The correct name of the EESR bit 8 is TRO (transmit retry over), not RTO.
Note that EESIPR bit 8, TROIP remained correct...
Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Salil Mehta says:
====================
Misc. bug fixes and cleanup for HNS3 driver
This patch-set presents miscellaneous bug fixes and cleanups found
during internal review, system testing and cleanup.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
When vf module is loading, the cmd queue initialization should
happen before misc interrupt initialization, otherwise the misc
interrupt handle will cause using uninitialized cmd queue problem.
There is also the same issue when vf module is unloading.
This patch fixes it by adjusting the location of some function.
Fixes: e2cb1dec97 ("net: hns3: Add HNS3 VF HCL(Hardware Compatibility Layer) Support")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If CONFIG_ARM_SMMU_V3 is enabled, arm64's dma_ops will replace
arm64_swiotlb_dma_ops with iommu_dma_ops. When releasing contiguous
dma memory, the new ops will call the vunmap function which cannot
be run in interrupt context.
Currently, spin_lock_bh is called before vunmap is executed. This
disables BH and causes the interrupt context to be detected to
generate a kernel panic like below:
[ 2831.573400] kernel BUG at mm/vmalloc.c:1621!
[ 2831.577659] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
...
[ 2831.699907] Process rmmod (pid: 1893, stack limit = 0x0000000055103ee2)
[ 2831.706507] Call trace:
[ 2831.708941] vunmap+0x48/0x50
[ 2831.711897] dma_common_free_remap+0x78/0x88
[ 2831.716155] __iommu_free_attrs+0xa8/0x1c0
[ 2831.720255] hclge_free_cmd_desc+0xc8/0x118 [hclge]
[ 2831.725128] hclge_destroy_cmd_queue+0x34/0x68 [hclge]
[ 2831.730261] hclge_uninit_ae_dev+0x90/0x100 [hclge]
[ 2831.735127] hnae3_unregister_ae_dev+0xb0/0x868 [hnae3]
[ 2831.740345] hns3_remove+0x3c/0x90 [hns3]
[ 2831.744344] pci_device_remove+0x48/0x108
[ 2831.748342] device_release_driver_internal+0x164/0x200
[ 2831.753553] driver_detach+0x4c/0x88
[ 2831.757116] bus_remove_driver+0x60/0xc0
[ 2831.761026] driver_unregister+0x34/0x60
[ 2831.764935] pci_unregister_driver+0x30/0xb0
[ 2831.769197] hns3_exit_module+0x10/0x978 [hns3]
[ 2831.773715] SyS_delete_module+0x1f8/0x248
[ 2831.777799] el0_svc_naked+0x30/0x34
This patch fixes it by using spin_lock instead of spin_lock_bh.
Fixes: 68c0a5c706 ("net: hns3: Add HNS3 IMP(Integrated Mgmt Proc) Cmd Interface Support")
Signed-off-by: Xi Wang <wangxi11@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The link status update function is called by timer every second. But
net_stop and net_open may be called with very short intervals. The link
status update function can not detect the link state has changed. It
causes the netdev not running problem.
This patch fixes it by updating the link state in ae_stop function.
Fixes: 46a3df9f97 ("net: hns3: Add HNS3 Acceleration Engine & Compatibility Layer Support")
Signed-off-by: Fuyun Liang <liangfuyun1@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch does bit of a clean-up by using already defined enums for
certain values in function hclge_is_special_opcode(). Below enums from
have been used as replacements for magic values:
enum hclge_opcode_type{
<snip>
HCLGE_OPC_STATS_64_BIT = 0x0030,
HCLGE_OPC_STATS_32_BIT = 0x0031,
HCLGE_OPC_STATS_MAC = 0x0032,
<snip>
};
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If the hns3 driver has been built into kernel and then loaded with
the same driver which built as KLM, it may trigger an error like
below:
[ 20.009555] hns3: Hisilicon Ethernet Network Driver for Hip08 Family - version
[ 20.016789] hns3: Copyright (c) 2017 Huawei Corporation.
[ 20.022100] Error: Driver 'hns3' is already registered, aborting...
[ 23.517397] Unable to handle kernel NULL pointer dereference at virtual address 00000000
...
[ 23.691583] Process insmod (pid: 1982, stack limit = 0x00000000cd5f21cb)
[ 23.698270] Call trace:
[ 23.700705] __list_del_entry_valid+0x2c/0xd8
[ 23.705049] hnae3_unregister_client+0x68/0xa8
[ 23.709487] hns3_init_module+0x98/0x1000 [hns3]
[ 23.714093] do_one_initcall+0x5c/0x170
[ 23.717918] do_init_module+0x64/0x1f4
[ 23.721654] load_module+0x1d14/0x24b0
[ 23.725390] SyS_init_module+0x158/0x208
[ 23.729300] el0_svc_naked+0x30/0x34
This patch fixes it by adding module version info.
Fixes: 38caee9d3e ("net: hns3: Add support of the HNAE3 framework")
Signed-off-by: Xi Wang <wangxi11@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch fixes the missing initialization of the client list node
in the hnae3_register_client() function.
Fixes: 76ad4f0ee7 ("net: hns3: Add support of HNS3 Ethernet Driver for hip08 SoC")
Signed-off-by: Xi Wang <wangxi11@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Removes the goto and directly returns in case of errors as part of the
cleanup.
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
During the RX leg driver needs to fetch the ethernet header
length from the RX'ed Buffer Descriptor. Currently, proprietary
version hns3_nic_get_headlen is being used to fetch the header
length which uses l234info present in the Buffer Descriptor
which might not be valid for the first Buffer Descriptor if the
packet is spanning across multiple descriptors.
Kernel default eth_get_headlen API does the job correctly.
Fixes: 76ad4f0ee7 ("net: hns3: Add support of HNS3 Ethernet Driver for hip08 SoC")
Signed-off-by: Peng Li <lipeng321@huawei.com>
Reviewed-by: Yisen Zhuang <yisen.zhuang@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch fixes the error reported by Intel's kbuild and fixes a
return value in one of the legs, caught during review of the original
patch sent by kbuild.
Fixes: fdb793670a00 ("net: hns3: Add support of .sriov_configure in HNS3 driver")
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This commit removed calls to rtl_set_rx_mode(). This is ok for the
standard path if the link is brought up, however it breaks system
resume from suspend. Link comes up but no network traffic.
Meanwhile common code from rtl_hw_start_8169/8101/8168() was moved
to rtl_hw_start(), therefore re-add the call to rtl_set_rx_mode()
there.
Due to adding this call we have to move definition of rtl_hw_start()
after definition of rtl_set_rx_mode().
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Fixes: 82d3ff6dd1 ("r8169: remove calls to rtl_set_rx_mode")
Signed-off-by: David S. Miller <davem@davemloft.net>
Before the patch, the erspan BSO bit (Bad/Short/Oversized) is not
handled. BSO has 4 possible values:
00 --> Good frame with no error, or unknown integrity
11 --> Payload is a Bad Frame with CRC or Alignment Error
01 --> Payload is a Short Frame
10 --> Payload is an Oversized Frame
Based the short/oversized definitions in RFC1757, the patch sets
the bso bit based on the mirrored packet's size.
Reported-by: Xiaoyan Jin <xiaoyanj@vmware.com>
Signed-off-by: William Tu <u9012063@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Johan Hedberg says:
====================
pull request: bluetooth-next 2018-05-18
Here's the first bluetooth-next pull request for the 4.18 kernel:
- Refactoring of the btbcm driver
- New USB IDs for QCA_ROME and LiteOn controllers
- Buffer overflow fix if the controller sends invalid advertising data length
- Various cleanups & fixes for Qualcomm controllers
Please let me know if there are any issues pulling. Thanks.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
This reverts commit 6710f970d9.
Gotta love when developers have offline discussions, thinking everyone
is reading their responses/dialog.
The change had the potential for a number of race conditions on
shutdown, which is why we are reverting the change.
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently we use non-NUMA aware allocation for TPD and RRD buffers,
this patch modifies to use NUMA friendly allocation.
Signed-off-by: Hemanth Puranik <hpuranik@codeaurora.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Sergei Shtylyov says:
====================
Add Renesas R8A77980 GEther support
Here's a set of 3 patches against DaveM's 'net-next.git' repo. They (gradually)
add R8A77980 GEther support to the 'sh_eth' driver, starting with couple new
register bits/values introduced with this chip, and ending with adding a new
'struct sh_eth_cpu_data' instance connected to the new DT "compatible" prop
value...
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Finally, add support for the DT probing of the R-Car V3H (AKA R8A77980) --
it's the only R-Car gen3 SoC having the GEther controller -- others have
only EtherAVB...
Based on the original (and large) patch by Vladimir Barinov.
Signed-off-by: Vladimir Barinov <vladimir.barinov@cogentembedded.com>
Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Reviewed-by: Simon Horman <horms+renesas@verge.net.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
The R-Car V3H (AKA R8A77980) GEther controller adds the DMA burst mode bit
(NBST) in EDMR and the manual tells to always set it before doing any DMA.
Based on the original (and large) patch by Vladimir Barinov.
Signed-off-by: Vladimir Barinov <vladimir.barinov@cogentembedded.com>
Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Reviewed-by: Simon Horman <horms+renesas@verge.net.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
The R-Car V3H (AKA R8A77980) GEther controller adds support for the RGMII
PHY interface mode as a new value for the RMII_MII register.
Based on the original (and large) patch by Vladimir Barinov.
Signed-off-by: Vladimir Barinov <vladimir.barinov@cogentembedded.com>
Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Marvell PPv2 Header Parser sets some bits in the 'result_info' field in
each lookup iteration, to identify different packet attributes such as
DSA / VLAN tag, protocol infos, etc. This is used in further
classification stages in the controller.
It's the DSA tag detection entry that is in charge of detecting when there
is a single VLAN tag.
This commits adds the missing update of the result_info in this case.
Signed-off-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jiri Pirko says:
====================
devlink: introduce port flavours and common phys_port_name generation
This patchset resolves 2 issues we have right now:
1) There are many netdevices / ports in the system, for port, pf, vf
represenatation but the user has no way to see which is which
2) The ndo_get_phys_port_name is implemented in each driver separatelly,
which may lead to inconsistent names between drivers.
This patchset introduces port flavours which should address the first
problem. In this initial patchset, I focus on DSA and their port
flavours. As a follow-up, I plan to add PF and VF representor flavours.
However, that needs additional dependencies in drivers (nfp, mlx5).
The common phys_port_name generation is used by mlxsw. An example output
for mlxsw looks like this:
...
pci/0000:03:00.0/59: type eth netdev enp3s0np4 flavour physical number 4
pci/0000:03:00.0/61: type eth netdev enp3s0np1 flavour physical number 1
pci/0000:03:00.0/63: type eth netdev enp3s0np2 flavour physical number 2
pci/0000:03:00.0/49: type eth netdev enp3s0np8s0 flavour physical number 8 split_group 8 subport 0
pci/0000:03:00.0/50: type eth netdev enp3s0np8s1 flavour physical number 8 split_group 8 subport 1
pci/0000:03:00.0/51: type eth netdev enp3s0np8s2 flavour physical number 8 split_group 8 subport 2
pci/0000:03:00.0/52: type eth netdev enp3s0np8s3 flavour physical number 8 split_group 8 subport 3
As you can see, the netdev names are generated according to the flavour
and port number. In case the port is split, the split subnumber is also
included.
An example output for dsa_loop testing module looks like this:
mdio_bus/fixed-0:1f/0: type eth netdev lan1 flavour physical number 0
mdio_bus/fixed-0:1f/1: type eth netdev lan2 flavour physical number 1
mdio_bus/fixed-0:1f/2: type eth netdev lan3 flavour physical number 2
mdio_bus/fixed-0:1f/3: type eth netdev lan4 flavour physical number 3
mdio_bus/fixed-0:1f/4: type notset
mdio_bus/fixed-0:1f/5: type notset flavour cpu number 5
mdio_bus/fixed-0:1f/6: type notset
mdio_bus/fixed-0:1f/7: type notset
mdio_bus/fixed-0:1f/8: type notset
mdio_bus/fixed-0:1f/9: type notset
mdio_bus/fixed-0:1f/10: type notset
mdio_bus/fixed-0:1f/11: type notset
---
RFC->v1:
-removed nfp patches, removed DSA patch that used name generation helper
-patch 1:
- Reduced the nfp change just to simply use newly created attr_set func
-patch 2:
- rebased
- removed pf/vf reps flavours
-patch 3:
- rebased
-patch 4:
- added missing break pointed out by Andrew
====================
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Tested-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since devlink knows the info needed to generate the physical port name
in a generic way for all devlink users, use the helper to do the job.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Set the attrs and allow to expose port flavour to user via devlink.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Each driver implements physical port name generation by itself. However
as devlink has all needed info, it can easily do the job for all its
users. So implement this helper in devlink.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Devlink ports can have specific flavour according to the purpose of use.
This patch extend attrs_set so the driver can say which flavour port
has. Initial flavours are:
physical, cpu, dsa
User can query this to see right away what is the purpose of each port.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Change existing setter for split port information into more generic
attrs setter. Alongside with that, allow to set port number and subport
number for split ports.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some HW specific setups, like sun8i, do not populate all the necessary
callbacks, which is what HWIF helpers were expecting.
Fix this by always trying to get the generic helpers and populate them
if they were not previously populated by HW specific setup.
Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Fixes: 5f0456b431 ("net: stmmac: Implement logic to automatically
select HW Interface")
Reported-by: Corentin Labbe <clabbe.montjoie@gmail.com>
Tested-by: Corentin Labbe <clabbe.montjoie@gmail.com>
Cc: Corentin Labbe <clabbe.montjoie@gmail.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For T6, collect info on queue mapping to corresponding PF/VF in SGE.
Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch on the Marvell PPv2 driver is only cosmetic. Two typos are
removed as well as other cosmetic fixes, such as extra new lines or tabs
vs spaces.
Suggested-by: Stefan Chulski <stefanc@marvell.com>
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Trivial fix to spelling mistake in printk message text
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fixes: 20b654dfe1 ("tcp: support DUPACK threshold in RACK")
Signed-off-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Ursula Braun says:
====================
net/smc: cleanups 2018-05-18
here are SMC patches for net-next providing restructuring and cleanup
in different areas.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch splits up the functions smc_connect_rdma and smc_listen_work
into smaller functions.
Signed-off-by: Hans Wippel <hwippel@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch changes the function smc_buf_free to use the SMC link group
instead of the link as function parameter. Also, it changes the order of
the other two parameters.
Signed-off-by: Hans Wippel <hwippel@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch consists of Christmas tree fixes and removal of an unneeded
function parameter.
Signed-off-by: Hans Wippel <hwippel@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch moves a CDC sanity check from smc_cdc_msg_recv_action() to
the other sanity checks in smc_cdc_rx_handler(). While doing this, it
simplifies smc_cdc_msg_recv() and removes unneeded function parameters.
Signed-off-by: Hans Wippel <hwippel@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
SMC connection and buffer handling belong to smc_core. So, this patch
moves this code from smc.h to smc_core.
Signed-off-by: Hans Wippel <hwippel@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, the write offset within the RMB is calculated on each write
operation although it is fixed for each connection. With this patch, the
offset is calculated once and stored in a connection specific variable.
Signed-off-by: Hans Wippel <hwippel@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The connection index is actually a RMBE index. So, this patch changes
the name accordingly.
Signed-off-by: Hans Wippel <hwippel@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch moves the global link group list to smc_core where the link
group functions are. To make this work, it moves code in af_smc and
smc_ib that operates on the link group list to smc_core as well.
While at it, the link group counter is integrated into the list
structure and initialized to zero.
Signed-off-by: Hans Wippel <hwippel@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In addition to the buffer references, SMC currently stores the sizes of
the receive and send buffers in each connection as separate variables.
This patch introduces a buffer length variable in the common buffer
descriptor and uses this length instead.
Signed-off-by: Hans Wippel <hwippel@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
From: Or Gerlitz <ogerlitz@mellanox.com>
This series addresses a regression introduced by the
shared block TC changes [1]. Currently, for VF->VF and uplink->VF rules, the
TC core (cls_api) attempts to offload the same flow multiple times into
the driver, as a side effect of the mlx5 registration to the egdev callback.
We use the flow cookie to ignore attempts to add such flows, we can't
reject them (return error), b/c this will fail the offload attempt, so we
ignore that.
The last patch of the series deals with exposing HW stats counters through
ethtool for the vport reps.
Dave - the regression that we are addressing was introduced in 4.15 [1] and applies
to nfp and mlx5. Jiri suggested to push driver side fixes to net-next, this is
already done for nfp [2][3]. Once this is upstream, we will submit a small/point
single patch fix for the TC core code which can serve for net and stable, but not
carried into net-next, b/c it might limit some future use-cases.
[1] 208c0f4b52 "net: sched: use tc_setup_cb_call to call per-block callbacks"
[2] c50647d "nfp: flower: ignore duplicate cb requests for same rule"
[3] 54a4a03 "nfp: flower: support offloading multiple rules with same cookie"
-----BEGIN PGP SIGNATURE-----
iQEcBAABAgAGBQJa/iMHAAoJEEg/ir3gV/o+6YcIAMIcUmH0Yga2CQLl1VGZr4v7
5Yo5z8upZ2pKVlBNtgeDonIckcFNbtPaUC7xTolFmOks6DgoGKKLvrIyq3tNG+42
ShF91BmLvrpl/+8GmsjNf5qvsmc6piOHfBknlaIl7XeoaKLMfy4ts7/Cryt0U24k
meE/zu7slOOam6H2RyXKLsJa0uP/SpxrCq1OAlAwhmVe60p9SRCVkutmRqU47OuO
Oc3XGtYbU3IVa3B0bdi+SdOyF1RykCH3PKSrChy2WhdfpSp29I+gydfWMX8/3+z4
mH3/LDi4CAoHiCqnUr3s5h6zGuYqwcpSYY3tUPUD3A48LnKjT70LJF85F/v7exY=
=QeXI
-----END PGP SIGNATURE-----
Merge tag 'mlx5e-updates-2018-05-17' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux
Saeed Mahameed says:
====================
mlx5e-updates-2018-05-17
From: Or Gerlitz <ogerlitz@mellanox.com>
This series addresses a regression introduced by the
shared block TC changes [1]. Currently, for VF->VF and uplink->VF rules, the
TC core (cls_api) attempts to offload the same flow multiple times into
the driver, as a side effect of the mlx5 registration to the egdev callback.
We use the flow cookie to ignore attempts to add such flows, we can't
reject them (return error), b/c this will fail the offload attempt, so we
ignore that.
The last patch of the series deals with exposing HW stats counters through
ethtool for the vport reps.
Dave - the regression that we are addressing was introduced in 4.15 [1] and applies
to nfp and mlx5. Jiri suggested to push driver side fixes to net-next, this is
already done for nfp [2][3]. Once this is upstream, we will submit a small/point
single patch fix for the TC core code which can serve for net and stable, but not
carried into net-next, b/c it might limit some future use-cases.
[1] 208c0f4b52 "net: sched: use tc_setup_cb_call to call per-block callbacks"
[2] c50647d "nfp: flower: ignore duplicate cb requests for same rule"
[3] 54a4a03 "nfp: flower: support offloading multiple rules with same cookie"
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
mlx5 core dirver updates for both net-next and rdma-next branches.
From Christophe JAILLET, first three patche to use kvfree where needed.
From: Or Gerlitz <ogerlitz@mellanox.com>
Next six patches from Roi and Co adds support for merged
sriov e-switch which comes to serve cases where both PFs, VFs set
on them and both uplinks are to be used in single v-switch SW model.
When merged e-switch is supported, the per-port e-switch is logically
merged into one e-switch that spans both physical ports and all the VFs.
This model allows to offload TC eswitch rules between VFs belonging
to different PFs (and hence have different eswitch affinity), it also
sets the some of the foundations needed for uplink LAG support.
-----BEGIN PGP SIGNATURE-----
iQEcBAABAgAGBQJa/fLEAAoJEEg/ir3gV/o+7jUH/3n5/Uw1LLt3TfeKArx6i0F1
3G4U5B0ha03qiDqXprwhyQ3I6lgYmRBmjcxnqmvcqOAqO4/hSsjtTR+A/mgbEDhJ
YtdekFNEX+72h/N2GIpZwChIWSE3EcMPaLYnV8TwLUgh9YSust2sCLSBbJCjxOKc
j78M8ept/bXZwTm/iJhEjtmqw0xl91rl011chCAua0iEpH3wxteDARmKABFHMQxl
I3N/x/e/astgcSCNgpO4uDf9zEIRkNdzcHPzSMJ6C2Oo5W9XiZEekfw7WKj9nXfa
G+eGckkAyCOQ/r2lZ9nA0ZUvQ2X6JISvxgohuaCNwTgsz3acTxbLnQK4YWHzQCQ=
=iHi6
-----END PGP SIGNATURE-----
Merge tag 'mlx5-updates-2018-05-17' of git://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux
Saeed Mahameed says:
====================
mlx5-updates-2018-05-17
mlx5 core dirver updates for both net-next and rdma-next branches.
From Christophe JAILLET, first three patche to use kvfree where needed.
From: Or Gerlitz <ogerlitz@mellanox.com>
Next six patches from Roi and Co adds support for merged
sriov e-switch which comes to serve cases where both PFs, VFs set
on them and both uplinks are to be used in single v-switch SW model.
When merged e-switch is supported, the per-port e-switch is logically
merged into one e-switch that spans both physical ports and all the VFs.
This model allows to offload TC eswitch rules between VFs belonging
to different PFs (and hence have different eswitch affinity), it also
sets the some of the foundations needed for uplink LAG support.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet says:
====================
tcp: implement SACK compression
When TCP receives an out-of-order packet, it immediately sends
a SACK packet, generating network load but also forcing the
receiver to send 1-MSS pathological packets, increasing its
RTX queue length/depth, and thus processing time.
Wifi networks suffer from this aggressive behavior, but generally
speaking, all these SACK packets add fuel to the fire when networks
are under congestion.
This patch series adds SACK compression, but the infrastructure
could be leveraged to also compress ACK in the future.
v2: Addressed Neal feedback.
Added two sysctls to allow fine tuning, or even disabling the feature.
v3: take rtt = min(srtt, rcv_rtt) as Yuchung suggested, because rcv_rtt
can be over estimated for RPC (or sender limited)
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
This per netns sysctl allows for TCP SACK compression fine-tuning.
This limits number of SACK that can be compressed.
Using 0 disables SACK compression.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This per netns sysctl allows for TCP SACK compression fine-tuning.
Its default value is 1,000,000, or 1 ms to meet TSO autosizing period.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This counter tracks number of ACK packets that the host has not sent,
thanks to ACK compression.
Sample output :
$ nstat -n;sleep 1;nstat|egrep "IpInReceives|IpOutRequests|TcpInSegs|TcpOutSegs|TcpExtTCPAckCompressed"
IpInReceives 123250 0.0
IpOutRequests 3684 0.0
TcpInSegs 123251 0.0
TcpOutSegs 3684 0.0
TcpExtTCPAckCompressed 119252 0.0
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When TCP receives an out-of-order packet, it immediately sends
a SACK packet, generating network load but also forcing the
receiver to send 1-MSS pathological packets, increasing its
RTX queue length/depth, and thus processing time.
Wifi networks suffer from this aggressive behavior, but generally
speaking, all these SACK packets add fuel to the fire when networks
are under congestion.
This patch adds a high resolution timer and tp->compressed_ack counter.
Instead of sending a SACK, we program this timer with a small delay,
based on RTT and capped to 1 ms :
delay = min ( 5 % of RTT, 1 ms)
If subsequent SACKs need to be sent while the timer has not yet
expired, we simply increment tp->compressed_ack.
When timer expires, a SACK is sent with the latest information.
Whenever an ACK is sent (if data is sent, or if in-order
data is received) timer is canceled.
Note that tcp_sack_new_ofo_skb() is able to force a SACK to be sent
if the sack blocks need to be shuffled, even if the timer has not
expired.
A new SNMP counter is added in the following patch.
Two other patches add sysctls to allow changing the 1,000,000 and 44
values that this commit hard-coded.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Acked-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Toke Høiland-Jørgensen <toke@toke.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>
As explained in commit 9f9843a751 ("tcp: properly handle stretch
acks in slow start"), TCP stacks have to consider how many packets
are acknowledged in one single ACK, because of GRO, but also
because of ACK compression or losses.
We plan to add SACK compression in the following patch, we
must therefore not call tcp_enter_quickack_mode()
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Sergei Shtylyov