NAPI tx mode improves TCP behavior by enabling TCP small queues (TSQ).
TSQ reduces queuing ("bufferbloat") and burstiness.
Previous measurements have shown significant improvement for
TCP_STREAM style workloads. Such as those in commit 86a5df1495
("Merge branch 'virtio-net-tx-napi'").
There has been uncertainty about smaller possible regressions in
latency due to increased reliance on tx interrupts.
The above results did not show that, nor did I observe this when
rerunning TCP_RR on Linux 5.1 this week on a pair of guests in the
same rack. This may be subject to other settings, notably interrupt
coalescing.
In the unlikely case of regression, we have landed a credible runtime
solution. Ethtool can configure it with -C tx-frames [0|1] as of
commit 0c465be183 ("virtio_net: ethtool tx napi configuration").
NAPI tx mode has been the default in Google Container-Optimized OS
(COS) for over half a year, as of release M70 in October 2018,
without any negative reports.
Link: https://marc.info/?l=linux-netdev&m=149305618416472
Link: https://lwn.net/Articles/507065/
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch does following things:
1. add the keyword const before some variables which won't be modified
in functions.
2. changes some variables from signed to unsigned to avoid bitwise
operation on signed variables.
3. adds or removes initialization of some variables.
4. defines a new structure to help parsing mailbox messages instead of
using an array which is harder to get the meaning of each element.
Signed-off-by: Weihang Li <liweihang@hisilicon.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Yufeng Mo <moyufeng@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The driver may not be able to disable the ring through firmware
when downing the netdev during reset process, which may cause
hardware accessing freed buffer problem.
This patch delays the ring buffer clearing to reset uninit
process because hardware will not access the ring buffer after
hardware reset is completed.
Fixes: bb6b94a896 ("net: hns3: Add reset interface implementation in client")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If hns3_nic_net_xmit does not return NETDEV_TX_BUSY when doing
a loopback selftest, the skb is not freed in hns3_clean_tx_ring
or hns3_nic_net_xmit, which causes skb not freed problem.
This patch fixes it by freeing skb when hns3_nic_net_xmit does
not return NETDEV_TX_OK.
Fixes: c39c4d98dc ("net: hns3: Add mac loopback selftest support in hns3 driver")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The netdev is dereferenced before null checking in the function
hns3_setup_tc.
This patch moves the dereferencing after the null checking.
Fixes: 76ad4f0ee7 ("net: hns3: Add support of HNS3 Ethernet Driver for hip08 SoC")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In hns3_nic_init_irq(), if request irq fail at index i,
the function return directly without releasing irq resources
that already requested, and nowhere else will release them.
Signed-off-by: Yonglong Liu <liuyonglong@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
IMP will set restting state for all function when PF FLR, driver
just clear the restting state in resetting progress, but don't do
it in initializing progress. As FLR is not created by driver,
it is necessary to clear restting state when initializing HW device.
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Function hclge_handle_all_hw_msix_error() contains four parts:
1. Query buffer descriptors for MSI-X errors.
2. Query and clear all main PF MSI-X errors.
3. Query and clear all PF MSI-X errors.
4. Handle mac tunnel interrupts.
Part 2 and part 3 handle errors of some different modules respectively,
this patch extracts them into dividual functions, which makes the logic
clearer.
Signed-off-by: Weihang Li <liweihang@hisilicon.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch modifies print message of rx_q_search_miss from error to dfx to
prevent misleading users, because this interrupt may occur if we receive
packets during initialization of HNS3 driver.
Otherwise, this patch masks 28th bit of PPU_MPF_ABNORMAL_SRC2 which is now
meaningless.
Signed-off-by: Weihang Li <liweihang@hisilicon.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Presently the HNS driver enables the HNS H/W error interrupts after
the dev initialization is completed. However some exceptions such as
NCSI errors can occur when the network port driver is not loaded
and those errors required reporting to the BMC.
Therefore the firmware enabled all the HNS ras error interrupts
before the driver is loaded. And in some cases, there will be some
H/W errors remained unclear before reboot. Thus the HNS driver needs
to process and recover those hw errors occurred before HNS driver is
initialized.
This patch adds processing of the HNS hw errors(RAS and MSI-X)
which occurred before the driver initialization. For RAS, because
they are enabled by firmware, so we can detect specific bits, then
log and clear them. But for MSI-X which can not be enabled before
open vector0 irq, we can't detect the specific error bits, so we
just write 1 to all interrupt source registers to clear.
Signed-off-by: Shiju Jose <shiju.jose@huawei.com>
Signed-off-by: Weihang Li <liweihang@hisilicon.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
HNS does not need to be reset when errors occur in some bits.
However presently the HNAE3_FUNC_RESET is set in this case and
as a result the default_reset is done when these errors are reported.
This patch fix this issue. Also patch does some optimization
in setting the reset level for the error recovery.
Reported-by: Weihang Li <liweihang@hisilicon.com>
Signed-off-by: Shiju Jose <shiju.jose@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Presently the error handling code sets the reset level required
for the recovery of the hw errors to the reset framework in the
error_detected AER callback. However the rest_event would be
called later from the slot_reset callback. This can cause issue
of using the wrong reset_level if a high priority reset request
occur before the slot_reset is called.
This patch delays setting of the reset level, required
for the hw errors, to the reset framework until the
slot_reset is called.
Reported-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: Shiju Jose <shiju.jose@huawei.com>
Signed-off-by: Weihang Li <liweihang@hisilicon.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The default window size is calculated for best performance based
on internal hw buffer sizes. The size differs between the
different chips and modes.
Fixes: 67b40dccc4 ("qed: Implement iWARP initialization, teardown and qp operations")
Signed-off-by: Ariel Elior <ariel.elior@marvell.com>
Signed-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The driver needs to assign a lossless traffic class for the MPA ll2
connection to ensure no packets are dropped when returning from the
driver as they will never be re-transmitted by the peer.
Fixes: ae3488ff37 ("qed: Add ll2 connection for processing unaligned MPA packets")
Signed-off-by: Ariel Elior <ariel.elior@marvell.com>
Signed-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Destroy QP waits for it's ep object state to be set to CLOSED
before proceeding. ep->state can be updated from a different
context. Add smp_store_release/READ_ONCE to synchronize.
Fixes: fc4c6065e6 ("qed: iWARP implement disconnect flows")
Signed-off-by: Ariel Elior <ariel.elior@marvell.com>
Signed-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The acpi_node_get_property_reference() doesn't return ACPI error codes,
it just returns regular negative kernel error codes. This patch doesn't
affect run time, it's just a clean up.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Ruslan Babayev <ruslan@babayev.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Apparently there are still cards in the wild with a very old
management FW. Let's make the error message in that case
indicate more clearly that management firmware has to be
updated.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The KSZ9477 series chips have a SYNCLKO pin which by default outputs a
25MHz clock, but some board setups require a 125MHz clock instead. Added
a microchip,synclko-125 device tree property to allow indicating a
125MHz clock output is required.
Signed-off-by: Robert Hancock <hancock@sedsystems.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>
The Silicon Errata and Data Sheet Clarification documents for the
KSZ9477 series of chips describe a number of otherwise undocumented PHY
register settings which are required to work around various chip errata.
Apply these settings when initializing the PHY ports on these chips.
Signed-off-by: Robert Hancock <hancock@sedsystems.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>
Switch stmmac_mdio_reset to use GPIO descriptors. GPIO core handles the
"snps,reset-gpio" for GPIO descriptors so we don't need to take care of
it inside the driver anymore.
The advantage of this is that we now preserve the GPIO flags which are
passed via devicetree. This is required on some newer Amlogic boards
which use an Open Drain pin for the reset GPIO. This pin can only output
a LOW signal or switch to input mode but it cannot output a HIGH signal.
There are already devicetree bindings for these special cases and GPIO
core already takes care of them but only if we use GPIO descriptors
instead of GPIO numbers.
Signed-off-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit "net: phy: Add detection of 1000BaseX link mode support" added
support for not filtering out 1000BaseX mode from the PHY's supported
modes in genphy_config_init, but we have to make a similar change in
genphy_read_abilities in order to actually detect it as a supported mode
in the first place. Add this in.
Signed-off-by: Robert Hancock <hancock@sedsystems.ca>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Only for consistency reasons, do it like in main cpsw.c module
and use ndev reference but not by means of slave.
Signed-off-by: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
We can avoid another indirect call per packet wrapping the rx
handler call with the proper helper.
To ensure that even the last listed direct call experience
measurable gain, despite the additional conditionals we must
traverse before reaching it, I tested reversing the order of the
listed options, with performance differences below noise level.
Together with the previous indirect call patch, this gives
~6% performance improvement in raw UDP tput.
v2 -> v3:
- use only the direct calls always available regardless of
the mlx5 build options
- drop the direct call list macro, to keep the code as simple
as possible for future rework
v1 -> v2:
- update the direct call list and use a macro to define it,
as per Saeed suggestion. An intermediated additional
macro is needed to allow arg list expansion
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We can avoid an indirect call per packet wrapping the skb creation
with the appropriate helper.
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Johannes Berg says:
====================
Many changes all over:
* HE (802.11ax) work continues
* WPA3 offloads
* work on extended key ID handling continues
* fixes to honour AP supported rates with auth/assoc frames
* nl80211 netlink policy improvements to fix some issues
with strict validation on new commands with old attrs
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
>From RTL8168e-vl the value in register MaxTxPacketSize is interpreted
differently, therefore use new helper rtl_is_8168evl_up to set this
register.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add helper rtl_is_8168evl_up to make the code better readable and to
simplify it.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There's no rate control algorithm that *doesn't* want to call
it internally, and calling it internally will let us modify
its behaviour in the future.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Require that each vendor command give a policy of its sub-attributes
in NL80211_ATTR_VENDOR_DATA, and then (stricly) check the contents,
including the NLA_F_NESTED flag that we couldn't check on the outer
layer because there we don't know yet.
It is possible to use VENDOR_CMD_RAW_DATA for raw data, but then no
nested data can be given (NLA_F_NESTED flag must be clear) and the
data is just passed as is to the command.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
tp->coalesce_info is used in rtl_coalesce_info() only, so we can
remove this member. In addition replace phy_ethtool_get_link_ksettings
with a direct access to tp->phydev->speed.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In case of a timeout currently ~0 is returned. Callers often just check
whether a certain bit is set and therefore may behave incorrectly.
So let's return -ETIMEDOUT in case of a timeout.
r8168_phy_ocp_read is used in r8168g_mdio_read only, therefore we can
apply the same change.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We were accessing the pointer returned from platform_get_resource before
checking if it was valid, causing an oops if it was not. Move this access
after the call to devm_ioremap_resource which does the validity check.
Signed-off-by: Robert Hancock <hancock@sedsystems.ca>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
We statically allocate 8 cmd buffers on the read channel, when the only
IO left that's still using them is the long-running READ.
Replace this with a single allocated cmd, that gets restarted whenever
the READ completed.
This introduces refcounting for allocated cmds, so that the READ cmd can
survive the IO completion.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The current IDX sequence first sends one WRITE cmd to activate the
device, and then sends a second cmd that READs the response.
Using qeth_alloc_cmd(), we can combine this into a single IO with two
command-chained CCWs.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The RCD code is the last remaining IO path that doesn't use the
qeth_send_control_data() infrastructure. Doing so allows us to remove
all sorts of custom state machinery and logic in the IRQ handler.
Instead of introducing statically allocated cmd buffers for this single
IO on the data channel, use the new qeth_alloc_cmd() helper.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
qeth currently uses a fixed set of statically allocated cmd buffers for
the read and write IO channels. This (1) doesn't play well with the single
RCD cmd we need to issue on the data channel, (2) doesn't provide the
necessary flexibility for certain IDX improvements, and (3) is also rather
wasteful since the buffers are idle most of the time.
Add a new type of cmd buffer that is dynamically allocated, and keeps
its ccw chain in the DMA data area. Since this touches most callers of
qeth_setup_ccw(), also add a new CCW flags parameter for future usage.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Each cmd buffer maintains a pointer to the IO channel that it was/will
be issued on. So when dealing with cmd buffers, we don't need to pass
around a separate channel pointer.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The vast majority of SETUP-classified trace entries can be moved to
their device-specific trace file. This reduces pollution of the global
SETUP file, and provides a consistent trace view of all activity on the
device.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
OSN currently provides a custom code path to submit IPA cmds, without
waiting for the cmd response. Replace it with qeth_send_ipa_cmd(), which
uses the common qeth_send_control_data() IO infrastructure.
By setting a custom iob->callback, we can now provide feedback to the
caller about whether the cmd has been successfully submitted to HW.
Since the callback then immediately wakes up the reply-waiter object, we
maintain the old behaviour of returning early without waiting for the
response.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The basic MPC initialization sequence is strictly sequential, and
waiting for an available cmd buffer should never be necessary.
So this change only affects the OSN path, where dangling waiters on an
unbounded wait_event() are not desirable. Switch to qeth_get_buffers(),
and let OSN callers deal with -ENOMEM.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When called from qeth_core_probe_device(), qeth_determine_capabilities()
initializes the device's BLKT defaults. From all other callers, the
ccw_device has already been set online and the BLKT setting is skipped.
Clean this up by extracting the BLKT setting into a separate helper that
gets called from the right place.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The completion of a pending READ cmd is processed via
qeth_issue_next_read_cb(). Let this callback also start the next READ
cmd, instead of hardcoding that step into the IRQ handler.
While at it remove the check of the channel state,
__qeth_issue_next_read() already does this.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When the tear down sequence in qeth_l?_stop_card() has finished, the
card is guaranteed to be in DOWN state and we don't have to check for
it again.
With this insight we can also remove the redundant setting of
card->state in qeth_l?_set_online()'s error path.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Slightly reduce the complexity of the core xmit path, by replacing some
open-coded logic with the corresponding helpers.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Current code suppresses debug entries when an TX buffer completes in
ERROR state with no error indication set in SBALF15.
This was introduced back with
commit 58490f1807 ("qeth: HiperSockets SIGA retry support on CC=2.").
But qeth no longer retries after CC=2, and this sort of suppression
make no sense anymore. Remove it.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
