currently, the davinci_emac reads the MAC from the DT, copies
it from pdata->mac_addr to priv->mac_addr, then blindly overwrites
it by reading from registers in the SoC, and falls back to a
random MAC if it's still not valid. This completely ignores any
MAC address in the device tree.
In order to use the local-mac-address, check to see if the contents
of priv->mac_addr are valid before falling back to reading from the
SoC when the MAC address is not valid.
Signed-off-by: Adam Ford <aford173@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Link: https://lore.kernel.org/r/20231022151911.4279-1-aford173@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The linux-distros list relaxed their rules to try to adapt better to
how the Linux kernel works. Let's update the Coordination part to
explain why and when to contact them or not to and how to avoid trouble
in the future.
Link: https://www.openwall.com/lists/oss-security/2023/09/08/4
Cc: Kees Cook <keescook@chromium.org>
Cc: Solar Designer <solar@openwall.com>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Acked-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Willy Tarreau <w@1wt.eu>
Link: https://lore.kernel.org/r/20231015130959.26242-1-w@1wt.eu
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
We are seeing systemd hang on its autofs direct mount at
/proc/sys/fs/binfmt_misc.
Historically this was due to a mismatch in the communication structure
size between a 64 bit kernel and a 32 bit user space and was fixed by
making the pipe communication record oriented.
During autofs v5 development I decided to stay with the existing usage
instead of changing to a packed structure for autofs <=> user space
communications which turned out to be a mistake on my part.
Problems arose and they were fixed by allowing for the 64 bit to 32
bit size difference in the automount(8) code.
Along the way systemd started to use autofs and eventually encountered
this problem too. systemd refused to compensate for the length
difference insisting it be fixed in the kernel. Fortunately Linus
implemented the packetized pipe which resolved the problem in a
straight forward and simple way.
In the autofs mount api conversion series I inadvertatly dropped the
packet pipe flag settings when adding the autofs_parse_fd() function.
This patch fixes that omission.
Fixes: 546694b8f6 ("autofs: add autofs_parse_fd()")
Signed-off-by: Ian Kent <raven@themaw.net>
Link: https://lore.kernel.org/r/20231023093359.64265-1-raven@themaw.net
Tested-by: Anders Roxell <anders.roxell@linaro.org>
Cc: Bill O'Donnell <bodonnel@redhat.com>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Dan Carpenter <dan.carpenter@linaro.org>
Cc: Anders Roxell <anders.roxell@linaro.org>
Cc: Naresh Kamboju <naresh.kamboju@linaro.org>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Reported-by: Linux Kernel Functional Testing <lkft@linaro.org>
Reported-by: Anders Roxell <anders.roxell@linaro.org>
Signed-off-by: Christian Brauner <brauner@kernel.org>
The calling code actually handles -ECHILD, so this BUG_ON
can be converted to WARN_ON_ONCE.
Signed-off-by: Bernd Schubert <bschubert@ddn.com>
Link: https://lore.kernel.org/r/20231023184718.11143-1-bschubert@ddn.com
Cc: Christian Brauner <brauner@kernel.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Amir Goldstein <amir73il@gmail.com>
Cc: Dharmendra Singh <dsingh@ddn.com>
Cc: Miklos Szeredi <miklos@szeredi.hu>
Cc: linux-fsdevel@vger.kernel.org
Signed-off-by: Christian Brauner <brauner@kernel.org>
Chen Yu reports a hackbench regression of cluster wakeup when
hackbench threads equal to the CPU number [1]. Analysis shows
it's because we wake up more on the target CPU even if the
prev_cpu is a good wakeup candidate and leads to the decrease
of the CPU utilization.
Generally if the task's prev_cpu is idle we'll wake up the task
on it without scanning. On cluster machines we'll try to wake up
the task in the same cluster of the target for better cache
affinity, so if the prev_cpu is idle but not sharing the same
cluster with the target we'll still try to find an idle CPU within
the cluster. This will improve the performance at low loads on
cluster machines. But in the issue above, if the prev_cpu is idle
but not in the cluster with the target CPU, we'll try to scan an
idle one in the cluster. But since the system is busy, we're
likely to fail the scanning and use target instead, even if
the prev_cpu is idle. Then leads to the regression.
This patch solves this in 2 steps:
o record the prev_cpu/recent_used_cpu if they're good wakeup
candidates but not sharing the cluster with the target.
o on scanning failure use the prev_cpu/recent_used_cpu if
they're recorded as idle
[1] https://lore.kernel.org/all/ZGzDLuVaHR1PAYDt@chenyu5-mobl1/
Closes: https://lore.kernel.org/all/ZGsLy83wPIpamy6x@chenyu5-mobl1/
Reported-by: Chen Yu <yu.c.chen@intel.com>
Signed-off-by: Yicong Yang <yangyicong@hisilicon.com>
Tested-and-reviewed-by: Chen Yu <yu.c.chen@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Link: https://lkml.kernel.org/r/20231019033323.54147-4-yangyicong@huawei.com
For platforms having clusters like Kunpeng920, CPUs within the same cluster
have lower latency when synchronizing and accessing shared resources like
cache. Thus, this patch tries to find an idle cpu within the cluster of the
target CPU before scanning the whole LLC to gain lower latency. This
will be implemented in 2 steps in select_idle_sibling():
1. When the prev_cpu/recent_used_cpu are good wakeup candidates, use them
if they're sharing cluster with the target CPU. Otherwise trying to
scan for an idle CPU in the target's cluster.
2. Scanning the cluster prior to the LLC of the target CPU for an
idle CPU to wakeup.
Testing has been done on Kunpeng920 by pinning tasks to one numa and two
numa. On Kunpeng920, Each numa has 8 clusters and each cluster has 4 CPUs.
With this patch, We noticed enhancement on tbench and netperf within one
numa or cross two numa on top of tip-sched-core commit
9b46f1abc6d4 ("sched/debug: Print 'tgid' in sched_show_task()")
tbench results (node 0):
baseline patched
1: 327.2833 372.4623 ( 13.80%)
4: 1320.5933 1479.8833 ( 12.06%)
8: 2638.4867 2921.5267 ( 10.73%)
16: 5282.7133 5891.5633 ( 11.53%)
32: 9810.6733 9877.3400 ( 0.68%)
64: 7408.9367 7447.9900 ( 0.53%)
128: 6203.2600 6191.6500 ( -0.19%)
tbench results (node 0-1):
baseline patched
1: 332.0433 372.7223 ( 12.25%)
4: 1325.4667 1477.6733 ( 11.48%)
8: 2622.9433 2897.9967 ( 10.49%)
16: 5218.6100 5878.2967 ( 12.64%)
32: 10211.7000 11494.4000 ( 12.56%)
64: 13313.7333 16740.0333 ( 25.74%)
128: 13959.1000 14533.9000 ( 4.12%)
netperf results TCP_RR (node 0):
baseline patched
1: 76546.5033 90649.9867 ( 18.42%)
4: 77292.4450 90932.7175 ( 17.65%)
8: 77367.7254 90882.3467 ( 17.47%)
16: 78519.9048 90938.8344 ( 15.82%)
32: 72169.5035 72851.6730 ( 0.95%)
64: 25911.2457 25882.2315 ( -0.11%)
128: 10752.6572 10768.6038 ( 0.15%)
netperf results TCP_RR (node 0-1):
baseline patched
1: 76857.6667 90892.2767 ( 18.26%)
4: 78236.6475 90767.3017 ( 16.02%)
8: 77929.6096 90684.1633 ( 16.37%)
16: 77438.5873 90502.5787 ( 16.87%)
32: 74205.6635 88301.5612 ( 19.00%)
64: 69827.8535 71787.6706 ( 2.81%)
128: 25281.4366 25771.3023 ( 1.94%)
netperf results UDP_RR (node 0):
baseline patched
1: 96869.8400 110800.8467 ( 14.38%)
4: 97744.9750 109680.5425 ( 12.21%)
8: 98783.9863 110409.9637 ( 11.77%)
16: 99575.0235 110636.2435 ( 11.11%)
32: 95044.7250 97622.8887 ( 2.71%)
64: 32925.2146 32644.4991 ( -0.85%)
128: 12859.2343 12824.0051 ( -0.27%)
netperf results UDP_RR (node 0-1):
baseline patched
1: 97202.4733 110190.1200 ( 13.36%)
4: 95954.0558 106245.7258 ( 10.73%)
8: 96277.1958 105206.5304 ( 9.27%)
16: 97692.7810 107927.2125 ( 10.48%)
32: 79999.6702 103550.2999 ( 29.44%)
64: 80592.7413 87284.0856 ( 8.30%)
128: 27701.5770 29914.5820 ( 7.99%)
Note neither Kunpeng920 nor x86 Jacobsville supports SMT, so the SMT branch
in the code has not been tested but it supposed to work.
Chen Yu also noticed this will improve the performance of tbench and
netperf on a 24 CPUs Jacobsville machine, there are 4 CPUs in one
cluster sharing L2 Cache.
[https://lore.kernel.org/lkml/Ytfjs+m1kUs0ScSn@worktop.programming.kicks-ass.net]
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Barry Song <song.bao.hua@hisilicon.com>
Signed-off-by: Yicong Yang <yangyicong@hisilicon.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Tim Chen <tim.c.chen@linux.intel.com>
Reviewed-by: Chen Yu <yu.c.chen@intel.com>
Reviewed-by: Gautham R. Shenoy <gautham.shenoy@amd.com>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Tested-and-reviewed-by: Chen Yu <yu.c.chen@intel.com>
Tested-by: Yicong Yang <yangyicong@hisilicon.com>
Link: https://lkml.kernel.org/r/20231019033323.54147-3-yangyicong@huawei.com
Add cpus_share_resources() API. This is the preparation for the
optimization of select_idle_cpu() on platforms with cluster scheduler
level.
On a machine with clusters cpus_share_resources() will test whether
two cpus are within the same cluster. On a non-cluster machine it
will behaves the same as cpus_share_cache(). So we use "resources"
here for cache resources.
Signed-off-by: Barry Song <song.bao.hua@hisilicon.com>
Signed-off-by: Yicong Yang <yangyicong@hisilicon.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Gautham R. Shenoy <gautham.shenoy@amd.com>
Reviewed-by: Tim Chen <tim.c.chen@linux.intel.com>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Tested-and-reviewed-by: Chen Yu <yu.c.chen@intel.com>
Tested-by: K Prateek Nayak <kprateek.nayak@amd.com>
Link: https://lkml.kernel.org/r/20231019033323.54147-2-yangyicong@huawei.com
Igor Raits and Bagas Sanjaya report a RQCF_ACT_SKIP leak warning.
This warning may be triggered in the following situations:
CPU0 CPU1
__schedule()
*rq->clock_update_flags <<= 1;* unregister_fair_sched_group()
pick_next_task_fair+0x4a/0x410 destroy_cfs_bandwidth()
newidle_balance+0x115/0x3e0 for_each_possible_cpu(i) *i=0*
rq_unpin_lock(this_rq, rf) __cfsb_csd_unthrottle()
raw_spin_rq_unlock(this_rq)
rq_lock(*CPU0_rq*, &rf)
rq_clock_start_loop_update()
rq->clock_update_flags & RQCF_ACT_SKIP <--
raw_spin_rq_lock(this_rq)
The purpose of RQCF_ACT_SKIP is to skip the update rq clock,
but the update is very early in __schedule(), but we clear
RQCF_*_SKIP very late, causing it to span that gap above
and triggering this warning.
In __schedule() we can clear the RQCF_*_SKIP flag immediately
after update_rq_clock() to avoid this RQCF_ACT_SKIP leak warning.
And set rq->clock_update_flags to RQCF_UPDATED to avoid
rq->clock_update_flags < RQCF_ACT_SKIP warning that may be triggered later.
Fixes: ebb83d84e4 ("sched/core: Avoid multiple calling update_rq_clock() in __cfsb_csd_unthrottle()")
Closes: https://lore.kernel.org/all/20230913082424.73252-1-jiahao.os@bytedance.com
Reported-by: Igor Raits <igor.raits@gmail.com>
Reported-by: Bagas Sanjaya <bagasdotme@gmail.com>
Suggested-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Hao Jia <jiahao.os@bytedance.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/all/a5dd536d-041a-2ce9-f4b7-64d8d85c86dc@gmail.com
