| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
When netdev->flags has IFF_PROMISC or IFF_ALLMULTI, set the
corresponding bits in the MAC Control Register (MACCR).
This change is based on code from the ftgmac100 driver, see
ftgmac100_start_hw() in ftgmac100.c
Signed-off-by: Sergei Antonov <saproj@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Per cpu entries are no longer used in consideration
for doing gc or not. Remove the extra per cpu entries
pull to directly check for time and perform gc.
Signed-off-by: Tanmay Bhushan <007047221b@gmail.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Siddharth Vadapalli says:
====================
Add PPS support to am65-cpts driver
The CPTS hardware doesn't support PPS signal generation. Using the GenFx
(periodic signal generator) function, it is possible to model a PPS signal
followed by routing it via the time sync router to the CPTS_HWy_TS_PUSH
(hardware time stamp) input, in order to generate timestamps at 1 second
intervals.
This series adds driver support for enabling PPS signal generation.
Additionally, the documentation for the am65-cpts driver is updated with
the bindings for the "ti,pps" property, which is used to inform the
pair [CPTS_HWy_TS_PUSH, GenFx] to the cpts driver.
Changes from v1:
1. Drop device-tree patches.
2. Address Roger's comments on the:
"net: ethernet: ti: am65-cpts: add pps support" patch.
3. Collect Reviewed-by tag from Rob Herring.
v1:
https://lore.kernel.org/r/20230111114429.1297557-1-s-vadapalli@ti.com/
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
When CPTS clock is sync/adjusted by running linuxptp (ptp4l) it will cause
PPS jitter as Genf running PPS is not adjusted.
The same PPM adjustment has to be applied to GenF as to PHC clock to
correct PPS length and keep them in sync.
Testing:
Master:
ptp4l -P -2 -H -i eth0 -l 6 -m -q -p /dev/ptp1 -f ptp.cfg &
testptp -d /dev/ptp1 -P 1
ppstest /dev/pps0
Slave:
linuxptp/ptp4l -P -2 -H -i eth0 -l 6 -m -q -p /dev/ptp1 -f ptp1.cfg -s &
<port 1: UNCALIBRATED to SLAVE on MASTER_CLOCK_SELECTED;>
testptp -d /dev/ptp1 -P 1
ppstest /dev/pps0
Master log:
source 0 - assert 620.000000689, sequence: 530
source 0 - assert 621.000000689, sequence: 531
source 0 - assert 622.000000689, sequence: 532
source 0 - assert 623.000000689, sequence: 533
source 0 - assert 624.000000689, sequence: 534
source 0 - assert 625.000000689, sequence: 535
source 0 - assert 626.000000689, sequence: 536
source 0 - assert 627.000000689, sequence: 537
source 0 - assert 628.000000689, sequence: 538
source 0 - assert 629.000000689, sequence: 539
source 0 - assert 630.000000689, sequence: 540
source 0 - assert 631.000000689, sequence: 541
source 0 - assert 632.000000689, sequence: 542
source 0 - assert 633.000000689, sequence: 543
source 0 - assert 634.000000689, sequence: 544
source 0 - assert 635.000000689, sequence: 545
Slave log:
source 0 - assert 620.000000706, sequence: 252
source 0 - assert 621.000000709, sequence: 253
source 0 - assert 622.000000707, sequence: 254
source 0 - assert 623.000000707, sequence: 255
source 0 - assert 624.000000706, sequence: 256
source 0 - assert 625.000000705, sequence: 257
source 0 - assert 626.000000709, sequence: 258
source 0 - assert 627.000000709, sequence: 259
source 0 - assert 628.000000707, sequence: 260
source 0 - assert 629.000000706, sequence: 261
source 0 - assert 630.000000710, sequence: 262
source 0 - assert 631.000000708, sequence: 263
source 0 - assert 632.000000705, sequence: 264
source 0 - assert 633.000000710, sequence: 265
source 0 - assert 634.000000708, sequence: 266
source 0 - assert 635.000000707, sequence: 267
Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: Siddharth Vadapalli <s-vadapalli@ti.com>
Reviewed-by: Roger Quadros <rogerq@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
CPTS doesn't have HW support for PPS ("pulse per second”) signal
generation, but it can be modeled by using Time Sync Router and routing
GenFx (periodic signal generator) output to CPTS_HWy_TS_PUSH (hardware time
stamp) input, and configuring GenFx to generate 1sec pulses.
+------------------------+
| CPTS |
| |
+--->CPTS_HW4_PUSH GENFx+---+
| | | |
| +------------------------+ |
| |
+--------------------------------+
Add corresponding support to am65-cpts driver. The DT property "ti,pps"
has to be used to enable PPS support and configure pair
[CPTS_HWy_TS_PUSH, GenFx].
Once enabled, PPS can be tested using ppstest tool:
# ./ppstest /dev/pps0
Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: Siddharth Vadapalli <s-vadapalli@ti.com>
Reviewed-by: Roger Quadros <rogerq@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add the ti,pps property used to indicate the pair of HWx_TS_PUSH input and
the TS_GENFy output.
Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: Siddharth Vadapalli <s-vadapalli@ti.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Clark Wang says:
====================
stmmac: Add eqos and fec support for imx93
This patchset add imx93 support for dwmac-imx glue driver.
There are some changes of GPR implement.
And add fec and eqos nodes for imx93 dts.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Enable FEC function for imx93-11x11-evk board.
Signed-off-by: Clark Wang <xiaoning.wang@nxp.com>
Reviewed-by: Peng Fan <peng.fan@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Enable EQoS function for imx93-11x11-evk board.
Signed-off-by: Clark Wang <xiaoning.wang@nxp.com>
Reviewed-by: Peng Fan <peng.fan@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add FEC node for imx93 platform.
Signed-off-by: Clark Wang <xiaoning.wang@nxp.com>
Reviewed-by: Peng Fan <peng.fan@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add EQoS node for imx93 platform.
Signed-off-by: Clark Wang <xiaoning.wang@nxp.com>
Reviewed-by: Peng Fan <peng.fan@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add mx93 compatible string for fec driver.
Signed-off-by: Clark Wang <xiaoning.wang@nxp.com>
Acked-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add mx93 compatible string for eqos driver.
Signed-off-by: Clark Wang <xiaoning.wang@nxp.com>
Acked-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add imx93 platform support for dwmac-imx driver.
Signed-off-by: Clark Wang <xiaoning.wang@nxp.com>
Reviewed-by: Peng Fan <peng.fan@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Introduce NFT_MSG_DESTROY* message type. The destroy operation performs a
delete operation but ignoring the ENOENT errors.
This is useful for the transaction semantics, where failing to delete an
object which does not exist results in aborting the transaction.
This new command allows the transaction to proceed in case the object
does not exist.
Signed-off-by: Fernando Fernandez Mancera <ffmancera@riseup.net>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Florian Westphal <fw@strlen.de>
|
|
nft_ct expression cannot be made builtin to nf_tables without also
forcing the conntrack itself to be builtin.
However, this can be avoided by splitting retrieval of a few
selector keys that only need to access the nf_conn structure,
i.e. no function calls to nf_conntrack code.
Many rulesets start with something like
"ct status established,related accept"
With this change, this no longer requires an indirect call, which
gives about 1.8% more throughput with a simple conntrack-enabled
forwarding test (retpoline thunk used).
Signed-off-by: Florian Westphal <fw@strlen.de>
|
|
objref expression is builtin, so avoid calls to it for
RETOLINE=y builds.
Signed-off-by: Florian Westphal <fw@strlen.de>
|
|
If CONFIG_RETPOLINE is enabled nf_tables avoids indirect calls for
builtin expressions.
On newer cpus indirect calls do not go through the retpoline thunk
anymore, even for RETPOLINE=y builds.
Just like with the new tc retpoline wrappers:
Add a static key to skip the if / else if cascade if the cpu
does not require retpolines.
Suggested-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
|
|
Marked as 'to be removed soon' since kernel 4.1 (2015).
Functionality was superseded by the 'cluster' match, added in kernel
2.6.30 (2009).
clusterip_tg_check still has races that can give
proc_dir_entry 'ipt_CLUSTERIP/10.1.1.2' already registered
followed by a WARN splat.
Remove it instead of trying to fix this up again.
clusterip uapi header is left as-is for now.
Signed-off-by: Florian Westphal <fw@strlen.de>
|
|
Move rcu_read_lock/unlock to nf_conntrack_find_get(), this avoids
nested rcu_read_lock call from resolve_normal_ct().
Signed-off-by: Florian Westphal <fw@strlen.de>
|
|
Compiler can't merge the two test_bit() calls, so load ct->status
once and use non-atomic accesses.
This is fine because IPS_EXPECTED or NAT_CLASH are either set at ct
creation time or not at all, but compiler can't know that.
Signed-off-by: Florian Westphal <fw@strlen.de>
|
|
Those are all useless or dubious.
getorigdst() is called via setsockopt, so return value/errno will
already indicate an appropriate error.
For other pr_debug calls there are better replacements, such as
slab/slub debugging or 'conntrack -E' (ctnetlink events).
Signed-off-by: Florian Westphal <fw@strlen.de>
|
|
The conntrack logging facilities include useful info such as in/out
interface names and packet headers.
Use those in more places instead of pr_debug calls.
Furthermore, several pr_debug calls can be removed, they are useless
on production machines due to the sheer volume of log messages.
Signed-off-by: Florian Westphal <fw@strlen.de>
|
|
Michael Walle says:
====================
net: mdio: Continue separating C22 and C45
I've picked this older series from Andrew up and rebased it onto
the latest net-next.
This is the third (and hopefully last) patch set in the series which
separates the C22 and C45 MDIO bus transactions at the API level to the
MDIO bus drivers.
The first patch is a newer version of the former "net: dsa: Separate C22
and C45 MDIO bus transaction methods", which only contains the mt7530
changes. Although posted as v1, because this is a new series, there
is a changelog included in the patch comment section.
The last patch is a new one, which isn't from Andrew's tree.
====================
Link: https://lore.kernel.org/r/20230116-net-next-c45-seperation-part-3-v1-0-0c53afa56aad@walle.cc
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The rswitch MDIO bus driver only supports C45 transfers. Update the
function names to make this clear, pass the mmd as a parameter, and
register the accessors to the _c45 ops of the bus driver structure.
Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The T1 MDIO bus driver can perform both C22 and C45 transfers. Create
separate functions for each and register the C45 versions using the
new API calls where appropriate.
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The sja1105 MDIO bus driver only supports C45 transfers. Update the
function names to make this clear, pass the mmd as a parameter, and
register the accessors to the _c45 ops of the bus driver structure.
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The xgbe driver reuses MII_ADDR_C45 for its own purpose. The values
derived with it are never passed to phylib or a linux MDIO bus driver.
In order that MII_ADDR_C45 can be removed, add an XGBE specific
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The xgbe MDIO bus driver can perform both C22 and C45 transfers, when
using its MDIO bus hardware. The SFP I2C mdio bus driver only supports
C22. Create separate functions for each and register the C45 versions
using the new API calls where appropriate.
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The hns MDIO bus driver can perform both C22 and C45 transfers.
Create separate functions for each and register the C45 versions using
the new API calls where appropriate.
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
When performing a C45 bus transaction, make use of the c45 variants of
the bus read/write helpers. The ability to pass a special register
value is being removed to clean up the mdio bus driver API.
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The ixgbe MDIO bus driver can perform both C22 and C45 transfers.
Create separate functions for each and register the C45 versions using
the new API calls where appropriate.
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The macb MDIO bus driver can perform both C22 and C45 transfers.
Create separate functions for each and register the C45 versions using
the new API calls where appropriate.
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The nixge MDIO bus driver can perform both C22 and C45 transfers.
Create separate functions for each and register the C45 versions using
the new API calls where appropriate.
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The sxgdb MDIO bus driver can perform both C22 and C45 transfers.
Create separate functions for each and register the C45 versions using
the new API calls where appropriate.
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
mt7530 does support C45, but its uses a mix of registering its MDIO
bus and providing its private MDIO bus to the DSA core, too. This makes
the change a bit more complex.
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Convert various mult-bit fields to be defined using GENMASK/FIELD_PREP.
Simplify the code with the ternary operator.
Signed-off-by: Mengyuan Lou <mengyuanlou@net-swift.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Link: https://lore.kernel.org/r/20230116103839.84087-1-mengyuanlou@net-swift.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
https://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux
Wolfram Sang says:
====================
Immutable branch adding fwnode API to the I2C core
I2C changes requested by Russell King.
This allows him to rework SFP code further.
* tag 'i2c-fwnode-api-2023017' of https://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
i2c: add fwnode APIs
====================
Link: https://lore.kernel.org/r/Y8ZhI4g0wsvpjokd@ninjato/
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
https://git.kernel.org/pub/scm/linux/kernel/git/broonie/regmap
Mark Brown says:
====================
regmap: Rework regmap_mdio_c45_{read|write} for new C45 API.
This reworks the regmap MDIO handling of C45 addresses in
preparation for some forthcoming updates to the networking code.
* tag 'regmap-mdio-c45-rework' of https://git.kernel.org/pub/scm/linux/kernel/git/broonie/regmap:
regmap: Rework regmap_mdio_c45_{read|write} for new C45 API.
====================
Link: https://lore.kernel.org/r/Y8VjkgcWHjR9TzNw@sirena.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The recent refactoring to split C22 and C45 introduced two unneeded
semiconons which the kernel test bot reported. Remove them.
Reported-by: kernel test robot <lkp@intel.com>
Fixes: 93641ecbaa1f ("net: mdio: cavium: Separate C22 and C45 transactions")
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Michael Walle <michael@walle.cc>
Link: https://lore.kernel.org/r/20230115164203.510615-1-andrew@lunn.ch
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
__inet_hash_connect() has a fast path taken if sk_head(&tb->owners) is
equal to the sk parameter.
sk_head() returns the hlist_entry() with respect to the sk_node field.
However entries in the tb->owners list are inserted with respect to the
sk_bind_node field with sk_add_bind_node().
Thus the check would never pass and the fast path never execute.
This fast path has never been executed or tested as this bug seems
to be present since commit 1da177e4c3f4 ("Linux-2.6.12-rc2"), thus
remove it to reduce code complexity.
Signed-off-by: Pietro Borrello <borrello@diag.uniroma1.it>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20230112-inet_hash_connect_bind_head-v3-1-b591fd212b93@diag.uniroma1.it
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
Lorenzo Bianconi says:
====================
net: ethernet: mtk_wed: introduce reset support
Introduce proper reset integration between ethernet and wlan drivers in order
to schedule wlan driver reset when ethernet/wed driver is resetting.
Introduce mtk_hw_reset_monitor work in order to detect possible DMA hangs.
====================
Link: https://lore.kernel.org/r/cover.1673715298.git.lorenzo@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
Introduce reset and reset_complete wlan callback to schedule WLAN driver
reset when ethernet/wed driver is resetting.
Tested-by: Daniel Golle <daniel@makrotopia.org>
Co-developed-by: Sujuan Chen <sujuan.chen@mediatek.com>
Signed-off-by: Sujuan Chen <sujuan.chen@mediatek.com>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
Introduce mtk_hw_check_dma_hang routine to monitor possible dma hangs.
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Tested-by: Daniel Golle <daniel@makrotopia.org>
Co-developed-by: Sujuan Chen <sujuan.chen@mediatek.com>
Signed-off-by: Sujuan Chen <sujuan.chen@mediatek.com>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
Avoid to power-down the ethernet chip during hw reset and align reset
procedure to vendor sdk.
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Tested-by: Daniel Golle <daniel@makrotopia.org>
Co-developed-by: Sujuan Chen <sujuan.chen@mediatek.com>
Signed-off-by: Sujuan Chen <sujuan.chen@mediatek.com>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
Introduce mtk_hw_warm_reset utility routine. This is a preliminary patch
to align reset procedure to vendor sdk and avoid to power down the chip
during hw reset.
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Tested-by: Daniel Golle <daniel@makrotopia.org>
Co-developed-by: Sujuan Chen <sujuan.chen@mediatek.com>
Signed-off-by: Sujuan Chen <sujuan.chen@mediatek.com>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
This is a preliminary patch to add Wireless Ethernet Dispatcher reset
support.
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Tested-by: Daniel Golle <daniel@makrotopia.org>
Co-developed-by: Sujuan Chen <sujuan.chen@mediatek.com>
Signed-off-by: Sujuan Chen <sujuan.chen@mediatek.com>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
Jesper Dangaard Brouer says:
====================
net: use kmem_cache_free_bulk in kfree_skb_list
The kfree_skb_list function walks SKB (via skb->next) and frees them
individually to the SLUB/SLAB allocator (kmem_cache). It is more
efficient to bulk free them via the kmem_cache_free_bulk API.
Netstack NAPI fastpath already uses kmem_cache bulk alloc and free
APIs for SKBs.
The kfree_skb_list call got an interesting optimization in commit
520ac30f4551 ("net_sched: drop packets after root qdisc lock is
released") that can create a list of SKBs "to_free" e.g. when qdisc
enqueue fails or deliberately chooses to drop . It isn't a normal data
fastpath, but the situation will likely occur when system/qdisc are
under heavy workloads, thus it makes sense to use a faster API for
freeing the SKBs.
E.g. the (often distro default) qdisc fq_codel will drop batches of
packets from fattest elephant flow, default capped at 64 packets (but
adjustable via tc argument drop_batch).
Performance measurements done in [1]:
[1] https://github.com/xdp-project/xdp-project/blob/master/areas/mem/kfree_skb_list01.org
====================
Link: https://lore.kernel.org/r/167361788585.531803.686364041841425360.stgit@firesoul
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
The kfree_skb_list function walks SKB (via skb->next) and frees them
individually to the SLUB/SLAB allocator (kmem_cache). It is more
efficient to bulk free them via the kmem_cache_free_bulk API.
This patches create a stack local array with SKBs to bulk free while
walking the list. Bulk array size is limited to 16 SKBs to trade off
stack usage and efficiency. The SLUB kmem_cache "skbuff_head_cache"
uses objsize 256 bytes usually in an order-1 page 8192 bytes that is
32 objects per slab (can vary on archs and due to SLUB sharing). Thus,
for SLUB the optimal bulk free case is 32 objects belonging to same
slab, but runtime this isn't likely to occur.
The expected gain from using kmem_cache bulk alloc and free API
have been assessed via a microbencmark kernel module[1].
The module 'slab_bulk_test01' results at bulk 16 element:
kmem-in-loop Per elem: 109 cycles(tsc) 30.532 ns (step:16)
kmem-bulk Per elem: 64 cycles(tsc) 17.905 ns (step:16)
More detailed description of benchmarks avail in [2].
[1] https://github.com/netoptimizer/prototype-kernel/tree/master/kernel/mm
[2] https://github.com/xdp-project/xdp-project/blob/master/areas/mem/kfree_skb_list01.org
V2: rename function to kfree_skb_add_bulk.
Reviewed-by: Saeed Mahameed <saeed@kernel.org>
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
The SKB drop reason uses __builtin_return_address(0) to give the call
"location" to trace_kfree_skb() tracepoint skb:kfree_skb.
To keep this stable for compilers kfree_skb_reason() is annotated with
__fix_address (noinline __noclone) as fixed in commit c205cc7534a9
("net: skb: prevent the split of kfree_skb_reason() by gcc").
The function kfree_skb_list_reason() invoke kfree_skb_reason(), which
cause the __builtin_return_address(0) "location" to report the
unexpected address of kfree_skb_list_reason.
Example output from 'perf script':
kpktgend_0 1337 [000] 81.002597: skb:kfree_skb: skbaddr=0xffff888144824700 protocol=2048 location=kfree_skb_list_reason+0x1e reason: QDISC_DROP
Patch creates an __always_inline __kfree_skb_reason() helper call that
is called from both kfree_skb_list() and kfree_skb_list_reason().
Suggestions for solutions that shares code better are welcome.
As preparation for next patch move __kfree_skb() invocation out of
this helper function.
Reviewed-by: Saeed Mahameed <saeed@kernel.org>
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
