Age | Commit message (Collapse) | Author |
|
|
|
Signed-off-by: Mandy Lavi <mandy.lavi@freescale.com>
Change-Id: Id1be0b2511e5c36fd785f8ee72f4189305c7ac71
Reviewed-on: http://git.am.freescale.net:8181/4105
Reviewed-by: Bucur Madalin-Cristian-B32716 <madalin.bucur@freescale.com>
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Rivera Jose-B46482 <German.Rivera@freescale.com>
|
|
'9e6c643 phy/fixed: use an unique MDIO bus name' changed the name of the
fixed PHY bus to "fixed-0". This update has been already done for the
non-dpaa SoC(s) in 'e5c7d1f of/mdio: fix fixed link bus name'
Signed-off-by: Shruti Kanetkar <Shruti@Freescale.com>
Change-Id: Ie2ead771bc74a2b62c26eca6b5618bdebbe1eb61
Reviewed-on: http://git.am.freescale.net:8181/4129
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Hamciuc Bogdan-BHAMCIU1 <bogdan.hamciuc@freescale.com>
Reviewed-by: Rivera Jose-B46482 <German.Rivera@freescale.com>
|
|
'9e6c643 phy/fixed: use an unique MDIO bus name' changed the name of the
fixed PHY bus to "fixed-0". This update has been already done for the
non-dpaa SoC(s) in 'e5c7d1f of/mdio: fix fixed link bus name'
Signed-off-by: Shruti Kanetkar <Shruti@Freescale.com>
Change-Id: Ie2ead771bc74a2b62c26eca6b5618bdebbe1eb61
Reviewed-on: http://git.am.freescale.net:8181/4129
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Hamciuc Bogdan-BHAMCIU1 <bogdan.hamciuc@freescale.com>
Reviewed-by: Rivera Jose-B46482 <German.Rivera@freescale.com>
|
|
|
|
Signed-off-by: Mandy Lavi <mandy.lavi@freescale.com>
Change-Id: If2fda57aaf42719d0a3aae2256d513856583913b
Reviewed-on: http://git.am.freescale.net:8181/3427
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
Changes open dma allocation to workaround erratum A006981
Also verify that it applies only to B4-rev1 while rev2 should have this erratum fixed
Signed-off-by: Mandy Lavi <mandy.lavi@freescale.com>
Change-Id: Icc0552269da9ad80763e8ad1394d6ad67ee143a8
Reviewed-on: http://git.am.freescale.net:8181/3428
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
Prevent sleeping function to be called from invalid context
Signed-off-by: Mandy Lavi <mandy.lavi@freescale.com>
Change-Id: I1fa7619489248c823d55a96b00cc4b7bd75bbf6d
Reviewed-on: http://git.am.freescale.net:8181/3430
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
FM_PCD_MatchTableGetMissStatistics and FM_PCD_HashTableGetMissStatistics
Signed-off-by: Mandy Lavi <mandy.lavi@freescale.com>
Change-Id: I56a02300f4603245344effc898ff9fdd137d8e4c
Reviewed-on: http://git.am.freescale.net:8181/3905
Reviewed-by: Schmitt Richard-B43082 <B43082@freescale.com>
Tested-by: Schmitt Richard-B43082 <B43082@freescale.com>
|
|
Apparently reset values of B4/T4 in fmd are not completely
reliable. will use this definition to force fman init
values.
Signed-off-by: Mandy Lavi <mandy.lavi@freescale.com>
Change-Id: Id0b16cf7faf8c53ab9741bcd370e27e91a9ab2a0
Reviewed-on: http://git.am.freescale.net:8181/3904
Reviewed-by: Chereji Marian-Cornel-R27762 <marian.chereji@freescale.com>
Tested-by: Schmitt Richard-B43082 <B43082@freescale.com>
Reviewed-by: Schmitt Richard-B43082 <B43082@freescale.com>
|
|
- miss entry in a CC Node
- miss entry in the hash table
Signed-off-by: Mandy Lavi <mandy.lavi@freescale.com>
Change-Id: Iacbebff6f8f7ef1fa9c97b24d1072515f5098860
Reviewed-on: http://git.am.freescale.net:8181/3828
Reviewed-by: Lavi Mandy-R52568 <Mandy.Lavi@freescale.com>
Reviewed-by: Chereji Marian-Cornel-R27762 <marian.chereji@freescale.com>
Reviewed-by: Schmitt Richard-B43082 <B43082@freescale.com>
Tested-by: Schmitt Richard-B43082 <B43082@freescale.com>
|
|
Allow Rx/Tx pause frame configuration via ethtool.
The gfar devices feature link autonegotioation by default.
The device is being configured with the new pause frame
settings during link autonegotiation (see adjust_link()).
Signed-off-by: Claudiu Manoil <claudiu.manoil@freescale.com>
Change-Id: Icd0b4155ad10e6f9174ffe17c8bf5bec748f9142
Reviewed-on: http://git.am.freescale.net:8181/3651
Reviewed-by: Bucur Madalin-Cristian-B32716 <madalin.bucur@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
Tested-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
MACCFG1 register bits to enable PAUSE frame handling are set by the
driver unconditionally. Per 802.3 standard, PAUSE frame handling is a
full duplex feature, and neither meaningful nor correct in half duplex.
Signed-off-by: Claudiu Manoil <claudiu.manoil@freescale.com>
Change-Id: I3066169b9e96f1c29391ac7636a26ca4152129c3
Reviewed-on: http://git.am.freescale.net:8181/3650
Reviewed-by: Bucur Madalin-Cristian-B32716 <madalin.bucur@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
Tested-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
Prevent sleeping function to be called from invalid context
Signed-off-by: Mandy Lavi <mandy.lavi@freescale.com>
Change-Id: I1fa7619489248c823d55a96b00cc4b7bd75bbf6d
Reviewed-on: http://git.am.freescale.net:8181/3430
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
Changes open dma allocation to workaround erratum A006981
Also verify that it applies only to B4-rev1 while rev2 should have this erratum fixed
Signed-off-by: Mandy Lavi <mandy.lavi@freescale.com>
Change-Id: Icc0552269da9ad80763e8ad1394d6ad67ee143a8
Reviewed-on: http://git.am.freescale.net:8181/3428
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
Signed-off-by: Mandy Lavi <mandy.lavi@freescale.com>
Change-Id: If2fda57aaf42719d0a3aae2256d513856583913b
Reviewed-on: http://git.am.freescale.net:8181/3427
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
When the skb recycling feature is disabled via module param
(gfar_skb_recycling_en), rec->local is being accessed while
not initialized.
Signed-off-by: Claudiu Manoil <claudiu.manoil@freescale.com>
Change-Id: I0f9d355302e694860cc3d46c296409126887b4a9
Reviewed-on: http://git.am.freescale.net:8181/3369
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
Signed-off-by: Mandy Lavi <mandy.lavi@freescale.com>
ENGR00270413
Change-Id: I9ca855698d1eaa93b7cb131d811f4c9dd6c72088
Reviewed-on: http://git.am.freescale.net:8181/3297
Reviewed-by: Bucur Madalin-Cristian-B32716 <madalin.bucur@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
Tested-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
Signed-off-by: Mandy Lavi <mandy.lavi@freescale.com>
ENGR00269928
Change-Id: I7fa7fec2f1eef4b8233ed0d0069e30557adb4bda
Reviewed-on: http://git.am.freescale.net:8181/3296
Reviewed-by: Bucur Madalin-Cristian-B32716 <madalin.bucur@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
Tested-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
Signed-off-by: Mandy Lavi <mandy.lavi@freescale.com>
ENGR00258538
Change-Id: If05f2a070c35c9884782d879566b04f40accfab8
Reviewed-on: http://git.am.freescale.net:8181/3295
Reviewed-by: Chereji Marian-Cornel-R27762 <marian.chereji@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
Tested-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
Signed-off-by: Mandy Lavi <mandy.lavi@freescale.com>
Change-Id: I69b5d2f92045daa71fd8dd6765596397fe736ef5
Reviewed-on: http://git.am.freescale.net:8181/3292
Reviewed-by: Bucur Madalin-Cristian-B32716 <madalin.bucur@freescale.com>
Reviewed-by: Hamciuc Bogdan-BHAMCIU1 <bogdan.hamciuc@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
Tested-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
If the system gets into an out-of-memory state, we are unable to
refill the default buffer pool with new buffers. While this lasts,
just release any incoming frame back to the pool without further
processing.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@freescale.com>
Signed-off-by: Bogdan Hamciuc <bogdan.hamciuc@freescale.com>
Change-Id: Ia7fe59c44aa371f81540b45d197aef896b6d6b3a
Reviewed-on: http://git.am.freescale.net:8181/3050
Reviewed-by: Bucur Madalin-Cristian-B32716 <madalin.bucur@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
Tested-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
When system memory is depleted, the Rx buffers are put back into
their pool. Bad accounting on that error path made it so the software
software counters remained behind the hardware counters, eventually
allowing for negative totals.
Fix this and keep the software pool counters in sync with the hardware
counters.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@freescale.com>
Signed-off-by: Bogdan Hamciuc <bogdan.hamciuc@freescale.com>
Change-Id: I46191bc38f12aa6a25378eba3e5891bc3f18cbc6
Reviewed-on: http://git.am.freescale.net:8181/3049
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Sovaiala Cristian-Constantin-B39531 <Cristian.Sovaiala@freescale.com>
Reviewed-by: Bucur Madalin-Cristian-B32716 <madalin.bucur@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
The initial patch was meant to allow the user to set any MTU she desires
without rebooting the system and changing the kernel bootargs.
This however led to problems when the user would manually choose
FSL_DPAA_ETH_OPTIMIZE_FOR_IPFWD, because the default MAX_FRAME_SIZE=9600
would remain selected and this wasn't all that obvious.
Reverting to the previous config selection is the best choice now.
This reverts commit 68dbabbdb1e28169a372530ca3712c3b63296322.
Signed-off-by: Bogdan Hamciuc <bogdan.hamciuc@freescale.com>
Change-Id: I71499e9b86f8909d049ff4347c4fe035e67d704a
Reviewed-on: http://git.am.freescale.net:8181/2971
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Sovaiala Cristian-Constantin-B39531 <Cristian.Sovaiala@freescale.com>
Reviewed-by: Schmitt Richard-B43082 <B43082@freescale.com>
Tested-by: Schmitt Richard-B43082 <B43082@freescale.com>
|
|
Move back to the "optimize for IPFWD" compile-time choice.
This effectively reverts commit id 692f5140.
Signed-off-by: Bogdan Hamciuc <bogdan.hamciuc@freescale.com>
Change-Id: I9735f489789b7568ae6f6137387d01a92bec9ee2
Reviewed-on: http://git.am.freescale.net:8181/2972
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Sovaiala Cristian-Constantin-B39531 <Cristian.Sovaiala@freescale.com>
Reviewed-by: Schmitt Richard-B43082 <B43082@freescale.com>
Tested-by: Schmitt Richard-B43082 <B43082@freescale.com>
|
|
On P3/P4/P5 platforms, as well as on P1023, FMan's default resource
allocation values are not enough to enable jumbo frame forwarding in the
absence of SG support. Enable the (previously deselected) FMan dynamic
resource allocation algorithm on those platforms, if the legacy
DPAA-Ethernet driver (missing SG support) is used.
Signed-off-by: Bogdan Hamciuc <bogdan.hamciuc@freescale.com>
Change-Id: I6dfc9e2c624c123febfc4d2cbe8fe1fda660525d
Reviewed-on: http://git.am.freescale.net:8181/2929
Reviewed-by: Bucur Madalin-Cristian-B32716 <madalin.bucur@freescale.com>
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Lavi Mandy-R52568 <Mandy.Lavi@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
Tested-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
Check the return value from the buffer-allocating function, in order to
prevent looping forever if memory depletes.
Signed-off-by: Bogdan Hamciuc <bogdan.hamciuc@freescale.com>
Change-Id: Ia321b33ebb38012679248ae4905d4dafaa8efd7d
Reviewed-on: http://git.am.freescale.net:8181/2928
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Sovaiala Cristian-Constantin-B39531 <Cristian.Sovaiala@freescale.com>
Reviewed-by: Radulescu Ruxandra Ioana-B05472 <ruxandra.radulescu@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
Tested-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
Until now we considered a buffer recycleable if its size didn't
exceed the buffer pool size by more than a certain amount.
Recent changes in the stack however allow the allocation of skb
buffers much larger than the requested size, which impacts our
ability to recycle forwarded frames, especially when jumbo frame
support is enabled.
Make the recycling size upper limit a fixed and sufficiently large
value, regardless of the default size of the buffer pool, in order
to simplify recycling logic and guard against unfavourable cases.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@freescale.com>
Change-Id: I311903357335492e821f51fac8a448d67fd73433
Reviewed-on: http://git.am.freescale.net:8181/2899
Reviewed-by: Bucur Madalin-Cristian-B32716 <madalin.bucur@freescale.com>
Reviewed-by: Sovaiala Cristian-Constantin-B39531 <Cristian.Sovaiala@freescale.com>
Reviewed-by: Hamciuc Bogdan-BHAMCIU1 <bogdan.hamciuc@freescale.com>
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
Tested-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
When computing the padding between the beginning of a FD buffer and
the head of the skb associated to it, we wrongly assume that this
padding is always lower than PAGE_SIZE.
In ip forward scenarios with jumbo frames support enabled and Rx
extra headroom not zero, the above assumption is proven incorrect.
Properly compute the padding size in order to avoid skb corruption.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@freescale.com>
Change-Id: I24f08e3cc782a73cbd12d99445c326c530e3d8f2
Reviewed-on: http://git.am.freescale.net:8181/2898
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Bucur Madalin-Cristian-B32716 <madalin.bucur@freescale.com>
Reviewed-by: Sovaiala Cristian-Constantin-B39531 <Cristian.Sovaiala@freescale.com>
Reviewed-by: Hamciuc Bogdan-BHAMCIU1 <bogdan.hamciuc@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
Tested-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
Changed allocation of liodnBase to the Host Command.
For Fmanv3 platforms the first (index 1) Offline Port is not used
therefore the liodnBase should be set starting with index 2 not 1
CR:ENGR00261726
Signed-off-by: Mandy Lavi <mandy.lavi@freescale.com>
|
|
Due to a bad iterator, the Txq initialization logic mangled the Tx
queues settings and their respective Tx Confirmation FQs assignments.
Aside from the obvious structural problem, this created a nasty
imbalance in transmit and confirmation FD processing, resulting in
sustained degradation of termination use-cases.
Signed-off-by: Bogdan Hamciuc <bogdan.hamciuc@freescale.com>
Change-Id: Id28a8499c16405d7d4df5d2e66f1ece241409144
Reviewed-on: http://git.am.freescale.net:8181/2861
Reviewed-by: Bucur Madalin-Cristian-B32716 <madalin.bucur@freescale.com>
Reviewed-by: Sovaiala Cristian-Constantin-B39531 <Cristian.Sovaiala@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
Tested-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
fixed-link does not work in 3.8.x, to make XFI work which has no PHY,
it's actually not necessary to use fixed-link, otherwise there will be
more work and code changes to make fixed-link work in 3.8.x. For now, if
driver does not find the phy_node of MAC, just set the mac_dev->phy_dev
= NULL and return 0; we can get XFI work and this won't break other
ethernet ports. It's only for SDK1.4, if we decide to bring fixed-link
back, it can be done after SDK1.4.
Signed-off-by: Shaohui Xie <Shaohui.Xie@freescale.com>
Change-Id: I930edbe76f5d739f7d6565d3a7f8a3aabf224db0
Reviewed-on: http://git.am.freescale.net:8181/2797
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Zang Tiefei-R61911 <tie-fei.zang@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
Tested-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
Tx Confirmation FQs tend, with originating traffic, to generate a high
number of portal interrupts. A machine doing only TCP transmits may well
end up having more portal interrupts than its Rx peer.
By setting the HOLDACTIVE flag on the Tx Confirmation FQs, we
significantly reduce the number of portal interrupts in termination
tests.
The quantitative effect is anywhere between 2% and 4% on Tx-mainly
cores (most visible in UDP_STREAM tests, least visible on TCP_STREAM).
Signed-off-by: Bogdan Hamciuc <bogdan.hamciuc@freescale.com>
Change-Id: I54e2e8287f72da0cfe231af91c624be805ddee17
Reviewed-on: http://git.am.freescale.net:8181/2815
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Sovaiala Cristian-Constantin-B39531 <Cristian.Sovaiala@freescale.com>
Reviewed-by: Bucur Madalin-Cristian-B32716 <madalin.bucur@freescale.com>
Reviewed-by: Radulescu Ruxandra Ioana-B05472 <ruxandra.radulescu@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
Tested-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
There is an upper limit for the size of a recycleable skb, in order
to prevent unbounded growth of the recycled buffers. This limit
was however too strict and could prevent recycling for some of the
skbs found on the ipforward path.
Relax the recycling condition, such as to also account for internal
buffer padding, skb alignment padding and sh_info size.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@freescale.com>
Change-Id: I0f0b8a48a564657dd107661d6bb5ad474ce8a7a8
Reviewed-on: http://git.am.freescale.net:8181/2807
Reviewed-by: Bucur Madalin-Cristian-B32716 <madalin.bucur@freescale.com>
Reviewed-by: Sovaiala Cristian-Constantin-B39531 <Cristian.Sovaiala@freescale.com>
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Hamciuc Bogdan-BHAMCIU1 <bogdan.hamciuc@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
Tested-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
For Mac-less interfaces we also export the index of the interface
so that it can be used in udev rules for interface renaming.
Signed-off-by: Cristian Sovaiala <cristian.sovaiala@freescale.com>
Change-Id: I522c563ff876cf8c59aad492b057fe91b1f601f5
Reviewed-on: http://git.am.freescale.net:8181/2783
Reviewed-by: Bucur Madalin-Cristian-B32716 <madalin.bucur@freescale.com>
Reviewed-by: Radulescu Ruxandra Ioana-B05472 <ruxandra.radulescu@freescale.com>
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Hamciuc Bogdan-BHAMCIU1 <bogdan.hamciuc@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
Tested-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
Print messages with error FD codes were identical on the Rx Default and
Rx Error paths. Add a small text differentiator on the Rx Error path.
Signed-off-by: Bogdan Hamciuc <bogdan.hamciuc@freescale.com>
Change-Id: I5a1367e2a25df6117ca5454b77862a42714a3112
Reviewed-on: http://git.am.freescale.net:8181/2760
Reviewed-by: Radulescu Ruxandra Ioana-B05472 <ruxandra.radulescu@freescale.com>
Reviewed-by: Sovaiala Cristian-Constantin-B39531 <Cristian.Sovaiala@freescale.com>
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Bucur Madalin-Cristian-B32716 <madalin.bucur@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
Tested-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
FSL_FM_MAX_FRAME_SIZE used to be conservatively set to 1522, to
accomodate the driver without S/G support. But in fact the default value
should depend on whether S/G support is available or not.
Using 9600 as the default max frame size, if S/G support is available in
the DPAA Ethernet driver.
Signed-off-by: Bogdan Hamciuc <bogdan.hamciuc@freescale.com>
Change-Id: I82d4c0b41f4a46e3464eb4d64239bb3d4a4158dc
Reviewed-on: http://git.am.freescale.net:8181/2759
Reviewed-by: Radulescu Ruxandra Ioana-B05472 <ruxandra.radulescu@freescale.com>
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Sovaiala Cristian-Constantin-B39531 <Cristian.Sovaiala@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
Tested-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
Factor number of skb fragments out of the for loop.
Signed-off-by: Bogdan Hamciuc <bogdan.hamciuc@freescale.com>
Change-Id: I750ebee25d92a9a609e1f17187c1e999d179836b
Reviewed-on: http://git.am.freescale.net:8181/2758
Reviewed-by: Radulescu Ruxandra Ioana-B05472 <ruxandra.radulescu@freescale.com>
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Sovaiala Cristian-Constantin-B39531 <Cristian.Sovaiala@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
Tested-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
Fragmented skbuffs are wrapped in SG frames before being enqueued to
FMan's Tx port. A new buffer needs to be allocated on Tx (and freed on
Tx Confirm) to contain the SGT. Up until now, this buffer was one full
page.
Using the page fragment APIs to get smaller buffers for the SGT.
Signed-off-by: Bogdan Hamciuc <bogdan.hamciuc@freescale.com>
Change-Id: I1da07141b119ca431e276025d833178cd73a1077
Reviewed-on: http://git.am.freescale.net:8181/2757
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Sovaiala Cristian-Constantin-B39531 <Cristian.Sovaiala@freescale.com>
Reviewed-by: Radulescu Ruxandra Ioana-B05472 <ruxandra.radulescu@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
Tested-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
Reorder fields in struct dpa_priv_s, based on the frequency of access on
the hotpaths.
Signed-off-by: Bogdan Hamciuc <bogdan.hamciuc@freescale.com>
Change-Id: I06a97e6a647b44a5b398c4dd759460175e1b6bba
Reviewed-on: http://git.am.freescale.net:8181/2756
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Sovaiala Cristian-Constantin-B39531 <Cristian.Sovaiala@freescale.com>
Reviewed-by: Radulescu Ruxandra Ioana-B05472 <ruxandra.radulescu@freescale.com>
Reviewed-by: Bucur Madalin-Cristian-B32716 <madalin.bucur@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
Tested-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
Signed-off-by: Bogdan Hamciuc <bogdan.hamciuc@freescale.com>
Change-Id: I5a8f34a48fe549d00a49270fa1b0442ead3bf5fa
Reviewed-on: http://git.am.freescale.net:8181/2755
Reviewed-by: Sovaiala Cristian-Constantin-B39531 <Cristian.Sovaiala@freescale.com>
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Radulescu Ruxandra Ioana-B05472 <ruxandra.radulescu@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
Tested-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
Signed-off-by: Mandy Lavi <mandy.lavi@freescale.com>
Change-Id: I5efcdbe646486c9b1b0863a714fc49bc7d7ccb3f
Reviewed-on: http://git.am.freescale.net:8181/2754
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Chereji Marian-Cornel-R27762 <marian.chereji@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
Tested-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
CR:ENGR00263170
The previous default liodn_offset is no longer valid in sdk1.4
Signed-off-by: Mandy Lavi <mandy.lavi@freescale.com>
Change-Id: I9acbe9433e4c569b197139d5d9201af740f318f0
Reviewed-on: http://git.am.freescale.net:8181/2732
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Bulie Radu-Andrei-B37577 <Radu.Bulie@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
Patch optimizes the skb clean up for packets processed by ASF IPsec
by using gfar_asf_reclaim_skb() instead of skb_recycle(). Patch also
re-cycles the SKB after DMA to eTsec, instead of recycling it when the
TX BD is used next time.
Change-Id: Iaa396ee2ac89a1ee8b5e8434870481f18674a6d9
Signed-off-by: Rajan Gupta <rajan.gupta@freescale.com>
Reviewed-on: http://git.am.freescale.net:8181/2713
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Manoil Claudiu-B08782 <claudiu.manoil@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
For 36 bit mode, TBASEH/RBASEH eTSEC regs need to be programmed
with the upper 32-bit part of the DMA base addr of the BD region.
This is needed when the BDs are allocated in L2 SRAM above the 4G
boundary.
For addresses below 4G, which is the default use case, these
registers just maintain their default reset value of 0.
Signed-off-by: Claudiu Manoil <claudiu.manoil@freescale.com>
Change-Id: I69e3b86c09ad11780d7124031dd582dc8cd697f0
Reviewed-on: http://git.am.freescale.net:8181/2696
Reviewed-by: Gupta Rajan-B15745 <rajan.gupta@freescale.com>
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
When this option is enabled the driver will try to allocate the
buffer descriptors (BDs) to the L2 SRAM memory, using the
powerpc/85xx cache-sram support API for the P1/P2 QorIQ platforms.
For this to work, the size and the address of the SRAM region
must be provided as kernel boot params.
Should the attempt to allocate in L2 SRAM fail, the driver falls
back to normal allocation of BDs as before (i.e. to DDR).
The gfar_l2sram_en driver module param is provided to allow to
disable the allocation to L2 SRAM at driver level (enabled by
default).
Sysfs stubs allow to enable/diable allocation of BDs to L2 SRAM
for individual gianfar devices, as the L2 SRAM is an expensive
resource. By default, each device has this feature enabled.
This feature also disables the BD L2 stasing feature by default.
Stasing however may still be configured per device via sysfs.
Signed-off-by: Claudiu Manoil <claudiu.manoil@freescale.com>
Change-Id: Ie967cd24c2c6d6654b33d175be0d9e3c785c324a
Reviewed-on: http://git.am.freescale.net:8181/2596
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Gupta Rajan-B15745 <rajan.gupta@freescale.com>
Reviewed-by: Sovaiala Cristian-Constantin-B39531 <Cristian.Sovaiala@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
With this patch, some unnecessary instructions are removed.
There is no need to know the CPU that executes the allocation
procedure. This will eliminate some of dpa_make_private_pool()
requirements (no need to disable thread migration or preemption).
The CPU whose counters will be updated must now be provided to
dpa_bp_add_8().
Change-Id: I09bc88a69649cd52ad1b0055c00117cff5313eac
Signed-off-by: Marian Rotariu <marian.rotariu@freescale.com>
Reviewed-on: http://git.am.freescale.net:8181/2689
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Bucur Madalin-Cristian-B32716 <madalin.bucur@freescale.com>
Reviewed-by: Radulescu Ruxandra Ioana-B05472 <ruxandra.radulescu@freescale.com>
Reviewed-by: Hamciuc Bogdan-BHAMCIU1 <bogdan.hamciuc@freescale.com>
Reviewed-by: Sovaiala Cristian-Constantin-B39531 <Cristian.Sovaiala@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
Signed-off-by: Mandy Lavi <mandy.lavi@freescale.com>
- PFC
Adjustments for PFC configuration constraints and limitations related to
port prefetch mode
- workaround ucode issues
Fix the following HW erratas regarding discard/error frames on V3:
FM_OP_NO_VSP_NO_RELEASE_ERRATA_FMAN_A006675 -
Description: OP without VSP will cause buffer leaks when instructed to discard a frame.
Workaround: FW will release the buffers.
FM_ERROR_VSP_NO_MATCH_SW006 -
Description: Any port with VSP enabled and multiple VSPs are configured
on this port can cause a situation where an error frame will
be enqueued to the error queue not with the default VSP.
Workaround: FW will replaced the current VSP with the default VSP
just before the frame is being enqueued to the error queue.
- Chosen-node new parameter support
errors-to-discard
Usage: optional
Value type: <u32>
Definition: Specifies which errors should be discarded.
Errors that are not in the mask, will not be discarded;
I.e. those errors will be enqueued and sent to the default error queue.
Change-Id: Ib468c67de88376e17d9c39ab5a0c8fc5b33b7b82
Reviewed-on: http://git.am.freescale.net:8181/2605
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Bucur Madalin-Cristian-B32716 <madalin.bucur@freescale.com>
Reviewed-by: Garg Vakul-B16394 <vakul@freescale.com>
Reviewed-by: Radulescu Ruxandra Ioana-B05472 <ruxandra.radulescu@freescale.com>
Reviewed-by: Chereji Marian-Cornel-R27762 <marian.chereji@freescale.com>
Reviewed-by: Wang Haiying-R54964 <Haiying.Wang@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
Patch enhances the performance for forwarded traffic by re-using the
skb in rx ring instead of allocating a new skb. Changes are done
under RX_TX_BUFF_XCHG patch which gets enabled for ASF enabled kernel only.
Patch also removes per-cpu variables used in RX_TX_BUFF_XCHG patch, replacing
with parameters in skbuff structures to enhance the performance.
Change-Id: I2b3c1ec80fe3ef21ade9ce881d5cb86695169518
Signed-off-by: Rajan Gupta <rajan.gupta@freescale.com>
Reviewed-on: http://git.am.freescale.net:8181/2678
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Manoil Claudiu-B08782 <claudiu.manoil@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
|
|
This is the 3.8.13 stable release
|