| Age | Commit message (Collapse) | Author |
|---|
|
Correct a single typo in a comment.
Misspelling found using checkpatch.pl.
Signed-off-by: R Veera Kumar <vkor@vkten.in>
Link: https://lore.kernel.org/r/20200319093301.GA2453@tulip.local
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Correct a single typo in a comment.
Misspelling found using checkpatch.pl.
Signed-off-by: R Veera Kumar <vkor@vkten.in>
Link: https://lore.kernel.org/r/20200319093715.GA2550@tulip.local
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Correct a single typo in a comment.
Misspelling found using checkpatch.pl.
Signed-off-by: R Veera Kumar <vkor@vkten.in>
Link: https://lore.kernel.org/r/20200319094043.GA2669@tulip.local
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Correct a single typo in a comment.
Misspelling found using checkpatch.pl.
Signed-off-by: R Veera Kumar <vkor@vkten.in>
Link: https://lore.kernel.org/r/20200319094358.GA2751@tulip.local
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Correct a long line in documentation to respect the
80 character line limit.
Found using checkpatch.pl.
Signed-off-by: R Veera Kumar <vkor@vkten.in>
Link: https://lore.kernel.org/r/20200319085751.GA1928@tulip.local
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Cleanup checkpatch.pl WARNINGS: Missing a blank line after declarations.
Signed-off-by: Sam Muhammed <jane.pnx9@gmail.com>
Link: https://lore.kernel.org/r/20200319132135.3362-1-jane.pnx9@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
This change supports limited multiple device modes by:
- At most 4 ports contains OTG/Device capability.
- One port run as device mode at a time.
Signed-off-by: Nagarjuna Kristam <nkristam@nvidia.com>
Acked-by: Felipe Balbi <balbi@kernel.org>
Signed-off-by: Thierry Reding <treding@nvidia.com>
|
|
When device mode is set/unset, VBUS override activity is done via
exported functions from padctl driver. Use phy_set_mode() instead.
Signed-off-by: Nagarjuna Kristam <nkristam@nvidia.com>
Acked-by: Felipe Balbi <balbi@kernel.org>
Signed-off-by: Thierry Reding <treding@nvidia.com>
|
|
usb-phy is used to get notified on the USB role changes. Get usb-phy from
the UTMI PHY.
Signed-off-by: Nagarjuna Kristam <nkristam@nvidia.com>
Acked-by: Felipe Balbi <balbi@kernel.org>
Signed-off-by: Thierry Reding <treding@nvidia.com>
|
|
Padctl driver will act as a central driver to receive USB role changes via
usb-role-switch. This is updated to corresponding host, device drivers.
Hence remove usb-role-switch from XUDC driver.
Signed-off-by: Nagarjuna Kristam <nkristam@nvidia.com>
Acked-by: Felipe Balbi <balbi@kernel.org>
[treding@nvidia.com: rebase onto Greg's usb-next branch]
Signed-off-by: Thierry Reding <treding@nvidia.com>
|
|
Get usb-phy's for availbale USB 2 phys. Register id notifiers for available
usb-phy's to receive role change notifications. Perform PP for the received
role change usb ports.
Signed-off-by: Nagarjuna Kristam <nkristam@nvidia.com>
[treding@nvidia.com: rebase onto Greg's usb-next branch]
Signed-off-by: Thierry Reding <treding@nvidia.com>
|
|
I have hit the following build error:
armv7a-hardfloat-linux-gnueabi-ld: drivers/phy/tegra/xusb.o: in function `tegra_xusb_port_unregister':
xusb.c:(.text+0x2ac): undefined reference to `usb_remove_phy'
armv7a-hardfloat-linux-gnueabi-ld: drivers/phy/tegra/xusb.o: in function `tegra_xusb_setup_ports':
xusb.c:(.text+0xf30): undefined reference to `usb_add_phy_dev'
PHY_TEGRA_XUSB should select USB_PHY because it uses symbols defined in
the code enabled by that.
Fixes: 23babe30fb45d ("phy: tegra: xusb: Add usb-phy support")
Signed-off-by: Corentin Labbe <clabbe@baylibre.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
|
|
The device-managed allocation API doesn't work well with the life-cycle
of device objects. Since ports have device objects allocated within, it
can lead to situations where these devices need to stay around until
after their parent pad controller has been unbound from its driver. The
device-managed memory allocated for the port objects will, however, get
freed when the pad controller unbinds from the driver. This can cause
use-after-free errors down the road.
Note that the device is deleted as part of the driver unbind operation,
so there isn't much that can be done with it after that point, but the
memory still needs to stay around to ensure none of the references are
invalidated.
One situation where this arises is when a VBUS supply is associated with
a USB 2 or 3 port. When that supply is released using regulator_put() an
SRCU call will queue the release of the device link connecting the port
and the regulator after a grace period. This means that the regulator is
going to keep on to the last reference of the port device even after the
pad controller driver was unbound (which is when the memory backing the
port device is freed).
Fix this by allocating port objects using non-device-managed memory. Add
release callbacks for these objects so that their memory gets freed when
the last reference goes away. This decouples the port devices' lifetime
from the "active" lifetime of the pad controller (i.e. the time during
which the pad controller driver owns the device).
Signed-off-by: Thierry Reding <treding@nvidia.com>
|
|
Devices are created for each port of the XUSB pad controller. Each USB 2
and USB 3 port can potentially have an associated VBUS power supply that
needs to be removed when the device is removed.
Since port devices never bind to a driver, the driver core will not get
to perform the cleanup of device-managed resources that usually happens
on driver unbind.
Now, the driver core will also perform device-managed resource cleanup
for driver-less devices when they are released. However, when a device
link is created between the regulator and the port device, as part of
regulator_get(), the regulator takes a reference to the port device and
prevents it from being released unless regulator_put() is called, which
will never happen.
Avoid this by using the non-device-managed API and manually releasing
the regulator reference when the port is unregistered.
Signed-off-by: Thierry Reding <treding@nvidia.com>
|
|
Probe deferral is an expected error condition that will usually be
recovered from. Print such error messages at debug level to make them
available for diagnostic purposes when building with debugging enabled
and hide them otherwise to not spam the kernel log with them.
Signed-off-by: Thierry Reding <treding@nvidia.com>
|
|
Deferred probe is an expected return value for tegra_fuse_readl().
Given that the driver deals with it properly, there's no need to
output a warning that may potentially confuse users.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Acked-by: Kishon Vijay Abraham I <kishon@ti.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
|
|
Add support for the XUSB pad controller found on Tegra194 SoCs. It is
mostly similar to the same IP found on Tegra186, but the number of
pads exposed differs, as do the programming sequences. Because most of
the Tegra194 XUSB PADCTL registers definition and programming sequence
are the same as Tegra186, Tegra194 XUSB PADCTL can share the same
driver, xusb-tegra186.c, with Tegra186 XUSB PADCTL.
Tegra194 XUSB PADCTL supports up to USB 3.1 Gen 2 speed, however, it
is possible for some platforms have long signal trace that could not
provide sufficient electrical environment for Gen 2 speed. This patch
adds a "maximum-speed" property to usb3 ports which can be used to
specify the maximum supported speed for any particular USB 3.1 port.
For a port that is not capable of SuperSpeedPlus, "maximum-speed"
property should carry "super-speed".
Signed-off-by: JC Kuo <jckuo@nvidia.com>
Acked-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
|
|
As xusb-tegra186.c will be reused for Tegra194, it would be good to
protect Tegra186 soc data with CONFIG_ARCH_TEGRA_186_SOC. This commit
also reshuffles Tegra186 soc data single CONFIG_ARCH_TEGRA_186_SOC
will be sufficient.
Signed-off-by: JC Kuo <jckuo@nvidia.com>
Acked-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
|
|
Add support for set_mode on UTMI phy. This allow XUSB host/device mode
drivers to configure the hardware to corresponding modes.
Signed-off-by: Nagarjuna Kristam <nkristam@nvidia.com>
Acked-by: Kishon Vijay Abraham I <kishon@ti.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
|
|
Add support for set_mode on USB 2 phy. This allow XUSB host/device mode
drivers to configure the hardware to corresponding modes.
Signed-off-by: Nagarjuna Kristam <nkristam@nvidia.com>
Acked-by: Kishon Vijay Abraham I <kishon@ti.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
|
|
Tegra XUSB host, device mode driver requires the USB 3 companion port
number for corresponding USB 2 port. Add API to retrieve the same.
Signed-off-by: Nagarjuna Kristam <nkristam@nvidia.com>
Reviewed-by: JC Kuo <jckuo@nvidia.com>
Acked-by: Kishon Vijay Abraham I <kishon@ti.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
|
|
For USB 2 ports that has usb-role-switch enabled, add usb-phy for
corresponding USB 2 phy. USB role changes from role switch are then
updated to corresponding host and device mode drivers via usb-phy notifier
block.
Signed-off-by: Nagarjuna Kristam <nkristam@nvidia.com>
Acked-by: Kishon Vijay Abraham I <kishon@ti.com>
[treding@nvidia.com: rebase onto Greg's usb-next branch]
Signed-off-by: Thierry Reding <treding@nvidia.com>
|
|
If usb-role-switch property is present in USB 2 port, register
usb-role-switch to receive usb role changes.
Signed-off-by: Nagarjuna Kristam <nkristam@nvidia.com>
Acked-by: Kishon Vijay Abraham I <kishon@ti.com>
[treding@nvidia.com: rebase onto Greg's usb-next branch]
Signed-off-by: Thierry Reding <treding@nvidia.com>
|
|
access
Before GICv4.1, all operations would be serialized with the affinity
changes by virtue of using the same ITS command queue. With v4.1, things
change, as invalidations (and a number of other operations) are issued
using the redistributor MMIO frame.
We must thus make sure that these redistributor accesses cannot race
against aginst the affinity change, or we may end-up talking to the
wrong redistributor.
To ensure this, we expand the irq_to_cpuid() helper to take a spinlock
when the LPI is mapped to a vLPI (a new per-VPE lock) on each operation
that requires mutual exclusion.
Signed-off-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Zenghui Yu <yuzenghui@huawei.com>
Link: https://lore.kernel.org/r/20200304203330.4967-4-maz@kernel.org
|
|
In a system that is only sparsly populated with CPUs, we can end-up with
redistributors structures that are not initialized. Let's make sure we
don't try and access those when iterating over them (in this case when
checking we have a L2 VPE table).
Fixes: 4e6437f12d6e ("irqchip/gic-v4.1: Ensure L2 vPE table is allocated at RD level")
Signed-off-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Zenghui Yu <yuzenghui@huawei.com>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Link: https://lore.kernel.org/r/20200304203330.4967-3-maz@kernel.org
|
|
To allow the direct injection of SGIs into a guest, the GICv4.1
architecture has to sacrifice the Active state so that SGIs look
a lot like LPIs (they are injected by the same mechanism).
In order not to break existing software, the architecture gives
offers guests OSs the choice: SGIs with or without an active
state. It is the hypervisors duty to honor the guest's choice.
For this, the architecture offers a discovery bit indicating whether
the GIC supports GICv4.1 SGIs (GICD_TYPER2.nASSGIcap), and another
bit indicating whether the guest wants Active-less SGIs or not
(controlled by GICD_CTLR.nASSGIreq).
A hypervisor not supporting GICv4.1 SGIs would leave nASSGIcap
clear, and a guest not knowing about GICv4.1 SGIs (or definitely
wanting an Active state) would leave nASSGIreq clear (both being
thankfully backward compatible with older revisions of the GIC).
Since Linux is perfectly happy without an active state on SGIs,
inform the hypervisor that we'll use that if offered.
Signed-off-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Zenghui Yu <yuzenghui@huawei.com>
Link: https://lore.kernel.org/r/20200304203330.4967-2-maz@kernel.org
|
|
This reverts commit 8ba88804bb3b877c841bc1864a8605111580cd0b as a better
version is already in Rafael's tree, sorry about that.
Reported-by: "Rafael J. Wysocki" <rjw@rjwysocki.net>
Cc: Madhuparna Bhowmik <madhuparnabhowmik10@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
https://git.linaro.org/people/daniel.lezcano/linux into timers/core
Pull clockevent/clocksource updates from Daniel Lezcano:
- Avoid creating dead devices by flagging the driver with OF_POPULATED
in order to prevent the platform to create another device (Saravana Kannan)
- Remove unused includes from imx family drivers (Anson Huang)
- timer-dm-ti rework to prepare for pwm and suspend support (Lokesh Vutla)
- Fix the rate for the global clock on the pit64b (Claudiu Beznea)
- Fix timer-cs5535 by requesting an irq with non-NULL dev_id (Afzal Mohammed)
- Replace setup_irq() by request_irq() (Afzal Mohammed)
- Add support for the TCU of X1000 (Zhou Yanjie)
- Drop the bogus omap_dm_timer_of_set_source() function (Suman Anna)
- Do not update the counter when updating the period in order to
prevent a disruption when the pwm is used (Lokesh Vutla)
- Improve owl_timer_init() failure messages (Matheus Castello)
- Add driver for the Ingenic JZ47xx OST (Maarten ter Huurne)
- Pass the interrupt and the shutdown callbacks in the init function
for ast2600 support (Joel Stanley)
- Add the ast2600 compatible string for the fttmr010 (Joel Stanley)
|
|
Enclose the chained handler with chained_irq_{enter,exit}(), so that the
muxed interrupts get properly acked.
This patch also fixes a reboot bug on OX820 SoC, where the jiffies timer
interrupt is never acked. The kernel waits a clock tick forever in
calibrate_delay_converge(), which leads to a boot hang.
Fixes: c41b16f8c9d9 ("ARM: integrator/versatile: consolidate FPGA IRQ handling code")
Signed-off-by: Sungbo Eo <mans0n@gorani.run>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20200319023448.1479701-1-mans0n@gorani.run
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux
Saeed Mahameed says:
====================
mlx5-updates-2020-03-17
1) Compiler warnings and cleanup for the connection tracking series
2) Bug fixes for the connection tracking series
3) Fix devlink port register sequence
4) Last five patches in the series, By Eli cohen
Add the support for forwarding traffic between two eswitch uplink
representors (Hairpin for eswitch), using mlx5 termination tables
to change the direction of a packet in hw from RX to TX pipeline.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
We precompute the static-static ECDH during configuration time, in order
to save an expensive computation later when receiving network packets.
However, not all ECDH computations yield a contributory result. Prior,
we were just not letting those peers be added to the interface. However,
this creates a strange inconsistency, since it was still possible to add
other weird points, like a valid public key plus a low-order point, and,
like points that result in zeros, a handshake would not complete. In
order to make the behavior more uniform and less surprising, simply
allow all peers to be added. Then, we'll error out later when doing the
crypto if there's an issue. This also adds more separation between the
crypto layer and the configuration layer.
Discussed-with: Mathias Hall-Andersen <mathias@hall-andersen.dk>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The situation in which we wind up hitting the default case here
indicates a major bug in earlier parsing code. It is not a usual thing
that should ever happen, which means a "friendly" message for it doesn't
make sense. Rather, replace this with a WARN_ON, just like we do earlier
in the file for a similar situation, so that somebody sends us a bug
report and we can fix it.
Reported-by: Fabian Freyer <fabianfreyer@radicallyopensecurity.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
We carry out checks to the effect of:
if (skb->protocol != wg_examine_packet_protocol(skb))
goto err;
By having wg_skb_examine_untrusted_ip_hdr return 0 on failure, this
means that the check above still passes in the case where skb->protocol
is zero, which is possible to hit with AF_PACKET:
struct sockaddr_pkt saddr = { .spkt_device = "wg0" };
unsigned char buffer[5] = { 0 };
sendto(socket(AF_PACKET, SOCK_PACKET, /* skb->protocol = */ 0),
buffer, sizeof(buffer), 0, (const struct sockaddr *)&saddr, sizeof(saddr));
Additional checks mean that this isn't actually a problem in the code
base, but I could imagine it becoming a problem later if the function is
used more liberally.
I would prefer to fix this by having wg_examine_packet_protocol return a
32-bit ~0 value on failure, which will never match any value of
skb->protocol, which would simply change the generated code from a mov
to a movzx. However, sparse complains, and adding __force casts doesn't
seem like a good idea, so instead we just add a simple helper function
to check for the zero return value. Since wg_examine_packet_protocol
itself gets inlined, this winds up not adding an additional branch to
the generated code, since the 0 return value already happens in a
mergable branch.
Reported-by: Fabian Freyer <fabianfreyer@radicallyopensecurity.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
git://anongit.freedesktop.org/drm/drm-misc into drm-fixes
- drm/lease: fix WARNING in idr_destroy
- Fix AVI frame colorimetry in the dw-hdmi bridge.
- Fix compiler warning in komeda by annotating functions as __maybe_unused.
- Downgrade bochs pci_request_region failure from error to warning to
workaround firmware fb.
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/7654ac39-deb8-c9ca-9fd5-ef77b2636380@linux.intel.com
|
|
At least some integrated PHY's in RTL8168/RTL8125 chip versions support
downshift, and the actual link speed can be read from a vendor-specific
register. Info about this register was provided by Realtek.
More details about downshift configuration (e.g. number of attempts)
aren't available, therefore the downshift tunable is not implemented.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Implement occupancy counting for counters and expose over devlink
resource API.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Put all init operations related to subpools into
mlxsw_sp_counter_sub_pools_init().
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Move the validation of subpools configuration, to avoid possible over
commitment to resource registration. Add WARN_ON to indicate bug
in the code.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Implement devlink resources support for counter pools. Move the subpool
sizes calculations into the new resources register function.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
query entry size
Add new field to subpool struct that would indicate which
resource id should be used to query the entry size for
the subpool from the device.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Currently, the global static array of subpools is used. Make it
per-instance as multiple instances of the mlxsw driver can have
different values.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The bank size is different between Spectrum versions. Also it is
a resource that can be queried. So instead of hard coding the value in
code, query it from the firmware.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
gro_cells_init() returns error if memory allocation is failed.
But the vxlan module doesn't check the return value of gro_cells_init().
Fixes: 58ce31cca1ff ("vxlan: GRO support at tunnel layer")`
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Chelsio NICs have 3 filter regions, in following order of priority:
1. High Priority (HPFILTER) region (Highest Priority).
2. HASH region.
3. Normal FILTER region (Lowest Priority).
Currently, there's a 1-to-1 mapping between the prio value passed
by TC and the filter region index. However, it's possible to have
multiple TC rules with the same prio value. In this case, if a region
is exhausted, no attempt is made to try inserting the rule in the
next available region.
So, rework and remove the 1-to-1 mapping. Instead, dynamically select
the region to insert the filter rule, as long as the new rule's prio
value doesn't conflict with existing rules across all the 3 regions.
Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
To check whether a netdevice has already been registered, look at
NETREG_REGISTERED to replace some hacks I added a while ago.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
qeth_do_ioctl() is only reached through our own net_device_ops, so we
can trust that dev->ml_priv still contains what we put there earlier.
qeth_bridgeport_an_set() is an internal function that doesn't require
such sanity checks.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Data addresses in the AOB are absolute, and need to be translated before
being fed into kmem_cache_free(). Currently this phys_to_virt() is a no-op.
Also see commit 2db01da8d25f ("s390/qdio: fill SBALEs with absolute addresses").
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Versions are meaningless for an in-kernel driver.
Instead use the UTS_RELEASE that is set by ethtool_get_drvinfo().
Cc: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This adds support for SOF_TIMESTAMPING_TX_SOFTWARE.
No support for non-IQD devices, since they orphan the skb in their xmit
path.
To play nice with TX bulking, set the timestamp when the buffer that
contains the skb(s) is actually flushed out to HW.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
For ucast traffic, qeth_iqd_select_queue() falls back to
netdev_pick_tx(). This will potentially use skb_tx_hash() to distribute
the flow over all active TX queues - so txq 0 is a valid selection, and
qeth_iqd_select_queue() needs to check for this and put it on some other
queue. As a result, the distribution for ucast flows is unbalanced and
hits QETH_IQD_MIN_UCAST_TXQ heavier than the other queues.
Open-coding a custom variant of skb_tx_hash() isn't an option, since
netdev_pick_tx() also gives us eg. access to XPS. But we can pull a
little trick: add a single TC class that excludes the mcast txq, and
thus encourage skb_tx_hash() to not pick the mcast txq.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
