| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
octeontx2 driver calls page_pool_create() during driver probe()
and fails if queue size > 32k. Page pool infra uses these buffers
as shock absorbers for burst traffic. These pages are pinned down
over time as working sets varies, due to the recycling nature
of page pool, given page pool (currently) don't have a shrinker
mechanism, the pages remain pinned down in ptr_ring.
Instead of clamping page_pool size to 32k at
most, limit it even more to 2k to avoid wasting memory.
This have been tested on octeontx2 CN10KA hardware.
TCP and UDP tests using iperf shows no performance regressions.
Fixes: b2e3406a38f0 ("octeontx2-pf: Add support for page pool")
Suggested-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Reviewed-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: Ratheesh Kannoth <rkannoth@marvell.com>
Acked-by: Jesper Dangaard Brouer <hawk@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The FEC driver supports the statistics for XDP actions except for
XDP_TX before, because the XDP_TX was not supported when adding
the statistics for XDP. Now the FEC driver has supported XDP_TX
since commit f601899e4321 ("net: fec: add XDP_TX feature support").
So it's reasonable and necessary to add statistics for XDP_TX.
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The ice hardware has a synchronization mechanism used to drive the
simultaneous application of commands on both PHY ports and the source timer
in the MAC.
When issuing a sync via ice_ptp_exec_tmr_cmd(), the hardware will
simultaneously apply the commands programmed for the main timer and each
PHY port. Neither the main timer command register, nor the PHY port command
registers auto clear on command execution.
During the execution of a timer command intended for a single port on E822
devices, such as those used to configure a PHY during link up, the driver
is not correctly clearing the previous commands.
This results in unintentionally executing the last programmed command on
the main timer and other PHY ports whenever performing reconfiguration on
E822 ports after link up. This results in unintended side effects on other
timers, depending on what command was previously programmed.
To fix this, the driver must ensure that the main timer and all other PHY
ports are properly initialized to perform no action.
The enumeration for timer commands does not include an enumeration value
for doing nothing. Introduce ICE_PTP_NOP for this purpose. When writing a
timer command to hardware, leave the command bits set to zero which
indicates that no operation should be performed on that port.
Modify ice_ptp_one_port_cmd() to always initialize all ports. For all ports
other than the one being configured, write their timer command register to
ICE_PTP_NOP. This ensures that no side effect happens on the timer command.
To fix this for the PHY ports, modify ice_ptp_one_port_cmd() to always
initialize all other ports to ICE_PTP_NOP. This ensures that no side
effects happen on the other ports.
Call ice_ptp_src_cmd() with a command value if ICE_PTP_NOP in
ice_sync_phy_timer_e822() and ice_start_phy_timer_e822().
With both of these changes, the driver should no longer execute a stale
command on the main timer or another PHY port when reconfiguring one of the
PHY ports after link up.
Fixes: 3a7496234d17 ("ice: implement basic E822 PTP support")
Signed-off-by: Siddaraju DH <siddaraju.dh@intel.com>
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Sunitha Mekala <sunithax.d.mekala@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Registers of mdio accessing are common defined in libwx, remove the
redundant macro definitions in ngbe driver.
Signed-off-by: Jiawen Wu <jiawenwu@trustnetic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Wangxun SP chip supports to connect with external PHY (marvell 88x3310),
which links to 10GBASE-T/1000BASE-T/100BASE-T. Add the identification of
media types from subsystem device IDs. For sp_media_copper, register mdio
bus for the external PHY.
Signed-off-by: Jiawen Wu <jiawenwu@trustnetic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Disable data path before PCS VR reset while switching PCS mode, to prevent
the blocking of data path. Enable AN interrupt for CL37 auto-negotiation.
Signed-off-by: Jiawen Wu <jiawenwu@trustnetic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Since XPCS device identifier is implemented in the firmware version
0x20010 and above, so add a warning to prompt the users to upgrade the
firmware to make sure the driver works.
Signed-off-by: Jiawen Wu <jiawenwu@trustnetic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
used as core library"
Recent merge had a conflict in:
drivers/net/ethernet/mellanox/mlx5/core/en_accel/macsec_fs.c
between commit:
aeb660171b06 ("net/mlx5e: fix double free in macsec_fs_tx_create_crypto_table_groups")
from Linus' tree and commit:
cb5ebe4896f9 ("net/mlx5e: Move MACsec flow steering operations to be used as core library")
from the mlx5-next tree. This was missed and the former commit
got lost, bring it back.
Fixes: 3c5066c6b0a5 ("Merge branch 'mlx5-next' of https://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux")
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Link: https://lore.kernel.org/r/20230815123725.4ef5b7b9@canb.auug.org.au
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Use a dynamic calculation to determine the shift value for the internal
timer cyclecounter that will lead to the highest precision frequency
adjustments. Previously used a constant for the shift value assuming all
devices supported by the driver had a nominal frequency of 1GHz. However,
there are devices that operate at different frequencies. The previous shift
value constant would break the PHC functionality for those devices.
Reported-by: Vadim Fedorenko <vadim.fedorenko@linux.dev>
Closes: https://lore.kernel.org/netdev/20230815151507.3028503-1-vadfed@meta.com/
Fixes: 6a4010927562 ("net/mlx5: Update cyclecounter shift value to improve ptp free running mode precision")
Signed-off-by: Rahul Rameshbabu <rrameshbabu@nvidia.com>
Tested-by: Vadim Fedorenko <vadim.fedorenko@linux.dev>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Acked-by: Saeed Mahameed <saeedm@nvidia.com>
Link: https://lore.kernel.org/r/20230821230554.236210-1-rrameshbabu@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Patch series "New page table range API", v6.
This patchset changes the API used by the MM to set up page table entries.
The four APIs are:
set_ptes(mm, addr, ptep, pte, nr)
update_mmu_cache_range(vma, addr, ptep, nr)
flush_dcache_folio(folio)
flush_icache_pages(vma, page, nr)
flush_dcache_folio() isn't technically new, but no architecture
implemented it, so I've done that for them. The old APIs remain around
but are mostly implemented by calling the new interfaces.
The new APIs are based around setting up N page table entries at once.
The N entries belong to the same PMD, the same folio and the same VMA, so
ptep++ is a legitimate operation, and locking is taken care of for you.
Some architectures can do a better job of it than just a loop, but I have
hesitated to make too deep a change to architectures I don't understand
well.
One thing I have changed in every architecture is that PG_arch_1 is now a
per-folio bit instead of a per-page bit when used for dcache clean/dirty
tracking. This was something that would have to happen eventually, and it
makes sense to do it now rather than iterate over every page involved in a
cache flush and figure out if it needs to happen.
The point of all this is better performance, and Fengwei Yin has measured
improvement on x86. I suspect you'll see improvement on your architecture
too. Try the new will-it-scale test mentioned here:
https://lore.kernel.org/linux-mm/20230206140639.538867-5-fengwei.yin@intel.com/
You'll need to run it on an XFS filesystem and have
CONFIG_TRANSPARENT_HUGEPAGE set.
This patchset is the basis for much of the anonymous large folio work
being done by Ryan, so it's received quite a lot of testing over the last
few months.
This patch (of 38):
Determine if a value lies within a range more efficiently (subtraction +
comparison vs two comparisons and an AND). It also has useful (under some
circumstances) behaviour if the range exceeds the maximum value of the
type. Convert all the conflicting definitions of in_range() within the
kernel; some can use the generic definition while others need their own
definition.
Link: https://lkml.kernel.org/r/20230802151406.3735276-1-willy@infradead.org
Link: https://lkml.kernel.org/r/20230802151406.3735276-2-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Add devices IDs for the next LOM generations that will be available on the
next Intel Client platforms. This patch provides the initial support for
these devices.
Signed-off-by: Sasha Neftin <sasha.neftin@intel.com>
Tested-by: Naama Meir <naamax.meir@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
With the 10us interval, we were seeing PTM transactions take around 12us.
Hardware team suggested this interval could be lowered to 1us which was
confirmed with PCIe sniffer. With the 1us interval, PTM dialogs took
around 2us.
Suggested-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Signed-off-by: Sasha Neftin <sasha.neftin@intel.com>
Tested-by: Muhammad Husaini Zulkifli <muhammad.husaini.zulkifli@intel.com>
Reviewed-by: Muhammad Husaini Zulkifli <muhammad.husaini.zulkifli@intel.com>
Tested-by: Naama Meir <naamax.meir@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
Add support for using the four sets of timestamping registers that
i225/i226 have available for TX.
In some workloads, where multiple applications request hardware
transmission timestamps, it was possible that some of those requests
were denied because the only in use register was already occupied.
This is also in preparation to future support for hardware
timestamping with multiple PTP domains. With multiple domains chances
of multiple TX timestamps being requested at the same time increase.
Before:
$ sudo ./ntpperf -i enp3s0 -m 10:22:22:22:22:21 -d 192.168.1.3 -s 172.18.0.0/16 -I -H -o 37
| responses | TX timestamp offset (ns)
rate clients | lost invalid basic xleave | min mean max stddev
1000 100 0.00% 0.00% 0.00% 100.00% +1 +41 +73 13
1500 150 0.00% 0.00% 0.00% 100.00% +9 +49 +87 15
2250 225 0.00% 0.00% 0.00% 100.00% +9 +42 +79 13
3375 337 0.00% 0.00% 0.00% 100.00% +11 +46 +81 13
5062 506 0.00% 0.00% 0.00% 100.00% +7 +44 +80 13
7593 759 0.00% 0.00% 0.00% 100.00% +9 +44 +79 12
11389 1138 0.00% 0.00% 0.00% 100.00% +14 +51 +87 13
17083 1708 0.00% 0.00% 0.00% 100.00% +1 +41 +80 14
25624 2562 0.00% 0.00% 0.00% 100.00% +11 +50 +5107 51
38436 3843 0.00% 0.00% 0.00% 100.00% -2 +36 +7843 38
57654 5765 0.00% 0.00% 0.00% 100.00% +4 +42 +10503 69
86481 8648 0.00% 0.00% 0.00% 100.00% +11 +54 +5492 65
129721 12972 0.00% 0.00% 0.00% 100.00% +31 +2680 +6942 2606
194581 16384 16.79% 0.00% 0.87% 82.34% +73 +4444 +15879 3116
291871 16384 35.05% 0.00% 1.53% 63.42% +188 +5381 +17019 3035
437806 16384 54.95% 0.00% 2.55% 42.50% +233 +6302 +13885 2846
After:
$ sudo ./ntpperf -i enp3s0 -m 10:22:22:22:22:21 -d 192.168.1.3 -s 172.18.0.0/16 -I -H -o 37
| responses | TX timestamp offset (ns)
rate clients | lost invalid basic xleave | min mean max stddev
1000 100 0.00% 0.00% 0.00% 100.00% -20 +12 +43 13
1500 150 0.00% 0.00% 0.00% 100.00% -23 +18 +57 14
2250 225 0.00% 0.00% 0.00% 100.00% -2 +33 +67 13
3375 337 0.00% 0.00% 0.00% 100.00% +1 +38 +76 13
5062 506 0.00% 0.00% 0.00% 100.00% +9 +52 +93 14
7593 759 0.00% 0.00% 0.00% 100.00% +11 +47 +82 13
11389 1138 0.00% 0.00% 0.00% 100.00% -9 +27 +74 13
17083 1708 0.00% 0.00% 0.00% 100.00% -13 +25 +66 14
25624 2562 0.00% 0.00% 0.00% 100.00% -8 +28 +65 13
38436 3843 0.00% 0.00% 0.00% 100.00% -13 +28 +69 13
57654 5765 0.00% 0.00% 0.00% 100.00% -11 +32 +71 14
86481 8648 0.00% 0.00% 0.00% 100.00% +2 +44 +83 14
129721 12972 15.36% 0.00% 0.35% 84.29% -2 +2248 +22907 4252
194581 16384 42.98% 0.00% 1.98% 55.04% -4 +5278 +65039 5856
291871 16384 54.33% 0.00% 2.21% 43.46% -3 +6306 +22608 5665
We can see that with 4 registers, as expected, we are able to handle a
increasing number of requests more consistently, but as soon as all
registers are in use, the decrease in quality of service happens in a
sharp step.
Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Muhammad Husaini Zulkifli <muhammad.husaini.zulkifli@intel.com>
Reviewed-by: Kurt Kanzenbach <kurt@linutronix.de>
Tested-by: Naama Meir <naamax.meir@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
https://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux
Leon Romanovsky says:
====================
mlx5 MACsec RoCEv2 support
From Patrisious:
This series extends previously added MACsec offload support
to cover RoCE traffic either.
In order to achieve that, we need configure MACsec with offload between
the two endpoints, like below:
REMOTE_MAC=10:70:fd:43:71:c0
* ip addr add 1.1.1.1/16 dev eth2
* ip link set dev eth2 up
* ip link add link eth2 macsec0 type macsec encrypt on
* ip macsec offload macsec0 mac
* ip macsec add macsec0 tx sa 0 pn 1 on key 00 dffafc8d7b9a43d5b9a3dfbbf6a30c16
* ip macsec add macsec0 rx port 1 address $REMOTE_MAC
* ip macsec add macsec0 rx port 1 address $REMOTE_MAC sa 0 pn 1 on key 01 ead3664f508eb06c40ac7104cdae4ce5
* ip addr add 10.1.0.1/16 dev macsec0
* ip link set dev macsec0 up
And in a similar manner on the other machine, while noting the keys order
would be reversed and the MAC address of the other machine.
RDMA traffic is separated through relevant GID entries and in case
of IP ambiguity issue - meaning we have a physical GIDs and a MACsec
GIDs with the same IP/GID, we disable our physical GID in order
to force the user to only use the MACsec GID.
v0: https://lore.kernel.org/netdev/20230813064703.574082-1-leon@kernel.org/
* 'mlx5-next' of https://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux:
RDMA/mlx5: Handles RoCE MACsec steering rules addition and deletion
net/mlx5: Add RoCE MACsec steering infrastructure in core
net/mlx5: Configure MACsec steering for ingress RoCEv2 traffic
net/mlx5: Configure MACsec steering for egress RoCEv2 traffic
IB/core: Reorder GID delete code for RoCE
net/mlx5: Add MACsec priorities in RDMA namespaces
RDMA/mlx5: Implement MACsec gid addition and deletion
net/mlx5: Maintain fs_id xarray per MACsec device inside macsec steering
net/mlx5: Remove netdevice from MACsec steering
net/mlx5e: Move MACsec flow steering and statistics database from ethernet to core
net/mlx5e: Rename MACsec flow steering functions/parameters to suit core naming style
net/mlx5: Remove dependency of macsec flow steering on ethernet
net/mlx5e: Move MACsec flow steering operations to be used as core library
macsec: add functions to get macsec real netdevice and check offload
====================
Link: https://lore.kernel.org/r/20230821073833.59042-1-leon@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Cross-merge networking fixes after downstream PR.
Conflicts:
include/net/inet_sock.h
f866fbc842de ("ipv4: fix data-races around inet->inet_id")
c274af224269 ("inet: introduce inet->inet_flags")
https://lore.kernel.org/all/679ddff6-db6e-4ff6-b177-574e90d0103d@tessares.net/
Adjacent changes:
drivers/net/bonding/bond_alb.c
e74216b8def3 ("bonding: fix macvlan over alb bond support")
f11e5bd159b0 ("bonding: support balance-alb with openvswitch")
drivers/net/ethernet/broadcom/bgmac.c
d6499f0b7c7c ("net: bgmac: Return PTR_ERR() for fixed_phy_register()")
23a14488ea58 ("net: bgmac: Fix return value check for fixed_phy_register()")
drivers/net/ethernet/broadcom/genet/bcmmii.c
32bbe64a1386 ("net: bcmgenet: Fix return value check for fixed_phy_register()")
acf50d1adbf4 ("net: bcmgenet: Return PTR_ERR() for fixed_phy_register()")
net/sctp/socket.c
f866fbc842de ("ipv4: fix data-races around inet->inet_id")
b09bde5c3554 ("inet: move inet->mc_loop to inet->inet_frags")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux
Saeed Mahameed says:
====================
mlx5-updates-2023-08-22
1) Patches #1..#13 From Jiri:
The goal of this patchset is to make the SF code cleaner.
Benefit from previously introduced devlink_port struct containerization
to avoid unnecessary lookups in devlink port ops.
Also, benefit from the devlink locking changes and avoid unnecessary
reference counting.
2) Patches #14,#15:
Add ability to configure proto both UDP and TCP selectors in RX and TX
directions.
* tag 'mlx5-updates-2023-08-22' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux:
net/mlx5e: Support IPsec upper TCP protocol selector
net/mlx5e: Support IPsec upper protocol selector field offload for RX
net/mlx5: Store vport in struct mlx5_devlink_port and use it in port ops
net/mlx5: Check vhca_resource_manager capability in each op and add extack msg
net/mlx5: Relax mlx5_devlink_eswitch_get() return value checking
net/mlx5: Return -EOPNOTSUPP in mlx5_devlink_port_fn_migratable_set() directly
net/mlx5: Reduce number of vport lookups passing vport pointer instead of index
net/mlx5: Embed struct devlink_port into driver structure
net/mlx5: Don't register ops for non-PF/VF/SF port and avoid checks in ops
net/mlx5: Remove no longer used mlx5_esw_offloads_sf_vport_enable/disable()
net/mlx5: Introduce mlx5_eswitch_load/unload_sf_vport() and use it from SF code
net/mlx5: Allow mlx5_esw_offloads_devlink_port_register() to register SFs
net/mlx5: Push devlink port PF/VF init/cleanup calls out of devlink_port_register/unregister()
net/mlx5: Push out SF devlink port init and cleanup code to separate helpers
net/mlx5: Rework devlink port alloc/free into init/cleanup
====================
Link: https://lore.kernel.org/all/20230823051012.162483-1-saeed@kernel.org/
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
Systems having 4 GiB of RAM and more require DMA addressing beyond the
current 32-bit limit. Starting from MT7988 the hardware now supports
36-bit DMA addressing, let's use that new capability in the driver to
avoid running into swiotlb on systems with 4 GiB of RAM or more.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Link: https://lore.kernel.org/r/95b919c98876c9e49761e44662e7c937479eecb8.1692721443.git.daniel@makrotopia.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
MT7981, MT7986 and MT7988 come with in-SoC SRAM dedicated for Ethernet
DMA rings. Support using the SRAM without breaking existing device tree
bindings, ie. only new SoC starting from MT7988 will have the SRAM
declared as additional resource in device tree. For MT7981 and MT7986
an offset on top of the main I/O base is used.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Link: https://lore.kernel.org/r/e45e0f230c63ad58869e8fe35b95a2fb8925b625.1692721443.git.daniel@makrotopia.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Add bits needed to reset the frame engine on MT7988.
Fixes: 445eb6448ed3 ("net: ethernet: mtk_eth_soc: add basic support for MT7988 SoC")
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Link: https://lore.kernel.org/r/89b6c38380e7a3800c1362aa7575600717bc7543.1692721443.git.daniel@makrotopia.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
More register macros need to be adjusted for the 3rd GMAC on MT7988.
Account for added bit in SYSCFG0_SGMII_MASK.
Fixes: 445eb6448ed3 ("net: ethernet: mtk_eth_soc: add basic support for MT7988 SoC")
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/1c8da012e2ca80939906d85f314138c552139f0f.1692721443.git.daniel@makrotopia.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
As we already added the exception tracing for XDP_TX, I think it is
necessary to add the exception tracing for other XDP actions, such
as XDP_REDIRECT, XDP_ABORTED and unknown error actions.
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Link: https://lore.kernel.org/r/20230822065255.606739-1-wei.fang@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Use the standard error pointer macro to shorten the code and simplify.
Signed-off-by: Yu Liao <liaoyu15@huawei.com>
Link: https://lore.kernel.org/r/20230822021455.205101-2-liaoyu15@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
When building for power4, newer binutils don't recognise the "dcbfl"
extended mnemonic.
dcbfl RA, RB is equivalent to dcbf RA, RB, 1.
Switch to "dcbf" to avoid the build error.
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add check for pf->vf not being NULL before dereferencing
pf->vf[vsi->vf_id] in updating VSI filter sync.
Add a similar check before dereferencing !pf->vf[vsi->vf_id].trusted
in the condition for clearing promisc mode bit.
Fixes: c87c938f62d8 ("i40e: Add VF VLAN pruning")
Signed-off-by: Andrii Staikov <andrii.staikov@intel.com>
Signed-off-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>
Tested-by: Rafal Romanowski <rafal.romanowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
All callers of build_skb() (*) in bnxt are in NAPI context.
The budget checking is somewhat convoluted because in the shared
completion queue cases Rx packets are discarded by netpoll
by forcing an error (E). But that happens before skb allocation.
Only a call chain starting at __bnxt_poll_work() can lead to
an skb allocation and it checks budget (b).
* bnxt_rx_multi_page_skb
* bnxt_rx_skb
` bp->rx_skb_func
* bnxt_tpa_end
` bnxt_rx_pkt
E bnxt_force_rx_discard
E bnxt_poll_nitroa0
b __bnxt_poll_work
Use napi_build_skb() to take advantage of the skb cache.
In iperf tests with HW-GRO enabled it barely makes a difference
but in cases where HW-GRO is not as effective (or disabled) it
can give even a >10% boost (20.7Gbps -> 23.1Gbps).
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
After the conversion to use the auxiliary bus, the custom device
management is not needed anymore and can be deleted.
Signed-off-by: Petr Pavlu <petr.pavlu@suse.com>
Tested-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Acked-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Use the auxiliary bus to perform device management of the infiniband
part of the mlx4 driver.
Signed-off-by: Petr Pavlu <petr.pavlu@suse.com>
Tested-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Acked-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Use the auxiliary bus to perform device management of the ethernet part
of the mlx4 driver.
Signed-off-by: Petr Pavlu <petr.pavlu@suse.com>
Tested-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Acked-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add an auxiliary virtual bus to model the mlx4 driver structure. The
code is added along the current custom device management logic.
Subsequent patches switch mlx4_en and mlx4_ib to the auxiliary bus and
the old interface is then removed.
Structure mlx4_priv gains a new adev dynamic array to keep track of its
auxiliary devices. Access to the array is protected by the global
mlx4_intf mutex.
Functions mlx4_register_device() and mlx4_unregister_device() are
updated to expose auxiliary devices on the bus in order to load mlx4_en
and/or mlx4_ib. Functions mlx4_register_auxiliary_driver() and
mlx4_unregister_auxiliary_driver() are added to substitute
mlx4_register_interface() and mlx4_unregister_interface(), respectively.
Function mlx4_do_bond() is adjusted to walk over the adev array and
re-adds a specific auxiliary device if its driver sets the
MLX4_INTFF_BONDING flag.
Signed-off-by: Petr Pavlu <petr.pavlu@suse.com>
Tested-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Acked-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The mlx4_core driver has a logic that allows a sub-driver to set the
MLX4_INTFF_BONDING flag which then causes that function mlx4_do_bond()
asks the sub-driver to fully re-probe a device when its bonding
configuration changes.
Performing this operation is disallowed in mlx4_register_interface()
when it is detected that any mlx4 device is multifunction (SRIOV). The
code then resets MLX4_INTFF_BONDING in the driver flags.
Move this check directly into mlx4_do_bond(). It provides a better
separation as mlx4_core no longer directly modifies the sub-driver flags
and it will allow to get rid of explicitly keeping track of all mlx4
devices by the intf.c code when it is switched to an auxiliary bus.
Signed-off-by: Petr Pavlu <petr.pavlu@suse.com>
Tested-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Acked-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Function mlx4_en_queue_bond_work() is used in mlx4_en to start a bond
reconfiguration. It gathers data about a new port map setting, takes
a reference on the netdev that triggered the change and queues a work
object on mlx4_en_priv.mdev.workqueue to perform the operation. The
scheduled work is mlx4_en_bond_work() which calls
mlx4_bond()/mlx4_unbond() and consequently mlx4_do_bond().
At the same time, function mlx4_change_port_types() in mlx4_core might
be invoked to change the port type configuration. As part of its logic,
it re-registers the whole device by calling mlx4_unregister_device(),
followed by mlx4_register_device().
The two operations can result in concurrent access to the data about
currently active interfaces on the device.
Functions mlx4_register_device() and mlx4_unregister_device() lock the
intf_mutex to gain exclusive access to this data. The current
implementation of mlx4_do_bond() doesn't do that which could result in
an unexpected behavior. An updated version of mlx4_do_bond() for use
with an auxiliary bus goes and locks the intf_mutex when accessing a new
auxiliary device array.
However, doing so can then result in the following deadlock:
* A two-port mlx4 device is configured as an Ethernet bond.
* One of the ports is changed from eth to ib, for instance, by writing
into a mlx4_port<x> sysfs attribute file.
* mlx4_change_port_types() is called to update port types. It invokes
mlx4_unregister_device() to unregister the device which locks the
intf_mutex and starts removing all associated interfaces.
* Function mlx4_en_remove() gets invoked and starts destroying its first
netdev. This triggers mlx4_en_netdev_event() which recognizes that the
configured bond is broken. It runs mlx4_en_queue_bond_work() which
takes a reference on the netdev. Removing the netdev now cannot
proceed until the work is completed.
* Work function mlx4_en_bond_work() gets scheduled. It calls
mlx4_unbond() -> mlx4_do_bond(). The latter function tries to lock the
intf_mutex but that is not possible because it is held already by
mlx4_unregister_device().
This particular case could be possibly solved by unregistering the
mlx4_en_netdev_event() notifier in mlx4_en_remove() earlier, but it
seems better to decouple mlx4_en more and break this reference order.
Avoid then this scenario by recognizing that the bond reconfiguration
operates only on a mlx4_dev. The logic to queue and execute the bond
work can be moved into the mlx4_core driver. Only a reference on the
respective mlx4_dev object is needed to be taken during the work's
lifetime. This removes a call from mlx4_en that can directly result in
needing to lock the intf_mutex, it remains a privilege of the core
driver.
Signed-off-by: Petr Pavlu <petr.pavlu@suse.com>
Tested-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Acked-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The mlx4_interface.activate callback was introduced in commit
79857cd31fe7 ("net/mlx4: Postpone the registration of net_device"). It
dealt with a situation when a netdev notifier received a NETDEV_REGISTER
event for a new net_device created by mlx4_en but the same device was
not yet visible to mlx4_get_protocol_dev(). The callback can be removed
now that mlx4_get_protocol_dev() is gone.
Signed-off-by: Petr Pavlu <petr.pavlu@suse.com>
Tested-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Acked-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Use a notifier to implement mlx4_dispatch_event() in preparation to
switch mlx4_en and mlx4_ib to be an auxiliary device.
A problem is that if the mlx4_interface.event callback was replaced with
something as mlx4_adrv.event then the implementation of
mlx4_dispatch_event() would need to acquire a lock on a given device
before executing this callback. That is necessary because otherwise
there is no guarantee that the associated driver cannot get unbound when
the callback is running. However, taking this lock is not possible
because mlx4_dispatch_event() can be invoked from the hardirq context.
Using an atomic notifier allows the driver to accurately record when it
wants to receive these events and solves this problem.
A handler registration is done by both mlx4_en and mlx4_ib at the end of
their mlx4_interface.add callback. This matches the current situation
when mlx4_add_device() would enable events for a given device
immediately after this callback, by adding the device on the
mlx4_priv.list.
Signed-off-by: Petr Pavlu <petr.pavlu@suse.com>
Tested-by: Leon Romanovsky <leonro@nvidia.com>
Acked-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Function mlx4_dispatch_event() takes an 'unsigned long' as its event
parameter. The actual value is none (MLX4_DEV_EVENT_CATASTROPHIC_ERROR),
a pointer to mlx4_eqe (MLX4_DEV_EVENT_PORT_MGMT_CHANGE), or a 32-bit
integer (remaining events).
In preparation to switch mlx4_en and mlx4_ib to be an auxiliary device,
the mlx4_interface.event callback is replaced with a notifier and
function mlx4_dispatch_event() gets updated to invoke
atomic_notifier_call_chain(). This requires forwarding the input 'param'
value from the former function to the latter. A problem is that the
notifier call takes 'void *' as its 'param' value, compared to
'unsigned long' used by mlx4_dispatch_event(). Re-passing the value
would need either punning it to 'void *' or passing down the address of
the input 'param'. Both approaches create a number of unnecessary casts.
Change instead the input 'param' of mlx4_dispatch_event() from
'unsigned long' to 'void *'. A mlx4_eqe pointer can be passed directly,
callers using an int value are adjusted to pass its address.
Signed-off-by: Petr Pavlu <petr.pavlu@suse.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Rename the mlx4_en_dev.nb notifier_block member to netdev_nb in
preparation to add a mlx4 core notifier_block.
Signed-off-by: Petr Pavlu <petr.pavlu@suse.com>
Tested-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Acked-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Simplify the mlx4 driver interface by removing mlx4_get_protocol_dev()
and the associated mlx4_interface.get_dev callbacks. This is done in
preparation to use an auxiliary bus to model the mlx4 driver structure.
The change is motivated by the following situation:
* The mlx4_en interface is being initialized by mlx4_en_add() and
mlx4_en_activate().
* The latter activate function calls mlx4_en_init_netdev() ->
register_netdev() to register a new net_device.
* A netdev event NETDEV_REGISTER is raised for the device.
* The netdev notififier mlx4_ib_netdev_event() is called and it invokes
mlx4_ib_scan_netdevs() -> mlx4_get_protocol_dev() ->
mlx4_en_get_netdev() [via mlx4_interface.get_dev].
This chain creates a problem when mlx4_en gets switched to be an
auxiliary driver. It contains two device calls which would both need to
take a respective device lock.
Avoid this situation by updating mlx4_ib_scan_netdevs() to no longer
call mlx4_get_protocol_dev() but instead to utilize the information
passed in net_device.parent and net_device.dev_port. This data is
sufficient to determine that an updated port is one that the mlx4_ib
driver should take care of and to keep mlx4_ib_dev.iboe.netdevs up to
date.
Following that, update mlx4_ib_get_netdev() to also not call
mlx4_get_protocol_dev() and instead scan all current netdevs to find
find a matching one. Note that mlx4_ib_get_netdev() is called early from
ib_register_device() and cannot use data tracked in
mlx4_ib_dev.iboe.netdevs which is not at that point yet set.
Finally, remove function mlx4_get_protocol_dev() and the
mlx4_interface.get_dev callbacks (only mlx4_en_get_netdev()) as they
became unused.
Signed-off-by: Petr Pavlu <petr.pavlu@suse.com>
Tested-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Acked-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Commit 8cd160a29415 ("qede: convert to new udp_tunnel_nic infra")
removed qede_udp_tunnel_{add,del}() but not the declarations.
Commit 0ebcebbef1cc ("qed: Read device port count from the shmem")
removed qed_device_num_engines() but not its declaration.
Commit 1e128c81290a ("qed: Add support for hardware offloaded FCoE.")
declared but never implemented qed_fcoe_set_pf_params().
Signed-off-by: Yue Haibing <yuehaibing@huawei.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Some of the newer silicon versions in CN10K series supports a feature
where in the current PTP timestamp in HW can be updated atomically
without losing any cpu cycles unlike read/modify/write register.
This patch uses this feature so that PTP accuracy can be improved
while adjusting the master offset in HW. There is no need for SW
timecounter when using this feature. So removed references to SW
timecounter wherever appropriate.
Signed-off-by: Sai Krishna <saikrishnag@marvell.com>
Signed-off-by: Naveen Mamindlapalli <naveenm@marvell.com>
Signed-off-by: Sunil Kovvuri Goutham <sgoutham@marvell.com>
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Support TCP as protocol selector for policy and state in IPsec
packet offload mode.
Example of state configuration is as follows:
ip xfrm state add src 192.168.25.3 dst 192.168.25.1 \
proto esp spi 1001 reqid 10001 aead 'rfc4106(gcm(aes))' \
0x54a7588d36873b031e4bd46301be5a86b3a53879 128 mode transport \
offload packet dev re0 dir in sel src 192.168.25.3 dst 192.168.25.1 \
proto tcp dport 9003
Acked-by: Raed Salem <raeds@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Support RX policy/state upper protocol selector field offload,
to enable selecting RX traffic for IPsec operation based on l4
protocol UDP with specific source/destination port.
Signed-off-by: Emeel Hakim <ehakim@nvidia.com>
Reviewed-by: Raed Salem <raeds@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Instead of using internal devlink_port->index to perform vport lookup in
every devlink port op, store the vport pointer to the container struct
mlx5_devlink_port and use it directly in port ops.
Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Since the follow-up patch is going to remove
mlx5_devlink_port_fn_get_vport() entirely, move the vhca_resource_manager
capability checking to individual ops. Add proper extack message
on the way.
Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
If called from port ops, it is not needed to perform the checks in
mlx5_devlink_eswitch_get(). The reason is devlink port would not be
registered if the checks are not true. Introduce relaxed version
mlx5_devlink_eswitch_nocheck_get() and use it in port ops.
Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Instead of initializing "err" variable, just return "-EOPNOTSUPP"
directly where it is needed.
Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
During devlink port init/cleanup and register/unregister calls, there
are many lookups of vport. Instead of passing vport_num as argument to
functions, pass the vport struct pointer directly and avoid repeated
lookups.
Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Struct devlink_port is usually embedded in a driver-specific struct
which allows to carry driver context to devlink port ops.
Introduce a container struct to include devlink_port struct
in preparation to also include driver context for devlink port ops.
Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Currently each PF/VF/SF devlink port op called into mlx5 code calls
is_port_function_supported() to check if the port is either
PF, VF or SF. So make sure that the ops are registered with devlink
port only for those and avoid the is_port_function_supported() checks
in ops.
Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Since the previous patch removed the only users of these functions,
remove them.
Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Similar to the PF/VF helpers, introduce a set of load/unload helpers
for SF vports. From there, call mlx5_eswitch_load/unload_vport() which
are common for PFs/VFs and newly introduced SF helpers.
Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Currently there is a separate set of functions used to
register/unregister the SF. The only difference is currently the ops
struct. Move the struct up and use it for SFs in
mlx5_esw_offloads_devlink_port_register().
Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
