The kaiser update made an interesting choice, never to free any shadow
page tables. Contention on global spinlock was worrying, particularly
with it held across page table scans when freeing. Something had to be
done: I was going to add refcounting; but simply never to free them is
an appealing choice, minimizing contention without complicating the code
(the more a page table is found already, the less the spinlock is used).
But leaking pages in this way is also a worry: can we get away with it?
At the very least, we need a count to show how bad it actually gets:
in principle, one might end up wasting about 1/256 of memory that way
(1/512 for when direct-mapped pages have to be user-mapped, plus 1/512
for when they are user-mapped from the vmalloc area on another occasion
(but we don't have vmalloc'ed stacks, so only large ldts are vmalloc'ed).
Add per-cpu stat NR_KAISERTABLE: including 256 at startup for the
shared pgd entries, and 1 for each intermediate page table added
thereafter for user-mapping - but leave out the 1 per mm, for its
shadow pgd, because that distracts from the monotonic increase.
Shown in /proc/vmstat as nr_overhead (0 if kaiser not enabled).
In practice, it doesn't look so bad so far: more like 1/12000 after
nine hours of gtests below; and movable pageblock segregation should
tend to cluster the kaiser tables into a subset of the address space
(if not, they will be bad for compaction too). But production may
tell a different story: keep an eye on this number, and bring back
lighter freeing if it gets out of control (maybe a shrinker).
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
While trying to get our gold link to work, four cleanups:
matched the gdt_page declaration to its definition;
in fiddling unsuccessfully with PERCPU_INPUT(), lined up backslashes;
lined up the backslashes according to convention in percpu-defs.h;
deleted the unused irq_stack_pointer addition to irq_stack_union.
Sad to report that aligning backslashes does not appear to help gold
align to 8192: but while these did not help, they are worth keeping.
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Kaiser only needs to map one page of the stack; and
kernel/fork.c did not build on powerpc (no __PAGE_KERNEL).
It's all cleaner if linux/kaiser.h provides kaiser_map_thread_stack()
and kaiser_unmap_thread_stack() wrappers around asm/kaiser.h's
kaiser_add_mapping() and kaiser_remove_mapping(). And use
linux/kaiser.h in init/main.c to avoid the #ifdefs there.
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Add the Rockchip ISP1 specific processing parameter format
V4L2_META_FMT_RK_ISP1_PARAMS and metadata format
V4L2_META_FMT_RK_ISP1_STAT_3A for 3A.
Signed-off-by: Shunqian Zheng <zhengsq@rock-chips.com>
Signed-off-by: Jacob Chen <jacob2.chen@rock-chips.com>
BUG=b:36227021
TEST=Camera works on Scarlet with the whole series
(am from https://patchwork.linuxtv.org/patch/46247/)
Conflicts:
drivers/media/v4l2-core/v4l2-ioctl.c
Change-Id: I9ede55b68c075e54cefb11edecf5ef68b5452ec4
Signed-off-by: Jacob Chen <jacob2.chen@rock-chips.com>
due to optee os bugfix patch for FIQ:
cf6a4d8 arm: sm: [bugfix] save/restore fiq core registers
we have to update sm_nsec_ctx to keep same with optee os, otherwise
FIQ debugger "bt" command is abnormal.
Change-Id: I950cfacd1c34abd88fbee2862c593b5fa59387bd
Signed-off-by: Joseph Chen <chenjh@rock-chips.com>
When the devfreq cooling device was designed, it was an oversight not to
pass a pointer to the struct devfreq as the first parameters of the
callbacks. The design patterns of the kernel suggest it for a good
reason.
By passing a pointer to struct devfreq, the driver can register one
function that works with multiple devices. With the current
implementation, a driver that can work with multiple devices has to
create multiple copies of the same function with different parameters so
that each devfreq_cooling_device can use the appropriate one. By
passing a pointer to struct devfreq, the driver can identify which
device it's referring to.
Change-Id: I384bf9aafd2391eccab2ca6a76e4e57f2740aa6b
Cc: Zhang Rui <rui.zhang@intel.com>
Cc: Eduardo Valentin <edubezval@gmail.com>
Reviewed-by: Punit Agrawal <punit.agrawal@arm.com>
Reviewed-by: Ørjan Eide <orjan.eide@arm.com>
Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
Signed-off-by: Javi Merino <javi.merino@arm.com>
Signed-off-by: Eduardo Valentin <edubezval@gmail.com>
Signed-off-by: Tao Huang <huangtao@rock-chips.com>
(cherry picked from commit 3aa5374376)
[ Upstream commit 513674b5a2 ]
sysctl.ip6.auto_flowlabels is default 1. In our hosts, we set it to 2.
If sockopt doesn't set autoflowlabel, outcome packets from the hosts are
supposed to not include flowlabel. This is true for normal packet, but
not for reset packet.
The reason is ipv6_pinfo.autoflowlabel is set in sock creation. Later if
we change sysctl.ip6.auto_flowlabels, the ipv6_pinfo.autoflowlabel isn't
changed, so the sock will keep the old behavior in terms of auto
flowlabel. Reset packet is suffering from this problem, because reset
packet is sent from a special control socket, which is created at boot
time. Since sysctl.ipv6.auto_flowlabels is 1 by default, the control
socket will always have its ipv6_pinfo.autoflowlabel set, even after
user set sysctl.ipv6.auto_flowlabels to 1, so reset packset will always
have flowlabel. Normal sock created before sysctl setting suffers from
the same issue. We can't even turn off autoflowlabel unless we kill all
socks in the hosts.
To fix this, if IPV6_AUTOFLOWLABEL sockopt is used, we use the
autoflowlabel setting from user, otherwise we always call
ip6_default_np_autolabel() which has the new settings of sysctl.
Note, this changes behavior a little bit. Before commit 42240901f7
(ipv6: Implement different admin modes for automatic flow labels), the
autoflowlabel behavior of a sock isn't sticky, eg, if sysctl changes,
existing connection will change autoflowlabel behavior. After that
commit, autoflowlabel behavior is sticky in the whole life of the sock.
With this patch, the behavior isn't sticky again.
Cc: Martin KaFai Lau <kafai@fb.com>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Tom Herbert <tom@quantonium.net>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit b5476022bb ]
IPv4 stack reacts to changes to small MTU, by disabling itself under
RTNL.
But there is a window where threads not using RTNL can see a wrong
device mtu. This can lead to surprises, in igmp code where it is
assumed the mtu is suitable.
Fix this by reading device mtu once and checking IPv4 minimal MTU.
This patch adds missing IPV4_MIN_MTU define, to not abuse
ETH_MIN_MTU anymore.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 9abffc6f2e upstream.
mcryptd_enqueue_request() grabs the per-CPU queue struct and protects
access to it with disabled preemption. Then it schedules a worker on the
same CPU. The worker in mcryptd_queue_worker() guards access to the same
per-CPU variable with disabled preemption.
If we take CPU-hotplug into account then it is possible that between
queue_work_on() and the actual invocation of the worker the CPU goes
down and the worker will be scheduled on _another_ CPU. And here the
preempt_disable() protection does not work anymore. The easiest thing is
to add a spin_lock() to guard access to the list.
Another detail: mcryptd_queue_worker() is not processing more than
MCRYPTD_BATCH invocation in a row. If there are still items left, then
it will invoke queue_work() to proceed with more later. *I* would
suggest to simply drop that check because it does not use a system
workqueue and the workqueue is already marked as "CPU_INTENSIVE". And if
preemption is required then the scheduler should do it.
However if queue_work() is used then the work item is marked as CPU
unbound. That means it will try to run on the local CPU but it may run
on another CPU as well. Especially with CONFIG_DEBUG_WQ_FORCE_RR_CPU=y.
Again, the preempt_disable() won't work here but lock which was
introduced will help.
In order to keep work-item on the local CPU (and avoid RR) I changed it
to queue_work_on().
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Some chips need adjust opp level by different chip-process, add
common functions to select opp level from device-tree, so modules
can select opp level easy.
Change-Id: Ifbd5f720e6a52a68f13697bbb37ac01ff4a87e3e
Signed-off-by: Liang Chen <cl@rock-chips.com>
For next Soc VOP only vopb win1 support AFBDC, so we need
add afbdc feature for every win.
Change-Id: Icbe5e26189d2147a6b81f2f75d0b855b2c35fd26
Signed-off-by: Sandy Huang <hjc@rock-chips.com>
This adds the necessary data for handling dmcfreq on the rk3328
Change-Id: If4cff5cc372f80b6776a7272a1bff54abef2cf33
Signed-off-by: CanYang He <hcy@rock-chips.com>
Add a helper function to get driver private data from CEC
adapter. This helps the readability a little bit and allows
to change the 'priv' field name to something else without
needing to touch all drivers.
Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
(cherry picked from commit 0b0b97dcc7)
Change-Id: I2691d8328350e968b74a90d1622b8e857cefc112
Signed-off-by: Algea Cao <algea.cao@rock-chips.com>
The kAPI cec-pin.h header also defined data structures that did
not belong here but were private to the CEC core code.
Split that part off into a cec-pin-priv.h header.
Signed-off-by: Hans Verkuil <hansverk@cisco.com>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
(cherry picked from commit 5bf24e08b6)
Change-Id: I8a0100eb9731a5657ef51939ef56d0c22e046023
Signed-off-by: Algea Cao <algea.cao@rock-chips.com>
Add support for two new low-level events: PIN_HPD_LOW and PIN_HPD_HIGH.
This is specifically meant for use with the upcoming cec-gpio driver
and makes it possible to trace when the HPD pin changes. Some HDMI
sinks do strange things with the HPD and this makes it easy to debug
this.
Note that this also moves the initialization of a devnode mutex and
list to the allocate_adapter function: if the HPD is high, then as
soon as the HPD interrupt is created an interrupt occurs and
cec_queue_pin_hpd_event() is called which requires that the devnode
mutex and list are initialized.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
(cherry picked from commit 333ef6bd10)
Change-Id: I2fc5ca673ed67247d6f1a60906e515caa828a064
Signed-off-by: Algea Cao <algea.cao@rock-chips.com>
The current CEC pin value (0 or 1) was part of the cec_pin struct,
but that assumes that CEC pin monitoring can only be used with
a driver that uses the low-level CEC pin framework.
But hardware that has both a high-level API and can monitor the
CEC pin at low-level at the same time does not need to depend on
the cec pin framework.
To support such devices remove the cur_value field from struct cec_pin
and add a cec_pin_is_high field to cec_adapter. This also makes it
possible to drop the '#ifdef CONFIG_CEC_PIN' in cec-api.c.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
(cherry picked from commit 28e11b15b6)
Change-Id: I7f144faf81e6eb3dfdaa7cca5978214f67eaa84d
Signed-off-by: Algea Cao <algea.cao@rock-chips.com>
The free_irq() function could be called from interrupt context,
which is invalid. Move this to the thread.
In the interrupt handler we just request that the thread disables
the irq. This is done through an atomic so we don't need to add
any spinlocks.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
(cherry picked from commit cb74749493)
Change-Id: Icb6eaa227e0799b33834758c6fa6d148859ec98f
Signed-off-by: Algea Cao <algea.cao@rock-chips.com>
The CEC_EVENT_PIN_LOW/HIGH defines and the cec_queue_pin_event() function
did not specify that these were about CEC pin events.
Since in the future there will also be HPD pin events it is wise to rename
the event defines and function to CEC_EVENT_PIN_CEC_LOW/HIGH and
cec_queue_pin_cec_event() now before these become part of the ABI.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
(cherry picked from commit 9a6b2a8740)
Change-Id: I292a2aca552575958e0705727ab7ea277839b479
Signed-off-by: Algea Cao <algea.cao@rock-chips.com>
Allow calling cec_notifier_set_phys_addr and
cec_notifier_set_phys_addr_from_edid with a NULL notifier, in which
case these functions do nothing.
Add a cec_notifier_phys_addr_invalidate helper function (the notifier
equivalent of cec_phys_addr_invalidate).
These changes simplify drm CEC driver support.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
(cherry picked from commit fc1ff45a07)
Change-Id: I301d15d4f7fd710ad8c532035492f7434a089fa7
Signed-off-by: Algea Cao <algea.cao@rock-chips.com>
Add support for CEC hardware that relies on low-level pin polling or
GPIO interrupts.
One example is the Allwinner SoC. But any GPIO-based CEC implementation can
use this as well.
A GPIO implementation is very suitable as well for debugging: it can use
interrupts to detect state changes and report it. Userspace can then verify
if the bus traffic is correct. This also makes error injection possible.
The disadvantage is that it is hard to get the timings right since linux
isn't a hard realtime system.
In general on an idle system it works quite well, but under load the timer
will miss its mark every so often.
The debugfs file /sys/kernel/debug/cec/cecX/status gives some statistics
with respect to the timer overruns.
When the adapter is unconfigured and the low-level driver supports
interrupts, then the interrupt will be used to detect changes. This should
be quite accurate. But when the adapter is configured a hrtimer has to be
used.
The hrtimer implements a state machine where for each state the code will
read the bus or drive the bus and go on to the next state. It will re-arm
the timer with a delay based on the next state.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Reviewed-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
(cherry picked from commit ea5c8ef296)
Change-Id: Iba76fc7d2a93bfe5fff1792afe82416ad4f02c8c
Signed-off-by: Algea Cao <algea.cao@rock-chips.com>
Add support for the new MONITOR_PIN mode.
Add the cec_pin_event function that the CEC pin code will call to queue pin
change events.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Reviewed-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
(cherry picked from commit b8d62f50b1)
Change-Id: I52267565092f0f3539e2a0f98328a360839a1d2f
Signed-off-by: Algea Cao <algea.cao@rock-chips.com>
Event handling was always fairly simplistic since there were only
two events. With the addition of pin events this needed to be redesigned.
The state_change and lost_msgs events are now core events with the
guarantee that the last state is always available. The new pin events
are a queue of events (up to 64 for each event) and the oldest event
will be dropped if the application cannot keep up. Lost events are
marked with a new event flag.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Reviewed-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
(cherry picked from commit 6b2bbb0874)
Change-Id: I922f9ef9ae718ed93e02556cc007b2d943c5b11d
Signed-off-by: Algea Cao <algea.cao@rock-chips.com>
Currently the transmit_(attempt_)done and received_msg functions set
the timestamp themselves. For the upcoming low-level pin API we need
to pass this as an argument instead. So make _ts variants that allow
the caller to specify the timestamp.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Reviewed-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
(cherry picked from commit 0861ad14c6)
Change-Id: I98991dc09741fd6a5c3593e25d70f08897f663a6
Signed-off-by: Algea Cao <algea.cao@rock-chips.com>
Kernel logging messes up the upcoming low-level CEC monitoring support
which is very time-sensitive. So change the debug level of this message
but keep a counter that is shown in the debugfs status log.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Reviewed-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
(cherry picked from commit bb789e03f2)
Change-Id: I4d40fc677c4f5d35db2176aad8846ae780bce53d
Signed-off-by: Algea Cao <algea.cao@rock-chips.com>
Add a new capability CEC_CAP_NEEDS_HPD. If this capability is set
then the hardware can only use CEC if the HDMI Hotplug Detect pin
is high. Such hardware cannot handle the corner case in the CEC specification
where it is possible to transmit messages even if no hotplug signal is
present (needed for some displays that turn off the HPD when in standby,
but still have CEC enabled).
Typically hardware that needs this capability have the HPD wired to the CEC
block, often to a 'power' or 'active' pin.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
(cherry picked from commit f902c1e95d)
Change-Id: I6bdbb202fb53ae2e8f14e3be15ec7613914f2bbb
Signed-off-by: Algea Cao <algea.cao@rock-chips.com>
A simpler variant of cec_transmit_done to be used where the HW does
just a single attempt at a transmit. So if the status indicates an
error, then the corresponding error count will always be 1 and this
function figures that out based on the status argument.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
(cherry picked from commit c94cdc1e0c)
Change-Id: I707f73b15a3634c7a5883764caf6747422daa888
Signed-off-by: Algea Cao <algea.cao@rock-chips.com>
This function simplifies the integration of CEC in DRM drivers.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
(cherry picked from commit 23111ec304)
Change-Id: Ibad64be8af1b0548978bd91b467e82bc719afca3
Signed-off-by: Algea Cao <algea.cao@rock-chips.com>
This config option is strictly speaking independent of the
media subsystem since it can be used by drm as well.
Besides, it looks odd when drivers select CEC_CORE and
MEDIA_CEC_NOTIFIER, that's inconsistent naming.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
(cherry picked from commit e94c32818d)
Change-Id: I6f89afe3b6f0bca110451ece5ef3ecb12c2922bb
Signed-off-by: Algea Cao <algea.cao@rock-chips.com>
Drop the separate cec-edid.h header and merge it into cec.h.
There was really no need to have a separate header for this.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
(cherry picked from commit ee7e987133)
Change-Id: Idf14df419f4663f24bcbb2de3603498651a469d4
Signed-off-by: Algea Cao <algea.cao@rock-chips.com>
The Kconfig options for the CEC subsystem were a bit messy. In
addition there were two cec sources (cec-edid.c and cec-notifier.c)
that were outside of the media/cec directory, which was weird.
Move those sources to media/cec as well.
The cec-edid and cec-notifier functionality is now part of the cec
module and these are no longer separate modules.
Also remove the MEDIA_CEC_EDID config option and include it with the
main CEC config option (which defined CEC_EDID anyway).
Added static inlines to cec-edid.h for dummy functions when CEC_CORE
isn't defined.
CEC drivers should now depend on CEC_CORE.
CEC drivers that need the cec-notifier functionality must explicitly
select CEC_NOTIFIER.
The s5p-cec and stih-cec drivers depended on VIDEO_DEV instead of
CEC_CORE, fix that as well.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Acked-by: Benjamin Gaignard <benjamin.gaignard@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
(cherry picked from commit 56a263aaa0)
Change-Id: I3c83c4d1aea534c2e8163422c73c8bc276f20f8f
Signed-off-by: Algea Cao <algea.cao@rock-chips.com>
Add support for CEC notifiers, which is used to convey CEC physical address
information from video drivers to their CEC counterpart driver(s).
Based on an earlier version from Russell King:
https://patchwork.kernel.org/patch/9277043/
The cec_notifier is a reference counted object containing the CEC physical address
state of a video device.
When a new notifier is registered the current state will be reported to
that notifier at registration time.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com>
Tested-by: Benjamin Gaignard <benjamin.gaignard@linaro.org>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
(cherry picked from commit 6917a7b774)
Change-Id: I048f69ccbf5c1de2f856a6820653d8bfc8455e11
Signed-off-by: Algea Cao <algea.cao@rock-chips.com>
when vpu reset, sometimes will failed when set power domain
idle
Change-Id: I38185cce616adb41e698b0c95793934abbd51f62
Signed-off-by: Jung Zhao <jung.zhao@rock-chips.com>
If an interrupt is marked NOAUTOEN then request_irq() installs the action,
but does not enable the interrupt via startup_irq(). The interrupt is
enabled via enable_irq() later from the driver. enable_irq() calls
irq_enable().
That means that for interrupts which have a irq_startup() callback this
callback is never invoked. Neither is irq_domain_activate_irq() invoked for
such interrupts.
If an interrupt depends on irq_startup() or irq_domain_activate_irq() then
the enable via irq_enable() is not enough.
Add a status flag IRQD_IRQ_STARTED_UP and use this to select the proper
mechanism in enable_irq(). Use the flag also to avoid pointless calls into
the low level functions.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Marc Zyngier <marc.zyngier@arm.com>
Cc: dianders@chromium.org
Cc: jeffy <jeffy.chen@rock-chips.com>
Cc: Brian Norris <briannorris@chromium.org>
Cc: tfiga@chromium.org
Link: http://lkml.kernel.org/r/20170531100212.130986205@linutronix.de
(cherry picked from commit 201d7f47f3)
Conflicts:
include/linux/irq.h
[due to missing upstream patches:
08d85f3 irqdomain: Avoid activating interrupts more than once
1a3d28a UPSTREAM: genirq: Introduce IRQD_AFFINITY_MANAGED flag
6297714 UPSTREAM: irq: Privatize irq_common_data::state_use_accessors]
Change-Id: Ie8d492c694171fa81e3df61e8561bec160ad37bb
Signed-off-by: Jeffy Chen <jeffy.chen@rock-chips.com>
Per listen(fd, backlog) rules, there is really no point accepting a SYN,
sending a SYNACK, and dropping the following ACK packet if accept queue
is full, because application is not draining accept queue fast enough.
This behavior is fooling TCP clients that believe they established a
flow, while there is nothing at server side. They might then send about
10 MSS (if using IW10) that will be dropped anyway while server is under
stress.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Acked-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 5ea8ea2cb7)
Backport for pass Android CTS CtsLibcoreTestCases:
libcore.javax.net.ServerSocketFactoryTest#testCreateServerSocketWithPortNoBacklog
Refer the comments on libcore/luni/src/test/java/libcore/javax/net/ServerSocketFactoryTest.java
// This test may fail on kernel versions between 4.4 and 4.9, due to a kernel implementation
// detail change. Backporting the following kernel change will fix the behavior.
// http://b/31960002
// https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=5ea8ea2cb7f1d0db15762c9b0bb9e7330425a071
public void testCreateServerSocketWithPortNoBacklog() throws IOException {
ServerSocket serverSocket = ServerSocketFactory.getDefault().createServerSocket(0, 1);
testSocket(serverSocket, 1);
}
Change-Id: I9bb3531480f4942447d8d4eded110aa31bc7c031
Signed-off-by: Tao Huang <huangtao@rock-chips.com>
