Currently the bridge doesn't notify the underlying devices about new
FDBs learned. The FDB sync is placed on the switchdev notifier chain
because devices may potentially learn FDB that are not directly related
to their ports, for example:
1. Mixed SW/HW bridge - FDBs that point to the ASICs external devices
should be offloaded as CPU traps in order to
perform forwarding in slow path.
2. EVPN - Externally learned FDBs for the vtep device.
Notification is sent only about static FDB add/del. This is done due
to fact that currently this is the only scenario supported by switch
drivers.
Signed-off-by: Arkadi Sharshevsky <arkadis@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Reviewed-by: Ivan Vecera <ivecera@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is done as a preparation stage before setting the bridge port flags
from the bridge code. Currently the device can be queried for the bridge
flags state, but the querier cannot distinguish if the flag is disabled
or if it is not supported at all. Thus, add new attr and a bit-mask which
include information regarding the support on a per-flag basis.
Drivers that support bridge offload but not support bridge flags should
return zeroed bitmask.
Signed-off-by: Arkadi Sharshevsky <arkadis@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Reviewed-by: Ivan Vecera <ivecera@redhat.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Ivan Vecera <ivecera@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Change the third parameter to be the required struct xen_dm_op_buf *
instead of a generic void * (which blindly accepts any pointer).
Signed-off-by: Sergey Dyasli <sergey.dyasli@citrix.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Signed-off-by: Juergen Gross <jgross@suse.com>
Move the reboot-mode.h include file into include/linux to allow drivers
outside drivers/power/reset to implement reboot-mode.
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.co.uk>
The PMU IRQ number is set through the VCPU device's KVM_SET_DEVICE_ATTR
ioctl handler for the KVM_ARM_VCPU_PMU_V3_IRQ attribute, but there is no
enforced or stated requirement that this must happen after initializing
the VGIC. As a result, calling vgic_valid_spi() which relies on the
nr_spis being set during the VGIC init can incorrectly fail.
Introduce irq_is_spi, which determines if an IRQ number is within the
SPI range without verifying it against the actual VGIC properties.
Signed-off-by: Christoffer Dall <cdall@linaro.org>
Reviewed-by: Marc Zyngier <marc.zyngier@arm.com>
When injecting an IRQ to the VGIC, you now have to present an owner
token for that IRQ line to show that you are the owner of that line.
IRQ lines driven from userspace or via an irqfd do not have an owner and
will simply pass a NULL pointer.
Also get rid of the unused kvm_vgic_inject_mapped_irq prototype.
Signed-off-by: Christoffer Dall <cdall@linaro.org>
Acked-by: Marc Zyngier <marc.zyngier@arm.com>
Previously there was no way to configure these chips in the event that the
defaults didn't match the battery in question.
For chips with RAM data memory (and also those with flash/NVM data memory
if CONFIG_BATTERY_BQ27XXX_DT_UPDATES_NVM is defined and the user has not
set module param dt_monitored_battery_updates_nvm=0) we now call
power_supply_get_battery_info(), check its values, and write battery
properties to chip data memory if there is a dm_regs table for the chip.
Signed-off-by: Matt Ranostay <matt@ranostay.consulting>
Signed-off-by: Liam Breck <kernel@networkimprov.net>
Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.co.uk>
Add these to enable read/write of chip data memory RAM/NVM/flash:
bq27xxx_battery_seal()
bq27xxx_battery_unseal()
bq27xxx_battery_set_cfgupdate()
bq27xxx_battery_soft_reset()
bq27xxx_battery_read_dm_block()
bq27xxx_battery_write_dm_block()
bq27xxx_battery_checksum_dm_block()
Signed-off-by: Matt Ranostay <matt@ranostay.consulting>
Signed-off-by: Liam Breck <kernel@networkimprov.net>
Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.co.uk>
David Howells says:
====================
rxrpc: Tx length parameter
Here's a set of patches that allows someone initiating a client call with
AF_RXRPC to indicate upfront the total amount of data that will be
transmitted. This will allow AF_RXRPC to encrypt directly from source
buffer to packet rather than having to copy into the buffer and only
encrypt when it's full (the encrypted portion of the packet starts with a
length and so we can't encrypt until we know what the length will be).
The three patches are:
(1) Provide a means of finding out what control message types are actually
supported. EINVAL is reported if an unsupported cmsg type is seen, so
we don't want to set the new cmsg unless we know it will be accepted.
(2) Consolidate some stuff into a struct to reduce the parameter count on
the function that parses the cmsg buffer.
(3) Introduce the RXRPC_TX_LENGTH cmsg. This can be provided on the first
sendmsg() that contributes data to a client call request or a service
call reply. If provided, the user must provide exactly that amount of
data or an error will be incurred.
Changes in version 2:
(*) struct rxrpc_send_params::tx_total_len should be s64 not u64. Thanks to
Julia Lawall for reporting this.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
DRAM supply shortage and poor memory pressure tracking in TCP
stack makes any change in SO_SNDBUF/SO_RCVBUF (or equivalent autotuning
limits) and tcp_mem[] quite hazardous.
TCPMemoryPressures SNMP counter is an indication of tcp_mem sysctl
limits being hit, but only tracking number of transitions.
If TCP stack behavior under stress was perfect :
1) It would maintain memory usage close to the limit.
2) Memory pressure state would be entered for short times.
We certainly prefer 100 events lasting 10ms compared to one event
lasting 200 seconds.
This patch adds a new SNMP counter tracking cumulative duration of
memory pressure events, given in ms units.
$ cat /proc/sys/net/ipv4/tcp_mem
3088 4117 6176
$ grep TCP /proc/net/sockstat
TCP: inuse 180 orphan 0 tw 2 alloc 234 mem 4140
$ nstat -n ; sleep 10 ; nstat |grep Pressure
TcpExtTCPMemoryPressures 1700
TcpExtTCPMemoryPressuresChrono 5209
v2: Used EXPORT_SYMBOL_GPL() instead of EXPORT_SYMBOL() as David
instructed.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In Tiny SRCU, __srcu_read_lock() is a trivial function, outweighed by
its EXPORT_SYMBOL_GPL(), and on many architectures, its call sequence.
This commit therefore moves it to srcutiny.h so that it can be inlined.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Commit d160a727c4 ("srcu: Make SRCU be built by default") in response
to build errors, which were caused by code that included srcu.h
despite !SRCU. However, srcutiny.o is almost 2K of code, which is not
insignificant for those attempting to run the Linux kernel on IoT devices.
This commit therefore makes SRCU be once again optional, and adjusts
srcu.h to allow error-free inclusion in !SRCU kernel builds.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Acked-by: Nicolas Pitre <nico@linaro.org>
This commit adds a rcupdate_announce_bootup_oddness() function to
print out non-default values of significant kernel boot parameter
settings to aid in debugging.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
This commit rearranges Tiny SRCU's srcu_struct structure, substitutes
u8 for bool, and shrinks counters down to short.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
There is material describing the ordering guarantees provided by
spin_unlock_wait(), but it is not necessarily easy to find. This commit
therefore adds a docbook header comment to this function informally
describing its semantics.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Currently rcu_barrier() uses call_rcu() to enqueue new callbacks
on each CPU with a non-empty callback list. This works, but means
that rcu_barrier() forces grace periods that are not otherwise needed.
The key point is that rcu_barrier() never needs to wait for a grace
period, but instead only for all pre-existing callbacks to be invoked.
This means that rcu_barrier()'s new callbacks should be placed in
the callback-list segment containing the last pre-existing callback.
This commit makes this change using the new rcu_segcblist_entrain()
function.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Linu Cherian reported a WARN in cleanup_srcu_struct() when shutting
down a guest running iperf on a VFIO assigned device. This happens
because irqfd_wakeup() calls srcu_read_lock(&kvm->irq_srcu) in interrupt
context, while a worker thread does the same inside kvm_set_irq(). If the
interrupt happens while the worker thread is executing __srcu_read_lock(),
updates to the Classic SRCU ->lock_count[] field or the Tree SRCU
->srcu_lock_count[] field can be lost.
The docs say you are not supposed to call srcu_read_lock() and
srcu_read_unlock() from irq context, but KVM interrupt injection happens
from (host) interrupt context and it would be nice if SRCU supported the
use case. KVM is using SRCU here not really for the "sleepable" part,
but rather due to its IPI-free fast detection of grace periods. It is
therefore not desirable to switch back to RCU, which would effectively
revert commit 719d93cd5f ("kvm/irqchip: Speed up KVM_SET_GSI_ROUTING",
2014-01-16).
However, the docs are overly conservative. You can have an SRCU instance
only has users in irq context, and you can mix process and irq context
as long as process context users disable interrupts. In addition,
__srcu_read_unlock() actually uses this_cpu_dec() on both Tree SRCU and
Classic SRCU. For those two implementations, only srcu_read_lock()
is unsafe.
When Classic SRCU's __srcu_read_unlock() was changed to use this_cpu_dec(),
in commit 5a41344a3d ("srcu: Simplify __srcu_read_unlock() via
this_cpu_dec()", 2012-11-29), __srcu_read_lock() did two increments.
Therefore it kept __this_cpu_inc(), with preempt_disable/enable in
the caller. Tree SRCU however only does one increment, so on most
architectures it is more efficient for __srcu_read_lock() to use
this_cpu_inc(), and any performance differences appear to be down in
the noise.
Cc: stable@vger.kernel.org
Fixes: 719d93cd5f ("kvm/irqchip: Speed up KVM_SET_GSI_ROUTING")
Reported-by: Linu Cherian <linuc.decode@gmail.com>
Suggested-by: Linu Cherian <linuc.decode@gmail.com>
Cc: kvm@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Roopa reported attempts to delete a bond device that is referenced in a
multipath route is hanging:
$ ifdown bond2 # ifupdown2 command that deletes virtual devices
unregister_netdevice: waiting for bond2 to become free. Usage count = 2
Steps to reproduce:
echo 1 > /proc/sys/net/ipv6/conf/all/ignore_routes_with_linkdown
ip link add dev bond12 type bond
ip link add dev bond13 type bond
ip addr add 2001:db8:2::0/64 dev bond12
ip addr add 2001:db8:3::0/64 dev bond13
ip route add 2001:db8:33::0/64 nexthop via 2001:db8:2::2 nexthop via 2001:db8:3::2
ip link del dev bond12
ip link del dev bond13
The root cause is the recent change to keep routes on a linkdown. Update
the check to detect when the device is unregistering and release the
route for that case.
Fixes: a1a22c1206 ("net: ipv6: Keep nexthop of multipath route on admin down")
Reported-by: Roopa Prabhu <roopa@cumulusnetworks.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
Acked-by: Roopa Prabhu <roopa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Having multiple devices being able to signal the same interrupt line is
very confusing and almost certainly guarantees a configuration error.
Therefore, introduce a very simple allocator which allows a device to
claim an interrupt line from the vgic for a given VM.
Signed-off-by: Christoffer Dall <cdall@linaro.org>
Acked-by: Marc Zyngier <marc.zyngier@arm.com>
First we define an ABI using the vcpu devices that lets userspace set
the interrupt numbers for the various timers on both the 32-bit and
64-bit KVM/ARM implementations.
Second, we add the definitions for the groups and attributes introduced
by the above ABI. (We add the PMU define on the 32-bit side as well for
symmetry and it may get used some day.)
Third, we set up the arch-specific vcpu device operation handlers to
call into the timer code for anything related to the
KVM_ARM_VCPU_TIMER_CTRL group.
Fourth, we implement support for getting and setting the timer interrupt
numbers using the above defined ABI in the arch timer code.
Fifth, we introduce error checking upon enabling the arch timer (which
is called when first running a VCPU) to check that all VCPUs are
configured to use the same PPI for the timer (as mandated by the
architecture) and that the virtual and physical timers are not
configured to use the same IRQ number.
Signed-off-by: Christoffer Dall <cdall@linaro.org>
Reviewed-by: Marc Zyngier <marc.zyngier@arm.com>
We currently initialize the arch timer IRQ numbers from the reset code,
presumably because we once intended to model multiple CPU or SoC types
from within the kernel and have hard-coded reset values in the reset
code.
As we are moving towards userspace being in charge of more fine-grained
CPU emulation and stitching together the pieces needed to emulate a
particular type of CPU, we should no longer have a tight coupling
between resetting a VCPU and setting IRQ numbers.
Therefore, move the logic to define and use the default IRQ numbers to
the timer code and set the IRQ number immediately when creating the
VCPU.
Signed-off-by: Christoffer Dall <cdall@linaro.org>
Reviewed-by: Marc Zyngier <marc.zyngier@arm.com>
We are about to need this define in the arch timer code as well so move
it to a common location.
Signed-off-by: Christoffer Dall <cdall@linaro.org>
Acked-by: Marc Zyngier <marc.zyngier@arm.com>
Since we got support for devices in userspace which allows reporting the
PMU overflow output status to userspace, we should actually allow
creating the PMU on systems without an in-kernel irqchip, which in turn
requires us to slightly clarify error codes for the ABI and move things
around for the initialization phase.
Signed-off-by: Christoffer Dall <cdall@linaro.org>
Reviewed-by: Marc Zyngier <marc.zyngier@arm.com>
We want to move some TCP sysctls to net namespaces in the future.
tcp_window_scaling, tcp_sack and tcp_timestamps being fetched
from tcp_parse_options(), we need to pass an extra parameter.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When generating bootable VM images certain systems (most notably
s390x) require devices with 4k blocksize. This patch implements
a new flag 'LO_FLAGS_BLOCKSIZE' which will set the physical
blocksize to that of the underlying device, and allow to change
the logical blocksize for up to the physical blocksize.
Signed-off-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Declare bus.write/read_bulk/write_bulk().
Add I2C write/read_bulk/write_bulk() to implement the above.
Add bq27xxx_write/read_block/write_block() helpers to call the above.
Signed-off-by: Matt Ranostay <matt@ranostay.consulting>
Signed-off-by: Liam Breck <kernel@networkimprov.net>
Acked-by: "Andrew F. Davis" <afd@ti.com>
Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.co.uk>
power_supply_get_battery_info() reads battery data from devicetree.
struct power_supply_battery_info provides battery data to drivers.
Its fields correspond to elements in enum power_supply_property.
Drivers may surface battery data in sysfs via corresponding
POWER_SUPPLY_PROP_* fields.
Signed-off-by: Matt Ranostay <matt@ranostay.consulting>
Signed-off-by: Liam Breck <kernel@networkimprov.net>
Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.co.uk>
We need to push the chain index down to the drivers, so they have the
information to which chain the rule belongs. For now, no driver supports
multichain offload, so only chain 0 is supported. This is needed to
prevent chain squashes during offload for now. Later this will be used
to implement multichain offload.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
KVM/ARM Fixes for v4.12-rc5 - Take 2
Changes include:
- Fix an issue with migrating GICv2 VMs on GICv3 systems.
- Squashed a bug for gicv3 when figuring out preemption levels.
- Fix a potential null pointer derefence in KVM happening under memory
pressure.
- Maintain RES1 bits in the SCTLR_EL2 to make sure KVM works on new
architecture revisions.
- Allow unaligned accesses at EL2/HYP
Instead of using the SKB queue with the fake pkt_type for the
offloaded RX BA session management, also handle this with the
normal aggregation state machine worker. This also makes the
use of this more reliable since it gets rid of the allocation
of the fake skb.
Combined with the previous patch, this finally allows us to
get rid of the pkt_type hack entirely, so do that as well.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
This brings in commit 7a7c0a6438 ("mac80211: fix TX aggregation
start/stop callback race") to allow the follow-up cleanup.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
The HID transport drivers either re-implement exactly the same logic
(usbhid, i2c-hid) or forget to implement it (usbhid) which causes issues
when the same device is accessed via multiple interfaces (for example input
device through evdev and also hidraw). Let's muve the locking logic into
HID core to make sure the serialized behavior is always enforced.
Also let's uninline and move hid_hw_start() and hid_hw_stop() into hid-core
as hid_hw_start() is somewhat large and do not believe we get any benefit
from these two being inline.
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Read port connector type from the firmware instead of caching it in the
driver metadata.
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Update struct mlx5_ifc_create(modify)_flow_table_bits according to
the last device specification.
Signed-off-by: Maor Gottlieb <maorg@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Apple currently supports three very common USB chargers:
https://www.apple.com/power-adapters/
These chargers implement a proprietary Apple method for advertising
1A, 2.1A, and 2.4A at 5V called "Brick ID".
In addition, 3rd parties implement the same charging method in many
charging accessories that work with iOS devices.
Devices that have charger detection chips such as the Pericom PI3USB9281,
eg. Google Chromebook Pixel 2015, are capable of detecting
these chargers, so let's add a type to facilicate passing that info
up to userspace.
This adds a separate power supply type for Apple's proprietary
"Brick ID" charging method.
Signed-off-by: Benson Leung <bleung@chromium.org>
Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.co.uk>
We have been facing some problems with self-suspending constrained
deadline tasks. The main reason is that the original CBS was not
designed for such sort of tasks.
One problem reported by Xunlei Pang takes place when a task
suspends, and then is awakened before the deadline, but so close
to the deadline that its remaining runtime can cause the task
to have an absolute density higher than allowed. In such situation,
the original CBS assumes that the task is facing an early activation,
and so it replenishes the task and set another deadline, one deadline
in the future. This rule works fine for implicit deadline tasks.
Moreover, it allows the system to adapt the period of a task in which
the external event source suffered from a clock drift.
However, this opens the window for bandwidth leakage for constrained
deadline tasks. For instance, a task with the following parameters:
runtime = 5 ms
deadline = 7 ms
[density] = 5 / 7 = 0.71
period = 1000 ms
If the task runs for 1 ms, and then suspends for another 1ms,
it will be awakened with the following parameters:
remaining runtime = 4
laxity = 5
presenting a absolute density of 4 / 5 = 0.80.
In this case, the original CBS would assume the task had an early
wakeup. Then, CBS will reset the runtime, and the absolute deadline will
be postponed by one relative deadline, allowing the task to run.
The problem is that, if the task runs this pattern forever, it will keep
receiving bandwidth, being able to run 1ms every 2ms. Following this
behavior, the task would be able to run 500 ms in 1 sec. Thus running
more than the 5 ms / 1 sec the admission control allowed it to run.
Trying to address the self-suspending case, Luca Abeni, Giuseppe
Lipari, and Juri Lelli [1] revisited the CBS in order to deal with
self-suspending tasks. In the new approach, rather than
replenishing/postponing the absolute deadline, the revised wakeup rule
adjusts the remaining runtime, reducing it to fit into the allowed
density.
A revised version of the idea is:
At a given time t, the maximum absolute density of a task cannot be
higher than its relative density, that is:
runtime / (deadline - t) <= dl_runtime / dl_deadline
Knowing the laxity of a task (deadline - t), it is possible to move
it to the other side of the equality, thus enabling to define max
remaining runtime a task can use within the absolute deadline, without
over-running the allowed density:
runtime = (dl_runtime / dl_deadline) * (deadline - t)
For instance, in our previous example, the task could still run:
runtime = ( 5 / 7 ) * 5
runtime = 3.57 ms
Without causing damage for other deadline tasks. It is note worthy
that the laxity cannot be negative because that would cause a negative
runtime. Thus, this patch depends on the patch:
df8eac8caf ("sched/deadline: Throttle a constrained deadline task activated after the deadline")
Which throttles a constrained deadline task activated after the
deadline.
Finally, it is also possible to use the revised wakeup rule for
all other tasks, but that would require some more discussions
about pros and cons.
Reported-by: Xunlei Pang <xpang@redhat.com>
Signed-off-by: Daniel Bristot de Oliveira <bristot@redhat.com>
[peterz: replaced dl_is_constrained with dl_is_implicit]
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Juri Lelli <juri.lelli@arm.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Luca Abeni <luca.abeni@santannapisa.it>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Romulo Silva de Oliveira <romulo.deoliveira@ufsc.br>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tommaso Cucinotta <tommaso.cucinotta@sssup.it>
Link: http://lkml.kernel.org/r/5c800ab3a74a168a84ee5f3f84d12a02e11383be.1495803804.git.bristot@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Without this the build will fail for !CONFIG_ACPI builds on x86.
Fixes: 94116f81 ("ACPI: Switch to use generic guid_t in acpi_evaluate_dsm()")
Signed-off-by: Christoph Hellwig <hch@lst.de>
Allow to skip writing odd chroma rows by setting the RDRW bit for
4:2:0 chroma subsampled formats for any IDMAC write channel. This
also allows to skip reading odd rows for the VDIC read channel.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Network devices can allocate reasources and private memory using
netdev_ops->ndo_init(). However, the release of these resources
can occur in one of two different places.
Either netdev_ops->ndo_uninit() or netdev->destructor().
The decision of which operation frees the resources depends upon
whether it is necessary for all netdev refs to be released before it
is safe to perform the freeing.
netdev_ops->ndo_uninit() presumably can occur right after the
NETDEV_UNREGISTER notifier completes and the unicast and multicast
address lists are flushed.
netdev->destructor(), on the other hand, does not run until the
netdev references all go away.
Further complicating the situation is that netdev->destructor()
almost universally does also a free_netdev().
This creates a problem for the logic in register_netdevice().
Because all callers of register_netdevice() manage the freeing
of the netdev, and invoke free_netdev(dev) if register_netdevice()
fails.
If netdev_ops->ndo_init() succeeds, but something else fails inside
of register_netdevice(), it does call ndo_ops->ndo_uninit(). But
it is not able to invoke netdev->destructor().
This is because netdev->destructor() will do a free_netdev() and
then the caller of register_netdevice() will do the same.
However, this means that the resources that would normally be released
by netdev->destructor() will not be.
Over the years drivers have added local hacks to deal with this, by
invoking their destructor parts by hand when register_netdevice()
fails.
Many drivers do not try to deal with this, and instead we have leaks.
Let's close this hole by formalizing the distinction between what
private things need to be freed up by netdev->destructor() and whether
the driver needs unregister_netdevice() to perform the free_netdev().
netdev->priv_destructor() performs all actions to free up the private
resources that used to be freed by netdev->destructor(), except for
free_netdev().
netdev->needs_free_netdev is a boolean that indicates whether
free_netdev() should be done at the end of unregister_netdevice().
Now, register_netdevice() can sanely release all resources after
ndo_ops->ndo_init() succeeds, by invoking both ndo_ops->ndo_uninit()
and netdev->priv_destructor().
And at the end of unregister_netdevice(), we invoke
netdev->priv_destructor() and optionally call free_netdev().
Signed-off-by: David S. Miller <davem@davemloft.net>
